paperID
stringlengths 36
36
| pwc_id
stringlengths 8
47
| arxiv_id
stringlengths 6
16
⌀ | nips_id
float64 | url_abs
stringlengths 18
329
| url_pdf
stringlengths 18
742
| title
stringlengths 8
325
| abstract
stringlengths 1
7.27k
⌀ | authors
stringlengths 2
7.06k
| published
stringlengths 10
10
⌀ | conference
stringlengths 12
47
⌀ | conference_url_abs
stringlengths 16
198
⌀ | conference_url_pdf
stringlengths 27
199
⌀ | proceeding
stringlengths 6
47
⌀ | taskID
stringlengths 7
1.44k
| areaID
stringclasses 688
values | embedding
stringlengths 9.26k
12.5k
| umap_embedding
stringlengths 29
44
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1e260fd9-8423-4ba1-8b2d-bc0c0e12ee9b
|
enhancing-balanced-graph-edge-partition-with
|
2012.09451
| null |
https://arxiv.org/abs/2012.09451v1
|
https://arxiv.org/pdf/2012.09451v1.pdf
|
Enhancing Balanced Graph Edge Partition with Effective Local Search
|
Graph partition is a key component to achieve workload balance and reduce job completion time in parallel graph processing systems. Among the various partition strategies, edge partition has demonstrated more promising performance in power-law graphs than vertex partition and thereby has been more widely adopted as the default partition strategy by existing graph systems. The graph edge partition problem, which is to split the edge set into multiple balanced parts to minimize the total number of copied vertices, has been widely studied from the view of optimization and algorithms. In this paper, we study local search algorithms for this problem to further improve the partition results from existing methods. More specifically, we propose two novel concepts, namely adjustable edges and blocks. Based on these, we develop a greedy heuristic as well as an improved search algorithm utilizing the property of the max-flow model. To evaluate the performance of our algorithms, we first provide adequate theoretical analysis in terms of the approximation quality. We significantly improve the previously known approximation ratio for this problem. Then we conduct extensive experiments on a large number of benchmark datasets and state-of-the-art edge partition strategies. The results show that our proposed local search framework can further improve the quality of graph partition by a wide margin.
|
['Kian-Lee Tan', 'Dongxiang Zhang', 'Yi Zhou', 'Mingyu Xiao', 'Zhenyu Guo']
|
2020-12-17
| null | null | null | null |
['novel-concepts']
|
['reasoning']
|
[ 1.53295079e-03 -1.82357803e-01 -4.02144790e-01 -1.08713266e-02
-2.24619031e-01 -3.52521926e-01 1.25578530e-02 4.93989617e-01
-2.04132229e-01 7.32795060e-01 -3.64816189e-01 -5.00411153e-01
-6.27393305e-01 -1.07240582e+00 -3.19950432e-01 -7.25526571e-01
-5.39496839e-02 7.43179262e-01 6.08691156e-01 -1.72906712e-01
3.62287402e-01 6.23771012e-01 -1.17553341e+00 -3.87986809e-01
9.60879207e-01 8.06081295e-01 2.51350313e-01 3.54999393e-01
-5.61252981e-02 4.58408266e-01 -4.31100905e-01 -2.09861726e-01
4.45163697e-01 -5.62317073e-01 -9.47420537e-01 3.05127680e-01
-5.16809702e-01 1.74651846e-01 -4.19112861e-01 9.67779994e-01
7.36059904e-01 3.90691608e-01 2.68439323e-01 -1.44022048e+00
-7.36407712e-02 7.40523577e-01 -1.21365941e+00 4.86077935e-01
1.41787767e-01 -1.22884445e-01 9.75993574e-01 -2.45688394e-01
5.56650400e-01 7.94767618e-01 3.72436523e-01 -1.78215262e-02
-1.18594432e+00 -6.00373328e-01 2.22569168e-01 4.61513162e-01
-1.78926945e+00 -1.53711811e-01 8.03728163e-01 -9.13598239e-02
9.31323111e-01 5.22805214e-01 5.76915205e-01 -7.82586932e-02
4.29942340e-01 3.64216983e-01 9.21290636e-01 -4.79105592e-01
-6.49767788e-03 -2.62980722e-02 3.68479490e-01 7.35711098e-01
7.65867114e-01 -2.56494462e-01 -3.43197405e-01 -2.39666656e-01
6.09979153e-01 -8.38304609e-02 -4.77919668e-01 -6.10815883e-01
-8.35597038e-01 5.56576192e-01 4.73756641e-01 8.76282454e-02
-3.89257610e-01 1.85854703e-01 4.67413247e-01 -2.01142579e-02
3.72272909e-01 1.93284467e-01 -1.07705794e-01 2.69003818e-03
-7.37363935e-01 2.48017713e-01 9.22548473e-01 1.05141211e+00
7.61703074e-01 -3.32691759e-01 -3.35567921e-01 6.74613237e-01
1.89273432e-01 2.78225720e-01 -1.02797098e-01 -3.90417188e-01
5.80237389e-01 8.55654895e-01 -1.54629245e-01 -1.51181400e+00
-5.38217783e-01 -6.95818484e-01 -1.08331001e+00 -3.66345227e-01
1.83654830e-01 6.98530823e-02 -5.12931585e-01 1.49089801e+00
6.39004707e-01 9.99239534e-02 -4.65955228e-01 7.97675550e-01
3.22799057e-01 7.01588333e-01 -1.07976004e-01 -7.74974644e-01
1.30035675e+00 -1.17384768e+00 -6.87383771e-01 3.32489647e-02
6.59203231e-01 -6.94882751e-01 5.68787515e-01 2.12918639e-01
-1.20914245e+00 -3.39850396e-01 -1.02163351e+00 2.27505073e-01
-4.85512689e-02 -2.45282307e-01 6.07124865e-01 7.37622023e-01
-9.65038598e-01 4.21148598e-01 -8.98998439e-01 -3.96994114e-01
-2.54891850e-02 4.67746675e-01 -1.05081417e-01 -1.15648784e-01
-8.07778358e-01 4.60958630e-01 4.98530537e-01 6.81836978e-02
-2.15840459e-01 -4.72494096e-01 -5.58505535e-01 5.25443077e-01
1.01748407e+00 -8.56115580e-01 8.46764684e-01 -9.02893990e-02
-1.12493670e+00 6.41382515e-01 -1.36358425e-01 -2.89715081e-01
2.29496062e-01 4.14521873e-01 -2.22706109e-01 1.67882293e-01
-3.27116549e-02 -1.44591808e-01 2.66035438e-01 -1.06171274e+00
-5.59926629e-01 -4.41405803e-01 4.61850464e-02 4.73015368e-01
-4.06007975e-01 2.83409655e-02 -9.21991885e-01 -3.80804956e-01
8.60052034e-02 -1.11187088e+00 -4.97220248e-01 -7.75574982e-01
-5.64440250e-01 -3.44960570e-01 2.67066866e-01 -1.18270673e-01
2.09940290e+00 -1.96761823e+00 1.58179328e-01 7.84939349e-01
6.25901937e-01 1.00100376e-01 3.68399888e-01 7.23074377e-01
2.74627171e-02 1.58349976e-01 -1.64820552e-02 7.09094852e-02
3.35166859e-03 3.28687276e-03 1.63745016e-01 5.68485677e-01
-5.09435892e-01 7.43009806e-01 -6.58063054e-01 -6.14135325e-01
-5.45265861e-02 -2.30112791e-01 -5.23694754e-01 1.39688671e-01
1.73100233e-01 7.29759689e-03 -5.51597893e-01 3.18464994e-01
1.04042947e+00 -5.70602119e-01 5.58830678e-01 -2.19957270e-02
-1.30117266e-03 4.63787802e-02 -1.46241701e+00 1.23059857e+00
-1.17942274e-01 5.86508475e-02 2.68171638e-01 -1.17624497e+00
1.04791999e+00 2.33485475e-02 7.80413508e-01 -5.30632436e-01
1.72970220e-01 8.86325985e-02 1.46108478e-01 -2.99068898e-01
6.55110240e-01 -4.75244001e-02 -1.92367211e-01 6.13026321e-01
-6.79803848e-01 1.50936604e-01 6.76214814e-01 3.64771008e-01
1.31220901e+00 -3.49144071e-01 5.73348939e-01 -7.17435062e-01
6.74583614e-01 8.36329907e-02 5.76165557e-01 6.55740798e-01
-2.17124358e-01 1.21371150e-01 8.29158783e-01 -2.60235161e-01
-6.84545755e-01 -6.91565096e-01 1.26969472e-01 9.41838980e-01
7.00871706e-01 -8.02139699e-01 -7.70166695e-01 -3.85299861e-01
-3.07359062e-02 2.20918939e-01 -2.43500426e-01 -2.35007390e-01
-5.39416134e-01 -9.64995146e-01 8.66649970e-02 4.59027171e-01
3.73847157e-01 -7.84969747e-01 -5.27357578e-01 1.41573861e-01
-1.49425402e-01 -1.03396678e+00 -7.50883639e-01 1.14261121e-01
-6.31982088e-01 -1.22035611e+00 -2.13530511e-01 -7.58776009e-01
8.18440557e-01 8.48323464e-01 1.07866788e+00 6.05305016e-01
-2.05159515e-01 1.11731760e-01 -3.45662802e-01 -2.15649098e-01
-4.62759733e-02 5.64251959e-01 -5.91410957e-02 -1.58558637e-01
1.78338826e-01 -4.29198712e-01 -6.61749005e-01 5.99791408e-01
-9.98143554e-01 2.55957872e-01 4.80376184e-01 6.13339067e-01
8.22500646e-01 8.55354548e-01 4.85282987e-01 -1.27220082e+00
9.86691415e-01 -4.83729094e-01 -6.68477714e-01 3.85926902e-01
-1.02332330e+00 6.84840381e-02 7.04267740e-01 1.47158399e-01
-8.29900146e-01 -2.08655685e-01 1.70817286e-01 -2.23391876e-01
3.64554882e-01 8.43741894e-01 -3.13338488e-01 -2.36086458e-01
1.37105018e-01 2.21815720e-01 -1.44690514e-01 -2.27962643e-01
9.01079178e-03 4.28620100e-01 2.73213416e-01 -8.06367636e-01
6.54089630e-01 2.98204094e-01 5.08846223e-01 -4.82192665e-01
-3.41237545e-01 -9.11504805e-01 -3.15919012e-01 -2.27368876e-01
5.00924289e-01 -3.94835770e-01 -1.13983059e+00 1.82484612e-01
-7.71735609e-01 -1.55948400e-01 1.87059999e-01 1.91306353e-01
-3.60104918e-01 6.97687030e-01 -6.26392782e-01 -8.45900893e-01
-3.58499974e-01 -1.31204796e+00 7.47960269e-01 2.95450419e-01
1.60839632e-01 -8.10587466e-01 1.53970987e-01 9.02040228e-02
2.39680663e-01 2.85578758e-01 9.98454511e-01 -6.19653702e-01
-7.27396727e-01 -9.25574750e-02 -4.44324106e-01 -2.97022641e-01
8.77304152e-02 -1.45875990e-01 -8.44913796e-02 -6.53727949e-01
-1.33008495e-01 2.87590742e-01 7.47876823e-01 2.18319550e-01
1.35950291e+00 6.60881624e-02 -9.41671073e-01 5.86004376e-01
1.74066532e+00 2.47507706e-01 5.53212345e-01 3.95975649e-01
6.83000028e-01 4.67474788e-01 6.84542835e-01 6.72133982e-01
3.71707380e-01 8.05193782e-01 2.02994436e-01 -8.83308649e-02
1.30610004e-01 7.49090826e-03 -2.97176123e-01 1.24385393e+00
-2.11089060e-01 -8.27679157e-01 -1.01017785e+00 3.68270487e-01
-2.09296799e+00 -5.83959341e-01 -3.40633780e-01 2.47818184e+00
5.11789262e-01 4.04076546e-01 3.66534740e-01 4.27529305e-01
1.11775887e+00 3.79080214e-02 -2.20477715e-01 -5.03942490e-01
3.96475881e-01 2.11635634e-01 8.61703396e-01 4.29889411e-01
-6.51579976e-01 6.42863691e-01 5.93121433e+00 1.14726138e+00
-6.76150382e-01 -1.32817015e-01 6.12643182e-01 1.25257865e-01
-2.40286916e-01 1.18477009e-01 -8.56552124e-01 4.90152180e-01
7.87964761e-01 -8.78514588e-01 7.13467479e-01 6.57773495e-01
2.33818620e-01 -1.57275632e-01 -7.89043486e-01 1.04731667e+00
-5.46925589e-02 -9.99609709e-01 -8.51865858e-02 3.68698061e-01
8.07158947e-01 -5.30094206e-01 -3.03504229e-01 1.50010064e-01
-7.53738880e-02 -7.89393127e-01 2.64743328e-01 2.51796722e-01
6.48434699e-01 -1.15951502e+00 8.27334940e-01 4.44922388e-01
-1.77488458e+00 1.93127263e-02 -4.29201424e-01 -2.10107431e-01
4.12104636e-01 7.34693706e-01 -4.61012721e-01 1.27788758e+00
3.96244735e-01 2.04811603e-01 -4.02568579e-01 1.42903256e+00
7.06720501e-02 4.55919921e-01 -5.32599509e-01 -6.71977252e-02
-1.75136868e-02 -5.07177055e-01 5.48588395e-01 8.90929937e-01
5.34430146e-02 3.01189870e-01 5.72684824e-01 4.77676630e-01
-2.32695282e-01 5.83774269e-01 -4.09464300e-01 4.54839086e-03
7.25991368e-01 1.32987380e+00 -1.38375533e+00 -2.24338546e-01
-3.00534189e-01 4.57688630e-01 4.10380155e-01 2.16380447e-01
-1.06723249e+00 -7.38164008e-01 4.00147825e-01 3.37447226e-01
3.05037826e-01 -2.65247136e-01 -2.36374229e-01 -6.65433884e-01
9.71187130e-02 -6.10925555e-01 5.93302369e-01 -2.35492155e-01
-9.55867648e-01 6.51275694e-01 1.99827507e-01 -8.34186912e-01
5.30657247e-02 -2.30436772e-01 -6.08784914e-01 7.50821590e-01
-1.27745068e+00 -5.65295219e-01 -5.11117995e-01 4.24969643e-01
1.58499360e-01 2.37350494e-01 2.67667830e-01 4.61845249e-01
-9.07325864e-01 5.51097870e-01 1.68610498e-01 -3.05994034e-01
5.93213916e-01 -1.00339544e+00 1.78904429e-01 1.11478364e+00
-3.06182206e-01 6.20390892e-01 6.83939934e-01 -7.98597813e-01
-1.52915251e+00 -7.59761453e-01 6.33148670e-01 1.83827609e-01
5.05318522e-01 -2.60904282e-01 -7.24951744e-01 4.57545280e-01
1.52744114e-01 -6.37704954e-02 5.17614365e-01 1.47927687e-01
3.40742052e-01 -4.27904993e-01 -8.67942154e-01 4.86150205e-01
1.17518377e+00 8.18284377e-02 6.21266775e-02 2.36763567e-01
6.89449966e-01 -7.03238308e-01 -8.19942474e-01 5.69150329e-01
2.75396824e-01 -9.37182784e-01 8.16538692e-01 -3.36041063e-01
1.56501681e-01 -4.22832936e-01 2.71683395e-01 -1.19158638e+00
-8.20144355e-01 -8.51700902e-01 -5.15794456e-02 1.17706501e+00
2.26004824e-01 -8.91920745e-01 9.23620522e-01 5.42133629e-01
-7.61420429e-02 -1.31490588e+00 -6.58798158e-01 -9.09873784e-01
-4.59788471e-01 4.90537519e-03 8.78625154e-01 6.67413890e-01
2.65899777e-01 6.14364028e-01 -2.29463682e-01 3.67122032e-02
4.65023309e-01 5.88343263e-01 9.45754766e-01 -9.59398210e-01
-4.73097622e-01 -6.45201564e-01 -3.61843437e-01 -1.16377199e+00
-2.61025243e-02 -1.03894603e+00 -4.48713861e-02 -1.81414771e+00
6.55585945e-01 -6.36735916e-01 -3.30963641e-01 2.08024949e-01
-5.34385383e-01 3.43996845e-02 1.94842517e-01 2.34333813e-01
-9.23193157e-01 3.55967194e-01 1.36667323e+00 1.99646667e-01
-4.16032970e-01 1.36499017e-01 -8.01382363e-01 1.45766169e-01
8.07915032e-01 -4.70183641e-01 -6.83895290e-01 -1.28626525e-01
4.56414312e-01 3.06168169e-01 -3.62733811e-01 -7.69478321e-01
4.45997298e-01 -4.00891989e-01 -2.61721969e-01 -5.17692983e-01
-1.63066596e-01 -9.19870436e-01 4.75046694e-01 5.65922856e-01
-7.37882033e-03 4.61392909e-01 1.98876634e-01 8.42236757e-01
-5.90769574e-02 -1.67300910e-01 5.51758170e-01 1.47760883e-01
-4.55643445e-01 5.74565530e-01 -7.56745711e-02 8.21720138e-02
1.46925890e+00 -2.69903481e-01 -5.11369169e-01 -1.44480482e-01
-3.94251406e-01 7.45276809e-01 5.26646078e-01 1.37505354e-02
2.46590719e-01 -1.27614379e+00 -6.52753651e-01 1.52609339e-02
1.40116781e-01 1.95477605e-02 3.46722722e-01 1.18306041e+00
-1.00205052e+00 6.02098525e-01 4.00404967e-02 -4.29890335e-01
-1.33833742e+00 1.07530797e+00 -1.46838799e-01 -8.20456624e-01
-5.26452601e-01 5.21131337e-01 4.20061976e-01 7.79895559e-02
-1.03362642e-01 2.85729021e-02 5.65940253e-02 -5.02433002e-01
1.73963591e-01 8.26136410e-01 1.87890083e-01 -3.63337040e-01
-5.49998105e-01 3.79125893e-01 -3.63968126e-03 3.58475298e-01
1.08103323e+00 -4.89908129e-01 -5.13164222e-01 1.13651734e-02
8.21399987e-01 2.98716068e-01 -5.09324014e-01 -1.39118344e-01
-1.25649527e-01 -7.65909612e-01 -4.97960299e-02 -1.98247820e-01
-1.21086895e+00 2.74626672e-01 -4.24454473e-02 9.28510725e-01
1.53002739e+00 -1.10534178e-02 1.01988387e+00 5.42724356e-02
4.32770252e-01 -1.03706682e+00 -3.22635770e-01 2.16316000e-01
2.78582990e-01 -7.05511928e-01 4.71457124e-01 -1.18893695e+00
-2.74359941e-01 8.25508654e-01 7.83719897e-01 -2.24415399e-02
6.70042455e-01 2.97520489e-01 -6.07306838e-01 -3.90202194e-01
-7.54053295e-01 -3.26429814e-01 1.25047475e-01 1.05732150e-01
2.89546281e-01 3.54996055e-01 -1.11069894e+00 6.31933630e-01
3.22906785e-02 -2.44239733e-01 3.81574839e-01 8.04995954e-01
-4.94271934e-01 -1.36460197e+00 -3.03749740e-01 6.21851742e-01
-4.11696702e-01 7.57234767e-02 -1.01069838e-01 8.22248340e-01
-1.82020411e-01 8.44467342e-01 -1.27255395e-01 -2.60310769e-01
3.59867483e-01 -3.52374882e-01 6.22840106e-01 -5.60967922e-01
-4.14628118e-01 1.06408797e-01 7.65982419e-02 -3.46555144e-01
-2.34327048e-01 -2.29845569e-01 -1.19801664e+00 -7.90420711e-01
-8.54284286e-01 5.40471613e-01 3.57545584e-01 7.81354964e-01
4.80251580e-01 9.26397443e-01 7.55711198e-01 -4.38649863e-01
-4.33755070e-01 -6.84567511e-01 -1.00100327e+00 2.39244074e-01
-5.20714104e-01 -7.29624331e-01 -5.81226498e-02 -4.64025795e-01]
|
[7.071807861328125, 5.1654229164123535]
|
b47e48fc-ad3c-4226-9bc5-5aefcd1843a2
|
automatic-environmental-sound-recognition
|
1607.04589
| null |
http://arxiv.org/abs/1607.04589v1
|
http://arxiv.org/pdf/1607.04589v1.pdf
|
Automatic Environmental Sound Recognition: Performance versus Computational Cost
|
In the context of the Internet of Things (IoT), sound sensing applications
are required to run on embedded platforms where notions of product pricing and
form factor impose hard constraints on the available computing power. Whereas
Automatic Environmental Sound Recognition (AESR) algorithms are most often
developed with limited consideration for computational cost, this article seeks
which AESR algorithm can make the most of a limited amount of computing power
by comparing the sound classification performance em as a function of its
computational cost. Results suggest that Deep Neural Networks yield the best
ratio of sound classification accuracy across a range of computational costs,
while Gaussian Mixture Models offer a reasonable accuracy at a consistently
small cost, and Support Vector Machines stand between both in terms of
compromise between accuracy and computational cost.
|
['Mark D. Plumbley', 'Sacha Krstulovic', 'Adam M. Stark', 'Siddharth Sigtia']
|
2016-07-15
| null | null | null | null |
['sound-classification']
|
['audio']
|
[ 2.60256648e-01 -3.36403996e-01 1.30702212e-01 -2.35641167e-01
-7.90340602e-01 -4.42880124e-01 3.87276679e-01 -9.48826298e-02
-5.71382403e-01 1.12982780e-01 -2.09091336e-01 -6.05795979e-01
-5.25513947e-01 -8.60163867e-01 -1.30407095e-01 -5.72143555e-01
2.63386250e-01 2.31017619e-02 -1.15647607e-01 2.30252430e-01
2.60269754e-02 4.80999708e-01 -1.74983013e+00 -3.43355089e-01
6.85643971e-01 1.61469769e+00 7.45706409e-02 7.10556090e-01
4.72922362e-02 4.66771334e-01 -5.54578066e-01 -4.58753526e-01
4.24866021e-01 -1.77216575e-01 -2.35359967e-01 -4.37949657e-01
1.86824068e-01 -4.55158263e-01 -1.89085543e-01 1.06007910e+00
9.24565554e-01 1.38251320e-01 6.02773845e-01 -1.31050968e+00
-3.12464863e-01 6.82352483e-01 -7.11111054e-02 2.29266167e-01
1.66062266e-01 8.27002525e-02 1.19619358e+00 -6.71722889e-01
-2.42868751e-01 6.07809186e-01 8.20935905e-01 3.00858796e-01
-1.19725072e+00 -8.27797592e-01 -3.30471694e-01 -1.25758529e-01
-1.47445369e+00 -8.19603324e-01 8.57024312e-01 -1.21344380e-01
1.42863011e+00 4.60077405e-01 6.21874511e-01 6.15085840e-01
3.26534390e-01 -1.34532545e-02 9.06506360e-01 -3.90561998e-01
6.81731761e-01 1.49782319e-02 1.07808840e-02 2.09376976e-01
6.90107822e-01 -2.66228348e-01 -5.89132667e-01 -4.36065167e-01
4.49685544e-01 -1.44251868e-01 1.41275793e-01 3.80554140e-01
-5.25203466e-01 6.62808239e-01 -1.07099481e-01 3.76364976e-01
-3.31367195e-01 5.36516368e-01 1.27316684e-01 2.42269814e-01
3.48411262e-01 5.26499867e-01 -4.29551393e-01 -4.98720646e-01
-1.10902679e+00 1.16017453e-01 8.74581277e-01 4.85771269e-01
3.91662508e-01 5.84992647e-01 5.54886818e-01 1.00021434e+00
5.35379231e-01 7.74865210e-01 4.80946034e-01 -1.11180437e+00
2.70343542e-01 2.37134501e-01 1.44119561e-01 -1.13325870e+00
-3.69135022e-01 -3.19235563e-01 -4.95412230e-01 2.09719464e-01
3.91979843e-01 -2.29951605e-01 -4.97722715e-01 1.28267252e+00
7.59624988e-02 -2.12967902e-01 8.90940949e-02 6.09901071e-01
5.36458015e-01 6.40209079e-01 2.87263125e-01 -9.39118490e-02
1.39615464e+00 -1.49018973e-01 -4.79644626e-01 -5.42996109e-01
2.72608489e-01 -8.04776788e-01 8.31483662e-01 4.27177489e-01
-9.90987241e-01 -3.65776777e-01 -1.32758284e+00 4.85016443e-02
-2.95943916e-01 -3.93785164e-02 8.03486526e-01 1.49097466e+00
-9.00164545e-01 5.85968435e-01 -1.07866204e+00 -4.12515737e-03
2.39170149e-01 3.60585630e-01 1.91208243e-01 1.83764264e-01
-7.38658667e-01 7.21901417e-01 -8.94430652e-02 3.49863797e-01
-1.64865121e-01 -7.96987116e-01 -5.25896013e-01 1.55220270e-01
-6.76543340e-02 -4.72954303e-01 1.22432232e+00 -7.05745637e-01
-1.56646061e+00 3.05174679e-01 3.95260990e-01 -5.43310642e-01
3.90880257e-01 -9.94260311e-02 -9.10936236e-01 2.01182216e-01
-1.96349114e-01 5.44656575e-01 9.40617859e-01 -5.96180081e-01
-6.33331120e-01 -3.50515306e-01 4.05846536e-02 -9.41909850e-02
-4.71982777e-01 1.64485246e-01 4.88846511e-01 -2.83233494e-01
2.83061504e-01 -1.16134882e+00 -1.29340842e-01 -1.97836280e-01
1.48404703e-01 -2.16456875e-01 6.91199839e-01 -6.72293365e-01
1.23129320e+00 -2.22988105e+00 -7.63498306e-01 2.62145609e-01
-4.35087942e-02 4.35255505e-02 2.75570422e-01 4.33232747e-02
4.58687730e-02 2.07982540e-01 1.50618404e-02 1.40555218e-01
3.34683031e-01 3.41980625e-03 -2.25859240e-01 5.37014842e-01
1.16271064e-01 3.72249275e-01 -5.15145779e-01 -3.23754638e-01
2.22990289e-01 5.51004887e-01 -6.94511414e-01 -9.92961526e-02
6.18600473e-02 -2.65480429e-01 -4.52541232e-01 7.21633375e-01
5.13484538e-01 5.51654119e-03 2.71962553e-01 -1.46395072e-01
-1.38014078e-01 5.68329573e-01 -1.61052811e+00 1.21772432e+00
-8.21386099e-01 4.86633003e-01 2.35891148e-01 -6.84768438e-01
9.82400775e-01 3.23955834e-01 6.06257856e-01 -7.56674945e-01
3.10555935e-01 5.84242404e-01 4.43302244e-01 -3.49588037e-01
7.43599951e-01 -2.60750324e-01 -1.91013180e-02 7.25214839e-01
-3.38059098e-01 -2.71582633e-01 -3.03164631e-01 -3.87594938e-01
1.17224157e+00 -5.12359627e-02 1.08762115e-01 -3.79583359e-01
-2.32862562e-01 -7.40010887e-02 2.55099982e-01 3.85594845e-01
-2.75692791e-01 1.37736708e-01 -1.13818213e-01 -3.72485042e-01
-9.96162057e-01 -6.76481307e-01 -4.34267133e-01 1.20026529e+00
-9.73119512e-02 2.94350926e-02 -7.80317545e-01 1.19935691e-01
-8.01648647e-02 8.82729232e-01 1.26417324e-01 5.34063280e-02
-3.24124664e-01 -9.04153407e-01 8.42186868e-01 5.96909940e-01
4.16582346e-01 -8.83866787e-01 -1.68231201e+00 1.60211787e-01
2.79979110e-01 -1.14008892e+00 -1.58020496e-01 4.83557671e-01
-6.07621670e-01 -6.41905725e-01 -3.63029093e-01 -2.40439817e-01
1.63666010e-01 2.59562880e-01 9.86602962e-01 -2.63380026e-03
-4.57208902e-01 8.45043659e-01 -2.99343199e-01 -7.02646673e-01
-4.13980931e-01 -1.81347653e-01 3.91023934e-01 -2.01928183e-01
5.53874850e-01 -9.65528011e-01 -7.62191594e-01 9.00159851e-02
-9.40560877e-01 -3.95418525e-01 4.08512384e-01 2.53794521e-01
2.99829751e-01 8.10010731e-01 6.53664887e-01 -2.40126718e-02
7.16626227e-01 -5.11289835e-01 -4.99415189e-01 -1.54425904e-01
-9.83072937e-01 -2.12578014e-01 4.78949577e-01 -5.84717810e-01
-9.12192822e-01 7.14126006e-02 -2.95206457e-01 -2.78306544e-01
-1.44579917e-01 4.03116018e-01 -2.45012194e-01 -1.20507859e-01
6.16619468e-01 -6.57460093e-02 -1.85343802e-01 -4.61219102e-01
1.15154654e-01 1.16963696e+00 3.58070076e-01 -6.07558966e-01
3.09421301e-01 2.25492254e-01 -7.04359636e-02 -1.07233262e+00
-4.90560621e-01 1.51966721e-01 -2.16265082e-01 -1.59776464e-01
7.65811861e-01 -8.63199055e-01 -6.98072731e-01 3.89638573e-01
-5.78598857e-01 -7.70824924e-02 -3.82834971e-01 6.97472572e-01
-3.54273170e-01 1.21983767e-01 -2.01651111e-01 -1.25304282e+00
-5.68646610e-01 -1.12222648e+00 7.33397186e-01 1.89812303e-01
-6.47644460e-01 -5.50051868e-01 -3.24732453e-01 3.38272005e-01
7.11755991e-01 2.81911455e-02 8.57533991e-01 -6.56991065e-01
-3.80918324e-01 -5.14131367e-01 -6.38308525e-02 4.20801431e-01
5.54430820e-02 4.16035764e-02 -1.25490654e+00 1.56430881e-02
2.09608451e-01 3.69019248e-02 3.18291217e-01 6.05176330e-01
1.04831278e+00 -3.52996916e-01 1.52828425e-01 3.25317740e-01
1.65106392e+00 5.28492868e-01 4.92244393e-01 2.27911502e-01
4.46589202e-01 3.78791660e-01 2.38452762e-01 6.21303558e-01
2.89886385e-01 4.88717109e-01 3.95113438e-01 4.59561110e-01
1.13839023e-01 -2.93839812e-01 2.44476169e-01 9.53012466e-01
-1.13573596e-01 -4.46105972e-02 -9.68224585e-01 3.43774348e-01
-1.19526434e+00 -7.70009339e-01 3.48134220e-01 2.22603154e+00
4.60102201e-01 2.82228678e-01 4.43908900e-01 8.96408617e-01
4.64967936e-01 2.02054426e-01 -5.78436077e-01 -6.92334533e-01
2.33616233e-01 4.42848921e-01 7.92235732e-01 1.08479366e-01
-8.01450431e-01 1.19004592e-01 7.42497587e+00 9.92491543e-01
-1.18678296e+00 7.35302866e-02 7.19364941e-01 -2.44772956e-01
-3.86660635e-01 -3.69330019e-01 -5.23153365e-01 7.89989173e-01
1.69063103e+00 -2.81292796e-02 5.35806656e-01 1.21891153e+00
8.02281797e-02 -1.95762798e-01 -9.55559373e-01 1.26412177e+00
-3.36677462e-01 -9.24403608e-01 -2.39049822e-01 2.72024959e-01
2.11235389e-01 -7.85411373e-02 3.27591389e-01 -1.18747450e-01
1.23785667e-01 -1.05164695e+00 1.07800317e+00 2.86400557e-01
9.19262946e-01 -9.21003282e-01 5.22876441e-01 1.80066586e-01
-1.30924678e+00 -5.08352220e-01 -3.08945954e-01 -5.94994485e-01
-1.68493867e-01 7.11458266e-01 -3.65626872e-01 -2.08869874e-02
1.07025349e+00 -2.94007063e-01 -8.97411481e-02 5.44468522e-01
3.41000289e-01 9.52491999e-01 -6.73973083e-01 -3.94640297e-01
-2.19565667e-02 -2.49956504e-01 2.27188364e-01 1.11842048e+00
7.56140888e-01 2.70936608e-01 -7.03679398e-02 7.88628399e-01
1.43569365e-01 1.15435749e-01 -3.90430003e-01 -3.60910594e-01
1.04514337e+00 1.21334493e+00 -1.09948552e+00 3.03644314e-02
-3.19958806e-01 1.93182915e-01 -4.90468144e-01 -1.06379427e-01
-6.69889688e-01 -4.10833031e-01 9.17330563e-01 2.95556873e-01
2.03715026e-01 -3.34477276e-01 -7.89358020e-01 -4.98290956e-01
5.70467524e-02 -6.21226788e-01 7.25205317e-02 -7.09023297e-01
-1.05409586e+00 3.91767085e-01 -2.57865816e-01 -1.45553708e+00
-2.73358405e-01 -5.24374664e-01 -3.56085390e-01 6.07884288e-01
-1.14259994e+00 -9.42777336e-01 -2.13630348e-02 1.64326340e-01
3.39428753e-01 -1.22391835e-01 9.21839297e-01 4.96517479e-01
-3.15070122e-01 7.20867038e-01 4.80878092e-02 -2.08658233e-01
-1.15717091e-01 -9.45414662e-01 2.41232321e-01 7.14936912e-01
1.78534895e-01 3.92026842e-01 1.06952190e+00 -3.10053974e-01
-1.81736326e+00 -6.40529394e-01 6.47286296e-01 -2.32271627e-01
7.50806034e-01 -1.39898667e-03 -3.98230195e-01 1.50041953e-01
-1.60453126e-01 6.44095838e-02 9.03512478e-01 -2.35948432e-02
-5.17756879e-01 -3.61071020e-01 -1.43291068e+00 3.89029920e-01
7.79711366e-01 -8.65635157e-01 -1.12949543e-01 -2.60726511e-01
5.52743196e-01 -9.19798762e-02 -1.32695544e+00 2.12382078e-01
1.01444435e+00 -7.24257588e-01 9.52024698e-01 3.53257135e-02
3.59522820e-01 -3.43458414e-01 -7.98230290e-01 -8.41609538e-01
-2.22139761e-01 -5.83496690e-01 1.78287700e-01 1.51589608e+00
4.57793355e-01 -7.69030929e-01 5.99632025e-01 1.42937326e+00
6.03449233e-02 -7.19262242e-01 -1.23585284e+00 -7.45379090e-01
-1.11249998e-01 -1.37789214e+00 8.84580255e-01 7.04414129e-01
-7.40478188e-02 1.83515608e-01 -1.30205110e-01 1.23748653e-01
4.92999434e-01 9.45103541e-02 3.40501785e-01 -1.32172763e+00
-3.11851650e-01 -5.60621202e-01 -7.22706854e-01 -4.74132419e-01
-2.56137609e-01 -6.06310666e-01 9.05190315e-03 -1.15075088e+00
5.18149743e-03 -8.29478681e-01 -3.14547479e-01 4.70294207e-01
2.49900565e-01 5.30282140e-01 3.09996456e-01 8.75953659e-02
-2.88354337e-01 -7.97330141e-02 3.41090411e-01 -1.57043040e-01
-3.39871138e-01 3.24096590e-01 -9.95663464e-01 9.43278193e-01
6.58552647e-01 -4.17455047e-01 -4.12243903e-01 -5.94104469e-01
6.24221563e-01 8.99923965e-02 2.26936296e-01 -1.23634303e+00
1.73171818e-01 -6.80179819e-02 3.50923479e-01 -1.46618500e-01
3.57099086e-01 -1.25858057e+00 7.67619014e-01 3.36224258e-01
-2.30482548e-01 -6.92875236e-02 1.07592255e-01 3.50474983e-01
2.64158130e-01 -2.31876224e-01 9.16615725e-01 1.43466905e-01
-3.01076204e-01 -2.32271440e-02 -4.38901424e-01 -3.32632512e-01
5.84974945e-01 -7.02288151e-01 -4.28457186e-02 -2.92603791e-01
-4.08759326e-01 -6.57115221e-01 2.33683318e-01 3.64602536e-01
2.29602396e-01 -1.15946913e+00 -2.07445234e-01 9.16722864e-02
-3.07732314e-01 -3.97531301e-01 4.17835265e-01 3.99875462e-01
-5.86269975e-01 3.63753915e-01 -1.40015185e-01 -3.58247936e-01
-1.08311605e+00 1.67172462e-01 4.22495902e-01 1.43682405e-01
-6.42081678e-01 7.58764565e-01 -5.06745100e-01 5.34382823e-04
1.09334648e-01 -4.11803514e-01 1.08739920e-01 1.84934914e-01
6.11252964e-01 9.44993436e-01 4.84331816e-01 -6.56304300e-01
-5.91876566e-01 4.38589454e-01 6.25132561e-01 -8.24544951e-02
1.36359155e+00 -4.64370511e-02 1.31712526e-01 3.21384966e-01
1.03796995e+00 1.14830937e-02 -9.78888273e-01 1.32877409e-01
-2.25336030e-01 -5.49840689e-01 5.08233130e-01 -6.04583323e-01
-1.09343863e+00 6.30056918e-01 1.08391631e+00 6.79879725e-01
1.27058709e+00 -1.16188169e-01 9.23371196e-01 2.42735207e-01
6.03183866e-01 -1.39304674e+00 -3.41569096e-01 -1.67374566e-01
2.92031556e-01 -8.06111753e-01 2.61056423e-01 -1.05037568e-02
-2.66143382e-01 8.36607754e-01 5.52782454e-02 -1.83660001e-01
1.23034477e+00 1.04444003e+00 -5.79989962e-02 2.64459401e-02
-4.17992532e-01 6.19509965e-02 -1.05586328e-01 5.81762195e-01
2.26841420e-01 4.78443265e-01 -3.30736130e-01 1.16766191e+00
-8.46560657e-01 -1.12825915e-01 3.06919843e-01 9.06145811e-01
-6.14536822e-01 -4.50036913e-01 -3.74375224e-01 8.66935432e-01
-1.08733737e+00 -1.12855785e-01 2.57622600e-01 3.19597334e-01
3.51373881e-01 1.45111406e+00 2.96528041e-01 -6.88203216e-01
2.49916211e-01 6.70130923e-02 3.15321475e-01 -1.18373610e-01
-4.13129151e-01 3.03621083e-01 2.80453742e-01 -2.70793200e-01
-5.08314252e-01 -6.46906137e-01 -1.06790304e+00 -5.01678109e-01
-5.62795639e-01 -1.67622156e-02 1.20522571e+00 1.07201028e+00
2.44229212e-01 3.58724594e-01 5.65840840e-01 -8.36228430e-01
-7.52613246e-01 -8.21092665e-01 -1.12494016e+00 -2.79324740e-01
-8.08691699e-03 -3.68363768e-01 -6.99641883e-01 -2.98043881e-02]
|
[14.573345184326172, 5.464269638061523]
|
c383d237-3f2f-4c15-b0d2-04bdb747e17e
|
itervm-iterative-vision-modeling-module-for
|
2204.0263
| null |
https://arxiv.org/abs/2204.02630v1
|
https://arxiv.org/pdf/2204.02630v1.pdf
|
IterVM: Iterative Vision Modeling Module for Scene Text Recognition
|
Scene text recognition (STR) is a challenging problem due to the imperfect imagery conditions in natural images. State-of-the-art methods utilize both visual cues and linguistic knowledge to tackle this challenging problem. Specifically, they propose iterative language modeling module (IterLM) to repeatedly refine the output sequence from the visual modeling module (VM). Though achieving promising results, the vision modeling module has become the performance bottleneck of these methods. In this paper, we newly propose iterative vision modeling module (IterVM) to further improve the STR accuracy. Specifically, the first VM directly extracts multi-level features from the input image, and the following VMs re-extract multi-level features from the input image and fuse them with the high-level (i.e., the most semantic one) feature extracted by the previous VM. By combining the proposed IterVM with iterative language modeling module, we further propose a powerful scene text recognizer called IterNet. Extensive experiments demonstrate that the proposed IterVM can significantly improve the scene text recognition accuracy, especially on low-quality scene text images. Moreover, the proposed scene text recognizer IterNet achieves new state-of-the-art results on several public benchmarks. Codes will be available at https://github.com/VDIGPKU/IterNet.
|
['Yongtao Wang', 'Xiaojie Chu']
|
2022-04-06
| null | null | null | null |
['scene-text-recognition']
|
['computer-vision']
|
[ 2.56463289e-01 -6.12184465e-01 -1.66465849e-01 -7.49170780e-02
-4.23119456e-01 -2.70835191e-01 7.12253332e-01 -1.18151635e-01
-3.97565216e-01 1.34888273e-02 1.32460594e-01 -2.79247165e-01
3.43447179e-01 -6.58294380e-01 -5.38499415e-01 -5.96045196e-01
9.02499437e-01 4.73842137e-02 4.06326741e-01 1.04326203e-01
5.68436205e-01 1.95320904e-01 -1.50372362e+00 6.01310492e-01
1.00432432e+00 8.98913026e-01 7.31746793e-01 6.20909870e-01
-7.25637317e-01 1.13566136e+00 -1.03075810e-01 -9.78042036e-02
-5.13110273e-02 -2.54286110e-01 -6.96125150e-01 5.39487839e-01
3.80239546e-01 -4.08606499e-01 -6.59419775e-01 1.17307007e+00
1.40045360e-01 1.86383218e-01 4.99568731e-01 -1.09407699e+00
-6.42286956e-01 3.81535709e-01 -8.75325441e-01 6.89686313e-02
2.82705009e-01 3.26777399e-01 7.74017811e-01 -1.40218055e+00
3.76229882e-01 1.19184363e+00 1.74558252e-01 1.66456342e-01
-9.52736616e-01 -5.67378998e-01 5.46649575e-01 5.07279694e-01
-1.67078960e+00 -5.77652097e-01 8.91930938e-01 -5.00521064e-01
8.26611340e-01 1.56253219e-01 3.49966139e-01 7.53627777e-01
-1.10554164e-02 1.30299461e+00 1.05592322e+00 -5.33213079e-01
-5.08897305e-02 2.13476658e-01 4.04361129e-01 1.06492949e+00
7.10242242e-02 -3.29620242e-01 -6.02579534e-01 1.60511389e-01
7.31156230e-01 3.24789792e-01 -3.08878958e-01 -1.24282867e-01
-1.22433710e+00 4.92769688e-01 4.34299082e-01 4.81722027e-01
-3.42342228e-01 6.33110339e-03 3.34813446e-01 -6.78200424e-02
1.40168816e-01 -2.80325741e-01 5.22335693e-02 -2.13181302e-02
-1.14318073e+00 -2.61008412e-01 4.94942009e-01 7.94693053e-01
9.15529966e-01 2.51391977e-01 -2.92973220e-01 1.02375579e+00
6.48098409e-01 7.44689405e-01 5.20039141e-01 -2.51702487e-01
7.59033978e-01 9.03819978e-01 -1.86652884e-01 -1.27292979e+00
-1.57475486e-01 -2.76818454e-01 -9.22437549e-01 -2.46323973e-01
2.04354040e-02 2.75038183e-01 -1.21801782e+00 9.75350976e-01
2.84164816e-01 3.55279267e-01 1.54074878e-01 9.92712259e-01
1.24357688e+00 1.17953336e+00 1.82688564e-01 -1.14745565e-01
1.26608825e+00 -1.44739342e+00 -5.01432717e-01 -5.05760372e-01
5.82911313e-01 -1.04443085e+00 1.25333858e+00 3.41174811e-01
-7.61815488e-01 -7.84548700e-01 -9.07664180e-01 -1.93465441e-01
-1.60296619e-01 8.99569750e-01 2.11053491e-01 3.31398875e-01
-8.30954671e-01 -1.10311024e-01 -9.41326618e-01 -5.49385726e-01
3.75221610e-01 1.04149252e-01 -1.32069230e-01 -4.51169461e-01
-6.08850896e-01 5.24051964e-01 4.81707871e-01 3.13565582e-01
-8.53101730e-01 -2.79442400e-01 -9.04341877e-01 1.06062051e-02
5.68082988e-01 -4.98399317e-01 1.02516758e+00 -1.01643729e+00
-1.44811857e+00 7.34765053e-01 -6.24038935e-01 -6.86256289e-02
4.71612096e-01 -1.65564463e-01 -3.16721052e-01 2.05470726e-01
4.56660474e-03 4.58461523e-01 1.05173290e+00 -1.39145541e+00
-6.18986249e-01 -1.70206711e-01 -2.65090972e-01 4.73122776e-01
-4.88002062e-01 -2.59847138e-02 -1.05276704e+00 -5.92653155e-01
2.86853403e-01 -5.12870431e-01 -5.79167679e-02 -1.27902264e-02
-5.91016769e-01 -2.24926844e-01 1.09527087e+00 -7.28640795e-01
1.20010793e+00 -2.33671403e+00 3.24585848e-02 -2.07630452e-02
2.17654452e-01 6.40592754e-01 -2.49753937e-01 3.59686941e-01
3.02307308e-01 -9.43603739e-02 -9.61320400e-02 -5.11249065e-01
-1.56058982e-01 -1.72404535e-02 -4.52188015e-01 4.05756563e-01
1.09839011e-02 1.03081799e+00 -4.26400483e-01 -8.31077039e-01
8.70632529e-01 3.61445844e-01 -2.27626354e-01 2.30117753e-01
-3.15124810e-01 2.34839216e-01 -6.40564740e-01 7.40539193e-01
9.43305910e-01 -5.90712249e-01 -1.10894285e-01 -4.01489764e-01
-3.59745294e-01 -3.83134373e-02 -1.14804435e+00 1.65983319e+00
-3.14666808e-01 6.63168728e-01 -1.85480639e-01 -1.04319322e+00
1.08672500e+00 -1.69781193e-01 2.14083984e-01 -8.25281382e-01
3.04792196e-01 -9.90344584e-02 -4.14609402e-01 -7.27482617e-01
6.42237067e-01 1.53601080e-01 1.84260935e-01 2.77274579e-01
-1.45545185e-01 5.04007041e-02 9.97895226e-02 3.58432978e-01
7.42896616e-01 1.22057632e-01 3.57748717e-01 8.65561068e-02
1.14142787e+00 2.71221220e-01 4.12333369e-01 7.09913135e-01
-1.38021305e-01 5.11586785e-01 1.12437099e-01 -3.12885165e-01
-8.86402190e-01 -1.02162683e+00 5.56734763e-02 9.91892397e-01
4.95502681e-01 -5.41886866e-01 -5.79694867e-01 -5.16358376e-01
-7.61252716e-02 7.38872051e-01 -3.15820098e-01 4.46722806e-02
-2.81928122e-01 -5.42792320e-01 5.64321101e-01 4.54766363e-01
1.07599819e+00 -7.82027781e-01 -3.31290811e-01 2.84954887e-02
-2.69434154e-01 -1.53214693e+00 -6.25330985e-01 -3.23492289e-01
-6.56186461e-01 -8.34541023e-01 -6.01854682e-01 -9.76441920e-01
7.73982108e-01 8.09203625e-01 4.49730545e-01 3.75828564e-01
-3.44529301e-01 4.55875993e-01 -5.66583753e-01 5.59638953e-03
-2.95774072e-01 -2.93648411e-02 -2.41838560e-01 5.04364550e-01
3.27997863e-01 -1.03369117e-01 -3.52916569e-01 2.50671685e-01
-1.00668287e+00 8.92077386e-01 6.81250453e-01 7.36943603e-01
6.68117642e-01 1.85840204e-01 7.07347840e-02 -5.06969690e-01
2.26307273e-01 -1.26243040e-01 -7.53628314e-01 5.14443576e-01
-2.82510549e-01 -6.28576949e-02 8.35675001e-01 -3.61316979e-01
-1.21005392e+00 3.18200171e-01 -4.01336551e-02 -6.37072742e-01
-3.02249312e-01 7.39738762e-01 -1.18349276e-01 -1.65539280e-01
1.79536343e-01 1.09334898e+00 -2.85968244e-01 -3.94524664e-01
2.73337960e-01 1.03329337e+00 4.85598117e-01 -3.80222023e-01
8.35105479e-01 5.94674230e-01 -2.97334999e-01 -1.33194029e+00
-9.28991616e-01 -6.61363721e-01 -5.63314199e-01 -2.70804226e-01
8.21019650e-01 -1.12794042e+00 -6.65664494e-01 7.98636079e-01
-1.13774955e+00 -3.98395389e-01 3.40265661e-01 4.43055063e-01
-2.43607938e-01 9.23517048e-01 -4.81331170e-01 -8.09774280e-01
-4.33068186e-01 -1.22309124e+00 1.23705661e+00 4.25358981e-01
4.79616374e-01 -8.96476090e-01 -2.38779411e-01 5.59395909e-01
1.75054550e-01 -2.80264825e-01 7.26762772e-01 -3.01219344e-01
-7.98757017e-01 6.24441765e-02 -8.10572803e-01 2.73746550e-01
1.38036624e-01 9.33252200e-02 -9.28272784e-01 -1.89848706e-01
-1.00031987e-01 -1.51590854e-01 1.19409657e+00 2.02352181e-01
1.12111735e+00 -7.98549876e-02 -2.99646467e-01 6.92100346e-01
1.58346212e+00 5.19665182e-02 5.80183923e-01 3.56088847e-01
1.36280334e+00 2.11694852e-01 7.40570724e-01 5.41841984e-01
6.93703294e-01 5.59201598e-01 1.55022293e-01 -1.53372332e-01
-2.51878798e-01 -3.46114427e-01 6.69760466e-01 1.15750051e+00
4.22333300e-01 -3.07905912e-01 -1.28556180e+00 3.31519276e-01
-2.17155409e+00 -7.43292451e-01 -2.89060205e-01 1.82853830e+00
4.23593521e-01 3.10597066e-02 -2.97649771e-01 -3.79228871e-03
8.34282696e-01 4.07333553e-01 -5.56994438e-01 -1.17893457e-01
-2.30086535e-01 -1.48862198e-01 3.41105193e-01 4.28176641e-01
-1.12254417e+00 1.49257064e+00 4.79562140e+00 1.09122467e+00
-1.36099958e+00 -8.08603615e-02 4.00600225e-01 7.14953691e-02
-8.61210898e-02 1.15583390e-01 -9.36900079e-01 3.88149142e-01
3.85009915e-01 -4.24110055e-01 5.08683443e-01 8.11268508e-01
4.85161573e-01 -2.55810738e-01 -8.11344028e-01 1.42990947e+00
6.13946736e-01 -1.35035968e+00 5.70619285e-01 -2.00675249e-01
6.57158673e-01 6.13394417e-02 9.81907174e-02 1.94717318e-01
-9.50260460e-02 -9.23734844e-01 6.56248093e-01 6.79485917e-01
7.65978456e-01 -6.29762769e-01 5.35161614e-01 5.70581555e-01
-1.75028014e+00 3.76772583e-02 -5.64548910e-01 1.63087308e-01
-2.95047020e-03 6.72165215e-01 -7.80208409e-01 7.24615335e-01
4.87654656e-01 1.13914502e+00 -8.87779534e-01 1.03049624e+00
-1.82565883e-01 6.44905508e-01 -2.28146672e-01 -1.97653010e-01
4.81934458e-01 -1.08684108e-01 4.35984373e-01 1.30940068e+00
1.75005510e-01 5.19719534e-02 7.06660211e-01 9.12782013e-01
-9.10941139e-02 5.27537167e-01 -4.95930254e-01 -1.82657778e-01
1.67548731e-01 1.30252695e+00 -7.40996599e-01 -5.52141488e-01
-6.07801557e-01 1.17826056e+00 2.46111125e-01 4.74446863e-01
-7.41932571e-01 -2.89307088e-01 2.89321512e-01 -2.90294558e-01
3.91974777e-01 -5.18101275e-01 -3.32670867e-01 -1.61910832e+00
1.62311450e-01 -8.53574514e-01 3.65220129e-01 -1.11902726e+00
-8.59326720e-01 3.77031326e-01 -3.32354337e-01 -1.14970315e+00
2.38955408e-01 -6.71230793e-01 -5.64738214e-01 8.82346094e-01
-1.63782382e+00 -1.54659915e+00 -7.60282218e-01 8.23242426e-01
1.15971959e+00 -1.80599689e-01 2.55208135e-01 1.04255266e-01
-8.39002550e-01 4.36416119e-01 1.91171542e-01 3.24663699e-01
4.52294558e-01 -6.64229810e-01 3.80124211e-01 1.12815666e+00
4.06950861e-01 3.64498526e-01 1.84483528e-01 -7.36298025e-01
-1.86184847e+00 -1.41207695e+00 5.34312546e-01 -8.73598829e-02
6.25066459e-01 -3.47626984e-01 -1.03166485e+00 6.06293619e-01
8.99609849e-02 -1.77844629e-01 1.82209015e-01 -4.48317975e-01
-3.92718792e-01 9.93225649e-02 -7.14631140e-01 8.00637066e-01
7.50699997e-01 -7.15155959e-01 -5.43963790e-01 1.53017491e-01
6.28344953e-01 -4.27671909e-01 -3.08732867e-01 2.53977090e-01
3.57228041e-01 -6.45482302e-01 7.29315996e-01 -1.16331629e-01
3.90803248e-01 -6.75542653e-01 -3.03766817e-01 -7.76025414e-01
-2.20624298e-01 -9.00727808e-02 9.10170525e-02 1.29125333e+00
6.44170791e-02 -5.54890811e-01 6.68873489e-01 1.66601002e-01
-9.63609442e-02 -4.69913572e-01 -6.34381056e-01 -5.69092035e-01
-1.67847857e-01 -7.33589530e-01 1.86481655e-01 9.52048481e-01
-9.03599188e-02 3.40683222e-01 -4.49940562e-01 3.69822919e-01
7.08562434e-01 4.19049472e-01 9.35100555e-01 -7.70051599e-01
-1.96885735e-01 -4.26513225e-01 -4.73161161e-01 -1.50551105e+00
2.67048150e-01 -1.15505457e+00 1.22925855e-01 -1.75264955e+00
6.32486105e-01 -7.00533092e-02 -1.31635413e-01 5.05274355e-01
-2.91588455e-01 -3.14655341e-02 6.28237605e-01 3.37918907e-01
-9.01970327e-01 8.31021547e-01 1.20993733e+00 -4.75962728e-01
-2.38689005e-01 -4.45881218e-01 -4.92436975e-01 7.73101807e-01
7.91918635e-01 -2.87241250e-01 -2.28922501e-01 -7.29825616e-01
-2.39049524e-01 -2.65550371e-02 6.41108036e-01 -9.49734569e-01
7.44229198e-01 -4.18683320e-01 4.37980890e-01 -1.08388340e+00
2.36303359e-01 -5.54118276e-01 -1.28297538e-01 4.26726431e-01
-2.57251322e-01 -2.54825115e-01 4.26275909e-01 4.53273684e-01
-2.21068650e-01 -1.90645799e-01 7.94939995e-01 1.46268029e-02
-1.10842574e+00 2.72949010e-01 -4.40489829e-01 -1.68403313e-01
7.93427527e-01 -2.12613061e-01 -5.61226428e-01 -1.05236903e-01
-4.63319600e-01 4.02648032e-01 5.65738916e-01 5.36837518e-01
1.08094084e+00 -1.08456492e+00 -7.87988544e-01 1.71589181e-01
3.83620828e-01 -6.58983514e-02 4.60225224e-01 9.86020327e-01
-5.48695385e-01 4.86656487e-01 1.27486438e-01 -9.78516102e-01
-1.48440468e+00 5.66583991e-01 1.99159116e-01 -2.21311435e-01
-7.88828135e-01 4.88746345e-01 5.68402588e-01 -2.25381762e-01
2.37621501e-01 -1.75226226e-01 -1.02495238e-01 -3.73321295e-01
6.06559217e-01 1.23225763e-01 -1.42308384e-01 -9.30823028e-01
-4.83780205e-01 9.98514056e-01 -4.01114911e-01 -5.19565754e-02
9.72202599e-01 -3.46800655e-01 -1.08638234e-01 4.85008985e-01
1.07793117e+00 -2.58733183e-01 -8.86122286e-01 -7.14524031e-01
-1.83694884e-01 -6.64957285e-01 3.72349441e-01 -5.17070353e-01
-1.04083622e+00 1.06255162e+00 5.16185343e-01 -3.30707163e-01
1.21480560e+00 -1.89140961e-01 7.45588005e-01 7.06334591e-01
1.85936689e-01 -1.06316912e+00 3.45820069e-01 6.59052908e-01
7.45100677e-01 -1.26890230e+00 1.46796405e-01 -5.27705014e-01
-8.12038481e-01 1.09455287e+00 6.96889460e-01 -2.39747334e-02
4.55557108e-01 1.51425600e-01 1.08723491e-01 -5.67486919e-02
-6.77947283e-01 -5.04156411e-01 3.22318763e-01 2.41481498e-01
2.55030453e-01 -3.97811718e-02 -1.18375257e-01 4.15241152e-01
2.51117438e-01 7.31024221e-02 4.75670248e-01 8.36705089e-01
-6.51760221e-01 -8.24871719e-01 -5.77253163e-01 5.12150466e-01
-7.68422931e-02 -3.63812387e-01 -4.15127128e-01 4.41123843e-01
-2.27224559e-01 1.12512517e+00 -9.77592990e-02 -6.04166508e-01
2.25667313e-01 -3.82582359e-02 2.67111063e-01 -5.41891277e-01
-3.30498576e-01 3.17018211e-01 -1.74900278e-01 -4.32329446e-01
-3.12796175e-01 -7.14176774e-01 -1.49760902e+00 -2.95913160e-01
-3.08238000e-01 -1.52338520e-01 7.07384944e-01 1.00580561e+00
2.34292611e-01 3.64460886e-01 7.66802967e-01 -6.32874846e-01
-1.18291058e-01 -7.92071044e-01 -3.59548360e-01 2.07629561e-01
2.18331829e-01 -4.09608603e-01 -2.48869583e-01 4.20219451e-01]
|
[11.801560401916504, 2.070963144302368]
|
f02dffed-3a32-4afe-b044-2c2c4380767b
|
temporal-action-proposal-generation-with-1
|
2112.07984
| null |
https://arxiv.org/abs/2112.07984v1
|
https://arxiv.org/pdf/2112.07984v1.pdf
|
Temporal Action Proposal Generation with Background Constraint
|
Temporal action proposal generation (TAPG) is a challenging task that aims to locate action instances in untrimmed videos with temporal boundaries. To evaluate the confidence of proposals, the existing works typically predict action score of proposals that are supervised by the temporal Intersection-over-Union (tIoU) between proposal and the ground-truth. In this paper, we innovatively propose a general auxiliary Background Constraint idea to further suppress low-quality proposals, by utilizing the background prediction score to restrict the confidence of proposals. In this way, the Background Constraint concept can be easily plug-and-played into existing TAPG methods (e.g., BMN, GTAD). From this perspective, we propose the Background Constraint Network (BCNet) to further take advantage of the rich information of action and background. Specifically, we introduce an Action-Background Interaction module for reliable confidence evaluation, which models the inconsistency between action and background by attention mechanisms at the frame and clip levels. Extensive experiments are conducted on two popular benchmarks, i.e., ActivityNet-1.3 and THUMOS14. The results demonstrate that our method outperforms state-of-the-art methods. Equipped with the existing action classifier, our method also achieves remarkable performance on the temporal action localization task.
|
['Hujie Huang', 'Hongxun Yao', 'Boyang xia', 'Sheng Jin', 'Lining Wang', 'Wenhao Wu', 'Haosen Yang']
|
2021-12-15
| null | null | null | null |
['temporal-action-proposal-generation', 'action-localization']
|
['computer-vision', 'computer-vision']
|
[ 2.93852329e-01 -1.93441305e-02 -4.28423643e-01 -1.62572518e-01
-6.54293716e-01 -3.92916203e-02 5.99388421e-01 -1.72818869e-01
-3.41358542e-01 6.74389780e-01 3.22067678e-01 1.35305017e-01
2.62674272e-01 -4.72832561e-01 -5.62549591e-01 -8.24309826e-01
2.02294856e-01 -8.71425346e-02 9.33006465e-01 -1.07934652e-02
1.36587813e-01 5.23611754e-02 -1.46752310e+00 3.94712359e-01
8.46853375e-01 1.24152279e+00 1.69909522e-01 1.74864009e-01
-9.94371716e-03 1.13554239e+00 -5.08472025e-01 -1.75108775e-01
2.66661674e-01 -7.05454707e-01 -3.85307133e-01 3.10065895e-01
3.64651084e-01 -5.21489322e-01 -4.31444407e-01 1.06102073e+00
3.63022804e-01 4.16768134e-01 2.56061435e-01 -1.35295355e+00
-6.19502105e-02 3.81559491e-01 -7.30190575e-01 5.75650930e-01
3.87806058e-01 4.95371938e-01 8.28947186e-01 -1.05655289e+00
6.87846422e-01 1.30691671e+00 3.39285821e-01 5.82867265e-01
-9.34563160e-01 -6.49829447e-01 9.30058300e-01 6.37426436e-01
-1.41351104e+00 -5.03628254e-01 8.07511508e-01 -3.33958685e-01
5.27159214e-01 1.44741729e-01 7.48724282e-01 1.16768694e+00
7.88252279e-02 1.21402788e+00 8.15458179e-01 -1.57335714e-01
3.64181459e-01 -2.72826701e-01 -2.79327482e-01 6.15944922e-01
-5.12917377e-02 -1.52552396e-01 -7.33126938e-01 5.40776849e-02
6.77710116e-01 -1.36954682e-02 -4.80299503e-01 -2.87799567e-01
-1.38629115e+00 4.11478549e-01 2.10189492e-01 2.27992639e-01
-4.59635139e-01 2.59057254e-01 4.83585954e-01 -3.49415660e-01
5.61162889e-01 -2.43941262e-01 -2.57722884e-01 -1.79411426e-01
-9.49837565e-01 1.37415767e-01 2.79896617e-01 1.11958086e+00
4.87334251e-01 6.32832795e-02 -9.58099008e-01 5.41433096e-01
3.72675925e-01 1.68443710e-01 4.71971065e-01 -1.05349183e+00
6.18932188e-01 6.46221042e-01 4.11393553e-01 -1.07382572e+00
-3.79627757e-02 -3.62069726e-01 -6.37665689e-01 1.51285604e-01
2.71922827e-01 6.22055456e-02 -8.75739038e-01 1.66487789e+00
7.61091053e-01 8.92478764e-01 -2.33370394e-01 1.09980345e+00
7.01081336e-01 5.82353830e-01 2.96003908e-01 -5.70978582e-01
1.09654379e+00 -1.39876699e+00 -8.90306830e-01 -1.99177057e-01
4.74372923e-01 -5.61760187e-01 9.34379995e-01 5.08789718e-01
-1.09469378e+00 -9.08549190e-01 -9.89796877e-01 1.99569404e-01
7.52090961e-02 5.14118969e-01 4.97493744e-01 2.49656975e-01
-5.06339967e-01 5.90252638e-01 -1.16072118e+00 -2.42538214e-01
6.12348914e-01 1.63853215e-03 -2.77752072e-01 -1.17560312e-01
-9.86839473e-01 6.94316268e-01 6.92562640e-01 4.20711339e-01
-1.11599910e+00 -3.70593667e-01 -7.19406486e-01 -4.95612472e-02
9.48240697e-01 -4.36315209e-01 1.23525405e+00 -1.23382735e+00
-1.62420642e+00 4.46728081e-01 -1.31804213e-01 -4.26238000e-01
8.48888814e-01 -3.59870821e-01 -4.01273519e-01 2.92846531e-01
2.00897306e-01 6.45092726e-01 8.38368595e-01 -9.26873565e-01
-1.02913141e+00 8.82663485e-03 2.06651136e-01 2.43506908e-01
-1.28128260e-01 9.31065828e-02 -1.03345704e+00 -8.49148929e-01
3.47605944e-01 -7.05592513e-01 -4.01777208e-01 1.53610542e-01
-3.11966181e-01 -5.11326790e-01 6.78816497e-01 -5.91587663e-01
1.59336066e+00 -2.15854526e+00 8.09955224e-02 -9.39120054e-02
1.68728773e-02 3.97645354e-01 -9.94750783e-02 -5.69783188e-02
1.30139768e-01 -1.87445566e-01 8.81620334e-04 -3.29357594e-01
-4.56220098e-02 1.50155440e-01 -1.21539466e-01 5.98991156e-01
3.10344994e-01 7.52293348e-01 -1.15214467e+00 -9.86729205e-01
4.17879850e-01 1.81777105e-01 -6.09749973e-01 2.55509555e-01
-5.60822129e-01 7.60459006e-01 -7.38713086e-01 7.86920846e-01
6.20311737e-01 -9.60929021e-02 8.99108797e-02 -3.52668166e-01
-2.54329383e-01 2.29560196e-01 -1.44005215e+00 1.95459414e+00
-9.69643742e-02 2.58979052e-01 -1.89824216e-02 -8.76655281e-01
7.05019116e-01 2.88224906e-01 6.41090512e-01 -6.02642536e-01
1.02095149e-01 3.25420238e-02 -2.37567704e-02 -5.26854396e-01
2.52205491e-01 1.37932792e-01 1.60376564e-01 5.52134551e-02
3.22415642e-02 4.35354352e-01 4.37750489e-01 1.31360009e-01
1.15566897e+00 8.25264215e-01 3.89536887e-01 9.11379308e-02
9.26917911e-01 -2.69106090e-01 1.25810635e+00 7.22544670e-01
-6.13646805e-01 5.87212443e-01 5.56231380e-01 -3.90823603e-01
-5.04662097e-01 -7.89238513e-01 3.21745574e-01 1.17202568e+00
5.05246699e-01 -6.17936850e-01 -6.34676933e-01 -1.05976701e+00
-3.91644895e-01 5.03189564e-01 -6.98126733e-01 -1.61930904e-01
-8.35172832e-01 -5.08309662e-01 2.21017137e-01 7.45721042e-01
7.32703209e-01 -1.20424592e+00 -6.04584634e-01 4.77318078e-01
-4.45104599e-01 -1.24445879e+00 -7.64259398e-01 -2.70575792e-01
-7.57976115e-01 -1.04954517e+00 -6.65427625e-01 -3.47453177e-01
4.80870754e-01 3.03889036e-01 9.14194107e-01 2.36221060e-01
-5.91001585e-02 3.32979172e-01 -6.20231509e-01 -1.03369899e-01
-9.54778194e-02 -2.78032720e-01 7.48573570e-04 4.38572466e-01
2.92579591e-01 -5.62655091e-01 -9.15329397e-01 6.65326238e-01
-8.11117649e-01 3.94609600e-01 6.95048988e-01 6.64846659e-01
1.02401567e+00 -6.69807103e-03 3.96533847e-01 -3.52967054e-01
-2.50021853e-02 -3.45153511e-01 -5.86099744e-01 4.21249449e-01
-2.21006006e-01 -1.90969080e-01 2.76213080e-01 -6.69733107e-01
-1.33649898e+00 2.06399068e-01 3.09845805e-02 -8.85957837e-01
-1.01992190e-01 3.17738712e-01 -6.07994497e-01 2.13028774e-01
2.82971233e-01 2.64224350e-01 -4.52345550e-01 -4.03297752e-01
3.38312507e-01 1.68488234e-01 6.98547482e-01 -5.26207447e-01
4.56139445e-01 7.19056726e-01 -4.97431383e-02 -2.55133837e-01
-1.12253666e+00 -6.40851915e-01 -5.97113788e-01 -5.92936575e-01
9.90212679e-01 -9.64364290e-01 -5.87278306e-01 3.52809131e-01
-1.26506579e+00 -3.51669282e-01 -1.92369819e-01 6.87253594e-01
-6.28470659e-01 6.38117969e-01 -2.88651764e-01 -1.04941821e+00
-1.15043953e-01 -1.02694035e+00 1.11617732e+00 2.64132768e-01
1.32950276e-01 -4.20725048e-01 -8.54386464e-02 2.86664993e-01
4.46907654e-02 4.64911252e-01 1.73838988e-01 -4.27715331e-01
-9.49743450e-01 -5.14575504e-02 -2.66250610e-01 4.38434839e-01
-1.52342105e-02 -1.35913091e-02 -8.17383766e-01 -8.44381005e-02
1.45984283e-02 -1.69128314e-01 1.19930780e+00 4.24446791e-01
1.34214520e+00 -2.59558827e-01 -4.68142122e-01 3.67919862e-01
1.04569793e+00 3.04975152e-01 8.54216874e-01 1.76173553e-01
6.27705872e-01 4.40307975e-01 1.44472373e+00 7.56497145e-01
1.30542204e-01 1.08771956e+00 5.72168052e-01 1.55832529e-01
-1.55814499e-01 -3.91962349e-01 5.42458117e-01 4.49977875e-01
-4.30791974e-01 -3.20676416e-01 -4.69913453e-01 2.76720375e-01
-2.42504120e+00 -1.16472900e+00 -4.80269045e-02 2.09904242e+00
6.13448501e-01 5.89852035e-01 1.77499130e-01 -1.04876474e-01
9.52676296e-01 3.49151075e-01 -5.90984344e-01 4.68017608e-01
-2.04827506e-02 -6.19340912e-02 1.95535123e-01 1.57970950e-01
-1.48133004e+00 9.36708689e-01 4.91751719e+00 1.19673121e+00
-7.35267162e-01 3.95243526e-01 6.35058641e-01 -2.02311948e-01
1.80052087e-01 2.63465494e-02 -8.98658633e-01 7.91917980e-01
2.51030266e-01 -3.41538601e-02 3.73551734e-02 9.73812640e-01
6.96052015e-01 -3.57327491e-01 -1.10936666e+00 8.64291906e-01
1.10250153e-01 -1.11645508e+00 -1.97983176e-01 -2.67143846e-01
6.08651042e-01 -2.67538130e-01 -2.08633378e-01 5.56425393e-01
-1.04689427e-01 -5.22119522e-01 1.01111472e+00 7.63114393e-01
5.18303216e-01 -5.15844405e-01 7.11513102e-01 3.31677765e-01
-1.58632588e+00 -1.44618437e-01 -3.40134323e-01 7.32970238e-02
4.60658908e-01 5.14458001e-01 -4.05719668e-01 6.40662730e-01
6.40239120e-01 1.07240212e+00 -4.39137399e-01 1.21506202e+00
-5.56351006e-01 6.55729175e-01 -1.88231200e-01 1.91392988e-01
2.48970538e-01 -1.26616254e-01 5.25280774e-01 1.14848912e+00
3.02527159e-01 4.65907931e-01 6.13122225e-01 9.01092947e-01
1.42831579e-01 9.15974006e-02 -4.22762483e-02 1.31883875e-01
2.61110455e-01 1.31527257e+00 -8.98055851e-01 -5.79958975e-01
-4.88660216e-01 1.02869236e+00 3.10576484e-02 2.62308300e-01
-1.37812948e+00 3.12195495e-02 4.37466890e-01 6.47567809e-02
5.84702313e-01 -5.21894321e-02 2.75374115e-01 -1.34753883e+00
4.24951613e-01 -6.85160160e-01 4.78030384e-01 -8.45607162e-01
-1.04039812e+00 3.31056893e-01 1.66994095e-01 -1.72436965e+00
1.39933079e-01 -2.80845731e-01 -9.23834860e-01 5.05153179e-01
-1.25510061e+00 -1.13889813e+00 -6.00511968e-01 5.76977432e-01
8.86393249e-01 1.76075473e-01 5.35056964e-02 4.55712914e-01
-8.63637209e-01 2.61714339e-01 -4.79036957e-01 4.90585156e-02
7.04943001e-01 -8.29444706e-01 6.46416396e-02 1.07265615e+00
8.90973806e-02 3.28682393e-01 5.26980817e-01 -8.87731731e-01
-9.03576732e-01 -1.32340050e+00 3.91855568e-01 -3.39638978e-01
5.98121226e-01 -1.54973090e-01 -8.16739023e-01 5.16501665e-01
-7.06254393e-02 4.34842557e-01 2.48510882e-01 -2.86193848e-01
-5.90426512e-02 -1.78794533e-01 -7.54803181e-01 6.31902397e-01
1.39995122e+00 -7.72755891e-02 -4.67902988e-01 4.16900456e-01
8.61207604e-01 -5.51913679e-01 -5.38891554e-01 6.59481108e-01
5.40874362e-01 -1.27811182e+00 9.06135738e-01 -5.32383263e-01
3.94181430e-01 -8.19443762e-01 -1.20027341e-01 -6.85100436e-01
-3.04901719e-01 -6.34024501e-01 -4.65845525e-01 1.31070590e+00
-4.87590171e-02 -5.52784018e-02 9.13450956e-01 3.68147612e-01
-4.23413962e-01 -9.08587277e-01 -1.20330548e+00 -7.95483887e-01
-6.74810588e-01 -5.85423172e-01 3.87271404e-01 5.66798747e-01
-1.27821848e-01 5.10783168e-03 -5.59670329e-01 6.50533214e-02
3.40426892e-01 -9.76546630e-02 7.78619409e-01 -8.41926217e-01
-4.95580852e-01 -4.14005756e-01 -5.28434098e-01 -1.32683694e+00
1.89828172e-01 -4.12776619e-01 3.64779413e-01 -1.43668938e+00
1.97553113e-01 -2.01029956e-01 -6.63969338e-01 5.26641250e-01
-4.75996912e-01 2.14922190e-01 2.63542324e-01 1.55521855e-01
-1.39909756e+00 8.97886217e-01 1.36317432e+00 -1.97170392e-01
-3.13421994e-01 1.10862911e-01 -1.59516498e-01 1.11479437e+00
6.40552819e-01 -4.36633110e-01 -3.66841793e-01 -3.46325748e-02
-3.94175537e-02 1.01802588e-01 5.18268406e-01 -1.29493737e+00
2.62888759e-01 -5.93319774e-01 2.50568360e-01 -9.40153658e-01
3.63659352e-01 -5.73329449e-01 -3.90810752e-03 3.10694009e-01
-2.39345267e-01 -2.76743352e-01 1.71511702e-03 1.05905664e+00
-2.27963746e-01 -9.70417410e-02 7.20341027e-01 -1.83769092e-01
-1.00394368e+00 4.68350440e-01 -2.18827263e-01 7.22560957e-02
1.20638371e+00 -1.82906702e-01 -2.32061207e-01 -1.30097836e-01
-8.46897304e-01 2.82614350e-01 2.01544359e-01 3.77304435e-01
7.26763904e-01 -1.63537264e+00 -6.28552437e-01 -9.10280868e-02
3.46552819e-01 -2.17049281e-04 4.31383401e-01 1.47300696e+00
-8.79881009e-02 1.04694173e-01 -2.75111981e-02 -6.62533343e-01
-1.21866488e+00 6.91719770e-01 3.10476363e-01 -3.46686184e-01
-7.12411523e-01 7.91729271e-01 4.90985900e-01 2.96648413e-01
4.49264795e-01 -5.32798946e-01 -2.44459853e-01 -5.22025265e-02
7.65449345e-01 5.10302126e-01 -2.70149618e-01 -6.31212056e-01
-4.78416800e-01 2.88125724e-01 1.59950122e-01 -1.27143100e-01
9.50834990e-01 -1.35170177e-01 3.69535647e-02 2.42706478e-01
6.99831605e-01 -8.16564038e-02 -1.69270754e+00 -2.24636003e-01
4.09420617e-02 -7.16360033e-01 -1.28963992e-01 -5.42249084e-01
-1.11386240e+00 6.95855916e-01 6.03496969e-01 -1.37474179e-01
1.24241507e+00 4.95838262e-02 5.47662675e-01 2.25176990e-01
4.05678779e-01 -1.34907842e+00 5.79789758e-01 1.75134748e-01
9.27270651e-01 -1.27591276e+00 1.99150816e-01 -5.03128171e-01
-6.52780831e-01 8.92903566e-01 1.15382254e+00 1.10749498e-01
4.12532687e-01 -1.38085514e-01 -3.53359222e-01 7.85784051e-02
-8.01803887e-01 -4.36488628e-01 4.07595545e-01 3.77310753e-01
2.32639536e-01 -1.73102006e-01 -6.59988165e-01 8.54763925e-01
7.43102431e-01 3.51228327e-01 3.11075956e-01 8.40486169e-01
-5.69881201e-01 -8.55688751e-01 -2.67724603e-01 2.35024527e-01
-3.64821434e-01 7.70492852e-02 -1.22960433e-01 6.26525819e-01
7.87212610e-01 1.05013382e+00 -1.51099116e-01 -3.71281922e-01
3.74897033e-01 -1.84283242e-01 4.38138336e-01 -6.19269967e-01
-2.86004126e-01 5.54325223e-01 1.28578693e-01 -1.05087602e+00
-1.02304780e+00 -7.43658841e-01 -1.17740333e+00 2.77337134e-01
-6.16566658e-01 -2.13895626e-02 6.70155808e-02 9.48077917e-01
2.19962373e-01 6.95226252e-01 3.54360163e-01 -1.04816294e+00
-2.46546805e-01 -1.04360592e+00 -5.09678125e-01 2.88517654e-01
-4.52392325e-02 -1.03591573e+00 -2.93856710e-01 1.83127031e-01]
|
[8.503103256225586, 0.55224609375]
|
b1bffea5-54d2-4996-b9e2-4db603b9449b
|
a-comparative-study-of-face-detection
|
2305.11077
| null |
https://arxiv.org/abs/2305.11077v1
|
https://arxiv.org/pdf/2305.11077v1.pdf
|
A Comparative Study of Face Detection Algorithms for Masked Face Detection
|
Contemporary face detection algorithms have to deal with many challenges such as variations in pose, illumination, and scale. A subclass of the face detection problem that has recently gained increasing attention is occluded face detection, or more specifically, the detection of masked faces. Three years on since the advent of the COVID-19 pandemic, there is still a complete lack of evidence regarding how well existing face detection algorithms perform on masked faces. This article first offers a brief review of state-of-the-art face detectors and detectors made for the masked face problem, along with a review of the existing masked face datasets. We evaluate and compare the performances of a well-representative set of face detectors at masked face detection and conclude with a discussion on the possible contributing factors to their performance.
|
['Subhankar Mishra', 'Danush Shekar', 'Sahel Mohammad Iqbal']
|
2023-05-18
| null | null | null | null |
['face-detection', 'occluded-face-detection']
|
['computer-vision', 'computer-vision']
|
[ 2.55283713e-01 -2.69751728e-01 6.35879040e-02 -3.87207031e-01
-4.24166441e-01 -5.40681601e-01 5.20860791e-01 -2.35012725e-01
-3.63688082e-01 3.54633421e-01 -3.67963850e-03 1.28098235e-01
2.54562289e-01 -3.16498041e-01 -2.78125674e-01 -7.74154723e-01
-3.49051297e-01 3.45576137e-01 4.09540124e-02 -8.17891434e-02
-2.50284262e-02 9.34216082e-01 -2.14129376e+00 2.61697829e-01
2.41889387e-01 7.13638008e-01 -7.11322650e-02 5.45360744e-01
1.67557463e-01 1.49859801e-01 -8.73006701e-01 -6.16375148e-01
2.93146163e-01 -4.09198940e-01 -2.96576142e-01 4.07903552e-01
8.56531143e-01 -6.39288187e-01 -1.22034281e-01 9.58633125e-01
8.30569267e-01 -4.65678304e-01 4.37505364e-01 -1.26555121e+00
-4.75203484e-01 2.34901085e-01 -9.00599122e-01 7.01764107e-01
4.71669436e-01 9.26627368e-02 1.63492382e-01 -1.44858658e+00
6.08618319e-01 1.80262446e+00 7.08451688e-01 7.27192283e-01
-1.06913996e+00 -9.86549258e-01 1.78056523e-01 6.95094094e-02
-1.87125242e+00 -9.91377771e-01 3.45295757e-01 -5.19945681e-01
7.57767379e-01 1.91549480e-01 3.90503556e-01 9.10313368e-01
-1.11121625e-01 2.10649326e-01 1.16134858e+00 -6.71328425e-01
-9.74660069e-02 4.38431978e-01 -2.25529253e-01 9.01540279e-01
7.44786620e-01 3.29307616e-01 -5.47844112e-01 -5.28599322e-01
5.97940683e-01 -3.05455565e-01 -2.85031013e-02 -1.57851905e-01
-4.88310337e-01 9.85750854e-01 -5.81415407e-02 3.48339379e-01
-1.05976745e-01 -4.42083210e-01 3.09682339e-01 6.02609068e-02
6.14026129e-01 -1.53193608e-01 -7.89261982e-02 6.42876208e-01
-1.00922298e+00 1.90700412e-01 4.51147228e-01 5.02758324e-01
5.44133186e-01 2.53213733e-01 -2.34733924e-01 7.38240063e-01
5.37336290e-01 6.37989044e-01 7.76147693e-02 -3.92142266e-01
-6.84801936e-02 3.74587089e-01 1.38107380e-02 -9.97279882e-01
-4.99532849e-01 -1.71957269e-01 -2.74120688e-01 4.97593790e-01
4.17928040e-01 -3.17999274e-01 -8.96594048e-01 1.63213742e+00
6.05256200e-01 1.97860032e-01 -5.13695478e-01 6.10896647e-01
1.09944022e+00 -7.74292722e-02 7.58574456e-02 -5.94689608e-01
1.83765185e+00 -3.68141443e-01 -8.05068791e-01 -4.62735683e-01
-7.59699121e-02 -1.20009255e+00 5.96598350e-02 1.01260595e-01
-7.94854462e-01 -6.40905499e-01 -1.02704036e+00 3.12916070e-01
-4.84211922e-01 3.10910463e-01 2.93671548e-01 1.70247054e+00
-1.11652851e+00 5.33371456e-02 -4.34798360e-01 -9.42894518e-01
5.47309518e-01 6.64218962e-01 -5.55780411e-01 -4.37985882e-02
-7.55804896e-01 1.18263507e+00 -3.22693214e-02 1.58291250e-01
-1.04022706e+00 -3.72274816e-01 -6.54648185e-01 -3.52991253e-01
4.52136129e-01 -3.04255515e-01 1.10896730e+00 -8.07310760e-01
-9.20220435e-01 1.73560512e+00 -5.85160315e-01 3.90016325e-02
3.71961623e-01 -7.80970901e-02 -1.02052557e+00 2.34496132e-01
7.58597925e-02 5.56388736e-01 1.33949709e+00 -1.08343530e+00
-4.40741062e-01 -8.55101824e-01 -4.09825832e-01 -7.96575621e-02
-1.13746695e-01 1.33061218e+00 -4.15752769e-01 -4.96960253e-01
-1.06209256e-01 -8.13755035e-01 1.45738572e-01 2.58098662e-01
-2.21175402e-01 -2.50984073e-01 9.75190580e-01 -5.38092017e-01
9.87284184e-01 -2.18656182e+00 -3.42163712e-01 -7.48706087e-02
2.29114845e-01 5.86534142e-01 -1.61145508e-01 4.26975757e-01
-3.74537438e-01 6.63990155e-02 1.04311025e-02 -6.18906260e-01
-1.88973427e-01 -8.00922886e-02 -2.01869011e-01 1.07607353e+00
4.06798393e-01 5.23645639e-01 -5.00967026e-01 -5.45757949e-01
1.35248616e-01 8.31372261e-01 -2.63300747e-01 1.13882460e-01
3.41492474e-01 4.03786272e-01 -1.08566076e-01 1.14628279e+00
1.10589349e+00 3.15500528e-01 2.21563920e-01 9.82829407e-02
-4.35759217e-01 -1.25953272e-01 -1.31247735e+00 7.19315827e-01
4.43797946e-01 4.89474595e-01 7.83531904e-01 -3.99288267e-01
8.13364208e-01 7.13313520e-01 3.16117734e-01 -6.86512962e-02
4.28847909e-01 2.35281616e-01 6.78888112e-02 -3.62325698e-01
1.53394833e-01 -2.48078912e-01 6.09200120e-01 3.10061753e-01
-7.74348453e-02 3.42688084e-01 2.05386713e-01 -1.67510167e-01
6.34609818e-01 -2.23597720e-01 6.46186829e-01 -3.15196902e-01
7.32528567e-01 -3.52182955e-01 4.13563848e-01 5.19183695e-01
-6.44412518e-01 6.30610645e-01 2.00588971e-01 -4.47617859e-01
-3.24112713e-01 -8.68016303e-01 -5.05683124e-01 1.28299403e+00
-3.22596937e-01 -2.22420335e-01 -1.06349659e+00 -5.84857106e-01
4.42667902e-01 -1.35139212e-01 -8.54403019e-01 9.85502303e-02
-4.60965753e-01 -1.18980157e+00 7.26265490e-01 3.33743751e-01
2.86227971e-01 -9.59222436e-01 -5.74669302e-01 -2.40176842e-01
2.05171123e-01 -1.22205043e+00 -2.93999016e-01 -7.81621784e-02
-5.82606137e-01 -1.31291783e+00 -6.05650127e-01 -1.01790154e+00
9.11569297e-01 6.63863719e-01 9.80251193e-01 5.38822412e-01
-1.02987981e+00 4.01883036e-01 3.48786935e-02 -8.00956368e-01
-3.28158408e-01 -5.20071566e-01 6.83367610e-01 2.79411405e-01
8.31955850e-01 1.07866321e-02 -5.49184442e-01 6.13655031e-01
-5.31450927e-01 -9.05833483e-01 3.75901550e-01 2.97023445e-01
2.04874218e-01 2.43725460e-02 6.00759208e-01 -7.79168367e-01
2.49689907e-01 -2.44063154e-01 -7.68162608e-01 1.91895798e-01
-3.17310721e-01 -6.94501519e-01 -2.47230455e-01 -2.37503201e-01
-1.17723334e+00 4.23321038e-01 5.61440829e-03 -1.97183862e-01
-3.26729864e-01 -3.62945050e-01 -3.93298835e-01 -6.32537842e-01
7.67431736e-01 -6.64054006e-02 3.18680912e-01 -5.64859390e-01
2.07830518e-01 7.06413865e-01 4.60741162e-01 -4.45998833e-02
9.72728491e-01 1.02482295e+00 -9.19751301e-02 -1.21756542e+00
-4.81300414e-01 -6.85135901e-01 -8.39510918e-01 -2.85461515e-01
8.40937555e-01 -9.20570254e-01 -7.41417348e-01 7.28457749e-01
-1.02771771e+00 3.13106358e-01 2.31367573e-01 2.44973749e-01
-1.12172812e-01 4.22660828e-01 -4.56635505e-01 -1.19722676e+00
-3.27264547e-01 -1.18850696e+00 1.15293479e+00 1.71945766e-01
-1.87177002e-01 -7.15145171e-01 4.30120863e-02 3.92023832e-01
4.07405555e-01 2.64795572e-01 4.80356187e-01 -2.27087572e-01
-2.80690908e-01 -4.70884740e-01 -1.79878533e-01 7.63847455e-02
5.46236634e-01 2.52729267e-01 -1.67370033e+00 -6.91683769e-01
1.42531142e-01 -1.95262395e-02 8.69004488e-01 5.66243410e-01
5.65030694e-01 8.27537104e-02 -7.93958902e-01 3.74537647e-01
1.13508904e+00 2.24650204e-01 3.98464203e-01 -1.12121388e-01
1.26132503e-01 8.72758806e-01 3.76461536e-01 3.15464318e-01
-2.16948926e-01 8.14138472e-01 4.28501606e-01 -1.80125326e-01
-4.48738784e-01 6.18622564e-02 2.67619759e-01 -7.25332797e-02
-8.98214430e-02 3.62297744e-02 -5.88013887e-01 2.43363857e-01
-1.05380702e+00 -9.66118693e-01 8.02799035e-03 2.25680327e+00
2.03250244e-01 -2.37306803e-01 6.76031291e-01 2.21900836e-01
1.33592415e+00 5.83235584e-02 -3.16333830e-01 -1.28495887e-01
-3.11422944e-01 1.97400317e-01 2.89394021e-01 3.99367213e-01
-1.37328005e+00 6.95497215e-01 8.51976299e+00 3.75073105e-01
-1.12322366e+00 2.47139737e-01 4.66365010e-01 -1.10803828e-01
6.55494332e-01 -3.50165546e-01 -1.35233867e+00 3.33148628e-01
4.71764535e-01 1.10837780e-01 1.41652986e-01 6.66408420e-01
1.28063872e-01 -2.28111833e-01 -9.45737779e-01 1.14728940e+00
6.53562903e-01 -7.66302466e-01 -2.65474528e-01 1.89932093e-01
5.60703278e-01 -1.83727160e-01 4.74426359e-01 -2.34077722e-02
-3.06166202e-01 -1.20424986e+00 3.98751944e-01 -1.42205507e-01
1.04241931e+00 -4.52689439e-01 4.87309843e-01 -1.73167586e-01
-1.43443346e+00 -2.24486426e-01 -3.60450983e-01 -1.81344509e-01
1.39405695e-03 3.74426067e-01 -9.61609304e-01 2.77416073e-02
6.45379782e-01 7.43513256e-02 -7.38815546e-01 1.16510510e+00
-1.19502999e-01 2.43361637e-01 -3.33521992e-01 3.31839114e-01
-3.30429673e-01 1.25394762e-01 4.67700303e-01 1.22832310e+00
8.07286799e-02 1.64055169e-01 2.65947014e-01 6.63896918e-01
5.45960013e-03 4.71454449e-02 -4.59159404e-01 3.34924944e-02
6.51229382e-01 1.50527871e+00 -9.42831695e-01 -7.64813647e-02
-8.98891926e-01 6.90272450e-01 1.30277351e-01 1.45310670e-01
-5.76638877e-01 2.53536314e-01 1.07509553e+00 4.69875216e-01
3.91899258e-01 6.64127693e-02 1.04170730e-02 -4.90798414e-01
2.80148280e-03 -1.10519767e+00 8.58051836e-01 -2.06315264e-01
-1.25466383e+00 6.95242524e-01 7.89466277e-02 -7.53581703e-01
-1.04249446e-02 -9.23383176e-01 -6.54344440e-01 1.01421010e+00
-1.32545233e+00 -1.04996848e+00 -4.15403008e-01 4.55457538e-01
4.30223733e-01 -4.37796503e-01 9.57008779e-01 3.90328169e-01
-9.38712656e-01 6.34639025e-01 -1.63163766e-01 1.58409178e-01
8.88559043e-01 -5.74642360e-01 5.41017473e-01 8.36506903e-01
-1.14744961e-01 7.16702640e-01 7.87691355e-01 -7.90594041e-01
-1.32560992e+00 -9.23770547e-01 7.70232081e-01 -6.92581117e-01
1.37918085e-01 -4.48187143e-01 -6.03330374e-01 7.45196521e-01
3.84672917e-02 1.62497163e-01 7.33646929e-01 -4.08954639e-03
-5.95583439e-01 -1.52987391e-01 -1.64554143e+00 1.95825934e-01
1.03242576e+00 -5.17503023e-01 -3.26330870e-01 5.31253815e-01
-1.58525109e-01 -1.17382705e-01 -1.78840593e-01 4.71333414e-01
6.91681087e-01 -1.16292274e+00 1.09445333e+00 -3.50666612e-01
-5.11640728e-01 -3.11831117e-01 2.36440841e-02 -6.84094012e-01
-5.76428711e-01 -7.74595261e-01 1.80550158e-01 1.26371372e+00
-1.46165714e-01 -7.80619740e-01 8.84314954e-01 2.65164763e-01
4.46784437e-01 -2.48414457e-01 -1.05601513e+00 -9.16586399e-01
-2.24881917e-01 2.34118104e-01 5.29175401e-01 8.03230047e-01
-2.69432157e-01 -2.25306358e-02 -3.44156981e-01 4.63305265e-01
8.83532405e-01 -2.53145024e-02 5.66521704e-01 -1.37366605e+00
9.50099826e-02 -4.92746115e-01 -7.04914212e-01 -2.32066095e-01
-1.19436383e-02 -5.60647488e-01 -2.08588332e-01 -9.97807086e-01
5.65994084e-01 1.68113589e-01 5.19105345e-02 3.32745790e-01
-3.48994315e-01 9.59555864e-01 -6.36639372e-02 5.68105988e-02
-3.50232348e-02 -1.36198431e-01 6.77884519e-01 2.19879895e-01
4.64662816e-03 -4.50046062e-02 -7.63444602e-01 8.03964019e-01
6.66436434e-01 -4.99964416e-01 -3.48248472e-03 -2.40361214e-01
-4.18944389e-01 -3.49431992e-01 2.34087661e-01 -9.08084989e-01
-1.26364147e-02 1.22432955e-01 8.35473537e-01 -6.40744865e-01
7.24984050e-01 -4.77476805e-01 2.58572996e-01 8.03960562e-01
5.40362179e-01 3.79963547e-01 5.38492680e-01 2.36658275e-01
1.17199384e-01 -1.93548530e-01 1.31507742e+00 -2.70571500e-01
-6.69403672e-01 2.03300148e-01 -6.45624042e-01 -2.45565742e-01
1.29070604e+00 -4.64877129e-01 -5.00825822e-01 -1.04049649e-02
-6.89118326e-01 -3.47365364e-02 4.89631414e-01 6.97631717e-01
4.35749859e-01 -9.18468833e-01 -1.14893043e+00 6.47229373e-01
1.80032969e-01 -9.85417724e-01 2.58379489e-01 8.43581438e-01
-3.67941529e-01 5.60412824e-01 -3.33831280e-01 -5.48991859e-01
-2.22550750e+00 9.54263926e-01 4.93160725e-01 5.39539158e-01
-1.71768159e-01 1.12424099e+00 3.91876668e-01 6.56852424e-02
3.14372391e-01 5.17070532e-01 -3.28514725e-01 4.38992620e-01
1.13176215e+00 6.94649696e-01 2.62437552e-01 -1.24404001e+00
-1.00902998e+00 7.39834249e-01 -9.34367850e-02 2.28145719e-01
6.82505727e-01 -6.07515089e-02 -4.44726318e-01 -1.01331286e-01
7.39300907e-01 2.55760401e-01 -7.00119317e-01 6.70961477e-03
-1.68822780e-01 -6.56342268e-01 -1.81260362e-01 -6.21515572e-01
-1.17331898e+00 6.82238042e-01 1.24677813e+00 -9.54772905e-02
1.05390728e+00 3.07503968e-01 -1.18425128e-03 -1.31714061e-01
3.66477460e-01 -7.02900469e-01 4.24789041e-02 1.31462082e-01
9.35332000e-01 -1.35459065e+00 2.86284298e-01 -1.13880777e+00
-1.86500549e-02 8.71678114e-01 5.02383769e-01 1.83776632e-01
8.19108009e-01 4.03861701e-01 5.04738271e-01 -4.12028849e-01
-2.68403500e-01 -6.00903451e-01 2.70244330e-01 1.00024629e+00
6.28349423e-01 8.88050720e-02 -2.15986237e-01 9.04768184e-02
-1.07909568e-01 -3.15688580e-01 1.00632995e-01 8.60864878e-01
-7.97976196e-01 -1.11639142e+00 -1.14375985e+00 3.32378328e-01
-6.75999224e-01 3.12245905e-01 -9.71561730e-01 6.74293637e-01
5.31521380e-01 1.40944767e+00 4.48575355e-02 -3.04806642e-02
1.67985737e-01 3.50365192e-01 7.85847008e-01 -1.04733336e+00
-5.34321368e-01 3.08750495e-02 -5.49237281e-02 -3.45587850e-01
-5.17027795e-01 -9.00527954e-01 -6.62443399e-01 -4.07195598e-01
-4.97353435e-01 -1.34755626e-01 5.67113101e-01 6.82922542e-01
2.10112214e-01 4.72113816e-03 4.72079068e-01 -8.61519873e-01
-4.59146470e-01 -1.01178563e+00 -9.09353018e-01 1.18958101e-01
4.67600077e-01 -1.05764532e+00 -4.46746826e-01 -1.76006168e-01]
|
[13.32138729095459, 0.7720433473587036]
|
cca59710-efe9-4a90-8bd7-2c1135240020
|
depth-aware-cnn-for-rgb-d-segmentation
|
1803.06791
| null |
http://arxiv.org/abs/1803.06791v1
|
http://arxiv.org/pdf/1803.06791v1.pdf
|
Depth-aware CNN for RGB-D Segmentation
|
Convolutional neural networks (CNN) are limited by the lack of capability to
handle geometric information due to the fixed grid kernel structure. The
availability of depth data enables progress in RGB-D semantic segmentation with
CNNs. State-of-the-art methods either use depth as additional images or process
spatial information in 3D volumes or point clouds. These methods suffer from
high computation and memory cost. To address these issues, we present
Depth-aware CNN by introducing two intuitive, flexible and effective
operations: depth-aware convolution and depth-aware average pooling. By
leveraging depth similarity between pixels in the process of information
propagation, geometry is seamlessly incorporated into CNN. Without introducing
any additional parameters, both operators can be easily integrated into
existing CNNs. Extensive experiments and ablation studies on challenging RGB-D
semantic segmentation benchmarks validate the effectiveness and flexibility of
our approach.
|
['Weiyue Wang', 'Ulrich Neumann']
|
2018-03-19
|
depth-aware-cnn-for-rgb-d-segmentation-1
|
http://openaccess.thecvf.com/content_ECCV_2018/html/Weiyue_Wang_Depth-aware_CNN_for_ECCV_2018_paper.html
|
http://openaccess.thecvf.com/content_ECCV_2018/papers/Weiyue_Wang_Depth-aware_CNN_for_ECCV_2018_paper.pdf
|
eccv-2018-9
|
['thermal-image-segmentation']
|
['computer-vision']
|
[ 1.03061907e-01 1.87361032e-01 1.43904999e-01 -5.48604012e-01
-3.57844263e-01 -6.81163967e-01 5.96679747e-01 3.29415679e-01
-6.61561847e-01 2.26891711e-01 -1.65078461e-01 -4.00709599e-01
1.85730159e-01 -1.22547698e+00 -6.86246812e-01 -2.94026226e-01
-4.23020124e-02 2.07670882e-01 7.12864578e-01 -1.81362107e-01
2.68165708e-01 1.08563995e+00 -1.56560290e+00 1.78648695e-01
8.94315004e-01 1.45253038e+00 8.38662311e-02 4.70361382e-01
-7.61528134e-01 1.57030940e-01 -2.82603234e-01 -9.45040509e-02
7.28754282e-01 -8.40806067e-02 -8.34259987e-01 3.47225033e-02
4.97591764e-01 -6.66296542e-01 -2.27568924e-01 8.19703937e-01
4.44674820e-01 5.76687567e-02 1.78804159e-01 -1.07936954e+00
-3.72958899e-01 1.83636829e-01 -5.99578559e-01 1.40323356e-01
1.80245355e-01 3.05637985e-01 5.36386013e-01 -8.11906159e-01
4.66298848e-01 1.09046221e+00 1.04539216e+00 4.50762808e-01
-1.13625765e+00 -3.90242636e-01 3.52648407e-01 -4.08522248e-01
-1.25812221e+00 1.84020493e-02 9.47043002e-01 -2.11027086e-01
1.32624972e+00 7.24054649e-02 9.39840853e-01 4.64774519e-01
-2.53387719e-01 6.82654738e-01 9.03618276e-01 -1.04903370e-01
4.14872408e-01 -2.28269786e-01 3.24435830e-02 8.00128400e-01
8.54832605e-02 5.41077442e-02 -3.83088082e-01 1.96206555e-01
1.36475194e+00 1.99570850e-01 -3.47604081e-02 -5.84993243e-01
-1.06058097e+00 7.28868842e-01 1.03803802e+00 8.05290565e-02
-2.60162562e-01 4.52715307e-01 2.46157542e-01 -3.07627432e-02
7.60331511e-01 4.40041214e-01 -6.41157210e-01 -1.04170449e-01
-1.08727360e+00 3.55579346e-01 5.25367141e-01 1.05438375e+00
1.14217877e+00 -2.15895638e-01 5.93431778e-02 3.46252322e-01
6.88309297e-02 2.81662643e-01 4.65999991e-02 -1.19164550e+00
4.28487271e-01 1.15390944e+00 -1.08024441e-01 -6.98299408e-01
-8.25933993e-01 -2.63674110e-01 -5.82158267e-01 5.30924380e-01
5.21360099e-01 6.23748973e-02 -1.49342263e+00 1.09642434e+00
4.87303942e-01 1.59655467e-01 -2.52518684e-01 8.93208563e-01
9.99871373e-01 2.70626754e-01 1.79723531e-01 5.00678897e-01
1.17710245e+00 -8.27182651e-01 -1.83568135e-01 -3.42125773e-01
6.47387743e-01 -5.05133152e-01 1.04725885e+00 5.51895723e-02
-1.39532220e+00 -3.65774721e-01 -1.07244313e+00 -6.80820286e-01
-6.58213735e-01 -2.20781296e-01 1.19397891e+00 7.30841696e-01
-1.38961804e+00 8.32468212e-01 -1.21113789e+00 -2.43460909e-01
9.40694809e-01 6.92665756e-01 -3.56021523e-01 -8.22910573e-03
-8.16442668e-01 5.24226367e-01 3.34522367e-01 1.61776736e-01
-3.58779520e-01 -1.06071234e+00 -1.01626730e+00 9.14124120e-03
1.13612987e-01 -9.23903525e-01 1.23915517e+00 -6.93544328e-01
-1.56624258e+00 9.91314113e-01 3.09078507e-02 -2.93718606e-01
7.06291854e-01 -1.63687348e-01 3.49693090e-01 5.13360143e-01
-6.44656494e-02 1.07498014e+00 4.73882079e-01 -1.14634264e+00
-5.47691643e-01 -6.18943334e-01 4.48010474e-01 1.59653112e-01
-1.82923019e-01 -4.13599014e-01 -8.01251709e-01 -4.84322429e-01
8.84404719e-01 -6.53321862e-01 -5.08444190e-01 6.14454091e-01
-2.76024342e-01 1.46463318e-02 9.60604727e-01 -3.39088947e-01
6.66356325e-01 -2.17778492e+00 -8.67577717e-02 2.20951900e-01
3.35891664e-01 1.85102254e-01 5.07586971e-02 1.60143301e-02
2.93229729e-01 4.30130810e-01 -4.64098394e-01 -6.30490720e-01
-2.45464146e-02 2.69910514e-01 -1.16799399e-02 3.91137987e-01
5.70891201e-01 1.21405923e+00 -8.86533022e-01 -4.40502316e-01
6.03096247e-01 7.37139642e-01 -7.85791636e-01 -8.46473798e-02
-4.00515258e-01 4.38096076e-01 -6.66549146e-01 8.67677748e-01
9.97728467e-01 -2.80625969e-01 -3.81425023e-01 -3.88231039e-01
-3.38456005e-01 6.27017677e-01 -9.14813817e-01 2.37412214e+00
-5.08835256e-01 3.59520853e-01 1.92401912e-02 -7.97196209e-01
6.79906309e-01 -8.42896849e-02 5.31361580e-01 -8.62341881e-01
2.25571826e-01 2.10227594e-01 -4.92237866e-01 -2.39202548e-02
5.62289298e-01 1.11221999e-01 4.87847067e-02 3.38773817e-01
-8.78090505e-03 -7.64682770e-01 -2.21853793e-01 -3.71746868e-02
1.08559775e+00 4.52682495e-01 -2.33423665e-01 -1.18221611e-01
9.89795327e-02 3.08018714e-01 3.25095206e-01 6.96868598e-01
-1.03336066e-01 1.03231144e+00 4.09862190e-01 -6.90768838e-01
-1.00908577e+00 -1.30623198e+00 -3.30988020e-01 7.43108511e-01
2.86464155e-01 -2.38031402e-01 -8.40362906e-01 -5.85595131e-01
1.69361979e-01 2.74823844e-01 -7.82312989e-01 1.43924832e-01
-7.10222602e-01 -6.18900239e-01 5.57555258e-01 1.08423746e+00
1.06645727e+00 -6.56745851e-01 -1.00394070e+00 3.43139917e-01
3.32410485e-01 -1.41804183e+00 1.09147478e-03 3.79333049e-01
-1.44927537e+00 -6.54817283e-01 -6.76593840e-01 -4.39721882e-01
7.46679068e-01 4.34436262e-01 1.11312282e+00 2.30224922e-01
-4.41867799e-01 2.35040605e-01 -1.85621843e-01 -3.06785315e-01
1.22134976e-01 4.35273707e-01 -6.50496244e-01 -6.77889526e-01
2.42511064e-01 -7.57926285e-01 -1.01463842e+00 1.11374609e-01
-9.40247834e-01 3.53293121e-01 4.13143396e-01 2.79484659e-01
7.37835467e-01 -2.48079598e-01 3.38743366e-02 -6.80899858e-01
1.27278134e-01 -1.67912051e-01 -7.66275048e-01 -2.51310259e-01
-3.52239609e-01 -2.29966082e-02 1.15808688e-01 -5.31807318e-02
-8.29561710e-01 4.11552638e-01 -3.18686187e-01 -2.79741585e-01
-3.51760894e-01 2.20617920e-01 -1.09399013e-01 -4.54564273e-01
2.87383318e-01 -9.85417664e-02 8.80397931e-02 -4.65122253e-01
5.65439224e-01 1.52212173e-01 4.72011685e-01 -4.87119913e-01
7.06300437e-01 9.78225172e-01 1.43227324e-01 -7.08623469e-01
-7.65022755e-01 -3.76718611e-01 -9.09256697e-01 7.19790906e-02
1.16514230e+00 -9.30276155e-01 -7.00309038e-01 8.04182529e-01
-1.23421609e+00 -6.93709195e-01 -2.68896103e-01 1.47321224e-01
-4.88191456e-01 5.34175970e-02 -7.87083030e-01 -4.46898848e-01
-2.12778956e-01 -1.20612133e+00 1.42333484e+00 3.97242010e-01
-4.53779362e-02 -1.00237608e+00 -4.83394504e-01 2.73622274e-01
5.76743424e-01 5.37099659e-01 8.44747841e-01 -2.35017747e-01
-1.16272390e+00 -2.47679606e-01 -7.26844192e-01 2.46662185e-01
9.49091092e-02 -1.09322928e-01 -1.21230960e+00 2.33717278e-01
-3.23181033e-01 -1.32204657e-02 1.01333058e+00 4.54943717e-01
1.59647048e+00 1.93059936e-01 -2.60820895e-01 1.27919722e+00
1.60081387e+00 -2.54159540e-01 7.06510484e-01 4.08548355e-01
1.07676542e+00 3.79763931e-01 1.81372538e-01 3.39787275e-01
5.96369088e-01 3.65990907e-01 7.89449692e-01 -5.85222006e-01
-1.96932748e-01 -9.88602936e-02 -4.33518916e-01 3.05026144e-01
-2.54107624e-01 -8.15143660e-02 -1.18942487e+00 4.64081436e-01
-1.60163164e+00 -5.13184488e-01 -1.80257484e-01 1.92940760e+00
7.55795717e-01 4.03966814e-01 -9.19971466e-02 2.56264955e-01
3.40461321e-02 6.99332654e-02 -6.24399245e-01 -3.51881683e-01
-1.05325058e-01 7.01334774e-01 1.01813722e+00 3.63214314e-01
-1.23254311e+00 1.20539808e+00 6.14465284e+00 4.42274272e-01
-1.32206488e+00 2.56356429e-02 6.76591933e-01 -3.57440799e-01
-3.86013418e-01 -1.87682733e-01 -7.18314052e-01 -2.95994943e-03
4.20620084e-01 3.73356760e-01 2.26039737e-01 7.26673782e-01
1.47274092e-01 -4.82799113e-01 -1.01979756e+00 8.94422650e-01
-3.38957697e-01 -1.65575278e+00 -2.11630799e-02 7.23110512e-02
7.43790984e-01 4.22426969e-01 5.75066619e-02 -1.03201143e-01
2.71022588e-01 -1.09688413e+00 7.55659342e-01 2.30778381e-01
6.33891523e-01 -7.94881999e-01 5.61721623e-01 -1.07452413e-02
-1.18840528e+00 1.01940088e-01 -2.25430131e-01 -3.39111239e-01
1.37502939e-01 7.75781631e-01 -5.56829751e-01 5.01892507e-01
1.02555490e+00 6.99862897e-01 -5.89813828e-01 9.18648064e-01
-1.90424442e-01 2.06843868e-01 -7.36682117e-01 1.36110008e-01
6.69071853e-01 4.88345735e-02 3.14257555e-02 1.10201633e+00
1.71240181e-01 2.09410384e-01 1.29970834e-01 1.29437935e+00
-5.92976846e-02 -1.89113840e-01 -3.62790465e-01 1.45698115e-01
3.59807670e-01 1.13106871e+00 -1.45196688e+00 -8.85959491e-02
-4.61924434e-01 1.18397200e+00 2.93105036e-01 4.05948430e-01
-6.88147664e-01 -3.46215695e-01 9.55470622e-01 3.17338437e-01
5.30639410e-01 -7.92775929e-01 -1.19467747e+00 -6.49656296e-01
9.77160260e-02 -1.11480281e-01 -1.25410527e-01 -6.54182076e-01
-1.08012378e+00 4.85850871e-01 2.71873605e-02 -8.20913851e-01
2.23747864e-01 -7.57406950e-01 -3.90741348e-01 1.01308334e+00
-1.91974211e+00 -1.21255982e+00 -7.83977628e-01 6.00034237e-01
2.06559926e-01 5.23240924e-01 6.68975830e-01 2.86119878e-01
-2.69593388e-01 2.36594826e-01 -5.49517870e-01 2.42827669e-01
1.88195780e-01 -1.40835035e+00 9.24630225e-01 5.98629177e-01
-4.18559223e-01 6.26685500e-01 2.16126740e-01 -4.73375529e-01
-1.30675304e+00 -1.14327741e+00 5.70073366e-01 -5.02090454e-01
3.78376782e-01 -6.34058237e-01 -8.73978436e-01 4.97143865e-01
-2.42257118e-01 4.96271044e-01 4.05915886e-01 -1.23209693e-01
-3.85425627e-01 7.27584958e-02 -1.31879187e+00 5.29011786e-01
1.36934900e+00 -5.82163632e-01 -2.56835192e-01 3.98777872e-02
1.09440780e+00 -9.17988360e-01 -8.60229492e-01 5.99411964e-01
2.66691536e-01 -1.34631515e+00 1.46537983e+00 -1.26594603e-01
5.85995197e-01 -3.79797876e-01 5.52173257e-02 -9.86431658e-01
8.76163393e-02 -2.89992422e-01 2.08515719e-01 8.65102530e-01
2.67867923e-01 -6.42147720e-01 1.03116500e+00 9.04852569e-01
-5.33856452e-01 -8.17729592e-01 -9.45341587e-01 -5.56276202e-01
2.23288298e-01 -8.08346808e-01 9.61786151e-01 8.22172284e-01
-4.03045744e-01 -3.30386490e-01 5.31362355e-01 3.02555859e-01
4.89268363e-01 1.58089474e-01 6.99374855e-01 -1.26857078e+00
6.30623400e-02 -8.09443653e-01 -5.72860897e-01 -1.29601538e+00
-1.52463302e-01 -7.42221892e-01 -1.94914488e-03 -1.95623994e+00
-3.88006449e-01 -9.12852049e-01 -4.95135449e-02 7.26082802e-01
-4.91630062e-02 7.10009933e-01 6.80013672e-02 -1.21607326e-01
-2.77623117e-01 3.99814844e-01 1.33744693e+00 9.32169110e-02
-4.97745007e-01 -2.54320323e-01 -3.56295913e-01 8.65154982e-01
9.44904983e-01 -1.06284656e-01 -4.47101206e-01 -9.07865167e-01
2.15234041e-01 -3.06296587e-01 8.73637795e-01 -1.06959963e+00
2.07924068e-01 1.17549829e-01 5.91773331e-01 -8.37603807e-01
4.44279313e-01 -7.60431349e-01 -2.41324142e-01 2.26425707e-01
6.52072802e-02 1.15915693e-01 6.55917227e-01 1.83243841e-01
-8.19389448e-02 1.96380258e-01 6.69038177e-01 -4.08229768e-01
-8.66828680e-01 6.19023740e-01 6.62068836e-03 -7.66386315e-02
9.29698765e-01 -7.49481559e-01 -8.27878863e-02 7.94825424e-03
-4.98853296e-01 2.51636207e-02 8.45633805e-01 2.93132931e-01
6.10810339e-01 -9.56545234e-01 -1.03480458e-01 3.61340880e-01
4.25936049e-03 9.07146454e-01 1.78540021e-01 7.36890733e-01
-1.12787318e+00 2.77655065e-01 -2.46271148e-01 -9.29729104e-01
-6.40000463e-01 2.44612947e-01 5.60039878e-01 5.60223088e-02
-8.15546393e-01 1.09261918e+00 1.63416654e-01 -4.63541985e-01
2.35690653e-01 -9.41186011e-01 4.00749594e-01 -2.07363665e-01
3.24033141e-01 1.36554569e-01 3.86915982e-01 -1.29811183e-01
-4.87315625e-01 6.98493302e-01 1.81269258e-01 -4.84913401e-02
1.51027369e+00 -7.81139061e-02 -9.47545916e-02 1.78203776e-01
1.17003989e+00 -4.06431705e-01 -1.76438272e+00 -1.29604088e-02
-3.45508941e-02 -4.44057614e-01 5.29221892e-01 -6.28675580e-01
-1.52137744e+00 1.14358044e+00 4.89912003e-01 5.96945211e-02
1.15309703e+00 9.06310230e-03 1.06149101e+00 1.02409078e-02
4.19858903e-01 -9.34125006e-01 -5.92245050e-02 5.49601555e-01
4.50222135e-01 -1.26708138e+00 5.52348979e-02 -9.39219773e-01
3.07697617e-02 1.21320581e+00 5.75248837e-01 -2.00280190e-01
8.69483829e-01 5.70201933e-01 1.75389528e-01 -4.01366740e-01
-1.01863332e-01 -4.02293921e-01 1.75805107e-01 6.30916417e-01
4.88453418e-01 -2.83521622e-01 5.82051873e-02 2.16013402e-01
-2.73175687e-01 1.71117391e-02 2.47074589e-01 1.26136053e+00
-3.09344232e-01 -9.32676613e-01 -3.07689682e-02 9.93355364e-02
-1.50371253e-01 -2.24201187e-01 -2.23136723e-01 8.41593385e-01
2.99232751e-01 5.91216087e-01 6.08726263e-01 -8.41835141e-02
3.81249189e-01 -1.42596543e-01 6.35107577e-01 -5.70638418e-01
-7.67872214e-01 -1.33537501e-01 -3.43517035e-01 -1.06776512e+00
-5.37371457e-01 -5.06431460e-01 -1.75441134e+00 -3.60241055e-01
-2.36016568e-02 -5.29310286e-01 1.08671987e+00 9.39225554e-01
6.09455585e-01 6.44171596e-01 1.80946454e-01 -1.32069182e+00
-5.12468703e-02 -4.99192923e-01 -3.06813359e-01 1.97983325e-01
2.84388036e-01 -4.98744339e-01 -1.53303161e-01 -1.64027199e-01]
|
[8.271322250366211, -3.02783203125]
|
f24acf82-3765-4b19-ad07-e2c6edfa306e
|
empirical-study-of-diachronic-word-embeddings
|
1909.01863
| null |
https://arxiv.org/abs/1909.01863v1
|
https://arxiv.org/pdf/1909.01863v1.pdf
|
Empirical Study of Diachronic Word Embeddings for Scarce Data
|
Word meaning change can be inferred from drifts of time-varying word embeddings. However, temporal data may be too sparse to build robust word embeddings and to discriminate significant drifts from noise. In this paper, we compare three models to learn diachronic word embeddings on scarce data: incremental updating of a Skip-Gram from Kim et al. (2014), dynamic filtering from Bamler and Mandt (2017), and dynamic Bernoulli embeddings from Rudolph and Blei (2018). In particular, we study the performance of different initialisation schemes and emphasise what characteristics of each model are more suitable to data scarcity, relying on the distribution of detected drifts. Finally, we regularise the loss of these models to better adapt to scarce data.
|
['Alexandre Allauzen', 'Syrielle Montariol']
|
2019-09-04
|
empirical-study-of-diachronic-word-embeddings-1
|
https://aclanthology.org/R19-1092
|
https://aclanthology.org/R19-1092.pdf
|
ranlp-2019-9
|
['diachronic-word-embeddings']
|
['natural-language-processing']
|
[-6.56683370e-02 -2.92504221e-01 -3.21729124e-01 -3.34352821e-01
-2.74561107e-01 -7.29025841e-01 9.39413011e-01 5.45028865e-01
-1.04087341e+00 7.74033248e-01 5.55281401e-01 -3.68579626e-01
-3.30632269e-01 -5.76411545e-01 -4.04854029e-01 -6.49703562e-01
-3.14615488e-01 4.16733474e-01 2.22826198e-01 -2.01728344e-01
9.24730152e-02 1.90361679e-01 -1.34619749e+00 -2.06873387e-01
8.18835318e-01 6.02285981e-01 2.20554039e-01 7.75904834e-01
-1.21112943e-01 6.73577711e-02 -5.00327647e-01 -2.55841881e-01
1.63225546e-01 -1.00616310e-02 -4.19835836e-01 -4.81737375e-01
1.86007112e-01 -1.55194700e-01 -3.91168118e-01 8.93159986e-01
6.67126954e-01 1.02185026e-01 7.89787948e-01 -8.23367596e-01
-1.01985526e+00 8.67982924e-01 -3.16805393e-01 9.59271252e-01
-4.12092209e-02 3.64310384e-01 1.03322470e+00 -8.20188522e-01
6.08374655e-01 1.18551230e+00 6.40693486e-01 6.78287983e-01
-1.36706626e+00 -4.45966274e-01 7.24305153e-01 3.84567171e-01
-1.17217314e+00 -4.85289514e-01 4.61992323e-01 -5.98023295e-01
9.31744039e-01 5.80772385e-03 6.24143183e-01 1.83421135e+00
2.21612051e-01 6.07898235e-01 7.12895930e-01 -3.89622480e-01
4.16122973e-01 -8.94282311e-02 3.63136351e-01 3.39934155e-02
7.76477695e-01 1.76454648e-01 -6.51039064e-01 -3.66504043e-01
2.37785831e-01 -3.09064109e-02 -2.34650344e-01 -3.54443729e-01
-8.53122830e-01 7.14534938e-01 2.92976320e-01 5.65241516e-01
-5.51111579e-01 2.60491043e-01 4.54768151e-01 3.24936628e-01
8.62190843e-01 2.84347475e-01 -7.99338341e-01 -7.30492294e-01
-7.74926007e-01 1.74472436e-01 4.19130862e-01 5.17251968e-01
5.18889010e-01 1.45026088e-01 -2.43406147e-01 8.09588134e-01
3.91008377e-01 4.26835179e-01 1.00827229e+00 -2.24922970e-01
2.35713318e-01 1.15033031e-01 2.88042426e-01 -4.78935182e-01
-3.66492361e-01 -2.33459234e-01 -4.78066087e-01 -3.51912528e-01
5.00791192e-01 -3.30174029e-01 -1.12300086e+00 1.96130288e+00
2.04115033e-01 4.72517997e-01 -1.36488080e-01 6.94912076e-01
2.17287853e-01 5.09268641e-01 2.74833143e-01 -1.50535971e-01
1.15207934e+00 -5.11049509e-01 -7.93773770e-01 -5.05036175e-01
7.57896245e-01 -4.09996331e-01 1.20177031e+00 2.66175002e-01
-6.18579566e-01 -3.18039536e-01 -8.48722875e-01 1.65358081e-01
-4.85324651e-01 -3.64911884e-01 4.43144470e-01 7.42482841e-01
-9.67655659e-01 6.16883159e-01 -1.08378911e+00 -5.98074615e-01
2.03509107e-01 2.38020327e-02 8.39146301e-02 -3.02519560e-01
-1.65108645e+00 9.53337848e-01 4.82736826e-01 8.92894194e-02
-7.51536667e-01 -9.70822752e-01 -6.47109210e-01 -3.00075442e-01
2.18619052e-02 -5.51772654e-01 1.02337348e+00 -8.24191749e-01
-1.20243442e+00 5.65653265e-01 -1.89862221e-01 -8.49930406e-01
5.21318853e-01 -4.03070658e-01 -3.88279885e-01 -4.49117452e-01
-1.53361931e-01 4.35534656e-01 8.42839777e-01 -8.82618845e-01
-5.47756195e-01 -4.19078737e-01 -2.19873011e-01 2.71864180e-02
-7.45430946e-01 -1.58469275e-01 -1.38425604e-01 -8.21364939e-01
-4.59428340e-01 -7.50621676e-01 -2.05073506e-01 -1.79449588e-01
3.41517538e-01 -4.76953447e-01 5.64627230e-01 -6.33398473e-01
1.67350769e+00 -2.24499035e+00 2.79199272e-01 -1.37312248e-01
-8.38808268e-02 3.46975148e-01 -5.92606008e-01 4.83679444e-01
6.76718354e-02 2.58127838e-01 -3.62352133e-01 -3.71411681e-01
3.27666961e-02 7.23582149e-01 -2.07165077e-01 6.09078586e-01
3.40028435e-01 7.95549333e-01 -1.17309844e+00 1.82739869e-01
2.40335502e-02 5.63938260e-01 -2.89635390e-01 -1.85218588e-01
-1.77131012e-01 1.71854615e-01 -4.98037003e-02 5.28383791e-01
3.44003737e-01 3.03638965e-01 1.88634723e-01 2.98472345e-01
-1.99107975e-01 5.90826988e-01 -1.03291976e+00 1.34683418e+00
-7.44613230e-01 8.65647256e-01 -1.51077539e-01 -7.82977581e-01
9.35102344e-01 1.65620297e-01 3.88413221e-01 -6.35778189e-01
1.10640846e-01 1.89701289e-01 3.75616878e-01 -4.44933951e-01
5.47530949e-01 -2.30890617e-01 -4.95436415e-02 4.36312854e-01
-8.20918307e-02 5.47202043e-02 4.45376456e-01 -4.97835800e-02
1.28957868e+00 -7.29486868e-02 2.25892160e-02 -3.08308125e-01
8.53397548e-02 -4.09061700e-01 7.15298474e-01 8.58862042e-01
-4.69949663e-01 4.33672726e-01 4.14210886e-01 -4.61639315e-01
-1.12343609e+00 -1.09682047e+00 -5.08219481e-01 1.20468926e+00
-2.40989044e-01 -5.04289508e-01 -2.48305202e-01 -4.79834348e-01
3.60304207e-01 8.94308805e-01 -9.74521518e-01 -5.71442604e-01
-4.19427425e-01 -1.30257809e+00 4.30175334e-01 4.74742204e-01
-3.19025517e-01 -1.00016654e+00 -6.90420091e-01 3.47048253e-01
1.07357219e-01 -5.30452073e-01 -5.09624720e-01 4.91314262e-01
-9.22338903e-01 -9.15939748e-01 -6.65256143e-01 -2.63033032e-01
3.81769240e-01 -5.80572598e-02 1.00550330e+00 -3.69264483e-01
-2.21774787e-01 5.83553493e-01 -6.61950409e-01 -6.52176976e-01
-2.46531457e-01 4.15236741e-01 5.85420489e-01 5.42445295e-02
7.02175736e-01 -3.77813995e-01 -5.23138344e-01 -1.76229164e-01
-1.20222425e+00 -7.81300545e-01 3.86927903e-01 8.69838297e-01
2.04912826e-01 -1.34678409e-01 4.70126510e-01 -6.38368070e-01
9.83365178e-01 -8.95927906e-01 -2.48606905e-01 -1.06312269e-02
-1.06715846e+00 3.09733659e-01 3.11434299e-01 -9.94087636e-01
-7.31973410e-01 -5.27880967e-01 5.77780837e-03 -2.73316085e-01
1.24655530e-01 5.01227260e-01 1.93528220e-01 5.02524376e-01
6.68486893e-01 2.11267006e-02 -3.00931543e-01 -6.61806822e-01
5.59383929e-01 8.05304408e-01 1.05474301e-01 -6.57592297e-01
6.22927129e-01 3.37232471e-01 -6.95939422e-01 -7.76132941e-01
-6.68028831e-01 -4.22536373e-01 -7.27362871e-01 -4.43310253e-02
5.23859441e-01 -9.07621741e-01 1.59740269e-01 4.35968637e-01
-1.00045967e+00 -6.56964362e-01 -6.22463584e-01 5.48871994e-01
-1.01379454e-01 2.21969217e-01 -4.77560669e-01 -8.95683110e-01
-8.35032314e-02 -6.69672668e-01 7.20981181e-01 1.55734479e-01
-5.16637146e-01 -1.34820592e+00 6.63679838e-01 -3.64297807e-01
6.40066564e-01 1.67356014e-01 6.79149866e-01 -6.95706606e-01
4.96214852e-02 -3.16587657e-01 8.41779932e-02 4.01127011e-01
5.41434228e-01 2.57991284e-01 -8.39570284e-01 -6.34160697e-01
-1.56418219e-01 8.01578388e-02 1.25446069e+00 3.97727519e-01
7.75568664e-01 -3.14908683e-01 -1.52383268e-01 3.91278207e-01
1.30220914e+00 1.52348235e-01 2.95668274e-01 7.18537331e-01
5.07134676e-01 4.72358406e-01 4.19086844e-01 5.35699785e-01
4.29767400e-01 4.32417929e-01 3.17168206e-01 5.24202704e-01
9.13423151e-02 -2.70278960e-01 8.13187063e-01 9.98703599e-01
1.63234219e-01 -2.36166298e-01 -1.21384919e+00 1.21906292e+00
-1.75363886e+00 -7.58876622e-01 -2.14700326e-01 2.37884331e+00
9.85053718e-01 5.08899510e-01 3.08302373e-01 -9.30395126e-02
5.83049893e-01 5.74060857e-01 -6.81164801e-01 -5.54760277e-01
-2.25928336e-01 2.62078464e-01 7.90701568e-01 6.06450498e-01
-7.37152755e-01 8.06963265e-01 6.41493320e+00 2.83957005e-01
-9.45958555e-01 4.12137896e-01 2.96935529e-01 -4.58941400e-01
-6.40270531e-01 4.86494824e-02 -7.04466462e-01 1.03887272e+00
1.54311728e+00 -2.25660905e-01 1.90089136e-01 3.41644883e-01
2.00859338e-01 5.45277596e-02 -1.02165771e+00 6.82315826e-01
1.15041181e-01 -7.56509542e-01 -1.12063088e-01 -4.70768809e-02
7.71041274e-01 5.17730951e-01 1.55738249e-01 3.17499548e-01
6.86946452e-01 -7.85257757e-01 6.83205605e-01 6.07582569e-01
3.30745369e-01 -5.18296242e-01 8.15602541e-01 2.97048569e-01
-8.07550788e-01 -3.73160094e-01 -3.18308860e-01 -2.05595180e-01
2.67023981e-01 9.04866219e-01 -5.26202202e-01 3.10435742e-01
7.93354154e-01 7.49630213e-01 -7.41818726e-01 8.92776191e-01
-2.45131537e-01 1.04487991e+00 -5.88683188e-01 -3.48519117e-01
4.30665225e-01 -1.64942026e-01 5.55167854e-01 1.23881984e+00
2.83844590e-01 -5.12625635e-01 -2.37383336e-01 3.66839379e-01
1.07015021e-01 -9.20561031e-02 -5.65029919e-01 -2.08286583e-01
8.03285837e-01 8.55623662e-01 -3.84891808e-01 -1.03069685e-01
-4.93543923e-01 8.08523059e-01 2.83750772e-01 4.64736342e-01
-4.99795675e-01 -1.09116673e-01 1.18242896e+00 1.59067929e-01
6.21430457e-01 -5.29015601e-01 -1.39319882e-01 -1.17753077e+00
9.71730426e-02 -5.22646546e-01 4.81237948e-01 -2.21435443e-01
-1.50549781e+00 3.47371519e-01 2.15223476e-01 -5.89820743e-01
-1.51527554e-01 -6.87791586e-01 -6.46696508e-01 8.34989130e-01
-1.62463021e+00 -6.78943336e-01 9.42382216e-02 -4.25528623e-02
5.92215002e-01 2.87787467e-01 3.95821095e-01 3.08310151e-01
-7.68117249e-01 4.30028141e-01 6.08347237e-01 -1.55821100e-01
9.45136249e-01 -1.39828897e+00 8.39449763e-01 7.99474001e-01
3.93926561e-01 6.89583242e-01 8.78863096e-01 -6.64268970e-01
-1.08728707e+00 -9.56247211e-01 9.44357455e-01 -9.24745023e-01
1.09935594e+00 -5.19994676e-01 -1.21653771e+00 6.63636625e-01
2.32481539e-01 -8.94642100e-02 5.39587855e-01 4.75010306e-01
-4.92285609e-01 -1.60059869e-01 -6.35695398e-01 5.97918808e-01
8.69420946e-01 -4.40334409e-01 -6.55285537e-01 9.01386142e-02
6.78032339e-01 7.32358545e-02 -7.27770984e-01 2.27460325e-01
4.54362929e-01 -5.99887788e-01 6.62794828e-01 -7.93271184e-01
-1.68742821e-01 -1.88101277e-01 -6.29118308e-02 -1.63190603e+00
-3.91658306e-01 -4.90273833e-01 -2.13520020e-01 1.44273126e+00
4.05526221e-01 -8.65159929e-01 1.97047606e-01 2.73597717e-01
1.91476256e-01 -6.54392660e-01 -1.12471306e+00 -9.28567171e-01
5.04984319e-01 -4.96348172e-01 6.17342532e-01 9.85018611e-01
-2.75952011e-01 2.39921525e-01 -3.44029784e-01 -1.03861012e-01
2.81112134e-01 -6.22328937e-01 4.73958820e-01 -1.37805521e+00
6.55464083e-02 -5.51389158e-01 -3.63146305e-01 -7.42066801e-01
2.30758965e-01 -7.17664659e-01 6.95914850e-02 -1.34679961e+00
-1.95986256e-02 -2.91213065e-01 -8.50642741e-01 2.83822358e-01
-5.71486771e-01 -5.05051874e-02 8.35848227e-02 5.30269034e-02
-3.73908311e-01 6.81489944e-01 3.80929708e-01 -1.30643234e-01
-2.57423192e-01 -1.96749717e-01 -5.40306568e-01 4.19045657e-01
9.22523916e-01 -6.38714910e-01 -4.14597631e-01 -7.22553313e-01
5.54647386e-01 -7.84439981e-01 1.29875153e-01 -5.06127477e-01
4.25242633e-02 -4.45090272e-02 6.08002208e-02 -2.71836221e-01
-1.47171706e-01 -5.55254459e-01 -2.81814281e-02 5.57649076e-01
-2.98511356e-01 4.66188282e-01 3.77320945e-01 9.79221940e-01
4.51807231e-02 -4.79518652e-01 5.70180953e-01 2.31852531e-02
-5.33416092e-01 2.00179681e-01 -4.91830230e-01 3.43413025e-01
7.12351322e-01 -6.70609474e-02 7.21348869e-03 -1.73429906e-01
-4.70950305e-01 4.04768080e-01 4.80974197e-01 9.61827219e-01
2.21506089e-01 -1.38983750e+00 -7.80889213e-01 8.56167451e-02
1.41454741e-01 -2.21400745e-02 1.47264645e-01 8.35445702e-01
-3.95657793e-02 2.05635577e-01 1.40993699e-01 -4.59493101e-01
-8.61018538e-01 5.65779150e-01 6.95361495e-02 -1.91120982e-01
-3.74595404e-01 7.09375620e-01 -3.13884139e-01 -2.60759979e-01
1.66360870e-01 -5.51522076e-01 -6.94317222e-02 8.12978625e-01
3.70980710e-01 2.57511616e-01 1.40545582e-02 -3.07207644e-01
-7.14597583e-01 2.95440823e-01 -3.33629489e-01 -2.86266893e-01
1.41871250e+00 -3.24644566e-01 9.57181603e-02 1.21755171e+00
9.35350835e-01 -2.86698639e-01 -1.47602940e+00 -4.39940959e-01
5.36050737e-01 -2.94543147e-01 8.66419228e-04 -3.72623771e-01
-4.16276485e-01 6.88259542e-01 8.29243600e-01 4.74729210e-01
7.65779197e-01 -1.92764521e-01 6.81132376e-01 -6.48544505e-02
1.08017601e-01 -1.32922876e+00 -6.64762408e-02 5.76295793e-01
4.33525026e-01 -1.02089775e+00 -7.56747648e-02 6.34327054e-01
-3.61730188e-01 9.81234312e-01 3.12924504e-01 3.91496792e-02
8.59099388e-01 1.54135562e-02 4.88243364e-02 1.02112159e-01
-1.25025415e+00 -5.44724345e-01 9.25289243e-02 7.15028822e-01
2.97026753e-01 -3.35062779e-02 -7.05750585e-01 4.08276677e-01
-3.43965292e-02 -1.31230354e-01 5.63765883e-01 8.31089318e-01
-3.42087716e-01 -1.19320428e+00 -2.55589664e-01 5.80189943e-01
-2.60637611e-01 -2.99996912e-01 -4.35442895e-01 6.38086677e-01
1.97068170e-01 6.57541573e-01 5.04545271e-01 -7.06407428e-02
5.13621628e-01 5.35757780e-01 2.54356682e-01 -6.88284934e-01
-3.75029504e-01 -2.18708888e-01 -3.37336995e-02 5.57915960e-03
-2.59606898e-01 -1.16115439e+00 -4.87264723e-01 -1.27534002e-01
-2.21000433e-01 -3.86363338e-03 5.66008389e-01 9.20807481e-01
3.48345548e-01 5.90974629e-01 4.25725311e-01 -3.70403379e-01
-8.00840914e-01 -1.42396295e+00 -3.81206185e-01 4.04560685e-01
4.98253018e-01 -6.75441027e-01 -8.15975308e-01 -1.13372646e-01]
|
[10.18155574798584, 8.752906799316406]
|
d577295f-e12d-44e6-baef-9a9bc1ad5dfe
|
linguistic-knowledge-in-data-augmentation-for
|
2111.14709
| null |
https://arxiv.org/abs/2111.14709v3
|
https://arxiv.org/pdf/2111.14709v3.pdf
|
Linguistic Knowledge in Data Augmentation for Natural Language Processing: An Example on Chinese Question Matching
|
To investigate the role of linguistic knowledge in data augmentation (DA) for Natural Language Processing (NLP), we designed two adapted DA programs and applied them to LCQMC (a Large-scale Chinese Question Matching Corpus) for a binary Chinese question matching classification task. The two DA programs produce augmented texts by five simple text editing operations (or DA techniques), largely irrespective of language generation rules, but one is enhanced with a pre-trained n-gram language model to fuse it with prior linguistic knowledge. We then trained four neural network models (BOW, CNN, LSTM, and GRU) and a pre-trained model (ERNIE-Gram) on the LCQMCs train sets of varying size as well as the related augmented train sets produced by the two DA programs. The results show that there are no significant performance differences between the models trained on the two types of augmented train sets, both when the five DA techniques are applied together or separately. Moreover, due to the inability of the five DA techniques to make strictly paraphrastic augmented texts, the results indicate the need of sufficient amounts of training examples for the classification models trained on them to mediate the negative impact of false matching augmented text pairs and improve performance, a limitation of random text editing perturbations used as a DA approach. Similar results were also obtained for English.
|
['Zhengxiang Wang']
|
2021-11-29
| null | null | null | null |
['text-augmentation', 'question-similarity']
|
['natural-language-processing', 'natural-language-processing']
|
[ 4.43213761e-01 2.84712076e-01 3.60438138e-01 -2.73527861e-01
-8.40307295e-01 -5.10556161e-01 7.19970942e-01 3.19700003e-01
-1.00504279e+00 4.74605143e-01 2.71438181e-01 -9.10027921e-01
2.13097438e-01 -9.22339380e-01 -7.64846742e-01 -1.64554209e-01
3.44730616e-01 3.90300274e-01 2.31745824e-01 -6.94652677e-01
1.19536415e-01 3.10350448e-01 -1.40592635e+00 8.32017839e-01
1.24096763e+00 6.92843318e-01 2.51413792e-01 9.84024346e-01
-5.75363696e-01 1.09665644e+00 -9.10443366e-01 -8.61732781e-01
2.70371050e-01 -6.03004336e-01 -9.79759216e-01 -2.98643380e-01
5.55301845e-01 -2.10672304e-01 -4.26034868e-01 8.27523828e-01
4.39155608e-01 2.58011729e-01 4.15235966e-01 -8.96788657e-01
-1.22817028e+00 7.77087808e-01 1.29142493e-01 2.33366117e-01
6.53520107e-01 -5.75411096e-02 9.15658355e-01 -1.03860366e+00
4.85187173e-01 1.14048457e+00 7.71259725e-01 7.11407304e-01
-1.04567266e+00 -6.21913075e-01 -2.65344590e-01 1.35853186e-01
-9.16325986e-01 -5.38938224e-01 5.66125333e-01 -9.91602018e-02
1.41424274e+00 4.10985380e-01 1.93985701e-01 9.15421009e-01
-6.43132403e-02 6.58642828e-01 9.36843693e-01 -1.34409237e+00
-5.57643510e-02 7.00740933e-01 5.79756141e-01 4.23860103e-01
6.35390729e-02 1.78517282e-01 8.65463018e-02 -2.51814157e-01
2.04638615e-01 -4.13764983e-01 -3.47113431e-01 2.86025733e-01
-9.42711890e-01 1.08904624e+00 2.43461132e-01 8.21550429e-01
-1.52395144e-01 -2.41002053e-01 5.43588281e-01 8.06877136e-01
1.88879341e-01 9.44470942e-01 -7.99138069e-01 7.85062686e-02
-4.27993059e-01 2.53383130e-01 8.88129354e-01 1.28695631e+00
6.66360319e-01 2.99961343e-02 -5.82401156e-01 9.98460948e-01
-1.95099413e-01 4.39729780e-01 1.18282712e+00 -4.39129919e-01
8.68258059e-01 8.91570389e-01 -4.04198430e-02 -1.16247308e+00
-2.63696611e-01 -8.80584121e-02 -6.01419628e-01 -4.83361274e-01
6.70846224e-01 -2.81596690e-01 -7.37688363e-01 1.74656749e+00
-1.27307594e-01 -4.09826070e-01 3.28763574e-01 2.93870598e-01
1.13306463e+00 7.76778519e-01 1.43507555e-01 2.07316995e-01
1.39647424e+00 -1.03790474e+00 -8.78008544e-01 -3.10977548e-01
1.42553306e+00 -9.13240969e-01 1.69243550e+00 -6.54059350e-02
-8.81771028e-01 -9.56216037e-01 -9.81544495e-01 -3.08115870e-01
-7.83655465e-01 3.45080107e-01 2.73042381e-01 1.09115994e+00
-9.39972281e-01 4.64082897e-01 -1.23181209e-01 -4.72420096e-01
-1.83938473e-01 1.33165419e-01 -3.18258137e-01 -2.70535976e-01
-1.65925431e+00 1.41684020e+00 4.23919111e-01 -4.07558717e-02
3.62575729e-03 -5.97629845e-01 -8.91975582e-01 2.01464787e-01
-2.71868892e-02 -5.17173260e-02 1.32216132e+00 -1.20299947e+00
-1.36244226e+00 8.93582463e-01 -1.05343841e-01 -4.30667758e-01
2.07033604e-01 -1.50565445e-01 -6.94810748e-01 -1.88130569e-02
-8.58524144e-02 7.09585428e-01 4.44819987e-01 -8.61266792e-01
-4.80494618e-01 -2.22786978e-01 -9.69015714e-03 -7.56679699e-02
-6.06513023e-01 5.70593894e-01 -2.53601223e-01 -8.39035809e-01
-2.85317332e-01 -6.66220248e-01 -1.35933489e-01 -4.28674400e-01
-2.11320817e-02 -3.11823845e-01 6.21431768e-01 -1.31687891e+00
1.40259850e+00 -1.95624614e+00 -4.28359032e-01 1.54227823e-01
-3.82572949e-01 7.92741776e-01 -7.84383357e-01 5.85269630e-01
-2.00145513e-01 4.00485575e-01 -2.49514759e-01 -2.54412860e-01
-5.28729148e-02 2.61195123e-01 -1.88224241e-01 -9.07479450e-02
5.24819434e-01 1.22575700e+00 -7.68845439e-01 -2.34218106e-01
4.83580772e-03 -2.94197630e-02 -5.17878711e-01 4.29839849e-01
-2.38127589e-01 -1.60529613e-01 -2.92278789e-02 2.68678248e-01
5.47642469e-01 8.62698629e-02 -3.67044434e-02 1.05417185e-01
3.09543282e-01 5.69139779e-01 -7.26113558e-01 1.17118609e+00
-5.67839921e-01 8.05923700e-01 -3.11965257e-01 -9.79127109e-01
1.25739622e+00 5.13008356e-01 -1.90155759e-01 -1.29061246e+00
2.22074568e-01 1.83202416e-01 2.41105095e-01 -6.09151661e-01
1.07645011e+00 -5.97391650e-02 -1.38247713e-01 6.47490621e-01
5.33962734e-02 -1.50190547e-01 3.49829167e-01 2.22452730e-01
1.18039513e+00 -2.70738333e-01 1.38707161e-01 -1.47897273e-01
1.00361788e+00 4.62542802e-01 3.23835582e-01 1.18853772e+00
-7.02574626e-02 4.39505577e-01 2.11065874e-01 -2.61381418e-01
-1.27539694e+00 -4.13333476e-01 -8.46582353e-02 1.34129429e+00
-4.31741357e-01 -2.86122531e-01 -8.57514918e-01 -8.03981960e-01
-7.18603209e-02 1.33433723e+00 -6.20202482e-01 -5.50923049e-01
-9.57021594e-01 -5.08628070e-01 1.15959871e+00 5.56898713e-01
3.16260785e-01 -1.55978012e+00 -3.25173050e-01 4.49457049e-01
-3.03181946e-01 -1.27737033e+00 -5.19537210e-01 1.60358280e-01
-6.82290554e-01 -1.08474922e+00 -4.97498810e-01 -1.23811471e+00
6.54237986e-01 1.70890898e-01 1.21174908e+00 6.28371060e-01
1.16179928e-01 3.89133364e-01 -9.54011500e-01 -3.74074787e-01
-1.29800510e+00 2.80146420e-01 -2.56201416e-01 -2.41409808e-01
9.76885617e-01 -3.26369591e-02 2.12145120e-01 2.79340148e-01
-1.16357541e+00 1.06528223e-01 5.40346861e-01 1.12183166e+00
-1.54166445e-01 -3.53243798e-01 8.63994956e-01 -9.37070549e-01
1.09025300e+00 -2.13273168e-01 -6.28586829e-01 6.91045046e-01
-4.55836743e-01 9.71688032e-02 7.93747246e-01 -8.86419535e-01
-1.13856566e+00 -2.08393976e-01 -3.73777807e-01 -1.21928506e-01
-2.06297502e-01 7.28732288e-01 -1.51167855e-01 -8.27440619e-02
1.06953490e+00 2.85152346e-01 8.18448737e-02 -6.21789753e-01
5.63843369e-01 1.08149171e+00 6.96050227e-01 -5.70910037e-01
6.32944226e-01 -2.87530333e-01 -8.63214672e-01 -6.64688051e-01
-5.57373226e-01 -3.39372188e-01 -6.12605572e-01 6.49033338e-02
8.29882562e-01 -6.18722320e-01 -3.42091322e-01 5.19258857e-01
-1.36556137e+00 -3.91285777e-01 -3.14197868e-01 3.43240380e-01
1.61925480e-02 4.93218571e-01 -1.03015661e+00 -5.26838720e-01
-6.64747059e-01 -8.66023540e-01 3.96804571e-01 -8.26300122e-03
-3.54201943e-01 -9.64370787e-01 4.63012420e-03 5.24524331e-01
6.76954329e-01 -3.10806513e-01 1.58052480e+00 -1.32611084e+00
1.39481556e-02 -6.35485291e-01 -1.96259066e-01 8.02109897e-01
1.93800554e-01 -1.11228913e-01 -9.20292377e-01 -2.46700555e-01
2.11936295e-01 -4.67382669e-01 4.38911170e-01 -2.99334943e-01
9.09158587e-01 -5.30643582e-01 2.61222303e-01 9.60338414e-02
9.93191838e-01 4.00811464e-01 8.33826482e-01 3.88584852e-01
3.58831227e-01 6.69847429e-01 3.12203616e-01 -2.29409993e-01
1.82196692e-01 6.50178850e-01 -1.69508174e-01 -5.16696908e-02
1.50182415e-02 -4.15563524e-01 4.81333613e-01 1.25073206e+00
6.07530773e-01 -2.52031267e-01 -9.69554007e-01 5.61670303e-01
-1.37275052e+00 -6.71719253e-01 -4.51523274e-01 2.19674683e+00
1.02957618e+00 -2.89108567e-02 -4.08484966e-01 2.79694051e-01
7.48672783e-01 -3.54693711e-01 -1.02166466e-01 -8.61471176e-01
-4.25717562e-01 6.80770934e-01 3.22137266e-01 3.64371598e-01
-8.79740477e-01 9.71309364e-01 6.39324427e+00 7.21931875e-01
-9.42939818e-01 6.05740882e-02 5.30640244e-01 4.12269294e-01
-5.14903724e-01 2.24873750e-03 -6.72321081e-01 3.02071422e-01
1.27468753e+00 -8.40721577e-02 3.39577198e-01 7.11254656e-01
-1.92456946e-01 -8.45494345e-02 -9.82755303e-01 4.27595675e-01
2.50367820e-01 -1.36832106e+00 2.68737495e-01 -4.11339790e-01
6.24735892e-01 -7.83328116e-02 -4.75415468e-01 9.76748645e-01
3.70964497e-01 -8.89534295e-01 5.62783062e-01 2.08880767e-01
5.86129427e-01 -3.67540896e-01 1.08040321e+00 5.60455322e-01
-5.36826551e-01 -2.02575967e-01 -6.94872320e-01 -1.65018305e-01
-6.84729666e-02 6.13607727e-02 -7.91446924e-01 5.59226096e-01
5.89986324e-01 -1.39372125e-02 -1.21282136e+00 5.17269492e-01
-1.30398825e-01 7.61181891e-01 -2.02516720e-01 -5.25325954e-01
2.34433725e-01 -6.60485625e-02 1.44800037e-01 1.42155898e+00
1.79549500e-01 -6.38756156e-02 -5.49288392e-01 8.00541461e-01
-1.91925675e-01 5.28403640e-01 -3.77310097e-01 -2.20376328e-01
6.47946656e-01 1.08375680e+00 -1.94963366e-01 -6.34721398e-01
-9.32389319e-01 8.63811493e-01 6.11480236e-01 4.12038296e-01
-5.06662786e-01 -6.33449852e-01 -5.01170941e-02 -6.04030229e-02
1.72751427e-01 2.24193096e-01 -5.30864835e-01 -1.26742291e+00
2.04649314e-01 -1.42737246e+00 5.23197651e-01 -1.08710635e+00
-1.48409677e+00 8.47454965e-01 -2.45494634e-01 -8.70571554e-01
-3.32752824e-01 -6.88587427e-01 -8.06995034e-01 1.21565247e+00
-1.19372880e+00 -8.99217606e-01 -3.07729781e-01 6.41851962e-01
3.19292307e-01 -4.32867467e-01 1.04135251e+00 5.75986564e-01
-5.57380438e-01 1.14658606e+00 1.47997603e-01 6.93800986e-01
7.29092419e-01 -9.13044512e-01 6.62057161e-01 1.05787456e+00
1.11465171e-01 7.21195042e-01 2.19865710e-01 -5.68397164e-01
-1.17156982e+00 -1.12741423e+00 1.64796007e+00 -4.93424773e-01
4.51773703e-01 -5.42419791e-01 -1.57818222e+00 6.88444972e-01
2.71043956e-01 -4.87583876e-01 6.58478439e-01 -5.57758696e-02
-2.16091707e-01 1.23683617e-01 -1.09520829e+00 5.54989755e-01
4.00773525e-01 -7.59968877e-01 -1.34632909e+00 7.66661465e-02
1.22008467e+00 -1.21677227e-01 -7.90388465e-01 4.60688531e-01
1.94984093e-01 -5.58575392e-01 5.44507384e-01 -1.13338792e+00
4.13461953e-01 8.88677910e-02 -2.24277854e-01 -9.71902609e-01
-3.19663733e-01 -9.17989761e-02 3.50074649e-01 1.46594167e+00
8.67629051e-01 -7.06814229e-01 6.43812239e-01 1.07495594e+00
-3.36741060e-01 -3.69824350e-01 -7.90893257e-01 -7.60444701e-01
5.87910295e-01 -5.60397387e-01 8.31749678e-01 1.34508276e+00
2.07525149e-01 5.04625618e-01 -5.53845279e-02 -2.77591556e-01
-3.45576376e-01 -4.56637025e-01 9.49930906e-01 -6.98984146e-01
-2.76366442e-01 -4.08196300e-01 6.21195361e-02 -8.70504498e-01
2.05019787e-01 -1.06772828e+00 9.21941176e-03 -1.07158971e+00
-1.01194970e-01 -6.74623370e-01 7.84082413e-02 5.63705087e-01
-7.79835045e-01 -1.96257517e-01 3.69812518e-01 3.58230956e-02
1.91822499e-01 6.35203958e-01 8.51999938e-01 4.39803228e-02
-3.58065248e-01 5.96217960e-02 -7.60166168e-01 2.55197018e-01
9.37532783e-01 -5.24178624e-01 -7.40315169e-02 -5.24528325e-01
2.09101602e-01 7.55894706e-02 1.59349680e-01 -7.21534908e-01
2.94363827e-01 3.14829022e-01 2.95310706e-01 -5.63445210e-01
-9.18496996e-02 -7.64224589e-01 -2.60575175e-01 5.34925640e-01
-7.79395461e-01 5.64813912e-01 6.36990666e-01 -8.82603228e-02
-3.30374867e-01 -7.92905450e-01 6.87906384e-01 -1.04537249e-01
-4.75111723e-01 -3.35578769e-01 -8.43828380e-01 2.28664130e-01
5.08558393e-01 -3.74789447e-01 -3.73590112e-01 -5.71686387e-01
-2.80581325e-01 1.33506432e-01 2.24417016e-01 8.36091101e-01
3.06184679e-01 -1.32362914e+00 -7.89690793e-01 5.39875269e-01
3.01720530e-01 -2.30656192e-01 -7.04657845e-03 4.12413508e-01
-5.90218306e-01 7.25152254e-01 -1.48761228e-01 -6.91469461e-02
-1.05819333e+00 8.33106041e-01 4.43860143e-01 -5.66891849e-01
-2.67736644e-01 5.15828431e-01 -2.61845767e-01 -1.31884134e+00
2.92313516e-01 -4.70750839e-01 -3.60119224e-01 -2.77891755e-01
5.22228241e-01 2.64629364e-01 5.28436601e-01 -4.53876436e-01
1.13599030e-02 -2.10362300e-02 -2.41167769e-01 -5.81356548e-02
9.00888562e-01 3.16495150e-02 -1.29523546e-01 -3.01856678e-02
1.33921504e+00 2.49996688e-02 -2.13566437e-01 -6.19097650e-01
4.74078476e-01 -1.01818457e-01 -2.94070512e-01 -7.83406436e-01
-6.74200296e-01 8.58400822e-01 3.05165052e-01 6.86162803e-03
1.05803645e+00 -3.14841181e-01 7.68051624e-01 8.53973627e-01
-1.36005297e-01 -1.13264978e+00 -6.27761558e-02 9.78809476e-01
1.00828099e+00 -1.16185999e+00 -4.85210776e-01 -2.63627231e-01
-6.10234141e-01 1.00301886e+00 9.05895352e-01 1.56800702e-01
4.07456309e-01 -2.37832051e-02 4.22068268e-01 1.60304382e-01
-7.37794995e-01 -8.14440623e-02 4.32304025e-01 6.90338373e-01
5.19874215e-01 -3.03079605e-01 -4.80264992e-01 8.36182535e-01
-5.12080193e-01 -9.03519690e-02 6.96227491e-01 8.83607984e-01
-1.75117269e-01 -1.26216209e+00 -4.49816704e-01 9.01895940e-01
-3.39368790e-01 -5.60473263e-01 -5.63120306e-01 1.18965483e+00
-1.02883928e-01 1.17967927e+00 3.57175559e-01 -5.87325156e-01
5.33083677e-01 4.54275191e-01 3.40507589e-02 -5.73309243e-01
-1.41391671e+00 -5.24374545e-01 4.35895354e-01 -1.22744650e-01
-3.59610170e-02 -4.62400526e-01 -9.37583148e-01 -3.29682201e-01
-7.85888314e-01 3.24513674e-01 4.44214046e-01 9.85037565e-01
1.87035665e-01 1.24655828e-01 4.08030778e-01 -6.69902787e-02
-8.10217261e-01 -1.56656861e+00 -1.18643202e-01 9.10423040e-01
-1.72754422e-01 1.66393489e-01 -4.31302607e-01 -5.49502708e-02]
|
[10.988883018493652, 9.440794944763184]
|
19788220-5804-4299-a5ed-a581b5c4080f
|
llvip-a-visible-infrared-paired-dataset-for
|
2108.10831
| null |
https://arxiv.org/abs/2108.10831v4
|
https://arxiv.org/pdf/2108.10831v4.pdf
|
LLVIP: A Visible-infrared Paired Dataset for Low-light Vision
|
It is very challenging for various visual tasks such as image fusion, pedestrian detection and image-to-image translation in low light conditions due to the loss of effective target areas. In this case, infrared and visible images can be used together to provide both rich detail information and effective target areas. In this paper, we present LLVIP, a visible-infrared paired dataset for low-light vision. This dataset contains 30976 images, or 15488 pairs, most of which were taken at very dark scenes, and all of the images are strictly aligned in time and space. Pedestrians in the dataset are labeled. We compare the dataset with other visible-infrared datasets and evaluate the performance of some popular visual algorithms including image fusion, pedestrian detection and image-to-image translation on the dataset. The experimental results demonstrate the complementary effect of fusion on image information, and find the deficiency of existing algorithms of the three visual tasks in very low-light conditions. We believe the LLVIP dataset will contribute to the community of computer vision by promoting image fusion, pedestrian detection and image-to-image translation in very low-light applications. The dataset is being released in https://bupt-ai-cz.github.io/LLVIP. Raw data is also provided for further research such as image registration.
|
['Wenli Zhou', 'ShengJie Liu', 'Wenqi Tang', 'Minzhen Li', 'Chuang Zhu', 'Xinyu Jia']
|
2021-08-24
| null | null | null | null |
['multispectral-object-detection', 'infrared-and-visible-image-fusion', 'low-light-pedestrian-detection', 'thermal-infrared-pedestrian-detection']
|
['computer-vision', 'computer-vision', 'computer-vision', 'computer-vision']
|
[ 3.70898336e-01 -9.03043926e-01 -1.08655736e-01 -2.84657925e-01
-4.93744701e-01 -5.71888745e-01 5.96690178e-01 -2.23656833e-01
-5.29444158e-01 6.66393280e-01 -1.09490314e-02 -3.05489570e-01
4.75483924e-01 -8.68100882e-01 -4.62777644e-01 -1.04063141e+00
5.12699544e-01 -2.50384748e-01 3.84035259e-01 -3.94501448e-01
-3.65303349e-05 4.37880456e-01 -1.84410906e+00 4.61800277e-01
6.67989314e-01 7.56188631e-01 3.67187798e-01 8.12121511e-01
1.39199287e-01 5.21268010e-01 -2.48368755e-01 -4.92644310e-01
7.81772435e-01 -2.60105342e-01 -2.79742658e-01 2.43927911e-01
1.08596969e+00 -2.77154744e-01 -4.95220184e-01 1.30813527e+00
8.84243548e-01 2.62977302e-01 3.82901728e-01 -1.50767577e+00
-8.50350320e-01 -1.20494448e-01 -1.10749722e+00 6.14390314e-01
6.11316204e-01 7.20213175e-01 2.37722397e-01 -1.02709031e+00
3.38510185e-01 1.32630742e+00 6.33165300e-01 3.06976378e-01
-9.30844665e-01 -9.23161983e-01 -2.45747223e-01 4.73193735e-01
-1.43373239e+00 -6.76772237e-01 6.93357587e-01 -4.22679633e-01
4.22470868e-01 5.70750594e-01 6.53567493e-01 9.58837330e-01
1.13773100e-01 3.52408826e-01 1.51202738e+00 -6.10277176e-01
-1.34525478e-01 9.70389545e-02 1.81017801e-01 5.52674413e-01
5.95636487e-01 7.38548458e-01 -4.72626448e-01 1.65141687e-01
5.45182884e-01 4.19282705e-01 -3.43523324e-01 3.19353968e-01
-1.16153955e+00 4.50054049e-01 6.97008431e-01 1.83464140e-01
-1.62475053e-02 -1.66447982e-01 2.06306186e-02 3.09943229e-01
5.72696149e-01 -3.76026064e-01 -9.64919329e-02 6.73236430e-01
-4.74579364e-01 1.57717228e-01 1.02481991e-02 7.55515397e-01
8.58144343e-01 -8.10946673e-02 -2.36474097e-01 9.01924312e-01
4.45794761e-01 1.11055267e+00 1.13467881e-02 -9.48693275e-01
6.82005465e-01 3.43931884e-01 2.95969427e-01 -9.75710511e-01
-3.59166831e-01 -1.64238945e-01 -1.13646221e+00 7.28584826e-01
5.11170864e-01 -1.53112531e-01 -8.26391280e-01 1.22825158e+00
6.04253471e-01 1.27226606e-01 9.53373387e-02 1.20765865e+00
1.48642075e+00 7.70807862e-01 6.92046955e-02 -3.33958805e-01
1.73742938e+00 -9.63475883e-01 -6.26121342e-01 -5.57941735e-01
1.18426278e-01 -1.38299310e+00 7.84655273e-01 1.03559375e-01
-1.01236594e+00 -1.16435492e+00 -7.20548928e-01 -2.00886056e-01
-4.96646821e-01 4.16171730e-01 4.76645380e-01 8.43858778e-01
-1.15484357e+00 -1.55017182e-01 -7.61817694e-02 -7.22248375e-01
3.55067730e-01 5.30342832e-02 -5.65852165e-01 -5.95804811e-01
-1.02517104e+00 7.53945589e-01 4.29736853e-01 2.75369316e-01
-5.34345686e-01 -4.51674998e-01 -8.00614715e-01 -5.82366705e-01
2.48109952e-01 -9.54140127e-01 6.61384284e-01 -5.83489835e-01
-8.83976340e-01 1.42344177e+00 -2.88147062e-01 -1.99001461e-01
5.65562427e-01 1.64848208e-01 -6.43794715e-01 4.79176976e-02
3.93307924e-01 6.00730836e-01 6.91694498e-01 -1.52192569e+00
-9.58503366e-01 -6.21975362e-01 -6.11068979e-02 3.27228010e-01
3.42026129e-02 5.03706694e-01 -6.81283593e-01 -4.06858653e-01
-9.21268836e-02 -8.76401305e-01 -1.42202944e-01 2.44528279e-01
-3.75349998e-01 -1.25623807e-01 1.02240145e+00 -7.95766175e-01
4.86377805e-01 -2.03791904e+00 -5.66313803e-01 -2.93996036e-01
2.44130448e-01 6.94663227e-01 -2.07888126e-01 1.92443535e-01
-9.15202349e-02 -2.28810549e-01 -5.15764430e-02 -2.34306097e-01
-6.22677207e-01 -5.62717542e-02 -4.52719778e-02 9.06523466e-01
-2.95007765e-01 9.61911142e-01 -9.12221014e-01 -7.84840882e-01
9.24415231e-01 6.37167573e-01 2.32664734e-01 1.34353951e-01
3.88115227e-01 7.66337931e-01 -2.11050645e-01 1.02484429e+00
1.28619182e+00 2.75630474e-01 -4.91080940e-01 -6.07183695e-01
-4.92518097e-01 -7.08038628e-01 -1.04558980e+00 1.16929829e+00
-2.81437010e-01 8.88936043e-01 -5.64534450e-04 -6.98933780e-01
1.02142179e+00 -1.62957553e-02 5.43860316e-01 -1.08489656e+00
2.33000711e-01 -8.98373500e-02 -3.22571605e-01 -6.71406031e-01
4.33225483e-01 1.25704631e-01 3.09854746e-01 1.05269514e-01
-3.76317531e-01 -4.33469787e-02 4.77966994e-01 -4.85558510e-02
3.94031495e-01 -1.08535841e-01 1.84667766e-01 1.46103829e-01
8.10005248e-01 -1.40753584e-02 3.36410403e-01 8.99349928e-01
-5.76368809e-01 8.03562105e-01 -5.37991047e-01 -5.60912192e-01
-1.11077654e+00 -1.31247902e+00 -3.58750403e-01 1.14242327e+00
6.89526439e-01 9.23824906e-02 -4.07833785e-01 -1.38013616e-01
-2.35293031e-01 4.18353438e-01 -3.21247101e-01 1.92114525e-02
-3.82311225e-01 -1.07302606e+00 3.24640721e-01 2.41102338e-01
1.29181731e+00 -8.59235942e-01 -4.20669436e-01 -3.41795057e-01
-5.67161560e-01 -1.43243945e+00 -5.42915404e-01 -4.33942109e-01
-4.35678810e-01 -1.37643313e+00 -8.27671766e-01 -9.39515889e-01
5.84175527e-01 1.50399375e+00 9.64779317e-01 2.62124151e-01
-7.83921480e-01 3.70503396e-01 -2.79741794e-01 -4.35555667e-01
-1.89911112e-01 -6.93688631e-01 9.03246030e-02 2.65302062e-01
4.32137877e-01 -1.20473385e-01 -1.01213443e+00 5.55868983e-01
-7.01574385e-01 1.91173568e-01 4.16714072e-01 7.02834785e-01
5.32409728e-01 2.80065417e-01 -1.91912577e-01 -2.95008034e-01
2.31027797e-01 -2.67232340e-02 -7.14910686e-01 3.49039555e-01
-1.60346374e-01 -6.13669753e-01 3.13501418e-01 -2.19567731e-01
-1.51543415e+00 2.28257075e-01 6.35477081e-02 -1.58955798e-01
-5.41405439e-01 -4.05844361e-01 -2.08673581e-01 -5.07210553e-01
9.79033649e-01 3.26360255e-01 -1.50176272e-01 -4.40271020e-01
4.53192264e-01 9.77247059e-01 1.08355558e+00 -1.33202255e-01
1.04252446e+00 8.80937219e-01 1.00709870e-01 -1.07013774e+00
-9.34754789e-01 -9.77973104e-01 -6.98349059e-01 -5.95109940e-01
1.07568789e+00 -1.29004252e+00 -7.65801132e-01 7.11689889e-01
-1.08442950e+00 -3.19602489e-02 -8.64693969e-02 6.58557475e-01
-1.47862807e-01 6.19634569e-01 -4.09582555e-01 -1.05780435e+00
-5.34459233e-01 -1.05095387e+00 1.18535876e+00 6.77568316e-01
6.71439111e-01 -7.50051320e-01 -1.17579214e-01 9.58303690e-01
2.68619686e-01 3.36942106e-01 -3.37778367e-02 2.96124607e-01
-4.40305054e-01 -1.48475170e-01 -8.07231665e-01 3.03388417e-01
1.20239936e-01 7.48461857e-02 -1.22041965e+00 -3.84831160e-01
-6.30424395e-02 -2.81130876e-02 1.32050216e+00 6.15952194e-01
7.79716909e-01 1.08506726e-02 -4.52986717e-01 7.38829672e-01
1.51916933e+00 1.96667612e-01 8.19541931e-01 4.19454902e-01
9.58661020e-01 6.66405618e-01 8.06144953e-01 1.29718408e-01
3.31940621e-01 1.07918143e+00 4.61111993e-01 -9.46380794e-01
-8.35968792e-01 2.15151563e-01 4.95452553e-01 1.08162329e-01
-5.91520607e-01 -9.28849280e-02 -9.37503755e-01 3.07531595e-01
-1.77240133e+00 -1.52644277e+00 -1.06969798e+00 2.33553457e+00
5.83091438e-01 -4.42102045e-01 2.64965385e-01 1.03281938e-01
1.21316922e+00 1.31491542e-01 -1.14051573e-01 1.23393007e-01
-6.21932387e-01 -1.98256522e-01 9.11152005e-01 5.62765360e-01
-1.35399437e+00 7.91855633e-01 5.83363485e+00 8.56790662e-01
-9.45669055e-01 4.32989836e-01 8.73248398e-01 1.68985054e-01
3.83890301e-01 -8.20587948e-02 -9.81265247e-01 6.92785323e-01
3.62282902e-01 -1.53771229e-02 2.78681308e-01 4.29551333e-01
5.68834662e-01 -5.18522799e-01 -5.09025574e-01 1.60569119e+00
3.16541523e-01 -9.42657888e-01 -2.17667505e-01 -1.12790717e-02
9.38736618e-01 1.75419345e-01 2.10055381e-01 -2.97335863e-01
2.02566609e-01 -9.08447146e-01 4.54376072e-01 6.20014429e-01
9.39120173e-01 -6.52742863e-01 7.55714715e-01 2.70915896e-01
-1.66222692e+00 -9.30379257e-02 -8.13958704e-01 5.80012938e-03
1.36497125e-01 6.08859062e-01 -4.56221282e-01 6.81807220e-01
9.85148370e-01 8.68177533e-01 -1.11597335e+00 1.28385818e+00
-1.51225820e-01 1.68581009e-01 -1.99652299e-01 3.19500089e-01
-1.97889626e-01 -6.43123090e-01 4.00224984e-01 1.12522352e+00
1.35561079e-01 4.83776294e-02 6.23536050e-01 7.67124772e-01
3.11800927e-01 -1.42390296e-01 -1.02761221e+00 6.42099142e-01
2.75087088e-01 1.52290177e+00 -5.75528383e-01 -3.66068006e-01
-5.81519067e-01 9.37705696e-01 -5.00588775e-01 6.01463437e-01
-9.53764617e-01 -1.56452686e-01 5.29188991e-01 2.40058735e-01
-2.06652731e-01 -2.35468075e-01 -3.56976479e-01 -1.10998583e+00
9.64555815e-02 -4.57719505e-01 5.28251648e-01 -1.29218864e+00
-1.34929907e+00 4.45447832e-01 1.27644032e-01 -1.37408352e+00
3.44720274e-01 -6.05273306e-01 -6.43581331e-01 9.94093657e-01
-1.77942669e+00 -1.69334781e+00 -9.65690970e-01 1.15486765e+00
6.10369265e-01 -2.74897277e-01 1.21224783e-01 4.76761818e-01
-5.97890019e-01 3.99255306e-01 2.89033026e-01 2.48318106e-01
8.36555779e-01 -6.55596972e-01 2.40487471e-01 1.28672767e+00
6.04781695e-02 1.50277689e-01 6.74648225e-01 -5.85185587e-01
-1.32595742e+00 -1.47126472e+00 6.36749864e-01 -5.67264855e-01
1.50813103e-01 -1.36319116e-01 -3.95427912e-01 3.60956967e-01
3.71343851e-01 4.62091744e-01 3.99296552e-01 -5.86531162e-01
-1.56247541e-01 -4.64047909e-01 -1.26231003e+00 4.79369670e-01
1.13423586e+00 -3.39844882e-01 -2.73979664e-01 8.19343090e-01
4.43165213e-01 -2.16737360e-01 -5.25146902e-01 3.50192010e-01
3.94481063e-01 -1.23485887e+00 1.86209512e+00 1.44282877e-01
8.97552893e-02 -7.67609537e-01 -2.24284545e-01 -8.50360155e-01
-3.29213083e-01 -2.03052223e-01 3.99312645e-01 1.39615035e+00
-2.16407329e-01 -8.30261707e-01 3.06770891e-01 2.81452090e-01
2.29057372e-01 4.39933091e-02 -7.70783246e-01 -6.70658588e-01
-3.79653573e-01 -3.20414782e-01 2.04492271e-01 7.50235915e-01
-8.09244692e-01 2.08605468e-01 -6.64801478e-01 2.66001046e-01
1.41873956e+00 2.17665061e-01 1.06346726e+00 -1.07781553e+00
-4.41098139e-02 -1.94207340e-01 -4.91287977e-01 -4.49840337e-01
-2.21154064e-01 -7.32625306e-01 -9.98719484e-02 -1.78234076e+00
6.19292498e-01 -1.71481520e-01 7.20303282e-02 5.38938284e-01
-5.25714636e-01 1.18454266e+00 3.48428279e-01 4.59778696e-01
-6.51017308e-01 2.20200583e-01 1.41906214e+00 -4.68857050e-01
1.28813228e-02 5.84259778e-02 -6.17287993e-01 6.97086930e-01
1.07153094e+00 -1.90196246e-01 -1.32674342e-02 -2.40759358e-01
-4.37647849e-01 -1.81594163e-01 9.60578322e-01 -1.13919199e+00
4.18525875e-01 -4.43532407e-01 1.01417923e+00 -7.88565516e-01
5.14072239e-01 -7.11724699e-01 2.87656069e-01 6.85370803e-01
1.38012946e-01 -1.41907647e-01 2.14416161e-02 5.00205100e-01
-4.21706289e-02 1.18166789e-01 1.19571376e+00 -2.87185937e-01
-1.07124019e+00 3.90633285e-01 -2.86147948e-02 -2.06246197e-01
1.47107399e+00 -5.13733685e-01 -9.68517900e-01 -1.35903433e-01
-2.41980225e-01 3.01657617e-02 6.15956187e-01 4.53648686e-01
7.69036829e-01 -1.47819054e+00 -1.18561375e+00 3.15097213e-01
3.89633298e-01 -3.18125337e-01 4.94660288e-01 1.01282489e+00
-6.22144282e-01 2.42860481e-01 -5.25586784e-01 -8.31730366e-01
-2.13895774e+00 7.08334744e-01 3.46130759e-01 3.43127787e-01
-7.57369816e-01 6.25494421e-01 4.32899892e-01 -1.45632684e-01
-5.17673604e-02 1.88932389e-01 -3.52616102e-01 -2.66574442e-01
1.06324530e+00 6.55338109e-01 -1.42922208e-01 -1.29917669e+00
-3.37837398e-01 9.88018394e-01 3.11132491e-01 1.92185521e-01
1.02096450e+00 -6.70033991e-01 -2.48266220e-01 1.30110577e-01
1.04767144e+00 -1.12238713e-01 -9.51118767e-01 -4.20171857e-01
-8.87993395e-01 -1.17857587e+00 1.71042919e-01 -5.87476373e-01
-1.26936150e+00 9.06998217e-01 1.40372443e+00 5.82643300e-02
1.40548325e+00 4.76955399e-02 6.73634171e-01 1.28813475e-01
3.81180733e-01 -7.73779631e-01 1.41419426e-01 1.64297208e-01
6.21201634e-01 -1.78244853e+00 2.27424473e-01 -6.02011263e-01
-5.19624650e-01 1.06367767e+00 6.92541957e-01 2.24946305e-01
2.10511386e-01 2.47476652e-01 3.93683106e-01 5.34753539e-02
-1.43267974e-01 -8.94448102e-01 2.90447533e-01 1.42408407e+00
3.35840613e-01 1.76120289e-02 -1.88712373e-01 -1.67741179e-01
4.44705561e-02 -1.90594107e-01 4.50735241e-01 6.66243911e-01
-7.05454707e-01 -9.45143700e-01 -1.29839730e+00 1.07912458e-01
-2.07097441e-01 -1.17962420e-01 -4.83032346e-01 3.67199779e-01
5.43928862e-01 1.63768196e+00 -2.25691721e-01 -3.45221698e-01
1.42745465e-01 -5.49412608e-01 4.63759959e-01 -1.20420210e-01
-4.41843987e-01 -1.11347884e-01 -1.75750002e-01 -4.08992946e-01
-8.59772265e-01 -3.53288800e-01 -6.57362103e-01 -7.77193546e-01
-3.54485482e-01 -3.62347811e-01 7.55214214e-01 4.45822954e-01
1.96110085e-02 1.93531901e-01 8.03771615e-01 -1.24939966e+00
3.32048871e-02 -7.02962518e-01 -5.62911689e-01 7.52968073e-01
2.77266443e-01 -5.80032170e-01 -2.82291591e-01 3.49750668e-01]
|
[10.006142616271973, -1.5978096723556519]
|
a39dc226-6e8c-4bcc-a712-47e7e17cee20
|
conciseness-an-overlooked-language-task
|
2211.04126
| null |
https://arxiv.org/abs/2211.04126v1
|
https://arxiv.org/pdf/2211.04126v1.pdf
|
Conciseness: An Overlooked Language Task
|
We report on novel investigations into training models that make sentences concise. We define the task and show that it is different from related tasks such as summarization and simplification. For evaluation, we release two test sets, consisting of 2000 sentences each, that were annotated by two and five human annotators, respectively. We demonstrate that conciseness is a difficult task for which zero-shot setups with large neural language models often do not perform well. Given the limitations of these approaches, we propose a synthetic data generation method based on round-trip translations. Using this data to either train Transformers from scratch or fine-tune T5 models yields our strongest baselines that can be further improved by fine-tuning on an artificial conciseness dataset that we derived from multi-annotator machine translation test sets.
|
['Shankar Kumar', 'Chris Alberti', 'Aashish Kumar', 'Felix Stahlberg']
|
2022-11-08
| null | null | null | null |
['synthetic-data-generation', 'synthetic-data-generation']
|
['medical', 'miscellaneous']
|
[ 4.50485080e-01 5.20530641e-01 -8.58782753e-02 -4.97149616e-01
-1.50401211e+00 -6.15002990e-01 8.42739224e-01 2.47462660e-01
-6.08039737e-01 1.26491904e+00 6.36577129e-01 -3.14286292e-01
2.27775007e-01 -4.21854019e-01 -8.66029084e-01 -1.70948625e-01
2.91273803e-01 9.38554466e-01 -6.79182038e-02 -5.66153765e-01
2.40766853e-01 1.78253382e-01 -8.17501366e-01 4.78056520e-01
1.19870222e+00 3.46979529e-01 3.31486091e-02 8.94490600e-01
7.60245398e-02 5.69376826e-01 -1.06485999e+00 -8.87702465e-01
1.59910083e-01 -4.29031998e-01 -1.19343448e+00 -4.13263664e-02
5.88768125e-01 -2.34929398e-01 -1.77139372e-01 7.95585752e-01
7.22179592e-01 1.81064412e-01 7.50456035e-01 -6.19209051e-01
-9.33513761e-01 1.05516171e+00 -5.94165511e-02 2.29995683e-01
3.72840941e-01 1.95912570e-01 1.08558536e+00 -7.66072452e-01
8.98161292e-01 1.29190040e+00 9.15324807e-01 8.81026566e-01
-1.62243497e+00 -1.88700244e-01 -1.88151509e-01 -1.63985327e-01
-1.04598069e+00 -9.04572129e-01 4.51273918e-01 2.38251295e-02
1.38035846e+00 5.03831863e-01 1.61683872e-01 1.60932958e+00
5.17620817e-02 7.04893112e-01 7.75795162e-01 -6.29448116e-01
2.26559248e-02 3.00592750e-01 6.47180900e-02 3.37346256e-01
2.94215143e-01 -2.98210084e-01 -2.38525942e-01 -8.48524421e-02
2.06982702e-01 -6.61922336e-01 -2.72622436e-01 1.61800627e-02
-1.48936892e+00 1.00089669e+00 3.21034268e-02 3.07008356e-01
-1.64443940e-01 6.08053282e-02 7.87210107e-01 7.53469467e-01
7.69425273e-01 1.38880801e+00 -4.81468767e-01 -3.73588264e-01
-1.26993752e+00 4.58775669e-01 1.18477142e+00 1.34137094e+00
4.34799522e-01 -1.04144146e-03 -6.98712468e-01 1.14082229e+00
-5.04378736e-01 3.71819347e-01 6.51885867e-01 -1.07913911e+00
9.63343501e-01 9.19857845e-02 3.78674895e-01 -4.80561078e-01
-2.17837617e-01 -2.69356549e-01 -7.46538520e-01 -3.98562580e-01
3.09222698e-01 -3.41993630e-01 -7.57231176e-01 1.87982595e+00
-3.06074440e-01 -5.25739431e-01 3.81972849e-01 4.23895180e-01
8.63930404e-01 7.69628346e-01 -1.27224475e-01 -4.89562690e-01
9.53401864e-01 -1.31348395e+00 -7.67007291e-01 -1.28889099e-01
1.05443645e+00 -1.03463173e+00 1.53134215e+00 3.47367018e-01
-1.55320251e+00 -4.00257915e-01 -9.89888847e-01 -5.63953102e-01
-2.23897621e-01 2.03774691e-01 3.98624837e-01 4.77254450e-01
-1.08722842e+00 1.05546927e+00 -8.19350064e-01 -5.21457732e-01
2.50171304e-01 1.95632368e-01 -4.71071571e-01 9.65499207e-02
-1.16851234e+00 1.48250270e+00 3.26407045e-01 -2.10571438e-01
-9.02511597e-01 -3.80470693e-01 -8.98137391e-01 7.39672128e-03
3.13983202e-01 -1.10097528e+00 1.79911935e+00 -9.98456299e-01
-1.66001010e+00 9.24027801e-01 -2.34807104e-01 -8.78895402e-01
6.12467885e-01 -4.36487496e-01 -6.75524920e-02 -1.74170472e-02
3.08878988e-01 7.02555776e-01 5.64966321e-01 -9.45589542e-01
-1.44932941e-01 2.49031305e-01 2.17370540e-01 1.46254614e-01
-3.31031382e-01 3.76138866e-01 -1.53507918e-01 -8.87319565e-01
-6.62779093e-01 -8.79099071e-01 -3.36682647e-01 -6.39638782e-01
-9.79956448e-01 -9.18715224e-02 2.21396357e-01 -7.73313761e-01
1.36407757e+00 -1.48361742e+00 4.82923687e-01 -3.26672882e-01
-8.07285458e-02 4.29080784e-01 -6.89026415e-01 8.07295620e-01
-4.87774890e-03 5.90515733e-01 -4.33356375e-01 -9.64284718e-01
6.00403845e-02 2.02198744e-01 -6.28636837e-01 2.26272829e-02
5.39271176e-01 1.14462018e+00 -8.68773103e-01 -5.09132743e-01
-2.90798336e-01 3.49279530e-02 -7.08820462e-01 3.30409139e-01
-4.95472193e-01 2.46970206e-01 -8.25002939e-02 2.66925663e-01
1.71515673e-01 6.94524497e-02 -1.21470869e-01 9.53785423e-03
4.14438844e-02 7.98375070e-01 -4.55617815e-01 1.94071841e+00
-7.90599465e-01 8.09961677e-01 -3.54603201e-01 -9.90738094e-01
8.01467240e-01 4.16375726e-01 -1.90331548e-01 -5.48825562e-01
5.68092465e-02 2.64445066e-01 5.63612580e-02 -6.49208426e-01
9.28993583e-01 -3.45950007e-01 -3.73697966e-01 6.73310757e-01
3.06046546e-01 -9.56874073e-01 8.16040874e-01 3.49350929e-01
1.16285920e+00 3.16091403e-02 3.19489479e-01 -2.55925894e-01
3.35771888e-01 1.72853559e-01 3.91075402e-01 9.27032292e-01
1.33629084e-01 1.03832972e+00 7.11095214e-01 -2.89766043e-01
-1.77317309e+00 -7.88768291e-01 8.61280411e-02 1.04397774e+00
-3.93572420e-01 -8.64625156e-01 -1.06277382e+00 -5.24766922e-01
-5.28658986e-01 1.29720712e+00 -5.46038747e-01 -2.33342528e-01
-8.33511829e-01 -7.14996457e-01 8.99384499e-01 3.80811274e-01
1.79750808e-02 -1.02931285e+00 -3.48996490e-01 3.01631540e-01
-4.94361728e-01 -1.15609241e+00 -7.77488172e-01 1.66445121e-01
-9.32006657e-01 -5.44541359e-01 -8.83756757e-01 -8.38315964e-01
4.67298687e-01 1.05799204e-02 1.78607821e+00 -9.66450125e-02
8.25197250e-02 -3.53002325e-02 -3.49508047e-01 -6.14971340e-01
-1.06596875e+00 7.44505405e-01 1.04282737e-01 -6.39326930e-01
-1.01556331e-02 -7.71941423e-01 -1.24830283e-01 7.07565993e-03
-8.65823746e-01 5.29013753e-01 6.64776146e-01 1.08105230e+00
4.65116829e-01 -5.32257617e-01 8.06188107e-01 -1.15260768e+00
1.37208557e+00 -1.38676718e-01 -1.50079727e-01 4.19497877e-01
-2.92220682e-01 4.58577305e-01 1.03770077e+00 -4.62164968e-01
-8.90204728e-01 -3.33377123e-01 -1.49257749e-01 -1.48166791e-01
-2.83602420e-02 4.17662621e-01 -1.63032971e-02 3.34976524e-01
9.69518661e-01 1.43429667e-01 -1.44482896e-01 -5.35291195e-01
8.27473998e-01 6.82118595e-01 5.56014240e-01 -8.14564407e-01
7.46417642e-01 -1.13359116e-01 -3.39347899e-01 -7.10084438e-01
-1.06272364e+00 5.40652275e-02 -5.71553946e-01 5.33880591e-01
6.47005498e-01 -8.15679193e-01 3.35541777e-02 -3.20898369e-02
-1.70188498e+00 -5.71821272e-01 -6.46906435e-01 2.28444770e-01
-9.39563394e-01 3.86239022e-01 -9.05174255e-01 -3.18917215e-01
-8.63818765e-01 -1.07386541e+00 1.23493040e+00 -6.25073463e-02
-8.68825316e-01 -9.51630294e-01 4.30190504e-01 2.50629932e-01
6.58060193e-01 2.19653413e-01 9.63538349e-01 -1.05540454e+00
-2.04517365e-01 -1.95999593e-01 3.58555838e-02 6.24484897e-01
1.00194022e-01 2.75489390e-01 -7.82277703e-01 -1.62477061e-01
2.11699195e-02 -9.43179309e-01 8.42002690e-01 1.82571843e-01
1.06298661e+00 -7.36759126e-01 5.46453334e-03 4.59613711e-01
1.01404333e+00 -3.14931303e-01 6.55344784e-01 3.72422844e-01
6.00562274e-01 5.38710713e-01 3.98907572e-01 1.51236787e-01
3.53264004e-01 7.69063175e-01 -1.31999448e-01 -6.19955286e-02
-2.02773735e-01 -3.16809952e-01 2.77042598e-01 1.24089026e+00
-1.62901714e-01 -5.05301952e-01 -6.81400776e-01 7.07435489e-01
-1.70084763e+00 -1.02367842e+00 2.09092140e-01 1.99309957e+00
1.59317613e+00 4.26557779e-01 1.97287112e-01 -1.68007940e-01
4.63537306e-01 1.00427389e-01 -1.11704350e-01 -1.11159992e+00
-2.95486540e-01 4.48551744e-01 2.65207797e-01 7.78258801e-01
-1.03923917e+00 1.20369887e+00 6.87051868e+00 8.29513431e-01
-8.88675511e-01 8.28117430e-02 8.52146029e-01 -4.96903807e-01
-6.05049491e-01 -1.28126651e-01 -7.12982833e-01 3.24852049e-01
1.41764367e+00 -5.84798574e-01 5.55933535e-01 6.31558895e-01
4.63527501e-01 2.53420532e-01 -1.72179484e+00 6.11946881e-01
3.14262092e-01 -1.36441398e+00 3.90994012e-01 -3.72501433e-01
1.09092367e+00 8.88841599e-02 -1.64962023e-01 6.39267862e-01
5.00544906e-01 -1.32375944e+00 7.50652671e-01 3.68442595e-01
9.08638775e-01 -6.36943519e-01 6.81136250e-01 6.10510230e-01
-4.85526860e-01 3.10653597e-01 -5.77271938e-01 -1.33431107e-01
3.98229837e-01 5.18919766e-01 -9.94476318e-01 5.74955285e-01
4.06290218e-02 5.31651020e-01 -8.01425457e-01 8.59192550e-01
-4.04126734e-01 5.73460519e-01 -1.88234508e-01 -2.03117847e-01
2.08525211e-01 -4.02986407e-02 6.99444413e-01 1.57125568e+00
3.31264555e-01 -2.65670419e-01 -1.79014921e-01 9.85428214e-01
-5.08892000e-01 1.45883277e-01 -8.97156596e-01 -2.33882621e-01
3.38030517e-01 1.17544961e+00 -2.41966814e-01 -6.55982375e-01
-2.94324547e-01 1.03545403e+00 7.95862317e-01 3.23229730e-01
-8.52096915e-01 -7.08936751e-01 2.32960030e-01 -1.36964098e-01
1.65316135e-01 -1.90891460e-01 -4.78558928e-01 -1.54531121e+00
1.33608937e-01 -1.28112507e+00 -1.33457869e-01 -7.33977973e-01
-1.25585532e+00 1.02702641e+00 1.51631862e-01 -1.09514511e+00
-7.07194984e-01 -2.55817056e-01 -9.19093609e-01 9.59740043e-01
-1.29702640e+00 -9.53488946e-01 1.83367223e-01 2.26181541e-02
1.00777996e+00 -5.82774878e-02 1.16440487e+00 1.02656804e-01
-6.63380980e-01 8.32880199e-01 2.18356133e-01 -6.05918169e-02
8.53197694e-01 -1.35057402e+00 1.02123392e+00 8.75642180e-01
2.14836121e-01 8.51065636e-01 1.14044499e+00 -3.57204616e-01
-9.81247127e-01 -1.21643925e+00 1.54797113e+00 -7.58088052e-01
6.22993410e-01 -5.72663069e-01 -8.67731273e-01 8.99294019e-01
5.77758729e-01 -4.69330370e-01 5.29120266e-01 1.87116295e-01
-7.31486306e-02 1.36492863e-01 -8.19476902e-01 8.42862725e-01
1.10157967e+00 -5.13297081e-01 -1.17763734e+00 8.96238148e-01
1.15464354e+00 -4.78155762e-01 -7.85198987e-01 3.75947297e-01
1.37663886e-01 -7.24979997e-01 6.66184843e-01 -1.10065365e+00
9.72836673e-01 2.86092311e-01 -2.95899203e-03 -1.91655159e+00
-5.51784933e-02 -9.77985740e-01 1.48000389e-01 1.49896455e+00
9.08923507e-01 -2.67308444e-01 3.51512343e-01 6.72520041e-01
-4.33744669e-01 -7.83001125e-01 -8.61071408e-01 -9.30730164e-01
6.46040678e-01 -1.18905470e-01 4.31203425e-01 6.72424972e-01
1.65219948e-01 9.95103061e-01 -4.12547767e-01 -6.49227619e-01
1.52903631e-01 4.42912355e-02 1.08522666e+00 -7.30823934e-01
-3.79855454e-01 -6.09936893e-01 2.56105155e-01 -1.05114615e+00
2.80164421e-01 -9.18713748e-01 2.20298618e-01 -1.46523738e+00
3.82385522e-01 -1.19041465e-01 4.03956741e-01 3.22488308e-01
-2.62972474e-01 1.44598305e-01 7.83987567e-02 1.21536776e-01
-7.53952861e-01 9.02844608e-01 1.32729352e+00 -3.45147908e-01
-1.36993170e-01 1.79588236e-02 -8.97516251e-01 4.76338685e-01
7.78978586e-01 -4.04326618e-01 -3.99394780e-01 -9.37661946e-01
1.69639587e-01 1.63264155e-01 -1.76825494e-01 -7.63607621e-01
-1.55279785e-01 -8.84965048e-05 1.01900827e-02 -1.88886717e-01
2.90951252e-01 -1.42017573e-01 -1.77919596e-01 2.35109329e-01
-1.05004120e+00 3.51993710e-01 2.65586317e-01 8.15852731e-02
-3.21971387e-01 -5.43369949e-01 7.09195912e-01 -4.11062777e-01
8.75524133e-02 3.80738378e-02 -3.11931610e-01 4.69631106e-01
6.42018437e-01 2.51712471e-01 -4.72913980e-01 -6.84703469e-01
-6.03466630e-01 1.83485419e-01 6.17006004e-01 2.27643400e-01
3.31470817e-01 -1.31788850e+00 -1.25391936e+00 -2.54874945e-01
3.45713273e-02 1.40292794e-01 -2.61775672e-01 6.98928952e-01
-6.70758545e-01 7.13505864e-01 -1.00587562e-01 -1.75034091e-01
-1.06031013e+00 5.96387863e-01 2.50068605e-01 -7.47349441e-01
-3.87915999e-01 7.73079216e-01 1.12419426e-02 -5.36112547e-01
1.52229548e-01 -6.91568077e-01 -9.65215713e-02 -1.78114936e-01
5.46172500e-01 2.20635414e-01 1.65869579e-01 -3.59016240e-01
1.87312178e-02 2.33885795e-01 -2.84723192e-01 -4.06583220e-01
1.38009298e+00 -1.96434379e-01 1.30098639e-02 4.33785617e-01
1.13410699e+00 1.29665762e-01 -7.20799387e-01 -1.65204063e-01
1.13686286e-01 -1.87052473e-01 -5.71006894e-01 -7.20831215e-01
-3.49548280e-01 9.11604226e-01 -2.61661291e-01 3.41711700e-01
8.92465591e-01 -9.81686860e-02 9.37725663e-01 8.93685699e-01
-9.76107549e-03 -1.16158962e+00 1.50903985e-01 8.50845695e-01
1.44180310e+00 -1.33434463e+00 1.34862531e-02 -1.63992435e-01
-7.64733195e-01 1.15607238e+00 4.37949359e-01 -1.86080247e-01
-2.32262298e-01 1.08168602e-01 -9.20586139e-02 1.73556700e-01
-1.30977988e+00 1.44964457e-01 1.90800756e-01 4.94424909e-01
7.15395331e-01 8.91687535e-03 -6.68093681e-01 6.56235576e-01
-8.82608771e-01 -2.53522005e-02 8.25742841e-01 5.40176094e-01
-2.77913153e-01 -1.25885081e+00 -4.96439002e-02 4.44971085e-01
-5.55693865e-01 -5.36870360e-01 -7.26890564e-01 7.19606638e-01
-3.68166089e-01 8.94463897e-01 -1.25726208e-01 -3.41031492e-01
4.85298812e-01 6.83801100e-02 6.91923559e-01 -1.01099992e+00
-6.47021294e-01 -3.08912843e-01 9.93223011e-01 -1.80529207e-01
-1.61718607e-01 -5.12988210e-01 -6.83876634e-01 -3.87867838e-01
-2.61787981e-01 3.92693341e-01 4.89616543e-01 1.00866258e+00
2.24125713e-01 4.28749412e-01 5.12790501e-01 -1.07204151e+00
-1.09247935e+00 -1.67758489e+00 2.52949540e-02 6.48225605e-01
2.17418864e-01 -7.28411973e-02 -2.44508505e-01 2.80297607e-01]
|
[11.73682975769043, 9.238557815551758]
|
581c48bb-138c-4d2d-aff4-731b35faa5e3
|
aug-ila-more-transferable-intermediate-level
| null | null |
https://openreview.net/forum?id=zKbMQ2NY1y
|
https://openreview.net/pdf?id=zKbMQ2NY1y
|
Aug-ILA: More Transferable Intermediate Level Attacks with Augmented References
|
An intriguing property of deep neural networks is that adversarial attacks can transfer across different models. Existing methods such as the Intermediate Level Attack (ILA) further improve black-box transferability by fine-tuning a reference adversarial attack, so as to maximize the perturbation on a pre-specified layer of the source model. In this paper, we revisit ILA and evaluate the effect of applying augmentation to the images before passing them to ILA. We start by looking into the effect of common image augmentation techniques and exploring novel augmentation with the aid of adversarial perturbations. Based on the observations, we propose Aug-ILA, an improved method that enhances the transferability of an existing attack under the ILA framework. Specifically, Aug-ILA has three main characteristics: typical image augmentation such as random cropping and resizing applied to all ILA inputs, reverse adversarial update on the clean image, and interpolation between two attacks on the reference image. Our experimental results show that Aug-ILA outperforms ILA and its subsequent variants, as well as state-of-the-art transfer-based attacks, by achieving $96.99\%$ and $87.84\%$ average attack success rates with perturbation budgets $0.05$ and $0.03$, respectively, on nine undefended models.
|
['Dit-yan Yeung', 'Chiu Wai Yan']
|
2021-09-29
| null | null | null | null |
['image-augmentation']
|
['computer-vision']
|
[ 4.81473982e-01 1.52491510e-01 2.06303462e-01 -1.48160994e-01
-8.01378965e-01 -8.48423243e-01 6.77413225e-01 -3.25279355e-01
-5.96077979e-01 5.95593929e-01 -1.77105352e-01 -5.23431063e-01
1.13775894e-01 -8.60004365e-01 -1.25774479e+00 -7.47906327e-01
-3.31034571e-01 -1.06836557e-01 1.47617444e-01 -5.75481474e-01
-1.22017734e-01 6.83122933e-01 -8.72419834e-01 3.06273431e-01
6.89075112e-01 1.02397203e+00 -4.20495927e-01 6.31915689e-01
1.80702090e-01 6.05336607e-01 -9.95383620e-01 -9.21391308e-01
7.36201048e-01 -4.49499846e-01 -7.55371332e-01 -2.78870583e-01
6.35103464e-01 -5.53350568e-01 -7.34724343e-01 1.36740756e+00
6.04800582e-01 8.95402133e-02 4.43925291e-01 -1.58504152e+00
-7.35227108e-01 8.43234777e-01 -7.68687546e-01 2.28128031e-01
-1.01920158e-01 4.87673610e-01 5.82840681e-01 -6.52086079e-01
2.40572378e-01 1.34275854e+00 7.36507475e-01 1.10518801e+00
-1.25678575e+00 -1.29750502e+00 4.08869654e-01 9.83388871e-02
-1.13978744e+00 -3.61744732e-01 7.34084427e-01 -2.15121165e-01
7.13552892e-01 3.22468847e-01 5.18847853e-02 1.41133106e+00
1.15891516e-01 5.24257898e-01 1.27054071e+00 -3.10140789e-01
1.93113029e-01 -2.86938827e-02 -1.40398860e-01 5.27617872e-01
1.98787972e-01 4.07042205e-01 -1.61581144e-01 -2.50858277e-01
6.31971061e-01 -1.59774438e-01 -4.01161015e-01 5.34641780e-02
-9.00173247e-01 6.77076161e-01 9.35872972e-01 -1.11993596e-01
-2.83505529e-01 3.65971923e-01 5.76826811e-01 4.64049846e-01
2.38347039e-01 6.48647308e-01 -4.62990135e-01 2.87669331e-01
-5.52038789e-01 2.27371991e-01 5.47062278e-01 8.41522932e-01
8.46892238e-01 4.85102028e-01 -2.72729844e-01 4.45279777e-01
2.57523619e-02 5.36460638e-01 3.62371743e-01 -8.36339056e-01
7.14480042e-01 1.09767511e-01 -2.30681270e-01 -9.07624006e-01
-3.79901119e-02 -4.92253155e-01 -1.15571284e+00 7.61085391e-01
3.56418967e-01 -5.87368906e-01 -1.33367908e+00 2.27014780e+00
1.47955358e-01 5.69808006e-01 3.12169492e-01 4.50166553e-01
5.63853025e-01 5.58333397e-01 1.73960209e-01 9.78398472e-02
1.02355826e+00 -9.77956831e-01 -5.69783807e-01 -3.86546999e-01
4.87921864e-01 -7.07269192e-01 1.05648947e+00 2.74062604e-01
-1.14993763e+00 -6.97130263e-01 -1.18769372e+00 4.27473336e-01
-5.18702567e-01 -2.54483104e-01 2.75285602e-01 9.55942631e-01
-9.38032210e-01 7.61666238e-01 -8.13778460e-01 2.64384359e-01
6.96875751e-01 5.05543709e-01 -5.79186559e-01 3.32198502e-03
-1.59198201e+00 8.19829166e-01 3.02175432e-01 1.92199603e-01
-1.30314469e+00 -1.21739662e+00 -9.44023848e-01 -1.06968742e-03
2.17721567e-01 -5.67405045e-01 9.89024341e-01 -1.10330415e+00
-1.53542769e+00 6.28782153e-01 2.98991710e-01 -9.17042613e-01
6.71987116e-01 -4.14879620e-01 -3.81410003e-01 -2.96102371e-02
-2.60182947e-01 7.12749541e-01 1.11149883e+00 -1.50713396e+00
-2.08616048e-01 -2.49881700e-01 4.15254325e-01 -8.19529146e-02
-5.59748292e-01 7.42196366e-02 -1.97224811e-01 -9.89607036e-01
-3.42519283e-01 -1.13892817e+00 -3.86308640e-01 7.69657865e-02
-7.49748111e-01 5.01876533e-01 1.06144035e+00 -6.14483297e-01
9.08339024e-01 -2.38693953e+00 6.57484904e-02 3.35743040e-01
2.28418946e-01 8.45029235e-01 -5.89098811e-01 3.31012785e-01
-5.74132144e-01 5.74931026e-01 -5.05957127e-01 -3.55669796e-01
-1.11630571e-03 1.84298635e-01 -8.24372590e-01 4.36980456e-01
3.85058582e-01 8.56660545e-01 -6.10731184e-01 -3.59112620e-02
2.29834262e-02 6.00117981e-01 -6.75567746e-01 3.86117518e-01
7.87866190e-02 3.84007871e-01 -6.73510730e-02 3.43938768e-01
1.04079175e+00 2.44273499e-01 -2.72447973e-01 -2.86236852e-01
2.65072197e-01 -5.78711405e-02 -1.01264739e+00 1.30243349e+00
-3.67032647e-01 5.85000098e-01 2.28455022e-01 -8.25765789e-01
8.03385198e-01 2.74991602e-01 5.42456843e-02 -3.38128716e-01
2.96359539e-01 3.59412841e-02 3.88743669e-01 -5.95121644e-02
3.24908532e-02 -6.34751022e-02 -2.06284046e-01 4.20927018e-01
1.55528978e-01 2.14158483e-02 -1.41538933e-01 3.84353310e-01
1.22685862e+00 -2.17012420e-01 7.27378130e-02 -1.06416725e-01
6.48922503e-01 -4.97040242e-01 4.44674432e-01 9.99763727e-01
-3.97159874e-01 5.17133355e-01 5.32502115e-01 -2.35130250e-01
-9.50965703e-01 -1.16233611e+00 1.50662601e-01 9.10685956e-01
7.06839487e-02 -1.40046239e-01 -1.20497680e+00 -1.04203475e+00
-5.53471558e-02 5.50994217e-01 -1.06152654e+00 -6.56094849e-01
-7.33927906e-01 -8.36756587e-01 1.09251392e+00 6.61773741e-01
1.13265324e+00 -1.05278349e+00 -1.07297778e-01 -2.82846779e-01
6.92016408e-02 -1.14929426e+00 -5.40605068e-01 1.74201921e-01
-5.07667959e-01 -8.47980559e-01 -6.45702839e-01 -6.71083510e-01
8.69680464e-01 1.03225954e-01 7.22958446e-01 8.73547643e-02
-7.34392852e-02 2.12563667e-02 -2.19895080e-01 -5.73508203e-01
-7.14173377e-01 9.56905410e-02 2.26559773e-01 1.64927095e-01
-1.70657098e-01 -8.71572852e-01 -5.12020230e-01 3.17321658e-01
-1.22100294e+00 -3.52862507e-01 7.18826830e-01 9.42525744e-01
3.57670188e-01 8.84042010e-02 5.58843434e-01 -1.07648540e+00
5.92687488e-01 -4.16558534e-01 -4.61205572e-01 2.57449728e-02
-2.51046985e-01 -1.91419851e-02 9.79657173e-01 -8.40999484e-01
-8.49277198e-01 -1.56694412e-01 -4.73324209e-01 -9.90678847e-01
-1.96828216e-01 1.61150768e-01 -5.45975029e-01 -6.11341655e-01
1.03349316e+00 1.89831331e-01 -1.16455577e-01 -2.11382538e-01
5.16379654e-01 1.65461570e-01 7.59991288e-01 -5.51170826e-01
1.74975383e+00 3.91166717e-01 3.95985357e-02 -5.10444880e-01
-6.44786239e-01 2.42527023e-01 -4.71461236e-01 -3.98802422e-02
5.78715444e-01 -7.66503215e-01 -5.97314477e-01 1.01655269e+00
-1.00536752e+00 -6.39839411e-01 -3.96458000e-01 1.56331018e-01
-3.54073882e-01 3.97659034e-01 -6.50270343e-01 -3.74694765e-01
-4.97223586e-01 -1.26942337e+00 4.45776910e-01 1.48548841e-01
1.17483899e-01 -7.79925942e-01 -1.67892888e-01 1.14241585e-01
6.86619222e-01 7.03125477e-01 8.35476220e-01 -9.43790019e-01
-3.64936352e-01 -3.35624933e-01 -1.92751423e-01 8.60154927e-01
2.42731780e-01 -1.51832417e-01 -1.18405604e+00 -6.86404943e-01
1.07456595e-01 -5.26308656e-01 8.43313515e-01 7.54049420e-02
1.39397156e+00 -7.20758438e-01 -4.07690704e-02 1.12089241e+00
1.32700455e+00 3.36665183e-01 9.22373533e-01 6.17472053e-01
8.02761614e-01 2.58289337e-01 2.00056091e-01 8.77942890e-02
-3.13153535e-01 3.34796280e-01 1.03742003e+00 -3.80152911e-01
-5.77866472e-02 -1.92314178e-01 5.33139586e-01 3.09412092e-01
1.37389541e-01 -1.68236226e-01 -6.97808623e-01 3.60201716e-01
-1.31114686e+00 -8.96498322e-01 4.26086575e-01 2.05555820e+00
1.00766683e+00 5.73004603e-01 -6.22389130e-02 3.35452110e-01
7.77105451e-01 4.25793409e-01 -6.98504627e-01 -5.69160342e-01
-3.21549699e-02 5.78431189e-01 7.20608890e-01 5.40376306e-01
-1.34482050e+00 1.01540172e+00 6.41211176e+00 8.52636933e-01
-1.19673228e+00 3.78501564e-02 7.72682011e-01 1.44057430e-03
-6.47339597e-02 -3.50513816e-01 -6.65298879e-01 3.16869676e-01
8.91269445e-01 -3.23472857e-01 6.08637989e-01 8.22717905e-01
-2.27103710e-01 6.04804575e-01 -9.60603952e-01 4.20803577e-01
2.85638161e-02 -1.25114167e+00 4.81340915e-01 1.28167160e-02
9.03641164e-01 -9.21164379e-02 5.83989143e-01 4.87965703e-01
5.81251502e-01 -1.08135712e+00 4.83897418e-01 -1.73850842e-02
9.11345661e-01 -1.04997098e+00 7.24775374e-01 1.40167698e-01
-9.03005660e-01 -5.29804565e-02 -1.35267019e-01 9.76005495e-02
-1.50006175e-01 5.93124256e-02 -5.42291582e-01 5.12526095e-01
6.89811707e-01 1.49185494e-01 -5.07955194e-01 5.64023376e-01
-5.70670247e-01 8.18818510e-01 -1.67413443e-01 4.84547406e-01
4.80679005e-01 2.88641870e-01 6.63947582e-01 1.22055387e+00
6.81110919e-02 1.90153718e-02 -3.17288451e-02 8.63044202e-01
-6.92311645e-01 -2.42667511e-01 -5.45009077e-01 3.01526815e-01
5.63140512e-01 1.10535204e+00 -2.01961398e-01 -1.62648126e-01
-2.31574759e-01 1.06430721e+00 1.83028102e-01 4.68066305e-01
-1.34385431e+00 -8.17821562e-01 1.15745211e+00 -5.26121072e-02
1.56227872e-01 1.38234586e-01 -1.85099021e-01 -7.70717084e-01
4.55080159e-02 -1.28144145e+00 2.56035119e-01 -4.78336632e-01
-1.27265429e+00 1.04170573e+00 4.43287231e-02 -1.17087233e+00
-7.72686377e-02 -5.36904931e-01 -1.03424120e+00 1.17013669e+00
-1.54248834e+00 -1.24679875e+00 -2.60877937e-01 9.76161838e-01
2.50476122e-01 -3.29070717e-01 8.40557516e-01 2.34022304e-01
-8.89818788e-01 1.56752336e+00 -1.74863160e-01 6.13746762e-01
7.02142239e-01 -1.08158910e+00 1.06621945e+00 1.45347679e+00
-2.27114726e-02 7.48648643e-01 6.86162412e-01 -3.19441468e-01
-1.08612621e+00 -1.61207283e+00 4.34776098e-02 -1.16165735e-01
8.50994468e-01 -3.79863173e-01 -1.21295536e+00 9.80379581e-01
3.93848658e-01 4.12752181e-01 4.18461621e-01 -3.14573288e-01
-8.61987472e-01 -2.31853426e-01 -1.48112190e+00 9.67059493e-01
8.60541523e-01 -5.93343318e-01 -3.39790553e-01 1.54629387e-02
1.02821064e+00 -6.14009082e-01 -7.41239667e-01 6.06347859e-01
1.76155567e-01 -7.19536483e-01 1.32920635e+00 -8.33988547e-01
4.48392540e-01 -1.19821414e-01 -1.18777357e-01 -1.54365540e+00
-2.74979323e-01 -1.09510708e+00 -3.02054062e-02 1.14768732e+00
4.17574227e-01 -1.00094962e+00 7.18310177e-01 4.88789678e-01
-2.26757646e-01 -7.15244114e-01 -9.15411472e-01 -9.76305544e-01
6.02250695e-01 -3.07478815e-01 6.78629518e-01 1.03091717e+00
-4.29609299e-01 -2.14844123e-01 -4.82362181e-01 8.13894451e-01
8.09846878e-01 -5.34833908e-01 1.02321720e+00 -5.28391838e-01
-4.16272908e-01 -3.43374878e-01 -4.34869856e-01 -5.96559525e-01
3.14862698e-01 -6.86877489e-01 -6.58923239e-02 -7.46500850e-01
-1.58806637e-01 -2.83513725e-01 -6.02636218e-01 8.88716221e-01
-4.98694777e-01 6.50842428e-01 5.28695464e-01 -2.97576934e-02
1.48286685e-01 3.74732912e-01 1.16846204e+00 -3.09711426e-01
1.32753439e-02 4.07698303e-02 -8.15190613e-01 8.37191582e-01
1.10906303e+00 -5.22392869e-01 -4.15437192e-01 -5.38120747e-01
-1.73289731e-01 -3.89003515e-01 4.96707141e-01 -1.17158449e+00
-1.61480214e-02 -8.67692903e-02 3.03297430e-01 -4.65401001e-02
4.25064445e-01 -6.86457992e-01 -4.41848077e-02 7.81210184e-01
-5.59596539e-01 4.41847630e-02 8.34847629e-01 4.41690773e-01
-1.57857716e-01 -1.24648318e-01 1.33203053e+00 6.41476586e-02
-4.82823789e-01 4.75492030e-01 -9.58055630e-02 2.26331875e-01
1.16234767e+00 9.33974832e-02 -6.48568153e-01 -2.61614799e-01
-7.22812772e-01 -6.27963394e-02 1.82647616e-01 3.63616168e-01
5.72334588e-01 -1.24813581e+00 -8.31172824e-01 5.23273647e-01
-2.54276395e-01 -1.49653941e-01 3.25178266e-01 3.42067748e-01
-3.11756045e-01 -6.97267503e-02 -5.17982602e-01 -1.36832058e-01
-1.36497343e+00 7.85428882e-01 5.92971265e-01 -4.23629850e-01
-4.57071275e-01 1.17742753e+00 6.00776672e-01 -3.25779408e-01
3.91697794e-01 -7.59251937e-02 1.04820020e-01 -3.85199457e-01
6.77762747e-01 2.41154492e-01 3.67515758e-02 -5.09515285e-01
-2.62173355e-01 3.48508477e-01 -3.85174811e-01 -1.05879873e-01
9.42202032e-01 2.53484905e-01 1.74976885e-02 -2.63968825e-01
1.30859876e+00 4.02779691e-02 -1.43343568e+00 -3.94140661e-01
-6.25452936e-01 -4.26547617e-01 -1.25294432e-01 -8.89836371e-01
-1.45918751e+00 9.60240006e-01 5.17186821e-01 1.60377085e-01
1.56078410e+00 -3.88794363e-01 8.03010881e-01 3.17955881e-01
1.20420352e-01 -3.41398895e-01 2.55169839e-01 4.75693196e-01
1.18378878e+00 -9.48315024e-01 -9.84046087e-02 -3.37146640e-01
-4.31175709e-01 6.02403164e-01 9.43475425e-01 -4.91694033e-01
6.53718829e-01 5.27570546e-01 2.69406378e-01 3.18578064e-01
-6.19366169e-01 1.90512776e-01 1.15142092e-01 7.47828305e-01
-2.29076460e-01 -2.65244156e-01 2.27476746e-01 4.28700238e-01
-3.02087724e-01 -4.93680388e-01 5.25549650e-01 8.74417603e-01
1.24195665e-02 -1.06165957e+00 -5.69822967e-01 1.01621270e-01
-9.00465369e-01 -2.12522253e-01 -4.26514685e-01 1.04984260e+00
1.04275823e-01 6.62319064e-01 -1.83256656e-01 -8.30323339e-01
5.08921146e-01 -7.08396882e-02 2.62997091e-01 -2.80249268e-01
-1.02767777e+00 -4.11340922e-01 -1.97758585e-01 -6.32575572e-01
-7.39313364e-02 -3.87266427e-01 -9.10027564e-01 -4.83877301e-01
-2.81693608e-01 -3.08204032e-02 5.08504391e-01 7.71943390e-01
2.61682063e-01 7.88858771e-01 1.02257919e+00 -1.16444278e+00
-8.84479761e-01 -1.02107334e+00 -1.44445807e-01 4.50423628e-01
5.76976061e-01 -3.40324193e-01 -8.44146013e-01 7.54565597e-02]
|
[5.578927516937256, 7.910180568695068]
|
b88898f4-e8be-4a2f-a473-e06a009a5a5e
|
approaching-neural-chinese-word-segmentation
|
2008.05348
| null |
https://arxiv.org/abs/2008.05348v3
|
https://arxiv.org/pdf/2008.05348v3.pdf
|
Approaching Neural Chinese Word Segmentation as a Low-Resource Machine Translation Task
|
Chinese word segmentation has entered the deep learning era which greatly reduces the hassle of feature engineering. Recently, some researchers attempted to treat it as character-level translation, which further simplified model designing, but there is a performance gap between the translation-based approach and other methods. This motivates our work, in which we apply the best practices from low-resource neural machine translation to supervised Chinese segmentation. We examine a series of techniques including regularization, data augmentation, objective weighting, transfer learning, and ensembling. Compared to previous works, our low-resource translation-based method maintains the effortless model design, yet achieves the same result as state of the art in the constrained evaluation without using additional data.
|
['Pin-zhen Chen', 'Kenneth Heafield']
|
2020-08-12
| null | null | null | null |
['low-resource-neural-machine-translation']
|
['natural-language-processing']
|
[ 2.36428723e-01 -1.43483758e-01 -3.76556724e-01 -4.96621400e-01
-1.13347602e+00 -4.26362395e-01 2.60359585e-01 -2.61793762e-01
-8.05370450e-01 6.95357740e-01 2.06811994e-01 -7.02897906e-01
4.89871591e-01 -5.47750711e-01 -5.46054006e-01 -3.03017467e-01
6.60002172e-01 5.93225658e-01 1.74071699e-01 -1.38680547e-01
2.51982480e-01 7.97457993e-02 -7.05264091e-01 2.08660662e-01
1.38085687e+00 7.34555304e-01 1.41169995e-01 3.39726388e-01
-5.17744541e-01 3.94019932e-01 -4.08561528e-01 -7.45013714e-01
2.43112609e-01 -5.32420814e-01 -9.70780909e-01 -1.24060795e-01
1.41511992e-01 -2.30095714e-01 -2.51323562e-02 1.08373785e+00
7.33184397e-01 8.11579973e-02 3.56685817e-01 -6.28363073e-01
-1.25484908e+00 9.06277299e-01 -5.22187769e-01 -2.08382726e-01
-1.55162871e-01 -1.89048633e-01 9.96616364e-01 -1.02100015e+00
4.24799562e-01 1.10677695e+00 9.75300908e-01 9.65892017e-01
-8.67795706e-01 -5.07613063e-01 4.46128607e-01 1.21095562e-02
-1.20640564e+00 -2.12445214e-01 5.53367257e-01 -1.44787595e-01
1.28803205e+00 1.62589207e-01 4.77211595e-01 9.62913752e-01
-9.92582068e-02 1.27884734e+00 1.12690079e+00 -8.55587065e-01
6.25856146e-02 1.12754583e-01 3.40468764e-01 6.29224360e-01
9.89334956e-02 -4.27160077e-02 -2.16309011e-01 8.58988613e-02
6.38144195e-01 -2.49261409e-01 1.25167146e-02 2.33942926e-01
-9.72140670e-01 9.54045892e-01 1.55567572e-01 3.60661179e-01
-4.27032635e-02 1.36607960e-01 2.80741870e-01 1.80005521e-01
9.16615963e-01 6.39817536e-01 -1.19981480e+00 -4.08053964e-01
-1.06412745e+00 -1.36191785e-01 8.54901254e-01 1.26009786e+00
8.38564694e-01 3.13662499e-01 -2.01417729e-01 9.24398661e-01
4.16153044e-01 3.17645580e-01 8.46323431e-01 -6.26459837e-01
6.12987518e-01 6.07401073e-01 -1.82639793e-01 -3.45638663e-01
-3.17899317e-01 -5.32135010e-01 -5.56987762e-01 -3.57269496e-01
2.00543612e-01 -7.83377349e-01 -1.34660816e+00 1.68403661e+00
-4.50454131e-02 -1.64098423e-02 -1.60396263e-01 8.57853591e-01
5.31925738e-01 5.57638228e-01 6.75208494e-02 -3.89061905e-02
9.26004291e-01 -1.65642142e+00 -8.93109202e-01 -4.67695743e-01
8.75129282e-01 -1.01948428e+00 1.45762861e+00 4.01453763e-01
-9.43020284e-01 -5.52623808e-01 -9.23475564e-01 -2.41848022e-01
-5.98578513e-01 3.63723636e-01 1.15053868e+00 1.21939790e+00
-9.53451633e-01 7.62536824e-01 -1.09318256e+00 -3.39418203e-01
3.28325063e-01 5.75489402e-01 5.94106503e-03 -2.46975254e-02
-1.36340749e+00 1.22540629e+00 3.57442856e-01 2.39793763e-01
-4.85738575e-01 -4.98232335e-01 -7.56868720e-01 -1.29894093e-01
3.17627817e-01 -4.72642422e-01 1.35461843e+00 -1.16983354e+00
-1.86243188e+00 5.54862916e-01 -1.85105175e-01 -2.60654479e-01
5.28035223e-01 -8.33702743e-01 -3.00456375e-01 -4.87077206e-01
-1.60487756e-01 8.29713404e-01 4.67279941e-01 -8.23667645e-01
-5.05505145e-01 6.56062551e-03 -1.23007752e-01 1.98808223e-01
-6.53728604e-01 4.10309464e-01 -1.08018076e+00 -6.62923634e-01
5.28845079e-02 -9.29856956e-01 -6.47001326e-01 -5.39361000e-01
-3.98722470e-01 -1.93973139e-01 5.80694854e-01 -6.91594839e-01
1.56363511e+00 -1.76258481e+00 1.49787232e-01 -2.73384720e-01
-1.01050101e-01 7.01730490e-01 -3.91698569e-01 3.07926089e-01
1.64085835e-01 6.67321980e-01 -5.14381230e-01 -7.44392037e-01
3.05413939e-02 2.42634967e-01 4.83935922e-02 2.78025568e-01
5.01141012e-01 1.31721509e+00 -6.67407274e-01 -6.93228960e-01
-1.77530646e-01 3.32550257e-01 -7.77034402e-01 1.55061960e-01
-4.03206766e-01 2.14029372e-01 -7.13375986e-01 9.06549633e-01
8.02975833e-01 6.42858669e-02 1.69371009e-01 4.65396196e-02
-3.86305422e-01 5.34973979e-01 -8.29643548e-01 1.96637046e+00
-3.48329574e-01 5.69804013e-01 -1.99237615e-01 -8.81854713e-01
9.20161843e-01 1.29418299e-01 3.42596620e-01 -7.10518539e-01
3.00917059e-01 4.28466469e-01 1.21636398e-01 -4.79548842e-01
6.16794765e-01 -4.02338132e-02 -1.67591721e-01 4.61933851e-01
7.86013752e-02 -1.32984400e-01 9.58392490e-03 -1.59916759e-01
6.97354555e-01 9.00159478e-01 -1.23279013e-01 -2.89288908e-01
1.80786192e-01 4.19636637e-01 9.78625417e-01 5.67623734e-01
-1.73102677e-01 9.08464015e-01 2.17892870e-01 -2.66095102e-01
-1.01444817e+00 -5.79208672e-01 -2.62130257e-02 1.29423070e+00
2.38658190e-02 -4.32296187e-01 -1.33516538e+00 -8.32353294e-01
-3.72915417e-01 5.99431038e-01 -5.17023921e-01 -2.06875503e-01
-1.02855992e+00 -1.31395209e+00 9.18775856e-01 8.28312635e-01
6.24964595e-01 -9.08683956e-01 -1.83851898e-01 2.98140585e-01
-2.38812506e-01 -9.87443924e-01 -7.92402685e-01 3.81571144e-01
-1.17157388e+00 -4.71328110e-01 -7.66626239e-01 -1.18732953e+00
6.59202099e-01 -8.58574361e-02 1.11577094e+00 1.72725648e-01
1.11519173e-01 -9.28376764e-02 -6.93650305e-01 -4.68369961e-01
-1.34132117e-01 7.31948853e-01 -1.36505812e-03 -2.06329480e-01
8.28441620e-01 -1.11036099e-01 -2.71269321e-01 1.13344230e-01
-7.82557964e-01 1.13613822e-01 9.80232298e-01 1.03966582e+00
6.38911545e-01 -4.51129884e-01 6.16743803e-01 -1.21192312e+00
9.93674755e-01 -1.15610644e-01 -4.84675169e-01 6.06981397e-01
-1.00453925e+00 5.45429587e-02 6.40558362e-01 -7.32933939e-01
-1.19038439e+00 2.24088043e-01 -3.06215793e-01 -6.37222081e-02
7.75936618e-02 9.75969434e-01 -3.89566690e-01 -9.04534012e-02
4.42172945e-01 1.04854144e-01 -4.27837253e-01 -1.01531422e+00
4.35505897e-01 9.02826250e-01 1.59317965e-03 -7.81684160e-01
6.06231749e-01 -1.20535977e-01 -5.73711812e-01 -5.23223400e-01
-9.16342795e-01 -1.37910023e-01 -1.14038908e+00 2.48506293e-01
9.19676006e-01 -7.51155376e-01 -4.22320701e-02 6.30358219e-01
-1.23781049e+00 -5.72722554e-01 -9.14352983e-02 6.81151807e-01
-2.24088252e-01 5.88892817e-01 -1.06056070e+00 -6.43241644e-01
-6.31505668e-01 -1.24849391e+00 6.78746581e-01 7.08801672e-02
-2.06240620e-02 -9.55390871e-01 1.00938596e-01 4.38619465e-01
7.93509185e-01 -2.02766865e-01 9.28541362e-01 -7.47823954e-01
-4.03201014e-01 -3.04352790e-01 -2.04933658e-01 6.43720388e-01
9.26359296e-02 3.03570479e-02 -8.70905280e-01 -7.00063705e-02
4.13238928e-02 -3.78888547e-01 8.28335404e-01 3.44443023e-01
1.00435865e+00 1.15472339e-01 -7.57866874e-02 8.21017802e-01
1.29204822e+00 2.84810185e-01 6.47094011e-01 3.37783277e-01
9.76449370e-01 3.25578213e-01 7.28911579e-01 -2.03943133e-01
3.40572447e-01 4.79614258e-01 5.66535891e-05 -5.73436201e-01
-9.00925472e-02 -2.00795501e-01 4.04774487e-01 1.56189990e+00
-1.21887848e-01 -2.45490491e-01 -1.08573520e+00 4.56947953e-01
-2.06166339e+00 -8.43946710e-02 -7.52243251e-02 1.88174546e+00
1.11923051e+00 2.89695412e-01 -2.88541496e-01 -4.61372316e-01
6.31092906e-01 -2.19483748e-01 -6.33042336e-01 -7.33026147e-01
-1.08588062e-01 4.60136741e-01 6.37068927e-01 5.89452863e-01
-1.27983248e+00 1.87590957e+00 7.26592541e+00 9.79205489e-01
-1.24043477e+00 4.60587502e-01 7.47345746e-01 2.22107604e-01
-3.08339208e-01 2.26151079e-01 -9.64417160e-01 2.20649227e-01
1.05621099e+00 6.03952482e-02 4.65921372e-01 8.16428244e-01
2.10942388e-01 1.83659494e-01 -9.94305372e-01 4.19162393e-01
1.75604925e-01 -1.14751816e+00 6.79987520e-02 -1.05628632e-01
9.39620852e-01 5.58925927e-01 7.12990463e-02 6.69497073e-01
3.23765486e-01 -1.03771520e+00 4.11180079e-01 2.37868816e-01
9.52350020e-01 -6.77398384e-01 8.91374171e-01 3.44515175e-01
-9.28276300e-01 2.33049780e-01 -4.57956374e-01 -2.49361217e-01
1.98671699e-01 2.45669514e-01 -3.73596609e-01 4.99191701e-01
3.10747623e-01 5.20088017e-01 -5.81341803e-01 8.59890819e-01
-2.60351956e-01 1.20614791e+00 -1.39698431e-01 -2.96274960e-01
6.55666351e-01 -6.28915548e-01 -6.01310469e-02 1.56538725e+00
1.20736212e-01 -6.02836534e-02 3.78016651e-01 7.56740332e-01
-2.61051297e-01 3.31610620e-01 -5.36577702e-02 -1.98391497e-01
2.04706833e-01 1.10481524e+00 -7.21368670e-01 -3.54990870e-01
-6.83771610e-01 1.06198728e+00 4.31340635e-01 3.41962576e-01
-9.78443682e-01 -4.22035754e-01 3.89611691e-01 -2.70892680e-01
2.03071639e-01 -4.60454673e-01 -7.64246404e-01 -1.50620532e+00
-9.28263143e-02 -9.82805789e-01 -8.62309486e-02 -2.83934057e-01
-1.11084735e+00 9.34837282e-01 -2.29333326e-01 -1.24778140e+00
-7.06928372e-02 -7.84818411e-01 -5.51059842e-01 1.08612621e+00
-1.64952302e+00 -1.48463929e+00 2.81518519e-01 2.08543018e-01
9.64016497e-01 -1.79806575e-01 1.04232049e+00 7.32143402e-01
-9.85391915e-01 9.74110007e-01 1.87195033e-01 4.36614841e-01
7.86264300e-01 -1.27782667e+00 9.41241264e-01 1.11826181e+00
1.02972135e-01 8.19341362e-01 1.94097787e-01 -7.77951539e-01
-1.36411083e+00 -1.11113560e+00 1.15572262e+00 -3.51420283e-01
5.53413928e-01 -5.42148590e-01 -9.17655051e-01 6.64992571e-01
4.08137947e-01 -3.29799473e-01 8.55359435e-01 2.58442760e-01
-6.84047416e-02 1.63094550e-01 -9.13352609e-01 9.01410520e-01
8.83181512e-01 -2.66263485e-01 -6.34690404e-01 2.62192041e-01
1.15083265e+00 -3.31954062e-01 -6.53445542e-01 4.16132241e-01
5.19879878e-01 -2.03077808e-01 4.87825722e-01 -1.15767813e+00
4.33578193e-01 2.48047267e-03 -1.47505999e-01 -1.35977113e+00
-4.47277337e-01 -8.26910734e-01 8.55467543e-02 1.37780344e+00
1.22369373e+00 -5.78588188e-01 9.09232616e-01 9.59099293e-01
-5.52347720e-01 -1.07391894e+00 -6.91067636e-01 -6.21245325e-01
9.16112542e-01 -4.62870449e-01 5.85977495e-01 1.13934994e+00
-5.22623584e-02 4.93041784e-01 -5.68576396e-01 -2.95719922e-01
9.30190310e-02 1.15765393e-01 3.49392861e-01 -9.02986348e-01
-3.02401125e-01 -5.30474663e-01 3.09791714e-01 -1.43318331e+00
1.19402535e-01 -7.25886226e-01 2.37552837e-01 -1.58166635e+00
1.43650949e-01 -4.71294910e-01 -4.48900551e-01 6.35202467e-01
-4.34039623e-01 3.21556777e-01 1.85673401e-01 1.25469398e-02
-5.44592798e-01 5.61952949e-01 1.41640854e+00 -5.80062456e-02
-3.34643781e-01 6.26660585e-02 -9.41220820e-01 8.43094110e-01
1.10706306e+00 -6.15339935e-01 -5.76892383e-02 -1.32429981e+00
2.34881401e-01 -4.01786953e-01 -6.57888055e-01 -6.79517210e-01
2.19378978e-01 -1.52827159e-01 3.13362122e-01 -5.86325705e-01
2.40096778e-01 -6.29902959e-01 -4.65246975e-01 4.61938113e-01
-4.02538419e-01 2.64063805e-01 2.44241968e-01 1.21100202e-01
-2.91432321e-01 -4.79135096e-01 5.46047211e-01 -4.30028766e-01
-6.22645438e-01 3.13301444e-01 -5.16124904e-01 1.30676076e-01
4.29502964e-01 -1.14143811e-01 -7.83705935e-02 -4.79190722e-02
-4.49516445e-01 2.09089532e-01 2.41710424e-01 4.36988086e-01
4.72509116e-01 -1.06608164e+00 -8.63558233e-01 1.65583268e-01
-3.23965371e-01 -4.84914929e-02 -3.30307245e-01 8.63232732e-01
-5.93793213e-01 3.95964801e-01 1.07101694e-01 -2.12141693e-01
-7.77029932e-01 3.70509326e-01 2.14637533e-01 -4.48692173e-01
-3.52137804e-01 9.08862352e-01 -3.06704551e-01 -9.54907596e-01
2.03149945e-01 -3.06898594e-01 -1.90952227e-01 -2.90568061e-02
5.64504974e-02 2.91154087e-01 1.57397106e-01 -3.62340599e-01
-2.72437304e-01 7.15704620e-01 -4.27069068e-01 -2.40242437e-01
1.25652206e+00 1.19449440e-02 -1.31358728e-01 2.77125329e-01
1.16238725e+00 -1.85522139e-02 -1.15957463e+00 -2.83820242e-01
3.90917927e-01 -2.12353155e-01 1.95201367e-01 -1.12052965e+00
-1.15947890e+00 1.22923958e+00 4.92908388e-01 -1.53370470e-01
1.15463817e+00 -4.48882371e-01 1.21683812e+00 5.96364379e-01
2.95725465e-01 -1.70939863e+00 -2.38302171e-01 1.11794353e+00
3.47047925e-01 -1.41795862e+00 -3.06083611e-03 -3.70733708e-01
-8.48449707e-01 1.00738466e+00 9.05152619e-01 -2.70279884e-01
6.47652149e-01 5.10811508e-01 4.51670945e-01 2.66873807e-01
-6.24842286e-01 -4.58059132e-01 2.41572425e-01 5.28279781e-01
9.36307907e-01 1.32649213e-01 -8.76423895e-01 9.28984940e-01
1.10247564e-02 -2.56919283e-02 3.15919250e-01 9.05538678e-01
-4.03871983e-01 -1.65946329e+00 9.89347044e-03 3.76564711e-01
-8.73977542e-01 -7.10738838e-01 -7.01223195e-01 7.98898637e-01
1.17730975e-01 9.51943994e-01 -1.93874061e-01 -5.36835790e-01
2.41328686e-01 2.30731890e-01 2.79712051e-01 -8.48173738e-01
-9.57330763e-01 5.45858681e-01 1.97050601e-01 -2.73963153e-01
-3.39361906e-01 -4.26379532e-01 -1.16562951e+00 2.47327238e-02
-9.06495869e-01 2.03974620e-01 7.88958251e-01 1.08819354e+00
2.77254075e-01 4.82477248e-01 5.21169722e-01 -6.80609941e-01
-6.97350144e-01 -1.41003263e+00 3.71480696e-02 2.54302606e-04
-4.64776643e-02 -1.32923141e-01 3.72022693e-03 3.53216678e-02]
|
[10.001847267150879, 10.162835121154785]
|
f2168ce1-2bc2-4c55-a46c-d6607481258a
|
lm-cppf-paraphrasing-guided-data-augmentation
|
2305.18169
| null |
https://arxiv.org/abs/2305.18169v3
|
https://arxiv.org/pdf/2305.18169v3.pdf
|
LM-CPPF: Paraphrasing-Guided Data Augmentation for Contrastive Prompt-Based Few-Shot Fine-Tuning
|
In recent years, there has been significant progress in developing pre-trained language models for NLP. However, these models often struggle when fine-tuned on small datasets. To address this issue, researchers have proposed various adaptation approaches. Prompt-based tuning is arguably the most common way, especially for larger models. Previous research shows that adding contrastive learning to prompt-based fine-tuning is effective as it helps the model generate embeddings that are more distinguishable between classes, and it can also be more sample-efficient as the model learns from positive and negative examples simultaneously. One of the most important components of contrastive learning is data augmentation, but unlike computer vision, effective data augmentation for NLP is still challenging. This paper proposes LM-CPPF, Contrastive Paraphrasing-guided Prompt-based Fine-tuning of Language Models, which leverages prompt-based few-shot paraphrasing using generative language models, especially large language models such as GPT-3 and OPT-175B, for data augmentation. Our experiments on multiple text classification benchmarks show that this augmentation method outperforms other methods, such as easy data augmentation, back translation, and multiple templates.
|
['Yadollah Yaghoobzadeh', 'Sascha Rothe', 'Amirhossein Abaskohi']
|
2023-05-29
| null | null | null | null |
['sentiment-analysis', 'linguistic-acceptability']
|
['natural-language-processing', 'natural-language-processing']
|
[ 2.58302987e-01 1.31853908e-01 -6.10163391e-01 -4.15783495e-01
-9.89498317e-01 -6.86879277e-01 8.91020417e-01 2.37220481e-01
-4.99578089e-01 7.87842572e-01 5.78290582e-01 -4.42532331e-01
3.04511786e-01 -7.43691981e-01 -6.93087161e-01 -3.41922522e-01
6.06377900e-01 1.02472270e+00 -1.16400875e-01 -5.63155115e-01
2.58116543e-01 2.80489922e-01 -1.01172638e+00 4.10133719e-01
1.02816343e+00 4.32195783e-01 1.03935502e-01 5.17799675e-01
-9.16226029e-01 3.98168296e-01 -5.93178928e-01 -4.83124405e-01
3.13311756e-01 -2.81525493e-01 -8.19182754e-01 -1.04566097e-01
5.28537750e-01 -1.56119496e-01 -6.32064864e-02 6.70016527e-01
4.88453865e-01 4.15097445e-01 6.67655587e-01 -1.46452999e+00
-1.21260309e+00 8.06949139e-01 -6.26241624e-01 2.20218524e-01
1.80469826e-01 3.04780573e-01 1.05130577e+00 -1.36934745e+00
2.88305610e-01 1.61365962e+00 6.75506473e-01 8.55717838e-01
-1.54801083e+00 -9.43996370e-01 2.03853101e-01 -1.24533735e-02
-9.89830136e-01 -4.01549190e-01 7.36649930e-01 -2.97105253e-01
1.24398172e+00 4.98061851e-02 5.72914541e-01 1.40763903e+00
-2.45981351e-01 9.81326342e-01 1.17463374e+00 -8.58537376e-01
-9.35757756e-02 4.28103268e-01 2.95919925e-01 3.35804969e-01
1.65295079e-01 -4.82987100e-03 -4.16886449e-01 -3.63079369e-01
6.98966563e-01 2.02820778e-01 -1.64049454e-02 -2.52743304e-01
-1.18987918e+00 1.18588841e+00 2.41896972e-01 2.52468675e-01
-1.67177141e-01 -9.61952433e-02 4.13209260e-01 2.36326650e-01
8.17108989e-01 1.17151558e+00 -9.24237847e-01 -2.91777074e-01
-9.78524029e-01 3.70082766e-01 8.00967991e-01 7.67563581e-01
7.33834505e-01 8.11583847e-02 -7.78286159e-01 1.25927365e+00
2.68760592e-01 4.03972536e-01 9.29166377e-01 -5.89611232e-01
8.60697389e-01 9.04667616e-01 -5.99350072e-02 -5.28686166e-01
-2.30094731e-01 -3.91818255e-01 -8.70995700e-01 -1.75181478e-01
1.21731453e-01 -1.46724954e-01 -1.35437250e+00 1.99426687e+00
1.58449337e-01 1.45354703e-01 1.24600761e-01 3.20934832e-01
7.74337351e-01 8.96089971e-01 3.94998223e-01 -2.02733260e-02
1.25460100e+00 -1.43482506e+00 -7.82143295e-01 -8.68855596e-01
7.70005882e-01 -9.21739936e-01 1.96181536e+00 9.95242316e-03
-8.28411222e-01 -6.34054124e-01 -7.08809853e-01 -3.07874858e-01
-5.80924392e-01 7.21896067e-02 6.09652460e-01 6.54048264e-01
-8.16951632e-01 2.13355184e-01 -6.14580214e-01 -5.45233488e-01
5.17210186e-01 1.10162072e-01 -3.53858382e-01 -2.27095455e-01
-1.25258446e+00 1.10440159e+00 5.64093769e-01 -5.03192663e-01
-5.64960897e-01 -1.12080932e+00 -8.83624077e-01 1.34224027e-01
2.87404805e-01 -8.61372411e-01 1.37816799e+00 -7.41793454e-01
-1.35485089e+00 8.03430974e-01 -3.75885516e-01 -5.34758508e-01
2.51886189e-01 -4.75957125e-01 -1.62932485e-01 -4.48379517e-01
1.72723815e-01 1.21870756e+00 9.68303800e-01 -1.13223302e+00
-2.66540527e-01 -1.31923646e-01 -4.14099805e-02 4.39376503e-01
-1.03018224e+00 6.20181710e-02 -3.43678594e-01 -1.01808536e+00
-3.00596982e-01 -7.79667497e-01 -3.97051930e-01 -2.38278940e-01
-3.70715916e-01 -4.48385179e-01 9.27364111e-01 -5.51318049e-01
1.53665769e+00 -1.87122512e+00 -1.69833019e-01 -9.62771773e-02
-1.64013170e-02 7.23957062e-01 -6.51749492e-01 4.79715526e-01
-1.93958640e-01 4.80144233e-01 -1.73526838e-01 -5.60700476e-01
-1.57369412e-02 4.84721452e-01 -6.99870765e-01 -4.34890717e-01
6.32641613e-01 1.31066453e+00 -8.35914016e-01 -3.87708455e-01
1.43009230e-01 3.12101036e-01 -7.15622246e-01 2.37149775e-01
-5.40530384e-01 3.98319364e-01 -1.43680453e-01 4.13507342e-01
4.21981484e-01 -4.55922365e-01 -2.82137275e-01 -8.77580512e-03
3.04728270e-01 5.14286518e-01 -9.78609145e-01 1.52083373e+00
-6.01058364e-01 5.82320988e-01 -4.71990377e-01 -9.72461820e-01
1.05035079e+00 3.13046694e-01 6.84219822e-02 -5.71797371e-01
-1.53849512e-01 -3.74861844e-02 -1.36586800e-01 -4.40806597e-01
7.71714211e-01 -1.75919950e-01 1.69098917e-02 7.36829937e-01
9.02179778e-02 -4.55396175e-01 4.54742044e-01 3.86975825e-01
1.00645864e+00 6.72793612e-02 5.14162600e-01 1.96418643e-01
3.01804274e-01 1.59827054e-01 3.41370851e-01 8.52150917e-01
-4.14354756e-04 5.68487406e-01 -7.52693638e-02 -1.66465715e-01
-1.17865074e+00 -8.53705585e-01 1.65350795e-01 1.49258375e+00
-5.22148609e-01 -5.62052608e-01 -5.82815945e-01 -8.22063684e-01
1.13401085e-01 1.04023254e+00 -6.62095726e-01 -4.55333650e-01
-5.83492935e-01 -9.49441493e-01 5.46713054e-01 8.38963687e-01
4.03496355e-01 -1.24009669e+00 2.74112791e-01 3.91861975e-01
-4.11946923e-01 -1.10624242e+00 -7.46384382e-01 2.93597162e-01
-1.03774154e+00 -6.83181882e-01 -6.49052680e-01 -8.87290180e-01
8.98465037e-01 4.76877600e-01 1.39527357e+00 -8.17753673e-02
-1.59438863e-01 4.14751261e-01 -4.83996391e-01 -7.94217348e-01
-6.97685003e-01 4.17146862e-01 3.62810493e-02 -3.32829744e-01
7.31215298e-01 -4.26379532e-01 -1.42177165e-01 1.78235516e-01
-8.65098655e-01 1.99379548e-01 9.11337674e-01 1.30069733e+00
6.64383233e-01 -4.66935456e-01 6.83308303e-01 -1.09990478e+00
1.33491707e+00 -3.35325539e-01 -9.93228480e-02 4.82786387e-01
-9.00828004e-01 2.34900713e-01 6.73897326e-01 -9.79275465e-01
-9.47040260e-01 -2.01199085e-01 -2.49575108e-01 -4.21084672e-01
-1.54313773e-01 4.72989023e-01 5.76843098e-02 -1.01276077e-02
1.14316404e+00 3.23540032e-01 1.80697367e-02 -4.83403534e-01
7.37458289e-01 6.24210298e-01 2.24509209e-01 -6.72221363e-01
1.11772585e+00 1.16675228e-01 -5.11426210e-01 -7.94888914e-01
-1.21391630e+00 -4.99727368e-01 -5.04210472e-01 3.82550985e-01
3.92861545e-01 -8.30791056e-01 1.52089328e-01 3.12542543e-02
-1.12173867e+00 -5.51615357e-01 -7.25771487e-01 2.82937497e-01
-2.72931188e-01 2.50460356e-01 -5.08807182e-01 -6.14373744e-01
-8.06400299e-01 -8.28896880e-01 9.98656929e-01 1.78964317e-01
-6.77380741e-01 -1.21083534e+00 4.00853992e-01 6.62385166e-01
6.99539959e-01 -3.87943566e-01 1.14729989e+00 -1.23485792e+00
-2.25348547e-01 -3.36186826e-01 -1.34015203e-01 5.35504818e-01
2.97478050e-01 -7.10285231e-02 -9.15349305e-01 -2.14479387e-01
-3.12865764e-01 -7.05162942e-01 7.66437173e-01 2.53633559e-01
1.14297116e+00 -6.05154991e-01 -3.81390393e-01 3.72747511e-01
9.45660770e-01 -3.43309678e-02 4.48500544e-01 4.36585516e-01
7.09805608e-01 3.42533767e-01 8.59456837e-01 1.29898369e-01
4.27911013e-01 6.63585782e-01 2.04403512e-03 -3.14180285e-01
-3.05508822e-01 -5.60452819e-01 3.99806231e-01 8.14947128e-01
4.75167453e-01 -1.36498407e-01 -1.02877438e+00 6.10733628e-01
-1.65436625e+00 -8.34750533e-01 2.22945184e-01 1.87436295e+00
1.36100090e+00 1.65736288e-01 -4.82154042e-02 6.08640388e-02
4.47504908e-01 -9.53956246e-02 -6.22892141e-01 -4.23644871e-01
-1.80633798e-01 5.45986831e-01 1.16136998e-01 4.63768095e-01
-9.71899807e-01 1.30727279e+00 6.45653868e+00 1.00689304e+00
-9.42341387e-01 3.44764799e-01 6.79446578e-01 -1.49993494e-01
-4.99029487e-01 4.19932380e-02 -1.28967285e+00 3.47823054e-01
7.74501562e-01 -3.22379202e-01 3.15190196e-01 1.03008974e+00
1.85219109e-01 3.14891219e-01 -1.19859552e+00 9.27783847e-01
3.21735412e-01 -1.51384127e+00 5.40607274e-01 -5.50497584e-02
1.03343165e+00 1.50558278e-01 1.23897500e-01 1.07040727e+00
5.22311032e-01 -9.85064566e-01 1.42397851e-01 5.51805049e-02
5.43165922e-01 -5.70722580e-01 5.36262691e-01 5.59108436e-01
-8.29751253e-01 -7.85494596e-02 -4.23751354e-01 -8.72329846e-02
1.40291512e-01 4.78784174e-01 -1.38964665e+00 3.45493783e-03
4.73631561e-01 4.00154769e-01 -8.28055084e-01 8.01062226e-01
-4.88806844e-01 8.14732075e-01 -2.11295560e-01 -2.04749301e-01
1.48140416e-01 5.50462641e-02 3.52120966e-01 1.32405126e+00
2.51718819e-01 -2.34559163e-01 4.06741947e-01 8.02127421e-01
-3.58669966e-01 3.04434657e-01 -6.14174426e-01 -5.39966762e-01
7.98145115e-01 1.38443160e+00 -2.79993087e-01 -6.13317609e-01
-4.53722000e-01 8.61414135e-01 5.83349288e-01 3.39056015e-01
-6.06450915e-01 -1.40869826e-01 6.14088237e-01 -5.11204940e-04
-1.40278220e-01 -1.35300934e-01 -5.31864047e-01 -1.24641979e+00
-1.24144241e-01 -1.15098464e+00 3.77798468e-01 -8.27196598e-01
-1.66631508e+00 3.74394327e-01 -6.91807270e-02 -1.00251925e+00
-4.29321915e-01 -5.29573917e-01 -5.83223879e-01 1.05769038e+00
-1.62704945e+00 -1.42599010e+00 -2.03143343e-01 4.44767326e-01
1.09490132e+00 -3.49503905e-01 1.26204443e+00 1.78488344e-01
-5.86362839e-01 8.70629251e-01 -2.56153252e-02 2.18645364e-01
1.18973494e+00 -1.30835497e+00 9.57675338e-01 8.09569299e-01
5.17753422e-01 8.77966464e-01 5.69687068e-01 -6.96065962e-01
-9.99922872e-01 -1.35188842e+00 9.88836110e-01 -7.84821928e-01
6.88634217e-01 -4.53635097e-01 -1.25708151e+00 8.17409754e-01
3.02191734e-01 -1.46392763e-01 9.21991587e-01 3.88895869e-01
-4.46721971e-01 1.24154054e-02 -1.01422834e+00 8.22059035e-01
7.69707382e-01 -5.18528938e-01 -1.06980646e+00 6.25099540e-01
1.02265763e+00 -3.78671974e-01 -5.76471090e-01 3.02080661e-01
1.80456102e-01 -1.58328310e-01 1.17886317e+00 -1.03923571e+00
4.61433172e-01 1.65676460e-01 1.52146816e-01 -1.70987570e+00
-3.95904601e-01 -5.39341807e-01 -3.78906637e-01 1.60152042e+00
7.75234401e-01 -6.59625828e-01 1.03322959e+00 7.30978608e-01
-1.64867379e-02 -7.84807563e-01 -4.52271491e-01 -9.59331155e-01
3.64898086e-01 -3.93222690e-01 5.07788002e-01 1.36882222e+00
-1.61034226e-01 8.54679823e-01 -2.77402192e-01 -2.43398920e-01
1.83676243e-01 -1.62817776e-01 1.26427114e+00 -1.28996575e+00
-2.95554757e-01 -4.56407160e-01 2.42891550e-01 -1.14568722e+00
2.02413410e-01 -1.09680116e+00 -9.90437046e-02 -1.79348719e+00
4.18567747e-01 -5.49396574e-01 -1.55322403e-01 1.02872634e+00
-6.12936854e-01 1.25327587e-01 4.30351384e-02 1.72959313e-01
-1.14462845e-01 6.24020398e-01 1.29708636e+00 -2.07721561e-01
-4.85513836e-01 -1.17483437e-01 -1.05171406e+00 5.92374325e-01
1.03955328e+00 -4.41072583e-01 -5.32427788e-01 -4.75954473e-01
2.90238947e-01 -6.05134666e-01 9.67089683e-02 -6.18963122e-01
1.45501941e-01 -2.78783023e-01 4.01375502e-01 -4.95875925e-01
6.24543905e-01 -5.51192403e-01 -4.78191674e-01 2.98292190e-01
-6.57285750e-01 2.78837681e-01 6.50506377e-01 4.42580074e-01
-1.85049012e-01 -2.78020591e-01 7.48996317e-01 -2.72934079e-01
-6.33743942e-01 4.28121656e-01 -8.80490988e-03 5.94441772e-01
6.83697224e-01 -1.19323157e-01 -4.55709696e-01 -2.53068924e-01
-4.45232362e-01 4.30206507e-01 1.85665667e-01 8.27222347e-01
5.35326302e-01 -1.47401476e+00 -8.39038968e-01 4.13768232e-01
4.49954420e-01 -8.56396481e-02 -5.23788743e-02 6.01909518e-01
1.29924655e-01 5.22818744e-01 6.95860153e-03 -4.93562996e-01
-1.35341132e+00 5.08576751e-01 -8.96008983e-02 -7.61355281e-01
-3.21827859e-01 1.10491490e+00 1.96536154e-01 -1.00739360e+00
1.99421003e-01 -3.63079756e-01 -2.63176620e-01 9.26936865e-02
6.33704007e-01 -1.37219690e-02 2.18027290e-02 -8.51818845e-02
5.27248941e-02 3.21280211e-01 -6.22476637e-01 -3.23601454e-01
1.30646455e+00 4.93394258e-03 1.21431882e-02 4.53364879e-01
7.81329513e-01 -1.61143959e-01 -9.22971547e-01 -8.05683017e-01
4.25668173e-02 -4.34703290e-01 -1.40688404e-01 -1.21681917e+00
-5.93969345e-01 1.19888461e+00 3.29605162e-01 1.09816857e-01
8.84452283e-01 -1.28105640e-01 8.39846075e-01 6.84270203e-01
-4.05306034e-02 -1.23097706e+00 6.82510972e-01 8.82019460e-01
1.13313198e+00 -1.59700787e+00 -2.53995597e-01 -2.03829139e-01
-7.13019133e-01 9.14425671e-01 1.07211840e+00 2.10601702e-01
2.89344132e-01 1.17745876e-01 5.54257073e-02 2.18197897e-01
-9.25230503e-01 4.14777063e-02 4.16137487e-01 5.57782233e-01
6.77247941e-01 -1.04069903e-01 -1.03057846e-01 5.62022388e-01
-3.95633459e-01 -3.33907665e-03 1.44309044e-01 8.92271757e-01
-5.07911146e-01 -1.51824033e+00 -4.23984766e-01 7.82189608e-01
-2.27547392e-01 -6.53579235e-01 -4.47421044e-01 7.94629335e-01
-2.17928201e-01 7.66884983e-01 -2.14710142e-02 -1.14189036e-01
2.96921283e-01 6.21216536e-01 4.88333255e-01 -1.31261528e+00
-6.05368257e-01 -2.40938723e-01 5.95016405e-02 -1.61069497e-01
-8.68844017e-02 -3.67456257e-01 -1.12861693e+00 -7.98525810e-02
-3.94469708e-01 2.20445842e-01 6.67342126e-01 1.03259099e+00
5.32898605e-01 5.35577476e-01 4.26849753e-01 -7.20830560e-01
-8.50025296e-01 -1.46234524e+00 5.81753664e-02 4.47082371e-01
7.56203681e-02 -5.41158080e-01 -3.15222561e-01 -1.07093662e-01]
|
[10.818796157836914, 8.280889511108398]
|
d9d7e011-ce0b-4bc6-9ad7-82e092c92d94
|
data-driven-segmentation-of-post-mortem-iris
|
1807.04154
| null |
http://arxiv.org/abs/1807.04154v1
|
http://arxiv.org/pdf/1807.04154v1.pdf
|
Data-Driven Segmentation of Post-mortem Iris Images
|
This paper presents a method for segmenting iris images obtained from the
deceased subjects, by training a deep convolutional neural network (DCNN)
designed for the purpose of semantic segmentation. Post-mortem iris recognition
has recently emerged as an alternative, or additional, method useful in
forensic analysis. At the same time it poses many new challenges from the
technological standpoint, one of them being the image segmentation stage, which
has proven difficult to be reliably executed by conventional iris recognition
methods. Our approach is based on the SegNet architecture, fine-tuned with
1,300 manually segmented post-mortem iris images taken from the
Warsaw-BioBase-Post-Mortem-Iris v1.0 database. The experiments presented in
this paper show that this data-driven solution is able to learn specific
deformations present in post-mortem samples, which are missing from alive
irises, and offers a considerable improvement over the state-of-the-art,
conventional segmentation algorithm (OSIRIS): the Intersection over Union (IoU)
metric was improved from 73.6% (for OSIRIS) to 83% (for DCNN-based presented in
this paper) averaged over subject-disjoint, multiple splits of the data into
train and test subsets. This paper offers the first known to us method of
automatic processing of post-mortem iris images. We offer source codes with the
trained DCNN that perform end-to-end segmentation of post-mortem iris images,
as described in this paper. Also, we offer binary masks corresponding to manual
segmentation of samples from Warsaw-BioBase-Post-Mortem-Iris v1.0 database to
facilitate development of alternative methods for post-mortem iris
segmentation.
|
['Mateusz Trokielewicz', 'Adam Czajka']
|
2018-07-11
| null | null | null | null |
['iris-segmentation']
|
['medical']
|
[ 2.87004799e-01 1.65580660e-01 4.79795970e-03 -3.78996849e-01
-6.66019261e-01 -5.08211493e-01 3.12832057e-01 1.88259214e-01
-6.26347959e-01 4.28368628e-01 -1.41402498e-01 -3.90015811e-01
-6.51211083e-01 -4.22894388e-01 -4.18765604e-01 -1.02316999e+00
-7.79533014e-02 9.23671961e-01 -2.54276991e-01 -2.94253640e-02
3.72308165e-01 1.08501041e+00 -1.69994926e+00 1.78132936e-01
7.90792644e-01 7.05179334e-01 -5.66315770e-01 8.00498307e-01
2.40351468e-01 1.49905428e-01 -7.44791269e-01 -5.37405372e-01
5.29130399e-01 -5.51605821e-01 -1.12099051e+00 2.63988435e-01
9.99484420e-01 -1.27764836e-01 -1.07502967e-01 9.67396617e-01
7.37221241e-01 1.93134204e-01 5.82689643e-01 -6.94822371e-01
-2.57344633e-01 3.38564724e-01 -6.97904587e-01 3.82752597e-01
1.02157444e-01 4.18053269e-01 3.84180546e-01 -2.59042561e-01
6.42449677e-01 7.52345502e-01 8.57565463e-01 8.38086545e-01
-1.19296873e+00 -3.27096283e-01 -7.90165305e-01 6.10676631e-02
-1.36474717e+00 -4.29876983e-01 4.79002953e-01 -8.01774025e-01
8.57647359e-01 6.25772774e-01 7.15490639e-01 5.01124382e-01
-1.90725196e-02 7.03983009e-01 1.36741090e+00 -4.37509418e-01
3.17411199e-02 -2.84743398e-01 1.60994783e-01 5.98767281e-01
-5.51096397e-04 6.49199784e-01 -1.86368078e-01 4.83507589e-02
6.61274314e-01 -1.84658229e-01 -1.97770432e-01 6.56296872e-03
-9.57017064e-01 5.02310336e-01 3.68462741e-01 6.41243696e-01
-3.26403320e-01 -4.24152166e-01 5.71468234e-01 3.52247477e-01
5.71969628e-01 5.48016787e-01 -3.66115898e-01 -2.94773486e-02
-1.76920378e+00 6.17608167e-02 4.29156899e-01 1.94363922e-01
5.87023973e-01 -1.13498554e-01 -2.85333425e-01 5.93531907e-01
2.27468431e-01 1.55040205e-01 5.96496046e-01 -4.18305665e-01
-4.67357524e-02 7.94292152e-01 -2.82378525e-01 -7.67726898e-01
-7.66029775e-01 -5.94496489e-01 -9.50000167e-01 4.01424438e-01
7.83672929e-01 -1.60151869e-01 -1.54000354e+00 1.04573870e+00
4.84575063e-01 2.08131015e-01 -4.48114658e-03 1.09343541e+00
1.03393483e+00 1.88164309e-01 -3.02444156e-02 -3.05833340e-01
1.31479716e+00 -6.83683574e-01 -2.84144491e-01 3.41558367e-01
3.71556222e-01 -1.01771665e+00 4.68985081e-01 6.97768748e-01
-9.54194605e-01 -5.24618506e-01 -7.47053564e-01 1.32583722e-01
-4.26574826e-01 4.47868615e-01 4.11801845e-01 1.01726127e+00
-1.13382578e+00 8.26893151e-01 -9.70634222e-01 -4.29521084e-01
6.04852021e-01 9.85888124e-01 -5.04338622e-01 3.00997704e-01
-6.08919740e-01 5.96379936e-01 5.45368671e-01 2.96804339e-01
-5.26981950e-01 -6.34660006e-01 -8.37178230e-01 -2.43727118e-01
7.11877644e-02 -4.20619607e-01 8.62003803e-01 -1.10576987e+00
-1.41476250e+00 1.74583292e+00 -9.27378833e-02 -8.01027894e-01
6.34835839e-01 2.77522027e-01 -4.46642905e-01 4.85487491e-01
-2.78356373e-01 5.33531785e-01 8.28947842e-01 -1.00390482e+00
-5.08860946e-01 -7.46041715e-01 -3.75595689e-01 -3.07370991e-01
2.65757293e-01 4.18464273e-01 -3.87047023e-01 -7.99669564e-01
-2.22638864e-02 -6.16210222e-01 -1.84145644e-01 -3.97426486e-01
-5.22646308e-01 -3.37906182e-01 5.20282269e-01 -1.12709618e+00
9.73839641e-01 -2.04145265e+00 2.21715104e-02 4.18330610e-01
3.43407065e-01 9.70093369e-01 -7.86098912e-02 3.17027979e-02
-5.83537579e-01 8.88349116e-02 -7.36181319e-01 -6.68644369e-01
-2.97931731e-01 3.45475040e-02 1.38053790e-01 1.00052559e+00
2.61169951e-02 7.23202527e-01 -5.50755382e-01 -5.87399185e-01
5.22045612e-01 4.87523288e-01 -3.18349004e-02 2.15948597e-01
6.51417673e-02 6.62829101e-01 -7.35146925e-03 9.89467084e-01
9.23385382e-01 2.84366965e-01 -4.08860922e-01 -1.01506390e-01
-1.51741207e-01 -3.00516456e-01 -9.62145984e-01 1.68455935e+00
1.00824080e-01 7.63001025e-01 2.16046512e-01 -1.42102778e+00
1.01452172e+00 6.35969937e-01 6.03732586e-01 -4.80348825e-01
4.42072153e-01 2.38445044e-01 -1.44518036e-02 -6.92685902e-01
3.80667299e-01 -5.76911569e-01 4.93867129e-01 2.92488933e-01
3.17179888e-01 8.85065794e-02 6.42335176e-01 -4.35012251e-01
4.75368857e-01 2.61488289e-01 6.32429868e-02 -2.14159742e-01
8.80733788e-01 1.62935689e-01 6.25236869e-01 3.53952169e-01
-4.54788893e-01 1.04538119e+00 4.09800678e-01 -9.44660544e-01
-9.78403032e-01 -5.88091314e-01 -6.61306262e-01 2.72952735e-01
-2.82025248e-01 7.01248646e-02 -1.31571329e+00 -7.80766964e-01
-1.67010635e-01 2.00766400e-01 -7.76981533e-01 4.06329244e-01
-5.70161104e-01 -1.27779222e+00 7.54317105e-01 -1.01010129e-01
4.06960756e-01 -1.18164146e+00 -4.53397185e-01 1.95130050e-01
1.95139810e-01 -7.02313602e-01 -1.83728382e-01 -4.12501097e-01
-1.05534899e+00 -1.72346365e+00 -1.03685808e+00 -6.73817754e-01
8.78137231e-01 -5.30735850e-01 9.61673081e-01 5.22249579e-01
-9.70025659e-01 1.87458098e-01 -1.10730834e-01 -3.15532148e-01
-5.16808212e-01 -3.20596155e-03 -5.50608784e-02 5.65453768e-01
7.57656574e-01 -3.58634830e-01 -6.17253304e-01 1.98495075e-01
-1.13479662e+00 -2.21867681e-01 4.38837618e-01 9.00680542e-01
7.42680430e-01 1.47640631e-01 2.27760151e-02 -7.60072112e-01
2.45567322e-01 1.23601906e-01 -8.33054364e-01 2.59340912e-01
-6.19371772e-01 -4.03430969e-01 4.14135426e-01 1.05683237e-01
-6.72396541e-01 2.09816888e-01 -5.97884595e-01 -3.41444403e-01
-1.01026773e+00 2.76457042e-01 2.69624799e-01 -5.65463543e-01
7.29754090e-01 2.95505941e-01 3.21005374e-01 -8.39890897e-01
4.28123400e-02 1.00633454e+00 9.72267985e-01 -4.40402925e-01
6.40587270e-01 4.91862178e-01 3.60248238e-01 -8.61518383e-01
-4.78333205e-01 -8.67975533e-01 -9.94717777e-01 -1.23072593e-02
1.05470109e+00 -3.84016365e-01 -8.96220684e-01 1.04767907e+00
-7.91853666e-01 -2.40806997e-01 -5.36648214e-01 4.67499048e-01
-5.86978257e-01 6.74644947e-01 -6.30947769e-01 -8.06104720e-01
-7.79933393e-01 -1.33939874e+00 9.16043103e-01 8.19975913e-01
2.33270135e-02 -1.01688647e+00 2.50575125e-01 9.76468444e-01
8.64450857e-02 7.90568650e-01 6.54731989e-01 -7.78828740e-01
-2.24235013e-01 -4.79902506e-01 -1.36646822e-01 6.27020657e-01
-3.12759122e-03 4.61982161e-01 -1.05248129e+00 -4.99206632e-01
-8.89574662e-02 1.00353658e-01 9.93800581e-01 6.31118000e-01
1.12692153e+00 -8.36963579e-02 -1.17700532e-01 1.17226577e+00
1.59015691e+00 2.98866481e-01 9.60144579e-01 4.45983917e-01
3.55245143e-01 7.81031668e-01 3.46504986e-01 3.57515439e-02
-1.39189646e-01 4.99869555e-01 2.96736151e-01 -7.09113896e-01
-3.05107564e-01 3.31464589e-01 -3.14540058e-01 8.59044343e-02
-6.54837132e-01 6.96593150e-02 -1.13029480e+00 9.76872146e-01
-1.39352596e+00 -8.59015942e-01 -3.37931335e-01 2.46245337e+00
8.68129551e-01 -1.75661772e-01 4.51029480e-01 3.34714800e-01
7.70675421e-01 -2.29373649e-01 -2.97232479e-01 -4.59943533e-01
-2.12852761e-01 9.99350429e-01 4.61640179e-01 6.07490122e-01
-1.36765647e+00 7.84447610e-01 5.60085297e+00 8.72863948e-01
-1.24803412e+00 -1.06552191e-01 9.03851569e-01 -5.80598451e-02
5.70981979e-01 -1.80422753e-01 -6.13464594e-01 4.98237044e-01
1.10579503e+00 4.18109208e-01 3.48394513e-01 9.60464701e-02
1.86907038e-01 -3.47578049e-01 -7.02831745e-01 1.05010915e+00
2.43894428e-01 -1.43544221e+00 -2.27549613e-01 2.33011782e-01
6.21378899e-01 1.66868158e-02 3.67395207e-02 -3.13174814e-01
-3.67835701e-01 -1.32965529e+00 5.89266494e-02 7.68289804e-01
1.08077729e+00 -1.06260955e+00 1.20600176e+00 5.77052087e-02
-6.36361361e-01 1.29470974e-01 -1.64195344e-01 2.89893329e-01
2.15025451e-02 6.56183064e-01 -6.84484541e-01 7.91715980e-01
6.02936208e-01 6.18044674e-01 -7.26497114e-01 1.59225416e+00
-1.21658586e-01 8.15219045e-01 -2.35549465e-01 6.38000429e-01
2.92000741e-01 -3.64741087e-01 7.71828949e-01 1.12759733e+00
1.23529993e-01 5.98672330e-02 -4.09926116e-01 9.36806321e-01
1.42137066e-01 1.79277167e-01 -2.70972941e-02 -1.20857820e-01
-3.55387390e-01 1.26303160e+00 -9.40262973e-01 -2.63882935e-01
1.02584623e-02 7.74031818e-01 -2.87862837e-01 2.52175510e-01
-1.67231053e-01 -5.92178106e-01 5.85257530e-01 2.04841793e-01
1.07036173e-01 1.64619491e-01 -3.82632017e-01 -9.67777610e-01
-1.89132109e-01 -1.08210742e+00 6.63200915e-01 -5.12787342e-01
-1.15503991e+00 7.22594798e-01 -1.27653390e-01 -9.38775837e-01
-2.34236673e-01 -8.50868821e-01 -6.53249145e-01 1.33714437e+00
-1.34945297e+00 -1.48851871e+00 2.10739542e-02 4.29622889e-01
1.49620920e-01 -6.17718220e-01 8.39842200e-01 3.12939495e-01
-9.22354758e-01 7.59388506e-01 1.40330970e-01 5.86812794e-01
5.52907884e-01 -1.38596785e+00 3.81137818e-01 1.15454972e+00
2.12732285e-01 6.24589682e-01 5.85362017e-01 -4.73451108e-01
-7.67816484e-01 -6.94115877e-01 1.07816756e+00 -2.47278512e-01
1.55246601e-01 3.28232110e-01 -7.83515096e-01 3.97873461e-01
2.31106877e-01 -6.98808581e-02 1.06923389e+00 -7.43186921e-02
7.83264041e-02 -8.37930292e-03 -1.82072616e+00 2.39578158e-01
2.77998954e-01 -4.22861159e-01 -8.33990455e-01 4.06592935e-01
8.41108486e-02 -8.51976693e-01 -1.17555416e+00 4.25748855e-01
3.95342618e-01 -1.35792196e+00 1.02051079e+00 -7.76854157e-01
2.65860766e-01 -4.67087746e-01 6.02806926e-01 -9.06966984e-01
2.52121627e-01 -1.07792580e+00 4.40882929e-02 1.10346103e+00
2.40964875e-01 -7.22227752e-01 1.17530870e+00 5.52005768e-01
-7.31334686e-02 -8.00476789e-01 -1.35514390e+00 -4.81282860e-01
1.48698390e-01 -1.58786759e-01 6.98652983e-01 9.39237058e-01
-4.94493157e-01 -4.27540332e-01 5.88461421e-02 1.26116350e-01
8.35108876e-01 1.67593688e-01 6.69389009e-01 -1.38139689e+00
-2.11009398e-01 -7.49158442e-01 -9.50509369e-01 -3.80687773e-01
8.85098428e-03 -7.64011741e-01 -3.74568403e-01 -1.33687556e+00
-2.06065580e-01 -2.27562189e-01 -2.20803499e-01 7.41847098e-01
6.20169230e-02 9.50542688e-01 -4.33682203e-02 2.47312188e-01
3.42987984e-01 -3.51657867e-01 1.19649661e+00 -2.14618012e-01
-3.40995759e-01 3.80852103e-01 -3.59419554e-01 5.46174645e-01
8.86495650e-01 -3.84863019e-01 3.01980078e-01 1.33334458e-01
-1.85299709e-01 2.67830081e-02 5.71777344e-01 -1.07995844e+00
2.36255884e-01 1.86617076e-01 3.26832503e-01 -7.53531158e-01
-7.67875090e-02 -6.20609760e-01 1.87360227e-01 4.48920727e-01
-3.08882799e-02 -4.46832448e-01 4.68300700e-01 -1.21139646e-01
-4.65421766e-01 -5.51231265e-01 1.14584494e+00 -1.47698790e-01
-4.93185937e-01 3.93718958e-01 -1.14286415e-01 -2.46486291e-01
9.33926463e-01 -7.10071683e-01 -1.61810637e-01 2.52443224e-01
-1.30210793e+00 -1.18608261e-02 6.72257781e-01 -8.04967284e-02
4.27821308e-01 -6.35420561e-01 -9.40975666e-01 7.42873371e-01
7.13931490e-03 2.25594521e-01 5.97605050e-01 1.36356866e+00
-1.13746488e+00 4.07765061e-01 -2.68311113e-01 -6.63507342e-01
-1.83298910e+00 5.34263432e-01 8.87567997e-01 -2.17030600e-01
-6.20518386e-01 9.65458870e-01 -5.26971817e-01 -4.96940404e-01
1.38052151e-01 -2.38842681e-01 -4.17843103e-01 2.66154408e-01
6.03334606e-01 3.12074512e-01 5.53101957e-01 -1.05833673e+00
-1.69088822e-02 1.05934465e+00 -6.25973269e-02 2.67185330e-01
1.36598170e+00 1.83266386e-01 -7.38300979e-01 -2.16835827e-01
9.41166282e-01 -2.25754127e-01 -8.02493274e-01 -1.91729236e-02
-1.50895953e-01 -6.02079988e-01 3.00939709e-01 -1.16265917e+00
-1.42435634e+00 9.41467643e-01 1.11974812e+00 1.31743357e-01
1.53001046e+00 -3.26617420e-01 8.46528172e-01 -1.35801183e-02
7.31342882e-02 -1.01498520e+00 -1.02164733e+00 1.79794226e-02
6.20752811e-01 -1.19264710e+00 4.67252471e-02 -1.27474025e-01
1.02212289e-02 1.46670270e+00 1.21070668e-01 1.82394758e-02
4.18411434e-01 -6.66538328e-02 4.26341414e-01 -4.45370615e-01
2.37227827e-01 -4.55959022e-01 8.32001269e-01 8.23593676e-01
5.50120354e-01 2.12719232e-01 -5.96583068e-01 4.54612523e-02
-2.25635856e-01 1.19655773e-01 4.11138922e-01 6.41529560e-01
6.63407221e-02 -1.37589741e+00 -7.24415839e-01 5.42886019e-01
-8.97049010e-01 -7.40927970e-03 -5.00826955e-01 9.34857965e-01
6.61644816e-01 7.16645002e-01 2.38239065e-01 -1.36443019e-01
1.52248874e-01 1.56402946e-01 5.34567833e-01 -3.50867510e-01
-1.23940992e+00 1.84214368e-01 -1.70062825e-01 -3.41430396e-01
-7.95767665e-01 -7.87565947e-01 -1.00440323e+00 -3.76268148e-01
-1.22257233e-01 1.87887743e-01 7.60923982e-01 1.01433647e+00
1.54685706e-01 2.42399603e-01 4.12802070e-01 -8.41375530e-01
-2.08931968e-01 -9.50116694e-01 -9.31766510e-01 3.76569211e-01
7.68072844e-01 -2.26712674e-01 -4.47159648e-01 2.92573541e-01]
|
[3.7429049015045166, -3.631761312484741]
|
e76a19db-e76a-4b77-8b4b-4bff2746838f
|
lexically-constrained-text-generation-through
|
2012.10813
| null |
https://arxiv.org/abs/2012.10813v1
|
https://arxiv.org/pdf/2012.10813v1.pdf
|
Lexically-constrained Text Generation through Commonsense Knowledge Extraction and Injection
|
Conditional text generation has been a challenging task that is yet to see human-level performance from state-of-the-art models. In this work, we specifically focus on the Commongen benchmark, wherein the aim is to generate a plausible sentence for a given set of input concepts. Despite advances in other tasks, large pre-trained language models that are fine-tuned on this dataset often produce sentences that are syntactically correct but qualitatively deviate from a human understanding of common sense. Furthermore, generated sequences are unable to fulfill such lexical requirements as matching part-of-speech and full concept coverage. In this paper, we explore how commonsense knowledge graphs can enhance model performance, with respect to commonsense reasoning and lexically-constrained decoding. We propose strategies for enhancing the semantic correctness of the generated text, which we accomplish through: extracting commonsense relations from Conceptnet, injecting these relations into the Unified Language Model (UniLM) through attention mechanisms, and enforcing the aforementioned lexical requirements through output constraints. By performing several ablations, we find that commonsense injection enables the generation of sentences that are more aligned with human understanding, while remaining compliant with lexical requirements.
|
['Alessandro Oltramari', 'Eric Nyberg', 'Kaixin Ma', 'Jonathan Francis', 'Har Simrat Singh', 'Varsha Kuppur Rajendra', 'Pulkit Goel', 'Yikang Li']
|
2020-12-19
| null | null | null | null |
['conditional-text-generation']
|
['natural-language-processing']
|
[ 9.18233514e-01 6.60512447e-01 -1.05291894e-02 -4.22089785e-01
-9.00248885e-01 -5.64328969e-01 8.20098996e-01 1.83628127e-01
-2.54790872e-01 9.33152616e-01 6.44051969e-01 -4.20636147e-01
2.33756661e-01 -9.52516019e-01 -7.31629193e-01 -3.42855044e-02
7.19791174e-01 6.16762280e-01 -1.17796317e-01 -6.66434586e-01
3.23223621e-01 -1.12146690e-01 -1.58506024e+00 6.12929285e-01
1.32771313e+00 4.76656139e-01 3.25913131e-01 4.89152551e-01
-3.29411864e-01 9.32008028e-01 -8.94899547e-01 -7.65417159e-01
-1.46653831e-01 -9.78602886e-01 -1.09384251e+00 -2.20046937e-01
3.75371724e-01 -3.81927490e-02 1.95328265e-01 1.17183149e+00
2.34813675e-01 1.90054566e-01 8.87312412e-01 -1.08524871e+00
-1.20256424e+00 1.24329150e+00 4.82335500e-03 5.65455146e-02
7.61468232e-01 3.26240063e-01 1.48659933e+00 -5.75326085e-01
8.21554363e-01 1.42790055e+00 3.66286606e-01 1.14659572e+00
-1.41547704e+00 -5.80033064e-01 3.20876479e-01 1.96342707e-01
-1.16459560e+00 -5.78140318e-01 7.49526322e-01 -2.77681738e-01
1.56148815e+00 2.53291458e-01 5.36602616e-01 1.58379626e+00
1.76318839e-01 5.72812974e-01 8.64747643e-01 -7.71948218e-01
3.07928145e-01 -4.40628529e-02 -4.69413139e-02 4.22965050e-01
3.95435542e-01 -5.26187643e-02 -7.81121433e-01 8.75734687e-02
2.42859498e-01 -5.28697014e-01 -3.89798284e-01 2.69465357e-01
-1.41223562e+00 1.05159819e+00 3.06488603e-01 5.19608676e-01
-4.00635421e-01 2.64323026e-01 3.11056256e-01 1.22295804e-01
3.39109510e-01 1.00417781e+00 -5.40312767e-01 -1.44011173e-02
-1.01204026e+00 5.53470910e-01 7.76033700e-01 1.14582837e+00
4.58444357e-01 1.55217335e-01 -4.83288199e-01 6.05480909e-01
1.80421710e-01 5.48330367e-01 8.49752724e-01 -8.31295848e-01
5.57995498e-01 6.23882055e-01 -2.06756204e-01 -7.32820690e-01
-2.00534597e-01 -3.51665884e-01 -6.41546309e-01 -2.48850554e-01
1.78607985e-01 -2.57514566e-01 -9.27787840e-01 2.35717440e+00
-5.90293258e-02 -9.11183432e-02 5.68930805e-01 6.45833790e-01
7.71597266e-01 6.63690209e-01 5.34273744e-01 -1.07034847e-01
1.39309311e+00 -5.99318862e-01 -7.07280636e-01 -7.72477686e-01
7.29846954e-01 -5.55261970e-01 1.54557371e+00 8.93230513e-02
-1.18953717e+00 -4.72701102e-01 -1.18437564e+00 -3.02575678e-01
-3.37512732e-01 -1.97231919e-01 6.38531089e-01 5.38111806e-01
-1.06019163e+00 4.57627445e-01 -4.40655500e-01 -3.00421417e-01
3.60375464e-01 -2.00922653e-01 -7.42290765e-02 -1.41264543e-01
-1.69572508e+00 1.39839721e+00 8.29209208e-01 -2.30130628e-01
-7.76440144e-01 -7.99692154e-01 -1.29549944e+00 2.30962127e-01
4.17487472e-01 -1.27450192e+00 1.43519211e+00 -1.06040406e+00
-1.30492282e+00 1.02606606e+00 -2.79190034e-01 -7.07563937e-01
1.50716379e-01 -1.33531705e-01 -3.36256057e-01 -9.79594439e-02
4.16302234e-01 1.04178417e+00 5.89047074e-01 -1.06741607e+00
-4.30707067e-01 -1.39025360e-01 1.10996880e-01 2.56369770e-01
-1.00509319e-02 2.34209877e-02 1.26746982e-01 -8.02574933e-01
2.77731027e-02 -7.21652806e-01 -1.80487245e-01 -5.43731034e-01
-9.05218780e-01 -3.13441098e-01 1.22479424e-01 -6.36275470e-01
1.06252360e+00 -1.69952917e+00 2.76689082e-01 2.72569004e-02
-1.19313382e-01 7.76584893e-02 -2.81152993e-01 3.33707541e-01
-1.03493139e-01 5.96385002e-01 -6.31081283e-01 -1.76354721e-01
3.77488226e-01 7.05811381e-02 -7.48209774e-01 -3.30317795e-01
6.57738626e-01 1.22067297e+00 -1.07610583e+00 -4.12235707e-01
7.07453769e-03 2.27444172e-01 -9.39134181e-01 5.49758263e-02
-9.04110610e-01 1.27165213e-01 -1.28263757e-01 2.92565137e-01
1.50445297e-01 -1.29926488e-01 2.97041327e-01 -5.10323755e-02
4.32015210e-01 9.62316453e-01 -6.60250664e-01 1.75591624e+00
-8.09589386e-01 5.00283182e-01 -5.85545838e-01 -7.52843738e-01
7.13802397e-01 3.01356792e-01 -2.21867815e-01 -6.72225237e-01
9.71410722e-02 1.94576979e-01 4.17571336e-01 -2.98287362e-01
4.95687783e-01 -8.58017683e-01 -2.04529777e-01 6.68309510e-01
2.98085451e-01 -7.34185278e-01 5.30916989e-01 5.23789823e-01
9.81297791e-01 1.71306193e-01 5.30744851e-01 -1.51592493e-01
4.52022880e-01 1.59119591e-01 3.53617728e-01 7.51586139e-01
1.63868517e-01 5.34131408e-01 5.20790815e-01 9.09591541e-02
-9.95410264e-01 -1.11799943e+00 1.74697414e-01 9.68768835e-01
-1.16206139e-01 -5.98662436e-01 -9.24831331e-01 -5.74118018e-01
-2.84713775e-01 1.90773511e+00 -5.01089096e-01 -4.92041916e-01
-4.09154505e-01 -5.34287512e-01 8.62823904e-01 4.70194757e-01
2.83962935e-01 -1.64836299e+00 -7.61313081e-01 4.02491003e-01
-7.11406589e-01 -1.39277375e+00 -3.15519154e-01 -1.87195614e-02
-5.11392832e-01 -1.08885539e+00 -2.77865857e-01 -6.33521914e-01
6.24725401e-01 -3.05903643e-01 1.58439326e+00 2.24357054e-01
-1.86156869e-01 3.41233164e-02 -4.98379588e-01 -6.48323715e-01
-1.01277316e+00 1.49377629e-01 -2.29778498e-01 -4.25427616e-01
4.97062951e-01 -5.23005843e-01 -5.96659593e-02 -3.06868196e-01
-1.12026024e+00 4.93592829e-01 4.38055485e-01 8.11429083e-01
3.80289018e-01 -7.05571696e-02 9.07890022e-01 -1.03574634e+00
1.19983685e+00 -3.81625295e-01 -7.48139247e-02 3.20695579e-01
-5.23505747e-01 3.49523723e-01 7.74945438e-01 -1.96654886e-01
-1.33352888e+00 -1.62664697e-01 -2.90418029e-01 1.79773450e-01
-2.10230634e-01 6.40435874e-01 -3.01234335e-01 7.78145790e-01
1.10270119e+00 4.78008151e-01 -2.13389993e-01 1.68479949e-01
8.94880831e-01 4.53696579e-01 6.36432409e-01 -9.78727162e-01
7.35822439e-01 1.14490770e-01 -4.00876433e-01 -5.82739115e-01
-1.45454574e+00 1.69960188e-03 -4.86870438e-01 2.22677901e-01
1.09015954e+00 -8.16290081e-01 -2.53026515e-01 5.33818491e-02
-1.58668137e+00 -4.34867263e-01 -4.90497351e-01 1.66167751e-01
-7.94626176e-01 1.09320655e-01 -4.13478106e-01 -6.75330758e-01
-6.41088068e-01 -8.53767693e-01 1.05745196e+00 -1.69635247e-02
-9.54982340e-01 -9.25712049e-01 -4.37664203e-02 4.28522974e-01
4.32726085e-01 1.85589179e-01 1.38363767e+00 -8.53885472e-01
-2.74884582e-01 1.29737109e-01 -1.08896099e-01 5.48147500e-01
7.04548135e-02 -1.03135049e-01 -9.21267331e-01 2.55794942e-01
-8.97586867e-02 -6.32132292e-01 9.43607032e-01 1.77223131e-01
9.46567774e-01 -5.26088297e-01 -1.93301812e-01 2.54357040e-01
1.17024326e+00 8.02743882e-02 6.70162916e-01 7.51808435e-02
3.79816711e-01 7.82437325e-01 2.91467488e-01 1.68899953e-01
5.86265624e-01 3.65601182e-01 1.28065869e-01 2.78364182e-01
-2.66770095e-01 -6.50968790e-01 4.49820995e-01 5.17397344e-01
2.57661253e-01 -4.16322947e-01 -9.46891606e-01 8.27879608e-01
-1.64037514e+00 -1.39119875e+00 4.32440080e-02 1.76364768e+00
1.54400373e+00 4.73790407e-01 -3.94550085e-01 9.14978981e-02
6.59002662e-01 7.61688873e-02 -4.36189771e-01 -5.86203575e-01
-3.21903795e-01 6.23115599e-01 -1.78272605e-01 7.50665486e-01
-6.73345208e-01 1.50436509e+00 5.78664684e+00 6.87226355e-01
-8.23510230e-01 1.18389003e-01 4.83880550e-01 -7.01628998e-02
-9.61454690e-01 2.52684385e-01 -7.85410643e-01 3.10132205e-01
9.53092635e-01 -5.57655275e-01 5.20104349e-01 6.63026452e-01
6.21636547e-02 7.28727207e-02 -1.42744553e+00 6.99538350e-01
5.41430950e-01 -1.45239103e+00 5.93732536e-01 -3.57774526e-01
8.79698217e-01 -1.80843502e-01 -2.80980378e-01 6.95594430e-01
5.67540467e-01 -1.32211828e+00 1.01304436e+00 2.37144887e-01
8.10486376e-01 -6.26559258e-01 5.56105435e-01 5.42184293e-01
-7.22943544e-01 1.95433110e-01 -3.43937486e-01 -2.09966063e-01
4.72990155e-01 7.17715859e-01 -9.31823432e-01 2.82659322e-01
1.48775682e-01 3.27353030e-01 -5.50893605e-01 4.17524546e-01
-9.85140383e-01 6.11352921e-01 -1.17726117e-01 -3.30612630e-01
1.33727372e-01 2.66575664e-01 3.94425482e-01 1.36374950e+00
3.24077129e-01 2.79410452e-01 -1.29015902e-02 1.66568911e+00
-3.96660924e-01 5.66694513e-02 -7.42599428e-01 -2.73897678e-01
5.14267206e-01 8.96370769e-01 -4.22999650e-01 -6.53343797e-01
-9.29356739e-02 1.09995675e+00 4.86589700e-01 2.14677170e-01
-7.04606414e-01 -4.24268395e-01 6.24054193e-01 -8.97895545e-02
-7.71162659e-02 5.61407991e-02 -5.77106059e-01 -1.32551742e+00
1.59836739e-01 -9.71106410e-01 2.12968349e-01 -1.14707375e+00
-1.37294745e+00 6.75164580e-01 -7.17513040e-02 -6.10235989e-01
-7.42279589e-01 -4.37371761e-01 -8.53298664e-01 1.07473218e+00
-1.42888343e+00 -1.15090978e+00 -4.82793190e-02 3.56204122e-01
8.81794930e-01 -4.32678573e-02 1.06369948e+00 -3.69864225e-01
-1.74269155e-01 4.31442708e-01 -7.32293367e-01 -1.83658786e-02
4.31861848e-01 -1.26050270e+00 8.92474115e-01 1.06776500e+00
3.52055371e-01 1.00967169e+00 9.56182897e-01 -9.99324977e-01
-7.33397484e-01 -1.20649838e+00 1.46103191e+00 -6.54414892e-01
6.59347594e-01 -3.76392543e-01 -8.84120882e-01 8.92629385e-01
4.05455768e-01 -4.98852551e-01 7.95646369e-01 2.31898129e-01
-7.07035542e-01 4.93624866e-01 -1.04296398e+00 9.72310305e-01
1.26794958e+00 -6.55279517e-01 -1.24165511e+00 5.60696423e-01
1.07622826e+00 -2.93129444e-01 -2.68229872e-01 2.04335898e-01
1.58418804e-01 -7.04992831e-01 7.00261414e-01 -1.13653147e+00
1.03681576e+00 -2.47233868e-01 -3.45599294e-01 -1.87392092e+00
-2.10574076e-01 -6.38166606e-01 1.13819905e-01 1.22367358e+00
9.87854540e-01 -2.88277924e-01 3.52814704e-01 6.52203679e-01
-3.07702601e-01 -5.29560387e-01 -7.62802064e-01 -7.40848124e-01
3.69836718e-01 -7.75448024e-01 6.56183362e-01 8.82207930e-01
4.88469213e-01 1.12246847e+00 6.02158811e-03 -1.90431580e-01
4.74919617e-01 3.55564244e-02 3.42793822e-01 -9.77234364e-01
-3.39836448e-01 -7.00616121e-01 -4.24645282e-02 -6.08681142e-01
7.98978806e-01 -1.38789046e+00 4.11394149e-01 -1.90927434e+00
3.43730390e-01 -1.03011437e-01 1.89482123e-01 6.97262824e-01
-5.27902782e-01 -3.21140401e-02 4.31771576e-01 -2.64864445e-01
-2.93755442e-01 7.23690450e-01 1.18211901e+00 -2.95083970e-01
-1.36355720e-02 -4.07551050e-01 -1.40726709e+00 6.99194014e-01
9.62842941e-01 -3.34202677e-01 -6.67553604e-01 -5.47447443e-01
6.11657500e-01 -1.60090983e-01 5.07611752e-01 -8.98785949e-01
5.65867918e-03 -3.58963847e-01 3.04096878e-01 -1.11788943e-01
1.44220680e-01 -2.56785810e-01 -2.08892316e-01 3.73492122e-01
-9.05432284e-01 3.28016765e-02 2.96210498e-01 2.84292758e-01
-1.59161121e-01 -4.51175064e-01 6.98018670e-01 -3.84289026e-01
-5.81462562e-01 -2.89012015e-01 -3.30970258e-01 8.68321478e-01
8.60777140e-01 4.41469885e-02 -4.98172909e-01 -4.42801565e-01
-5.93417525e-01 5.31623652e-03 2.97227770e-01 5.94424784e-01
7.53800929e-01 -1.11553228e+00 -1.23978341e+00 5.76381804e-03
4.03046310e-01 -7.54998922e-02 1.08164847e-01 1.47259861e-01
-2.11397931e-01 6.57299101e-01 -5.47715686e-02 -1.15858115e-01
-7.37534225e-01 4.68073428e-01 3.22967082e-01 -2.45136768e-01
-3.83913040e-01 8.95656645e-01 1.84762895e-01 -7.45447338e-01
-1.47438005e-01 -5.33920944e-01 1.21627972e-02 -2.92254061e-01
5.67519903e-01 -2.80562431e-01 8.18204507e-02 -4.66687441e-01
-3.97411138e-01 1.51759371e-01 3.30745615e-02 -4.16131467e-01
9.50623155e-01 1.51157334e-01 -5.33416979e-02 1.55081823e-01
6.71931982e-01 7.17787668e-02 -7.33229995e-01 -4.82639894e-02
2.60558147e-02 -1.26478598e-02 -2.45872781e-01 -1.34534526e+00
-4.82873082e-01 8.92529309e-01 -3.04554254e-01 9.63730924e-03
8.35578024e-01 1.46798193e-01 9.72258627e-01 5.41382492e-01
1.81648210e-01 -1.06110585e+00 1.35340318e-01 9.02717650e-01
1.20679092e+00 -1.10450232e+00 -3.63583237e-01 -5.13318837e-01
-9.40645933e-01 7.47614503e-01 8.65734279e-01 5.83942756e-02
1.05786845e-01 2.34308317e-02 -1.18817560e-01 -1.41272232e-01
-1.14725792e+00 -4.40368056e-01 2.69583583e-01 8.16666782e-01
8.21950436e-01 2.43745327e-01 -3.53445590e-01 8.71613979e-01
-9.76672232e-01 -1.65842742e-01 7.12242365e-01 5.45719802e-01
-6.14088595e-01 -9.10650134e-01 -1.09515212e-01 5.43956339e-01
-3.35430592e-01 -8.69990051e-01 -7.65160739e-01 5.64371943e-01
7.80868754e-02 1.15904510e+00 -1.14902824e-01 -9.07905772e-02
3.87556881e-01 4.48209733e-01 5.99480569e-01 -1.27906990e+00
-5.07544279e-01 -5.20578146e-01 4.97547537e-01 -3.14935148e-01
-1.11632757e-02 -5.30740976e-01 -1.73084366e+00 -9.88955870e-02
-1.69664487e-01 5.88403009e-02 5.10104179e-01 1.39025784e+00
2.40586430e-01 7.18742967e-01 -6.09908067e-02 -4.23809886e-01
-7.98410475e-01 -1.14816713e+00 -3.45827341e-02 7.06642985e-01
-1.29759114e-03 -2.43318141e-01 -3.19197387e-01 3.89128834e-01]
|
[11.2650728225708, 8.849175453186035]
|
2e68935b-16ea-4853-9bdc-c14562998ea9
|
centam-creation-and-validation-of-a-new
| null | null |
https://aclanthology.org/2020.bucc-1.10
|
https://aclanthology.org/2020.bucc-1.10.pdf
|
cEnTam: Creation and Validation of a New English-Tamil Bilingual Corpus
|
Natural Language Processing (NLP), is the field of artificial intelligence that gives the computer the ability to interpret, perceive and extract appropriate information from human languages. Contemporary NLP is predominantly a data driven process. It employs machine learning and statistical algorithms to learn language structures from textual corpus. While application of NLP in English, certain European languages such as Spanish, German, etc. and Chinese, Arabic has been tremendous, it is not so, in many Indian languages. There are obvious advantages in creating aligned bilingual and multilingual corpora. Machine translation, cross-lingual information retrieval, content availability and linguistic comparison are a few of the most sought after applications of such parallel corpora. This paper explains and validates a parallel corpus we created for English-Tamil bilingual pair.
|
['Soman Kp', 'Premjith B', 'Sanjanasri JP', 'Vijay Krishna Menon']
|
2020-05-01
| null | null | null |
lrec-2020-5
|
['cross-lingual-information-retrieval']
|
['natural-language-processing']
|
[ 1.02891788e-01 -5.48450984e-02 -3.75120103e-01 -3.08976233e-01
-6.84136629e-01 -1.05663598e+00 9.26108956e-01 6.95015132e-01
-6.64460957e-01 1.29269743e+00 3.16923410e-01 -7.65268683e-01
-5.79743795e-02 -4.77847725e-01 -2.33650312e-01 -2.37589896e-01
1.64241150e-01 9.63912010e-01 2.45605726e-02 -5.74261606e-01
6.38907671e-01 7.34603465e-01 -9.02164936e-01 3.34932029e-01
9.43148375e-01 3.16258609e-01 5.05125940e-01 3.81750673e-01
-8.49525571e-01 7.87994564e-01 -1.22706980e-01 -4.86147016e-01
2.46168822e-01 -4.12934721e-01 -1.26438022e+00 -5.48002161e-02
-3.25672865e-01 3.08115274e-01 3.77701670e-01 9.98652577e-01
1.59152403e-01 -3.36913854e-01 6.86718702e-01 -9.13403332e-01
-5.94003201e-01 5.96488774e-01 -7.27013290e-01 2.88929969e-01
7.81258941e-01 -1.87749371e-01 6.44603431e-01 -1.12895048e+00
9.60040808e-01 1.44200659e+00 3.85703743e-01 4.19424064e-02
-8.74045134e-01 -3.99369121e-01 -4.66060281e-01 -1.13248937e-01
-1.28713346e+00 -1.51576668e-01 5.34019649e-01 -4.59082007e-01
1.16767251e+00 -6.36272281e-02 3.68319720e-01 4.87928391e-01
6.42148674e-01 5.28846025e-01 1.59381342e+00 -1.30807924e+00
-2.02655137e-01 7.65121758e-01 1.15776680e-01 3.84545445e-01
1.05524622e-01 -2.60869768e-02 -5.16671062e-01 -1.46970851e-02
2.93645740e-01 -2.47707784e-01 1.28228456e-01 3.13667923e-01
-1.41549289e+00 8.03552210e-01 -3.12032849e-01 1.03818369e+00
-6.41022146e-01 -6.70619130e-01 7.74880946e-01 6.11338615e-01
3.04013759e-01 4.30412531e-01 -8.54260266e-01 -4.71243560e-01
-8.61344278e-01 -5.09691350e-02 1.07065439e+00 1.04849279e+00
7.91502357e-01 -1.45830870e-01 5.82547724e-01 8.40289652e-01
4.21418369e-01 8.21338415e-01 7.95530558e-01 -4.03796971e-01
5.92440069e-01 9.16062176e-01 1.02526871e-02 -1.08988881e+00
-3.65415961e-01 2.72493631e-01 -5.19398272e-01 -1.19073600e-01
3.44165146e-01 -3.47074717e-01 -5.92008471e-01 1.20720851e+00
2.14915454e-01 -9.02295172e-01 6.69131637e-01 5.65927804e-01
5.14848351e-01 9.97372746e-01 2.37193137e-01 -5.59709370e-01
1.59003031e+00 -6.22545600e-01 -7.03188360e-01 -4.29776132e-01
3.34062546e-01 -1.54257178e+00 9.74268794e-01 4.35517997e-01
-1.03347433e+00 -3.45432490e-01 -7.37764657e-01 -2.40607828e-01
-7.75655091e-01 8.56255740e-03 5.19343197e-01 5.20554006e-01
-9.01613593e-01 2.04012588e-01 -4.69066978e-01 -9.73715246e-01
-2.22174942e-01 4.22280967e-01 -8.15754890e-01 -1.06937811e-01
-1.23966014e+00 1.22234154e+00 9.55183744e-01 -2.35823035e-01
-1.13208286e-01 -7.50631606e-03 -7.78413773e-01 -4.09790635e-01
2.58297712e-01 -1.43443629e-01 1.09471917e+00 -1.33698440e+00
-1.17165315e+00 1.42683756e+00 -5.52509069e-01 -4.61071074e-01
1.29761426e-02 -1.85745712e-02 -8.22820306e-01 1.62359968e-01
4.20807749e-01 3.94689441e-01 3.05307299e-01 -9.62579787e-01
-9.00532842e-01 -4.97491717e-01 -5.00694454e-01 1.90235734e-01
1.63469300e-01 8.47853661e-01 -2.56648958e-01 -5.62612653e-01
1.86323330e-01 -7.54716218e-01 -1.79036856e-01 -5.78188360e-01
-3.60729210e-02 -2.88619041e-01 6.03433490e-01 -1.15910935e+00
8.99789035e-01 -2.00114870e+00 -2.37043485e-01 3.42009634e-01
-2.05147326e-01 9.16147083e-02 1.85916796e-01 1.30478382e+00
1.02202944e-01 2.26993397e-01 -1.28773004e-01 4.78918761e-01
-1.33426890e-01 5.81253946e-01 -1.64567962e-01 1.44666806e-01
3.80230516e-01 6.15886807e-01 -8.39171946e-01 -1.00441623e+00
1.26020685e-01 1.82707265e-01 1.71631463e-02 -2.43798926e-01
-4.50470485e-02 5.27219653e-01 -6.61437094e-01 8.13794971e-01
3.64940137e-01 2.04960555e-01 4.14535224e-01 1.67833760e-01
-7.43096411e-01 4.84961450e-01 -9.12625253e-01 1.43461215e+00
-4.49232846e-01 9.61033821e-01 -5.34041263e-02 -1.05265260e+00
1.19980419e+00 7.22289324e-01 3.45671654e-01 -9.80090201e-01
1.46677420e-01 6.45132184e-01 7.29021356e-02 -9.57861722e-01
6.27390504e-01 -5.59141636e-01 -2.45066971e-01 4.65787470e-01
5.89583591e-02 -2.29244724e-01 6.83908284e-01 -2.09510457e-02
2.93662161e-01 1.84481621e-01 1.08862591e+00 -6.40695930e-01
9.59398985e-01 7.31397569e-01 6.57280505e-01 1.70573875e-01
-7.35732913e-02 1.55446723e-01 2.30120465e-01 -6.59358203e-01
-1.34023845e+00 -8.07433128e-01 -3.02656293e-01 8.60065222e-01
-1.70067802e-01 -1.57358065e-01 -4.82491076e-01 -2.43139669e-01
-4.69210595e-01 7.27924109e-01 1.35357410e-01 5.55361986e-01
-6.62672758e-01 -4.68396962e-01 3.28301251e-01 -7.93755800e-02
5.32247663e-01 -1.51602197e+00 -2.97891319e-01 5.15904546e-01
-1.80935413e-01 -1.42591977e+00 -1.25403434e-01 1.19913407e-01
-7.98873305e-01 -1.05119908e+00 -2.66222388e-01 -1.48161232e+00
6.18110716e-01 -7.62586892e-02 1.14660859e+00 -3.04113895e-01
-9.93236601e-02 1.41560897e-01 -5.58575749e-01 -9.77094293e-01
-9.93999720e-01 5.98776080e-02 3.53286445e-01 -4.06788200e-01
1.23937523e+00 -4.71411675e-01 9.43508744e-02 -1.86032541e-02
-9.10412610e-01 1.15246691e-01 8.61129284e-01 5.48649371e-01
5.42918980e-01 3.21844459e-01 6.01390302e-01 -1.04404771e+00
8.89050305e-01 -3.25217158e-01 -3.25633675e-01 4.94369298e-01
-6.15610957e-01 1.32006064e-01 7.20218778e-01 -1.03635080e-01
-1.15818620e+00 1.13369241e-01 -1.41025826e-01 6.60957754e-01
-5.87665796e-01 9.39995646e-01 -1.02479033e-01 6.11149482e-02
8.73074889e-01 5.09239316e-01 -1.77773014e-01 -4.14212465e-01
1.86104789e-01 1.08719146e+00 5.12322664e-01 -7.62173951e-01
7.81431556e-01 5.02985567e-02 -2.06161857e-01 -1.15804684e+00
-1.67348042e-01 -6.77185655e-01 -1.26264131e+00 -7.99028575e-02
7.61315048e-01 -7.89446414e-01 -1.48813814e-01 -5.95557783e-03
-1.26907468e+00 1.50835022e-01 -1.32050782e-01 8.51226211e-01
-1.62013710e-01 1.54960468e-01 -3.95488024e-01 -8.32197666e-01
-6.04635179e-01 -8.45327973e-01 6.93561018e-01 3.98596317e-01
-4.56560016e-01 -1.21337914e+00 2.71775007e-01 2.76373923e-01
1.49961039e-01 2.64218837e-01 1.12023306e+00 -1.01285350e+00
9.20806825e-02 -3.29817444e-01 -2.00383246e-01 2.16016829e-01
5.59416831e-01 1.41683489e-01 -3.72351348e-01 -1.13105855e-03
1.38040379e-01 -3.24707896e-01 3.31002730e-03 -1.10047586e-01
-1.64555937e-01 -5.79968929e-01 -2.44392350e-01 -2.23441646e-01
1.70729017e+00 9.54170167e-01 5.15445590e-01 5.55201411e-01
2.53640354e-01 1.05956638e+00 7.74356246e-01 -7.95925502e-03
4.03983623e-01 9.31757167e-02 -6.32374585e-01 -7.72106871e-02
3.52127314e-01 -8.33084881e-02 4.28687841e-01 1.44421387e+00
5.47019579e-02 1.54500037e-01 -1.65745807e+00 8.72875929e-01
-1.44633055e+00 -7.14336991e-01 -2.17108414e-01 1.88999212e+00
1.25566435e+00 2.56375343e-01 1.61189947e-03 1.64520487e-01
5.33289254e-01 -3.80003273e-01 7.42722675e-02 -1.04120648e+00
-4.33488697e-01 3.57439160e-01 3.79164189e-01 6.30596280e-01
-7.47770250e-01 1.28195620e+00 5.89913082e+00 6.37111008e-01
-1.21932065e+00 -4.03394960e-02 4.27167773e-01 6.39670730e-01
-1.24619111e-01 3.22811961e-01 -6.37684643e-01 2.68424392e-01
1.00213802e+00 -6.20506227e-01 5.20824075e-01 4.72815633e-01
7.05368936e-01 -3.57584327e-01 -9.62387204e-01 9.03299153e-01
7.94221833e-02 -1.03075850e+00 1.63121924e-01 -1.37757331e-01
8.27335536e-01 2.95056313e-01 -4.95420665e-01 3.78579013e-02
2.87025988e-01 -1.04325426e+00 5.11140466e-01 1.69508740e-01
3.54830444e-01 -9.02909875e-01 8.95445228e-01 6.17546916e-01
-8.55991244e-01 2.67898887e-01 -3.80549282e-01 -2.28891358e-01
2.19557494e-01 3.44078630e-01 -1.07954168e+00 6.22000635e-01
5.54573715e-01 2.81559855e-01 -4.89456624e-01 5.38368106e-01
-2.90384233e-01 5.96297503e-01 -3.14008862e-01 -2.66618133e-01
5.97471356e-01 -6.43086255e-01 4.89294708e-01 1.46750367e+00
2.26498768e-01 1.65348053e-01 2.85688102e-01 1.63461119e-01
2.66359955e-01 1.07961583e+00 -9.79407430e-01 -4.79926437e-01
2.60284364e-01 9.77269053e-01 -9.08823371e-01 -3.99984986e-01
-7.05090582e-01 7.24607527e-01 -2.48643067e-02 3.05176407e-01
-9.94652137e-03 -4.18675721e-01 -7.99333304e-02 1.62010238e-01
-3.16335589e-01 -4.80937511e-01 -4.04549688e-01 -9.28735793e-01
1.06370911e-01 -1.48458874e+00 3.55752230e-01 -6.35007501e-01
-1.26259315e+00 8.69521618e-01 1.35244891e-01 -1.08753216e+00
-7.53333211e-01 -8.89026999e-01 -3.03884476e-01 1.07531190e+00
-1.12553251e+00 -1.23185182e+00 4.48332667e-01 5.96256137e-01
7.89225996e-01 -7.91150331e-01 8.97674143e-01 1.72223434e-01
-3.71890999e-02 -1.46380842e-01 2.83185035e-01 3.38954568e-01
7.12727368e-01 -1.11068714e+00 1.95189267e-01 7.96076417e-01
4.31782812e-01 7.70318806e-01 8.79749656e-01 -6.38828993e-01
-1.20109010e+00 -4.34800804e-01 1.96710598e+00 -7.86413923e-02
9.43947971e-01 -1.18704267e-01 -5.93105495e-01 6.49514616e-01
9.52497423e-01 -7.80230820e-01 8.15045476e-01 -1.35650948e-01
4.97615784e-02 -1.29538074e-01 -1.18165898e+00 6.62737250e-01
1.09197035e-01 -7.68316507e-01 -1.14019966e+00 6.69053853e-01
3.78100991e-01 7.12063909e-03 -8.68158698e-01 -4.32570167e-02
4.51920152e-01 -5.84268570e-01 6.30446196e-01 -6.15898550e-01
3.65452707e-01 -4.41114545e-01 -7.65623599e-02 -1.04791009e+00
5.14277332e-02 -6.60080254e-01 1.07136643e+00 1.50559616e+00
1.02359009e+00 -8.26468110e-01 4.82743949e-01 7.18922496e-01
2.40743071e-01 -1.45034298e-01 -6.36110902e-01 -6.29475594e-01
3.32283348e-01 -4.68005478e-01 4.15364504e-01 1.36407232e+00
5.83652914e-01 7.59176850e-01 -1.48276553e-01 -1.00250997e-01
3.11350286e-01 1.45008355e-01 5.50078690e-01 -1.26416504e+00
2.05464177e-02 -1.70229465e-01 -2.37082586e-01 -4.01761472e-01
7.39058852e-02 -9.17482793e-01 -1.30603895e-01 -1.68694019e+00
6.12918660e-02 -1.69215858e-01 1.88218743e-01 5.51310718e-01
1.85823277e-01 -1.68812356e-03 7.61491731e-02 5.41862726e-01
-1.02986237e-02 -8.75351056e-02 1.14872444e+00 -2.87295412e-02
-4.53598797e-01 -2.97646195e-01 -5.60896933e-01 7.38295794e-01
1.25389636e+00 -7.96176136e-01 -1.92622676e-01 -4.47855711e-01
4.92380679e-01 5.00617251e-02 -4.29927945e-01 -5.60745478e-01
3.06744516e-01 -6.09627962e-01 2.55451947e-01 -4.93174255e-01
-3.74498814e-01 -1.08455575e+00 1.78895265e-01 4.35897976e-01
-5.77428788e-02 7.20272958e-01 3.71508211e-01 -3.60811651e-02
-7.19148695e-01 -4.23073351e-01 5.67770302e-01 -5.03534794e-01
-1.07527840e+00 -1.12466000e-01 -8.34261656e-01 7.86631554e-02
1.05571854e+00 -3.35613579e-01 7.88131058e-02 -1.29346997e-01
-2.70869553e-01 1.95957288e-01 4.73690540e-01 5.07779717e-01
3.30800265e-01 -1.22275162e+00 -9.37381506e-01 1.08160228e-01
1.12011053e-01 -1.09325491e-01 -4.49207187e-01 7.32727528e-01
-1.32370734e+00 9.14445698e-01 -5.59641242e-01 -1.74501717e-01
-1.19985139e+00 7.33814418e-01 -2.61289209e-01 -4.46294308e-01
-2.98808485e-01 7.67251402e-02 -1.56021401e-01 -6.47095919e-01
-3.36877823e-01 4.53912541e-02 -5.81846118e-01 -2.05528602e-01
4.00330603e-01 -1.12091161e-01 -1.18077390e-01 -1.26985312e+00
-5.46300173e-01 6.01602256e-01 -1.49191380e-01 -7.11511433e-01
1.15817690e+00 -3.91868055e-01 -7.17545509e-01 6.73559129e-01
1.00817251e+00 4.10768121e-01 -6.64772689e-02 -3.04370672e-01
7.91068494e-01 -2.29340136e-01 -3.03310633e-01 -9.49959338e-01
-3.90371323e-01 6.87763691e-01 2.84078717e-01 2.56984144e-01
1.15684903e+00 -5.77604286e-02 6.92201555e-01 6.09411716e-01
6.80990517e-01 -1.52529669e+00 -7.09835589e-01 7.26298153e-01
8.25745225e-01 -1.47206008e+00 -3.56465727e-02 -2.15888098e-01
-7.61356831e-01 1.39149475e+00 5.85498400e-02 2.19263434e-01
7.22339869e-01 4.29940730e-01 5.57890415e-01 -2.24707320e-01
-5.16881585e-01 3.90026672e-03 1.75256103e-01 5.64337194e-01
1.03698766e+00 -1.43616097e-02 -1.27146888e+00 2.53880680e-01
-5.34108341e-01 6.55998513e-02 2.89952457e-01 1.31496334e+00
-4.27011579e-01 -1.66243792e+00 -7.35013604e-01 1.13157965e-01
-1.06593502e+00 -3.88062000e-01 -7.70092070e-01 1.14349949e+00
1.91831842e-01 1.11179733e+00 -9.89866257e-02 7.49020875e-02
9.12967920e-02 2.25313485e-01 2.14507625e-01 -4.75234926e-01
-7.34248340e-01 1.48458391e-01 3.80878896e-01 1.43613264e-01
-8.91177416e-01 -7.52687871e-01 -1.37288415e+00 -3.83841157e-01
1.16514944e-01 7.93534458e-01 9.41527426e-01 1.38033032e+00
-9.32028666e-02 -2.76486427e-01 5.71047366e-01 -3.46736968e-01
1.57794073e-01 -9.98097777e-01 -5.46695411e-01 2.79678732e-01
-1.10026829e-01 8.48116875e-02 1.17388777e-02 6.00998998e-01]
|
[10.610873222351074, 10.025845527648926]
|
723f1334-0d76-40c0-99ad-86f3a5947cf5
|
challenges-in-clinical-natural-language
| null | null |
https://www.sciencedirect.com/science/article/pii/S1532046415001501?via%3Dihub
|
https://www.sciencedirect.com/science/article/pii/S1532046415001501/pdfft?md5=0f078fadd8924b8ec74b9a861e96863f&pid=1-s2.0-S1532046415001501-main.pdf
|
Challenges in clinical natural language processing for automated disorder normalization
|
Background
Identifying key variables such as disorders within the clinical narratives in electronic health records has wide-ranging applications within clinical practice and biomedical research. Previous research has demonstrated reduced performance of disorder named entity recognition (NER) and normalization (or grounding) in clinical narratives than in biomedical publications. In this work, we aim to identify the cause for this performance difference and introduce general solutions.
Methods
We use closure properties to compare the richness of the vocabulary in clinical narrative text to biomedical publications. We approach both disorder NER and normalization using machine learning methodologies. Our NER methodology is based on linear-chain conditional random fields with a rich feature approach, and we introduce several improvements to enhance the lexical knowledge of the NER system. Our normalization method – never previously applied to clinical data – uses pairwise learning to rank to automatically learn term variation directly from the training data.
Results
We find that while the size of the overall vocabulary is similar between clinical narrative and biomedical publications, clinical narrative uses a richer terminology to describe disorders than publications. We apply our system, DNorm-C, to locate disorder mentions and in the clinical narratives from the recent ShARe/CLEF eHealth Task. For NER (strict span-only), our system achieves precision = 0.797, recall = 0.713, f-score = 0.753. For the normalization task (strict span + concept) it achieves precision = 0.712, recall = 0.637, f-score = 0.672. The improvements described in this article increase the NER f-score by 0.039 and the normalization f-score by 0.036. We also describe a high recall version of the NER, which increases the normalization recall to as high as 0.744, albeit with reduced precision.
Discussion
We perform an error analysis, demonstrating that NER errors outnumber normalization errors by more than 4-to-1. Abbreviations and acronyms are found to be frequent causes of error, in addition to the mentions the annotators were not able to identify within the scope of the controlled vocabulary.
Conclusion
Disorder mentions in text from clinical narratives use a rich vocabulary that results in high term variation, which we believe to be one of the primary causes of reduced performance in clinical narrative. We show that pairwise learning to rank offers high performance in this context, and introduce several lexical enhancements – generalizable to other clinical NER tasks – that improve the ability of the NER system to handle this variation. DNorm-C is a high performing, open source system for disorders in clinical text, and a promising step toward NER and normalization methods that are trainable to a wide variety of domains and entities. (DNorm-C is open source software, and is available with a trained model at the DNorm demonstration website: http://www.ncbi.nlm.nih.gov/CBBresearch/Lu/Demo/tmTools/#DNorm.)
|
['Robert Leaman', 'Zhiyong Lu', 'Ritu Khare']
|
2015-07-14
| null | null | null |
journal-of-biomedical-informatics-2015-7
|
['medical-named-entity-recognition']
|
['natural-language-processing']
|
[ 1.12379879e-01 2.20633537e-01 -2.36746117e-01 -1.92269936e-01
-1.26146817e+00 -6.83102787e-01 3.29380184e-01 8.37584376e-01
-9.39742923e-01 9.21046436e-01 7.14799047e-01 -2.96448022e-01
-5.13735771e-01 -5.85852146e-01 -3.53591084e-01 -3.77956152e-01
1.36253029e-01 5.12998283e-01 -1.91795796e-01 -2.12981939e-01
1.21338047e-01 3.20872068e-01 -9.41102266e-01 4.16475028e-01
6.63305163e-01 3.85967880e-01 -5.41312769e-02 6.93418026e-01
-2.77143866e-01 8.27422976e-01 -7.57470131e-01 -2.98125625e-01
-2.46055976e-01 -3.33314955e-01 -9.66725647e-01 -4.30264950e-01
1.74497306e-01 1.30777925e-01 -8.69921595e-03 6.57860339e-01
8.16840649e-01 -1.57766923e-01 8.01447093e-01 -6.42879665e-01
-4.63464439e-01 6.77583933e-01 -1.28809035e-01 1.71455920e-01
7.19069600e-01 -3.84266227e-02 8.37990046e-01 -3.86309415e-01
1.18271637e+00 7.40690708e-01 1.26150441e+00 7.32777297e-01
-1.34007001e+00 -7.81556666e-01 -2.61119485e-01 -2.84322798e-01
-1.62543714e+00 -4.82152164e-01 -1.32515728e-01 -7.01198101e-01
1.49823880e+00 4.90489841e-01 4.01847690e-01 9.23822165e-01
4.54113066e-01 -1.46510988e-01 7.80201316e-01 -5.31172395e-01
1.41614228e-01 2.80263841e-01 1.55046940e-01 4.55096215e-01
4.55119193e-01 -1.79352611e-01 -2.88276851e-01 -6.37996674e-01
4.37301338e-01 -2.47779377e-02 -3.25791597e-01 5.85078001e-01
-1.32030058e+00 5.60294092e-01 5.91283068e-02 9.15045261e-01
-3.85414362e-01 -1.91936433e-01 7.45275140e-01 1.62210584e-01
2.91296244e-01 1.01114225e+00 -7.91633189e-01 -3.55107635e-01
-1.17608464e+00 1.35419965e-01 1.20179379e+00 9.05241489e-01
3.18219662e-02 -5.64249575e-01 -2.78476477e-01 9.82336760e-01
-3.46407145e-02 3.21950078e-01 7.46712685e-01 -6.47936344e-01
2.83753395e-01 6.15700483e-01 -1.29993647e-01 -1.01657009e+00
-9.70623493e-01 -3.66046757e-01 -8.39715362e-01 -4.88062024e-01
3.54303092e-01 -1.94810733e-01 -8.88252318e-01 1.80249357e+00
-5.32358848e-02 -3.97002965e-01 2.37273812e-01 3.02191138e-01
1.12543488e+00 2.04545602e-01 7.79424667e-01 -3.94784272e-01
1.81099486e+00 -1.33824229e-01 -1.13546968e+00 2.22554117e-01
1.33852148e+00 -1.03342557e+00 6.16801262e-01 2.08975554e-01
-7.86929369e-01 6.50506690e-02 -7.91203499e-01 -6.31662458e-02
-6.04727387e-01 -6.15834668e-02 3.03317249e-01 6.66656673e-01
-8.80711675e-01 6.73583150e-01 -8.71178150e-01 -9.05303657e-01
3.62472564e-01 4.80530769e-01 -7.73219526e-01 -2.32308991e-02
-1.41548824e+00 1.29719138e+00 4.55038249e-01 -4.78021264e-01
1.48652289e-02 -1.17802715e+00 -8.43196392e-01 -1.35814235e-01
8.38951468e-02 -7.26813912e-01 1.07775307e+00 -3.35949302e-01
-6.83463454e-01 1.12420535e+00 -2.03699693e-01 -2.60005802e-01
3.12744170e-01 1.98735058e-01 -9.12141144e-01 1.39364213e-01
5.75418949e-01 3.94745231e-01 -3.38764071e-01 -7.03765273e-01
-6.09604537e-01 -2.18241379e-01 -3.90875101e-01 -1.13531299e-01
-3.00277412e-01 4.35181737e-01 -2.79005110e-01 -7.56446123e-01
-3.37682590e-02 -7.60800183e-01 -4.45712507e-01 -3.97150397e-01
-3.43197197e-01 -1.68033972e-01 -1.41657546e-01 -9.58699703e-01
1.84060860e+00 -1.86956954e+00 -4.61807907e-01 2.77416438e-01
5.27682602e-01 1.15075491e-01 2.48126715e-01 6.91574037e-01
-4.84276652e-01 8.16754162e-01 -3.67081940e-01 4.04152796e-02
-2.82937288e-01 2.49851525e-01 2.33076707e-01 6.18769884e-01
3.71561944e-01 7.47301221e-01 -9.05552626e-01 -6.43679678e-01
-1.78665131e-01 6.11139059e-01 -5.64621210e-01 -1.31624281e-01
2.70008862e-01 1.71677083e-01 -1.98782712e-01 5.48916578e-01
3.67685765e-01 -2.48175621e-01 3.63403261e-01 -3.29063654e-01
-3.41348261e-01 5.09338140e-01 -1.09275210e+00 1.56881118e+00
-2.75246501e-01 3.15114141e-01 -1.91290915e-01 -6.31553411e-01
6.94526136e-01 7.26860344e-01 8.96314740e-01 -4.01800334e-01
5.18678874e-02 4.55031455e-01 1.96790263e-01 -8.19975972e-01
2.88043529e-01 -6.39003515e-01 -1.99677497e-01 -3.16108987e-02
2.57575661e-01 1.10036202e-01 3.66145223e-01 3.75572175e-01
1.71516979e+00 -4.26501125e-01 9.10325289e-01 -3.79967213e-01
3.89178813e-01 3.86375070e-01 6.73536897e-01 7.03374743e-01
3.43357436e-02 7.01451540e-01 7.69737482e-01 8.65157414e-03
-9.68722224e-01 -5.90003967e-01 -9.08964157e-01 4.97988284e-01
-6.86971247e-01 -1.05335903e+00 -6.42656803e-01 -5.15212297e-01
-8.63522217e-02 5.82782626e-01 -6.76954746e-01 -7.55870938e-02
-3.51055264e-01 -1.05097950e+00 1.19007945e+00 4.19049233e-01
-1.19959950e-01 -9.60167944e-01 -3.78394365e-01 5.13332784e-01
-3.04993272e-01 -1.20568466e+00 -4.81522620e-01 3.68893743e-01
-6.28542840e-01 -1.25732398e+00 -7.36264586e-01 -6.39476538e-01
4.92988855e-01 -7.98150480e-01 1.12219799e+00 3.45418416e-02
-5.75927734e-01 5.95818162e-01 -4.06649172e-01 -4.29686308e-01
-7.18817472e-01 2.05060005e-01 4.77586538e-02 -7.13538647e-01
7.95105040e-01 -1.12263143e-01 -3.64067912e-01 -6.74492028e-03
-1.14032912e+00 -5.07307589e-01 7.29023457e-01 9.61865783e-01
5.81755519e-01 -1.65569827e-01 5.83754897e-01 -1.30124688e+00
6.96186423e-01 -7.44271159e-01 -4.69079502e-02 2.22658776e-02
-1.05424047e+00 8.15716237e-02 3.31761241e-01 -4.27440405e-01
-5.82925200e-01 -2.96305194e-02 -5.06411910e-01 2.59584308e-01
-3.85333061e-01 8.69950652e-01 6.23542964e-02 2.41846755e-01
1.04239297e+00 -4.32076842e-01 -4.00291607e-02 -5.66764653e-01
-5.44737466e-02 8.93267274e-01 4.49631035e-01 -3.06193829e-01
1.42651334e-01 1.43037200e-01 8.75685178e-03 -7.04830050e-01
-5.38179755e-01 -9.64732766e-01 -3.36981475e-01 4.73592073e-01
1.20916724e+00 -9.79641914e-01 -7.05028415e-01 -2.08547458e-01
-9.75816250e-01 -8.10579024e-03 -2.88598359e-01 8.62465262e-01
-1.52275592e-01 1.96870826e-02 -7.48593628e-01 -4.84192312e-01
-4.90821481e-01 -9.63891745e-01 8.87202501e-01 -8.61952305e-02
-1.03025126e+00 -1.12040687e+00 4.00508493e-01 4.76739556e-02
1.98352724e-01 5.59632063e-01 1.06640863e+00 -1.39177251e+00
5.55308938e-01 -3.45404297e-01 -1.48530215e-01 -1.54710114e-02
4.58164871e-01 -1.42663419e-01 -6.90571129e-01 -2.00233422e-02
-9.85839218e-02 3.39322597e-01 7.58574188e-01 2.88462996e-01
7.10271358e-01 -2.70081997e-01 -7.05484450e-01 2.87014723e-01
1.52374744e+00 3.03636014e-01 6.02570295e-01 5.32820165e-01
6.47949994e-01 6.33338928e-01 2.01083004e-01 2.06189960e-01
3.16669047e-01 5.49119055e-01 -3.62592578e-01 -2.77352810e-01
-1.05491027e-01 2.35431436e-02 8.30980018e-02 7.24759102e-01
-1.22686969e-02 -1.08875912e-02 -1.53410435e+00 8.07951927e-01
-1.40894508e+00 -7.85291791e-01 -3.82372200e-01 2.10699463e+00
1.32548368e+00 6.08370416e-02 1.08811632e-02 -1.78116169e-02
7.13548422e-01 -5.98371327e-01 -6.25996217e-02 -5.55348158e-01
-1.02910742e-01 5.74239016e-01 9.05931473e-01 3.84196579e-01
-7.94123828e-01 5.78766465e-01 6.12984085e+00 6.13057196e-01
-8.50947142e-01 1.00339517e-01 2.65239596e-01 -1.50014460e-01
-1.46392230e-02 -1.76155418e-01 -9.93970871e-01 4.32914108e-01
1.53299284e+00 -3.70143473e-01 -2.05744822e-02 3.25488865e-01
1.75926030e-01 8.44311565e-02 -1.26916659e+00 9.84471083e-01
7.06873760e-02 -1.38849604e+00 -1.25417978e-01 2.42498428e-01
5.64156592e-01 -1.12063438e-02 -2.85627961e-01 2.47930899e-01
8.30506533e-02 -1.30862927e+00 2.79613972e-01 7.00108349e-01
1.36559379e+00 -6.13352060e-01 1.30860877e+00 -7.29080662e-02
-8.76298189e-01 1.58407539e-01 -1.00670032e-01 1.66114688e-01
8.71699378e-02 8.08151901e-01 -1.28005433e+00 6.44713938e-01
7.13497281e-01 6.36587024e-01 -3.28654885e-01 9.52830672e-01
1.31221935e-01 5.45198441e-01 -2.24145785e-01 7.14613497e-02
7.92271197e-02 5.53091824e-01 4.75553095e-01 1.87024629e+00
2.55093277e-01 5.04329264e-01 9.94597748e-02 5.87048054e-01
-3.64522547e-01 6.48881972e-01 -3.34430218e-01 -3.73259366e-01
4.11474794e-01 1.16647589e+00 -6.90586030e-01 -3.74474287e-01
-3.92707825e-01 5.02587259e-01 -5.04985787e-02 -1.97322313e-02
-5.14608324e-01 -7.41560102e-01 5.17507195e-01 3.02527368e-01
8.31410289e-05 4.07733351e-01 -3.33465695e-01 -9.60562706e-01
-1.80263564e-01 -1.02465880e+00 7.70836353e-01 -3.65860462e-01
-1.53617954e+00 6.80412769e-01 -1.60149373e-02 -1.15848410e+00
-3.21519554e-01 -5.67435265e-01 8.42652023e-02 1.19236112e+00
-9.99456644e-01 -7.93872416e-01 1.63674414e-01 3.05881262e-01
2.08121892e-02 -6.69000205e-03 1.40472007e+00 6.02130413e-01
-3.46309602e-01 8.63766432e-01 1.22259788e-01 5.70690691e-01
1.43001735e+00 -1.28991914e+00 8.94160569e-02 1.62526309e-01
-2.50480264e-01 1.29436243e+00 5.38072050e-01 -1.05071402e+00
-6.65485740e-01 -9.92696583e-01 1.66306031e+00 -6.79564118e-01
7.25223899e-01 -4.15840745e-02 -1.03745651e+00 4.76333529e-01
1.83725338e-02 -3.23183745e-01 1.41858649e+00 3.59942675e-01
-3.64606619e-01 2.05864072e-01 -1.49424350e+00 4.22420233e-01
8.35918427e-01 -6.33068681e-01 -9.59493041e-01 4.27858293e-01
6.37594342e-01 -3.78384888e-01 -1.69734514e+00 3.56423676e-01
8.03101182e-01 -1.02153048e-01 6.89205766e-01 -8.02185178e-01
3.99271876e-01 -1.49723783e-01 -1.73217297e-01 -1.04470372e+00
-3.70886028e-01 -4.53892529e-01 4.70519006e-01 1.49736452e+00
8.56112897e-01 -7.45334268e-01 2.57934034e-01 9.87341285e-01
-3.98179516e-02 -6.04639828e-01 -8.55473638e-01 -4.65452939e-01
4.48723823e-01 -4.27279681e-01 3.74246687e-01 1.37069893e+00
6.42649651e-01 3.25746775e-01 2.32960612e-01 8.90750438e-02
1.67767167e-01 -7.31526196e-01 6.50121123e-02 -1.35424650e+00
2.54185423e-02 -3.74735951e-01 -5.80899239e-01 -6.54078126e-02
-1.56102866e-01 -1.22488141e+00 -1.49805278e-01 -1.79680431e+00
4.37578738e-01 -4.78652298e-01 -4.91137177e-01 8.62380624e-01
-6.11957796e-02 1.65563166e-01 -3.03290151e-02 2.86280692e-01
-2.67964363e-01 -4.50096339e-01 6.03105903e-01 -2.79397378e-03
-3.72953445e-01 -5.26368618e-01 -1.22632062e+00 6.06215119e-01
6.80190444e-01 -1.10501754e+00 2.02068865e-01 -7.79785262e-03
4.62684453e-01 -1.17874764e-01 5.94564527e-02 -7.93657839e-01
4.21594024e-01 2.06603602e-01 5.72127938e-01 -2.53833443e-01
-3.86096895e-01 -6.72818303e-01 3.79192770e-01 6.76282585e-01
-5.77735960e-01 1.58893496e-01 6.53798342e-01 2.50504911e-01
3.36538926e-02 -3.18230093e-01 4.18991417e-01 -4.51336920e-01
-2.33637094e-01 -1.40712351e-01 -7.01686919e-01 3.34759504e-01
4.62653279e-01 9.61095318e-02 -2.99791157e-01 1.89119503e-02
-1.19979167e+00 6.84216172e-02 3.62512052e-01 2.07680330e-01
4.19891961e-02 -1.00382006e+00 -8.83366525e-01 -2.05576107e-01
3.76450509e-01 -3.46421182e-01 1.99751079e-01 1.14246690e+00
-5.10295391e-01 6.54889643e-01 7.31867924e-02 -4.28254545e-01
-1.37335420e+00 4.35694784e-01 3.57064426e-01 -6.36349082e-01
-6.17522061e-01 5.35847068e-01 -2.64991224e-01 -4.60722029e-01
5.93459532e-02 -5.08401513e-01 -4.00426388e-01 4.31406736e-01
6.71504438e-01 3.53080392e-01 6.08671606e-01 -7.15184808e-01
-8.35483968e-01 5.15121460e-01 -3.31186056e-01 -2.65681952e-01
1.37803185e+00 1.53848916e-01 -2.82829762e-01 3.97932798e-01
1.35223365e+00 4.73300993e-01 8.92055333e-02 1.69472560e-01
3.91956031e-01 1.69655412e-01 4.73956317e-02 -1.34178948e+00
-4.17615533e-01 1.77022487e-01 6.10497057e-01 1.06504530e-01
1.06589901e+00 1.64916962e-01 4.68674362e-01 3.06266248e-01
-9.10545290e-02 -1.02090251e+00 -8.08909416e-01 4.23940718e-01
6.19850636e-01 -8.91510069e-01 2.10940599e-01 -2.65670955e-01
-5.51614702e-01 9.64641869e-01 5.81421480e-02 2.39537075e-01
5.55325985e-01 6.71325803e-01 1.73008800e-01 -4.00901854e-01
-5.62026441e-01 -1.21844746e-01 3.45122278e-01 4.86294568e-01
1.07578063e+00 8.97839479e-03 -1.04639912e+00 1.22067964e+00
-3.32178771e-01 2.38816574e-01 4.99381810e-01 7.73420036e-01
2.14149117e-01 -1.38530016e+00 -4.28260088e-01 8.07639182e-01
-1.20094705e+00 -4.63792086e-01 -3.30839604e-01 8.96642566e-01
5.23792267e-01 1.11778820e+00 -9.72336531e-03 -2.79894441e-01
6.62372649e-01 4.31373894e-01 2.02541649e-01 -8.46358538e-01
-1.25596571e+00 2.85320401e-01 4.29710269e-01 -4.74497497e-01
-6.38536513e-01 -7.25010514e-01 -1.61923695e+00 -1.15320250e-01
-4.76768345e-01 5.16289592e-01 5.55918157e-01 1.00390804e+00
4.77578878e-01 9.43331063e-01 -1.85209841e-01 1.46192357e-01
-1.17709145e-01 -9.24922585e-01 -5.44083595e-01 3.31232190e-01
3.59889865e-01 -4.19735014e-01 -4.68750775e-01 2.96557784e-01]
|
[8.430915832519531, 8.743837356567383]
|
e927908f-0a40-42d6-af7a-48fec7c07f19
|
cup-a-conservative-update-policy-algorithm-1
|
2202.07565
| null |
https://arxiv.org/abs/2202.07565v1
|
https://arxiv.org/pdf/2202.07565v1.pdf
|
CUP: A Conservative Update Policy Algorithm for Safe Reinforcement Learning
|
Safe reinforcement learning (RL) is still very challenging since it requires the agent to consider both return maximization and safe exploration. In this paper, we propose CUP, a Conservative Update Policy algorithm with a theoretical safety guarantee. We derive the CUP based on the new proposed performance bounds and surrogate functions. Although using bounds as surrogate functions to design safe RL algorithms have appeared in some existing works, we develop them at least three aspects: (i) We provide a rigorous theoretical analysis to extend the surrogate functions to generalized advantage estimator (GAE). GAE significantly reduces variance empirically while maintaining a tolerable level of bias, which is an efficient step for us to design CUP; (ii) The proposed bounds are tighter than existing works, i.e., using the proposed bounds as surrogate functions are better local approximations to the objective and safety constraints. (iii) The proposed CUP provides a non-convex implementation via first-order optimizers, which does not depend on any convex approximation. Finally, extensive experiments show the effectiveness of CUP where the agent satisfies safe constraints. We have opened the source code of CUP at https://github.com/RL-boxes/Safe-RL.
|
['Gang Pan', 'Pengfei Li', 'Yu Zhang', 'Juntao Dai', 'Jiaming Ji', 'Long Yang']
|
2022-02-15
|
cup-a-conservative-update-policy-algorithm
|
https://openreview.net/forum?id=2wiaitACS_O
|
https://openreview.net/pdf?id=2wiaitACS_O
| null |
['safe-exploration']
|
['robots']
|
[-5.48695326e-01 1.86380550e-01 -5.06209850e-01 -2.08022855e-02
-1.02206779e+00 -6.03460550e-01 1.32430390e-01 1.44139886e-01
-6.52639508e-01 1.08100533e+00 9.85179842e-02 -3.48637968e-01
-3.35132569e-01 -6.64217889e-01 -9.93605793e-01 -8.50710094e-01
-5.07275939e-01 5.36840148e-02 2.80249864e-01 -2.76356965e-01
2.50554174e-01 2.08436206e-01 -1.17377210e+00 -5.28972208e-01
1.10516965e+00 1.20405924e+00 7.61559373e-03 3.87315691e-01
7.64100254e-01 6.55189216e-01 -3.84579897e-01 -1.41878143e-01
7.22590804e-01 -3.94406706e-01 -5.32304406e-01 -3.53561282e-01
-7.35720620e-02 -8.57491016e-01 -1.49637267e-01 1.20460522e+00
4.57344502e-01 4.35239673e-01 2.15412647e-01 -1.54464948e+00
-4.54072744e-01 6.80745780e-01 -7.56539106e-01 -1.84694260e-01
1.52554169e-01 2.87557751e-01 1.02334297e+00 -3.47586691e-01
3.26778650e-01 1.17537880e+00 5.04629314e-01 6.54834867e-01
-8.52337778e-01 -8.65690887e-01 5.46203196e-01 1.52179211e-01
-1.37735319e+00 -2.56734252e-01 5.63285589e-01 -6.22272007e-02
6.01710796e-01 3.95396560e-01 6.73293114e-01 9.21988487e-01
2.13739172e-01 9.57427084e-01 1.18009210e+00 -2.00877443e-01
7.19030499e-01 2.44643852e-01 5.66281825e-02 8.52930546e-01
5.69544137e-01 6.55154943e-01 -2.22301424e-01 -3.19378376e-01
7.54833639e-01 -1.38382658e-01 -5.65452278e-01 -5.80023348e-01
-9.56104815e-01 1.18023920e+00 5.36975324e-01 -1.99385062e-01
-3.92959714e-01 5.29338598e-01 4.44313824e-01 1.33684143e-01
3.21047843e-01 3.69074315e-01 -2.03652740e-01 -2.57400662e-01
-4.67236996e-01 5.87816119e-01 8.21780503e-01 8.96315932e-01
2.58291900e-01 2.19351172e-01 -1.99599564e-01 5.71456671e-01
5.16542792e-01 4.62174088e-01 2.19967335e-01 -1.15995216e+00
5.05185068e-01 1.28865734e-01 6.56451285e-01 -9.35698807e-01
-3.61036181e-01 -6.72934830e-01 -4.63313818e-01 4.18954611e-01
4.09632921e-01 -4.67028141e-01 -2.18675137e-01 2.03508329e+00
5.53494334e-01 8.64358768e-02 1.53658807e-01 1.13090992e+00
5.72881885e-02 6.24403119e-01 -1.93521678e-01 -5.79994798e-01
9.83318508e-01 -1.05226362e+00 -8.67493689e-01 -1.15247935e-01
4.96555001e-01 -2.56518036e-01 1.25794673e+00 5.10577679e-01
-1.00042605e+00 8.33144262e-02 -1.33670175e+00 4.31841761e-01
8.79978612e-02 -1.22238370e-02 8.69999766e-01 7.25345731e-01
-7.72302508e-01 4.98843163e-01 -9.50338840e-01 -8.09994712e-02
3.08862299e-01 3.57752353e-01 9.00648311e-02 3.95689875e-01
-1.21689880e+00 8.06045413e-01 5.33528805e-01 2.93338522e-02
-1.14814973e+00 -6.33299053e-01 -9.13187981e-01 3.15787271e-02
1.02482295e+00 -5.03055930e-01 1.53566349e+00 -5.47563016e-01
-1.98840261e+00 7.21594505e-03 2.13991016e-01 -7.28779554e-01
8.98976624e-01 -5.76570690e-01 1.70993358e-01 1.87987939e-01
-1.89940244e-01 2.55011559e-01 5.85440278e-01 -1.33238530e+00
-7.20290959e-01 -2.08127394e-01 5.35664201e-01 2.78648406e-01
-5.01853704e-01 -2.11389467e-01 -2.47522682e-01 -6.09923661e-01
-3.61722648e-01 -9.90325809e-01 -5.10913193e-01 5.11762835e-02
-3.67228985e-01 -8.27459693e-02 4.42986369e-01 -4.45068747e-01
1.52963710e+00 -1.97856879e+00 -8.79044160e-02 2.29750797e-01
1.58263855e-02 1.85597092e-01 8.55835378e-02 3.87521803e-01
3.64290416e-01 9.83666778e-02 -2.54948258e-01 -1.24072015e-01
3.91791970e-01 6.70683309e-02 -6.49035633e-01 8.58216345e-01
-3.57722938e-01 5.73911548e-01 -1.03687310e+00 -2.57563293e-01
2.58135140e-01 1.62485734e-01 -8.22442293e-01 2.68035501e-01
-1.60230741e-01 1.92720085e-01 -7.77495325e-01 3.94168973e-01
7.80727565e-01 2.18550086e-01 1.47018403e-01 -5.36183678e-02
-2.70608187e-01 1.27643466e-01 -1.46472526e+00 1.21600521e+00
-3.92596215e-01 -1.47643068e-03 4.01801169e-01 -9.34687436e-01
6.34928226e-01 1.06521040e-01 4.87129927e-01 -5.20973325e-01
3.58200312e-01 2.80534923e-01 -2.62053698e-01 -1.95397183e-01
2.85672784e-01 -7.77026266e-02 -2.08310619e-01 4.72056895e-01
-3.57926726e-01 2.02288151e-01 3.20629597e-01 6.66921139e-02
8.49251270e-01 3.49223793e-01 5.21790564e-01 -5.15799940e-01
4.87854749e-01 -2.63538241e-01 9.18309510e-01 8.33527565e-01
-6.00318491e-01 -5.03500402e-02 7.18335748e-01 -1.60255358e-01
-5.97429514e-01 -9.01731074e-01 -1.78864002e-01 9.81210530e-01
4.84104455e-01 -3.66836667e-01 -7.51362085e-01 -8.31266403e-01
1.63775265e-01 9.37546790e-01 -6.07903779e-01 -2.92036086e-01
-4.58037257e-01 -7.21643209e-01 4.81557786e-01 4.98029947e-01
6.06982231e-01 -5.07924020e-01 -1.10286546e+00 1.35230035e-01
1.30539732e-02 -7.92852044e-01 -9.24003959e-01 1.21180624e-01
-7.20720768e-01 -1.10904455e+00 -5.06058156e-01 -3.03543836e-01
6.69140756e-01 1.85544759e-01 5.24425328e-01 -1.39368819e-02
1.52458340e-01 4.40726787e-01 -4.36392009e-01 -5.37167132e-01
-1.23332338e-02 -1.74312308e-01 4.09068972e-01 -1.29862010e-01
-2.91495085e-01 -4.31982547e-01 -7.39447594e-01 3.36593628e-01
-7.13681161e-01 -4.35192809e-02 3.97203773e-01 7.85763144e-01
6.90746486e-01 6.45596832e-02 7.63530791e-01 -6.68265164e-01
9.92940128e-01 -4.03073251e-01 -1.37477803e+00 1.48975134e-01
-7.81819522e-01 3.59578967e-01 8.76209497e-01 -4.24835742e-01
-9.70440805e-01 -1.33828018e-02 -1.82384685e-01 -4.83086020e-01
3.80506694e-01 2.64110804e-01 -3.23666222e-02 -2.09655389e-01
4.12625015e-01 5.72859906e-02 1.63122445e-01 -3.19230884e-01
3.86860996e-01 4.26647037e-01 2.17598230e-01 -1.00398111e+00
8.09209168e-01 3.86529028e-01 -2.61239037e-02 -2.66550183e-01
-9.46725368e-01 -8.38138461e-02 1.71595082e-01 -1.18193820e-01
3.12175870e-01 -7.44455814e-01 -1.58667171e+00 -1.07717685e-01
-6.00210786e-01 -6.51181400e-01 -1.46916017e-01 7.55998611e-01
-8.45160663e-01 4.60770369e-01 -5.71019053e-01 -1.32380009e+00
-5.93064845e-01 -1.21824086e+00 6.56281292e-01 4.04168576e-01
8.48888680e-02 -8.98299813e-01 1.44453153e-01 4.20976058e-02
3.10816526e-01 6.13253355e-01 3.30226988e-01 -3.60433459e-01
-5.91115952e-01 3.82448807e-02 1.28536984e-01 2.95686692e-01
-6.11363128e-02 -1.02819130e-01 -5.96205413e-01 -7.96066344e-01
2.35748872e-01 -5.61237991e-01 6.26002610e-01 4.48001444e-01
1.27440941e+00 -9.77028847e-01 -1.67116910e-01 8.52295518e-01
1.46464169e+00 5.17309785e-01 1.97593912e-01 8.23843896e-01
2.41635472e-01 2.90229887e-01 1.22168207e+00 1.12367308e+00
4.46100265e-01 5.97323060e-01 7.92745173e-01 1.86150730e-01
5.64006627e-01 -3.67099732e-01 8.37352335e-01 5.08376002e-01
4.65857908e-02 4.22800109e-02 -4.91845459e-01 3.30990613e-01
-2.34067583e+00 -8.98417950e-01 2.71057248e-01 2.63446856e+00
1.08041465e+00 9.60207283e-02 3.06669712e-01 -2.18217731e-01
4.76191610e-01 3.86590250e-02 -9.29548681e-01 -5.24809301e-01
3.84335279e-01 -5.14405640e-03 9.60106850e-01 7.86283851e-01
-1.04650652e+00 6.63747251e-01 5.89543200e+00 9.70789611e-01
-1.00106108e+00 1.69663042e-01 3.95273954e-01 -4.71678734e-01
-2.04580888e-01 5.69901504e-02 -8.45832586e-01 6.38803422e-01
6.21431351e-01 -6.37153447e-01 7.51380086e-01 1.32291031e+00
4.29528594e-01 -2.57310897e-01 -1.03055751e+00 7.62396336e-01
-2.78236926e-01 -1.00886667e+00 -3.22370797e-01 8.53362381e-02
5.90534270e-01 -3.80809247e-01 1.17209271e-01 6.48975194e-01
6.05581045e-01 -8.01616073e-01 1.10021174e+00 1.42797738e-01
3.66420597e-01 -1.32394612e+00 6.60077810e-01 5.09332180e-01
-1.07642722e+00 -3.35350603e-01 -5.16841531e-01 1.21669397e-02
7.71501940e-03 3.25525880e-01 -5.16397834e-01 6.37061715e-01
5.14225721e-01 3.22674572e-01 -1.48362249e-01 1.18587220e+00
-5.61093330e-01 5.18520713e-01 -5.13343215e-01 -2.69861400e-01
4.29869205e-01 -3.34118634e-01 6.32954359e-01 7.76042998e-01
2.07605824e-01 -5.84230311e-02 5.52683651e-01 9.18136716e-01
1.43345922e-01 1.19777232e-01 -2.72703767e-01 8.48231763e-02
6.62864268e-01 1.22221422e+00 -3.37353855e-01 3.03722993e-02
-1.79365337e-01 4.71945018e-01 3.83944482e-01 3.50329578e-01
-1.37238002e+00 -5.77945530e-01 7.98006415e-01 -3.18174064e-01
1.16416335e-01 -2.26331219e-01 -1.26187682e-01 -1.04086244e+00
1.65043220e-01 -9.70625162e-01 6.82086170e-01 -1.42811731e-01
-1.01684892e+00 4.21044797e-01 1.82944968e-01 -1.28820550e+00
-2.40665495e-01 -3.42164904e-01 -5.01948178e-01 5.06862223e-01
-1.47724342e+00 -7.87319481e-01 -3.23953331e-02 6.68229640e-01
3.09261918e-01 1.31045550e-01 5.14455855e-01 2.48720884e-01
-9.47840929e-01 9.16736960e-01 2.57062048e-01 -3.07221919e-01
6.42206550e-01 -1.17006040e+00 -3.74864221e-01 9.63596404e-01
-4.01835293e-01 8.38258266e-01 9.05333221e-01 -5.25019765e-01
-1.63144934e+00 -9.34069276e-01 -2.02977240e-01 -4.98058349e-02
7.18912899e-01 -2.89275974e-01 -5.27006209e-01 8.61920953e-01
8.17647800e-02 -1.99089646e-02 4.14504826e-01 -1.48977740e-02
-8.80370438e-02 -4.18925136e-01 -1.17806828e+00 8.98126245e-01
7.72044659e-01 2.64657103e-02 -3.00871521e-01 4.01958406e-01
8.19684982e-01 -7.08672464e-01 -8.72945905e-01 4.50193256e-01
5.37287951e-01 -9.61388230e-01 7.39268959e-01 -2.88691044e-01
-9.18851197e-02 -5.01847625e-01 -7.42114037e-02 -1.34838617e+00
-1.72484051e-02 -9.37496722e-01 -5.32868028e-01 1.00726962e+00
1.91553697e-01 -1.06826222e+00 4.84860599e-01 7.34915912e-01
-9.64042544e-02 -1.35346043e+00 -9.20072258e-01 -1.32977211e+00
8.79370570e-02 -4.17851776e-01 5.24354577e-01 6.38814151e-01
2.90984362e-01 -1.73640162e-01 -7.61396408e-01 3.60173136e-01
8.91446590e-01 2.23784551e-01 6.32248402e-01 -4.44914132e-01
-5.76455057e-01 -4.16297913e-01 2.54042745e-01 -1.04125953e+00
1.10273674e-01 -5.40039182e-01 1.87276885e-01 -1.28619528e+00
2.08767995e-01 -6.62787795e-01 -4.61696237e-01 6.90926969e-01
-2.33130082e-01 -4.11624163e-01 3.22284997e-01 8.68717209e-02
-8.99756491e-01 1.15402722e+00 1.21349621e+00 1.78243592e-01
-4.51639414e-01 1.38774917e-01 -9.04025495e-01 7.97843218e-01
1.16143799e+00 -4.72907841e-01 -7.15378582e-01 -5.68287969e-02
1.74129918e-01 2.25538746e-01 1.73734501e-01 -7.39810646e-01
-1.02307843e-02 -5.68025768e-01 -1.00908957e-01 -1.87843591e-01
1.34377256e-01 -8.60469103e-01 -5.05494028e-02 9.49588358e-01
-4.52546060e-01 -2.05605462e-01 2.26418525e-01 6.17370486e-01
1.64126799e-01 -3.74742299e-01 1.07445550e+00 8.98201987e-02
-4.28194821e-01 4.58137751e-01 -1.46204814e-01 1.12762682e-01
1.38643968e+00 2.31259122e-01 -3.60331863e-01 -6.55534565e-01
-1.72659189e-01 7.90731132e-01 3.87431204e-01 2.70601392e-01
5.26600778e-01 -1.35447168e+00 -3.12260598e-01 -3.00278991e-01
-2.01255694e-01 -1.43942118e-01 -2.69055124e-02 1.09792590e+00
-4.35441256e-01 5.42387009e-01 -1.03222758e-01 -1.28203273e-01
-9.17825520e-01 7.79617906e-01 3.36804032e-01 -4.61891115e-01
-6.25850022e-01 6.49373353e-01 3.23821574e-01 -2.75181741e-01
6.81523204e-01 -4.04818565e-01 3.29677239e-02 -2.35604689e-01
6.03632033e-01 4.75752652e-01 -3.90336722e-01 -2.47812793e-02
-4.54055190e-01 3.46441984e-01 -3.83535326e-02 -1.98487997e-01
1.41372836e+00 -1.23645723e-01 2.33870551e-01 1.29728958e-01
9.41001773e-01 2.13015899e-01 -1.65503561e+00 9.30403247e-02
-1.08034074e-01 -7.27048159e-01 1.96099818e-01 -6.22837186e-01
-1.08485496e+00 3.81679177e-01 6.40281916e-01 -1.08048528e-01
1.20421720e+00 -4.90776062e-01 8.21718097e-01 4.28475887e-01
8.40725839e-01 -1.34412980e+00 -2.25680526e-02 4.03302729e-01
1.13711643e+00 -9.81118202e-01 3.01670104e-01 -2.52705574e-01
-8.14401567e-01 9.13844585e-01 8.60321522e-01 -3.02723944e-01
3.28751773e-01 3.05737734e-01 -3.18592221e-01 2.48253986e-01
-8.46055448e-01 -2.85405815e-01 -1.20883375e-01 2.87677109e-01
7.11734220e-02 6.60470352e-02 -9.23773587e-01 9.42402065e-01
-1.79289505e-01 -9.62751976e-04 5.36520839e-01 1.02898288e+00
-5.03991306e-01 -9.95374560e-01 -4.04180110e-01 1.06755644e-01
-5.55317044e-01 2.36012831e-01 1.26716733e-01 9.57109690e-01
-3.37340295e-01 8.96908700e-01 -4.24415648e-01 -1.61118627e-01
3.54634881e-01 -5.33581078e-01 4.28293586e-01 -1.21833861e-01
-3.95101875e-01 1.39619373e-02 1.24813735e-01 -9.82203186e-01
3.02761048e-02 -3.96779448e-01 -1.50355411e+00 -4.20223147e-01
-3.48551840e-01 4.04128790e-01 4.50641066e-01 6.35167658e-01
2.36305520e-01 3.69922876e-01 9.12315130e-01 -5.60344815e-01
-1.38643146e+00 -5.43876469e-01 -6.10812426e-01 1.00741677e-01
3.35317492e-01 -9.77017403e-01 -4.85530019e-01 -5.74149668e-01]
|
[4.296194076538086, 2.3286256790161133]
|
ba3cf05a-5f05-446c-b01e-e294b8512f12
|
entity-tracking-improves-cloze-style-reading
|
1810.02891
| null |
http://arxiv.org/abs/1810.02891v1
|
http://arxiv.org/pdf/1810.02891v1.pdf
|
Entity Tracking Improves Cloze-style Reading Comprehension
|
Reading comprehension tasks test the ability of models to process long-term
context and remember salient information. Recent work has shown that relatively
simple neural methods such as the Attention Sum-Reader can perform well on
these tasks; however, these systems still significantly trail human
performance. Analysis suggests that many of the remaining hard instances are
related to the inability to track entity-references throughout documents. This
work focuses on these hard entity tracking cases with two extensions: (1)
additional entity features, and (2) training with a multi-task tracking
objective. We show that these simple modifications improve performance both
independently and in combination, and we outperform the previous state of the
art on the LAMBADA dataset, particularly on difficult entity examples.
|
['Alexander M. Rush', 'Sam Wiseman', 'Luong Hoang']
|
2018-10-05
|
entity-tracking-improves-cloze-style-reading-1
|
https://aclanthology.org/D18-1130
|
https://aclanthology.org/D18-1130.pdf
|
emnlp-2018-10
|
['lambada']
|
['natural-language-processing']
|
[ 8.64676237e-02 2.35172004e-01 5.31978197e-02 -1.68618232e-01
-7.91141570e-01 -6.18495941e-01 1.04132593e+00 7.15987742e-01
-9.22138631e-01 8.62281144e-01 4.18685019e-01 -2.34628141e-01
-2.79506952e-01 -5.28578460e-01 -7.28509307e-01 -1.82484791e-01
-1.21398382e-01 5.75851500e-01 7.60730863e-01 -3.12731683e-01
6.62315667e-01 1.30525619e-01 -1.56218326e+00 4.97381426e-02
1.03390074e+00 4.43699479e-01 4.10470992e-01 7.68291354e-01
-3.34757477e-01 1.04742897e+00 -7.95796275e-01 -5.26080966e-01
-1.94652632e-01 -1.26122043e-01 -1.10028076e+00 -6.81290150e-01
1.00754023e+00 -4.16122794e-01 -2.79071480e-01 7.99709439e-01
6.08333528e-01 5.39886415e-01 5.95519066e-01 -8.91284108e-01
-1.19928598e+00 7.19585001e-01 -4.97881889e-01 1.11294234e+00
5.97113132e-01 -3.63028571e-02 1.18160796e+00 -1.05516005e+00
6.76042438e-01 9.74438250e-01 8.50688696e-01 5.06747365e-01
-9.90079641e-01 -4.03268933e-01 5.99830508e-01 5.83173215e-01
-1.17863679e+00 -7.42187798e-01 4.20569889e-02 -3.02764833e-01
1.80663157e+00 1.48185775e-01 2.68562198e-01 1.11968863e+00
2.28819683e-01 1.05179000e+00 1.04601681e+00 -6.76234424e-01
-5.29833697e-02 -2.69054383e-01 1.08104503e+00 6.30933225e-01
5.20857573e-01 9.29481834e-02 -8.16258788e-01 -1.15523813e-03
1.74718693e-01 -1.64690956e-01 -6.50728285e-01 -8.22650036e-04
-1.42677331e+00 4.71272796e-01 3.56922984e-01 5.44171154e-01
-4.59544510e-01 1.29267871e-01 7.50586316e-02 1.82255507e-01
3.36227894e-01 9.99749720e-01 -6.36993289e-01 -2.34824494e-01
-1.17112529e+00 5.16537786e-01 1.07582152e+00 1.02357674e+00
3.34468305e-01 -2.99974442e-01 -7.89340913e-01 6.10007703e-01
1.39492050e-01 4.30702090e-01 5.41608751e-01 -5.44275939e-01
7.03808248e-01 1.25564754e-01 2.07622394e-01 -7.68362284e-01
-8.40479612e-01 -6.95140123e-01 -1.67699978e-01 -1.00517616e-01
8.07001114e-01 2.75694504e-02 -9.89063501e-01 2.06977558e+00
-9.33161601e-02 1.98632516e-02 -1.49016500e-01 6.22048020e-01
8.27621520e-01 4.27371979e-01 7.44745970e-01 6.90087602e-02
1.57774508e+00 -1.23745871e+00 -1.05767298e+00 -6.48968399e-01
5.17814219e-01 -7.10380435e-01 1.00331092e+00 6.01985864e-02
-1.69955707e+00 -4.72353309e-01 -1.14956129e+00 -6.33580685e-01
-9.67189670e-01 -2.91304328e-02 6.85360253e-01 3.15714478e-01
-1.25542009e+00 6.80278480e-01 -7.49904692e-01 -5.52183330e-01
4.43399310e-01 1.35736346e-01 -1.36376336e-01 -5.38711548e-02
-1.51421976e+00 1.71282148e+00 3.75628263e-01 -1.69604942e-02
-4.35732961e-01 -8.24404478e-01 -8.70299339e-01 6.69784963e-01
4.85430509e-01 -1.04981947e+00 1.54018402e+00 -3.85912150e-01
-8.22686911e-01 9.91307974e-01 -4.88449097e-01 -4.82517987e-01
2.13985801e-01 -7.98630297e-01 -3.41718197e-01 8.63159001e-02
3.04497391e-01 6.58078671e-01 4.50333387e-01 -7.75567770e-01
-8.44632447e-01 -2.70431668e-01 4.33829278e-02 3.87350053e-01
-3.40470374e-01 1.79809123e-01 -4.28552359e-01 -7.39717066e-01
-2.39394486e-01 -6.24241829e-01 2.95640409e-01 -1.40002385e-01
-3.44357789e-01 -6.56084716e-01 3.54555607e-01 -1.02611625e+00
1.43248510e+00 -1.74218118e+00 1.14654571e-01 -1.16159149e-01
5.16414702e-01 2.66425073e-01 -2.10625619e-01 3.27099144e-01
1.98385780e-04 2.61444926e-01 2.64027510e-02 -4.47303623e-01
2.41877973e-01 -2.68101156e-01 -1.91674948e-01 8.12493712e-02
3.05451512e-01 1.46123028e+00 -9.66818094e-01 -4.26801205e-01
-3.30944479e-01 1.89374313e-01 -2.26501197e-01 -1.85478330e-01
-2.90965617e-01 -2.57736325e-01 -1.96052700e-01 5.86169004e-01
3.53573948e-01 -7.44109809e-01 -2.60551482e-01 1.72623694e-01
-1.94476783e-01 8.21999848e-01 -8.38178813e-01 1.62407613e+00
-2.49562990e-02 1.03608513e+00 -3.85016985e-02 -5.21915972e-01
1.37815714e-01 1.90263525e-01 -1.98078364e-01 -9.69042242e-01
-1.50709838e-01 -3.54060368e-03 2.21287906e-01 -5.53951561e-01
8.95051301e-01 2.73311764e-01 1.54387027e-01 6.36455953e-01
2.17317387e-01 4.97039914e-01 2.69481808e-01 5.85624039e-01
1.35021925e+00 1.65245503e-01 4.24795955e-01 -4.25540805e-01
1.89011142e-01 4.43727449e-02 6.00423664e-02 1.29911804e+00
-3.28962266e-01 4.41936374e-01 5.72776720e-02 3.71900387e-02
-9.26838994e-01 -1.03050041e+00 -9.41655487e-02 1.71943569e+00
1.75157934e-01 -4.08882499e-01 -6.94144130e-01 -7.39951372e-01
1.83908626e-01 1.24832404e+00 -7.42211998e-01 -2.00097114e-01
-7.25496829e-01 -6.46898270e-01 5.40859818e-01 1.14875555e+00
4.77155536e-01 -1.11996162e+00 -7.09773719e-01 3.26533467e-01
-1.66230857e-01 -9.30407226e-01 -4.58299488e-01 4.93422925e-01
-5.49847364e-01 -9.15291131e-01 -9.69328761e-01 -9.46313024e-01
3.19310367e-01 4.22806621e-01 1.74926269e+00 5.63106596e-01
-1.57148883e-01 5.93904376e-01 -8.90916064e-02 -6.22844994e-01
1.33013383e-01 6.17037654e-01 -3.55456471e-01 -9.93079603e-01
8.83913279e-01 -4.99948189e-02 -4.91242409e-01 1.15561657e-01
-4.18654412e-01 -1.80462189e-02 7.99317181e-01 1.01711178e+00
3.01130682e-01 -4.15773958e-01 9.74454224e-01 -1.14134717e+00
9.60958481e-01 -6.88632548e-01 -2.14015320e-01 7.48646200e-01
-9.09582376e-01 1.93758905e-01 1.83974087e-01 -3.54546905e-01
-1.23400354e+00 -4.50135976e-01 -1.28037751e-01 7.85936639e-02
-2.10341007e-01 6.29325330e-01 2.03675777e-01 7.29443431e-02
7.73010373e-01 1.87154353e-01 -4.18122649e-01 -4.78052974e-01
3.14170957e-01 1.87661752e-01 7.30973780e-01 -6.51814699e-01
7.13868439e-01 -1.42255142e-01 -2.66908407e-01 -6.50611162e-01
-1.32034683e+00 -4.61311191e-01 -5.43527365e-01 1.07753821e-01
1.00178194e+00 -9.63151932e-01 -6.53255403e-01 4.32874560e-01
-1.00564468e+00 -4.83400822e-01 -1.11961633e-01 3.97558749e-01
-2.48108804e-01 2.94409752e-01 -8.27133775e-01 -4.89565521e-01
-3.94081950e-01 -5.82239509e-01 9.95470881e-01 6.13435864e-01
-5.39736271e-01 -1.06757689e+00 -2.02532485e-02 1.28482908e-01
8.72716546e-01 -1.18391067e-01 1.05192673e+00 -1.19354200e+00
-7.24291265e-01 -3.38527113e-02 -3.29914421e-01 -2.93312341e-01
-4.15416569e-01 -2.05146492e-01 -1.02497888e+00 -1.48455560e-01
-2.61147350e-01 -3.23296338e-01 1.42874002e+00 2.87076652e-01
8.52408409e-01 -4.76405844e-02 -7.20364571e-01 2.39666760e-01
1.01011896e+00 -5.77372909e-02 4.55377728e-01 6.50055349e-01
5.04910767e-01 5.58032990e-01 3.61260176e-01 -2.29286939e-01
9.29724395e-01 5.99013388e-01 4.01188135e-02 1.41801476e-01
-5.01221895e-01 -2.51196980e-01 4.06162590e-02 4.98835206e-01
-6.79573417e-02 -5.73739707e-01 -1.16230011e+00 9.02139664e-01
-1.96373200e+00 -1.26740825e+00 -2.73695707e-01 1.85701454e+00
8.12394977e-01 4.01830822e-01 -1.82839543e-01 -2.42491171e-01
6.75823212e-01 1.72046527e-01 -6.15685880e-01 -9.52935070e-02
-4.42876846e-01 4.65496868e-01 4.24836606e-01 4.42199796e-01
-1.14903724e+00 9.67164159e-01 7.55362368e+00 5.00995219e-01
-5.54955482e-01 1.68341488e-01 1.83594272e-01 -1.28915444e-01
-2.34313384e-01 -1.95232928e-01 -1.22956622e+00 3.30122024e-01
1.11971748e+00 -3.01127642e-01 2.30400398e-01 5.91696501e-01
-5.67692757e-01 -4.31458354e-01 -1.40554476e+00 4.96668667e-01
5.33012748e-01 -1.19792283e+00 -1.64319247e-01 -2.25067511e-01
6.10743940e-01 1.73909962e-01 7.20051676e-02 7.68914044e-01
2.97269613e-01 -1.10300016e+00 8.55638087e-01 9.62973714e-01
2.35982969e-01 -3.33242744e-01 5.69338620e-01 4.13602173e-01
-8.05118799e-01 -4.04324681e-02 -3.74363273e-01 -1.48404902e-02
2.20130801e-01 2.02407330e-01 -5.55374622e-01 1.73733309e-01
7.70380676e-01 6.04478598e-01 -1.31225967e+00 1.49328887e+00
-5.74694455e-01 5.50384521e-01 -2.63678700e-01 -4.27247107e-01
7.99436569e-02 6.91956699e-01 7.02793241e-01 1.46509218e+00
1.86933577e-01 2.13111475e-01 -1.82355553e-01 9.12620962e-01
-2.80129641e-01 -5.68761565e-02 -3.75465631e-01 -1.78451762e-02
7.47007370e-01 1.00001538e+00 -5.02911150e-01 -5.21492541e-01
-6.88463748e-01 8.47748518e-01 9.57249224e-01 3.58080447e-01
-6.04094565e-01 -6.23465776e-01 2.34505355e-01 -4.44223434e-02
6.06711507e-01 -3.16038013e-01 -4.21530396e-01 -1.20897937e+00
9.69134867e-02 -5.60028553e-01 7.76679456e-01 -1.07625127e+00
-1.51000488e+00 2.85690516e-01 -7.31998160e-02 -3.90540332e-01
-2.81388611e-01 -6.11847758e-01 -6.66947901e-01 1.14341116e+00
-2.02345967e+00 -1.10793734e+00 -4.75136876e-01 5.74065745e-01
5.86546063e-01 -3.73944901e-02 9.07815933e-01 1.89445227e-01
-5.73862672e-01 8.41305315e-01 -2.41163485e-02 -5.05621254e-04
9.86847341e-01 -1.79363930e+00 6.56367838e-01 1.08777952e+00
2.88144827e-01 1.22334182e+00 6.22194231e-01 -6.98437452e-01
-9.78521585e-01 -5.65459132e-01 1.48253798e+00 -1.15698183e+00
4.96554434e-01 -3.77315372e-01 -1.40789175e+00 1.11937654e+00
8.21927547e-01 -3.78079504e-01 6.97733641e-01 7.60405302e-01
-4.63086963e-01 5.44652343e-01 -8.92002642e-01 5.70532262e-01
1.22946703e+00 -5.24900079e-01 -1.52434027e+00 3.04817587e-01
6.57842219e-01 -5.72097540e-01 -8.98071051e-01 7.86187947e-02
4.96030122e-01 -5.70283115e-01 1.08092737e+00 -9.90635991e-01
4.93518829e-01 -2.26489305e-01 1.99734434e-01 -1.44052458e+00
-6.81865156e-01 -4.70330447e-01 -7.86496222e-01 1.31596160e+00
6.93014145e-01 -5.76618969e-01 2.63657779e-01 9.81449783e-01
-2.78398454e-01 -5.27479589e-01 -5.65087438e-01 -7.63874054e-01
3.82149518e-01 2.27120399e-01 4.44191724e-01 1.10160768e+00
1.65834695e-01 7.68130064e-01 1.02461301e-01 1.41869679e-01
3.56960624e-01 -4.24364321e-02 1.94148540e-01 -1.35451841e+00
-1.36240646e-01 -8.30856085e-01 -9.83994734e-03 -1.21852696e+00
3.66644621e-01 -8.45737696e-01 1.28036901e-01 -1.78333986e+00
4.23519820e-01 -2.41181105e-01 -4.95614052e-01 6.56818390e-01
-1.10091925e+00 -2.81012952e-01 2.55870312e-01 3.55455548e-01
-1.11466086e+00 1.74272254e-01 1.03930688e+00 6.47645593e-02
1.55591190e-01 -1.85795575e-01 -1.12909365e+00 5.32435298e-01
5.22152781e-01 -2.68974632e-01 -1.26630008e-01 -8.28098714e-01
4.76585239e-01 -2.73425728e-01 3.49078923e-01 -1.07417762e+00
7.14465439e-01 2.67693728e-01 1.00128806e+00 -7.25843966e-01
8.19381252e-02 -3.82847190e-01 -5.72995842e-01 2.37055615e-01
-7.97245562e-01 5.27988851e-01 5.56803167e-01 6.84296846e-01
-3.19461487e-02 -4.72061783e-01 3.75304013e-01 -1.69447586e-01
-1.18561769e+00 -2.28049219e-01 -2.87560791e-01 6.17179036e-01
8.57229650e-01 -6.80036694e-02 -1.38009346e+00 -2.51836896e-01
-8.99156153e-01 6.11516297e-01 3.87324005e-01 5.01215875e-01
1.69575289e-01 -9.33715403e-01 -7.56673574e-01 -1.82464674e-01
1.94720581e-01 -2.14154422e-01 9.55860987e-02 8.90248179e-01
-1.54597014e-01 7.15758026e-01 -1.86841547e-01 -2.66090989e-01
-9.74787593e-01 7.23388016e-01 2.76555866e-01 -4.39942956e-01
-6.04953170e-01 1.09602630e+00 3.92705128e-02 6.31011948e-02
5.18092275e-01 -1.94364145e-01 -7.28705704e-01 4.13528085e-01
9.23500478e-01 4.22209948e-01 3.19430292e-01 -3.68460000e-01
-4.46709305e-01 3.58262688e-01 -6.63509011e-01 6.33089840e-02
1.33733237e+00 -4.08086270e-01 2.52385706e-01 2.62118191e-01
6.84261084e-01 2.28962600e-01 -9.85319257e-01 -3.70481640e-01
5.53933203e-01 -1.52230397e-01 1.37342989e-01 -1.43965840e+00
-5.11672258e-01 9.10220265e-01 3.30632389e-01 3.63171786e-01
7.90723085e-01 -2.47767065e-02 8.27939451e-01 7.72060931e-01
1.18722588e-01 -1.15332472e+00 -2.05399811e-01 8.59544456e-01
6.45448923e-01 -1.24720764e+00 2.02899173e-01 4.73954435e-03
-5.32581031e-01 1.00249112e+00 1.02679205e+00 7.15059936e-02
3.49849343e-01 1.58452272e-01 -2.34027937e-01 -5.62431753e-01
-1.08974314e+00 -4.97074246e-01 5.86662769e-01 5.20699263e-01
8.16163361e-01 -3.59615862e-01 -4.39211786e-01 7.55424261e-01
-2.35722467e-01 -1.49642631e-01 4.12813604e-01 1.23263955e+00
-6.05746567e-01 -6.27748787e-01 -8.09153691e-02 7.61393964e-01
-7.86106706e-01 -6.01797521e-01 -4.01398718e-01 1.04503167e+00
-3.61256719e-01 7.65791893e-01 2.13865921e-01 2.73255259e-01
6.44298673e-01 7.50000775e-01 5.68381965e-01 -7.67628372e-01
-9.56617653e-01 -3.50204468e-01 1.78292140e-01 -5.39770961e-01
-2.69407719e-01 -9.94044840e-01 -1.12115264e+00 -2.43993819e-01
-6.40751541e-01 -1.06773548e-01 3.13538492e-01 9.58470583e-01
5.55556059e-01 8.50613475e-01 -3.91775072e-01 -5.00146627e-01
-7.94031680e-01 -1.19594002e+00 -2.20455766e-01 5.68936348e-01
5.38986444e-01 -8.80782127e-01 -2.91954637e-01 -1.02524057e-01]
|
[11.041815757751465, 8.194320678710938]
|
4acfb409-f9f3-4ee4-930a-b75382787234
|
robust-depth-completion-with-uncertainty
|
2112.07895
| null |
https://arxiv.org/abs/2112.07895v2
|
https://arxiv.org/pdf/2112.07895v2.pdf
|
Robust Depth Completion with Uncertainty-Driven Loss Functions
|
Recovering a dense depth image from sparse LiDAR scans is a challenging task. Despite the popularity of color-guided methods for sparse-to-dense depth completion, they treated pixels equally during optimization, ignoring the uneven distribution characteristics in the sparse depth map and the accumulated outliers in the synthesized ground truth. In this work, we introduce uncertainty-driven loss functions to improve the robustness of depth completion and handle the uncertainty in depth completion. Specifically, we propose an explicit uncertainty formulation for robust depth completion with Jeffrey's prior. A parametric uncertain-driven loss is introduced and translated to new loss functions that are robust to noisy or missing data. Meanwhile, we propose a multiscale joint prediction model that can simultaneously predict depth and uncertainty maps. The estimated uncertainty map is also used to perform adaptive prediction on the pixels with high uncertainty, leading to a residual map for refining the completion results. Our method has been tested on KITTI Depth Completion Benchmark and achieved the state-of-the-art robustness performance in terms of MAE, IMAE, and IRMSE metrics.
|
['Guangming Shi', 'Xin Li', 'Jinjian Wu', 'Leida Li', 'Weisheng Dong', 'Yufan Zhu']
|
2021-12-15
| null | null | null | null |
['depth-completion']
|
['computer-vision']
|
[ 3.76852036e-01 1.85433120e-01 1.80187821e-01 -5.36908746e-01
-1.28333592e+00 -7.01391920e-02 2.84933537e-01 1.07722253e-01
-1.61829054e-01 7.28065908e-01 3.64900291e-01 4.34063584e-01
-2.53092080e-01 -7.63057053e-01 -6.63656235e-01 -6.61951959e-01
2.23599538e-01 5.88997543e-01 2.12147385e-01 2.79659152e-01
4.11366910e-01 5.23114622e-01 -1.57669687e+00 -1.65440604e-01
1.48183441e+00 1.21268034e+00 3.39873224e-01 2.49413326e-01
-9.17113051e-02 6.63553715e-01 -1.50514841e-01 -2.64520466e-01
4.09767032e-01 1.18245400e-01 -2.74104804e-01 1.86299115e-01
7.40299642e-01 -5.73476851e-01 -4.95189399e-01 1.17779160e+00
5.96480012e-01 2.04839826e-01 6.16171956e-01 -9.62921023e-01
-2.05836117e-01 1.51707441e-01 -9.87863302e-01 -3.28105360e-01
3.44288141e-01 4.51288708e-02 6.88413978e-01 -1.24968767e+00
6.23029947e-01 1.37677240e+00 5.45111299e-01 2.43703827e-01
-1.29949594e+00 -6.44626677e-01 2.53560215e-01 1.43876404e-01
-1.52012718e+00 -3.95171553e-01 8.82868528e-01 -6.13358200e-01
4.01233166e-01 -1.03056848e-01 4.43651199e-01 7.13786900e-01
2.54243225e-01 6.37957513e-01 9.93470550e-01 -6.70667961e-02
4.55244362e-01 -3.38225901e-01 -2.73268849e-01 6.19660437e-01
3.39806497e-01 2.61202931e-01 -7.84081817e-01 -8.39697663e-04
7.39013553e-01 8.42461586e-02 -4.42416221e-01 -5.08844554e-01
-1.05494606e+00 7.03406811e-01 6.38131857e-01 -5.25790632e-01
-4.67798769e-01 2.62277633e-01 3.00391973e-03 -2.88043708e-01
7.88159728e-01 8.99932832e-02 -9.15986001e-02 -7.38137751e-04
-1.06196558e+00 2.98592746e-01 3.46747100e-01 9.01876628e-01
1.25813425e+00 1.44569129e-01 -3.70726079e-01 7.28967607e-01
6.02806926e-01 7.14525938e-01 -1.49048179e-01 -1.39787102e+00
7.80750096e-01 5.67774355e-01 2.12455183e-01 -9.74626720e-01
-2.07773864e-01 -5.05279601e-01 -9.09286141e-01 5.36990523e-01
2.45713636e-01 1.23741247e-01 -1.09080446e+00 1.63544297e+00
5.52628994e-01 5.53754270e-01 -5.55186756e-02 1.03862858e+00
5.83136559e-01 5.37112951e-01 -2.22100183e-01 -1.19228378e-01
7.41120934e-01 -6.69612169e-01 -7.46551633e-01 -4.39950079e-01
-7.86004066e-02 -6.96753442e-01 7.43035614e-01 6.09775186e-01
-1.17026556e+00 -4.27075386e-01 -1.14338100e+00 -3.61848295e-01
4.05073136e-01 8.92060697e-02 3.51728201e-01 2.51109153e-01
-8.09832990e-01 6.76199317e-01 -9.73687232e-01 9.07398909e-02
4.79525685e-01 9.06521156e-02 -3.24025214e-01 -6.11966014e-01
-7.55231380e-01 6.27992034e-01 1.94117069e-01 3.00800264e-01
-1.18268085e+00 -1.02559364e+00 -1.12637579e+00 -2.79433072e-01
3.56045574e-01 -7.84737408e-01 7.10102379e-01 -2.72982508e-01
-1.48743534e+00 5.13416469e-01 -2.39680007e-01 -2.32878238e-01
7.74571598e-01 -4.32662815e-01 1.64092138e-01 3.12913924e-01
4.69015062e-01 9.03719604e-01 1.04429591e+00 -1.61615205e+00
-5.38950324e-01 -6.79073155e-01 -2.74777204e-01 2.31874332e-01
2.18056843e-01 -8.02515745e-01 -6.45601273e-01 -7.06683397e-01
9.50052679e-01 -7.11562991e-01 -4.16545480e-01 4.86146033e-01
-3.76905710e-01 3.55159461e-01 5.14248729e-01 -7.65113652e-01
9.11198139e-01 -2.29196429e+00 4.47220147e-01 4.33017075e-01
2.59539425e-01 -4.71256733e-01 -1.57726392e-01 1.78643450e-01
3.42589825e-01 -1.87982082e-01 -8.71691823e-01 -1.03346598e+00
-5.72185684e-03 3.69397074e-01 -2.38533393e-01 6.99977398e-01
3.21739167e-01 4.49053556e-01 -9.21717763e-01 -4.14568216e-01
5.79221904e-01 7.57549882e-01 -5.99819064e-01 3.80087614e-01
-3.65018159e-01 9.13722157e-01 -4.69256908e-01 8.87351334e-01
1.28494465e+00 6.66562393e-02 -5.09121895e-01 -3.42588693e-01
-1.98412180e-01 -9.97234434e-02 -1.50425327e+00 2.41920829e+00
-5.99718809e-01 2.86884755e-01 4.57914889e-01 -2.00288624e-01
1.14585984e+00 -1.24421239e-01 5.23203552e-01 -4.35560167e-01
-2.44082093e-01 4.01659697e-01 -5.41563630e-01 -2.40752652e-01
6.56605840e-01 -1.39876842e-01 1.83512688e-01 -1.64258350e-02
-2.71817565e-01 -1.01286435e+00 -2.45842978e-01 2.15333328e-01
9.13679183e-01 4.83565003e-01 -2.74615824e-01 -1.06313288e-01
4.81810093e-01 -2.74976134e-01 9.56126153e-01 3.38521749e-01
3.42590734e-02 1.34203184e+00 2.10987464e-01 -5.35355955e-02
-8.82357478e-01 -1.38276899e+00 -3.51276189e-01 2.10544139e-01
5.28911948e-01 -4.13220041e-02 -3.61763388e-01 -2.01649845e-01
1.86008111e-01 6.71603799e-01 -5.04889846e-01 -6.93306774e-02
-2.85807282e-01 -4.19860423e-01 1.52091190e-01 3.85844499e-01
6.06889665e-01 -4.74087775e-01 -2.46778980e-01 3.07896614e-01
-3.80984396e-01 -1.29557633e+00 -4.41980928e-01 5.81147783e-02
-1.18884444e+00 -9.17503595e-01 -6.89072013e-01 -2.78233737e-01
7.71044791e-01 -1.40518378e-02 9.65196609e-01 -3.42360318e-01
-3.89903545e-01 3.58607769e-01 -2.13009939e-01 -1.84012726e-01
7.06507042e-02 -3.94844651e-01 -1.53994024e-01 2.86685169e-01
-2.77090937e-01 -7.41206765e-01 -8.65540147e-01 2.20323160e-01
-1.07574213e+00 8.18146486e-03 4.85426903e-01 7.95126498e-01
1.03915250e+00 -3.72773558e-02 3.26366484e-01 -4.07393277e-01
6.71944991e-02 -4.10199374e-01 -8.72199535e-01 -1.43390894e-01
-6.83990896e-01 3.18803877e-01 3.22276130e-02 -9.55245569e-02
-1.29541039e+00 3.97354186e-01 -2.19969839e-01 -8.16905677e-01
2.07838669e-01 4.32512790e-01 -3.95304352e-01 -3.10528040e-01
3.41034502e-01 9.55988541e-02 -3.27453241e-02 -5.57344913e-01
3.07709664e-01 1.48769751e-01 7.04942286e-01 -8.65597725e-01
9.87664044e-01 9.42270756e-01 4.11674142e-01 -6.41798437e-01
-7.97495067e-01 -4.96824622e-01 -3.70137841e-01 -2.02233762e-01
7.04882860e-01 -1.43553829e+00 -3.26246977e-01 6.75762117e-01
-1.17042208e+00 -2.37219512e-01 -3.89943570e-01 5.58993995e-01
-5.05461574e-01 8.62372398e-01 -5.31703293e-01 -8.97630274e-01
-2.44565621e-01 -1.23167944e+00 1.40281796e+00 1.85627773e-01
2.23435894e-01 -6.77797437e-01 2.22070053e-01 4.11343366e-01
2.24983424e-01 5.92730880e-01 6.15461171e-01 4.53197002e-01
-1.17750442e+00 -1.25663623e-01 -3.79833162e-01 5.14746785e-01
-1.48316324e-02 1.44385904e-01 -1.17985153e+00 -1.95386603e-01
1.19025009e-02 -2.76191831e-01 1.12665379e+00 5.96474707e-01
1.21211267e+00 1.43076286e-01 -1.25272900e-01 9.69301343e-01
1.76094139e+00 -3.07387620e-01 9.88654613e-01 5.92890801e-03
8.27370107e-01 7.50852764e-01 9.96653318e-01 9.56054091e-01
4.94154066e-01 5.61097562e-01 9.96707261e-01 2.26100922e-01
-1.31534368e-01 -4.37174588e-01 1.61258489e-01 4.69793051e-01
1.25626370e-01 -8.38522390e-02 -7.65743434e-01 6.25262201e-01
-1.84688115e+00 -4.07172322e-01 -1.61938488e-01 2.53530240e+00
6.22077346e-01 1.01523831e-01 -5.38039625e-01 -5.14345467e-02
7.07760751e-01 2.58167595e-01 -7.09453583e-01 1.92989409e-01
-1.73882395e-01 3.17088038e-01 5.54776669e-01 9.73402560e-01
-7.77112663e-01 7.69974530e-01 5.18290949e+00 9.86303389e-01
-7.05074131e-01 5.43433838e-02 5.71798325e-01 -1.13810547e-01
-7.95712769e-01 6.78995252e-02 -7.04002321e-01 4.10024941e-01
2.66038120e-01 1.16711818e-01 1.61750168e-01 7.44063258e-01
4.52295184e-01 -6.20840073e-01 -9.67337370e-01 1.18297994e+00
6.87937019e-03 -1.10948944e+00 -1.12618983e-01 8.56323391e-02
1.17124236e+00 1.25740737e-01 5.55033796e-02 -1.49379790e-01
1.41007423e-01 -8.57481182e-01 8.71783495e-01 8.27617526e-01
8.30403566e-01 -7.69892573e-01 5.42371213e-01 2.14796543e-01
-1.30592525e+00 4.89216708e-02 -5.50730824e-01 4.25799899e-02
4.63311642e-01 1.42383027e+00 -2.93422312e-01 8.06023955e-01
5.21059036e-01 1.00566828e+00 -4.01674211e-01 1.35555899e+00
-6.12362504e-01 -1.33820586e-02 -4.52850133e-01 6.06906295e-01
-1.43567231e-02 -5.36512256e-01 6.86024249e-01 6.69020951e-01
6.00947797e-01 7.63445646e-02 2.20382676e-01 1.10195017e+00
-4.14477959e-02 -1.00040182e-01 -3.21430415e-01 4.90419328e-01
5.31733096e-01 1.13125598e+00 -3.91299188e-01 -3.78479087e-03
-1.61996797e-01 1.03730762e+00 2.07299784e-01 4.71048206e-01
-6.27133667e-01 -3.58435139e-03 9.11171615e-01 2.76112467e-01
5.43595627e-02 -3.02461952e-01 -7.46435761e-01 -1.22016430e+00
2.79427797e-01 -3.79294634e-01 4.48990539e-02 -8.31228614e-01
-1.37869382e+00 3.71668398e-01 -8.95885304e-02 -1.42486346e+00
-2.56876927e-03 -2.52427399e-01 -6.05668783e-01 1.07873476e+00
-1.93313444e+00 -9.36093748e-01 -8.24606478e-01 4.94913608e-01
4.96534467e-01 3.03683192e-01 4.07533586e-01 4.72916514e-01
-4.56743985e-01 1.23151906e-01 -6.14902657e-03 -4.02380794e-01
8.26523423e-01 -1.12527394e+00 1.28897324e-01 9.64659333e-01
-4.65303004e-01 1.74785793e-01 6.94748342e-01 -9.28863227e-01
-1.34963715e+00 -1.25054812e+00 3.57478678e-01 -1.05402827e-01
4.69308555e-01 -9.42192972e-02 -8.93260956e-01 3.38167369e-01
-3.41631651e-01 2.95778394e-01 1.46658897e-01 -4.55275476e-01
-2.86138773e-01 -3.70644331e-01 -1.47334373e+00 3.57611209e-01
9.75714087e-01 -5.22257328e-01 -1.60200685e-01 1.54440090e-01
8.27003539e-01 -7.54534781e-01 -8.36032331e-01 8.68247747e-01
4.47493225e-01 -1.07859862e+00 1.11813498e+00 4.78567362e-01
6.70837045e-01 -6.32149577e-01 -5.15373468e-01 -1.09322011e+00
-8.35806057e-02 -3.08449268e-01 -2.05325097e-01 1.21950424e+00
1.37891129e-01 -4.56630677e-01 9.27887559e-01 8.45674455e-01
-5.67766786e-01 -5.40210724e-01 -1.29823613e+00 -5.17417312e-01
-2.31558606e-01 -6.31658137e-01 3.10614973e-01 4.36238796e-01
-5.82877278e-01 -2.38477722e-01 -4.35317457e-01 6.12879455e-01
1.34325051e+00 -4.85789664e-02 7.66177595e-01 -1.14245749e+00
-2.55880266e-01 -6.62660673e-02 -4.75857586e-01 -1.35212100e+00
1.23963527e-01 -4.92268890e-01 4.93891895e-01 -1.80832124e+00
-1.78402998e-02 -5.96404433e-01 4.11160626e-02 7.00815469e-02
-2.65611380e-01 3.13389510e-01 5.94168203e-03 1.13230444e-01
-2.42992520e-01 1.11337364e+00 1.25018096e+00 -3.80284101e-01
-2.20823497e-01 -2.26236805e-02 -2.66719520e-01 6.23851776e-01
4.42158073e-01 -5.08641362e-01 -5.25641382e-01 -5.87254822e-01
3.22529763e-01 1.97990209e-01 4.19039726e-01 -1.28855240e+00
1.92330226e-01 -2.01388180e-01 3.13294947e-01 -1.07563066e+00
8.51445735e-01 -8.98730218e-01 2.67906427e-01 2.11051911e-01
5.32498881e-02 -3.95021617e-01 9.05396193e-02 9.11424100e-01
-3.84049177e-01 -1.31135792e-01 1.00156724e+00 8.25994983e-02
-7.67883897e-01 8.07376921e-01 2.65675634e-01 3.16284448e-02
8.04771483e-01 -2.62032092e-01 -3.19871083e-02 -4.64224011e-01
-6.73570096e-01 4.47428465e-01 7.93791234e-01 3.27315815e-02
1.24811053e+00 -1.22977555e+00 -1.02930844e+00 2.44030789e-01
4.11332786e-01 6.88875437e-01 5.77932060e-01 7.18802392e-01
-6.48301959e-01 -2.02342525e-01 6.41093999e-02 -9.44705844e-01
-6.93037987e-01 1.50450572e-01 2.28530943e-01 -1.90959677e-01
-6.96920812e-01 9.73562479e-01 1.88564852e-01 -4.29674178e-01
4.44792837e-01 -5.66986918e-01 2.46599361e-01 -2.85835359e-02
2.72011220e-01 5.80182850e-01 3.69106755e-02 -4.23973680e-01
-3.31005007e-01 1.02748168e+00 2.51233906e-01 -3.15031946e-01
1.20544064e+00 -4.77171630e-01 -1.08966030e-01 3.43874484e-01
9.97015715e-01 1.76741675e-01 -1.92661965e+00 -2.82506615e-01
-2.28798524e-01 -1.04653132e+00 3.03819656e-01 -5.57486355e-01
-1.20518041e+00 9.41619873e-01 7.07871497e-01 -6.38618648e-01
1.00003684e+00 -2.77561277e-01 7.65992343e-01 -8.93403590e-02
5.47378242e-01 -1.00653791e+00 2.56675899e-01 3.83984208e-01
1.06717205e+00 -1.58353770e+00 4.39509928e-01 -7.24435627e-01
-4.51845974e-01 9.30102587e-01 5.92375159e-01 -1.98584363e-01
6.60902917e-01 1.92835987e-01 -1.02131829e-01 5.93388593e-03
-3.23629051e-01 -5.30148260e-02 1.81213021e-01 5.82337976e-01
-3.81038301e-02 -6.49177954e-02 -1.08171828e-01 1.57315925e-01
1.92063168e-01 8.94348770e-02 4.79665518e-01 6.98821008e-01
-5.39468527e-01 -8.02125454e-01 -5.62779009e-01 2.06945837e-01
5.90089932e-02 -1.85137451e-01 2.22236738e-01 2.48179510e-01
1.68011874e-01 9.88807976e-01 -1.09029494e-01 -2.79869199e-01
2.13261560e-01 -3.12647641e-01 4.60099041e-01 -7.98781157e-01
5.37945032e-02 2.18774989e-01 -2.06720322e-01 -8.80125940e-01
-2.94874609e-01 -5.89499474e-01 -1.33614445e+00 -1.08046643e-01
-1.08175896e-01 -1.09612800e-01 9.08579350e-01 6.91286623e-01
2.64572084e-01 3.37409735e-01 7.58901834e-01 -1.17369044e+00
-3.51802498e-01 -8.69551420e-01 -7.94364452e-01 1.79989815e-01
4.68470842e-01 -7.11822927e-01 -7.46018946e-01 -4.18131381e-01]
|
[8.88848876953125, -2.6954450607299805]
|
c0d69dd4-5e28-42d9-bb52-c174e8867a72
|
orex-object-reconstruction-from-planner-cross
|
2211.12886
| null |
https://arxiv.org/abs/2211.12886v3
|
https://arxiv.org/pdf/2211.12886v3.pdf
|
OReX: Object Reconstruction from Planar Cross-sections Using Neural Fields
|
Reconstructing 3D shapes from planar cross-sections is a challenge inspired by downstream applications like medical imaging and geographic informatics. The input is an in/out indicator function fully defined on a sparse collection of planes in space, and the output is an interpolation of the indicator function to the entire volume. Previous works addressing this sparse and ill-posed problem either produce low quality results, or rely on additional priors such as target topology, appearance information, or input normal directions. In this paper, we present OReX, a method for 3D shape reconstruction from slices alone, featuring a Neural Field as the interpolation prior. A modest neural network is trained on the input planes to return an inside/outside estimate for a given 3D coordinate, yielding a powerful prior that induces smoothness and self-similarities. The main challenge for this approach is high-frequency details, as the neural prior is overly smoothing. To alleviate this, we offer an iterative estimation architecture and a hierarchical input sampling scheme that encourage coarse-to-fine training, allowing the training process to focus on high frequencies at later stages. In addition, we identify and analyze a ripple-like effect stemming from the mesh extraction step. We mitigate it by regularizing the spatial gradients of the indicator function around input in/out boundaries during network training, tackling the problem at the root. Through extensive qualitative and quantitative experimentation, we demonstrate our method is robust, accurate, and scales well with the size of the input. We report state-of-the-art results compared to previous approaches and recent potential solutions, and demonstrate the benefit of our individual contributions through analysis and ablation studies.
|
['Amit H. Bermano', 'Amir Vaxman', 'Haim Sawdayee']
|
2022-11-23
| null |
http://openaccess.thecvf.com//content/CVPR2023/html/Sawdayee_OReX_Object_Reconstruction_From_Planar_Cross-Sections_Using_Neural_Fields_CVPR_2023_paper.html
|
http://openaccess.thecvf.com//content/CVPR2023/papers/Sawdayee_OReX_Object_Reconstruction_From_Planar_Cross-Sections_Using_Neural_Fields_CVPR_2023_paper.pdf
|
cvpr-2023-1
|
['3d-shape-reconstruction', 'object-reconstruction']
|
['computer-vision', 'computer-vision']
|
[ 3.79203260e-01 3.87509227e-01 4.29416038e-02 -2.66218275e-01
-8.55746746e-01 -4.00202841e-01 4.35346186e-01 1.57927275e-01
-7.89850205e-02 5.81145048e-01 4.57297385e-01 -1.09327868e-01
-2.38368794e-01 -8.41544926e-01 -8.13101470e-01 -4.49090332e-01
-4.13803518e-01 3.73052269e-01 2.66416878e-01 -1.84817277e-02
1.06038101e-01 8.92861962e-01 -1.13505864e+00 4.25610840e-01
9.25507724e-01 1.06880355e+00 -8.40959623e-02 3.92589808e-01
-6.84684813e-02 3.67379576e-01 -2.98947304e-01 7.59478435e-02
3.40752780e-01 -2.18595505e-01 -7.98850298e-01 4.64930609e-02
4.76725847e-01 -3.39669257e-01 -2.60964483e-02 7.71541715e-01
5.52817225e-01 3.65566723e-02 8.75213802e-01 -6.40000403e-01
-4.20911103e-01 1.64485484e-01 -7.24694788e-01 -3.86149362e-02
2.02228934e-01 1.61034092e-01 5.21020234e-01 -1.20841956e+00
8.21526885e-01 8.71745288e-01 1.38531137e+00 2.49048442e-01
-1.58894920e+00 -3.47434819e-01 1.09635673e-01 -5.61868668e-01
-1.46520579e+00 -3.94054741e-01 1.00922453e+00 -6.28150582e-01
7.03116298e-01 4.36644018e-01 7.30267584e-01 6.18940175e-01
1.70777068e-01 4.11311775e-01 8.06791067e-01 -4.65685993e-01
3.25665087e-01 3.12198307e-02 -2.35844150e-01 6.65082455e-01
5.39744459e-02 1.69118151e-01 -3.60261261e-01 -4.21839058e-01
1.57469964e+00 -1.64130181e-01 -5.29164016e-01 -5.84665298e-01
-8.97151113e-01 6.73552155e-01 7.00474620e-01 3.77877653e-01
-6.40431583e-01 8.80181715e-02 3.01458035e-03 -1.51589185e-01
8.81070018e-01 4.71073598e-01 -3.52909505e-01 3.02611619e-01
-1.38757658e+00 1.85603991e-01 6.88260734e-01 6.83371305e-01
8.82277727e-01 -3.96088175e-02 -2.56900370e-01 8.88591111e-01
2.60640234e-01 5.06917015e-02 -2.22435202e-02 -1.26205385e+00
9.52017680e-02 5.17765939e-01 5.61388656e-02 -1.08643579e+00
-6.49591506e-01 -6.56517684e-01 -1.08928168e+00 6.14957333e-01
6.42976224e-01 -2.24880204e-01 -1.08058500e+00 1.56319714e+00
7.31526852e-01 3.21058661e-01 -4.86325711e-01 9.06938076e-01
7.74994075e-01 2.66454577e-01 -6.86527342e-02 -7.19145909e-02
1.17172265e+00 -5.22069931e-01 -3.80986124e-01 -2.30911911e-01
4.24350858e-01 -6.95827544e-01 9.51289475e-01 2.16815144e-01
-1.53527761e+00 -3.28559995e-01 -7.78634310e-01 -1.20426558e-01
-8.31001848e-02 2.39964426e-01 4.77475345e-01 2.77491271e-01
-1.22606194e+00 1.06399488e+00 -9.53763545e-01 -8.09711521e-04
7.02598870e-01 3.83870900e-01 -1.82559684e-01 -1.31277612e-03
-8.50091815e-01 7.05747485e-01 -9.73778144e-02 1.59036502e-01
-3.44741762e-01 -1.49850738e+00 -9.11767960e-01 1.11185022e-01
-6.28000498e-02 -8.78348291e-01 8.46648514e-01 -8.62523437e-01
-1.27317286e+00 9.15151119e-01 -1.30925611e-01 -2.02317774e-01
5.96271634e-01 1.37992352e-01 7.54322633e-02 3.29909205e-01
1.30480781e-01 6.51622951e-01 7.58428395e-01 -1.50110507e+00
-7.10312799e-02 -2.14456737e-01 -2.55095422e-01 5.45318164e-02
-9.21318531e-02 -2.71191686e-01 -5.03141582e-01 -1.10409343e+00
4.05004770e-01 -5.17014086e-01 -6.31470859e-01 4.49669451e-01
-2.98553944e-01 2.94414669e-01 5.35178065e-01 -1.00680566e+00
1.12462223e+00 -2.21124649e+00 -9.62763131e-02 7.70709276e-01
4.02693719e-01 -1.69620737e-01 5.36937118e-02 9.03340131e-02
-1.10413104e-01 8.63472298e-02 -8.15420151e-01 -4.80437815e-01
-4.29649383e-01 -4.45553698e-02 -1.07970648e-01 6.39455616e-01
4.92705107e-01 7.51725376e-01 -7.04438210e-01 -3.56164306e-01
1.26261488e-01 9.94115949e-01 -8.59076858e-01 3.01376302e-02
-9.48556066e-02 7.91461706e-01 -3.76734108e-01 5.86733222e-01
8.54257226e-01 -5.55970371e-01 -5.38523085e-02 -5.57467461e-01
-2.56556720e-01 2.49914244e-01 -1.38003600e+00 1.95676601e+00
-5.56041121e-01 3.01737040e-01 6.39306903e-01 -9.76667702e-01
8.68809283e-01 4.43475872e-01 7.78748512e-01 -4.12818223e-01
-2.94170119e-02 4.22528297e-01 -3.12311977e-01 -2.87446856e-01
1.23238750e-01 -1.84642032e-01 2.84480155e-01 3.14987004e-01
-4.51046042e-03 -4.31276411e-01 -2.06505939e-01 5.52373789e-02
1.04610169e+00 4.01465178e-01 1.95480183e-01 -5.59937775e-01
3.09160143e-01 1.34461448e-01 3.79487336e-01 5.32155156e-01
2.85709471e-01 1.29041290e+00 3.82733554e-01 -5.77302873e-01
-1.11726713e+00 -9.85897899e-01 -6.82068706e-01 4.63285446e-01
-9.09724608e-02 -1.95410848e-01 -8.56483042e-01 -4.76517409e-01
7.35445246e-02 4.41177934e-01 -8.12672257e-01 1.60323054e-01
-8.80753160e-01 -7.00005174e-01 3.09431672e-01 5.45225024e-01
1.82976380e-01 -9.12548304e-01 -5.94334662e-01 4.07971501e-01
1.19892456e-01 -1.00921392e+00 -4.15128857e-01 2.38143370e-01
-1.18856370e+00 -8.05185080e-01 -1.03328693e+00 -6.60184443e-01
1.12743855e+00 -3.27822149e-01 1.34020066e+00 2.01307446e-01
-3.91785502e-01 1.46500856e-01 2.31449261e-01 -1.13386504e-01
-2.34293088e-01 -1.24897905e-01 -2.67071605e-01 -1.50426906e-02
-3.89650851e-01 -1.07769489e+00 -8.63128364e-01 1.15216590e-01
-7.51992285e-01 2.56568938e-01 4.85656381e-01 7.70175993e-01
7.74617732e-01 -2.85529733e-01 4.37548548e-01 -8.73123467e-01
3.59250188e-01 -5.04626036e-01 -5.07588685e-01 -6.85292855e-02
-1.92884073e-01 1.05700761e-01 6.63531899e-01 -4.45955396e-01
-9.43657875e-01 3.14229727e-01 -4.00514454e-01 -5.44722736e-01
-3.16659570e-01 5.69053411e-01 2.10320652e-01 -3.81796181e-01
1.00181234e+00 -5.98567463e-02 2.48465493e-01 -7.00087786e-01
1.67435080e-01 9.61702541e-02 5.68846107e-01 -6.92888558e-01
6.10170424e-01 8.56039941e-01 2.71203786e-01 -8.53438735e-01
-8.04357290e-01 -2.61186302e-01 -6.88872099e-01 -2.86074042e-01
5.50965309e-01 -6.18095875e-01 -3.93648833e-01 1.49732322e-01
-1.28005874e+00 -6.10087454e-01 -6.40571594e-01 3.79780322e-01
-4.71589327e-01 6.78888112e-02 -6.29346609e-01 -5.44752061e-01
-4.24011290e-01 -1.12059772e+00 1.20996904e+00 5.59559725e-02
-4.88911271e-01 -1.26202106e+00 6.41707182e-02 -1.33556545e-01
7.31064379e-01 7.29924142e-01 9.66606140e-01 -1.92436591e-01
-6.04425788e-01 4.06595580e-02 -2.63799429e-01 1.96812645e-01
9.59701762e-02 -1.79588825e-01 -1.01855278e+00 -6.92743361e-02
1.23428807e-01 -1.93642303e-02 5.89149356e-01 9.89340007e-01
1.33303308e+00 -3.43903840e-01 -3.60436380e-01 1.04089355e+00
1.46479440e+00 -3.19827825e-01 5.38495660e-01 4.32103463e-02
7.25744963e-01 7.43511498e-01 -5.14530875e-02 3.97764266e-01
5.61653376e-02 4.84117091e-01 2.76763856e-01 -7.32186198e-01
-4.69945103e-01 -1.00368835e-01 -2.92821556e-01 3.41281384e-01
-2.65902430e-01 4.02344346e-01 -1.02611029e+00 5.85336983e-01
-1.53948951e+00 -5.31599879e-01 -1.93423450e-01 2.31287074e+00
1.08803666e+00 8.10618103e-02 -4.94348705e-02 2.38259826e-02
2.62672484e-01 -5.91762848e-02 -4.81843054e-01 -1.13369599e-01
6.20317906e-02 5.59023380e-01 3.40246052e-01 7.96249092e-01
-9.64652658e-01 4.70648199e-01 6.91179180e+00 8.53511751e-01
-1.44102430e+00 -9.13074240e-02 1.04105222e+00 -1.28290653e-01
-6.55404627e-01 -1.20156705e-01 -3.62119377e-01 3.01191628e-01
4.54777032e-01 3.16243172e-01 3.51184011e-01 5.63792884e-01
3.35862696e-01 -1.46050051e-01 -9.16805923e-01 5.69187105e-01
-1.70040786e-01 -1.92346334e+00 -2.22644672e-01 6.36684289e-03
8.50081861e-01 1.02076426e-01 -1.31324798e-01 -2.11076573e-01
9.76399556e-02 -1.43704343e+00 6.09297812e-01 6.22901499e-01
1.23107886e+00 -5.78901052e-01 2.94279486e-01 4.23208177e-01
-1.16134501e+00 3.91621321e-01 2.36549787e-03 1.16083631e-02
3.55665773e-01 1.05413234e+00 -7.46205449e-01 5.33037066e-01
7.76571572e-01 6.05241776e-01 -1.44363269e-01 1.09987283e+00
-3.90799942e-05 5.64843893e-01 -7.62249768e-01 4.92447734e-01
3.86843160e-02 -2.86502272e-01 6.20056689e-01 1.20348847e+00
2.97789037e-01 3.45893472e-01 1.95253745e-01 1.46615553e+00
3.97963375e-02 6.55363798e-02 -6.29186451e-01 6.82725608e-01
2.37079903e-01 1.28644991e+00 -8.85630488e-01 -2.16050863e-01
-3.57715786e-01 5.67320943e-01 3.74741286e-01 6.75418079e-01
-3.27758372e-01 -1.16828419e-01 3.73589665e-01 7.66459584e-01
3.13115478e-01 -1.68377578e-01 -1.18260479e+00 -7.35132217e-01
2.74490416e-01 -5.23855567e-01 2.45473891e-01 -5.65748930e-01
-1.17624652e+00 5.05837858e-01 5.25349453e-02 -1.13965487e+00
-2.15135291e-01 -3.82366002e-01 -8.25386703e-01 1.24124265e+00
-1.48222649e+00 -1.12333345e+00 -2.72495806e-01 2.97622114e-01
1.25425622e-01 3.28542262e-01 8.32342148e-01 4.38960403e-01
-6.63806200e-02 3.75388980e-01 -1.89959884e-01 1.10392004e-01
3.81952047e-01 -9.82736230e-01 2.89792925e-01 7.33274460e-01
-1.93431184e-01 7.01830328e-01 5.29061258e-01 -8.33505273e-01
-1.06494939e+00 -9.81593907e-01 6.66519105e-01 -3.65057707e-01
3.26758385e-01 -2.99986809e-01 -1.20483494e+00 5.69621444e-01
-9.19220522e-02 5.09440422e-01 3.10492992e-01 -2.03776974e-02
9.01188105e-02 2.07676962e-01 -1.44748521e+00 4.66411382e-01
8.10844719e-01 -1.85991734e-01 -2.36110285e-01 1.24023370e-01
4.31762099e-01 -8.90992880e-01 -1.06678617e+00 7.49808013e-01
5.46490014e-01 -1.03242278e+00 1.35865998e+00 -1.80477127e-01
6.22021973e-01 -2.98450500e-01 2.45646611e-01 -1.22454035e+00
-3.96356463e-01 -7.13051200e-01 -2.69700289e-01 8.90587687e-01
3.98070812e-01 -3.42862129e-01 1.06567824e+00 7.56329358e-01
-4.46998388e-01 -1.34639394e+00 -1.06774926e+00 -3.12640399e-01
3.37655336e-01 -3.72630119e-01 4.25484329e-01 1.07090604e+00
-1.81597382e-01 1.13739166e-02 6.01817295e-02 3.00014347e-01
7.89617002e-01 -2.59004813e-02 4.08658206e-01 -1.23379064e+00
-1.82585746e-01 -4.99191612e-01 1.21714488e-01 -1.15982068e+00
-3.04792345e-01 -8.91368926e-01 6.53845370e-02 -1.66039181e+00
-2.88355678e-01 -9.38995242e-01 2.01366767e-01 4.40383583e-01
-1.01701081e-01 4.68455315e-01 -2.15934262e-01 1.70077354e-01
2.67893225e-01 3.58020335e-01 1.57473230e+00 1.09075636e-01
-5.18679500e-01 -4.80173156e-02 -5.82056046e-01 1.04379320e+00
5.88711739e-01 -3.17036122e-01 -1.65431589e-01 -5.26284695e-01
7.07374960e-02 1.96510479e-01 7.36131728e-01 -8.84108365e-01
2.89641351e-01 -7.20188618e-02 7.55305588e-01 -5.65449536e-01
4.37935978e-01 -8.83128643e-01 3.52637410e-01 1.56471953e-01
-2.18190387e-01 -2.84554511e-01 4.05011535e-01 1.14878323e-02
-5.14569730e-02 -1.41495511e-01 1.01817012e+00 -2.33840942e-01
-7.22278953e-02 3.61777931e-01 -5.69143370e-02 2.23960280e-01
5.71049571e-01 -3.68385941e-01 2.08956346e-01 -2.63353884e-01
-9.45692539e-01 5.29116616e-02 5.51443815e-01 -1.93751231e-01
6.19471967e-01 -1.41897154e+00 -8.19646776e-01 6.62311673e-01
-4.01652336e-01 6.51965201e-01 3.72766703e-01 1.06738985e+00
-5.73833406e-01 -9.77410525e-02 2.00740751e-02 -8.38930428e-01
-6.97840095e-01 1.21174291e-01 6.71969295e-01 -3.70916963e-01
-1.08847034e+00 8.03877234e-01 3.11780572e-01 -5.52562416e-01
2.60785937e-01 -5.53150237e-01 -9.86158028e-02 -1.12640798e-01
3.80821437e-01 1.43552065e-01 2.78991967e-01 -4.23392266e-01
-3.36960196e-01 8.47613394e-01 3.09710711e-01 -1.54983565e-01
1.63648498e+00 2.42205083e-01 -1.20589815e-01 8.89034122e-02
1.25310135e+00 2.99430758e-01 -1.54599369e+00 -1.44795820e-01
-2.64849812e-01 -3.96244407e-01 2.47681782e-01 -7.57260859e-01
-1.17480540e+00 6.29242241e-01 2.30885029e-01 2.18054399e-01
1.00204110e+00 8.87259394e-02 7.40089476e-01 -3.82447690e-01
-8.88286042e-04 -8.30089808e-01 -1.92156345e-01 4.43717510e-01
1.16458070e+00 -8.52388263e-01 2.81490117e-01 -8.10461164e-01
-1.15644991e-01 1.12411106e+00 2.75797457e-01 -4.53789145e-01
1.04099417e+00 8.30162048e-01 3.88256498e-02 -4.49422538e-01
-2.32467473e-01 1.80110022e-01 7.49209762e-01 5.74280918e-01
6.32776856e-01 -3.35937709e-01 -1.19124763e-01 4.72378165e-01
-1.42395779e-01 1.55712739e-01 7.59032890e-02 9.32112098e-01
-1.02518842e-01 -8.91847372e-01 -6.29529536e-01 5.75638771e-01
-5.02716243e-01 -2.16872290e-01 4.11772495e-03 6.22077286e-01
6.19837977e-02 2.52735347e-01 2.95394957e-01 1.09536193e-01
4.41537142e-01 -2.11918265e-01 4.57568079e-01 -5.94022691e-01
-7.06424832e-01 4.32223111e-01 2.07973178e-02 -7.12003291e-01
-1.80788621e-01 -6.73640549e-01 -1.43056607e+00 1.10825256e-01
-4.07434180e-02 -1.10815726e-01 5.69531679e-01 7.22539902e-01
6.22304440e-01 6.76049292e-01 5.09610951e-01 -1.49581933e+00
-2.43185699e-01 -6.89346075e-01 -2.37003759e-01 4.73147959e-01
6.00263059e-01 -5.48230827e-01 -4.68265146e-01 1.24474347e-01]
|
[9.220884323120117, -3.2005867958068848]
|
c67ed0e0-16e2-4965-bcd6-0349b8194567
|
compositional-generalization-without-trees
|
2305.16954
| null |
https://arxiv.org/abs/2305.16954v1
|
https://arxiv.org/pdf/2305.16954v1.pdf
|
Compositional Generalization without Trees using Multiset Tagging and Latent Permutations
|
Seq2seq models have been shown to struggle with compositional generalization in semantic parsing, i.e. generalizing to unseen compositions of phenomena that the model handles correctly in isolation. We phrase semantic parsing as a two-step process: we first tag each input token with a multiset of output tokens. Then we arrange the tokens into an output sequence using a new way of parameterizing and predicting permutations. We formulate predicting a permutation as solving a regularized linear program and we backpropagate through the solver. In contrast to prior work, our approach does not place a priori restrictions on possible permutations, making it very expressive. Our model outperforms pretrained seq2seq models and prior work on realistic semantic parsing tasks that require generalization to longer examples. We also outperform non-tree-based models on structural generalization on the COGS benchmark. For the first time, we show that a model without an inductive bias provided by trees achieves high accuracy on generalization to deeper recursion.
|
['Ivan Titov', 'Alexander Koller', 'Matthias Lindemann']
|
2023-05-26
| null | null | null | null |
['semantic-parsing']
|
['natural-language-processing']
|
[ 8.69514942e-01 6.36634052e-01 7.34770484e-03 -6.42592430e-01
-9.22133923e-01 -1.15992951e+00 3.62430364e-01 6.65098131e-02
-3.23196173e-01 6.88925624e-01 3.93964559e-01 -6.08052015e-01
1.65843338e-01 -9.56188679e-01 -1.01321304e+00 -4.38986242e-01
2.91829649e-02 8.38392138e-01 2.16844067e-01 -1.90605238e-01
1.58257782e-01 1.41729847e-01 -1.52974963e+00 7.24002123e-01
7.15830445e-01 5.47938824e-01 3.50248754e-01 8.20765138e-01
-6.76768005e-01 6.32669866e-01 -5.38127840e-01 -3.62529516e-01
1.80039644e-01 -6.12915456e-01 -1.40177512e+00 -2.84203947e-01
4.99601156e-01 1.39488995e-01 3.15215528e-01 1.09986556e+00
1.91627771e-01 1.34917036e-01 3.00732911e-01 -9.31495190e-01
-7.53146350e-01 1.26406610e+00 -3.54139432e-02 1.11524403e-01
3.82240772e-01 -9.72614717e-03 1.52443218e+00 -5.35261691e-01
8.14259529e-01 1.41555679e+00 8.28626215e-01 1.09947741e+00
-1.59400284e+00 -5.65411806e-01 6.31968558e-01 -2.25885317e-01
-5.60617805e-01 -1.22991793e-01 3.20509493e-01 -1.99854285e-01
1.47067320e+00 4.27380621e-01 3.66777390e-01 1.24961972e+00
-2.60889888e-01 8.32411766e-01 1.00663412e+00 -3.52808505e-01
3.72977108e-01 -3.22741747e-01 4.92013544e-01 5.83660185e-01
2.32676528e-02 4.79228795e-03 -5.56741893e-01 -1.35152265e-01
3.73287976e-01 -3.87268335e-01 1.56541258e-01 -3.81522626e-02
-1.06283414e+00 7.36743212e-01 3.98669928e-01 2.49579802e-01
-1.06774541e-02 4.42155808e-01 6.33902013e-01 2.81514764e-01
4.22787696e-01 8.97267401e-01 -1.06358182e+00 -2.87070096e-01
-6.36519074e-01 3.78895968e-01 1.04822886e+00 1.00365067e+00
7.39901483e-01 -1.65284023e-01 -1.19441569e-01 7.89904356e-01
-1.96477368e-01 1.70484290e-01 6.52557492e-01 -1.27994931e+00
5.47645271e-01 3.54113907e-01 -9.92718264e-02 -2.49984339e-01
-6.10104680e-01 -5.24960995e-01 -2.31521964e-01 -1.99056312e-01
6.00807786e-01 -2.71390259e-01 -1.09119833e+00 2.31417942e+00
1.59178078e-02 3.32367182e-01 3.78695130e-01 6.73452854e-01
6.66364014e-01 7.79230773e-01 6.92190528e-01 1.08285911e-01
1.40710509e+00 -1.06440473e+00 -2.13295341e-01 -7.46020257e-01
1.07106996e+00 -3.88852805e-01 1.24336457e+00 4.53570932e-01
-1.21082067e+00 -3.46302539e-01 -7.27133155e-01 -3.24115098e-01
-4.90591437e-01 -3.68292719e-01 1.10504627e+00 6.28790200e-01
-1.18869340e+00 9.30341244e-01 -9.11454678e-01 -4.30395246e-01
4.04705137e-01 4.41265106e-01 -1.34796739e-01 -2.85217427e-02
-1.13901305e+00 7.59517491e-01 6.68641150e-01 1.64344441e-02
-4.93978530e-01 -1.09710777e+00 -1.01653099e+00 2.02900529e-01
3.44041586e-01 -1.04797864e+00 1.78338635e+00 -1.36506760e+00
-1.53242791e+00 1.07120049e+00 -5.68635404e-01 -5.57947755e-01
1.30895719e-01 -2.13167563e-01 -2.77664214e-02 -3.24994326e-01
2.68857032e-01 1.08443666e+00 1.55806661e-01 -1.16355395e+00
-5.95716417e-01 -3.32454056e-01 3.38074446e-01 2.33095825e-01
4.11946662e-02 1.79858804e-01 -1.39988676e-01 -5.14119148e-01
4.81343806e-01 -8.93216014e-01 -4.94931579e-01 -6.39008939e-01
-6.00965858e-01 -3.39601517e-01 3.26005787e-01 -4.23194140e-01
7.45530009e-01 -1.89969242e+00 3.70826095e-01 -6.45619184e-02
-9.54301208e-02 1.06620602e-01 -4.00420755e-01 3.60343039e-01
-2.60469556e-01 5.00184715e-01 -6.49449944e-01 -4.84615326e-01
3.71519119e-01 7.61114180e-01 -6.65958941e-01 -2.32960492e-01
5.33124864e-01 1.14751983e+00 -1.30950677e+00 -9.69486013e-02
-3.46816391e-01 -8.13705772e-02 -1.15511322e+00 -2.96338573e-02
-9.03371871e-01 3.58790755e-01 -4.21585321e-01 3.63683283e-01
5.57812989e-01 -3.11239511e-01 5.27070403e-01 2.50234365e-01
-5.51843010e-02 1.06965899e+00 -8.43407452e-01 2.03841639e+00
-6.26754522e-01 1.68077454e-01 -2.60685503e-01 -1.25324905e+00
6.50442064e-01 -9.75750014e-02 1.33102424e-02 -4.53418106e-01
-2.22568408e-01 3.14525127e-01 1.19466975e-01 -3.56578559e-01
5.34579813e-01 -6.05762720e-01 -4.48881567e-01 5.33626139e-01
2.07081854e-01 -1.10683948e-01 2.05299869e-01 1.62853584e-01
1.33457172e+00 5.82231164e-01 -9.22597796e-02 -4.15708423e-01
3.05213660e-01 3.47391397e-01 8.62357080e-01 9.56814647e-01
4.62934464e-01 5.51752567e-01 9.69545782e-01 -4.10252869e-01
-1.03111577e+00 -1.20717382e+00 1.88191924e-02 1.58189487e+00
-1.09665364e-01 -5.01076460e-01 -9.20619369e-01 -9.24540281e-01
2.69754753e-02 1.03325105e+00 -5.82503200e-01 1.07555976e-03
-9.94294167e-01 -9.93646145e-01 8.26786280e-01 9.71703529e-01
1.11187465e-01 -1.34050512e+00 -4.39331532e-01 4.73333836e-01
-1.26429290e-01 -1.24699855e+00 -1.61853075e-01 9.13279653e-01
-1.10196137e+00 -8.23768318e-01 -6.96045160e-02 -1.12703240e+00
7.31098235e-01 -4.11263138e-01 1.51536620e+00 -2.45750006e-02
-2.23696917e-01 1.80539489e-03 -2.68660247e-01 -3.06019902e-01
-5.86943328e-01 4.68473673e-01 -4.15529042e-01 -6.22588396e-01
5.40155709e-01 -6.28655434e-01 -2.15823919e-01 -1.27092317e-01
-7.43693471e-01 3.89306754e-01 4.38080847e-01 1.01825869e+00
5.81018746e-01 -2.47411206e-01 7.01642096e-01 -1.65936685e+00
3.51869196e-01 -6.49863183e-01 -5.64615309e-01 1.68018073e-01
-1.61197528e-01 4.85800266e-01 9.12707031e-01 -2.34529138e-01
-1.09195411e+00 3.07350844e-01 -2.21001297e-01 2.52981991e-01
-5.10013103e-01 4.44614410e-01 -2.22149953e-01 4.77290511e-01
5.27701139e-01 1.64776474e-01 -3.06505114e-01 -5.72970569e-01
5.40037453e-01 1.14460476e-01 6.77047968e-01 -1.28546751e+00
6.63002253e-01 1.41690835e-01 1.83845282e-01 -4.64365512e-01
-1.22652197e+00 -1.24549940e-01 -6.01433337e-01 5.73097289e-01
1.00668895e+00 -5.99072516e-01 -7.38844991e-01 2.81794786e-01
-1.33039486e+00 -7.22365737e-01 -4.81923074e-01 5.10504767e-02
-6.69387221e-01 2.96585083e-01 -7.61730254e-01 -5.52906632e-01
-1.38515964e-01 -8.43230605e-01 1.17211783e+00 -8.23774338e-02
-5.30097008e-01 -1.16178811e+00 -9.79812220e-02 3.46478224e-01
3.63087565e-01 -2.11695256e-03 1.32161427e+00 -1.13608766e+00
-6.22725427e-01 3.74216735e-01 -1.46846816e-01 2.82800376e-01
-1.31572977e-01 -4.88101691e-01 -1.07143831e+00 1.45584941e-01
-1.02721192e-01 -4.10747141e-01 1.02576363e+00 1.49064949e-02
1.60719085e+00 -2.92131215e-01 -2.33123034e-01 7.80660689e-01
1.28374696e+00 -3.08340527e-02 2.89128125e-01 5.32325745e-01
5.95585287e-01 7.35715926e-01 2.15439498e-01 -1.12340905e-01
4.50247705e-01 2.11232126e-01 4.04388428e-01 1.31161317e-01
9.61396545e-02 -5.90089619e-01 3.87849838e-01 3.92778009e-01
2.42244601e-01 -3.42914492e-01 -1.01527548e+00 5.80541670e-01
-1.77461660e+00 -7.78306544e-01 -2.24797726e-02 1.76869249e+00
9.89051402e-01 2.43594632e-01 -1.11905642e-01 -1.16865471e-01
5.91594279e-01 3.42135318e-02 -6.55193031e-01 -8.70740950e-01
-1.66544214e-01 9.63671029e-01 6.94336891e-01 6.49163604e-01
-9.19377387e-01 1.69098961e+00 6.68223953e+00 2.89829224e-01
-9.03549492e-01 -1.02014728e-02 5.66076338e-01 -1.52531326e-01
-7.95201182e-01 2.86568642e-01 -9.85951543e-01 2.38234907e-01
1.05883217e+00 3.34368423e-02 7.06612229e-01 5.63798189e-01
-2.52690881e-01 8.51043314e-02 -1.61999762e+00 2.26660997e-01
-1.21062122e-01 -1.22746134e+00 2.29007199e-01 -3.30843598e-01
5.82720101e-01 1.69365004e-01 -1.72517374e-02 4.23147500e-01
1.02941704e+00 -1.39388108e+00 5.82699001e-01 3.72120701e-02
3.19369286e-01 -4.13791269e-01 5.28108895e-01 3.10877711e-01
-8.11405241e-01 -2.96791613e-01 -3.95957023e-01 -3.85176688e-01
2.03723714e-01 1.55628279e-01 -7.22425342e-01 4.31071937e-01
5.94971240e-01 7.08841443e-01 -3.12178463e-01 5.77601790e-01
-6.05462551e-01 8.92887890e-01 -3.58240157e-01 -6.89707324e-02
5.59091747e-01 -1.16890140e-01 3.12881857e-01 1.49849486e+00
2.82660246e-01 3.45656425e-01 1.19003475e-01 1.04274917e+00
-5.10176308e-02 -5.46093434e-02 -3.42522651e-01 -7.59400055e-02
4.18955594e-01 9.59813356e-01 -6.12610579e-01 -4.12180245e-01
-3.92967492e-01 1.03448486e+00 6.09886765e-01 4.03443098e-01
-4.30812061e-01 -8.73041898e-02 7.92999089e-01 -1.93589076e-01
6.04070842e-01 -4.55989391e-02 -6.53401196e-01 -1.16924846e+00
1.07302302e-02 -7.16325581e-01 6.76321507e-01 -8.16335261e-01
-1.31265199e+00 4.66047078e-01 -5.29950112e-02 -4.11146760e-01
-4.16359484e-01 -9.75910604e-01 -6.19746983e-01 1.16967297e+00
-1.64063990e+00 -1.09636223e+00 1.97099626e-01 -3.67962606e-02
7.27389216e-01 2.03639686e-01 1.15281618e+00 -1.64133891e-01
-4.64916527e-01 5.29807746e-01 -4.18393493e-01 7.21071810e-02
3.25718939e-01 -1.74433434e+00 1.09738457e+00 8.69892836e-01
-7.21565075e-03 8.30243647e-01 8.21831703e-01 -5.19591212e-01
-1.37393844e+00 -1.19954407e+00 1.32630837e+00 -5.95201194e-01
9.01413441e-01 -7.96191990e-01 -1.00597286e+00 1.28794003e+00
-6.40711188e-02 -2.34771222e-02 6.61169648e-01 5.86579382e-01
-7.04300880e-01 3.06858540e-01 -8.21119726e-01 3.83503735e-01
1.85871291e+00 -4.95605677e-01 -1.07563150e+00 3.71978611e-01
1.07134306e+00 -6.94514275e-01 -5.73925972e-01 2.50413209e-01
2.92677194e-01 -5.69600224e-01 9.05019164e-01 -1.40663409e+00
7.56913066e-01 1.03377271e-02 -3.45461130e-01 -1.47613013e+00
-3.58242512e-01 -5.44178843e-01 4.57139969e-01 1.18620288e+00
9.28421497e-01 -8.27707589e-01 1.13509905e+00 5.51133096e-01
-5.45788884e-01 -6.11088276e-01 -6.61699891e-01 -9.25422192e-01
6.19556606e-01 -6.19291842e-01 9.03065205e-01 8.56857598e-01
2.03915581e-01 3.25847656e-01 2.12956041e-01 3.90831769e-01
5.22975624e-01 4.24436092e-01 3.80389571e-01 -1.18076181e+00
-6.11828923e-01 -5.17498910e-01 -2.13283911e-01 -1.30574310e+00
8.20377767e-01 -1.58171225e+00 2.61302859e-01 -1.41903687e+00
3.10603857e-01 -7.79184759e-01 -4.03404266e-01 8.18326235e-01
-2.87973702e-01 5.24456464e-02 -6.24795482e-02 -2.74380535e-01
-4.99998391e-01 1.20328054e-01 9.84346449e-01 7.30511621e-02
7.70831108e-02 -2.09092900e-01 -1.16205478e+00 7.06338823e-01
9.84884858e-01 -6.80759847e-01 -4.54401523e-01 -8.85906160e-01
6.37090743e-01 -1.51691427e-02 3.59315991e-01 -5.38633406e-01
1.93311810e-01 -2.06853047e-01 2.18751337e-02 -2.21012160e-01
1.76730409e-01 -4.03623372e-01 -7.36181810e-02 4.91898596e-01
-7.78063238e-01 1.79109141e-01 4.64794189e-01 3.40488642e-01
-2.22477280e-02 -4.80360419e-01 3.23820651e-01 -4.73767966e-01
-7.63160527e-01 -1.55332591e-02 -3.54098082e-01 5.77164531e-01
5.05172551e-01 -1.48017302e-01 -5.61276555e-01 1.04362451e-01
-1.11018288e+00 3.55689853e-01 5.24689496e-01 4.96641785e-01
9.06645209e-02 -7.80541182e-01 -5.25207698e-01 1.75427303e-01
-8.25677589e-02 3.49418551e-01 -2.41577812e-02 2.15465680e-01
-3.72186273e-01 4.74756718e-01 -4.36846763e-02 -4.15581405e-01
-1.07839823e+00 4.63891327e-01 3.84034216e-01 -2.14895561e-01
-5.84180772e-01 1.06153440e+00 5.37357032e-01 -1.10456598e+00
1.15181960e-01 -7.53354907e-01 8.01057294e-02 -1.85544401e-01
1.27835795e-01 3.27947810e-02 6.56845048e-02 -2.71329612e-01
-3.20513129e-01 6.04489565e-01 -3.69524285e-02 -1.09095156e-01
1.59635663e+00 3.13545883e-01 -3.21766138e-01 5.06663263e-01
1.18835402e+00 -1.50153309e-01 -1.00272250e+00 -1.95738345e-01
4.68146414e-01 1.03929743e-01 -4.11584973e-01 -1.09775186e+00
-7.42482185e-01 7.66328335e-01 -2.47883141e-01 3.32132205e-02
8.99465740e-01 3.99833411e-01 9.80791271e-01 6.43893898e-01
1.89456984e-01 -8.56044710e-01 -1.83307022e-01 1.02823019e+00
4.18314517e-01 -9.28192616e-01 -5.53329349e-01 -7.81618774e-01
-5.46448827e-01 1.13378000e+00 6.65515423e-01 -1.51212007e-01
1.79161772e-01 6.10406041e-01 -2.95710261e-03 -1.26622587e-01
-1.17916059e+00 -1.22781552e-01 -1.34754047e-01 4.83602524e-01
5.94745815e-01 1.67089567e-01 -2.75797009e-01 9.39006329e-01
-8.94957542e-01 -9.20629799e-02 3.80776227e-01 9.34550464e-01
-5.05362630e-01 -1.55157995e+00 3.15833613e-02 3.45908582e-01
-7.00245321e-01 -5.06717622e-01 -3.79706085e-01 6.09476328e-01
1.61426917e-01 6.11553252e-01 4.10893500e-01 -1.08300172e-01
3.15823019e-01 6.59608424e-01 5.80933750e-01 -1.34086418e+00
-9.36437011e-01 -4.51617837e-01 5.21375060e-01 -6.36197746e-01
-1.02018632e-01 -1.14127898e+00 -1.68069458e+00 1.28859475e-01
1.25004649e-01 3.22948188e-01 6.78214729e-01 9.50974107e-01
3.33160877e-01 6.82422101e-01 1.95965335e-01 -6.09155953e-01
-6.65368855e-01 -6.13688767e-01 -1.72242656e-01 5.77533126e-01
8.98222327e-02 -8.42960775e-02 -4.60622281e-01 5.88105060e-04]
|
[10.506134986877441, 9.148557662963867]
|
bbdbaf74-3905-4d3b-b594-8fa4c18a6d97
|
visualizing-representational-dynamics-with
|
1906.09264
| null |
https://arxiv.org/abs/1906.09264v2
|
https://arxiv.org/pdf/1906.09264v2.pdf
|
Visualizing Representational Dynamics with Multidimensional Scaling Alignment
|
Representational similarity analysis (RSA) has been shown to be an effective framework to characterize brain-activity profiles and deep neural network activations as representational geometry by computing the pairwise distances of the response patterns as a representational dissimilarity matrix (RDM). However, how to properly analyze and visualize the representational geometry as dynamics over the time course from stimulus onset to offset is not well understood. In this work, we formulated the pipeline to understand representational dynamics with RDM movies and Procrustes-aligned Multidimensional Scaling (pMDS), and applied it to neural recording of monkey IT cortex. Our results suggest that the the multidimensional scaling alignment can genuinely capture the dynamics of the category-specific representation spaces with multiple visualization possibilities, and that object categorization may be hierarchical, multi-staged, and oscillatory (or recurrent).
|
['Marieke Mur', 'Tim Kietzmann', 'Nikolaus Kriegeskorte', 'Baihan Lin']
|
2019-06-21
| null | null | null | null |
['object-categorization']
|
['computer-vision']
|
[ 3.27873200e-01 -4.91899908e-01 3.66602361e-01 -4.82771963e-01
3.19983102e-02 -1.07266140e+00 9.68773067e-01 3.47314894e-01
-2.29439124e-01 6.05796017e-02 4.53380495e-01 -1.06257200e-01
-6.99850559e-01 -3.01572263e-01 -2.04003885e-01 -6.75993860e-01
-5.67826867e-01 1.03628509e-01 -1.09113425e-01 -2.60639697e-01
7.65532374e-01 9.17492151e-01 -1.60951579e+00 6.66569889e-01
4.38356608e-01 8.68638873e-01 2.87633613e-02 5.21934986e-01
-1.95457228e-02 2.91671991e-01 -5.43244243e-01 1.62951753e-01
1.93975538e-01 -4.81563509e-01 -5.71525455e-01 -4.81924087e-01
5.52794039e-01 2.53738254e-01 -2.31057689e-01 1.04003739e+00
2.51300097e-01 1.45130545e-01 1.02736211e+00 -1.25298798e+00
-9.22181368e-01 2.60842741e-01 -6.99900270e-01 9.25357342e-01
2.05865070e-01 -6.64689913e-02 6.86708510e-01 -9.94022667e-01
9.26987767e-01 1.50469792e+00 2.64696509e-01 2.49670550e-01
-1.89228761e+00 -4.53586161e-01 5.45719713e-02 2.76614159e-01
-1.31538868e+00 -1.22935563e-01 7.07191169e-01 -1.23203325e+00
8.76963377e-01 4.14554328e-01 9.70956624e-01 1.05464411e+00
5.59709013e-01 2.66391933e-01 1.62434065e+00 -6.37570471e-02
3.28694999e-01 -2.20958129e-01 5.58954477e-01 1.90444142e-01
1.65118828e-01 -7.12024122e-02 -5.45663357e-01 -3.23171020e-01
8.55353117e-01 2.15942413e-01 -1.15055129e-01 -3.86536896e-01
-1.46886432e+00 8.01776469e-01 5.74142992e-01 5.99323571e-01
-4.00428683e-01 6.52526468e-02 4.92665201e-01 4.02108908e-01
4.39862967e-01 8.66874039e-01 -6.05360419e-02 -2.77975947e-01
-9.83763218e-01 1.14611797e-01 -5.18344864e-02 2.25404710e-01
5.16568601e-01 6.34650290e-02 -8.75122026e-02 8.65422666e-01
5.11082523e-02 -3.75268050e-03 1.00420761e+00 -1.12827706e+00
-3.08665959e-03 9.26228404e-01 -4.63034779e-01 -1.43992877e+00
-9.09848034e-01 -2.90675610e-01 -1.09044921e+00 4.18753147e-01
4.44635808e-01 4.37937289e-01 -3.97404671e-01 1.97825754e+00
4.90115536e-03 -7.16328472e-02 -3.92870992e-01 1.03111649e+00
3.92739326e-01 3.97939175e-01 1.74941987e-01 -3.24528456e-01
1.52974176e+00 -1.60400927e-01 -7.64099121e-01 1.55206844e-01
5.58165133e-01 -2.97547907e-01 9.41497087e-01 1.40209228e-01
-1.08488691e+00 -7.58196115e-01 -1.11220360e+00 -8.99378061e-02
-5.42758524e-01 -5.19335009e-02 5.81038356e-01 2.58157045e-01
-1.00847530e+00 8.80668342e-01 -1.08959293e+00 -5.69252551e-01
5.24375081e-01 -2.85983961e-02 -6.76368475e-01 3.47283363e-01
-8.18485737e-01 9.43713486e-01 1.40601814e-01 3.74004245e-02
-6.63908899e-01 -7.70908952e-01 -4.59566593e-01 6.37736246e-02
-7.12104917e-01 -4.35840011e-01 4.57374752e-01 -7.38490939e-01
-1.24979138e+00 1.08922935e+00 -9.89211798e-02 -2.16855798e-02
-2.83968616e-02 2.63266385e-01 -3.30294698e-01 2.19459087e-01
-2.29615495e-01 5.49756110e-01 6.71363294e-01 -9.31688190e-01
1.62075981e-01 -1.04205251e+00 -4.35044408e-01 1.66002139e-01
-3.64086568e-01 1.91842422e-01 6.51779056e-01 -8.70193839e-01
8.60483527e-01 -8.36301386e-01 1.28549755e-01 3.03930879e-01
4.60143238e-02 -9.36294869e-02 6.15524828e-01 -5.81640959e-01
1.37315071e+00 -2.73654819e+00 6.65407479e-01 2.92822748e-01
4.32992876e-01 -9.98596624e-02 -3.23172867e-01 5.03392637e-01
-7.18351603e-01 2.98177391e-01 -3.89082074e-01 2.20411811e-02
-1.27731547e-01 -1.63872689e-01 -5.21461189e-01 9.78596509e-01
2.49630466e-01 9.30746257e-01 -7.39404857e-01 5.40332720e-02
-9.77218375e-02 5.46762824e-01 -4.04014319e-01 -6.63652867e-02
2.98070371e-01 6.30257010e-01 9.08646584e-02 2.09639043e-01
7.46834993e-01 -5.08821271e-02 3.95996809e-01 -4.59958285e-01
-4.83478904e-01 1.31430224e-01 -1.01160955e+00 1.79312027e+00
1.00258514e-01 1.37071896e+00 -3.23486477e-01 -1.17682123e+00
1.14425659e+00 -1.08709909e-01 4.95642364e-01 -9.28237379e-01
8.99039581e-02 -3.54610980e-02 5.70000172e-01 -2.83038378e-01
1.74798697e-01 2.18246013e-01 1.60472423e-01 7.09060311e-01
2.57353961e-01 1.17287949e-01 -2.68964656e-02 1.44445673e-01
9.58824277e-01 -1.00350052e-01 4.06528562e-01 -6.85514987e-01
-1.01325689e-02 -1.61329851e-01 7.41421059e-02 2.68675119e-01
-1.94927141e-01 6.43211246e-01 9.93425250e-01 -4.39325482e-01
-1.06812048e+00 -1.26858687e+00 -6.32518649e-01 9.59432065e-01
-8.97424892e-02 -4.42510843e-01 -6.06650770e-01 3.77648026e-01
9.30020411e-04 5.56544423e-01 -1.07450163e+00 -5.54679334e-01
-4.12976563e-01 -8.01266193e-01 5.34045994e-01 3.56271416e-01
-3.06832567e-02 -7.54432380e-01 -9.71748888e-01 1.10204384e-01
2.49015093e-01 -6.73330486e-01 -1.50413305e-01 1.95889682e-01
-1.07471490e+00 -9.37657237e-01 -4.29453641e-01 -4.90373284e-01
5.76744795e-01 3.70172620e-01 5.27391613e-01 -3.98082137e-01
-9.03610051e-01 2.44908750e-01 7.14123100e-02 6.25783354e-02
-2.28356086e-02 -4.85291421e-01 3.97670239e-01 1.19862445e-01
1.30178779e-01 -1.05350971e+00 -9.75283563e-01 3.48598510e-01
-9.79130268e-01 3.59304734e-02 1.29920751e-01 4.36450183e-01
6.79709136e-01 -2.96723634e-01 4.72900480e-01 -3.04595262e-01
1.21598327e+00 -7.81623185e-01 -2.93780535e-01 1.61902979e-01
-4.92045939e-01 9.42519903e-02 3.70066911e-01 -7.05703855e-01
-7.09839940e-01 -2.77895659e-01 6.55458331e-01 -4.36023653e-01
-2.24191502e-01 5.63244998e-01 4.73979592e-01 4.48254980e-02
1.02008832e+00 3.13043267e-01 2.67875522e-01 -4.86708164e-01
6.57705963e-01 5.28107643e-01 4.11528736e-01 -4.20737624e-01
4.04988855e-01 7.30475843e-01 1.72749996e-01 -8.28021586e-01
-1.45031095e-01 -3.06131750e-01 -1.01008987e+00 -2.59576112e-01
1.01715684e+00 -7.97151029e-01 -9.78366256e-01 2.23593101e-01
-1.10666788e+00 -1.57267734e-01 -1.58296362e-01 6.30917549e-01
-4.64182556e-01 2.84770787e-01 -4.41062659e-01 -7.08478332e-01
-2.32655466e-01 -8.20207655e-01 7.01612413e-01 2.95684077e-02
-7.66758561e-01 -7.04343975e-01 6.92332387e-01 -2.70455807e-01
5.61219156e-01 6.22134507e-01 1.29456997e+00 -6.09272718e-01
-6.13454580e-02 -1.43547297e-01 -2.28685081e-01 -1.60223678e-01
-5.68128601e-02 5.81780910e-01 -9.76910293e-01 -2.46961161e-01
1.24290936e-01 8.27328488e-02 5.83518207e-01 5.13970852e-01
1.32166588e+00 -2.02569798e-01 -1.11403957e-01 6.59727335e-01
1.14104128e+00 4.01003391e-01 4.44947541e-01 2.10160464e-01
2.51923740e-01 1.01590252e+00 -1.50014115e-02 3.84776384e-01
7.97787979e-02 7.57704139e-01 2.16405064e-01 2.18490288e-01
1.15015253e-01 -5.22701554e-02 3.00228477e-01 9.51399863e-01
-1.36533648e-01 4.17937964e-01 -8.10258746e-01 2.37244144e-01
-1.83891129e+00 -1.29500592e+00 -4.80646431e-01 2.15036201e+00
3.30320567e-01 -1.67024389e-01 2.68538713e-01 3.21781412e-02
6.28243327e-01 9.66330916e-02 -7.64111936e-01 -7.62236834e-01
-4.54030603e-01 -3.87215093e-02 -1.97932661e-01 3.71879749e-02
-5.19908607e-01 4.02009308e-01 7.23559570e+00 3.31338912e-01
-1.32822871e+00 3.13457362e-02 1.36576220e-01 -3.49133670e-01
-2.62329966e-01 -5.26269227e-02 1.06170274e-01 3.67127538e-01
9.83578324e-01 -4.23663944e-01 8.13573062e-01 6.35849535e-01
2.73461640e-01 -2.71002147e-02 -1.19977403e+00 1.23631740e+00
-1.55277506e-01 -1.45788193e+00 -2.28258930e-02 1.50586218e-01
3.48815769e-01 1.13019451e-01 4.27800983e-01 -1.54773593e-01
-7.26192221e-02 -9.79823828e-01 6.84874296e-01 7.55816460e-01
8.55510652e-01 -1.82559818e-01 -5.73623441e-02 2.31587440e-01
-1.30478585e+00 -1.75938621e-01 -5.09045720e-01 -1.58109441e-01
-1.61925524e-01 2.01379254e-01 -4.58613992e-01 -1.26605947e-02
7.54590392e-01 9.49938774e-01 -9.17792976e-01 9.30307567e-01
3.68662298e-01 2.43527442e-01 4.28158753e-02 1.56412512e-01
-2.60534793e-01 -6.31945550e-01 6.43564403e-01 1.10844696e+00
4.10158038e-01 1.15057521e-01 -5.98424852e-01 1.32743025e+00
5.96928656e-01 2.44240724e-02 -7.42530107e-01 -5.86689174e-01
5.70620000e-01 1.55955040e+00 -9.90504205e-01 -2.34310851e-02
7.94918314e-02 7.21463978e-01 5.59561431e-01 4.75074261e-01
-3.92290682e-01 -1.75304592e-01 1.11804271e+00 1.27244413e-01
-2.99359828e-01 -8.97236466e-01 -4.30818498e-01 -1.15718794e+00
-1.84117436e-01 -4.83179688e-01 4.45250481e-01 -9.71863270e-01
-1.43750274e+00 7.43029356e-01 1.56016290e-01 -1.26266372e+00
9.40313712e-02 -7.03431785e-01 -6.86840534e-01 8.95758927e-01
-7.34647870e-01 -4.84457582e-01 -3.08704525e-01 5.27658522e-01
2.36075506e-01 -4.74568643e-02 1.03837097e+00 5.93194515e-02
-7.03933895e-01 1.57930940e-01 4.01331067e-01 -2.82001376e-01
4.25406605e-01 -1.12875164e+00 3.09200704e-01 3.08481961e-01
2.97967136e-01 1.16524875e+00 6.25486791e-01 -1.24314137e-01
-1.37879097e+00 -5.14783263e-01 1.20626181e-01 -5.61888039e-01
9.09888864e-01 -6.29985511e-01 -1.30910838e+00 3.79831195e-01
7.97138438e-02 -4.48501185e-02 1.22826493e+00 -1.08332880e-01
-7.03651309e-01 8.40663463e-02 -9.03423905e-01 6.96829736e-01
1.24000287e+00 -9.05083001e-01 -6.82063282e-01 3.27632040e-01
3.56943548e-01 1.29960850e-01 -1.29537368e+00 1.70627818e-01
8.38959634e-01 -9.15973604e-01 9.36418056e-01 -8.88254523e-01
2.13609874e-01 -4.80988055e-01 -4.05713350e-01 -1.53614521e+00
-9.29217100e-01 -1.85637131e-01 6.00349307e-02 9.61691022e-01
4.54502366e-03 -8.07207167e-01 -1.23686574e-01 5.20423412e-01
2.02252716e-02 -5.56145191e-01 -1.20621872e+00 -5.80808640e-01
1.06586635e-01 -1.61286384e-01 3.47101271e-01 1.14704919e+00
6.61716700e-01 7.08048418e-02 3.59549135e-01 -1.19462632e-01
3.20072979e-01 2.29385331e-01 2.10145473e-01 -1.48129690e+00
1.87055990e-01 -9.89492178e-01 -9.05580044e-01 -1.88441709e-01
7.44028166e-02 -1.41732991e+00 -6.09129488e-01 -1.33317542e+00
2.82322049e-01 5.42948209e-02 -4.97988403e-01 1.35730952e-01
2.99544394e-01 -9.99977142e-02 2.25840598e-01 8.27772975e-01
-1.85122490e-01 4.81648892e-01 1.10670006e+00 1.20275348e-01
-1.96796641e-01 -4.98640627e-01 -5.44880569e-01 4.51187491e-01
7.35451818e-01 -4.38216984e-01 -4.10764456e-01 -2.94744402e-01
4.16290432e-01 -9.58509594e-02 4.46085304e-01 -8.91002357e-01
1.08486310e-01 -1.06082313e-01 5.42184412e-01 -3.74277622e-01
3.75286251e-01 -6.34981155e-01 5.87105334e-01 5.33929706e-01
-7.22730339e-01 5.59980214e-01 2.75058925e-01 6.16996646e-01
-2.57808626e-01 2.08192959e-01 9.25899208e-01 7.17562363e-02
-4.29964840e-01 1.05227955e-01 -8.18929493e-01 -1.36333957e-01
9.89436805e-01 -4.44793075e-01 -7.52062440e-01 6.75087795e-02
-6.91759348e-01 -3.32736105e-01 4.01872218e-01 6.53451979e-01
6.86460316e-01 -1.68398881e+00 -3.73899966e-01 3.83095384e-01
1.31212398e-01 -7.19260097e-01 6.15457535e-01 9.15155828e-01
-3.44102800e-01 2.40287289e-01 -1.01454067e+00 -8.64891887e-01
-1.07124174e+00 6.15525365e-01 2.99775958e-01 2.48847783e-01
-5.02118528e-01 5.22826552e-01 2.98323542e-01 -1.67779312e-01
-1.49386093e-01 -3.80425751e-01 -8.01984370e-01 8.73110175e-01
5.72536349e-01 5.16476154e-01 -1.82311088e-01 -7.35901296e-01
-4.80806500e-01 9.86899674e-01 2.72720188e-01 -2.26048276e-01
1.46195209e+00 -1.83406785e-01 -6.02162004e-01 1.08041692e+00
1.56920218e+00 -6.58340812e-01 -1.09822583e+00 -5.32667041e-02
9.12780911e-02 -3.84251148e-01 -2.91691154e-01 -2.93375760e-01
-7.65775323e-01 1.34371650e+00 1.06211495e+00 6.34500980e-01
9.17173386e-01 -1.26352474e-01 -1.45104930e-01 2.04599261e-01
5.48803285e-02 -8.97740006e-01 4.06668007e-01 2.34893143e-01
1.53647494e+00 -6.84592545e-01 -4.29092869e-02 1.30117029e-01
-6.35549426e-01 1.54600453e+00 4.67893809e-01 -3.78632933e-01
9.33693767e-01 -1.60674095e-01 -2.20263585e-01 -6.44820333e-01
-7.83177197e-01 2.92531818e-01 6.98124468e-01 4.17503685e-01
6.15944505e-01 2.94536054e-01 -4.19529498e-01 4.82729644e-01
-3.38273823e-01 -5.28982103e-01 3.09568226e-01 6.04793191e-01
-2.77858973e-01 -4.81963396e-01 -2.55311221e-01 4.32169974e-01
7.07383975e-02 1.16943143e-01 -3.95975947e-01 4.30879742e-01
-1.11507595e-01 5.34099400e-01 5.94706297e-01 -5.73357642e-01
3.48119676e-01 3.49985272e-01 4.32589978e-01 -5.91957748e-01
-4.01738614e-01 -9.06530917e-02 -5.05306304e-01 -7.60888696e-01
-3.79809380e-01 -9.01347637e-01 -1.23273349e+00 -7.15747923e-02
7.86432251e-02 -3.85116518e-01 1.07842553e+00 5.41229904e-01
8.25963438e-01 4.66839194e-01 6.63260341e-01 -9.00147915e-01
-1.90951377e-01 -1.00622797e+00 -6.93027556e-01 6.41620159e-01
3.42072606e-01 -8.38458598e-01 -6.32166266e-01 -2.13955026e-02]
|
[7.951205730438232, 3.9365410804748535]
|
ece31ae3-148e-4c9b-8f46-aab13ed5b07b
|
a-little-birdie-told-me-inductive-biases-for
| null | null |
https://aclanthology.org/2020.wnut-1.31
|
https://aclanthology.org/2020.wnut-1.31.pdf
|
“A Little Birdie Told Me ... ” - Inductive Biases for Rumour Stance Detection on Social Media
|
The rise in the usage of social media has placed it in a central position for news dissemination and consumption. This greatly increases the potential for proliferation of rumours and misinformation. In an effort to mitigate the spread of rumours, we tackle the related task of identifying the stance (Support, Deny, Query, Comment) of a social media post. Unlike previous works, we impose inductive biases that capture platform specific user behavior. These biases, coupled with social media fine-tuning of BERT allow for better language understanding, thus yielding an F1 score of 58.7 on the SemEval 2019 task on rumour stance detection.
|
['Vidhisha Balachandran', 'Sharanya Chakravarthy', 'Tushar Kanakagiri', 'Karthik Radhakrishnan']
| null | null | null | null |
emnlp-wnut-2020-11
|
['rumour-detection']
|
['natural-language-processing']
|
[-1.90093666e-01 2.74100661e-01 -5.98264217e-01 -7.58034587e-02
-3.27076018e-01 -6.28797948e-01 1.22435796e+00 8.95187616e-01
-5.14692724e-01 7.79407203e-01 9.76159155e-01 -4.73646194e-01
3.59578699e-01 -8.03180099e-01 -4.35035944e-01 -1.01252295e-01
1.35844080e-02 1.35604233e-01 3.66482735e-01 -5.98739207e-01
7.57533550e-01 -1.19826056e-01 -1.16136980e+00 6.67406857e-01
6.58532858e-01 6.95076227e-01 -7.51353055e-02 3.08079660e-01
-2.63063490e-01 1.39455497e+00 -7.43604720e-01 -6.25409603e-01
-3.42086196e-01 -3.62749726e-01 -8.79056990e-01 -4.92707342e-02
2.79310465e-01 -1.61919519e-01 -1.59455299e-01 1.16315615e+00
3.20556402e-01 -5.04635796e-02 5.43208182e-01 -8.80183041e-01
-6.61696494e-01 1.15427971e+00 -7.92119920e-01 8.01659226e-01
5.75793862e-01 -3.82567406e-01 1.22615778e+00 -7.46390522e-01
8.60843003e-01 1.16225708e+00 7.49372244e-01 3.00351484e-03
-1.21463776e+00 -5.92102647e-01 -2.53084265e-02 -6.97054341e-02
-1.08232212e+00 -4.11126286e-01 5.17317176e-01 -8.18796277e-01
5.14509499e-01 2.78980881e-01 4.56431508e-01 1.55693471e+00
1.93840846e-01 5.83755612e-01 1.43847942e+00 -3.27783853e-01
8.27911496e-02 5.35461545e-01 3.72157156e-01 3.07215840e-01
3.91889125e-01 -3.30171913e-01 -8.00936043e-01 -6.30969107e-01
2.31435359e-01 -7.40216896e-02 -1.22820333e-01 4.19517130e-01
-1.07331777e+00 1.18871176e+00 6.44797981e-01 4.41373438e-01
-7.75068223e-01 -4.81141210e-01 6.80574834e-01 7.03965783e-01
1.11557293e+00 7.56249487e-01 -8.36718455e-02 -3.54253165e-02
-1.08399153e+00 6.21554673e-01 1.13843477e+00 3.74039918e-01
3.68435264e-01 -2.93990374e-01 -2.76474386e-01 9.92749751e-01
2.06042275e-01 4.52928364e-01 5.45091748e-01 -4.20383543e-01
3.46627682e-01 5.43652892e-01 3.76979738e-01 -1.59926987e+00
-5.48397720e-01 -7.27053940e-01 -4.21913177e-01 -2.68309891e-01
2.85375446e-01 -2.85983145e-01 -3.44657779e-01 1.35422647e+00
2.79181004e-01 -2.82808185e-01 -2.93914080e-01 6.66207075e-01
6.24948204e-01 5.72259963e-01 1.89240724e-02 -3.17171514e-01
1.56863046e+00 -5.64144194e-01 -7.98304617e-01 -3.72046947e-01
4.59457338e-01 -1.35182548e+00 8.55436563e-01 2.71452039e-01
-9.59556162e-01 1.40474081e-01 -9.99417007e-01 9.56244394e-02
-2.91129917e-01 -6.63827479e-01 3.00940335e-01 5.06452799e-01
-5.62820196e-01 4.58035737e-01 -2.01626971e-01 -3.72897744e-01
3.92617732e-01 -4.33364749e-01 2.61368930e-01 9.31858420e-02
-1.54504669e+00 1.14078248e+00 1.68946028e-01 -3.40852618e-01
-5.37163854e-01 -6.02373779e-01 -2.71645010e-01 -5.08081436e-01
6.17018938e-01 -3.59174758e-01 1.37075424e+00 -9.22646999e-01
-1.07125974e+00 1.08386815e+00 2.94073541e-02 -8.32668006e-01
8.99159491e-01 -2.83911914e-01 -6.96583629e-01 -1.41653180e-01
4.29524064e-01 -4.49069701e-02 9.23038900e-01 -8.43430161e-01
-6.39056146e-01 -2.24800691e-01 -2.85191536e-02 -2.91644167e-02
-2.84991443e-01 6.91936970e-01 3.01261425e-01 -6.63878381e-01
7.71443173e-02 -8.84465396e-01 1.66415989e-01 -7.64430404e-01
-6.53537452e-01 -1.60811618e-01 6.80510223e-01 -8.86706710e-01
1.48809326e+00 -1.87044048e+00 -2.71911323e-01 2.74932861e-01
5.76864898e-01 5.74299954e-02 5.39305687e-01 7.94290602e-01
3.67686778e-01 5.16942680e-01 2.29255974e-01 -1.56766415e-01
-1.55824333e-01 -1.80280849e-01 -5.49832046e-01 6.01631880e-01
-2.30496913e-01 6.27201438e-01 -1.10735381e+00 -3.66918534e-01
-5.32539964e-01 3.05823743e-01 -5.61564863e-01 -2.17272881e-02
-2.35789970e-01 3.58819604e-01 -4.34031546e-01 3.68869841e-01
4.49451923e-01 -6.06411278e-01 -9.64862294e-03 2.52337426e-01
-6.51308000e-01 1.20185697e+00 -4.63466585e-01 7.05605268e-01
-3.72541577e-01 9.19569314e-01 3.57689917e-01 -2.25296348e-01
7.96359718e-01 3.49045783e-01 3.42831790e-01 -6.10738218e-01
4.14075285e-01 1.08609907e-01 -1.12894349e-01 -4.70205277e-01
7.77851224e-01 -3.58468562e-01 -1.71939582e-01 8.81547153e-01
-5.44124246e-01 1.92555115e-01 5.65795181e-03 7.14685559e-01
6.92279756e-01 -6.39804959e-01 5.81031263e-01 -5.92875719e-01
1.88318491e-01 6.50973320e-02 3.76336947e-02 7.89998412e-01
-2.57641282e-02 2.27780312e-01 8.27499092e-01 -5.80833517e-02
-9.60068107e-01 -4.82353061e-01 -2.31609836e-01 1.34550142e+00
-1.96292996e-01 -4.13990736e-01 -5.61663091e-01 -5.23896337e-01
3.62069666e-01 9.39713717e-01 -6.90971851e-01 7.80129358e-02
-3.42862636e-01 -7.97858000e-01 5.47171474e-01 -1.53583929e-01
5.70784926e-01 -1.00910735e+00 -3.16906989e-01 3.84608924e-01
-7.64446318e-01 -1.09720409e+00 -3.81146044e-01 -2.55329072e-01
-4.94875103e-01 -9.10932660e-01 -7.10476756e-01 -2.87763387e-01
2.04589233e-01 4.40482974e-01 1.31237555e+00 3.84793103e-01
3.28716695e-01 -4.83399332e-02 -4.97077286e-01 -5.59634089e-01
-8.07719588e-01 2.86690801e-01 7.01221451e-02 1.08111590e-01
2.77771801e-01 -5.21758199e-01 -3.91944766e-01 3.20685983e-01
-8.04260552e-01 -3.53462771e-02 8.75423849e-02 5.32721281e-01
-3.28667104e-01 -2.30407834e-01 9.45906937e-01 -1.47667634e+00
1.23188663e+00 -1.21051621e+00 -3.51296753e-01 -5.65879226e-01
-9.49362516e-01 -4.16767836e-01 1.50083572e-01 -3.73439729e-01
-9.65803683e-01 -9.50240135e-01 -5.40105440e-02 6.09285951e-01
1.90143645e-01 1.11858380e+00 6.78466201e-01 1.88246772e-01
1.26149797e+00 -1.26968563e-01 1.24308832e-01 -6.27295136e-01
1.71099439e-01 1.06396663e+00 1.34062782e-01 1.29620479e-02
5.82418561e-01 4.31135058e-01 -7.18899369e-01 -1.17439950e+00
-1.51651800e+00 -7.06133783e-01 -6.75795600e-02 -1.91383779e-01
5.32263041e-01 -9.33725357e-01 -5.69590151e-01 3.35739136e-01
-1.13755143e+00 -1.31133096e-02 4.34731603e-01 1.73434556e-01
1.24438051e-02 2.84293473e-01 -8.74262154e-01 -9.13957655e-01
-2.85319984e-01 -4.07007426e-01 4.12557228e-03 -2.55286574e-01
-1.07789326e+00 -1.03424489e+00 1.51250541e-01 8.33841622e-01
8.68734300e-01 3.69603962e-01 5.53437233e-01 -1.03983903e+00
-1.34312406e-01 -4.22795534e-01 -3.87340307e-01 1.33316681e-01
5.44810109e-02 -2.32160389e-01 -7.90888846e-01 -2.02551991e-01
2.05493554e-01 -3.77467304e-01 8.79970372e-01 1.61801204e-01
3.31367940e-01 -1.00474846e+00 -2.14588001e-01 -3.12573642e-01
8.77972662e-01 -4.91710842e-01 3.66831511e-01 8.20429981e-01
3.25493842e-01 7.16589689e-01 3.58774632e-01 8.02217901e-01
4.56370652e-01 4.82833803e-01 4.86913919e-01 1.36406615e-01
1.87257335e-01 -4.75379825e-01 4.60307986e-01 6.13564491e-01
-1.31273596e-02 -3.25532198e-01 -1.10610390e+00 5.09687781e-01
-1.58525395e+00 -1.18840694e+00 -5.00649214e-01 2.08776879e+00
9.48023081e-01 5.79390943e-01 6.42608166e-01 5.99967130e-02
8.47860813e-01 7.49118745e-01 1.58194751e-01 -3.96876335e-01
-1.63271725e-01 -5.49605310e-01 7.87099659e-01 8.28112662e-01
-8.41386259e-01 6.73171639e-01 6.84001255e+00 5.77278212e-02
-1.23517156e+00 3.04042608e-01 3.19352448e-01 -9.27817374e-02
-3.03940594e-01 -1.87752187e-01 -8.70243609e-01 8.03800464e-01
1.12699187e+00 -5.34980893e-01 4.19024140e-01 5.52135170e-01
5.23131967e-01 -2.87601680e-01 -6.27484143e-01 2.97858387e-01
2.75888532e-01 -1.40837049e+00 -1.30161688e-01 1.55004561e-01
8.76742065e-01 6.44416392e-01 2.76462492e-02 2.83489585e-01
4.14966732e-01 -6.90808594e-01 9.29951668e-01 1.41063511e-01
3.59561332e-02 -4.28979307e-01 7.03788638e-01 8.91004980e-01
-1.18547931e-01 2.18151081e-02 -1.39766678e-01 -5.28467655e-01
5.62061310e-01 1.09959924e+00 -1.42822993e+00 2.91236192e-02
3.87742400e-01 6.49014294e-01 -3.65930200e-01 7.87928939e-01
-3.97874564e-01 1.20015812e+00 -9.52972844e-02 -4.39359903e-01
3.92622888e-01 3.02945107e-01 9.42536354e-01 1.46434867e+00
-1.66121826e-01 -2.80015141e-01 2.62880772e-01 7.56529450e-01
-4.90169108e-01 2.71455944e-01 -4.73004311e-01 -4.67913002e-01
4.99607444e-01 9.92417336e-01 -5.69676936e-01 -4.07851189e-01
-2.44464025e-01 7.57333815e-01 3.08317453e-01 8.81455392e-02
-7.48187244e-01 1.92393929e-01 3.52521092e-01 8.82098973e-01
-2.79205255e-02 -2.39041612e-01 -3.27496111e-01 -1.08964717e+00
-1.94731310e-01 -1.12736392e+00 4.27726507e-01 -2.52549112e-01
-1.41961741e+00 4.14695203e-01 -2.59874254e-01 -6.14803374e-01
-1.39502808e-01 -3.89073752e-02 -5.63598990e-01 8.99810791e-01
-1.64000726e+00 -6.70674980e-01 -1.72121018e-01 2.04365298e-01
3.81589860e-01 1.28033921e-01 4.53079313e-01 5.14685035e-01
-2.48732030e-01 2.10674927e-01 -2.00003430e-01 9.96874645e-02
1.01614618e+00 -9.27653730e-01 4.82805997e-01 3.99676502e-01
-1.87161505e-01 9.53800380e-01 1.45421219e+00 -9.93826509e-01
-7.22458601e-01 -8.82942140e-01 1.45854557e+00 -6.85958982e-01
1.43463635e+00 -2.39261538e-01 -1.09807956e+00 7.26105809e-01
4.27963376e-01 -6.68526828e-01 7.89737344e-01 7.19251573e-01
-7.38914609e-01 5.41489303e-01 -1.05375385e+00 5.57580948e-01
6.90025747e-01 -6.33648098e-01 -8.36337090e-01 6.90515280e-01
4.53900069e-01 -8.86521786e-02 -6.67513669e-01 -1.99267313e-01
4.34727460e-01 -1.02072835e+00 7.09834516e-01 -8.03454101e-01
9.12048340e-01 1.53053105e-01 1.40988097e-01 -1.41424942e+00
-5.23617208e-01 -8.68540108e-01 7.64204375e-03 1.16080177e+00
8.78784180e-01 -9.65121448e-01 5.12960792e-01 3.86414289e-01
2.22788930e-01 -3.06232989e-01 -4.40396845e-01 -4.27537233e-01
9.08892229e-02 -1.21256644e-02 1.35792047e-01 1.43549132e+00
4.65478480e-01 6.60682678e-01 -6.99799955e-01 -2.17396189e-02
5.52485645e-01 3.97523232e-02 6.36156082e-01 -1.44462514e+00
-1.02069721e-01 -7.48451173e-01 -8.42960700e-02 -1.06483793e+00
-2.19410241e-01 -8.60485435e-01 -1.95518062e-01 -1.26590240e+00
3.62421840e-01 -1.98691294e-01 -2.00474814e-01 -5.15952595e-02
5.24780266e-02 3.53380412e-01 -1.21137388e-01 6.09947145e-01
-7.65094519e-01 6.15714751e-02 9.95759249e-01 9.23323557e-02
-1.77172273e-01 1.53876305e-01 -1.13668132e+00 9.55333233e-01
7.94069350e-01 -7.85202026e-01 1.63403705e-01 -3.35700177e-02
1.14393151e+00 -1.10859916e-01 2.65895963e-01 -4.18941706e-01
1.60333112e-01 -1.57352671e-01 -1.46365792e-01 -5.39282620e-01
3.18662636e-02 -2.22152099e-01 -2.87806809e-01 4.44185793e-01
-7.78731585e-01 6.63060099e-02 -1.63535506e-01 6.24860764e-01
-2.47150898e-01 -2.11005926e-01 7.21613526e-01 -1.72697797e-01
-1.38539001e-01 -1.88248500e-01 -9.41131473e-01 3.54406446e-01
5.65651178e-01 9.64710712e-02 -5.95689416e-01 -7.79028833e-01
-6.49581790e-01 1.72875732e-01 4.78941649e-01 6.18183792e-01
3.00802588e-02 -7.39234209e-01 -1.27369750e+00 -1.07882708e-01
1.63083330e-01 -6.75015748e-01 -4.53667007e-02 1.30432081e+00
-1.97339028e-01 2.55914778e-01 -1.36116326e-01 -1.35421872e-01
-1.23376310e+00 2.09746957e-01 1.85504649e-02 -8.69132280e-02
-6.43851340e-01 6.62299752e-01 -3.94522011e-01 -7.07195699e-02
-1.14784099e-01 2.03913257e-01 -5.37890971e-01 7.16813862e-01
9.81171370e-01 7.86468029e-01 1.51512623e-01 -7.68487275e-01
-2.08918720e-01 -5.91907918e-01 -5.40666103e-01 -2.80141294e-01
1.18657851e+00 -4.79918212e-01 -5.23521185e-01 8.80168080e-01
1.10140002e+00 7.55656302e-01 -4.28061992e-01 -5.26389122e-01
5.69032669e-01 -5.48911572e-01 4.14983362e-01 -9.77747977e-01
-3.44155312e-01 1.58997819e-01 -3.72542471e-01 1.22353005e+00
5.04006557e-02 1.96024433e-01 9.04848635e-01 1.30513370e-01
1.37032732e-01 -9.84903991e-01 -4.33387645e-02 9.38760161e-01
1.03979206e+00 -1.33095253e+00 3.88644069e-01 -4.65373874e-01
-7.58867562e-01 6.28731847e-01 -6.47977963e-02 -7.71311671e-02
5.75252473e-01 1.09178767e-01 3.00288558e-01 -5.38620293e-01
-6.97095871e-01 2.58069962e-01 1.80556968e-01 -9.63657275e-02
1.07432497e+00 1.77115574e-01 -7.76986718e-01 4.48964149e-01
-4.58479196e-01 -1.39204025e-01 9.07797337e-01 8.14702094e-01
-9.52712238e-01 -3.52721512e-01 -3.47693115e-01 5.28500497e-01
-1.24581122e+00 -2.05656260e-01 -9.03783917e-01 5.12687624e-01
-5.20693481e-01 1.08298087e+00 -1.58924833e-01 -2.37387225e-01
2.32793484e-02 1.48169190e-01 -1.68174714e-01 -6.99962318e-01
-9.28285897e-01 -1.69874951e-01 8.92027080e-01 -1.55650452e-01
-2.21881598e-01 -9.52025056e-01 -7.93177485e-01 -9.10935760e-01
-5.68650484e-01 3.23475301e-01 7.05741644e-01 9.16179478e-01
2.96214879e-01 2.25363925e-01 6.33246124e-01 -2.72889882e-01
-9.21246231e-01 -1.28119838e+00 -3.50916028e-01 6.14564478e-01
5.66506267e-01 -6.33049607e-01 -6.50555670e-01 -2.45669708e-01]
|
[8.340951919555664, 10.09502124786377]
|
8a701d03-fa97-4b8f-aa88-314a9dc23719
|
dual-variational-generation-for-low-shot
|
1903.10203
| null |
https://arxiv.org/abs/1903.10203v3
|
https://arxiv.org/pdf/1903.10203v3.pdf
|
Dual Variational Generation for Low-Shot Heterogeneous Face Recognition
|
Heterogeneous Face Recognition (HFR) is a challenging issue because of the large domain discrepancy and a lack of heterogeneous data. This paper considers HFR as a dual generation problem, and proposes a novel Dual Variational Generation (DVG) framework. It generates large-scale new paired heterogeneous images with the same identity from noise, for the sake of reducing the domain gap of HFR. Specifically, we first introduce a dual variational autoencoder to represent a joint distribution of paired heterogeneous images. Then, in order to ensure the identity consistency of the generated paired heterogeneous images, we impose a distribution alignment in the latent space and a pairwise identity preserving in the image space. Moreover, the HFR network reduces the domain discrepancy by constraining the pairwise feature distances between the generated paired heterogeneous images. Extensive experiments on four HFR databases show that our method can significantly improve state-of-the-art results. The related code is available at https://github.com/BradyFU/DVG.
|
['Yibo Hu', 'Xiang Wu', 'Huaibo Huang', 'Ran He', 'Chaoyou Fu']
|
2019-03-25
| null | null | null | null |
['heterogeneous-face-recognition']
|
['computer-vision']
|
[-4.14783210e-02 -1.08481109e-01 3.20169926e-02 -2.77686626e-01
-1.06962276e+00 -3.41217488e-01 4.63523835e-01 -7.78554380e-01
7.61283785e-02 7.89801300e-01 2.50655711e-01 2.78213203e-01
-1.45947754e-01 -7.68317282e-01 -7.49304235e-01 -1.10220206e+00
6.19662464e-01 5.48680484e-01 -3.09492916e-01 -1.48359194e-01
-8.05519819e-02 2.48259634e-01 -1.62919986e+00 2.73305327e-01
7.65158832e-01 1.00593531e+00 2.93534279e-01 -1.76423714e-02
1.58109009e-01 5.21100819e-01 -5.00836015e-01 -6.21901035e-01
5.35906851e-01 -5.92446148e-01 -3.43516350e-01 2.71716535e-01
5.52422345e-01 -2.77120233e-01 -6.88234508e-01 1.41367912e+00
8.41762424e-01 1.53875172e-01 6.31099761e-01 -1.63605881e+00
-1.25807548e+00 3.92013550e-01 -6.84013188e-01 -3.18961263e-01
3.09324205e-01 -5.50249703e-02 7.45059609e-01 -1.13371611e+00
9.25642014e-01 1.41452217e+00 4.08681065e-01 8.66929770e-01
-1.35211802e+00 -1.02968597e+00 1.09039478e-01 4.20705199e-01
-1.90627193e+00 -4.89101768e-01 1.04370642e+00 -4.67117846e-01
2.08057985e-01 2.00997859e-01 4.92253333e-01 1.51913846e+00
-3.12651515e-01 6.71530604e-01 9.84506905e-01 -1.60211116e-01
1.43534750e-01 1.73303019e-02 -3.24148327e-01 5.06090522e-01
1.94665805e-01 -4.36292477e-02 -6.05065286e-01 -3.41786355e-01
7.68281221e-01 1.98799133e-01 -4.83804405e-01 -5.22383153e-01
-1.10973024e+00 9.91672456e-01 8.93424079e-02 2.87485003e-01
-4.10161555e-01 -3.25082183e-01 1.60900980e-01 4.69347648e-03
4.83326882e-01 -1.69670895e-01 7.17936233e-02 4.28385466e-01
-6.97380424e-01 3.10258418e-01 4.89624798e-01 1.00298274e+00
8.89215171e-01 2.29054391e-02 -3.53047043e-01 1.12459373e+00
5.13345599e-01 6.83273256e-01 4.64229465e-01 -1.36406553e+00
6.05900228e-01 3.13981295e-01 -8.58568121e-03 -1.18626559e+00
3.51522356e-01 -1.13337345e-01 -1.18760240e+00 7.26628900e-02
1.83765620e-01 1.24880951e-02 -7.57407665e-01 1.97790110e+00
5.75593948e-01 4.73763198e-01 2.99692750e-01 8.61844242e-01
1.01411104e+00 8.57859731e-01 -2.96626300e-01 -3.67723107e-01
1.09919310e+00 -7.86851048e-01 -9.12971377e-01 1.37176827e-01
-1.15218870e-01 -8.13136220e-01 7.38508880e-01 1.62730440e-01
-1.06301355e+00 -6.40461147e-01 -9.23391879e-01 -1.68751642e-01
8.07637051e-02 2.14685962e-01 -2.99138506e-03 4.31634605e-01
-9.39448178e-01 3.82898480e-01 -4.57836121e-01 7.17965141e-02
5.57760239e-01 2.28819340e-01 -7.54792333e-01 -5.53103387e-01
-1.24735832e+00 3.40087891e-01 1.72344595e-01 3.66785824e-01
-9.72847342e-01 -9.49365914e-01 -9.54922855e-01 -1.43811300e-01
2.83144265e-01 -7.92911470e-01 7.86671162e-01 -7.58623898e-01
-1.49404728e+00 9.43532407e-01 -2.04334617e-01 1.46798700e-01
9.86443758e-01 2.96735018e-01 -3.56840730e-01 3.47655006e-02
2.00459108e-01 5.17575443e-01 1.00508082e+00 -1.59683323e+00
-3.65976781e-01 -7.13921845e-01 -4.08805966e-01 1.34709656e-01
-3.01484764e-01 -3.18941087e-01 -8.92238975e-01 -8.63239467e-01
1.35178968e-01 -1.12365198e+00 -3.90670560e-02 -1.85624108e-01
-3.39527100e-01 -2.48895079e-01 8.19329321e-01 -9.33476269e-01
9.02036607e-01 -2.51592731e+00 5.39242327e-01 2.99842209e-01
4.33479488e-01 -2.62460876e-02 -4.81465608e-01 7.38018006e-02
-1.23006061e-01 -8.65836367e-02 -3.87746513e-01 -5.03390133e-01
2.30263054e-01 9.00385231e-02 -2.68455714e-01 5.92924654e-01
-4.50015925e-02 6.67759418e-01 -6.71789706e-01 -5.41343570e-01
5.76982275e-02 9.79984999e-01 -5.07634521e-01 4.73630488e-01
-1.00276008e-01 6.16207540e-01 -3.64592046e-01 6.00426376e-01
1.25076473e+00 -1.02930859e-01 4.08567131e-01 -5.70669293e-01
1.18604168e-01 -5.16780853e-01 -1.33631635e+00 1.60260534e+00
-2.27787971e-01 4.09717947e-01 1.06977761e-01 -9.57657337e-01
1.04458356e+00 4.76433963e-01 7.34082282e-01 -8.45894396e-01
1.52182460e-01 3.36841077e-01 -2.80286610e-01 -3.87005836e-01
2.31506199e-01 -2.49566913e-01 8.20633173e-02 2.26836175e-01
7.89943188e-02 1.78704515e-01 9.01864842e-02 4.08134833e-02
5.52689970e-01 1.05997622e-01 -5.01674414e-02 -9.69595239e-02
5.75756431e-01 -4.87684041e-01 1.12671351e+00 4.48510915e-01
-3.18160892e-01 1.14527988e+00 3.77860487e-01 -3.13120425e-01
-1.21143913e+00 -1.37963057e+00 -1.27293691e-01 4.52882677e-01
3.67898881e-01 -1.23809122e-01 -8.08497965e-01 -5.38627148e-01
1.02540515e-01 2.81416088e-01 -8.63767803e-01 -9.03610960e-02
-4.87203598e-01 -7.65478671e-01 3.78377140e-01 4.81919676e-01
7.65796363e-01 -8.36774111e-01 4.06979233e-01 -1.63441792e-01
-7.11387753e-01 -1.06498587e+00 -6.27721906e-01 -5.91508031e-01
-3.42024297e-01 -8.81602526e-01 -1.27994001e+00 -1.08402431e+00
8.02750766e-01 3.17180455e-01 8.34545672e-01 -1.33329704e-01
-3.50906670e-01 1.87544242e-01 -2.07704052e-01 2.32687443e-01
-3.72239858e-01 -2.84097403e-01 1.38756096e-01 5.08928835e-01
3.04503500e-01 -5.50386429e-01 -5.91324508e-01 4.98029649e-01
-9.40655649e-01 -2.91814818e-03 4.11177307e-01 1.17710090e+00
1.00597107e+00 2.22314540e-02 3.57585430e-01 -7.48337746e-01
4.15365428e-01 -6.07529342e-01 -5.83283901e-01 4.81815279e-01
-1.92322761e-01 -4.93917130e-02 5.27492404e-01 -5.88590264e-01
-1.29890263e+00 -2.53523067e-02 -1.35879129e-01 -1.04415607e+00
-9.40996483e-02 1.18176684e-01 -8.08192134e-01 -5.37935793e-02
3.25873852e-01 3.70251358e-01 1.50931403e-01 -3.59174073e-01
3.47681969e-01 7.72294819e-01 4.33386296e-01 -6.97943270e-01
9.12155449e-01 6.54328287e-01 -1.61895975e-01 -7.40722895e-01
-4.84177470e-01 -1.65376902e-01 -3.07555377e-01 -5.65144271e-02
7.88412929e-01 -1.19796824e+00 -7.89860487e-01 6.27495885e-01
-9.43836212e-01 -2.11439043e-01 -9.18303430e-02 4.36393619e-01
-5.59118092e-01 3.94852072e-01 -4.80175376e-01 -4.95972425e-01
-1.32767633e-01 -1.36285591e+00 1.05140436e+00 2.50617653e-01
2.84736902e-01 -5.88245988e-01 1.61979482e-01 7.03474820e-01
7.68829510e-02 2.37548947e-01 6.82964623e-01 -2.30442956e-01
-6.01680458e-01 1.59516782e-02 -2.49089211e-01 5.08418620e-01
1.99164867e-01 1.55834621e-02 -9.03159022e-01 -4.16999251e-01
4.09399271e-02 -1.48372963e-01 7.32287228e-01 3.93952996e-01
1.13957083e+00 -2.54057318e-01 -1.90600201e-01 7.89510965e-01
1.39077795e+00 3.57076108e-01 7.99878597e-01 1.05563095e-02
9.14709210e-01 8.66693377e-01 4.49791431e-01 6.11530602e-01
3.57061595e-01 8.50733697e-01 4.24787104e-02 1.15238950e-01
-2.47012928e-01 -4.06517059e-01 3.06537777e-01 8.27991486e-01
-8.13880339e-02 -2.54478008e-01 -6.60366416e-01 5.50328851e-01
-1.92551804e+00 -1.22416210e+00 3.31584752e-01 2.16340685e+00
6.76731408e-01 -6.97193325e-01 -5.58486991e-02 -1.87595814e-01
1.37486792e+00 1.91306338e-01 -6.38806522e-01 2.69262344e-01
-4.00202751e-01 -1.84547514e-01 5.81957847e-02 6.76076651e-01
-7.52780259e-01 6.53169453e-01 5.34668827e+00 1.10053599e+00
-8.49104106e-01 2.80761540e-01 7.92933285e-01 -1.80019751e-01
-6.79872274e-01 -3.25231582e-01 -8.71657193e-01 7.10273325e-01
2.79542774e-01 -3.49864036e-01 6.70781493e-01 7.48577237e-01
-1.49449304e-01 5.63270390e-01 -7.76848733e-01 1.52797484e+00
4.40193623e-01 -1.34865630e+00 3.19368035e-01 2.74951100e-01
1.05366707e+00 -2.64111966e-01 5.40928423e-01 1.58345938e-01
2.80469537e-01 -7.50038803e-01 7.59985089e-01 4.29369479e-01
7.95722067e-01 -9.60429370e-01 6.51805282e-01 -5.78794852e-02
-1.20050848e+00 -1.02880811e-02 -5.67649364e-01 5.83387494e-01
3.63132544e-02 5.36744297e-01 -1.92167759e-01 6.99270189e-01
8.55921149e-01 7.76960254e-01 -4.90259945e-01 4.68033522e-01
-5.89069873e-02 5.36769144e-02 -8.10704380e-02 5.47177434e-01
-3.52550268e-01 -7.16122806e-01 5.43525636e-01 4.79533523e-01
5.55613339e-01 2.73644999e-02 7.77652189e-02 9.69260097e-01
-4.91565377e-01 1.25684217e-01 -6.53753996e-01 8.00090358e-02
7.73091435e-01 1.06242812e+00 -1.34386748e-01 -2.22034946e-01
-3.51867259e-01 1.31549346e+00 2.91388214e-01 5.90697765e-01
-8.93627584e-01 -1.70815215e-01 8.07567716e-01 -2.72509187e-01
3.33246768e-01 2.00615957e-01 2.20825803e-02 -1.61470008e+00
4.86741871e-01 -1.17354095e+00 3.76203328e-01 -6.48448110e-01
-1.58382297e+00 7.92140722e-01 -1.98593810e-01 -1.29140592e+00
-2.08383903e-01 -3.17138284e-01 -2.42642328e-01 9.21729147e-01
-1.19934964e+00 -1.33520067e+00 -6.06660426e-01 8.92965853e-01
1.80522755e-01 -4.74537998e-01 6.58328414e-01 7.13662446e-01
-8.55183244e-01 8.69960368e-01 4.11563426e-01 2.49464527e-01
7.87328720e-01 -8.09902072e-01 1.38109624e-01 7.27427363e-01
-9.61477309e-03 7.27195978e-01 5.06766498e-01 -6.33151412e-01
-1.38018465e+00 -1.15387976e+00 4.87190157e-01 -2.23907843e-01
2.63879418e-01 -3.92726570e-01 -9.44331408e-01 7.46081114e-01
1.67173043e-01 3.97573978e-01 9.21705723e-01 -1.68339148e-01
-8.22468877e-01 -2.84906298e-01 -1.29046643e+00 7.47939825e-01
1.18822837e+00 -7.08996236e-01 -2.55359709e-01 6.64581880e-02
6.95872545e-01 -2.01990083e-01 -1.12511170e+00 5.67235470e-01
8.06549966e-01 -9.45976019e-01 1.04479361e+00 -2.46819928e-01
4.33883667e-01 -6.22114182e-01 -6.38945401e-01 -1.26440823e+00
-4.79709715e-01 -4.15478259e-01 1.35099478e-02 1.60155165e+00
1.05709665e-01 -7.35729635e-01 7.59481549e-01 4.90529090e-01
9.97987166e-02 -4.99155730e-01 -9.95895922e-01 -8.77278805e-01
6.04596473e-02 8.38637128e-02 7.85499752e-01 9.83882308e-01
-5.81482112e-01 6.71546087e-02 -5.26655376e-01 3.34422678e-01
1.09548771e+00 3.43032420e-01 6.62926614e-01 -9.92711246e-01
-3.38312745e-01 -1.73809916e-01 -4.19775218e-01 -6.52421534e-01
6.91847503e-01 -7.79070020e-01 1.40546769e-01 -1.19994140e+00
5.90344965e-01 -2.64582098e-01 -2.46402800e-01 2.13710874e-01
-1.49917617e-01 6.83516264e-01 3.55584651e-01 2.66243726e-01
-4.09079045e-01 1.09713781e+00 1.34922302e+00 -2.73709416e-01
-2.05771849e-02 -3.97313565e-01 -7.20758438e-01 3.89630407e-01
6.71613514e-01 -2.55571961e-01 -3.68070453e-01 -5.78832805e-01
-1.88106909e-01 1.09490626e-01 2.44659036e-01 -8.22887123e-01
1.13000974e-01 -2.29847521e-01 5.38582623e-01 -4.86231089e-01
3.93583089e-01 -7.06563592e-01 7.10066617e-01 1.87196955e-01
-1.82181865e-01 -8.68921131e-02 -1.47765707e-02 5.32270014e-01
-5.49577057e-01 1.57652602e-01 1.10681319e+00 3.21887434e-02
-5.79412460e-01 8.05956781e-01 1.50108933e-01 2.54203156e-02
1.18475115e+00 2.81614903e-02 -3.56591582e-01 -1.98428601e-01
-6.97687626e-01 9.50714275e-02 6.64887607e-01 6.19911849e-01
8.09373081e-01 -1.91463923e+00 -9.94035900e-01 5.20323217e-01
2.64735281e-01 5.65397926e-03 1.04877222e+00 5.20420074e-01
-3.61968040e-01 1.17200892e-02 -3.97543818e-01 -4.37455952e-01
-1.39993787e+00 6.58585668e-01 3.24106663e-01 4.37114313e-02
-5.84846139e-01 8.19359541e-01 5.80437183e-01 -5.80015779e-01
1.39738962e-01 5.95992684e-01 2.11616326e-03 2.46862322e-01
7.67526627e-01 4.13802862e-01 -1.29970491e-01 -1.04962921e+00
-3.22756797e-01 7.42413163e-01 -1.01888873e-01 -2.74090499e-01
1.26027310e+00 -2.93411702e-01 -3.45989972e-01 4.45720889e-02
1.64268124e+00 6.78026751e-02 -1.45808446e+00 -2.01103359e-01
-6.16576016e-01 -8.28855753e-01 -8.38376656e-02 -1.95973471e-01
-1.50201058e+00 7.74745822e-01 7.94794619e-01 -2.17717737e-01
1.20927763e+00 7.86925852e-03 7.71919310e-01 -1.46477446e-02
2.71863759e-01 -9.56611991e-01 2.14403957e-01 1.06462993e-01
1.10144269e+00 -1.25379646e+00 -2.68735886e-01 -4.85336721e-01
-7.26585329e-01 6.98649943e-01 7.25109398e-01 -6.34874627e-02
6.47737026e-01 -4.33027409e-02 1.82030588e-01 1.79935068e-01
-4.65166807e-01 -2.42345706e-02 3.15893181e-02 9.20488656e-01
1.52853101e-01 4.17534634e-02 -1.96060821e-01 6.10668898e-01
2.77875420e-02 -5.10400906e-02 2.53345847e-01 4.40515101e-01
2.37751931e-01 -1.32148993e+00 -5.41589618e-01 -2.35815663e-02
-3.07818860e-01 1.38528943e-01 -1.50615215e-01 4.52832043e-01
3.89705181e-01 8.34932804e-01 4.52181846e-02 -5.59497595e-01
2.45682061e-01 -1.53430820e-01 6.63462222e-01 -3.76321286e-01
2.49383822e-02 1.74685806e-01 -4.77641553e-01 -4.84888136e-01
-2.08347768e-01 -7.60888457e-01 -8.44294608e-01 -5.33334434e-01
-2.72372831e-02 2.71440238e-01 3.33979577e-01 6.00999475e-01
6.81504130e-01 2.82540500e-01 9.71156061e-01 -7.96045303e-01
-5.07331729e-01 -8.04168105e-01 -8.74979913e-01 9.38397050e-01
2.88053721e-01 -7.71404862e-01 -4.61940557e-01 2.04342127e-01]
|
[13.102635383605957, 0.26999735832214355]
|
9fa664c5-80ad-43e8-897c-99e46c85145e
|
taming-visually-guided-sound-generation
|
2110.08791
| null |
https://arxiv.org/abs/2110.08791v1
|
https://arxiv.org/pdf/2110.08791v1.pdf
|
Taming Visually Guided Sound Generation
|
Recent advances in visually-induced audio generation are based on sampling short, low-fidelity, and one-class sounds. Moreover, sampling 1 second of audio from the state-of-the-art model takes minutes on a high-end GPU. In this work, we propose a single model capable of generating visually relevant, high-fidelity sounds prompted with a set of frames from open-domain videos in less time than it takes to play it on a single GPU. We train a transformer to sample a new spectrogram from the pre-trained spectrogram codebook given the set of video features. The codebook is obtained using a variant of VQGAN trained to produce a compact sampling space with a novel spectrogram-based perceptual loss. The generated spectrogram is transformed into a waveform using a window-based GAN that significantly speeds up generation. Considering the lack of metrics for automatic evaluation of generated spectrograms, we also build a family of metrics called FID and MKL. These metrics are based on a novel sound classifier, called Melception, and designed to evaluate the fidelity and relevance of open-domain samples. Both qualitative and quantitative studies are conducted on small- and large-scale datasets to evaluate the fidelity and relevance of generated samples. We also compare our model to the state-of-the-art and observe a substantial improvement in quality, size, and computation time. Code, demo, and samples: v-iashin.github.io/SpecVQGAN
|
['Esa Rahtu', 'Vladimir Iashin']
|
2021-10-17
| null | null | null | null |
['audio-generation']
|
['audio']
|
[ 3.53298217e-01 -2.17192873e-01 4.44700181e-01 9.47247222e-02
-1.45122159e+00 -5.90982914e-01 3.98213506e-01 1.13082558e-01
-6.55937269e-02 5.31428158e-01 4.77520823e-01 7.00247660e-02
3.26179266e-01 -8.83242905e-01 -9.35407579e-01 -5.01296699e-01
-1.77952975e-01 -1.51589602e-01 1.71936750e-01 1.27579093e-01
8.37640315e-02 7.40818232e-02 -2.11528254e+00 5.54094374e-01
6.42670572e-01 1.28199875e+00 3.50844234e-01 1.32779729e+00
2.36462355e-01 5.93074322e-01 -8.14337671e-01 -2.51913875e-01
3.63426954e-01 -9.21526611e-01 -5.69470644e-01 -7.75614604e-02
6.10416532e-01 -5.58863521e-01 -1.52501971e-01 7.67208576e-01
8.53547871e-01 2.75450051e-01 5.09054184e-01 -1.13993740e+00
-3.77188891e-01 5.60252726e-01 -4.99642044e-02 -2.82249111e-03
8.67186069e-01 6.44591093e-01 1.07000887e+00 -8.57491851e-01
5.98396063e-01 1.07125902e+00 7.41407335e-01 4.08899188e-01
-1.24140763e+00 -7.48056829e-01 -4.13144022e-01 9.47759002e-02
-1.49330783e+00 -6.24156654e-01 7.14856923e-01 -5.08156896e-01
7.19424307e-01 5.47562778e-01 1.01146853e+00 1.21802998e+00
1.07820518e-02 3.37425143e-01 8.31226766e-01 -6.13748491e-01
5.18604755e-01 -1.56846315e-01 -6.46045566e-01 4.83001202e-01
-1.42985791e-01 4.64397550e-01 -7.67821133e-01 -3.24568748e-01
8.56686890e-01 -5.25771141e-01 -5.77081144e-01 2.61892751e-02
-1.17962682e+00 8.73886764e-01 2.15632722e-01 3.54876369e-02
-4.13399428e-01 4.98296648e-01 4.55184877e-01 1.73750401e-01
5.91412544e-01 4.08866107e-01 -5.11519425e-02 -6.73813462e-01
-1.42970610e+00 3.35998833e-01 6.91771269e-01 7.46506929e-01
6.74369574e-01 4.46194708e-01 -4.98700321e-01 7.33894289e-01
2.17763066e-01 6.88024580e-01 4.62157339e-01 -1.13687301e+00
1.78654969e-01 -1.28531620e-01 -1.41365841e-01 -9.36443448e-01
3.08788102e-03 -2.12168530e-01 -5.46966791e-01 4.34124917e-01
3.22928458e-01 -2.70340443e-01 -6.28917813e-01 1.58480930e+00
3.89190197e-01 5.03306031e-01 -1.61993057e-01 1.04888880e+00
7.43718982e-01 1.05596459e+00 -3.41597319e-01 -7.03094378e-02
1.16568351e+00 -1.04500723e+00 -4.98649329e-01 2.62948006e-01
2.32408091e-01 -1.06340277e+00 1.56415105e+00 6.43196404e-01
-1.36384583e+00 -9.52622414e-01 -1.10472107e+00 -8.98812488e-02
6.77118748e-02 2.01881155e-01 1.84440851e-01 8.41020525e-01
-1.15121841e+00 7.93399215e-01 -5.77632546e-01 -4.22436595e-02
3.86182249e-01 -2.18669504e-01 3.74813261e-03 1.18623003e-01
-8.73205066e-01 1.70583978e-01 5.06437682e-02 -5.42190313e-01
-1.48769295e+00 -1.12370121e+00 -8.52793336e-01 1.55798987e-01
5.58185130e-02 -6.60789847e-01 1.34400129e+00 -9.51673150e-01
-1.96892726e+00 3.24160546e-01 1.24507815e-01 -5.23581982e-01
5.04879057e-01 -2.64590532e-01 -4.66399640e-01 6.66269720e-01
1.47168608e-02 7.50739932e-01 1.29873121e+00 -1.05101895e+00
-5.66438317e-01 3.11831295e-01 3.87407430e-02 -1.32413045e-01
-2.78997838e-01 -3.34864203e-03 -3.34146500e-01 -1.04892099e+00
-5.62178075e-01 -7.87149251e-01 9.79337320e-02 1.82637602e-01
-2.45905429e-01 2.94669718e-01 5.26913226e-01 -8.16194654e-01
1.30998778e+00 -2.37071514e+00 -1.66369140e-01 -4.67363223e-02
4.09666672e-02 2.14566171e-01 -4.44111049e-01 4.21074480e-01
4.05289605e-02 8.17347243e-02 -3.34715188e-01 -3.52104664e-01
1.42198903e-02 -4.05137390e-01 -5.76364219e-01 2.03020424e-01
1.77109823e-01 4.56823498e-01 -1.05503523e+00 -3.39150697e-01
2.25822955e-01 8.15098763e-01 -1.14453101e+00 4.93958831e-01
-2.79987901e-01 5.21376133e-01 1.37649223e-01 3.39805007e-01
5.84847987e-01 1.13778934e-01 -1.78689286e-01 -2.59754866e-01
-4.83366698e-02 4.18858737e-01 -1.18361139e+00 2.06686282e+00
-8.70636582e-01 8.51025760e-01 -1.93920657e-01 -3.15723717e-01
9.12687898e-01 5.25313675e-01 1.60492167e-01 -4.87643838e-01
1.82346940e-01 1.97403997e-01 -2.43334442e-01 -4.38675702e-01
5.70537865e-01 1.18174173e-01 8.51430148e-02 6.48629427e-01
3.53866071e-01 -5.59731126e-01 3.54405344e-01 1.31190389e-01
1.18856812e+00 4.32195008e-01 1.37451202e-01 1.25786692e-01
2.81606764e-01 -3.73472750e-01 8.48096684e-02 5.95195353e-01
1.55343190e-01 1.22921824e+00 2.42200017e-01 3.47920917e-02
-1.28332722e+00 -1.21647525e+00 8.75249878e-02 1.08290780e+00
-4.27676469e-01 -8.34913611e-01 -1.25193095e+00 -3.18786532e-01
-3.34723651e-01 6.74309909e-01 -3.43687177e-01 -1.90823302e-01
-2.63871819e-01 -1.09178931e-01 9.15827334e-01 2.57733762e-01
9.79723632e-02 -1.15792656e+00 -1.10615838e+00 2.82179356e-01
-3.38370383e-01 -1.03825200e+00 -7.62183130e-01 -2.92193174e-01
-5.85277379e-01 -8.85231793e-01 -8.81820917e-01 -4.53158915e-01
3.65924567e-01 -6.86711818e-02 1.26682913e+00 -1.14481106e-01
-3.65696549e-01 3.88133705e-01 -8.00808311e-01 -3.18120003e-01
-6.88465178e-01 -2.13236079e-01 -2.83050421e-03 2.78929144e-01
-4.09822673e-01 -8.02833915e-01 -6.96911812e-01 -5.99767305e-02
-1.05554628e+00 2.42797345e-01 2.17314154e-01 7.89547920e-01
6.81418777e-01 -1.45200863e-01 4.49287504e-01 -2.39078104e-01
6.55727506e-01 -3.25412035e-01 -5.09919524e-01 -2.74389535e-01
-1.30668506e-01 -6.72444776e-02 9.75415409e-01 -6.72491670e-01
-8.07841957e-01 -6.89114705e-02 -2.88523257e-01 -7.07796514e-01
-1.83510259e-01 2.79027402e-01 3.38546410e-02 7.61366189e-02
8.66807282e-01 2.22933471e-01 -1.43146530e-01 -5.62557280e-01
6.42051995e-01 7.58013129e-01 7.41312206e-01 -5.54534197e-01
6.76110327e-01 4.10012424e-01 -2.85377532e-01 -8.51245761e-01
-5.18998384e-01 5.24886437e-02 7.45675862e-02 -3.67027879e-01
6.51709259e-01 -9.57806885e-01 -5.50632179e-01 4.60417092e-01
-1.11869502e+00 -6.80391252e-01 -8.67926419e-01 5.75824201e-01
-8.61673057e-01 2.50569850e-01 -6.27626419e-01 -8.84588182e-01
-5.43711841e-01 -1.06612551e+00 1.14649069e+00 -2.36413651e-03
-3.34739447e-01 -4.66540277e-01 4.34640914e-01 4.67450768e-02
4.08171237e-01 3.71915847e-01 4.67827082e-01 -7.59161934e-02
-4.66781259e-01 -5.15701585e-02 1.53306462e-02 5.62997639e-01
1.40686939e-02 3.35372657e-01 -1.29233229e+00 -3.12863886e-01
-1.20177515e-01 -4.96006250e-01 6.62717283e-01 5.16393423e-01
1.23094571e+00 -4.25657004e-01 2.89440274e-01 8.55779171e-01
1.29785895e+00 1.62716627e-01 6.31720543e-01 -2.04335019e-01
4.64816868e-01 2.46972799e-01 7.13129044e-01 9.71154928e-01
1.42829940e-01 7.41185427e-01 3.18648487e-01 8.08986500e-02
-7.38461316e-01 -6.72263980e-01 6.86438859e-01 1.06544507e+00
-8.35485905e-02 -2.77264625e-01 -5.51961899e-01 6.40699506e-01
-1.28074193e+00 -9.19494033e-01 1.19244739e-01 2.33719039e+00
1.15205371e+00 -5.65268360e-02 3.69088769e-01 5.91862142e-01
5.85053504e-01 1.44907624e-01 -5.59895411e-02 -4.34658319e-01
1.47935539e-01 8.72399807e-01 -1.11660786e-01 4.99179780e-01
-8.03612411e-01 5.77823222e-01 6.11535978e+00 1.29123926e+00
-1.44846594e+00 3.12612176e-01 5.66822231e-01 -6.24873340e-01
-3.82630169e-01 1.12022161e-02 -3.71186256e-01 7.45051622e-01
1.51287854e+00 -1.12059548e-01 8.19448352e-01 7.03463972e-01
4.49766189e-01 -1.11270599e-01 -1.17743897e+00 1.22163391e+00
1.50378883e-01 -1.63374424e+00 1.63209394e-01 -7.63475709e-03
7.04869866e-01 -2.86719710e-01 2.09002703e-01 1.49311289e-01
-5.62548339e-02 -9.47744906e-01 1.36438406e+00 4.90053147e-01
1.54139233e+00 -8.18525970e-01 2.54139960e-01 -9.19619948e-02
-1.46144545e+00 9.09099355e-02 -2.21035123e-01 -2.81921983e-01
2.84187585e-01 7.85633266e-01 -8.25477004e-01 4.47152466e-01
7.66805649e-01 3.94621283e-01 -4.90960002e-01 1.08886147e+00
-1.89278662e-01 9.96595025e-01 -3.34310442e-01 7.24803731e-02
1.55275427e-02 1.54610991e-01 6.42583370e-01 1.28282094e+00
8.51451099e-01 -2.11457342e-01 -8.15932602e-02 1.09820807e+00
-5.24726696e-03 1.84721932e-01 -3.78948063e-01 -1.09063283e-01
5.99860668e-01 1.22852814e+00 -4.36352104e-01 -3.72869074e-01
-1.62993550e-01 8.87569189e-01 -1.41035035e-01 6.24595582e-02
-1.18599641e+00 -6.61818266e-01 6.05418265e-01 2.08172053e-01
5.80495715e-01 1.37726823e-02 -3.27184238e-02 -9.28143561e-01
3.06253433e-02 -1.17254877e+00 1.00313269e-01 -1.03350222e+00
-8.70512009e-01 8.85065317e-01 -1.95667356e-01 -1.79549551e+00
-5.83478987e-01 2.68273745e-02 -5.06097615e-01 6.67643726e-01
-1.16456437e+00 -8.94745767e-01 -5.11920273e-01 5.26228607e-01
6.08726859e-01 -1.44922554e-01 1.14761412e+00 3.63573253e-01
-8.34966898e-02 7.40465999e-01 2.05890667e-02 -6.57114610e-02
6.98935151e-01 -8.84705007e-01 7.75021911e-01 9.18457627e-01
3.82848203e-01 -8.69194325e-03 7.47842193e-01 -3.91596377e-01
-1.23833108e+00 -1.22743309e+00 4.37575996e-01 -1.75037503e-01
3.45928371e-01 -5.38002074e-01 -5.67017794e-01 9.92273986e-02
2.11775303e-01 4.15105820e-01 7.66844809e-01 -5.37911713e-01
-3.82191151e-01 -1.68325588e-01 -1.16306162e+00 3.84633780e-01
9.91035342e-01 -7.13581145e-01 -1.33345842e-01 -8.61059725e-02
1.01083648e+00 -4.82726038e-01 -9.22797620e-01 -1.12967178e-01
6.51563108e-01 -1.25213647e+00 9.02947485e-01 -2.99570076e-02
8.68296564e-01 -3.87210935e-01 -3.40447009e-01 -1.49683750e+00
-6.50943145e-02 -1.15587127e+00 -2.14352459e-01 1.38919139e+00
1.72009721e-01 -2.15202257e-01 3.74459743e-01 -1.56515464e-01
-1.85052320e-01 -6.02440894e-01 -8.66311014e-01 -8.12306702e-01
-2.64987111e-01 -9.97859418e-01 8.39521945e-01 4.51662838e-01
4.54656892e-02 1.19136393e-01 -4.98122305e-01 -3.44226733e-02
5.05403936e-01 1.34774908e-01 1.10340416e+00 -8.09731126e-01
-7.47280836e-01 -1.09491371e-01 -4.77817774e-01 -8.14233601e-01
-3.50765228e-01 -6.86556101e-01 1.99732453e-01 -1.10418463e+00
-1.55789033e-01 -3.24195474e-01 -8.51952285e-02 1.15136400e-01
7.89200887e-02 7.33479977e-01 6.58033431e-01 -1.22782357e-01
-3.43517631e-01 7.53422618e-01 1.14563394e+00 -2.50024982e-02
-2.70614028e-01 -2.59080946e-01 -3.94109726e-01 6.10247135e-01
5.30750036e-01 -6.55010283e-01 -4.15575802e-01 -2.51745075e-01
1.10010408e-01 4.10253853e-01 6.29070818e-01 -1.74613917e+00
-2.37368479e-01 1.68694049e-01 1.96843773e-01 -2.85924077e-01
6.82758033e-01 -4.87916648e-01 4.57933307e-01 3.60800594e-01
-4.13752466e-01 -3.00176531e-01 2.54582733e-01 3.65662098e-01
-2.16016352e-01 -1.50837645e-01 8.46164227e-01 1.40012547e-01
-2.74140269e-01 8.63864943e-02 -3.90583366e-01 4.17878211e-01
6.58049524e-01 -2.20382109e-01 2.19195355e-02 -8.15844536e-01
-5.07498384e-01 -7.26846397e-01 4.72693145e-01 2.67560095e-01
7.95338094e-01 -1.52180743e+00 -9.69951451e-01 3.77788216e-01
-2.68147364e-02 -1.97152957e-01 5.12257814e-01 4.56475139e-01
-7.34067202e-01 1.24678738e-01 -3.08009386e-01 -6.55814528e-01
-1.12547994e+00 3.98323655e-01 3.44670080e-02 -1.22619607e-01
-6.12967789e-01 8.19566190e-01 2.26722762e-01 4.32433970e-02
7.98902661e-02 -7.59428382e-01 1.45407215e-01 -4.32991050e-02
8.43995154e-01 5.35816491e-01 1.01058975e-01 -5.22938490e-01
-1.21331476e-01 4.07892197e-01 8.02438200e-01 -6.31991267e-01
1.22959590e+00 1.59676328e-01 3.20195645e-01 4.24635530e-01
1.29761767e+00 4.17289287e-01 -1.61831999e+00 1.35937259e-01
-7.68848419e-01 -7.79232979e-01 1.00858703e-01 -6.56605780e-01
-1.11362898e+00 9.57893252e-01 8.42281401e-01 2.94301838e-01
1.39538002e+00 -3.06846410e-01 1.15518963e+00 -4.01323497e-01
4.27013308e-01 -8.63203108e-01 3.89976919e-01 3.49237949e-01
1.10287893e+00 -5.75523615e-01 -1.85309529e-01 -1.93997771e-01
-5.90625167e-01 1.06897330e+00 1.74408123e-01 -2.11827427e-01
4.86019969e-01 5.23967147e-01 3.99570875e-02 3.41116250e-01
-7.69354105e-01 -1.02281913e-01 3.34852338e-01 9.10385609e-01
3.57029349e-01 1.28635094e-01 2.38874014e-02 7.72713780e-01
-9.53481913e-01 1.73820600e-01 6.97110891e-01 6.09661400e-01
-2.04082280e-01 -7.21784294e-01 -5.85499227e-01 2.31395885e-01
-4.52642500e-01 -3.56623203e-01 9.93773937e-02 9.53155309e-02
1.79202110e-01 1.24356484e+00 2.78190792e-01 -6.86921775e-01
9.21224132e-02 -1.55624747e-01 5.90305805e-01 -4.73481655e-01
-7.66786158e-01 2.99063206e-01 9.50554982e-02 -7.48951018e-01
-1.81798711e-01 -4.82745260e-01 -1.09447467e+00 -2.65077263e-01
-2.33467788e-01 6.42032027e-02 7.09095955e-01 4.37563568e-01
6.32731676e-01 8.55612934e-01 8.21762681e-01 -1.38426077e+00
-3.72713804e-01 -9.67959404e-01 -4.41127539e-01 4.90661323e-01
3.87097746e-01 -2.92400181e-01 -4.51549500e-01 4.59336340e-01]
|
[15.565764427185059, 5.72912073135376]
|
3491aee0-8467-40c3-9e61-27d7208d7920
|
alphadesign-a-graph-protein-design-method-and
|
2202.01079
| null |
https://arxiv.org/abs/2202.01079v2
|
https://arxiv.org/pdf/2202.01079v2.pdf
|
AlphaDesign: A graph protein design method and benchmark on AlphaFoldDB
|
While DeepMind has tentatively solved protein folding, its inverse problem -- protein design which predicts protein sequences from their 3D structures -- still faces significant challenges. Particularly, the lack of large-scale standardized benchmark and poor accuray hinder the research progress. In order to standardize comparisons and draw more research interest, we use AlphaFold DB, one of the world's largest protein structure databases, to establish a new graph-based benchmark -- AlphaDesign. Based on AlphaDesign, we propose a new method called ADesign to improve accuracy by introducing protein angles as new features, using a simplified graph transformer encoder (SGT), and proposing a confidence-aware protein decoder (CPD). Meanwhile, SGT and CPD also improve model efficiency by simplifying the training and testing procedures. Experiments show that ADesign significantly outperforms previous graph models, e.g., the average accuracy is improved by 8\%, and the inference speed is 40+ times faster than before.
|
['Stan Z. Li', 'Cheng Tan', 'Zhangyang Gao']
|
2022-02-01
| null | null | null | null |
['protein-design']
|
['medical']
|
[ 3.09924185e-02 -1.92165300e-02 -3.99367273e-01 -4.96482760e-01
-3.62514734e-01 -3.43623489e-01 -1.74941286e-01 1.32844880e-01
-9.04893428e-02 1.06581008e+00 4.55663316e-02 -6.80292308e-01
3.12026948e-01 -6.92362845e-01 -1.21443808e+00 -7.07817554e-01
4.23042066e-02 4.39890385e-01 2.42203727e-01 -1.56177506e-01
1.90459043e-01 3.48468482e-01 -9.38246131e-01 2.89791286e-01
1.23136973e+00 7.41929173e-01 5.83907604e-01 2.33288035e-01
-1.01940125e-01 6.17462099e-01 -4.42299038e-01 -6.39993548e-01
3.20946872e-02 -6.41761005e-01 -8.43555808e-01 -2.43393093e-01
1.11266285e-01 -2.88119495e-01 -2.94304907e-01 1.01571894e+00
5.87982237e-01 -2.17776537e-01 2.89815933e-01 -1.02021825e+00
-1.06562543e+00 4.27866817e-01 -5.90016305e-01 -2.04631761e-01
5.10616064e-01 4.53859091e-01 1.05621231e+00 -7.45319128e-01
9.21862423e-01 1.17955625e+00 1.03053737e+00 6.13087118e-01
-1.45947933e+00 -6.03056490e-01 1.28267363e-01 5.27712941e-01
-1.33673525e+00 -1.23469874e-01 6.25544250e-01 -1.50341883e-01
1.65783691e+00 -1.04553044e-01 7.31605232e-01 1.06253397e+00
6.08787477e-01 6.03508055e-01 8.49252760e-01 -1.25530332e-01
7.03336149e-02 -4.69837725e-01 1.52326971e-01 1.08751321e+00
4.63374257e-01 2.24453155e-02 -4.63796526e-01 -2.48348191e-01
5.28748512e-01 -7.95638710e-02 -5.08125186e-01 -5.99563062e-01
-1.24967039e+00 6.91645622e-01 6.45899177e-01 -1.44353220e-02
-3.80710304e-01 -2.91880071e-02 3.97346169e-01 3.06257099e-01
4.02126610e-01 4.81686026e-01 -7.82755971e-01 -1.85978133e-02
-4.96196449e-01 5.87061286e-01 8.89254689e-01 1.27614248e+00
7.24947214e-01 -2.52454966e-01 8.60894769e-02 7.00946152e-01
4.95089293e-01 3.29690814e-01 1.28637478e-01 -5.60015082e-01
6.21912420e-01 9.34615433e-01 -1.41390890e-01 -1.04573596e+00
-6.22556627e-01 -3.13132048e-01 -8.63997221e-01 -2.03120798e-01
3.18280697e-01 4.34930287e-02 -8.62402678e-01 1.82140994e+00
3.70822221e-01 -9.80150551e-02 1.30277440e-01 1.11371315e+00
8.67335737e-01 7.87276089e-01 2.88089037e-01 -4.80863154e-01
1.26005161e+00 -9.10597682e-01 -5.85136831e-01 6.07924648e-02
9.65042293e-01 -5.12362778e-01 1.08857727e+00 5.24357259e-01
-8.56344342e-01 -5.28832257e-01 -1.31260407e+00 -4.71490830e-01
-1.87131226e-01 1.00945048e-02 8.57692063e-01 1.72722980e-01
-6.99150562e-01 8.58168483e-01 -8.30866218e-01 -3.27801138e-01
3.65200043e-01 4.48973924e-01 -6.30272806e-01 -2.36936525e-01
-1.27975345e+00 1.05395174e+00 5.93702137e-01 -1.57381997e-01
-6.51417613e-01 -8.49072635e-01 -7.57737756e-01 -8.36198851e-02
2.49024823e-01 -8.73603046e-01 9.68092084e-01 -4.62299049e-01
-1.46232057e+00 5.59297740e-01 -4.49670166e-01 -4.39325809e-01
1.40275881e-01 -1.83756754e-01 -3.90031517e-01 5.98982628e-03
-2.64849998e-02 5.84354103e-01 2.12432563e-01 -8.76287341e-01
-1.41018271e-01 -5.72611749e-01 2.57712211e-02 1.19277485e-01
4.87310551e-02 -2.03094125e-01 -5.08560419e-01 -5.80072343e-01
1.60024717e-01 -7.95855045e-01 -4.26830411e-01 1.24699585e-01
-5.67048252e-01 -2.56096959e-01 5.32724380e-01 -8.65157008e-01
1.38426661e+00 -1.74927211e+00 5.44625461e-01 8.95187408e-02
7.19156802e-01 4.34153378e-01 -1.25627533e-01 6.01274967e-01
-2.41886139e-01 -2.03895926e-01 -5.78946769e-01 2.02817574e-01
-2.02177510e-01 5.18358052e-01 6.71251491e-02 4.24011827e-01
3.24407578e-01 1.19450009e+00 -8.04885268e-01 -4.41643238e-01
1.66881830e-02 4.32118446e-01 -9.48731124e-01 2.57237107e-01
-7.39177763e-01 2.95707256e-01 -5.51837087e-01 6.54082060e-01
9.22686279e-01 -8.42919767e-01 7.53048897e-01 -5.95008850e-01
5.83564630e-03 4.06844348e-01 -6.53348565e-01 1.87117457e+00
1.67624518e-01 1.23418450e-01 -3.27216506e-01 -1.19262803e+00
1.19236541e+00 -2.46665496e-02 4.80251163e-01 -6.32139385e-01
-1.28220335e-01 1.61184236e-01 1.09212331e-01 -6.17865443e-01
1.73630849e-01 7.75103122e-02 3.69794875e-01 5.13566248e-02
9.24879313e-02 3.10657680e-01 1.62354231e-01 1.62949622e-01
1.31295145e+00 7.52679288e-01 4.54825163e-01 -3.70503813e-01
5.58359146e-01 3.98687780e-01 1.14727569e+00 -2.20648758e-02
-1.02057755e-01 3.98231357e-01 8.50185156e-01 -6.73965931e-01
-1.13955081e+00 -7.44694591e-01 4.92773354e-02 8.49836409e-01
4.31647122e-01 -9.26826179e-01 -9.45260167e-01 -9.46015358e-01
3.11226308e-01 4.82582718e-01 -2.77633816e-01 -2.60763645e-01
-7.95526445e-01 -1.31444418e+00 4.49077845e-01 4.31499481e-01
4.37721014e-01 -8.01008940e-01 -3.63210477e-02 5.22599995e-01
-2.42143184e-01 -7.73422599e-01 -5.84303498e-01 3.71562511e-01
-9.93926227e-01 -1.40931106e+00 -5.80525160e-01 -1.02203679e+00
8.02505136e-01 2.98284292e-01 1.27985930e+00 1.32425845e-01
-2.20329061e-01 -7.21752226e-01 -4.26646471e-01 -1.67569518e-01
-4.45120096e-01 1.84024066e-01 5.22008613e-02 -5.67531526e-01
8.19327414e-01 -6.19668126e-01 -7.64793456e-01 4.27228898e-01
-4.62379366e-01 4.70868200e-01 7.75999248e-01 9.48589683e-01
9.10071552e-01 -4.99935418e-01 7.51633883e-01 -1.01043308e+00
5.60712159e-01 -2.77882636e-01 -7.33955681e-01 5.27653992e-01
-1.34654760e+00 4.88482147e-01 1.06847703e+00 -6.49667904e-02
-7.25318015e-01 2.11375669e-01 -5.82649171e-01 -2.19086066e-01
2.39264056e-01 5.95211089e-01 -4.96801496e-01 -3.09318423e-01
6.17212117e-01 3.12921166e-01 3.77610385e-01 -8.16025972e-01
2.47327417e-01 5.01320243e-01 3.92662972e-01 -4.98590082e-01
3.27726096e-01 -2.26256356e-01 1.54849291e-01 -4.00849611e-01
-6.33178413e-01 -1.66240022e-01 -5.02194345e-01 2.08192945e-01
8.10522854e-01 -9.69735861e-01 -1.28536510e+00 2.95081675e-01
-1.34820402e+00 -3.01246922e-02 3.58338982e-01 4.72351223e-01
-5.41151524e-01 9.77504790e-01 -9.12053049e-01 -5.84333241e-02
-6.40757978e-01 -1.48605824e+00 8.98740888e-01 -9.92082208e-02
-1.59649238e-01 -6.37278855e-01 1.69036224e-01 4.15414482e-01
1.17655687e-01 2.52858549e-01 1.36059213e+00 -4.51212913e-01
-6.30060554e-01 3.55563581e-01 -4.84300733e-01 4.48192596e-01
2.26707794e-02 -8.36882368e-02 -3.57893288e-01 -3.11364979e-01
-3.19969177e-01 -3.48146200e-01 7.90827870e-01 2.45699644e-01
1.35048306e+00 -3.40045601e-01 -6.72903776e-01 9.60158527e-01
1.36566257e+00 2.05301359e-01 6.12176418e-01 2.13384479e-01
8.93084347e-01 1.94887519e-01 5.95838189e-01 1.88633636e-01
5.96146464e-01 7.09848344e-01 4.29256558e-01 -1.20238967e-01
-1.40406564e-01 -6.96496248e-01 3.07584614e-01 1.37542963e+00
-3.86342317e-01 -2.87922680e-01 -6.38410687e-01 -1.02345027e-01
-2.02407002e+00 -7.14114189e-01 -3.30212861e-01 1.89546740e+00
1.20303071e+00 -2.74378806e-02 1.62031706e-02 -1.38509631e-01
3.66128534e-01 -2.78729722e-02 -9.72643137e-01 -1.53389379e-01
-1.87886924e-01 2.26648554e-01 7.26166546e-01 3.95127177e-01
-7.86130965e-01 9.75279272e-01 6.71295404e+00 6.88710093e-01
-9.28984404e-01 -2.88089931e-01 4.90376562e-01 3.66957754e-01
-2.72811264e-01 1.67564303e-01 -9.38390672e-01 6.33973241e-01
9.08536494e-01 -1.95289165e-01 4.47553247e-01 1.09886658e+00
1.93501279e-01 4.06725407e-01 -1.12581587e+00 9.30230856e-01
5.13673499e-02 -1.65923965e+00 1.48519829e-01 8.15113410e-02
2.30243981e-01 3.94933186e-02 -4.83536154e-01 2.70453364e-01
2.53185987e-01 -9.55419362e-01 2.26315811e-01 5.76521873e-01
8.21424425e-01 -7.93653905e-01 6.49008214e-01 2.25867510e-01
-1.14879858e+00 3.67960691e-01 -8.15424144e-01 2.54441109e-02
8.47418830e-02 5.97702861e-01 -8.48827839e-01 9.06266510e-01
4.23058391e-01 1.14192462e+00 -3.69052649e-01 7.39502132e-01
-1.77814290e-01 6.36674643e-01 -3.56260873e-02 -3.33925575e-01
-9.29825380e-02 -4.80365455e-01 1.85734615e-01 1.15876937e+00
1.19288780e-01 3.03701460e-01 3.40250134e-01 8.38768601e-01
-3.13383400e-01 1.82672516e-01 -2.82876939e-01 -1.27249911e-01
4.35909152e-01 9.53478813e-01 -1.50953159e-01 -1.75722703e-01
-5.63985765e-01 1.08592999e+00 7.03075111e-01 2.38465175e-01
-1.09116232e+00 -6.90711915e-01 9.72749352e-01 -7.37675978e-03
2.30479449e-01 -3.16833645e-01 -1.53252315e-02 -1.24629951e+00
4.22959626e-02 -1.34642315e+00 -2.71965470e-02 -7.76215434e-01
-1.40844488e+00 4.73445415e-01 -3.64374250e-01 -9.71350551e-01
4.48650233e-02 -1.05417120e+00 -5.37899993e-02 9.60650027e-01
-1.39176905e+00 -9.69763339e-01 -1.03976063e-01 2.33488530e-01
2.70963103e-01 -1.75647169e-01 8.97126913e-01 5.00102162e-01
-8.46239746e-01 7.75919795e-01 2.76747823e-01 -1.43845767e-01
6.24924839e-01 -1.16667843e+00 1.03341496e+00 5.16045690e-01
-2.59123683e-01 8.24341476e-01 7.20876336e-01 -8.43296230e-01
-2.10377574e+00 -1.28515506e+00 1.13030851e+00 -3.55043828e-01
3.81661892e-01 -5.21894634e-01 -1.36706758e+00 4.42819178e-01
-1.66306078e-01 -2.06920326e-01 6.77946329e-01 1.40640363e-02
-4.37310874e-01 -9.39483196e-02 -1.04603922e+00 5.54166257e-01
1.72510719e+00 -3.40966702e-01 -5.49391747e-01 6.61949337e-01
1.30743289e+00 -4.22780126e-01 -1.33070099e+00 6.96892023e-01
5.51500082e-01 -8.41177225e-01 9.31981564e-01 -9.10579920e-01
4.82913643e-01 -5.33499420e-01 -1.34827763e-01 -1.15896142e+00
-6.84725940e-01 -4.91140991e-01 -4.58532065e-01 6.98019445e-01
5.81333339e-01 -7.49832749e-01 9.70325232e-01 1.17963925e-01
-5.01881301e-01 -1.32246232e+00 -6.48832262e-01 -7.76717186e-01
-6.06397912e-02 -6.36075056e-05 9.77949619e-01 8.83190751e-01
2.64368504e-01 6.80123568e-01 -5.33974528e-01 -1.01074740e-01
5.77751458e-01 3.14768612e-01 8.86032999e-01 -1.06509852e+00
-5.08582652e-01 2.30593774e-02 -4.16457534e-01 -1.66008770e+00
1.90318692e-02 -1.15738320e+00 -3.40210237e-02 -1.58887088e+00
3.90148699e-01 -1.38807207e-01 -2.46231005e-01 6.43484652e-01
-3.10521603e-01 -2.26200402e-01 -2.94573575e-01 1.99797943e-01
-6.15058661e-01 7.17159867e-01 1.64560652e+00 2.88817231e-02
1.10548526e-01 -4.19132352e-01 -8.00015926e-01 3.32114577e-01
7.29267061e-01 -2.39458501e-01 -3.68983328e-01 -2.87335724e-01
2.51252055e-01 4.97255735e-02 -3.55336145e-02 -7.65893519e-01
-3.30385752e-02 -1.85387403e-01 4.88931924e-01 -6.83253467e-01
6.61893561e-02 -5.02711177e-01 4.07300979e-01 7.64138997e-01
-2.13780388e-01 4.55055922e-01 -6.19563349e-02 7.72196054e-01
-5.62942494e-03 1.83373287e-01 5.52553236e-01 9.14278999e-03
-5.90409100e-01 6.75051570e-01 3.72503214e-02 -6.92422092e-02
9.14302289e-01 -1.09295942e-01 -5.76385260e-01 2.74534822e-01
-4.21096534e-01 3.27936202e-01 7.30431199e-01 1.62561253e-01
6.50435746e-01 -1.11057556e+00 -4.50473011e-01 3.98909926e-01
2.99717158e-01 -1.18454479e-01 1.79105643e-02 5.49252212e-01
-1.03313923e+00 5.99564254e-01 -2.58635342e-01 -4.62051243e-01
-1.38951528e+00 8.20640326e-01 1.76235646e-01 -2.99824774e-01
-7.74590909e-01 6.40862882e-01 1.86811030e-01 -5.57806253e-01
2.67365333e-02 -4.02696133e-01 7.31452703e-02 -5.97534776e-01
2.64045417e-01 1.47842959e-01 1.77363008e-01 -3.66580129e-01
-4.37747717e-01 4.64333713e-01 -3.69754642e-01 8.26436460e-01
1.43118167e+00 2.64819384e-01 -4.16056991e-01 -8.61596987e-02
1.23385048e+00 -4.89288896e-01 -1.22311342e+00 -1.58283606e-01
1.94569379e-01 -2.83761114e-01 -3.31647545e-01 -8.97473037e-01
-7.35627174e-01 6.41467690e-01 4.73351359e-01 -3.32746774e-01
1.02483666e+00 -2.97391236e-01 1.17360795e+00 8.07628572e-01
6.99289739e-01 -7.53116667e-01 -2.80990213e-01 5.66309810e-01
7.76089370e-01 -1.14199889e+00 2.14615867e-01 -5.78257143e-01
-5.11098504e-01 1.05229211e+00 7.75394499e-01 3.19476910e-02
4.00244385e-01 1.22812346e-01 -1.67378813e-01 -2.63727427e-01
-1.04705751e+00 1.38246447e-01 2.10419446e-01 6.67881846e-01
8.81206572e-01 1.14584774e-01 -7.96593666e-01 7.98479438e-01
-1.25045180e-01 3.67060483e-01 -1.00427292e-01 7.99625635e-01
-6.29390955e-01 -1.43452740e+00 1.71959415e-01 6.06006980e-01
-3.25526506e-01 -2.38713712e-01 -5.10284841e-01 6.04987383e-01
1.21043712e-01 6.27120554e-01 -4.74267781e-01 -8.75667632e-01
5.08476973e-01 5.29877841e-02 6.71873152e-01 -3.67450804e-01
-3.80698115e-01 -3.32536608e-01 1.72576085e-01 -7.48681903e-01
-2.47827455e-01 -1.27705753e-01 -1.43889332e+00 -6.82414532e-01
-4.91357625e-01 4.80666071e-01 4.69701886e-01 6.97852135e-01
9.79853153e-01 5.29298961e-01 4.78252321e-01 -3.00665200e-01
-8.19635212e-01 -7.82987416e-01 -3.74555826e-01 2.69495577e-01
-7.25051016e-02 -7.33580709e-01 1.54330030e-01 3.58941592e-02]
|
[4.722708702087402, 5.646328449249268]
|
871fb506-8c4f-43fb-ab4e-18faec96c27f
|
lifespan-age-transformation-synthesis
|
2003.09764
| null |
https://arxiv.org/abs/2003.09764v2
|
https://arxiv.org/pdf/2003.09764v2.pdf
|
Lifespan Age Transformation Synthesis
|
We address the problem of single photo age progression and regression-the prediction of how a person might look in the future, or how they looked in the past. Most existing aging methods are limited to changing the texture, overlooking transformations in head shape that occur during the human aging and growth process. This limits the applicability of previous methods to aging of adults to slightly older adults, and application of those methods to photos of children does not produce quality results. We propose a novel multi-domain image-to-image generative adversarial network architecture, whose learned latent space models a continuous bi-directional aging process. The network is trained on the FFHQ dataset, which we labeled for ages, gender, and semantic segmentation. Fixed age classes are used as anchors to approximate continuous age transformation. Our framework can predict a full head portrait for ages 0-70 from a single photo, modifying both texture and shape of the head. We demonstrate results on a wide variety of photos and datasets, and show significant improvement over the state of the art.
|
['Ira Kemelmacher-Shlizerman', 'Ohad Fried', 'Roy Or-El', 'Soumyadip Sengupta', 'Eli Shechtman']
|
2020-03-21
|
lifespan-age-transformation-synthesis-1
|
https://www.ecva.net/papers/eccv_2020/papers_ECCV/html/88_ECCV_2020_paper.php
|
https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123510732.pdf
|
eccv-2020-8
|
['image-to-video', 'multimodal-unsupervised-image-to-image', 'face-age-editing', 'human-aging']
|
['computer-vision', 'computer-vision', 'computer-vision', 'miscellaneous']
|
[ 2.99611270e-01 2.96633095e-01 -1.72756866e-01 -5.87387800e-01
-4.79206830e-01 -1.84385315e-01 5.14688313e-01 -2.31713519e-01
-4.90879655e-01 7.87538528e-01 5.73060870e-01 1.85597315e-01
6.46989703e-01 -8.59211564e-01 -8.38480175e-01 -5.84984601e-01
1.45685524e-01 5.94777465e-01 -9.57699642e-02 5.12705892e-02
-2.63578027e-01 1.74307078e-01 -1.68552434e+00 -6.47792071e-02
1.03287756e+00 7.85624087e-01 -6.43867910e-01 8.44227076e-01
4.13526803e-01 3.86916041e-01 -5.42897940e-01 -9.17216063e-01
9.48744491e-02 -1.57770142e-01 -5.43314338e-01 1.94789723e-01
1.43772364e+00 -8.01796138e-01 -7.02740431e-01 8.00828516e-01
8.33558619e-01 -2.40366980e-01 1.04913080e+00 -1.47369587e+00
-1.23589194e+00 3.96819323e-01 -6.48260057e-01 -1.80521488e-01
4.89709526e-01 1.06420135e-02 4.40902203e-01 -5.81112921e-01
5.32296479e-01 1.59838378e+00 9.37981486e-01 1.30269384e+00
-1.24674070e+00 -8.33258450e-01 3.37583810e-01 2.89231598e-01
-1.12670815e+00 -5.33257782e-01 7.02355027e-01 -5.50371230e-01
3.23581934e-01 -3.88857760e-02 9.94139969e-01 1.79441679e+00
3.70096445e-01 8.48245502e-01 1.31200159e+00 -3.92262906e-01
2.00757325e-01 -5.52863240e-01 3.46205868e-02 7.36133218e-01
2.24667981e-01 2.23883018e-02 -6.27337933e-01 3.96916233e-02
6.24593496e-01 -8.91695246e-02 1.29561201e-01 -3.32898527e-01
-8.30941439e-01 4.66093093e-01 2.78682530e-01 -1.74762607e-01
-1.94221258e-01 4.15278584e-01 3.93984206e-02 2.40588367e-01
9.11900938e-01 -2.00246990e-01 -3.11520070e-01 1.33162560e-02
-1.25654757e+00 5.66514313e-01 4.58223104e-01 7.76305258e-01
3.87189209e-01 1.57206103e-01 -1.50019363e-01 8.07204425e-01
8.11661705e-02 8.31264853e-01 4.13128644e-01 -1.31266606e+00
1.83413535e-01 4.37997788e-01 1.92529671e-02 -4.70973045e-01
-5.41700006e-01 -7.58246779e-02 -7.58220017e-01 4.46239591e-01
8.57281089e-01 -2.44630709e-01 -1.67043769e+00 2.07551742e+00
3.65083963e-01 -9.31470841e-03 -2.02491060e-01 4.23831969e-01
6.36880457e-01 1.97069108e-01 5.05428851e-01 -1.10217579e-01
1.22535944e+00 -9.06879246e-01 -4.43893641e-01 -8.69568765e-01
1.54370740e-01 -4.37351018e-01 1.14744389e+00 4.10417050e-01
-1.51282203e+00 -5.98683417e-01 -1.00142014e+00 -3.74040544e-01
-2.77404785e-01 9.68996249e-03 5.52711070e-01 8.73308063e-01
-1.41369426e+00 7.57472754e-01 -1.08945847e+00 -7.54697978e-01
5.92324257e-01 5.84517717e-01 -4.27120090e-01 -3.74141306e-01
-9.96003687e-01 8.17566931e-01 -2.31749192e-01 -3.12305987e-01
-8.56675208e-01 -1.07885444e+00 -8.64835143e-01 -2.99723268e-01
-1.80664033e-01 -1.21286976e+00 1.41803265e+00 -1.20725322e+00
-1.06883013e+00 1.36298084e+00 -2.32530743e-01 -4.19701964e-01
7.11752594e-01 -5.56151748e-01 -5.39076388e-01 1.43267497e-01
1.72033593e-01 1.27196431e+00 1.37939942e+00 -1.06800830e+00
-3.64235252e-01 -1.02071333e+00 -2.92754918e-01 2.59491652e-01
-6.07166171e-01 6.01145141e-02 -4.21709001e-01 -8.84616077e-01
-5.19103855e-02 -1.26211047e+00 -3.47384959e-02 5.01840770e-01
-1.16376594e-01 1.24789432e-01 5.98430812e-01 -1.33052039e+00
1.16966879e+00 -1.82873964e+00 2.48129547e-01 -3.71823847e-01
2.31670007e-01 -2.88667053e-01 -4.48314957e-02 2.88531240e-02
-2.90961862e-01 1.08023860e-01 -4.96671051e-02 -8.92341673e-01
6.57655597e-02 2.91527957e-01 1.26040921e-01 5.64612508e-01
-1.87646329e-01 8.77897084e-01 -6.98602736e-01 -7.03829706e-01
-3.24193805e-01 6.00827575e-01 -5.61164796e-01 1.20896675e-01
-3.24930958e-02 4.51569557e-01 -1.06748663e-01 9.00600076e-01
6.64994240e-01 1.76410228e-01 -2.57954061e-01 -1.12231798e-01
2.59728938e-01 -2.87854493e-01 -4.12668735e-01 1.68674469e+00
-2.72516996e-01 2.51307487e-01 -1.51951790e-01 -3.85937124e-01
7.68236339e-01 1.76145941e-01 6.09402180e-01 -6.74510360e-01
-5.62582053e-02 -2.92717498e-02 -1.48473009e-01 -2.29886562e-01
2.99046218e-01 -1.83121771e-01 -2.52150804e-01 3.12715173e-01
-1.30632058e-01 -1.38391733e-01 6.58387765e-02 5.05684167e-02
1.11337698e+00 2.75106609e-01 -1.01309836e-01 1.46002039e-01
-1.02587948e-02 -4.52250272e-01 7.21644402e-01 3.82322550e-01
-5.46352923e-01 9.94274676e-01 2.67688990e-01 -5.92284203e-01
-1.59848702e+00 -1.54153526e+00 1.03685006e-01 1.24434960e+00
-3.48640949e-01 6.37799203e-02 -1.14289808e+00 -7.69702852e-01
3.69614810e-02 7.11308658e-01 -1.12311673e+00 -4.97054756e-01
-6.24698639e-01 -7.09879220e-01 5.12131453e-01 8.52406740e-01
4.85178649e-01 -8.94159675e-01 -2.00416651e-02 -2.29944158e-02
-6.23558648e-02 -1.01823699e+00 -9.22153950e-01 -6.32974863e-01
-9.64537680e-01 -8.21852148e-01 -1.37999737e+00 -7.66767740e-01
9.95651305e-01 -5.71115017e-01 1.29418123e+00 -1.57153472e-01
-3.63489419e-01 9.96601999e-01 -5.91921285e-02 -5.08760035e-01
-6.44210756e-01 2.07479179e-01 4.12031621e-01 -6.07014336e-02
2.16587335e-01 -9.41975117e-01 -1.27364111e+00 1.37746856e-01
-4.93087113e-01 5.65533750e-02 2.87249863e-01 5.60722351e-01
2.38344818e-01 -3.17470610e-01 3.89606535e-01 -6.55780077e-01
1.67190656e-01 -1.77106157e-01 1.27372339e-01 3.54356855e-01
-1.06349361e+00 -5.80739826e-02 2.07651109e-01 -8.77526224e-01
-1.14638662e+00 2.63466358e-01 -1.32266343e-01 -3.26336771e-01
-2.82746494e-01 -3.02639157e-01 -2.60484457e-01 1.40878499e-01
6.45561397e-01 6.73881173e-03 2.40311958e-02 -4.55618560e-01
3.90162289e-01 2.74156690e-01 1.16961324e+00 -7.90727317e-01
8.88739705e-01 5.46533287e-01 9.21089575e-02 -4.44160581e-01
-8.89716148e-01 2.04521284e-01 -9.73288000e-01 -3.75703126e-01
1.05194616e+00 -1.07467997e+00 -2.18847290e-01 1.18030119e+00
-6.38347149e-01 -5.84421158e-01 -2.84916341e-01 1.26974201e-02
-7.33084559e-01 4.25004184e-01 -8.67580116e-01 -5.16717434e-01
-7.72007227e-01 -7.34550297e-01 1.22499931e+00 5.55535257e-01
-4.36960161e-01 -1.05044830e+00 8.63810182e-02 6.32355452e-01
9.26604867e-03 4.94539320e-01 1.07134724e+00 -1.10613041e-01
-6.31469637e-02 -3.18963468e-01 8.51573497e-02 4.02955472e-01
5.10998182e-02 1.63051546e-01 -8.24544549e-01 -5.13295829e-01
-6.76307976e-01 -5.07108867e-01 8.30340922e-01 7.69074976e-01
1.23908484e+00 -2.71508247e-01 -3.13345730e-01 6.56390250e-01
1.06087446e+00 9.71770734e-02 1.06504679e+00 1.68648601e-01
8.81717205e-01 5.34570038e-01 3.85948062e-01 4.08088386e-01
8.39140117e-01 4.53814566e-01 3.04265827e-01 -2.55821586e-01
-4.88737345e-01 -5.84691703e-01 5.80750704e-01 5.03484309e-01
-3.61977100e-01 -1.66046154e-02 -9.22012269e-01 6.70468748e-01
-1.33496511e+00 -8.14621627e-01 2.27858305e-01 2.18701291e+00
9.41782713e-01 1.86316967e-01 4.76891786e-01 6.25578538e-02
7.45967925e-01 1.52922794e-01 -8.83134425e-01 -3.58212799e-01
-6.56152815e-02 5.20088911e-01 5.44179738e-01 2.07317322e-01
-1.08549476e+00 8.65547121e-01 7.75183439e+00 3.48720908e-01
-8.16540480e-01 1.09733172e-01 1.13111281e+00 -1.90092653e-01
-2.30198920e-01 -2.55918413e-01 -9.51154828e-01 4.62929964e-01
9.26229894e-01 -2.13971157e-02 4.51746583e-01 7.25028872e-01
-2.51899683e-03 8.89177695e-02 -1.24857593e+00 8.84322166e-01
3.82880181e-01 -4.87279147e-01 -1.89553693e-01 6.84742332e-02
8.84568930e-01 -5.44755518e-01 6.76647484e-01 3.00294578e-01
4.75494325e-01 -1.03975415e+00 8.95081580e-01 9.59012449e-01
1.26408172e+00 -6.10736310e-01 -9.32815075e-02 -4.82519232e-02
-9.04605091e-01 -2.61433005e-01 5.65943792e-02 1.23155057e-01
3.51661928e-02 2.49660075e-01 -6.03789389e-01 -8.30971152e-02
9.21364605e-01 5.06513417e-01 -1.09671557e+00 7.29608595e-01
-3.69646698e-01 6.60489202e-01 -1.94885269e-01 6.18414402e-01
-5.47029018e-01 1.96256563e-02 1.37525439e-01 6.78669155e-01
7.23636985e-01 -6.58314005e-02 -1.33567661e-01 3.93082529e-01
-3.18067789e-01 -6.62037209e-02 -3.14689189e-01 -3.05822436e-02
4.09186661e-01 8.98186445e-01 -3.96332085e-01 -2.52463222e-01
-5.76675117e-01 1.34530854e+00 2.01340228e-01 4.20456141e-01
-7.72360802e-01 2.67700225e-01 7.17683196e-01 7.08021700e-01
3.56683582e-02 -2.18944997e-01 -4.68340784e-01 -1.06818664e+00
-1.15227267e-01 -9.86616254e-01 5.37521243e-01 -1.10209739e+00
-1.42562199e+00 4.43558432e-02 4.86316271e-02 -6.66075647e-01
-4.45213556e-01 -3.14154446e-01 -6.93158567e-01 6.21466339e-01
-7.91596770e-01 -1.90294909e+00 -4.49692935e-01 4.15075809e-01
5.61196268e-01 -2.67856773e-02 6.59005821e-01 4.00605768e-01
-6.26491904e-01 1.13107586e+00 -1.63707197e-01 3.80253717e-02
1.33914053e+00 -1.55728900e+00 9.40567911e-01 9.67780650e-01
-5.26214659e-01 1.69993430e-01 8.14109206e-01 -1.04102910e+00
-8.80015850e-01 -1.04940987e+00 8.04840147e-01 -5.13780296e-01
4.30551499e-01 -2.38840804e-01 -6.49684966e-01 8.98281991e-01
9.25270617e-02 -7.06577897e-02 4.24115300e-01 2.27898836e-01
-4.56655860e-01 -3.70127171e-01 -1.22311556e+00 9.58979368e-01
1.42338443e+00 -2.98493981e-01 -3.22782576e-01 -3.70478593e-02
6.91907823e-01 -3.99067730e-01 -1.23958039e+00 5.60377896e-01
1.09257519e+00 -7.11247325e-01 1.36476898e+00 -6.02116287e-01
6.86102867e-01 3.38250071e-01 9.98002663e-02 -1.39097762e+00
-3.17829549e-01 -4.19357955e-01 -4.70978498e-01 1.53846598e+00
1.27430230e-01 -3.13843966e-01 1.26781642e+00 1.28415644e+00
-9.65691954e-02 -6.89276695e-01 -7.80553162e-01 -4.28149968e-01
5.48708379e-01 -8.79371818e-03 7.97081888e-01 5.19393981e-01
-7.19128072e-01 5.65999560e-02 -6.48144782e-01 1.52358592e-01
1.03106630e+00 -3.73736352e-01 7.61371851e-01 -1.29293323e+00
-6.34511784e-02 -1.90046847e-01 -4.44011271e-01 -5.68676114e-01
3.04001659e-01 -3.23996991e-01 -2.38918409e-01 -1.52063620e+00
2.85604954e-01 -1.76484853e-01 -4.32980694e-02 6.35537982e-01
-4.22408700e-01 4.20989245e-01 -6.76284432e-02 -1.93212315e-01
-8.58603716e-02 4.89612371e-01 1.53126299e+00 -4.68014836e-01
8.94512013e-02 3.39103580e-01 -6.19671941e-01 9.04247761e-01
7.01013982e-01 -2.67444849e-01 -4.87530977e-01 -2.81492054e-01
2.48869568e-01 -2.15004105e-02 4.36265707e-01 -1.36011147e+00
-5.50347827e-02 -2.86772419e-02 1.12195456e+00 -5.09863317e-01
7.39797711e-01 -3.81969362e-01 2.12425888e-01 4.76803094e-01
-2.17290103e-01 2.33282879e-01 -1.77526712e-01 3.69003862e-01
3.63072246e-01 -6.17955476e-02 9.97609794e-01 2.75527295e-02
-5.29481292e-01 1.01235354e+00 -1.25501007e-01 2.94568390e-01
7.41389275e-01 -3.35308820e-01 -3.80286276e-01 -7.02729166e-01
-1.42318881e+00 1.44498304e-01 1.17259347e+00 5.24142563e-01
3.70588750e-01 -1.51555645e+00 -7.40543008e-01 1.03017092e-01
-5.34051843e-02 -2.06114486e-01 3.75837326e-01 3.82834405e-01
-4.58898306e-01 -3.80351126e-01 -5.61442494e-01 -1.84416607e-01
-1.69463563e+00 7.20794082e-01 3.23679566e-01 -2.04788625e-01
-4.04122740e-01 7.32384801e-01 3.34218472e-01 -6.53993487e-02
3.69616926e-01 -1.73656583e-01 -9.22224671e-02 1.34134829e-01
3.16458851e-01 6.02838695e-01 -3.51563811e-01 -6.05827391e-01
-2.40724422e-02 7.50483811e-01 -2.28523985e-01 -2.27869332e-01
1.21386695e+00 -4.26160127e-01 1.52806357e-01 4.28916603e-01
8.55893612e-01 -1.30138412e-01 -1.70542872e+00 -1.13573605e-02
-4.25941437e-01 -2.21116245e-01 -3.15545499e-01 -8.45996737e-01
-1.12770283e+00 8.23571503e-01 1.19952798e+00 -3.08906585e-01
1.42288744e+00 5.30534275e-02 1.16185260e+00 -3.16927016e-01
2.06583291e-01 -1.22342682e+00 3.72688502e-01 2.93894857e-02
7.64871657e-01 -1.10630310e+00 2.93699712e-01 -2.62564510e-01
-5.55583775e-01 8.07816148e-01 9.67387676e-01 8.80917651e-04
4.94484752e-01 1.43028155e-01 4.28152084e-03 2.46525794e-01
-5.43455720e-01 3.70131955e-02 4.98869121e-01 1.00712502e+00
2.97289163e-01 2.79271632e-01 -2.40064561e-01 4.56661165e-01
-5.13683140e-01 1.05135389e-01 2.85963178e-01 7.49366701e-01
-2.09452808e-01 -1.52981985e+00 -6.16602600e-01 6.22211635e-01
-6.83231056e-01 1.37701839e-01 -3.43761355e-01 7.16227829e-01
6.12590194e-01 4.36526775e-01 2.40610659e-01 -2.59484798e-01
2.30842486e-01 5.12560368e-01 1.12451637e+00 -3.77460629e-01
-4.80744876e-02 -1.90108463e-01 2.52033442e-01 -3.60591263e-01
-2.29469627e-01 -1.23469436e+00 -8.44327331e-01 -4.70840961e-01
4.72669095e-01 -6.56178713e-01 5.57236731e-01 6.67441249e-01
1.04272189e-02 2.07181394e-01 4.82827306e-01 -8.77526820e-01
-3.63115102e-01 -1.01825964e+00 -6.35997772e-01 6.37856305e-01
2.84541219e-01 -6.72251761e-01 -1.34932429e-01 6.56388998e-01]
|
[13.185416221618652, 0.4367968440055847]
|
790fcb8e-c8bf-4bb9-8ee3-b93e0432902c
|
homography-estimation-from-the-common-self
| null | null |
http://openaccess.thecvf.com/content_cvpr_2016/html/Huang_Homography_Estimation_From_CVPR_2016_paper.html
|
http://openaccess.thecvf.com/content_cvpr_2016/papers/Huang_Homography_Estimation_From_CVPR_2016_paper.pdf
|
Homography Estimation From the Common Self-Polar Triangle of Separate Ellipses
|
How to avoid ambiguity is a challenging problem for conic-based homography estimation. In this paper, we address the problem of homography estimation from two separate ellipses. We find that any two ellipses have a unique common self-polar triangle, which can provide three line correspondences. Furthermore, by investigating the location features of the common self-polar triangle, we show that one vertex of the triangle lies outside of both ellipses, while the other two vertices lies inside the ellipses separately. Accordingly, one more line correspondence can be obtained from the intersections of the conics and the common self-polar triangle. Therefore, four line correspondences can be obtained based on the common self-polar triangle, which can provide enough constraints for the homography estimation. The main contributions in this paper include: (1) A new discovery on the location features of the common self-polar triangle of separate ellipses. (2) A novel approach for homography estimation. Simulate experiments and real experiments are conducted to demonstrate the feasibility and accuracy of our approach.
|
['Yiu-ming Cheung', 'HUI ZHANG', 'Haifei Huang']
|
2016-06-01
| null | null | null |
cvpr-2016-6
|
['homography-estimation']
|
['computer-vision']
|
[-4.44446243e-02 2.51642048e-01 -6.03970401e-02 1.06220908e-01
-5.79285100e-02 -6.94556057e-01 4.06063586e-01 -3.57106626e-01
1.47716984e-01 5.80945969e-01 -2.12948620e-01 -2.15201259e-01
-1.58601761e-01 -6.70812488e-01 -7.79852986e-01 -6.26552939e-01
9.52461064e-02 7.67958760e-01 2.96868622e-01 -2.72076577e-01
2.74440050e-01 7.08592534e-01 -1.01274788e+00 -5.79516888e-01
1.05317461e+00 7.36656606e-01 -4.58698794e-02 3.97866845e-01
2.09048957e-01 -4.18914586e-01 -5.64807594e-01 -3.41506600e-01
6.15286291e-01 -3.91852319e-01 -4.41318125e-01 2.76530176e-01
5.18289626e-01 -3.28419209e-01 -2.03340352e-01 1.21805179e+00
7.44315311e-02 -2.18924761e-01 6.70332730e-01 -1.54397786e+00
-2.03376874e-01 1.05288118e-01 -1.07609785e+00 -5.20510852e-01
5.76926470e-01 -3.53080392e-01 8.40028286e-01 -6.60591543e-01
8.71644080e-01 1.12982059e+00 6.20968282e-01 -1.65772647e-01
-9.11058247e-01 -6.43265128e-01 -3.92126888e-01 -2.07910135e-01
-1.71622396e+00 -3.13968152e-01 9.72737610e-01 -3.27667564e-01
4.32720482e-01 3.89567822e-01 8.21283877e-01 3.15919876e-01
2.92365342e-01 3.29234004e-01 1.10680068e+00 -8.01593721e-01
-3.13152969e-01 3.83054554e-01 6.10659085e-02 9.15183783e-01
5.02777100e-01 2.05591515e-01 5.78828109e-03 -2.36266077e-01
1.32664466e+00 -1.50771812e-01 -2.74714023e-01 -9.83898401e-01
-1.28900445e+00 7.40393579e-01 -1.95613857e-02 4.30303097e-01
1.46481618e-01 -2.64992058e-01 -3.08132917e-01 -1.63335562e-01
-6.93135858e-02 4.22645360e-01 1.06537677e-01 -1.11508332e-01
-6.93169236e-01 1.01008885e-01 1.07723427e+00 1.63052976e+00
1.00181246e+00 -9.52979177e-02 6.41306758e-01 6.34228587e-01
4.21601713e-01 8.96286964e-01 2.44125593e-02 -8.43501389e-01
5.35719037e-01 5.58125913e-01 1.47770211e-01 -1.60274148e+00
-5.21214008e-01 -3.75706464e-01 -7.92266548e-01 -1.67882144e-01
7.74662852e-01 -3.97241652e-01 -3.72159123e-01 1.31221902e+00
3.70460659e-01 1.04268722e-01 -6.17510406e-03 8.17133904e-01
4.99461323e-01 4.15219009e-01 -8.26020956e-01 -3.66361380e-01
1.71837711e+00 -9.09336865e-01 -7.75131106e-01 -3.40352133e-02
4.17269289e-01 -1.17363274e+00 4.35720235e-01 6.69576111e-04
-1.03540075e+00 -2.35118315e-01 -1.46290112e+00 1.37809426e-01
-2.25481228e-03 6.67199671e-01 4.30893958e-01 5.71848392e-01
-6.58389449e-01 3.58581752e-01 -6.45688593e-01 -4.45995182e-01
-6.39817894e-01 4.65153277e-01 -5.17287135e-01 6.13219798e-01
-9.42498028e-01 7.98267603e-01 4.48904008e-01 1.37761667e-01
2.63634771e-01 -2.42919862e-01 -8.23811948e-01 1.06597379e-01
6.53042078e-01 -6.71257377e-01 8.34032893e-01 -4.55895960e-01
-1.58619201e+00 8.47239196e-01 -3.05285722e-01 -6.56578615e-02
7.09045827e-01 8.06487128e-02 -5.10760844e-01 2.77366608e-01
2.13617086e-01 2.63615876e-01 7.45622873e-01 -1.46641982e+00
-6.67648196e-01 -4.52186108e-01 -3.03232729e-01 3.04469317e-01
-8.81135091e-02 -7.98973590e-02 -8.58061433e-01 -3.65096420e-01
9.36012685e-01 -1.41554177e+00 1.45991789e-02 -3.80320489e-01
-1.09170103e+00 -7.46938214e-02 9.36822116e-01 -5.90635955e-01
1.48823380e+00 -2.14251113e+00 -7.62774283e-03 8.16136658e-01
3.88843387e-01 5.25605679e-02 4.09782112e-01 4.79975432e-01
-1.87146276e-01 2.11000606e-01 -5.46557307e-02 2.33502626e-01
-3.83266509e-02 -1.16414301e-01 -3.55440825e-01 8.30270529e-01
-1.83360532e-01 4.18265730e-01 -4.92177784e-01 -4.75958765e-01
6.65277466e-02 6.41438141e-02 -3.03338431e-02 3.53353880e-02
3.49825531e-01 2.36638159e-01 -4.51729834e-01 2.88390696e-01
1.11085665e+00 1.87500894e-01 3.76397610e-01 -6.20299578e-01
-5.84898531e-01 -1.97187498e-01 -1.83920550e+00 6.61600053e-01
-6.43831044e-02 5.56418121e-01 7.26269409e-02 -4.95337933e-01
1.43287683e+00 1.97142586e-01 6.94925427e-01 -3.46356779e-01
2.42294654e-01 4.81845915e-01 9.02865082e-02 -2.60565411e-02
7.03740656e-01 1.58342034e-01 9.65723842e-02 3.53282779e-01
-1.68877944e-01 -4.24032450e-01 3.55004013e-01 -1.93874277e-02
1.30718976e-01 -1.08660601e-01 7.08471715e-01 -4.76959735e-01
7.17990398e-01 -3.60256344e-01 7.02540100e-01 3.21092486e-01
-2.04582177e-02 6.20775700e-01 8.93993437e-01 -3.50262582e-01
-1.31343591e+00 -8.11882436e-01 -5.68804741e-01 -2.39112213e-01
8.70559573e-01 -6.67100728e-01 -7.65811980e-01 -2.07130611e-01
1.64466843e-01 2.35805422e-01 -1.76145315e-01 1.96702898e-01
-7.53108382e-01 -2.37277612e-01 1.53689936e-01 2.61291444e-01
6.30310059e-01 -9.76909027e-02 -5.89187205e-01 -1.90141335e-01
-2.02822402e-01 -1.39484799e+00 -6.66142166e-01 -3.18342686e-01
-7.99705088e-01 -1.45309508e+00 -8.06114495e-01 -7.74044991e-01
9.90301430e-01 7.00771391e-01 5.84867358e-01 3.20337899e-02
2.44273424e-01 3.54744703e-01 1.48577914e-01 -2.30712861e-01
-2.58283317e-01 -1.81066722e-01 1.58746347e-01 6.83847815e-02
1.35282308e-01 -4.56362844e-01 -3.59640360e-01 1.31683552e+00
-2.00220138e-01 2.44779736e-01 1.95984483e-01 5.24136901e-01
7.82325089e-01 2.10990921e-01 3.90586653e-03 -4.84395087e-01
5.42311728e-01 -1.53931588e-01 -1.31859636e+00 4.26236242e-01
-4.25619781e-01 5.86454049e-02 5.61537087e-01 -2.48147026e-01
-8.76894534e-01 2.81156898e-01 3.47178072e-01 -4.46301579e-01
-3.39393504e-02 6.68013990e-02 -4.30429786e-01 -4.69265193e-01
1.63166553e-01 5.05850129e-02 1.22546680e-01 -4.08535868e-01
4.51543421e-01 5.89478374e-01 6.91554368e-01 -5.07298946e-01
1.15629971e+00 6.49389446e-01 6.52271211e-01 -1.30924416e+00
-1.07600845e-01 -6.12558782e-01 -1.09706414e+00 -1.70486122e-01
7.00571656e-01 -5.29586315e-01 -9.81828570e-01 3.32603097e-01
-1.37482297e+00 5.65957546e-01 1.48234189e-01 7.52033949e-01
-7.28959262e-01 1.09342539e+00 -2.48736203e-01 -6.99767351e-01
-1.60351872e-01 -1.42248583e+00 1.03942609e+00 4.92706299e-01
-2.64839560e-01 -1.00986469e+00 2.07668766e-01 8.20672140e-02
-3.94633681e-01 2.35699639e-01 6.19061291e-01 -5.48590183e-01
-7.48700857e-01 -3.29604089e-01 -2.21217349e-01 -2.36648932e-01
-4.57261689e-02 6.25099480e-01 -2.06483647e-01 -2.16555685e-01
8.48364681e-02 5.26603341e-01 3.65950167e-01 3.61176550e-01
2.78058708e-01 -2.34499305e-01 -8.74618769e-01 1.21992981e+00
1.28431618e+00 7.19232976e-01 7.20367789e-01 5.72955012e-01
7.16574907e-01 5.12292743e-01 7.20737100e-01 2.22115204e-01
2.91852862e-01 1.05824876e+00 9.80571732e-02 -6.90679252e-02
1.67904183e-01 -4.23376739e-01 4.70377365e-03 9.31695044e-01
-2.75264889e-01 2.22176854e-02 -7.78527975e-01 2.96634018e-01
-1.86075032e+00 -4.83744591e-01 -7.65080452e-01 2.50230789e+00
3.24804246e-01 -6.36884645e-02 1.70215964e-01 -2.44937430e-04
1.00902748e+00 -7.45081380e-02 -1.38546333e-01 -4.36858505e-01
-2.56334454e-01 -4.91063178e-01 6.80392861e-01 8.09190869e-01
-1.12595427e+00 9.57502127e-01 6.46974277e+00 5.95342815e-01
-9.57118511e-01 -6.20334625e-01 -2.21489787e-01 7.39666939e-01
-3.24638754e-01 7.23379433e-01 -1.39721668e+00 4.90155965e-01
5.70360757e-02 -4.85832393e-01 5.11671929e-03 7.72683501e-01
-2.44018361e-01 -5.01801372e-01 -9.21692491e-01 1.08635557e+00
2.57905573e-01 -9.80828047e-01 -7.71469921e-02 4.10776615e-01
9.35783625e-01 -8.14138889e-01 5.21488376e-02 -4.51684147e-01
-2.19518468e-01 -4.28110421e-01 3.71324658e-01 3.48702550e-01
9.29559946e-01 -6.15090907e-01 6.71848059e-01 2.91822404e-01
-1.48898959e+00 3.71355593e-01 -4.59477305e-01 1.82979405e-01
3.70613813e-01 4.64215696e-01 -1.02605176e+00 9.99832451e-01
1.45556256e-01 4.68023092e-01 -3.39535832e-01 1.17713523e+00
-3.66361052e-01 -6.54565543e-02 -6.86652005e-01 1.43443555e-01
-1.79715212e-02 -1.29319680e+00 1.12788391e+00 8.24211538e-01
6.88142657e-01 2.25398481e-01 1.14461325e-01 1.12529767e+00
1.54369563e-01 3.39901477e-01 -8.19518209e-01 8.68400484e-02
6.31183147e-01 1.17361605e+00 -8.39089513e-01 -1.40338898e-01
-3.98958743e-01 4.98831689e-01 -2.36745119e-01 3.16648811e-01
-6.75950348e-01 -8.51855874e-01 3.72988492e-01 1.01329729e-01
1.03145048e-01 -5.23587763e-01 -4.93065715e-01 -1.26054919e+00
1.57378510e-01 -7.80242503e-01 -1.75063610e-02 -7.04142272e-01
-5.11463821e-01 4.84713852e-01 2.24906296e-01 -1.45216393e+00
-4.26252693e-01 -6.52708948e-01 -7.69346118e-01 9.67313528e-01
-8.26608896e-01 -1.15089273e+00 -2.26677194e-01 4.99664873e-01
-2.14929491e-01 -1.97573856e-01 6.85760558e-01 -6.63165003e-02
-4.90321249e-01 5.44997334e-01 4.91294831e-01 1.06832467e-01
7.12865591e-01 -1.13780117e+00 2.10559562e-01 9.20682251e-01
1.35666564e-01 9.28954065e-01 5.27397811e-01 -9.91292298e-01
-1.41557336e+00 -1.76671475e-01 9.40998137e-01 -2.02815145e-01
7.48468876e-01 -3.14615667e-01 -6.73580825e-01 9.77173448e-01
-1.71213355e-02 -6.29373252e-01 2.66341716e-01 9.21979919e-02
-6.18936233e-02 -5.89442812e-03 -6.72413290e-01 7.30355263e-01
6.62205040e-01 -1.82205111e-01 -5.78471482e-01 3.76059413e-01
3.13845187e-01 -7.19268918e-01 -9.09471452e-01 5.30569375e-01
9.44944263e-01 -1.07453096e+00 9.41025138e-01 -2.01592252e-01
8.98702368e-02 -6.12362862e-01 1.52879506e-01 -1.15449429e+00
-9.61626694e-02 -1.00830412e+00 1.57008618e-01 1.14859629e+00
1.39503226e-01 -9.27938044e-01 7.24776924e-01 3.18152755e-01
3.06150187e-02 -5.81200600e-01 -7.94256687e-01 -9.96183157e-01
-7.73771554e-02 2.83701748e-01 6.38030291e-01 9.24096882e-01
2.95553863e-01 4.93886918e-01 -5.99579453e-01 3.84115845e-01
6.62351668e-01 6.65384173e-01 1.21386397e+00 -1.34215713e+00
-1.13690123e-01 -3.87159586e-01 -4.95262474e-01 -1.58992350e+00
-8.34705234e-02 -3.76810104e-01 -3.65134776e-01 -1.12656999e+00
3.88690531e-02 -5.35574853e-01 6.89819694e-01 2.31806072e-03
2.74231941e-01 -9.43915453e-03 3.37715358e-01 5.84680319e-01
9.30880569e-03 3.26411128e-01 1.66763663e+00 2.84723222e-01
-5.12015462e-01 1.90977797e-01 -2.71723539e-01 1.22449648e+00
6.92467749e-01 -9.72641259e-02 -1.79717913e-01 -5.54820001e-02
1.00287437e-01 5.00986218e-01 -7.85993338e-02 -8.10409784e-01
4.39663947e-01 -1.69258583e-02 2.23286375e-01 -1.02624965e+00
4.35607165e-01 -9.74251091e-01 4.09798384e-01 3.32289845e-01
6.16436839e-01 1.22684084e-01 3.11926398e-02 3.19788098e-01
-2.19255328e-01 -4.98033851e-01 6.29271030e-01 1.12586439e-01
-2.23373055e-01 -1.60751939e-02 3.97034138e-02 -2.58097529e-01
1.35938799e+00 -7.58638203e-01 -4.71384257e-01 -6.24220014e-01
-6.84101462e-01 3.13411921e-01 8.55038404e-01 2.11356152e-02
4.39654380e-01 -1.40807736e+00 -2.52215147e-01 6.86483085e-01
-9.05571952e-02 -1.39361590e-01 -6.10953458e-02 1.07156813e+00
-8.33001792e-01 9.60765362e-01 -4.07642990e-01 -8.23254645e-01
-1.53085101e+00 4.68859106e-01 5.22854507e-01 9.00395289e-02
-3.89102787e-01 2.61626989e-01 4.08881068e-01 -3.69937569e-01
-5.61976843e-02 -1.12580910e-01 -3.35672021e-01 -4.08928879e-02
1.50228105e-02 7.76255906e-01 -3.97018850e-01 -1.27299237e+00
-2.34300882e-01 1.40640283e+00 1.32803425e-01 -1.96127772e-01
8.50992084e-01 -1.75088987e-01 -4.27593797e-01 5.78125045e-02
1.28654480e+00 7.51029253e-01 -8.17954361e-01 2.24807616e-02
-1.99135944e-01 -7.68092036e-01 -6.31387413e-01 -2.66056448e-01
-7.94769585e-01 8.01196098e-01 1.61429733e-01 3.01661104e-01
7.67684340e-01 -1.23515785e-01 7.16254771e-01 2.52979696e-01
2.65170664e-01 -1.05503750e+00 -2.80571640e-01 4.60705429e-01
9.72390890e-01 -6.63241386e-01 2.37243846e-01 -1.30622125e+00
-4.14140224e-01 1.80676615e+00 6.43966913e-01 -2.44717181e-01
5.00347435e-01 6.35350347e-02 -1.98868766e-01 -7.36291558e-02
9.42381769e-02 9.09089074e-02 6.70873344e-01 3.98641616e-01
1.98155984e-01 1.05116338e-01 -8.71621907e-01 4.06942606e-01
-6.82080805e-01 -6.17463887e-01 9.25358176e-01 5.65789878e-01
-5.37786722e-01 -1.30988169e+00 -9.17015374e-01 -8.53279755e-02
7.66875520e-02 3.56606036e-01 -6.99465394e-01 1.31676066e+00
7.42473975e-02 7.74476707e-01 3.10005426e-01 -3.94286960e-01
4.77640867e-01 -3.32060486e-01 5.72712183e-01 -2.75893901e-02
2.09159434e-01 6.57712460e-01 1.36365652e-01 -3.35204899e-01
-8.35614577e-02 -5.20500600e-01 -1.18894768e+00 -4.44818377e-01
-7.55337417e-01 3.60057265e-01 6.95294261e-01 6.07277393e-01
2.19321087e-01 -1.66731775e-01 8.27892661e-01 -2.57133067e-01
-6.26966298e-01 -4.93468225e-01 -9.20128584e-01 1.93393007e-01
1.02411501e-01 -8.42274010e-01 -5.66892743e-01 -2.14460626e-01]
|
[7.997897624969482, -2.3076441287994385]
|
92e72a0e-b742-40a9-a8a2-edeb3d0d20a0
|
bsn-boundary-sensitive-network-for-temporal
|
1806.02964
| null |
http://arxiv.org/abs/1806.02964v3
|
http://arxiv.org/pdf/1806.02964v3.pdf
|
BSN: Boundary Sensitive Network for Temporal Action Proposal Generation
|
Temporal action proposal generation is an important yet challenging problem,
since temporal proposals with rich action content are indispensable for
analysing real-world videos with long duration and high proportion irrelevant
content. This problem requires methods not only generating proposals with
precise temporal boundaries, but also retrieving proposals to cover truth
action instances with high recall and high overlap using relatively fewer
proposals. To address these difficulties, we introduce an effective proposal
generation method, named Boundary-Sensitive Network (BSN), which adopts "local
to global" fashion. Locally, BSN first locates temporal boundaries with high
probabilities, then directly combines these boundaries as proposals. Globally,
with Boundary-Sensitive Proposal feature, BSN retrieves proposals by evaluating
the confidence of whether a proposal contains an action within its region. We
conduct experiments on two challenging datasets: ActivityNet-1.3 and THUMOS14,
where BSN outperforms other state-of-the-art temporal action proposal
generation methods with high recall and high temporal precision. Finally,
further experiments demonstrate that by combining existing action classifiers,
our method significantly improves the state-of-the-art temporal action
detection performance.
|
['Ming Yang', 'Tianwei Lin', 'Haisheng Su', 'Xu Zhao', 'Chongjing Wang']
|
2018-06-08
|
bsn-boundary-sensitive-network-for-temporal-1
|
http://openaccess.thecvf.com/content_ECCV_2018/html/Tianwei_Lin_BSN_Boundary_Sensitive_ECCV_2018_paper.html
|
http://openaccess.thecvf.com/content_ECCV_2018/papers/Tianwei_Lin_BSN_Boundary_Sensitive_ECCV_2018_paper.pdf
|
eccv-2018-9
|
['temporal-action-proposal-generation']
|
['computer-vision']
|
[ 4.63551342e-01 -6.77311122e-02 -6.04116201e-01 -3.97948548e-02
-8.90833616e-01 -2.34355554e-01 7.37309396e-01 -9.06959176e-02
-5.11435628e-01 8.68564725e-01 5.74529588e-01 2.92904437e-01
2.66257469e-02 -7.15324223e-01 -4.23069507e-01 -6.43475711e-01
-2.55178332e-01 1.78740278e-01 1.41681600e+00 -3.07929981e-02
3.08690280e-01 2.90780872e-01 -1.58951199e+00 5.14544487e-01
4.21999902e-01 1.20917046e+00 -7.05568716e-02 2.55277157e-01
1.56449109e-01 9.92795944e-01 -7.17294753e-01 -1.44497022e-01
3.44108313e-01 -4.98140693e-01 -8.71286333e-01 1.01912126e-01
3.59930187e-01 -6.88173413e-01 -3.65835279e-01 8.25883269e-01
3.76047313e-01 4.64242309e-01 4.86432850e-01 -1.40807319e+00
-6.87266979e-03 6.26794279e-01 -6.99541748e-01 5.59748769e-01
6.26025200e-01 5.36419749e-01 8.98709834e-01 -6.52026176e-01
9.54123795e-01 1.24773228e+00 5.51565349e-01 6.00963891e-01
-8.70330989e-01 -7.78473318e-01 6.63456619e-01 4.78208423e-01
-1.37617636e+00 -4.36602116e-01 6.66272283e-01 -1.49839535e-01
9.28650498e-01 2.11860165e-01 9.29383814e-01 1.32038498e+00
1.31764993e-01 1.33033824e+00 7.13237584e-01 1.22191235e-02
4.04703856e-01 -5.42757750e-01 -4.57903773e-01 4.74619240e-01
-2.36009564e-02 7.68792629e-02 -6.68752670e-01 -2.34575607e-02
9.78324294e-01 1.43284485e-01 -1.84042603e-01 -2.62014300e-01
-1.80419886e+00 3.89834642e-01 2.95712233e-01 5.28832376e-01
-6.53925955e-01 4.12578493e-01 7.30511129e-01 -2.32869610e-01
3.58549863e-01 1.35856077e-01 -3.52262199e-01 -5.06723881e-01
-1.06025660e+00 4.39561009e-01 1.77077174e-01 9.18094754e-01
4.01267081e-01 -1.33278698e-01 -9.95006621e-01 5.87557912e-01
5.81390671e-02 2.32125089e-01 6.03987038e-01 -1.28012073e+00
7.00698137e-01 6.73679411e-01 4.19255286e-01 -9.59382534e-01
-1.85838565e-01 -1.43063024e-01 -7.43501484e-01 -1.14332713e-01
3.77332777e-01 1.54692858e-01 -8.65333378e-01 1.63010621e+00
6.61817849e-01 4.55949306e-01 -1.95111439e-01 9.13754225e-01
6.78496122e-01 6.60162926e-01 4.58312988e-01 -5.40316463e-01
1.29007292e+00 -1.17053246e+00 -7.77392626e-01 -1.78406864e-01
6.04491651e-01 -4.76777852e-01 6.70509756e-01 3.28439862e-01
-1.09624374e+00 -6.49619579e-01 -7.44348049e-01 2.66117662e-01
6.04662076e-02 3.92831862e-01 7.06546366e-01 3.79231796e-02
-8.61449540e-01 6.70868039e-01 -9.26994741e-01 -5.54271340e-01
7.90923655e-01 6.72700927e-02 -4.43734527e-01 -5.01473509e-02
-1.18061590e+00 6.52773321e-01 8.57198775e-01 3.95517275e-02
-1.16001153e+00 -2.86278188e-01 -8.76544774e-01 -1.12110242e-01
9.22823608e-01 -3.76388460e-01 1.53215551e+00 -8.21694016e-01
-1.28938222e+00 2.97691345e-01 -1.96529254e-01 -7.14834154e-01
9.45502222e-01 -1.37795851e-01 -4.99809802e-01 5.10595858e-01
6.05852365e-01 1.28617740e+00 8.26138377e-01 -6.17048800e-01
-1.16614902e+00 1.34829566e-01 1.45742655e-01 3.00045282e-01
-6.44499110e-03 1.37692466e-01 -6.71227813e-01 -8.72562230e-01
4.94339854e-01 -6.11347675e-01 -4.71654773e-01 3.32957447e-01
-2.12647602e-01 -8.47114861e-01 8.35710227e-01 -4.76304263e-01
1.41028988e+00 -1.96495652e+00 -2.38479540e-01 -5.92243485e-02
-9.10306126e-02 3.33977968e-01 -2.10907757e-01 2.35578820e-01
1.69954032e-01 -8.69935304e-02 -1.55930683e-01 4.11438979e-02
-3.86408158e-02 6.22686185e-02 -2.38122791e-01 2.70051479e-01
3.18226516e-01 9.74718630e-01 -1.43618309e+00 -1.03597105e+00
3.83695006e-01 4.27490175e-02 -3.59267384e-01 -7.57396221e-02
-4.17808473e-01 3.39911342e-01 -7.78914809e-01 9.60028768e-01
3.09623420e-01 -2.12894782e-01 2.79895961e-01 -2.88976818e-01
-4.42477800e-02 3.02636564e-01 -1.42004776e+00 1.74548733e+00
2.52756745e-01 3.46749723e-01 -4.90914971e-01 -9.05727446e-01
7.53271043e-01 4.05959398e-01 9.48681712e-01 -9.72089648e-01
-4.29164916e-02 5.13876639e-02 -3.32819015e-01 -5.53650320e-01
6.73425198e-01 4.68398593e-02 -1.04984671e-01 5.97547412e-01
-7.86061808e-02 3.08324963e-01 6.54050767e-01 3.60017896e-01
1.63753355e+00 8.45501065e-01 6.22003138e-01 3.26973438e-01
5.38645566e-01 -1.17150448e-01 1.05710256e+00 8.25592458e-01
-9.37829912e-01 5.35143197e-01 6.04997873e-01 -6.06568873e-01
-6.51921809e-01 -9.99504983e-01 3.06986690e-01 9.70284462e-01
3.66281152e-01 -4.20359850e-01 -5.07258058e-01 -1.33859384e+00
-3.26852828e-01 3.53219748e-01 -6.36510849e-01 -1.11808904e-01
-8.67837667e-01 -3.45394492e-01 3.88762504e-01 8.55434716e-01
1.00328946e+00 -1.76631868e+00 -8.83781016e-01 5.60388327e-01
-7.71717727e-01 -1.16525257e+00 -6.52326882e-01 -4.13993090e-01
-1.08461773e+00 -1.23372257e+00 -9.41456497e-01 -3.71028185e-01
6.14381909e-01 4.90960985e-01 1.05733562e+00 1.02232853e-02
-2.48156145e-01 2.24030852e-01 -6.92495704e-01 2.40804255e-02
-8.57983679e-02 -2.22821310e-01 -7.82824703e-04 5.11812456e-02
4.04525161e-01 -4.22307730e-01 -9.11785305e-01 8.76089871e-01
-8.53166044e-01 4.65833880e-02 8.13092768e-01 3.99302661e-01
8.27033579e-01 2.28762552e-01 5.64117789e-01 -1.10866949e-01
2.39025384e-01 -2.77203798e-01 -3.69053185e-01 2.77157843e-01
-1.55993281e-02 -1.45133808e-01 1.48519889e-01 -6.88791454e-01
-1.19132781e+00 1.99791491e-01 1.19962730e-01 -4.18930650e-01
-2.35653967e-01 5.27850632e-03 7.57565573e-02 3.96623969e-01
7.30900228e-01 4.57113177e-01 -2.67931163e-01 -2.55898356e-01
2.60061890e-01 9.52960104e-02 5.84385753e-01 -3.78915906e-01
4.28793520e-01 8.55137467e-01 -6.03935122e-02 -3.20175856e-01
-8.66924405e-01 -8.02095950e-01 -7.64849663e-01 -6.56186819e-01
8.03997695e-01 -8.02385032e-01 -6.54740453e-01 4.41954523e-01
-1.14290714e+00 -2.84269035e-01 -4.76981968e-01 6.46351933e-01
-7.39287794e-01 5.81152797e-01 -3.97370726e-01 -8.87606680e-01
-2.23427996e-01 -8.64844918e-01 1.19897282e+00 7.59560689e-02
-4.05077845e-01 -3.70260447e-01 7.89593980e-02 3.21084946e-01
1.38243124e-01 2.34996632e-01 -1.79416519e-02 -4.49611872e-01
-9.53236520e-01 -2.50970244e-01 -3.98798198e-01 1.06190503e-01
2.00674564e-01 -2.38649407e-03 -5.53499937e-01 -5.92971407e-02
-4.36433494e-01 -3.18312287e-01 1.09791660e+00 5.77830136e-01
1.18456399e+00 -3.60429198e-01 -7.10734367e-01 -1.03136100e-01
9.95850682e-01 2.58344352e-01 8.95801723e-01 3.18259627e-01
2.46657446e-01 5.61829925e-01 1.54515553e+00 7.30379641e-01
1.65688172e-01 9.58674192e-01 4.63321030e-01 1.84143543e-01
-1.03779547e-01 -3.26821178e-01 5.39961874e-01 1.74227610e-01
-4.48675990e-01 -2.07371056e-01 -4.11773503e-01 8.67646754e-01
-2.45162821e+00 -1.60703766e+00 1.03879392e-01 2.09060788e+00
7.86295950e-01 4.24735367e-01 5.84084570e-01 2.02572852e-01
8.72159302e-01 5.11298001e-01 -4.52953786e-01 4.05283898e-01
7.80996680e-02 -7.28239864e-02 2.76624411e-01 -1.76745728e-01
-1.58683980e+00 1.08239520e+00 6.15432262e+00 1.31298923e+00
-6.56856298e-01 2.20495299e-01 5.72946489e-01 -1.80289894e-01
2.64641583e-01 -3.60285081e-02 -8.59497130e-01 6.15499437e-01
3.95422310e-01 1.47903904e-01 -1.42475754e-01 8.80106688e-01
7.06952274e-01 -6.29682779e-01 -9.56382275e-01 7.83398807e-01
-2.61489078e-02 -1.34931374e+00 -1.79771837e-02 -1.61281362e-01
7.56704926e-01 -2.23101497e-01 -4.35775995e-01 5.24276853e-01
3.26129586e-01 -4.80227679e-01 9.86972094e-01 6.86506748e-01
4.99086618e-01 -4.01342958e-01 7.67313123e-01 1.83520615e-01
-1.76523936e+00 -1.21973298e-01 -3.43270600e-01 2.60212198e-02
6.05539083e-01 3.99521679e-01 -7.16870189e-01 4.04606551e-01
8.53055716e-01 1.25595808e+00 -4.34031188e-01 1.45671856e+00
-5.14005423e-01 4.79418695e-01 -3.21975976e-01 -1.48291677e-01
4.16119128e-01 1.06852025e-01 5.19362211e-01 1.05944550e+00
4.39033031e-01 4.67153102e-01 4.22731847e-01 6.52525365e-01
1.45633981e-01 2.90695541e-02 -4.28707868e-01 5.07469811e-02
4.47707266e-01 1.15168548e+00 -1.16898525e+00 -7.76355982e-01
-9.14689600e-02 8.05729747e-01 3.08991801e-02 2.14820087e-01
-1.28135419e+00 -5.42519987e-03 3.68314624e-01 1.47241503e-01
4.91165906e-01 3.50095543e-05 3.30202639e-01 -9.25492704e-01
4.22917455e-01 -5.87938726e-01 7.04582274e-01 -1.00611496e+00
-9.53628838e-01 2.10219815e-01 2.91812807e-01 -2.08602214e+00
-1.35642439e-01 -7.94566721e-02 -7.16785550e-01 1.22510254e-01
-1.05095291e+00 -1.10687649e+00 -3.40630919e-01 4.57713306e-01
1.07273853e+00 1.77795514e-01 1.79442316e-01 2.37661079e-01
-5.04691541e-01 2.13591903e-01 -6.43269062e-01 1.31111115e-01
6.31054103e-01 -7.59073198e-01 3.50036293e-01 9.07928824e-01
1.96318794e-02 1.56383112e-01 3.58580470e-01 -1.00780797e+00
-5.39786756e-01 -1.40853095e+00 9.64401305e-01 -4.37688380e-01
5.62206566e-01 1.80654690e-01 -6.82276964e-01 4.94788319e-01
-1.85564548e-01 2.58225709e-01 5.26168309e-02 -4.63463902e-01
-3.37383561e-02 -1.15363464e-01 -1.07142615e+00 8.02800119e-01
1.70798409e+00 -3.52927744e-02 -6.61484838e-01 5.74044168e-01
7.50107706e-01 -2.27582783e-01 -5.26678801e-01 6.45235717e-01
7.10557103e-01 -1.32318723e+00 9.90028203e-01 -4.34601754e-01
4.24983442e-01 -6.64487898e-01 9.05082151e-02 -6.40020967e-01
-3.45019639e-01 -6.76271141e-01 -3.34952354e-01 1.03132641e+00
1.42071277e-01 -1.79964945e-01 9.06253159e-01 1.77249059e-01
-1.56760275e-01 -8.04508567e-01 -1.22124863e+00 -1.06736028e+00
-8.34671497e-01 -7.09265709e-01 5.44171214e-01 6.12520158e-01
2.41085608e-03 -2.75310487e-01 -4.56462830e-01 -3.16149324e-01
3.45955640e-01 -1.83051988e-01 7.46762276e-01 -8.28585148e-01
2.01230850e-02 -5.59960067e-01 -7.04533219e-01 -1.28364277e+00
-1.21776834e-01 -2.70735383e-01 4.60603088e-01 -1.67990422e+00
3.22393030e-01 -3.13937843e-01 -4.45577025e-01 9.22928810e-01
-1.73202381e-01 5.48630357e-01 4.62231524e-02 2.88024366e-01
-1.54424584e+00 8.37540507e-01 1.42028749e+00 -2.70862043e-01
-3.61035109e-01 2.05607459e-01 -1.07313879e-01 8.33651006e-01
5.40728748e-01 -5.94394326e-01 -3.97863209e-01 1.90309614e-01
7.08453730e-03 3.23812407e-03 3.95350963e-01 -1.42367256e+00
1.23580687e-01 -6.36834860e-01 3.46669763e-01 -1.09548306e+00
4.19095516e-01 -5.76909423e-01 5.32733910e-02 5.50561666e-01
-2.89360374e-01 -1.44249454e-01 2.30155773e-02 8.43769550e-01
-2.47392699e-01 2.29396997e-03 5.35783947e-01 -4.03630018e-01
-1.29510725e+00 5.02243876e-01 -4.20109540e-01 9.60497931e-03
1.35959685e+00 -5.77084482e-01 -2.54922152e-01 -2.56109029e-01
-6.43856645e-01 2.65494764e-01 1.42412305e-01 5.82356811e-01
8.06791246e-01 -1.71193480e+00 -5.15548825e-01 -1.67953730e-01
3.33742589e-01 -6.24175109e-02 4.60186720e-01 1.32340968e+00
-2.52443850e-01 4.25016463e-01 -1.91171065e-01 -7.72673905e-01
-1.31956613e+00 5.24879932e-01 1.53931007e-01 -3.78616393e-01
-8.51878643e-01 8.66110146e-01 2.54386246e-01 6.56506717e-02
4.04720366e-01 -5.47817171e-01 -3.79910111e-01 1.33922338e-01
6.53976977e-01 6.03285551e-01 -3.72728050e-01 -7.20944762e-01
-4.85444188e-01 4.81915087e-01 1.21978402e-01 -2.62508392e-01
9.29593563e-01 1.65466562e-01 1.58085406e-01 -6.75711920e-03
5.33329785e-01 -5.23875415e-01 -1.68559849e+00 -3.87044311e-01
-5.09645231e-02 -7.96253979e-01 -3.30741405e-01 -6.37772202e-01
-9.92186844e-01 4.38462168e-01 4.35475618e-01 1.84523556e-02
1.19592786e+00 6.74645752e-02 8.87048721e-01 2.37914801e-01
7.18203247e-01 -1.41684532e+00 7.95105457e-01 2.47834399e-01
9.48171854e-01 -1.25729370e+00 3.23003799e-01 -3.76418471e-01
-8.24907959e-01 7.92614758e-01 1.05977035e+00 1.57322571e-01
2.98417747e-01 -4.13785636e-01 -3.21305424e-01 -1.86492890e-01
-8.31415653e-01 -3.86619508e-01 4.69576746e-01 3.90778095e-01
-5.41992784e-02 -4.90262061e-02 -7.19232321e-01 3.17122251e-01
4.91527766e-01 2.83205003e-01 2.17560813e-01 1.25831866e+00
-6.96425736e-01 -8.84549677e-01 -2.80880272e-01 4.92635489e-01
-2.64503211e-01 2.55896091e-01 -2.81880587e-01 8.15985322e-01
4.07734931e-01 8.16006958e-01 1.14310116e-01 -3.10726613e-01
2.72224605e-01 -2.11200312e-01 3.42382222e-01 -4.41957951e-01
-2.98814416e-01 2.94116676e-01 2.56727904e-01 -1.20584381e+00
-9.75928128e-01 -9.61330235e-01 -1.34525788e+00 5.61838150e-02
-3.23721975e-01 -5.92707507e-02 7.55316839e-02 1.12854993e+00
3.20704073e-01 5.01914620e-01 3.69610608e-01 -9.76957321e-01
-3.81539196e-01 -1.19534409e+00 -3.39519322e-01 4.49808687e-01
-1.16912328e-01 -8.95227015e-01 -1.70119941e-01 1.33577734e-01]
|
[8.315140724182129, 0.4194517731666565]
|
d509a3b0-5ba9-41e7-ad71-47fdfbe4b785
|
joint-classification-and-prediction-cnn
|
1805.06546
| null |
http://arxiv.org/abs/1805.06546v3
|
http://arxiv.org/pdf/1805.06546v3.pdf
|
Joint Classification and Prediction CNN Framework for Automatic Sleep Stage Classification
|
Correctly identifying sleep stages is important in diagnosing and treating
sleep disorders. This work proposes a joint classification-and-prediction
framework based on CNNs for automatic sleep staging, and, subsequently,
introduces a simple yet efficient CNN architecture to power the framework.
Given a single input epoch, the novel framework jointly determines its label
(classification) and its neighboring epochs' labels (prediction) in the
contextual output. While the proposed framework is orthogonal to the widely
adopted classification schemes, which take one or multiple epochs as contextual
inputs and produce a single classification decision on the target epoch, we
demonstrate its advantages in several ways. First, it leverages the dependency
among consecutive sleep epochs while surpassing the problems experienced with
the common classification schemes. Second, even with a single model, the
framework has the capacity to produce multiple decisions, which are essential
in obtaining a good performance as in ensemble-of-models methods, with very
little induced computational overhead. Probabilistic aggregation techniques are
then proposed to leverage the availability of multiple decisions. We conducted
experiments on two public datasets: Sleep-EDF Expanded with 20 subjects, and
Montreal Archive of Sleep Studies dataset with 200 subjects. The proposed
framework yields an overall classification accuracy of 82.3% and 83.6%,
respectively. We also show that the proposed framework not only is superior to
the baselines based on the common classification schemes but also outperforms
existing deep-learning approaches. To our knowledge, this is the first work
going beyond the standard single-output classification to consider multitask
neural networks for automatic sleep staging. This framework provides avenues
for further studies of different neural-network architectures for automatic
sleep staging.
|
['Oliver Y. Chén', 'Navin Cooray', 'Fernando Andreotti', 'Huy Phan', 'Maarten De Vos']
|
2018-05-16
| null | null | null | null |
['sleep-stage-detection', 'sleep-staging', 'automatic-sleep-stage-classification']
|
['medical', 'medical', 'medical']
|
[ 3.21436852e-01 -6.31507160e-03 -4.48287815e-01 -7.52763867e-01
-7.11613297e-01 -1.04300007e-01 3.84566903e-01 1.02804322e-02
-6.64055407e-01 8.16881597e-01 5.78906238e-02 -3.49634826e-01
-1.03604458e-01 -3.31476271e-01 -1.89987168e-01 -9.30513442e-01
-2.87207048e-02 3.54106814e-01 1.57436460e-01 2.29118422e-01
8.34389105e-02 2.72860490e-02 -1.75894535e+00 2.53497243e-01
9.46022868e-01 1.46818936e+00 2.03476876e-01 5.19072950e-01
1.64582804e-01 6.86535716e-01 -7.14823484e-01 -2.14036286e-01
-6.44811466e-02 -2.62882769e-01 -8.69160533e-01 -6.98454529e-02
2.71222323e-01 -5.56012467e-02 1.28823534e-01 6.36767805e-01
4.42552149e-01 2.77981699e-01 5.21126330e-01 -1.20588744e+00
-6.71928942e-01 2.50404686e-01 -3.46247822e-01 7.29883909e-01
-2.78664771e-02 -6.93708956e-02 1.16737354e+00 -4.70930189e-01
-1.51472792e-01 6.31134689e-01 8.50501835e-01 8.46027374e-01
-1.29771602e+00 -8.39565456e-01 1.72521725e-01 2.95471668e-01
-1.33513451e+00 -8.24783146e-01 3.79472136e-01 -2.69035071e-01
1.19632363e+00 2.27844477e-01 6.92753494e-01 1.12795436e+00
4.12629277e-01 7.38166094e-01 1.40078068e+00 -3.52205038e-01
3.35296720e-01 -2.05137134e-02 6.39353156e-01 7.59692788e-01
3.74511406e-02 -4.12801951e-02 -5.75176537e-01 -2.84593571e-02
1.10892177e-01 3.54710460e-01 1.01768665e-01 2.14079306e-01
-1.02266467e+00 6.32645190e-01 4.57279474e-01 4.31291640e-01
-2.26527095e-01 1.57317936e-01 3.33130121e-01 1.40525728e-01
8.91339362e-01 1.97445408e-01 -6.11362338e-01 -2.88195193e-01
-1.89090204e+00 -1.52550295e-01 7.06667364e-01 5.20694196e-01
7.06794024e-01 -1.43270373e-01 -4.63109165e-01 9.20977235e-01
3.03013861e-01 1.05781145e-01 6.99254572e-01 -9.37760413e-01
2.35293597e-01 7.85377204e-01 1.60077144e-03 -4.62459147e-01
-9.73633647e-01 -7.57334769e-01 -9.77575541e-01 8.70500654e-02
1.49657965e-01 1.28350896e-03 -1.02395272e+00 1.92254531e+00
-7.91865438e-02 4.63466644e-01 5.51555597e-04 5.35594821e-01
8.16438138e-01 2.50240058e-01 2.83479035e-01 -2.54044592e-01
1.74784219e+00 -1.21470678e+00 -7.85129666e-01 -4.99356896e-01
2.99819857e-01 -4.47898775e-01 8.72967064e-01 6.54089034e-01
-1.15981126e+00 -7.04221070e-01 -1.15372467e+00 -3.27182919e-01
-2.92844623e-01 5.89382470e-01 6.20491683e-01 7.74783373e-01
-1.70071936e+00 5.18427789e-01 -1.30058694e+00 -5.60190797e-01
7.37026930e-01 9.63873982e-01 -8.17039385e-02 4.38404769e-01
-9.06025410e-01 9.85073566e-01 6.91896230e-02 2.68973798e-01
-8.09237421e-01 -5.43897331e-01 -6.30885243e-01 2.00621784e-01
3.92278992e-02 -9.61683333e-01 1.47446573e+00 -9.90598142e-01
-1.32511342e+00 9.48223114e-01 -8.88857961e-01 -7.61642516e-01
-3.49700786e-02 -2.49595508e-01 -4.99941021e-01 -8.14620331e-02
2.93903142e-01 8.08932960e-01 7.54968345e-01 -6.81551099e-01
-8.68801117e-01 -4.45209503e-01 1.55594543e-01 1.60934143e-02
-5.33516824e-01 1.28110960e-01 -3.98761690e-01 -3.36255342e-01
-1.69978097e-01 -1.01997232e+00 -2.85817295e-01 -2.14396775e-01
-5.85060835e-01 -5.52171767e-01 3.39152664e-01 -3.09540540e-01
1.45741653e+00 -1.98778665e+00 5.17455786e-02 -3.04964751e-01
6.05143607e-01 1.20903991e-01 3.07973862e-01 1.13299288e-01
7.29130805e-02 7.66188353e-02 -3.00747931e-01 -1.31940114e+00
1.97503604e-02 4.62412864e-01 2.21710950e-02 5.16385972e-01
1.14501700e-01 8.45665991e-01 -5.70672691e-01 -4.14698660e-01
6.19588010e-02 4.26748127e-01 -4.63262796e-01 3.92240249e-02
2.09746525e-01 3.34372550e-01 -2.44926855e-01 6.94277346e-01
2.60932535e-01 -5.50967872e-01 -2.75154002e-02 -2.71692127e-02
-2.07043514e-01 7.37524748e-01 -6.06527209e-01 1.68353045e+00
-7.60950804e-01 5.36629975e-01 -2.07787171e-01 -1.04531336e+00
5.60000837e-01 3.45038772e-01 3.76512706e-01 -4.63876486e-01
1.44728944e-01 5.02691343e-02 -9.69750807e-03 -1.91957340e-01
4.53006089e-01 -3.20036262e-01 -2.97008842e-01 6.77141607e-01
3.80475938e-01 6.31988049e-01 2.57005483e-01 -8.47797543e-02
1.30106723e+00 -5.50686680e-02 4.86572951e-01 -4.24862236e-01
4.98366296e-01 -4.05070692e-01 8.99953306e-01 8.06895554e-01
-5.03889024e-01 4.27309364e-01 4.73872781e-01 -5.48101306e-01
-5.26807070e-01 -9.01623070e-01 -3.47748280e-01 1.25744259e+00
2.47447435e-02 -7.14454114e-01 -6.56506717e-01 -7.65863299e-01
-4.33220744e-01 5.21213710e-01 -1.09683633e+00 -1.99238241e-01
-1.71584263e-01 -1.16281605e+00 7.99234152e-01 7.86798418e-01
4.81692255e-01 -1.17567718e+00 -7.06430376e-01 2.62815841e-02
-2.73731172e-01 -1.06945002e+00 -4.07163531e-01 6.80941701e-01
-1.07479382e+00 -1.09961760e+00 -3.70156556e-01 -6.59426868e-01
4.48135018e-01 2.16394499e-01 1.26318133e+00 3.73360276e-01
-5.33846989e-02 6.92213885e-03 -1.77450582e-01 -3.74114394e-01
-5.47412038e-02 6.25647545e-01 3.30944031e-01 2.26990938e-01
7.47711062e-01 -1.04020965e+00 -1.09508801e+00 2.64595270e-01
-6.01752281e-01 1.96093306e-01 7.66610086e-01 7.86403954e-01
6.38402402e-01 3.64145637e-02 7.08436430e-01 -8.12451720e-01
4.85387146e-01 -8.45736980e-01 -1.09189279e-01 6.53087199e-02
-1.03876162e+00 -8.30123499e-02 7.41004646e-01 -1.75722569e-01
-8.61386538e-01 -9.53946784e-02 -2.55587041e-01 -1.85067803e-01
-5.15710771e-01 1.52762741e-01 2.40067139e-01 3.32612932e-01
4.32472646e-01 3.58772010e-01 -1.39028385e-01 -6.50013685e-01
1.08173462e-02 8.13325822e-01 3.29125375e-01 -2.03064144e-01
3.49637866e-01 5.05084634e-01 -5.60283326e-02 -4.89448190e-01
-1.30471659e+00 -7.75165260e-01 -8.42442393e-01 1.48259073e-01
1.17896271e+00 -1.16248751e+00 -6.96764350e-01 4.30689454e-01
-7.30373919e-01 -5.11106193e-01 -3.89862433e-03 1.37466341e-01
-4.36905473e-01 1.76914558e-01 -6.37601912e-01 -8.31044078e-01
-7.90605307e-01 -1.28683412e+00 1.32229829e+00 4.40762371e-01
-4.82792914e-01 -1.19737554e+00 1.01982445e-01 6.35087252e-01
3.77401263e-01 -1.93579838e-01 8.79080594e-01 -8.37402642e-01
-1.84878320e-01 1.06556401e-01 -4.77327667e-02 3.89478952e-01
2.78413951e-01 -2.97370046e-01 -1.37950695e+00 -2.08342344e-01
6.02043532e-02 -3.76881540e-01 1.19995213e+00 5.22501588e-01
1.31627297e+00 -1.14135370e-01 -4.22394037e-01 6.23829186e-01
1.21360302e+00 1.22372024e-01 4.14585769e-01 4.39539284e-01
3.12365144e-01 2.28478640e-01 1.50707111e-01 2.47791871e-01
9.08186376e-01 7.64518738e-01 3.45768720e-01 -3.54198396e-01
-1.21554151e-01 4.08479512e-01 5.42602956e-01 7.64496386e-01
-1.94614902e-01 -9.85873044e-02 -6.28913462e-01 6.02906227e-01
-1.83192348e+00 -9.00332272e-01 -6.31047785e-02 1.94341898e+00
7.88557291e-01 1.22841977e-01 3.77741754e-01 2.83917785e-01
2.37712070e-01 2.64031738e-01 -5.87618768e-01 -5.95049858e-01
2.52555221e-01 8.42712402e-01 1.73577636e-01 1.52934119e-01
-1.31119323e+00 6.79890692e-01 6.46277761e+00 6.33962452e-01
-7.98369348e-01 7.83173800e-01 6.87764466e-01 -6.88311577e-01
3.36282760e-01 -2.94650495e-01 -1.06991386e+00 8.05902600e-01
1.50872445e+00 3.71577412e-01 5.25405526e-01 7.34014750e-01
4.27557796e-01 -3.43044668e-01 -1.14237678e+00 7.48164833e-01
1.43005073e-01 -1.00636005e+00 -4.98227179e-01 2.26477042e-01
5.92804968e-01 3.83540273e-01 1.47735342e-01 5.63661635e-01
1.16300359e-01 -1.07923424e+00 6.35511458e-01 5.26374519e-01
7.04176247e-01 -6.79609537e-01 9.16427255e-01 3.72854918e-01
-1.16280460e+00 -3.91927660e-01 -2.47922987e-01 -2.80272603e-01
-8.84384289e-02 5.78290164e-01 -5.58669865e-01 4.17486608e-01
1.07695663e+00 9.15450871e-01 -1.05128217e+00 1.14771867e+00
-2.17559814e-01 1.01085639e+00 -1.61532223e-01 1.17031656e-01
2.00066522e-01 1.11714033e-02 -3.18569317e-02 1.18842387e+00
2.47591481e-01 -2.50591725e-01 1.63747817e-01 7.86728859e-01
-1.30609959e-01 -1.58605918e-01 -1.68760002e-01 3.50069195e-01
2.05789641e-01 1.62862563e+00 -9.74413991e-01 -3.23578119e-01
-6.12601399e-01 9.15027618e-01 6.19171143e-01 1.57207668e-01
-7.92376459e-01 -7.77513236e-02 7.88826108e-01 -1.03246838e-01
2.93200701e-01 6.92451745e-03 -6.07129753e-01 -1.09409606e+00
-4.38447408e-02 -5.13041735e-01 4.78984743e-01 -4.56055433e-01
-1.40692544e+00 8.18533719e-01 -5.64067550e-02 -9.39486802e-01
-9.59557965e-02 -3.52415085e-01 -1.02071011e+00 7.98108518e-01
-1.77909160e+00 -1.30981386e+00 -2.22969666e-01 5.73030114e-01
8.84814739e-01 -8.95296186e-02 9.61032569e-01 5.15274763e-01
-1.06488490e+00 7.36730874e-01 4.82942387e-02 -1.55573010e-01
6.04675114e-01 -1.65803528e+00 2.92771488e-01 7.27287889e-01
2.08994403e-01 9.14565325e-01 3.29764217e-01 -2.18012601e-01
-6.86918020e-01 -1.24502492e+00 1.14862883e+00 -7.13523507e-01
4.66385365e-01 -4.78555918e-01 -6.70299768e-01 7.36198604e-01
3.79049361e-01 -2.25393698e-01 1.38351524e+00 7.77356565e-01
2.49161184e-01 -3.87360424e-01 -7.81961977e-01 3.30350906e-01
9.14436519e-01 -4.61351395e-01 -6.94507539e-01 3.05549860e-01
3.65623057e-01 -1.83983430e-01 -7.29437113e-01 2.99542248e-01
6.71776354e-01 -1.26183760e+00 7.95531273e-01 -4.12619442e-01
3.73967022e-01 -2.46744335e-01 -4.09561880e-02 -1.05330420e+00
-4.63538140e-01 -4.96711582e-01 -4.99869704e-01 1.22904539e+00
4.99941200e-01 -5.82316816e-01 7.66396523e-01 6.05840564e-01
-5.10554910e-01 -1.24783003e+00 -1.19985306e+00 -4.71205682e-01
-3.98603752e-02 -6.14916623e-01 5.56210756e-01 3.91458362e-01
-3.70707214e-01 4.59088773e-01 -4.40519184e-01 2.10768402e-01
3.51541430e-01 2.51597881e-01 4.45266902e-01 -1.38184857e+00
-1.83188364e-01 -4.69837934e-01 -2.14247331e-01 -8.17431390e-01
4.18417901e-01 -1.10164320e+00 9.26297978e-02 -1.54224765e+00
5.23181915e-01 -5.33847451e-01 -1.18251443e+00 9.63812649e-01
-2.55929470e-01 8.76135647e-01 -9.50255468e-02 4.08310413e-01
-1.13023317e+00 4.73482728e-01 7.05840826e-01 1.81282952e-01
-2.80132025e-01 6.47655487e-01 -1.18910611e+00 8.97928357e-01
1.00868201e+00 -7.31491327e-01 -5.07218719e-01 -1.96099937e-01
9.09638777e-02 -3.46801311e-01 4.25985813e-01 -1.26243150e+00
3.51822615e-01 3.02738130e-01 6.22508943e-01 -5.47427952e-01
6.64511800e-01 -4.36714768e-01 -3.29566486e-02 2.71014392e-01
-3.18585217e-01 1.53628454e-01 2.66769361e-02 6.64254010e-01
8.80079269e-02 -7.06187487e-02 7.29263842e-01 -1.83752291e-02
-4.87031937e-01 1.67957470e-01 -7.18909442e-01 -1.39656022e-01
6.59405470e-01 -3.29650730e-01 -2.71871507e-01 -8.26302618e-02
-1.00681138e+00 2.77591974e-01 2.60477662e-01 3.95999968e-01
3.33931565e-01 -1.01936162e+00 -9.64281186e-02 3.53822768e-01
2.14557536e-02 -1.87170684e-01 2.24580109e-01 1.48692787e+00
5.55744879e-02 5.36224723e-01 -4.62370589e-02 -7.08704710e-01
-1.27983534e+00 2.67203540e-01 4.07378316e-01 -6.85331404e-01
-4.80869710e-01 7.54398942e-01 3.07363868e-01 -2.07978874e-01
2.31124967e-01 -7.16171086e-01 -4.40448105e-01 2.18250036e-01
5.18369734e-01 2.71726191e-01 4.77388740e-01 -6.47302151e-01
-4.17867154e-01 2.57145882e-01 -1.88328058e-01 2.35162228e-01
1.39225805e+00 -2.73592144e-01 -1.88391462e-01 7.31106162e-01
8.37228358e-01 -1.95066154e-01 -1.03514528e+00 -1.15027741e-01
4.96411920e-02 1.45700008e-01 3.53042632e-01 -1.13573313e+00
-1.14720333e+00 1.02490067e+00 7.68384695e-01 2.97996163e-01
1.50269341e+00 2.51741707e-02 1.01422846e+00 3.07395816e-01
1.28524646e-01 -7.71259189e-01 -1.77717328e-01 2.14385569e-01
2.71354914e-01 -1.26799536e+00 3.94362248e-02 1.97394997e-01
-5.40983260e-01 9.80257571e-01 6.86501920e-01 -5.24398647e-02
6.27620578e-01 4.80454974e-02 4.48588096e-02 -3.72751385e-01
-1.10789955e+00 -4.89506096e-01 6.06194556e-01 3.25399995e-01
3.70931387e-01 8.58203843e-02 -1.87263072e-01 1.22763872e+00
-3.29849601e-01 2.78757900e-01 2.27761671e-01 7.20615327e-01
-2.96552688e-01 -1.33322001e+00 1.78053611e-04 9.86443937e-01
-1.09698212e+00 -4.60728705e-01 -1.35779008e-01 6.34778440e-01
6.40558183e-01 1.39429510e+00 2.25678280e-01 -5.29226720e-01
-2.17732694e-02 3.77323985e-01 1.55284882e-01 -9.98664677e-01
-9.79408264e-01 -1.51251838e-01 6.57944381e-02 -6.28697932e-01
-8.66504371e-01 -5.88649392e-01 -9.54109550e-01 -7.75926486e-02
-3.28883708e-01 2.31398448e-01 3.20729882e-01 1.30325782e+00
6.03572786e-01 7.05817461e-01 3.95048827e-01 -1.00198722e+00
-1.47387370e-01 -1.04660165e+00 -5.71038306e-01 -4.05774713e-02
6.82324767e-01 -9.90129769e-01 -3.05959851e-01 -1.72775164e-02]
|
[13.513830184936523, 3.5224344730377197]
|
07bfa2b1-fc1a-4851-b04b-c76e72cfe63f
|
actor-and-action-modular-network-for-text
|
2011.00786
| null |
https://arxiv.org/abs/2011.00786v2
|
https://arxiv.org/pdf/2011.00786v2.pdf
|
Actor and Action Modular Network for Text-based Video Segmentation
|
Text-based video segmentation aims to segment an actor in video sequences by specifying the actor and its performing action with a textual query. Previous methods fail to explicitly align the video content with the textual query in a fine-grained manner according to the actor and its action, due to the problem of \emph{semantic asymmetry}. The \emph{semantic asymmetry} implies that two modalities contain different amounts of semantic information during the multi-modal fusion process. To alleviate this problem, we propose a novel actor and action modular network that individually localizes the actor and its action in two separate modules. Specifically, we first learn the actor-/action-related content from the video and textual query, and then match them in a symmetrical manner to localize the target tube. The target tube contains the desired actor and action which is then fed into a fully convolutional network to predict segmentation masks of the actor. Our method also establishes the association of objects cross multiple frames with the proposed temporal proposal aggregation mechanism. This enables our method to segment the video effectively and keep the temporal consistency of predictions. The whole model is allowed for joint learning of the actor-action matching and segmentation, as well as achieves the state-of-the-art performance for both single-frame segmentation and full video segmentation on A2D Sentences and J-HMDB Sentences datasets.
|
['Zhanyu Ma', 'Linjiang Huang', 'Liang Wang', 'Kai Niu', 'Yan Huang', 'Jianhua Yang']
|
2020-11-02
| null | null | null | null |
['action-understanding', 'referring-expression-segmentation']
|
['computer-vision', 'computer-vision']
|
[ 3.27370822e-01 4.32445928e-02 -3.29760134e-01 -4.75187510e-01
-8.55255485e-01 -5.45283675e-01 5.95857978e-01 -1.54223129e-01
-4.48934644e-01 2.19246984e-01 1.84106514e-01 8.71405825e-02
1.27429113e-01 -3.80347788e-01 -7.52062261e-01 -7.44623721e-01
1.99953124e-01 5.67013204e-01 8.86959076e-01 5.94937652e-02
7.02294633e-02 3.38538677e-01 -1.57060790e+00 8.66826713e-01
5.18696308e-01 1.44978368e+00 6.21303558e-01 6.27225220e-01
-4.42357540e-01 1.02281404e+00 -6.42239630e-01 -4.41165835e-01
2.79261887e-01 -6.41151488e-01 -1.19498324e+00 5.57570696e-01
4.72996384e-01 -4.66123492e-01 -2.90344298e-01 9.84115243e-01
3.76012623e-01 1.27439097e-01 4.53509748e-01 -1.22402215e+00
-7.14909434e-02 6.53369486e-01 -7.10730195e-01 2.52260029e-01
5.40218472e-01 1.64524212e-01 1.10807610e+00 -6.20700538e-01
8.79340947e-01 1.24163842e+00 2.22099558e-01 5.29428124e-01
-9.95847344e-01 -4.03693765e-01 5.70106328e-01 2.77239144e-01
-1.30170405e+00 -5.84412932e-01 8.73418570e-01 -4.88524556e-01
6.11013710e-01 2.15820879e-01 8.19509804e-01 1.00100255e+00
-2.60440074e-02 1.30124879e+00 5.81562936e-01 -9.76894274e-02
4.81498521e-03 -1.44206405e-01 -1.64440647e-01 6.67987108e-01
-6.86273277e-01 -2.89473474e-01 -5.32987893e-01 3.06436598e-01
7.23881841e-01 6.29872605e-02 -1.55769750e-01 -2.46421948e-01
-1.57595241e+00 3.24229479e-01 2.34048013e-02 5.48626721e-01
-5.51575482e-01 2.00951502e-01 5.57694018e-01 -5.00364602e-02
3.22562456e-01 2.17925925e-02 -6.55499101e-01 5.69134310e-04
-1.45587003e+00 1.95727095e-01 5.70962369e-01 9.98348296e-01
6.79061234e-01 -1.62848189e-01 -4.99385476e-01 5.72984278e-01
3.68584573e-01 2.13746175e-01 2.54930645e-01 -1.24737704e+00
6.06701076e-01 5.89224458e-01 2.22863227e-01 -9.32019472e-01
-3.34948927e-01 -2.06563532e-01 -6.19402289e-01 -1.62087679e-01
5.06005049e-01 5.39104603e-02 -1.07714427e+00 1.72663999e+00
6.41430795e-01 4.10456806e-01 -2.92356592e-02 1.04899549e+00
9.47408736e-01 7.37177134e-01 4.02185857e-01 -4.31495845e-01
1.70452273e+00 -1.18918788e+00 -8.34220946e-01 -3.81840467e-01
5.59878826e-01 -8.25132012e-01 6.29686236e-01 5.26037440e-02
-1.31702518e+00 -8.10691535e-01 -6.58699572e-01 -2.23567083e-01
3.34447026e-02 3.57326806e-01 2.43338346e-01 1.27320543e-01
-8.63453269e-01 2.22720653e-01 -8.08126390e-01 -2.57259995e-01
4.59546238e-01 3.66072923e-01 -3.66905689e-01 2.12230176e-01
-1.14259481e+00 5.34293115e-01 7.14661121e-01 2.53152072e-01
-9.88811016e-01 -5.36394954e-01 -9.33713615e-01 5.00802845e-02
6.53821409e-01 -7.02061117e-01 1.20430815e+00 -1.64079869e+00
-1.40007126e+00 1.01891983e+00 -3.68182659e-01 -5.42969644e-01
5.59097528e-01 -2.92592552e-02 -3.20746362e-01 6.17638171e-01
2.78340518e-01 1.15696907e+00 1.04606760e+00 -1.09063935e+00
-1.17448306e+00 -4.17146474e-01 3.16275448e-01 4.20589983e-01
6.21419623e-02 2.67646819e-01 -1.40005136e+00 -6.96502149e-01
3.30605298e-01 -6.61584437e-01 -6.57807738e-02 -4.70858254e-02
-3.95168304e-01 -3.70143801e-01 1.06388724e+00 -7.46859848e-01
1.38034189e+00 -2.17574501e+00 4.64112580e-01 -1.78857237e-01
1.86966881e-01 1.25588909e-01 -1.28160328e-01 7.10268915e-02
-9.36513990e-02 -1.63883045e-01 -2.08422914e-01 -6.23374701e-01
-2.39537045e-01 1.93580180e-01 -1.78606167e-01 5.64060450e-01
1.17276356e-01 1.01826847e+00 -6.69881999e-01 -1.15172935e+00
3.69740874e-01 2.77759522e-01 -4.20631349e-01 4.23506707e-01
-6.71108246e-01 7.20546007e-01 -7.37716496e-01 8.03504348e-01
4.57696080e-01 -2.86955208e-01 2.25632504e-01 -7.04130769e-01
-1.51521310e-01 1.32522330e-01 -1.34982932e+00 2.05650854e+00
-6.32341877e-02 4.39098299e-01 4.63222831e-01 -1.30425692e+00
5.40981054e-01 6.05730891e-01 1.16939640e+00 -7.56196260e-01
1.31532058e-01 1.71490848e-01 -3.04513842e-01 -8.71053636e-01
3.23729903e-01 3.88304368e-02 -1.68778449e-01 2.61337548e-01
1.61162674e-01 -2.63625495e-02 4.52504575e-01 1.48044348e-01
6.92126274e-01 5.98616242e-01 9.48531181e-02 4.90968637e-02
8.67255628e-01 -7.23648146e-02 6.86031520e-01 5.28140843e-01
-3.36024940e-01 6.57510161e-01 6.91968322e-01 -3.27781081e-01
-9.36833680e-01 -6.96702719e-01 1.41631186e-01 1.19708478e+00
6.56973779e-01 -2.19028816e-01 -8.82033765e-01 -9.71636057e-01
-4.04107273e-01 5.04418731e-01 -5.79057455e-01 2.67844766e-01
-7.33877659e-01 -1.59310579e-01 4.31281209e-01 4.47619975e-01
6.88351274e-01 -1.13782167e+00 -6.81132495e-01 2.18837857e-01
-7.65721619e-01 -1.56497204e+00 -8.33443105e-01 1.01778381e-01
-5.49441099e-01 -1.14590621e+00 -5.91183603e-01 -9.16012168e-01
5.00608742e-01 1.76243216e-01 8.94205689e-01 -5.99443354e-02
1.92271844e-01 3.35777789e-01 -3.56653392e-01 1.90995634e-01
-4.52952951e-01 -7.24829510e-02 -4.14892346e-01 5.44655025e-01
1.51177213e-01 -2.43195310e-01 -8.39056671e-01 7.07752466e-01
-1.06388342e+00 4.81661320e-01 3.95201802e-01 5.01047552e-01
7.66623795e-01 3.92987626e-03 2.35744432e-01 -5.35403132e-01
-4.38207351e-02 -2.85503238e-01 -6.31604016e-01 4.81376529e-01
2.07481533e-02 -2.44875737e-02 4.12416905e-01 -4.18522328e-01
-1.13341153e+00 6.17044091e-01 -2.75423527e-01 -7.27420509e-01
-4.15832341e-01 1.09981947e-01 -3.66569251e-01 3.74944717e-01
5.76707833e-02 2.58839756e-01 -2.64595687e-01 -2.99849033e-01
4.30634677e-01 5.13862789e-01 6.70409143e-01 -5.52885056e-01
3.66873145e-01 6.85516119e-01 -1.49072155e-01 -5.31002283e-01
-9.31689143e-01 -7.07852066e-01 -8.90591502e-01 -4.91678596e-01
1.64908767e+00 -9.09406185e-01 -8.29558194e-01 4.88456547e-01
-1.60779250e+00 -1.65784195e-01 -2.62609601e-01 4.60511208e-01
-8.11087787e-01 5.28244257e-01 -4.95366693e-01 -5.47218621e-01
-1.88390285e-01 -1.60773110e+00 1.60339880e+00 8.75520781e-02
-5.18250577e-02 -6.74734950e-01 -3.80111456e-01 7.32283890e-01
-1.09354839e-01 5.27158529e-02 5.75242102e-01 -7.66087651e-01
-8.01165402e-01 -1.12601295e-01 -3.42582643e-01 2.09640056e-01
-1.15152016e-01 1.31798506e-01 -7.45404720e-01 1.21102549e-01
1.18466735e-01 -4.01127227e-02 8.15707505e-01 7.18240082e-01
1.17065907e+00 -1.84641495e-01 -3.50086689e-01 6.11701608e-01
1.11211574e+00 4.81607169e-01 5.55466354e-01 3.40309441e-02
7.12648749e-01 8.57996285e-01 7.58996308e-01 3.07145059e-01
3.01367611e-01 1.02748215e+00 5.70009172e-01 -1.36796579e-01
-1.60425246e-01 -3.83143947e-02 3.48733902e-01 3.17762434e-01
2.71757934e-02 -5.05166709e-01 -6.25285625e-01 5.07656813e-01
-2.17893267e+00 -1.36655140e+00 -2.69537605e-02 1.81922829e+00
8.00144851e-01 2.51944400e-02 2.92234182e-01 -2.04644203e-01
9.62165892e-01 5.51225007e-01 -4.11123455e-01 1.01996876e-01
-2.83908218e-01 -2.13794097e-01 3.04157138e-01 3.70763659e-01
-1.42305207e+00 1.12129676e+00 5.75966167e+00 1.22022903e+00
-1.08075225e+00 2.67494321e-01 8.19004774e-01 -1.04226373e-01
-1.28840193e-01 6.28345162e-02 -8.76747727e-01 4.80747998e-01
6.08744383e-01 2.78358459e-01 2.48151258e-01 5.20372987e-01
5.88415623e-01 -3.14077139e-01 -1.27922761e+00 8.93027723e-01
3.90618965e-02 -1.58566201e+00 3.01625226e-02 -2.43523210e-01
5.55234730e-01 -2.62564689e-01 -2.90724516e-01 -2.67052185e-02
-3.32449555e-01 -6.72931314e-01 1.42761147e+00 5.42840362e-01
6.78571701e-01 -3.68290156e-01 5.66936076e-01 3.36019486e-01
-1.67562473e+00 -1.27585098e-01 2.34078035e-01 3.08912605e-01
6.10891521e-01 1.41701654e-01 -5.05896688e-01 5.85399091e-01
6.07350230e-01 8.13765168e-01 -3.35211635e-01 6.68104827e-01
-1.59217104e-01 2.44189620e-01 -2.24264070e-01 3.70012701e-01
3.58419865e-01 -2.53001124e-01 5.29387891e-01 1.20243049e+00
1.73645675e-01 1.45271318e-02 5.09432077e-01 7.75849402e-01
-2.21735965e-02 6.18034378e-02 -1.29513592e-01 -1.36615977e-01
2.66405523e-01 1.14373934e+00 -1.01305449e+00 -6.28297985e-01
-6.16167188e-01 1.19553459e+00 -3.29839513e-02 5.19060612e-01
-1.30625927e+00 1.49933398e-01 4.02678549e-01 2.21341372e-01
7.45019615e-01 -9.08281803e-02 5.41301118e-03 -1.12855506e+00
1.24233611e-01 -8.49653840e-01 5.36037624e-01 -9.03648138e-01
-7.69152522e-01 6.01291180e-01 1.45027861e-01 -1.36491084e+00
-1.32259920e-01 -2.06568897e-01 -3.30691129e-01 5.28271735e-01
-1.12311447e+00 -1.42038286e+00 -1.58134595e-01 9.02364194e-01
1.05134809e+00 1.37006566e-01 2.84975469e-01 6.17282331e-01
-5.93299091e-01 1.72146618e-01 -4.46863204e-01 2.74769008e-01
6.45791888e-01 -8.46064031e-01 -3.95843871e-02 1.02980459e+00
2.68681616e-01 1.81027398e-01 6.38882101e-01 -7.53406227e-01
-1.08932972e+00 -1.16684914e+00 7.99659073e-01 -2.44571209e-01
5.29170632e-01 -1.63267791e-01 -6.62212491e-01 6.73268914e-01
2.63089865e-01 8.48726034e-02 4.91668694e-02 -5.30055523e-01
-1.34766534e-01 -2.59465277e-01 -8.73828351e-01 4.91572857e-01
9.98904645e-01 -6.31852210e-01 -6.14726126e-01 5.44062495e-01
1.09362006e+00 -6.13084376e-01 -6.91023946e-01 4.55467731e-01
4.58050191e-01 -1.15999150e+00 1.01903117e+00 -4.95978177e-01
6.02037907e-01 -5.57875812e-01 -2.35747293e-01 -3.76450330e-01
8.32452029e-02 -5.62980413e-01 -1.41574144e-01 1.23758447e+00
2.20594183e-01 7.46547207e-02 7.69235194e-01 4.86736506e-01
-2.98792094e-01 -7.82460392e-01 -1.15217900e+00 -3.14382821e-01
-5.35752475e-01 -8.02320838e-01 3.87457967e-01 6.97128654e-01
-2.59370983e-01 2.73937494e-01 -5.58584750e-01 1.94624275e-01
3.40235263e-01 3.80016685e-01 6.60501003e-01 -7.52779603e-01
-2.41688818e-01 -6.45813704e-01 -3.77493531e-01 -1.65412283e+00
4.79030639e-01 -6.22557998e-01 1.43958449e-01 -1.54357588e+00
1.72069639e-01 -3.27095278e-02 3.49852955e-03 4.23185408e-01
-1.02891689e-02 2.68274665e-01 2.49777794e-01 3.94271314e-01
-1.16327620e+00 3.51434112e-01 1.40523720e+00 -3.19225013e-01
-3.13526094e-01 1.82554886e-01 -1.44665897e-01 7.65816391e-01
2.60959238e-01 -2.69178838e-01 -3.49525899e-01 -4.79792267e-01
-1.28056198e-01 6.68706596e-01 5.05074739e-01 -8.13456416e-01
4.60883886e-01 -2.79930949e-01 2.83460945e-01 -8.67111087e-01
5.23288608e-01 -1.16990113e+00 3.10555458e-01 2.58976281e-01
-4.82199162e-01 -1.12986699e-01 -8.54869932e-03 6.29344285e-01
-4.83470261e-01 -1.35649428e-01 8.49949598e-01 -1.47560745e-01
-8.77652228e-01 5.36124408e-01 -2.89505601e-01 -4.36995551e-02
1.37734222e+00 -4.37390804e-01 -2.19700914e-02 -2.40387306e-01
-1.03436673e+00 4.63826001e-01 3.09299737e-01 4.28851366e-01
5.77001035e-01 -1.05988312e+00 -4.18598115e-01 1.36549339e-01
-6.90228045e-02 1.35497347e-01 6.22561336e-01 1.40657866e+00
-3.49728674e-01 3.79877836e-01 -7.31722498e-03 -1.04499686e+00
-1.37708986e+00 6.28518403e-01 6.69419765e-01 -2.98212796e-01
-5.50352871e-01 8.31473768e-01 5.33371747e-01 1.02114804e-01
4.54035372e-01 -2.78844953e-01 -2.74802923e-01 4.26531076e-01
2.95437604e-01 -1.62413791e-02 -2.34047011e-01 -1.16640604e+00
-3.92593950e-01 6.79736435e-01 1.06735058e-01 -1.90129772e-01
9.96046960e-01 -4.96158391e-01 -2.34953612e-01 3.28867316e-01
1.25510585e+00 -2.08629414e-01 -1.51917112e+00 -2.49261186e-01
-2.18049645e-01 -3.58069628e-01 -2.07914278e-01 -4.32066262e-01
-1.29943299e+00 6.82514191e-01 2.89129853e-01 1.51918992e-01
1.29955792e+00 4.38060611e-01 9.00473893e-01 -9.91979688e-02
4.93446402e-02 -1.15074980e+00 2.06392273e-01 2.71383613e-01
5.73901832e-01 -1.05969429e+00 -7.26022050e-02 -3.65809172e-01
-7.59008169e-01 1.13391125e+00 7.02243209e-01 3.44291449e-01
4.26338315e-01 1.87620834e-01 -1.27010584e-01 -2.68585563e-01
-5.86081445e-01 -3.44896108e-01 5.95462024e-01 2.12260321e-01
2.28922218e-01 -2.80253619e-01 -3.49786580e-01 5.09430707e-01
4.86250132e-01 -1.23878822e-01 -2.44820099e-02 5.97903311e-01
-4.66600329e-01 -9.15756762e-01 -2.60534316e-01 1.13007642e-01
-5.85220218e-01 8.13393444e-02 -1.76422566e-01 6.34142816e-01
5.36261201e-01 8.74416888e-01 3.19990993e-01 -2.92233050e-01
2.29649931e-01 1.63474128e-01 2.39994988e-01 -3.87574822e-01
-7.49843001e-01 7.92202353e-01 1.72116328e-02 -1.01878643e+00
-1.06963313e+00 -6.66850388e-01 -1.55805600e+00 1.51238307e-01
-1.41260669e-01 1.19821690e-01 4.22381341e-01 1.29588830e+00
3.02191108e-01 6.10183239e-01 4.39955652e-01 -9.21179116e-01
3.38610671e-02 -6.02963567e-01 -3.58563751e-01 6.77237988e-01
2.69411445e-01 -3.91015977e-01 -2.54890621e-01 5.17418623e-01]
|
[9.661460876464844, 0.5223012566566467]
|
5af6ec5b-06b4-4175-8c29-12dd7383475e
|
cyclic-generative-adversarial-networks-with
|
2211.08424
| null |
https://arxiv.org/abs/2211.08424v1
|
https://arxiv.org/pdf/2211.08424v1.pdf
|
Cyclic Generative Adversarial Networks With Congruent Image-Report Generation For Explainable Medical Image Analysis
|
We present a novel framework for explainable labeling and interpretation of medical images. Medical images require specialized professionals for interpretation, and are explained (typically) via elaborate textual reports. Different from prior methods that focus on medical report generation from images or vice-versa, we novelly generate congruent image--report pairs employing a cyclic-Generative Adversarial Network (cycleGAN); thereby, the generated report will adequately explain a medical image, while a report-generated image that effectively characterizes the text visually should (sufficiently) resemble the original. The aim of the work is to generate trustworthy and faithful explanations for the outputs of a model diagnosing chest x-ray images by pointing a human user to similar cases in support of a diagnostic decision. Apart from enabling transparent medical image labeling and interpretation, we achieve report and image-based labeling comparable to prior methods, including state-of-the-art performance in some cases as evidenced by experiments on the Indiana Chest X-ray dataset
|
['Dwarikanath Mahapatra']
|
2022-11-16
| null | null | null | null |
['medical-report-generation']
|
['medical']
|
[ 1.04357100e+00 1.29621124e+00 -6.94683641e-02 -6.00468636e-01
-1.24815047e+00 -6.65845752e-01 5.03766418e-01 -7.21877366e-02
3.08718592e-01 9.53662395e-01 2.39134222e-01 -7.16947198e-01
1.59661725e-01 -4.82333511e-01 -8.61541271e-01 -4.74254757e-01
1.54459774e-01 6.14727497e-01 -5.35806179e-01 4.51812029e-01
-1.06641032e-01 2.07986429e-01 -7.64902294e-01 7.36092687e-01
5.46177924e-01 5.63575923e-01 -5.56554683e-02 1.21336532e+00
1.27223119e-01 1.44829011e+00 -9.33818579e-01 -9.00692701e-01
-1.61560848e-02 -1.29704583e+00 -1.16850126e+00 1.06647933e+00
4.87553030e-01 -6.50430620e-01 -2.75722593e-01 1.00994611e+00
1.43187225e-01 -4.26370203e-01 8.73041987e-01 -1.17309570e+00
-1.44425845e+00 9.29578722e-01 -5.38884878e-01 -4.01672795e-02
6.18665278e-01 3.45414340e-01 6.37941837e-01 -5.03352523e-01
1.03956246e+00 1.05811882e+00 4.93450880e-01 1.06077862e+00
-1.41538799e+00 -5.22900105e-01 -8.98760557e-02 -3.13944876e-01
-1.18389273e+00 -3.37704003e-01 6.74307585e-01 -2.22603679e-01
3.48748922e-01 6.98174238e-01 5.86341143e-01 1.40761054e+00
6.26527727e-01 6.33141041e-01 1.21920407e+00 -3.30177933e-01
9.93287116e-02 3.86971772e-01 -3.03296566e-01 1.25439250e+00
3.50127190e-01 4.56422120e-02 -2.43686542e-01 -2.17564374e-01
8.99097860e-01 1.25768319e-01 -7.00123847e-01 1.13830641e-01
-1.52235496e+00 9.13764656e-01 5.14922380e-01 1.96159214e-01
-6.03525460e-01 3.83571923e-01 -6.41566068e-02 6.95542395e-02
4.47840154e-01 5.33104897e-01 2.79282957e-01 5.69206774e-01
-1.11810338e+00 6.60223886e-02 6.94159210e-01 1.00064278e+00
2.60061264e-01 2.66846359e-01 -4.57274377e-01 -7.20350668e-02
2.25866273e-01 7.51082063e-01 2.99359530e-01 -1.01854146e+00
4.14050579e-01 3.42042655e-01 -2.88109928e-02 -1.11026216e+00
-2.77663141e-01 -5.29373586e-01 -1.22118664e+00 3.68090302e-01
9.67223048e-02 -8.26409981e-02 -1.28450072e+00 1.40460050e+00
1.37824148e-01 1.17134720e-01 4.75737423e-01 9.98393476e-01
1.20855021e+00 4.00396615e-01 8.83514062e-02 -3.42870504e-01
1.58051217e+00 -9.62859333e-01 -9.95010972e-01 -1.62960291e-01
3.96831751e-01 -8.74098122e-01 9.88704443e-01 2.37865746e-01
-1.56706858e+00 -4.90151465e-01 -1.04795754e+00 -7.77226388e-02
2.44814560e-01 1.62366390e-01 5.67493141e-01 7.21100569e-01
-1.18739057e+00 1.53213754e-01 -6.04780495e-01 -1.66730974e-02
6.60921156e-01 1.58026680e-01 -6.26258314e-01 -3.85345072e-01
-6.83419228e-01 8.23911071e-01 4.13277596e-01 -1.87869206e-01
-1.07419789e+00 -8.23357224e-01 -1.06220210e+00 -1.01078093e-01
2.28162840e-01 -1.36978388e+00 1.38941956e+00 -1.16834307e+00
-1.14647138e+00 1.61344099e+00 -2.91335192e-02 -6.11667633e-01
8.57410431e-01 4.12091225e-01 -5.62046230e-01 8.87852311e-01
2.60772407e-01 1.16767085e+00 1.23944306e+00 -1.89853013e+00
-2.67530024e-01 -1.61179844e-02 1.52898163e-01 -2.02207178e-01
2.89212853e-01 -3.59781235e-01 -2.99697906e-01 -9.18837905e-01
1.07534952e-01 -1.09080088e+00 -3.91952008e-01 2.05020249e-01
-1.07743037e+00 5.02593219e-01 7.33714104e-01 -8.89278412e-01
7.13942707e-01 -1.95457542e+00 -2.43845344e-01 1.14232898e-01
8.24462116e-01 -1.92346424e-01 -4.44016121e-02 2.91946325e-02
-3.81768256e-01 4.90655929e-01 -7.31907487e-01 -7.66292632e-01
-1.50462225e-01 4.51043993e-01 -5.59390366e-01 6.14644527e-01
5.12090147e-01 1.43011439e+00 -8.90787601e-01 -9.34243858e-01
1.83711573e-01 6.46809399e-01 -3.78122598e-01 4.35079157e-01
-2.40128830e-01 9.95143831e-01 -3.17252576e-01 6.03468657e-01
5.38952589e-01 -9.61439192e-01 2.31924310e-01 -1.05087861e-01
7.02101648e-01 -1.72376081e-01 -4.20929968e-01 1.54807961e+00
-4.55550104e-01 6.38917089e-01 -1.66482031e-01 -4.41678017e-01
6.13158882e-01 8.47549736e-01 2.21446931e-01 -2.85789907e-01
2.19617225e-02 1.22263236e-03 -2.43613929e-01 -6.25362813e-01
4.02191937e-01 -4.70190853e-01 -7.80765116e-02 1.16775000e+00
-3.70633453e-01 -6.56088054e-01 -2.35693440e-01 5.47991216e-01
1.13314068e+00 -2.27999203e-02 5.15651822e-01 2.10902542e-01
4.12876993e-01 3.93579274e-01 -1.45740092e-01 9.16729450e-01
2.02273186e-02 1.46136689e+00 4.25626904e-01 -5.67701519e-01
-1.30461109e+00 -1.24344039e+00 1.27996951e-01 1.05031386e-01
5.24580330e-02 1.27152383e-01 -9.39444005e-01 -1.14940739e+00
-3.75161529e-01 1.12028742e+00 -1.04860854e+00 -1.32485284e-02
-3.94713700e-01 -3.30086678e-01 6.36318147e-01 5.50867021e-01
3.56644601e-01 -1.27505577e+00 -7.39165008e-01 1.12353906e-01
-3.95644158e-01 -1.37618089e+00 -8.25732768e-01 -1.29154801e-01
-5.82319736e-01 -1.24477768e+00 -1.03664505e+00 -6.79616749e-01
1.56820488e+00 7.65714375e-03 1.52091014e+00 6.61043048e-01
-4.91046488e-01 7.21493602e-01 -1.75104737e-01 -4.28915381e-01
-1.40064847e+00 -2.90850908e-01 -4.38920617e-01 -1.22018538e-01
-4.21641767e-01 -5.70305623e-02 -7.30508447e-01 -6.57329410e-02
-1.56184447e+00 6.92908049e-01 7.17580914e-01 9.60854530e-01
7.71815956e-01 -1.86514229e-01 3.25369835e-01 -1.55804431e+00
6.39116585e-01 -3.52799505e-01 -1.47310570e-02 4.61020887e-01
-6.18266761e-01 3.80765498e-02 6.50385261e-01 -3.87733221e-01
-1.05655730e+00 6.28636256e-02 6.15568273e-02 -5.64580321e-01
-3.88832569e-01 9.78118554e-02 3.82011652e-01 1.05186939e-01
6.02214158e-01 3.15221548e-01 -5.90637550e-02 3.48399729e-01
5.59933543e-01 3.55517954e-01 1.21696830e+00 -7.06273392e-02
9.33943450e-01 8.35350454e-01 -3.11469845e-03 -3.77208740e-02
-9.93956923e-01 2.22798586e-01 -4.17545021e-01 -1.87782660e-01
1.25783443e+00 -6.50810003e-01 -6.43352926e-01 -5.13240039e-01
-1.46277320e+00 1.81276590e-01 -5.41849077e-01 3.52063626e-01
-7.50641048e-01 2.58663654e-01 -7.37471104e-01 -7.23051786e-01
-6.52371049e-01 -1.33876383e+00 1.25595760e+00 -3.49735096e-02
-7.26042628e-01 -1.19815862e+00 -3.79496098e-01 6.07306719e-01
7.80462697e-02 9.11983550e-01 1.13808954e+00 -3.45143139e-01
-7.48126566e-01 -2.78711647e-01 -3.17146301e-01 1.36415109e-01
2.90707886e-01 -1.51367009e-01 -1.19204736e+00 -1.42045319e-01
3.05737436e-01 -2.95395881e-01 5.42476416e-01 4.80581194e-01
1.10356820e+00 -7.41408467e-01 -3.21071088e-01 6.02611601e-01
1.12438476e+00 8.35470017e-03 7.19383955e-01 -1.32895365e-01
6.84692562e-01 4.74545449e-01 1.84213758e-01 1.16805233e-01
3.45026016e-01 4.33377959e-02 5.76946974e-01 -8.85079384e-01
-5.38036287e-01 -5.70200920e-01 -4.23357747e-02 4.68626946e-01
1.55251861e-01 -6.44102275e-01 -5.54053843e-01 4.86739725e-01
-1.49984884e+00 -8.37880611e-01 -3.17112744e-01 1.54259932e+00
7.74767220e-01 -1.19829200e-01 -3.49138528e-01 1.14226028e-01
6.77655876e-01 4.93504740e-02 -5.74621618e-01 -3.55496615e-01
-8.81791040e-02 2.06306830e-01 3.44000250e-01 3.90795022e-01
-7.48963118e-01 4.09033030e-01 7.39513588e+00 2.78466880e-01
-1.02843976e+00 2.52453595e-01 1.36354828e+00 1.69968843e-01
-8.02391112e-01 -3.91326189e-01 -1.66700291e-03 1.55123904e-01
6.91529572e-01 -6.12351894e-02 -6.77985772e-02 7.57236481e-01
1.08450139e-02 5.66047691e-02 -1.37251496e+00 1.07847381e+00
5.63650429e-01 -1.91312575e+00 5.07215798e-01 1.48793995e-01
1.18569469e+00 -8.61500800e-01 6.42881095e-01 -3.89263779e-01
4.08135772e-01 -1.45047998e+00 9.02787447e-01 5.26596010e-01
1.40317929e+00 -6.23957872e-01 9.02024329e-01 7.85047710e-02
-5.58147311e-01 4.27749872e-01 7.04169571e-02 4.91254896e-01
3.32415432e-01 4.64080930e-01 -1.53985536e+00 6.93429887e-01
1.68228522e-01 2.21968189e-01 -3.20725113e-01 2.99361527e-01
-6.27890110e-01 6.50053203e-01 3.45798880e-01 4.14923072e-01
3.81264269e-01 2.80959964e-01 3.66934806e-01 1.04350626e+00
2.66941845e-01 4.81160372e-01 1.56254753e-01 1.45023513e+00
-4.11925375e-01 -2.06896350e-01 -8.08764458e-01 -1.33886382e-01
-2.13733211e-01 1.17012191e+00 -1.03034484e+00 -7.52394676e-01
-8.97904038e-02 1.38509738e+00 -3.69278818e-01 3.50896060e-01
-1.00902867e+00 2.16842189e-01 -1.43219709e-01 3.04888368e-01
1.13359109e-01 3.10260057e-01 -7.64006853e-01 -8.58221412e-01
4.01009358e-02 -9.99451280e-01 3.63276571e-01 -1.40804303e+00
-1.13923204e+00 1.26706100e+00 -2.41561711e-01 -1.40192270e+00
-8.49667013e-01 -1.51339069e-01 -3.94564599e-01 8.31469059e-01
-1.40963590e+00 -1.74005687e+00 -4.16090429e-01 4.46574926e-01
5.24424016e-01 -1.12885721e-02 1.23107409e+00 -2.46666074e-01
1.40060827e-01 5.67590773e-01 -5.68478942e-01 6.37639090e-02
4.06131238e-01 -1.34846187e+00 5.90815008e-01 9.20433581e-01
3.90086204e-01 4.80770826e-01 8.66817892e-01 -7.46346354e-01
-1.13225842e+00 -1.32141733e+00 6.96389496e-01 -7.42603600e-01
2.27415681e-01 1.14738777e-01 -7.81423986e-01 1.04835141e+00
7.30631173e-01 1.19509324e-02 7.30070174e-01 -9.19418752e-01
-1.91106811e-01 5.42272210e-01 -1.71786785e+00 7.02343702e-01
7.17027605e-01 -5.83172321e-01 -5.73573887e-01 6.78910911e-01
9.33147132e-01 -7.97277272e-01 -7.34883785e-01 8.41675326e-02
3.80302906e-01 -8.28086853e-01 8.48659217e-01 -5.47264099e-01
1.10062706e+00 -4.43978101e-01 1.20755605e-01 -1.09465909e+00
-1.21703245e-01 -8.00127745e-01 1.87678665e-01 7.77395725e-01
5.20418823e-01 -4.12933648e-01 7.09545612e-01 6.71045899e-01
-2.11959258e-01 -5.09654820e-01 -4.89078611e-01 -1.64310828e-01
-3.85503501e-01 -4.66637075e-01 5.77820897e-01 1.02893388e+00
-2.86175013e-01 2.21613869e-02 -6.04850829e-01 4.49541509e-01
7.73758769e-01 2.64434159e-01 7.34942079e-01 -3.93961549e-01
-4.81147259e-01 -9.11298767e-03 -3.14968973e-01 -6.15423501e-01
1.92229912e-01 -1.00255537e+00 1.53919801e-01 -1.75254512e+00
4.97708470e-01 5.97729161e-02 3.19631666e-01 5.50802708e-01
-4.10530329e-01 1.00914609e+00 3.22067767e-01 3.29558164e-01
-4.35278445e-01 -2.28211030e-01 1.80232120e+00 -5.40468097e-01
3.18096489e-01 -3.50719430e-02 -1.08521664e+00 6.98662400e-01
3.33004385e-01 -8.30644429e-01 -4.77111399e-01 -3.92105520e-01
-3.48301069e-03 6.20570838e-01 7.29025364e-01 -6.71320021e-01
4.61776257e-02 -3.85251711e-03 8.18708003e-01 -6.60352588e-01
1.64862365e-01 -9.10577416e-01 7.15187311e-01 8.36340129e-01
-7.53340244e-01 1.71654895e-01 -1.95624568e-02 5.74911296e-01
-1.92044780e-01 -2.38782004e-01 6.08081400e-01 -6.42396092e-01
-1.18944921e-01 1.56319335e-01 -2.78063059e-01 -1.91779882e-01
1.13026929e+00 -2.54095733e-01 -4.84499484e-01 -8.28427196e-01
-9.06128943e-01 -2.16762215e-01 4.87540483e-01 8.03671125e-03
1.19603801e+00 -1.17361927e+00 -1.16680002e+00 1.66654840e-01
2.26953939e-01 1.10485427e-01 3.96247625e-01 6.22716129e-01
-8.91334116e-01 3.22611421e-01 1.62388697e-01 -7.87849605e-01
-1.32624531e+00 6.69184804e-01 2.37661600e-01 -5.62600791e-01
-1.15312314e+00 5.56894004e-01 7.01685786e-01 -6.91268519e-02
-1.61739647e-01 -5.93576491e-01 4.49591607e-01 -7.98140526e-01
9.27795410e-01 -3.69531840e-01 -1.58387363e-01 -5.03226519e-01
-9.79021043e-02 3.38556588e-01 -1.18136190e-01 -3.29099447e-01
1.06561530e+00 -2.76393384e-01 1.23028576e-01 1.94778830e-01
1.01544285e+00 -1.17116496e-02 -1.00442421e+00 1.15411311e-01
-4.73877251e-01 -3.00449669e-01 -1.84016407e-01 -1.14441991e+00
-1.20266581e+00 6.06789589e-01 4.41266268e-01 3.90104800e-01
1.17273760e+00 4.35939133e-01 8.70946467e-01 -5.44552971e-03
1.14316039e-01 -1.88353106e-01 3.27263176e-01 -4.79988903e-01
1.14272058e+00 -1.33735466e+00 2.64943510e-01 -4.55951452e-01
-1.20217836e+00 1.10185862e+00 -2.85883192e-02 1.56826284e-02
-6.36921730e-03 4.64723170e-01 6.30789995e-01 -4.64059621e-01
-5.90263188e-01 2.38490909e-01 6.00414693e-01 9.55155134e-01
3.67945701e-01 3.46069962e-01 2.13380188e-01 3.85011554e-01
-4.26729918e-01 2.22682133e-02 8.63641918e-01 7.67347038e-01
1.55197889e-01 -6.01193547e-01 -9.49715972e-01 2.73617595e-01
-9.06716347e-01 -2.61348963e-01 -4.74150062e-01 9.26981568e-01
-6.94494788e-03 1.13804984e+00 8.89289454e-02 -1.29938081e-01
1.42552052e-02 -2.01314658e-01 5.16417980e-01 -7.44343102e-01
-7.45079517e-01 -1.13255538e-01 2.11655032e-02 -2.59155393e-01
-6.51477873e-01 -9.11331549e-02 -1.55030155e+00 -3.04898202e-01
-6.50215745e-02 -7.25765154e-02 7.33012736e-01 9.75437403e-01
4.19465229e-02 1.12688994e+00 5.18743932e-01 -5.27276158e-01
-1.31166086e-01 -7.29771256e-01 -5.44682026e-01 8.69675279e-01
6.73197985e-01 1.61995009e-01 -2.23304182e-01 1.00662243e+00]
|
[14.982868194580078, -1.4365657567977905]
|
4c02fed0-18a9-4032-83e8-186616e15390
|
crunchgpt-a-chatgpt-assisted-framework-for
|
2306.15551
| null |
https://arxiv.org/abs/2306.15551v1
|
https://arxiv.org/pdf/2306.15551v1.pdf
|
CrunchGPT: A chatGPT assisted framework for scientific machine learning
|
Scientific Machine Learning (SciML) has advanced recently across many different areas in computational science and engineering. The objective is to integrate data and physics seamlessly without the need of employing elaborate and computationally taxing data assimilation schemes. However, preprocessing, problem formulation, code generation, postprocessing and analysis are still time consuming and may prevent SciML from wide applicability in industrial applications and in digital twin frameworks. Here, we integrate the various stages of SciML under the umbrella of ChatGPT, to formulate CrunchGPT, which plays the role of a conductor orchestrating the entire workflow of SciML based on simple prompts by the user. Specifically, we present two examples that demonstrate the potential use of CrunchGPT in optimizing airfoils in aerodynamics, and in obtaining flow fields in various geometries in interactive mode, with emphasis on the validation stage. To demonstrate the flow of the CrunchGPT, and create an infrastructure that can facilitate a broader vision, we built a webapp based guided user interface, that includes options for a comprehensive summary report. The overall objective is to extend CrunchGPT to handle diverse problems in computational mechanics, design, optimization and controls, and general scientific computing tasks involved in SciML, hence using it as a research assistant tool but also as an educational tool. While here the examples focus in fluid mechanics, future versions will target solid mechanics and materials science, geophysics, systems biology and bioinformatics.
|
['George Em Karniadakis', 'Khemraj Shukla', 'Adar Kahana', 'Leonard Gleyzer', 'Varun Kumar']
|
2023-06-27
| null | null | null | null |
['code-generation', 'geophysics']
|
['computer-code', 'miscellaneous']
|
[-2.72315353e-01 -4.43095148e-01 5.60424685e-01 1.65078461e-01
4.26571630e-02 -6.98840261e-01 5.64510763e-01 4.23465967e-01
5.00952452e-02 7.28252411e-01 -3.23631793e-01 -8.56315374e-01
-4.92244959e-01 -7.74322033e-01 -5.70072711e-01 -6.82857275e-01
-2.45425373e-01 5.31281233e-01 -2.58911431e-01 -2.16679975e-01
2.79948473e-01 9.52165484e-01 -1.84156704e+00 -3.03854253e-02
1.18247616e+00 7.75230944e-01 3.06588501e-01 8.52568150e-01
-2.68547773e-01 6.73233151e-01 -3.39102060e-01 3.11417490e-01
1.45274520e-01 -2.53092974e-01 -6.54328763e-01 -8.48292410e-02
1.23567306e-01 2.51733691e-01 4.05856490e-01 5.56451321e-01
5.73170364e-01 6.88264549e-01 5.21096885e-01 -9.69484210e-01
-1.05775788e-01 7.29395822e-02 -5.75159043e-02 -9.31739807e-02
4.10103142e-01 6.56777918e-01 4.12040055e-01 -7.97341943e-01
5.24560690e-01 1.03976071e+00 6.58882737e-01 2.21866578e-01
-1.07166290e+00 -4.31656629e-01 -1.69780284e-01 5.15854266e-03
-1.01433229e+00 -1.60756066e-01 6.49437606e-01 -1.00985754e+00
8.95770729e-01 8.10719311e-01 1.00169671e+00 5.00146747e-01
3.98503482e-01 2.55558312e-01 1.09014213e+00 -4.28607613e-01
6.65354252e-01 3.38894725e-01 -1.68667044e-02 4.53217894e-01
2.51996160e-01 3.80770594e-01 -4.46608096e-01 -1.70834243e-01
7.26657987e-01 -3.52632672e-01 -1.79890931e-01 -4.63236809e-01
-9.57852960e-01 8.56862366e-01 -1.76309645e-02 2.63629466e-01
-3.38765204e-01 -1.58658713e-01 3.10951799e-01 3.56167138e-01
6.39710963e-01 9.03789997e-01 -9.05339539e-01 -3.34670901e-01
-9.66283143e-01 7.50370145e-01 1.25258720e+00 5.53349078e-01
6.30209208e-01 4.55535591e-01 2.20018998e-01 6.21614933e-01
4.07957435e-01 3.74557495e-01 1.81718975e-01 -1.21172178e+00
-1.20959431e-01 5.94261885e-01 2.38572001e-01 -7.58234143e-01
-4.25812602e-01 -5.35891593e-01 -7.33741164e-01 8.50835681e-01
1.03801854e-01 -6.41162097e-01 -5.29161334e-01 9.54079688e-01
1.07237816e+00 2.15058148e-01 -1.00377593e-02 1.17970586e+00
1.16490865e+00 9.09380257e-01 2.91121155e-01 -2.59971708e-01
1.03721690e+00 -8.11309695e-01 -4.37262237e-01 2.36681495e-02
1.02351534e+00 -9.72246647e-01 9.58339691e-01 5.63057840e-01
-1.19747508e+00 -7.56793380e-01 -8.35900128e-01 6.87439926e-03
-8.96748006e-01 -1.03161614e-02 9.31847394e-01 4.76272374e-01
-8.62823486e-01 1.08071315e+00 -9.43014324e-01 -2.42624268e-01
-5.39816059e-02 1.44359037e-01 -1.11450233e-01 3.47741425e-01
-8.80441904e-01 1.12642264e+00 2.37456501e-01 6.08263128e-02
-4.55807030e-01 -1.71579409e+00 -1.04253399e+00 -8.11935589e-02
-1.81436422e-03 -1.16669655e+00 1.04174435e+00 -4.41701263e-01
-1.79924130e+00 4.79118496e-01 6.67600930e-02 -1.05379909e-01
7.64879763e-01 -3.68555814e-01 -2.45651811e-01 -3.77149671e-01
-1.84167013e-01 2.59902120e-01 2.95520037e-01 -1.18454504e+00
-4.55610633e-01 -3.40028736e-03 -1.51182413e-01 2.91892380e-01
2.12389737e-01 1.70335770e-02 -1.14334106e-01 -5.74547291e-01
-4.50147569e-01 -6.80464625e-01 -4.80686635e-01 -2.23223306e-02
-2.42157295e-01 3.40851247e-02 1.29253531e+00 -6.59800529e-01
1.10463905e+00 -1.79834938e+00 4.34290349e-01 1.32561967e-01
-1.70943066e-01 5.18937945e-01 5.38154125e-01 8.02197814e-01
-3.44849795e-01 -7.45249540e-02 -5.79140306e-01 -1.43548876e-01
-1.14725577e-02 2.13616453e-02 -1.15369469e-01 3.32076132e-01
-7.30332136e-02 5.72925508e-01 -7.52072990e-01 -1.76639214e-01
9.75445926e-01 6.18991017e-01 -4.72805768e-01 2.85601884e-01
-5.81671655e-01 1.15742791e+00 -3.58118802e-01 5.32941341e-01
7.67713249e-01 -1.97775960e-01 9.06189345e-03 4.85745817e-02
-9.09121275e-01 8.36159810e-02 -1.69383538e+00 1.73123050e+00
-9.17868257e-01 3.87743115e-01 8.61694932e-01 -1.14112103e+00
7.72806585e-01 3.01111013e-01 6.88985944e-01 -2.01089069e-01
1.59386039e-01 1.39290512e-01 -4.76133190e-02 -8.07138503e-01
3.64287555e-01 -6.10959567e-02 3.74888927e-01 6.15321040e-01
-1.18788697e-01 -8.75133097e-01 3.65768671e-01 -1.40658095e-01
4.64899749e-01 7.55974889e-01 1.65916100e-01 -6.64265752e-01
9.16379333e-01 4.32090193e-01 -6.51235459e-04 3.53841871e-01
4.13914949e-01 2.86306530e-01 4.35063057e-02 -5.48370361e-01
-9.04064238e-01 -6.65052176e-01 -5.21005392e-01 1.07222903e+00
-3.35764199e-01 -5.20363331e-01 -6.14320517e-01 -7.88326636e-02
2.86188722e-01 8.07639837e-01 -3.14345926e-01 3.83829653e-01
-5.39936841e-01 -1.02563477e+00 -2.22356990e-01 -4.61279079e-02
2.12031409e-01 -1.21275699e+00 -8.66572917e-01 2.46427372e-01
5.43949485e-01 -9.20397282e-01 2.24839449e-01 1.56710949e-02
-1.02330816e+00 -1.11870480e+00 -3.39915693e-01 -4.90925580e-01
3.62678885e-01 -1.97867885e-01 1.25344121e+00 2.55636930e-01
-7.16046929e-01 5.22286057e-01 -3.09196323e-01 -7.11533606e-01
-6.19182169e-01 1.72818359e-02 -5.92230260e-02 -5.19565701e-01
-3.55801374e-01 -8.46746743e-01 -4.14275229e-01 -3.43503468e-02
-9.00943160e-01 5.39130151e-01 -2.01732703e-02 3.37325960e-01
3.69747788e-01 5.49146421e-02 4.45744216e-01 -9.31949139e-01
6.17352724e-01 -5.88041961e-01 -1.02820587e+00 4.44170693e-03
-6.63524628e-01 -3.64941537e-01 1.00623858e+00 -3.76352146e-02
-9.32365775e-01 -3.22207715e-03 -5.23947060e-01 -3.90903562e-01
-6.18463397e-01 9.61038589e-01 -1.11811928e-01 -4.03959870e-01
4.60014492e-01 8.58188644e-02 3.66980463e-01 -8.73394907e-01
2.41536230e-01 4.00303990e-01 2.58720294e-02 -8.97761941e-01
7.87106633e-01 5.57128377e-02 4.00850803e-01 -1.25143099e+00
-3.43300998e-01 -2.26034313e-01 -6.14706516e-01 -3.55911344e-01
5.51779449e-01 -5.08456886e-01 -9.39394295e-01 3.24168980e-01
-9.19889092e-01 -4.90790844e-01 -5.38441122e-01 6.57397151e-01
-3.16967905e-01 2.31515303e-01 -2.73152143e-01 -7.36773729e-01
-5.62422037e-01 -1.31767976e+00 7.68538833e-01 4.29988354e-01
-3.48094106e-01 -1.66056693e+00 1.40763491e-01 3.05787057e-01
8.45228314e-01 6.97097421e-01 1.04913771e+00 -3.85112882e-01
-5.06451130e-01 -5.37973344e-02 3.36606085e-01 1.40111476e-01
6.49264455e-02 5.47101021e-01 -9.20918047e-01 -1.34120390e-01
-1.59046218e-01 -7.80205205e-02 1.92475423e-01 3.53003561e-01
1.51079786e+00 -2.07720593e-01 -2.42926180e-01 8.44946325e-01
1.43265200e+00 1.52019367e-01 1.16743155e-01 1.07213594e-01
7.76540220e-01 8.97822261e-01 5.33218563e-01 5.40456593e-01
-1.12077698e-01 4.62782502e-01 2.62664676e-01 -4.06077027e-01
8.68722722e-02 3.56797606e-01 5.61959594e-02 9.96818185e-01
-3.65845203e-01 1.68639004e-01 -1.01948833e+00 7.61844069e-02
-1.58709955e+00 -6.43043339e-01 -6.76842809e-01 2.10784507e+00
8.04781437e-01 -2.62567490e-01 -2.22393289e-01 5.42220473e-02
1.66289508e-01 -2.07469109e-02 -4.40132707e-01 -1.00599849e+00
3.48626107e-01 7.55746365e-01 1.11627251e-01 9.29682314e-01
-1.06345081e+00 7.16846585e-01 6.32892799e+00 7.41316617e-01
-1.53614795e+00 -8.74499232e-02 4.05046672e-01 -8.25044662e-02
-1.05938375e-01 6.54929653e-02 -5.49617589e-01 4.04486299e-01
1.12505436e+00 -3.85883451e-01 6.17392838e-01 7.99971104e-01
1.00314581e+00 -2.53967643e-01 -9.46322799e-01 3.58302623e-01
-5.11791229e-01 -1.79418981e+00 -1.30219115e-02 -1.61389709e-01
6.45041168e-01 2.23660208e-02 -3.34440798e-01 1.31111041e-01
2.34696835e-01 -1.18785214e+00 3.96538526e-01 6.50588274e-01
6.20916069e-01 -5.71477234e-01 3.80186945e-01 5.21864235e-01
-9.63123143e-01 3.86654556e-01 8.17549601e-02 -5.82757950e-01
2.74136096e-01 7.72193491e-01 -7.07036257e-01 1.02957916e+00
6.44005954e-01 6.39166057e-01 -2.09177375e-01 1.30517876e+00
3.50981414e-01 6.63444817e-01 -3.47574919e-01 -1.99922230e-02
-7.36465231e-02 -7.96675622e-01 7.12789595e-01 1.52622139e+00
3.89367610e-01 1.40498862e-01 2.32662067e-01 1.07762754e+00
6.51059866e-01 3.96560758e-01 -4.03911263e-01 1.11699514e-02
1.12420306e-01 1.54224193e+00 -6.01738334e-01 -2.82104909e-01
-1.77815586e-01 1.91235244e-01 -3.22487742e-01 5.11094272e-01
-8.00807059e-01 -3.27488810e-01 9.87863958e-01 2.83253938e-01
8.14188868e-02 -5.18962741e-01 -7.46952832e-01 -8.08349490e-01
-4.82788801e-01 -8.27080011e-01 1.02729537e-01 -9.45330977e-01
-9.57890511e-01 1.43434867e-01 2.40832999e-01 -9.12711859e-01
-3.29300016e-01 -8.62343729e-01 -9.26025808e-01 1.27160680e+00
-1.25802267e+00 -1.07757437e+00 -4.61997509e-01 1.78004920e-01
4.54176515e-01 -2.17030011e-02 1.07796168e+00 3.61100078e-01
-7.30329752e-01 -2.61344343e-01 4.01890129e-01 -5.60545743e-01
2.37627923e-01 -1.34230399e+00 1.92864239e-01 5.85857332e-01
-3.58301342e-01 6.98328555e-01 1.20807171e+00 -8.02566469e-01
-1.78581274e+00 -1.11190879e+00 5.31558812e-01 -2.98907042e-01
6.69353545e-01 -1.27728254e-01 -1.13888168e+00 1.09971389e-01
3.15361798e-01 -4.16464321e-02 6.48453772e-01 5.67533970e-02
5.76225460e-01 -3.67545933e-02 -9.41011310e-01 3.08200002e-01
5.48570275e-01 -1.92868542e-02 -2.35458419e-01 6.24584496e-01
3.95048529e-01 -9.02932942e-01 -1.22298193e+00 6.67305589e-01
3.43029529e-01 -8.13985229e-01 1.07971585e+00 -6.12509310e-01
5.76715469e-01 -5.34187913e-01 5.07897019e-01 -1.32653105e+00
9.02614649e-03 -8.78372490e-01 -2.45934233e-01 1.07989466e+00
4.47133183e-01 -6.39044523e-01 5.84005535e-01 6.64418519e-01
-4.89474654e-01 -8.39014947e-01 -7.80692875e-01 -2.83908069e-01
4.70048130e-01 -8.35984468e-01 4.50819880e-01 1.14909959e+00
-5.77117577e-02 -3.16987410e-02 -2.04286221e-02 2.64477909e-01
3.53947133e-01 4.71646249e-01 9.40339923e-01 -1.50750256e+00
-3.45162392e-01 -4.44103479e-01 1.62618443e-01 -4.93653029e-01
-4.13170755e-02 -7.72436023e-01 -3.37839872e-01 -1.65705180e+00
-5.69814265e-01 -6.89407647e-01 3.41881007e-01 1.12201571e-01
-7.55593088e-03 -5.71371987e-02 3.12895142e-02 1.60154492e-01
4.93959524e-02 2.80679494e-01 1.60849714e+00 3.07647586e-01
-5.03152430e-01 9.25613344e-02 -3.84016603e-01 6.98195696e-01
8.02837670e-01 -4.82344739e-02 -2.77257115e-01 -2.19082579e-01
1.67555273e-01 -1.29192397e-01 5.29236257e-01 -1.32013643e+00
2.81136751e-01 -6.04514241e-01 4.78132069e-01 -3.41418713e-01
2.05336779e-01 -9.58746672e-01 6.39763117e-01 4.65973735e-01
1.46642879e-01 4.23455276e-02 7.97689617e-01 -1.09484211e-01
2.97514834e-02 -1.90789878e-01 7.31110275e-01 -4.84061927e-01
-3.74011129e-01 1.41728610e-01 -5.88380992e-01 -1.46583870e-01
1.23497570e+00 -1.87685490e-01 -6.51973784e-02 -8.67911428e-03
-9.43477631e-01 3.53505403e-01 6.00168645e-01 1.47806793e-01
-1.03853242e-02 -5.29524922e-01 -5.57028651e-01 6.72549009e-01
-5.54422915e-01 4.04636770e-01 4.92978722e-01 7.66521871e-01
-1.20413113e+00 5.53775728e-01 -1.15435205e-01 -4.91991520e-01
-1.14102399e+00 3.10317069e-01 7.11578786e-01 -8.17261115e-02
-5.59172750e-01 6.68910980e-01 -7.02452194e-03 -7.37788379e-01
-7.38692284e-02 -3.73369128e-01 -4.09432709e-01 -9.37127098e-02
2.72549987e-01 5.98889410e-01 6.09384060e-01 -3.09538633e-01
-5.36369048e-02 7.29413092e-01 5.58237493e-01 9.99253839e-02
1.44395781e+00 4.91451845e-02 -3.10712904e-01 5.33585727e-01
5.23292601e-01 1.90651223e-01 -1.18850887e+00 5.31297207e-01
-3.03222358e-01 -1.92883715e-01 2.61017323e-01 -1.10883856e+00
-9.88609552e-01 8.05934906e-01 3.21905106e-01 4.22653824e-01
9.13825810e-01 -3.96380723e-01 5.52162945e-01 -2.80629694e-02
1.17045199e-03 -9.98494148e-01 -7.67536819e-01 7.30630696e-01
1.02315497e+00 -8.88131678e-01 4.02356714e-01 -5.26897073e-01
-2.26625592e-01 1.33282423e+00 7.43375480e-01 2.23574638e-01
1.10830081e+00 6.62043631e-01 1.96059823e-01 -2.88338631e-01
-6.51104867e-01 -1.08944587e-02 3.12274903e-01 4.26218182e-01
7.71552384e-01 -2.01426268e-01 -5.35442531e-01 3.04222375e-01
-4.42525536e-01 3.43586922e-01 2.18244672e-01 1.29692590e+00
-3.22605640e-01 -1.31759608e+00 -7.55167842e-01 3.80824149e-01
-4.69121456e-01 -3.97437662e-02 -4.11778204e-02 7.96810150e-01
5.91013491e-01 6.96766019e-01 -1.66272409e-02 -3.87049578e-02
2.26654723e-01 9.02370661e-02 1.11808181e-01 -5.72532773e-01
-1.08098769e+00 -1.03751875e-01 2.65799552e-01 -1.82875007e-01
-3.08134884e-01 -5.31390429e-01 -1.20925510e+00 -7.38215029e-01
-2.82899123e-02 7.87988663e-01 1.07982504e+00 1.04128110e+00
5.75953364e-01 7.96274841e-01 1.43686444e-01 -1.31102836e+00
7.60342181e-02 -1.04079270e+00 -4.25556362e-01 -7.02766329e-02
1.60130799e-01 -6.44345224e-01 -4.87453133e-01 2.55681306e-01]
|
[6.385396480560303, 3.275775909423828]
|
d4829767-40d6-48f3-8767-d9159afc1acb
|
lexicographic-multi-objective-reinforcement
|
2212.13769
| null |
https://arxiv.org/abs/2212.13769v1
|
https://arxiv.org/pdf/2212.13769v1.pdf
|
Lexicographic Multi-Objective Reinforcement Learning
|
In this work we introduce reinforcement learning techniques for solving lexicographic multi-objective problems. These are problems that involve multiple reward signals, and where the goal is to learn a policy that maximises the first reward signal, and subject to this constraint also maximises the second reward signal, and so on. We present a family of both action-value and policy gradient algorithms that can be used to solve such problems, and prove that they converge to policies that are lexicographically optimal. We evaluate the scalability and performance of these algorithms empirically, demonstrating their practical applicability. As a more specific application, we show how our algorithms can be used to impose safety constraints on the behaviour of an agent, and compare their performance in this context with that of other constrained reinforcement learning algorithms.
|
['Alessandro Abate', 'Charlie Griffin', 'Lewis Hammond', 'Joar Skalse']
|
2022-12-28
| null | null | null | null |
['multi-objective-reinforcement-learning']
|
['methodology']
|
[ 3.53180975e-01 1.80532277e-01 -5.86713791e-01 -7.52715336e-04
-4.82802540e-01 -6.57747447e-01 5.21810949e-01 1.06522053e-01
-8.91304135e-01 1.33685684e+00 -6.05453961e-02 -4.22459692e-01
-5.90119123e-01 -5.97006381e-01 -5.11467755e-01 -8.07922542e-01
-4.33749765e-01 5.66933692e-01 3.21337402e-01 -4.98368800e-01
6.37266397e-01 4.19040889e-01 -1.37540746e+00 -1.96653455e-01
5.17254829e-01 8.63438666e-01 2.22594619e-01 8.52780044e-01
4.11137283e-01 8.92835319e-01 -5.31248510e-01 1.65076002e-01
3.52395952e-01 -5.54647207e-01 -1.16796255e+00 2.50934362e-01
-2.22525239e-01 -3.89509469e-01 2.52371907e-01 9.11164105e-01
4.86335248e-01 4.15427983e-01 3.87195379e-01 -1.48009181e+00
-8.77750292e-02 3.07447642e-01 -5.21794498e-01 2.22628266e-01
6.33611023e-01 3.35321933e-01 1.12777174e+00 8.05133954e-02
5.27562499e-01 1.25220776e+00 1.04490258e-01 7.46446609e-01
-1.39934230e+00 -2.12799788e-01 4.88930434e-01 1.62829831e-01
-6.73933327e-01 -2.54872978e-01 3.58029157e-01 -2.17746794e-02
1.11246371e+00 2.27543965e-01 8.17743599e-01 8.17899704e-01
3.37019950e-01 7.79880285e-01 1.29583466e+00 -5.03145993e-01
6.32756889e-01 -9.04021859e-02 -4.10003066e-01 7.57006347e-01
7.06507766e-04 6.98238909e-01 -2.63697475e-01 -2.19785765e-01
5.35684943e-01 -3.74856174e-01 -1.49857923e-01 -6.54074609e-01
-9.74075139e-01 1.16260421e+00 1.30299125e-02 1.15777075e-01
-5.90659082e-01 5.41342676e-01 3.42011303e-01 6.79114819e-01
-6.09212592e-02 7.31312335e-01 -4.62127477e-01 -2.15457812e-01
-5.31013370e-01 6.79883301e-01 9.67680871e-01 5.22966385e-01
5.63302994e-01 1.96455181e-01 -5.90847619e-02 5.95465183e-01
3.73406976e-01 4.31258500e-01 3.55187744e-01 -1.37647653e+00
3.05804163e-01 3.38576809e-02 6.20492756e-01 -3.67096692e-01
-5.07093906e-01 2.33595315e-02 -1.20247990e-01 9.73370850e-01
3.12979847e-01 -5.58948100e-01 -6.09258354e-01 1.82746112e+00
4.02330428e-01 1.75576180e-01 3.23110700e-01 7.76900470e-01
-3.27839553e-01 6.87276006e-01 8.23144764e-02 -7.38672137e-01
8.00854743e-01 -7.08062291e-01 -5.52014410e-01 -4.01063502e-01
7.12269366e-01 -4.91326660e-01 9.10925150e-01 5.60453117e-01
-1.28350627e+00 1.02434874e-01 -1.00711596e+00 7.42997706e-01
9.44923237e-03 -3.83696914e-01 6.91451550e-01 6.20358527e-01
-1.15135515e+00 1.03191543e+00 -7.47188330e-01 -1.44336581e-01
-3.48602198e-02 6.55771852e-01 2.45923132e-01 2.98720449e-01
-1.19382346e+00 1.21064997e+00 6.43908322e-01 -1.05193868e-01
-1.27498412e+00 -2.10689843e-01 -7.28030324e-01 8.40769932e-02
1.02198184e+00 -3.21584046e-01 1.85711789e+00 -1.03782177e+00
-1.97298563e+00 4.59766835e-01 2.60477245e-01 -6.29267335e-01
5.24940610e-01 -3.92145105e-02 4.76571769e-02 7.31810108e-02
2.36004204e-01 4.25165206e-01 8.75540197e-01 -1.05232644e+00
-1.16721201e+00 -4.72626016e-02 5.31730533e-01 5.80885053e-01
-2.12283030e-01 2.92841583e-01 6.93530142e-02 -1.99219614e-01
-5.19061029e-01 -9.98241782e-01 -7.90145099e-01 -3.99257183e-01
-1.37852013e-01 -4.49897557e-01 4.89066631e-01 8.34580325e-03
1.09863889e+00 -1.73736334e+00 5.58030546e-01 4.93692189e-01
-3.72533947e-01 4.37805206e-02 -3.39971364e-01 5.99069476e-01
2.22792342e-01 8.95349458e-02 -3.33203167e-01 2.14657873e-01
1.41227871e-01 6.46872759e-01 -1.67585567e-01 4.88653988e-01
1.67947412e-01 6.10150695e-01 -1.12576067e+00 -3.31733197e-01
1.29052490e-01 -3.22344482e-01 -6.76171362e-01 2.55856305e-01
-5.04020274e-01 3.28704625e-01 -6.64292097e-01 2.74638861e-01
3.30573088e-03 1.13217987e-01 7.46143401e-01 6.37256205e-01
-3.34491581e-01 2.08665296e-01 -1.41608047e+00 9.38435018e-01
-4.70412105e-01 6.24834150e-02 3.89154732e-01 -1.34042394e+00
5.74607670e-01 2.46104568e-01 9.39279437e-01 -8.38669598e-01
1.33304924e-01 1.32605866e-01 3.52522614e-03 -4.48393494e-01
3.31144392e-01 -4.35399652e-01 -2.36183867e-01 7.20210195e-01
-2.43487224e-01 -5.35457253e-01 6.23211861e-01 1.66647397e-02
9.72249627e-01 2.90072978e-01 6.16918266e-01 -3.54738355e-01
7.39446759e-01 -3.65919061e-02 5.62739193e-01 9.42007959e-01
-1.96339563e-01 -4.17811751e-01 1.08490777e+00 -2.51302749e-01
-9.83613670e-01 -6.76002741e-01 2.22202897e-01 1.36114597e+00
2.10882679e-01 -1.81809455e-01 -5.79975247e-01 -8.26621294e-01
3.46330225e-01 7.09309876e-01 -5.42809367e-01 -1.63206205e-01
-7.65490174e-01 -6.85984612e-01 9.30572227e-02 3.69376510e-01
1.63336530e-01 -1.35561728e+00 -1.40263116e+00 5.10041296e-01
2.88503587e-01 -6.51484251e-01 -4.23739135e-01 4.22040761e-01
-7.66122520e-01 -1.45251489e+00 -4.90022838e-01 -7.55300224e-01
5.51850975e-01 -1.44239694e-01 8.95620942e-01 2.99689978e-01
1.90897025e-02 6.81989312e-01 -2.57615834e-01 -2.98031956e-01
-5.61972260e-01 -2.04979599e-01 3.11971694e-01 -2.90687919e-01
-2.99902290e-01 -2.86862999e-01 -3.99339318e-01 4.72699046e-01
-1.01808047e+00 -3.15515250e-01 3.24475527e-01 9.26053345e-01
5.71207643e-01 1.56907246e-01 7.77109087e-01 -6.52188957e-01
1.22831821e+00 -4.39404219e-01 -1.24378920e+00 4.31520581e-01
-8.34311545e-01 4.85344827e-01 8.49447489e-01 -5.11066914e-01
-6.74502790e-01 2.94672847e-01 -5.17167933e-02 3.04568186e-02
1.15345754e-01 3.81824195e-01 1.56643778e-01 -2.29989141e-01
4.96846437e-01 2.14197859e-01 2.34238312e-01 -1.10412464e-01
3.10276538e-01 4.16282564e-01 1.70862526e-01 -9.39570129e-01
4.56230670e-01 1.59141019e-01 3.85275751e-01 -5.60265362e-01
-6.23618364e-01 -2.57699460e-01 -1.10276893e-01 -3.92957747e-01
5.47522366e-01 -2.31496140e-01 -1.29605484e+00 8.66361707e-02
-6.14125907e-01 -8.49844575e-01 -2.82643318e-01 3.91353607e-01
-1.42064250e+00 3.60265762e-01 -2.37032637e-01 -1.16449821e+00
5.50103001e-02 -1.14928031e+00 5.95702708e-01 2.19847545e-01
-8.24110880e-02 -1.00745678e+00 3.09588492e-01 -1.65106907e-01
2.23620966e-01 2.48980686e-01 6.44442201e-01 -4.47420835e-01
-3.84274811e-01 3.22012722e-01 4.02237087e-01 1.67690262e-01
-6.07638471e-02 -2.25644246e-01 -9.26926434e-02 -7.60926068e-01
-4.66849878e-02 -8.32943797e-01 5.38520992e-01 4.66502190e-01
8.56478512e-01 -6.94264889e-01 -1.62531659e-01 1.93881437e-01
1.59702349e+00 6.59383118e-01 2.36767158e-01 8.26608956e-01
-4.30580089e-03 6.35557592e-01 1.01435566e+00 6.73710525e-01
6.83143586e-02 7.08065569e-01 8.72768939e-01 3.00859123e-01
5.92509508e-01 6.76404638e-03 4.83485222e-01 -1.82285905e-02
-1.52321935e-01 -1.38791293e-01 -7.04765201e-01 6.43080354e-01
-2.27876306e+00 -1.07087505e+00 4.25295085e-01 2.26648760e+00
9.11262214e-01 2.76730686e-01 5.72160184e-01 1.16033047e-01
6.18193030e-01 1.10189699e-01 -9.11263227e-01 -1.11709034e+00
4.33775663e-01 5.09919584e-01 8.24409723e-01 8.12824965e-01
-9.97311175e-01 1.03277886e+00 8.00426579e+00 5.44883251e-01
-7.75137544e-01 -2.49401346e-01 3.31134379e-01 -4.46017861e-01
-1.41440749e-01 8.11747462e-02 -5.56729734e-01 3.13364655e-01
9.74058688e-01 -4.58164275e-01 1.04878509e+00 8.74830663e-01
4.59430903e-01 -5.99751413e-01 -1.05444467e+00 3.62438589e-01
-4.76032168e-01 -9.15960729e-01 -4.10261750e-01 1.65219828e-01
7.80134261e-01 -4.34902251e-01 -3.94728519e-02 4.81678694e-01
9.71897066e-01 -1.04311085e+00 5.61125100e-01 9.08311531e-02
3.45086873e-01 -1.31613564e+00 3.56928438e-01 5.05646765e-01
-8.99950922e-01 -5.83503723e-01 -1.68419212e-01 -4.63344723e-01
3.59806448e-01 -2.33912971e-02 -6.55527711e-01 2.92559683e-01
3.48807365e-01 4.16178316e-01 2.73077130e-01 1.15414512e+00
-7.53947198e-01 1.96966916e-01 -4.63650554e-01 -8.51706147e-01
9.36211526e-01 -2.29651064e-01 4.49660271e-01 8.79126608e-01
-9.80585366e-02 6.10654429e-02 7.07913756e-01 5.23289263e-01
2.62517393e-01 9.52245668e-02 -5.55289209e-01 1.52113453e-01
1.72250718e-01 9.64864373e-01 -7.03022718e-01 -2.46816739e-01
-2.62334436e-01 8.03079903e-01 5.89489996e-01 3.67842287e-01
-8.20313215e-01 -3.98095101e-01 9.70097542e-01 -3.86348397e-01
4.11420703e-01 -2.12380841e-01 2.17104509e-01 -6.97141826e-01
-1.83720842e-01 -1.00785697e+00 6.01402462e-01 -2.56224453e-01
-8.42366219e-01 9.67413262e-02 2.04976529e-01 -8.13876987e-01
-8.51692498e-01 -4.98021364e-01 -6.04716718e-01 5.54977298e-01
-1.68157053e+00 -2.99865484e-01 5.94978094e-01 6.59475267e-01
4.81922507e-01 -6.31017685e-02 6.07304573e-01 -2.36562029e-01
-4.86030072e-01 1.09129496e-01 3.53185296e-01 -4.43592817e-01
1.27770215e-01 -1.50274742e+00 -1.00682817e-01 6.75657928e-01
-2.02334836e-01 6.93730488e-02 8.03767681e-01 -5.73661625e-01
-1.40748572e+00 -6.95747495e-01 3.54788780e-01 1.24258265e-01
8.83465469e-01 1.96136400e-01 -2.63639480e-01 6.75439596e-01
2.67519057e-01 -4.02248353e-01 2.19695300e-01 1.71850566e-02
4.34088558e-01 1.88494846e-01 -1.20692575e+00 6.09709978e-01
6.94603264e-01 2.92557366e-02 -5.54379761e-01 4.17570442e-01
2.66668528e-01 -4.91915315e-01 -6.79325044e-01 1.85084879e-01
4.17947203e-01 -8.14332068e-01 8.39480639e-01 -1.20136929e+00
2.71922439e-01 -1.33557290e-01 1.15702011e-01 -1.81293356e+00
-4.04931992e-01 -1.02694201e+00 -1.07249893e-01 3.91597271e-01
2.86114007e-01 -6.86615765e-01 7.67779171e-01 3.78185481e-01
1.45997748e-01 -1.12828195e+00 -1.24085343e+00 -1.17221737e+00
2.91292548e-01 -1.02025524e-01 3.38888884e-01 6.43091559e-01
4.46117848e-01 1.02213658e-01 -7.94122398e-01 3.21922526e-02
5.72308004e-01 1.60001934e-01 3.11587185e-01 -6.29174888e-01
-5.54187655e-01 -5.85456550e-01 -1.46820014e-02 -9.28765655e-01
4.22507137e-01 -5.94758987e-01 1.89556181e-01 -1.52730393e+00
-1.63795620e-01 -3.89257550e-01 -3.46201897e-01 7.48937309e-01
-6.02447577e-02 -2.62741745e-01 4.02000219e-01 -1.48783356e-01
-8.33599150e-01 5.05420446e-01 1.27504408e+00 4.66717444e-02
-5.19880712e-01 2.17955202e-01 -6.51646793e-01 5.46829343e-01
1.37886584e+00 -5.95796347e-01 -4.40343142e-01 2.90334653e-02
4.25197959e-01 4.70993042e-01 1.81413032e-02 -6.06525779e-01
2.48901714e-02 -1.04693210e+00 -7.92667493e-02 -6.74360096e-02
3.50600965e-02 -6.75538123e-01 -3.64060342e-01 1.11976063e+00
-5.00817716e-01 1.28849477e-01 1.01759322e-01 4.19130594e-01
1.32220849e-01 -8.13204288e-01 9.87895727e-01 -3.05872679e-01
-7.67659485e-01 1.53942809e-01 -9.80106294e-01 1.66865110e-01
1.63356709e+00 1.67668417e-01 1.64421424e-01 -6.32117152e-01
-6.51081800e-01 9.90436912e-01 2.24367023e-01 2.18091071e-01
6.18827224e-01 -1.14502406e+00 -6.16768956e-01 -1.46896422e-01
-3.36550832e-01 -4.96958643e-01 -4.54513341e-01 6.07812941e-01
-1.53710350e-01 4.59109753e-01 -5.11214435e-01 -1.58422589e-01
-1.31256366e+00 7.68016636e-01 6.79015875e-01 -6.89264655e-01
-2.80421704e-01 3.54297936e-01 -3.66456568e-01 -2.95278311e-01
3.48977685e-01 -7.55706802e-02 -2.58762270e-01 -1.19304053e-01
5.53869069e-01 4.73822057e-01 -4.35579121e-01 -5.22081442e-02
-3.90120536e-01 5.62090516e-01 1.67616401e-02 -6.74588442e-01
1.35804605e+00 3.07031795e-02 2.92124897e-02 -1.05458975e-01
7.57574618e-01 -2.77278781e-01 -1.41693652e+00 5.20099923e-02
2.63110310e-01 -4.78586465e-01 -4.42893058e-02 -8.47697377e-01
-8.48753512e-01 2.01772362e-01 3.95344406e-01 6.21299148e-01
1.19343865e+00 -2.64178246e-01 4.01886344e-01 6.74875140e-01
4.41850841e-01 -1.62462413e+00 2.29620233e-01 6.20970368e-01
6.93318844e-01 -8.58757615e-01 2.01385826e-01 9.91133526e-02
-8.96894991e-01 1.23642623e+00 5.38713276e-01 -4.26651001e-01
1.28549844e-01 2.36569792e-01 -3.09960008e-01 4.35486734e-02
-9.76243377e-01 -6.43760622e-01 -3.41979682e-01 6.55478954e-01
-8.23833123e-02 -2.81336531e-02 -9.16967392e-01 -2.60162413e-01
2.51802877e-02 8.11778530e-02 6.83163226e-01 1.39054930e+00
-8.87990057e-01 -1.62244582e+00 -5.18242359e-01 3.59745681e-01
-5.01781285e-01 2.60182559e-01 -4.99470949e-01 6.68123007e-01
-3.31320435e-01 1.17946017e+00 -2.45020062e-01 1.46776671e-02
4.09117699e-01 -1.11815974e-01 7.60132611e-01 -4.22148079e-01
-6.32805586e-01 7.59858713e-02 3.97143304e-01 -9.32564497e-01
-4.99921024e-01 -7.36407757e-01 -1.44698012e+00 -1.87374219e-01
-2.51264218e-02 3.89523923e-01 5.82922041e-01 8.98358405e-01
-1.80854470e-01 5.40338457e-01 1.15289772e+00 -6.97673857e-01
-1.30929339e+00 -4.04728889e-01 -5.64512432e-01 1.82425201e-01
5.15136898e-01 -7.09334731e-01 -2.73924679e-01 -5.30686677e-01]
|
[4.205593585968018, 2.3066813945770264]
|
aa318500-b17b-4674-899b-919b192f2a40
|
context-dependent-sentiment-analysis-in-user
| null | null |
https://aclanthology.org/P17-1081
|
https://aclanthology.org/P17-1081.pdf
|
Context-Dependent Sentiment Analysis in User-Generated Videos
|
Multimodal sentiment analysis is a developing area of research, which involves the identification of sentiments in videos. Current research considers utterances as independent entities, i.e., ignores the interdependencies and relations among the utterances of a video. In this paper, we propose a LSTM-based model that enables utterances to capture contextual information from their surroundings in the same video, thus aiding the classification process. Our method shows 5-10{\%} performance improvement over the state of the art and high robustness to generalizability.
|
['Louis-Philippe Morency', 'Amir Zadeh', 'Soujanya Poria', 'Navonil Majumder', 'Erik Cambria', 'Devamanyu Hazarika']
|
2017-07-01
| null | null | null |
acl-2017-7
|
['multimodal-emotion-recognition', 'emotion-recognition-in-conversation', 'multimodal-emotion-recognition']
|
['computer-vision', 'natural-language-processing', 'speech']
|
[ 4.39840667e-02 -2.55586982e-01 -5.71340062e-02 -6.79761827e-01
-2.70233214e-01 -5.19796073e-01 3.11976850e-01 9.46103185e-02
-4.72306550e-01 5.16222537e-01 5.07832646e-01 -2.57320870e-02
3.97288471e-01 -3.59477341e-01 -6.78229094e-01 -7.80203044e-01
1.03883579e-01 -3.33387047e-01 -3.95953432e-02 -3.28484714e-01
5.30720502e-02 8.58934373e-02 -1.32905412e+00 7.03153074e-01
-2.44834311e-02 1.18408072e+00 7.37974718e-02 7.20435858e-01
-3.97892743e-02 1.34088373e+00 -8.62342358e-01 -4.94723499e-01
-4.06482846e-01 -2.93533057e-01 -6.56405687e-01 2.85491705e-01
4.29659754e-01 -5.38330078e-01 -4.81023222e-01 8.49181712e-01
4.57631797e-01 2.33712614e-01 3.85862648e-01 -1.27366531e+00
-7.02899575e-01 3.36651236e-01 -4.17309433e-01 2.08940908e-01
6.02435350e-01 -4.70043421e-01 8.57786953e-01 -8.47875655e-01
4.29971874e-01 1.12286532e+00 4.20295089e-01 4.66780752e-01
-4.92622912e-01 -6.26579583e-01 6.40323997e-01 2.98920125e-01
-1.24479580e+00 -6.51492119e-01 9.20634747e-01 -3.92699927e-01
1.02798843e+00 1.22726336e-01 4.42556262e-01 1.42156315e+00
3.32869738e-01 1.17000413e+00 8.25418115e-01 -4.70789284e-01
2.01173842e-01 4.26334500e-01 2.29845256e-01 5.90897739e-01
-2.92056054e-01 -5.10575950e-01 -1.01034248e+00 1.05920382e-01
5.74044764e-01 -3.32041411e-03 -1.78683713e-01 1.37556624e-02
-1.19798708e+00 7.44381845e-01 2.15057716e-01 2.58802086e-01
-4.10689741e-01 1.76963285e-01 7.18596339e-01 3.74560952e-01
5.18638909e-01 1.24002229e-02 -3.50488514e-01 -3.51145089e-01
-5.56638360e-01 -2.05681890e-01 8.64986062e-01 1.16382241e+00
3.14012229e-01 7.58153200e-02 9.79952663e-02 4.86442089e-01
5.27314901e-01 5.85335910e-01 4.40937489e-01 -8.56258154e-01
4.78022605e-01 5.05090415e-01 1.51068315e-01 -1.43453670e+00
-5.39390445e-01 2.04963442e-02 -4.59843487e-01 -3.72285604e-01
-2.19983198e-02 -7.22928405e-01 -6.45990551e-01 1.70930791e+00
2.54255444e-01 1.62677035e-01 3.36348504e-01 8.17380250e-01
1.27590275e+00 7.24885583e-01 2.72344232e-01 -3.58588129e-01
1.19027281e+00 -9.13643241e-01 -1.12997437e+00 -1.80358291e-01
5.24111927e-01 -7.31760144e-01 7.31910348e-01 3.80352408e-01
-7.40927041e-01 -5.48157990e-01 -8.83029044e-01 2.03285292e-01
-5.38047314e-01 2.19292000e-01 5.99976361e-01 5.34848571e-01
-1.18169940e+00 -4.38292436e-02 -7.14589834e-01 -5.21423042e-01
5.31001873e-02 5.40957391e-01 -5.19759297e-01 1.55388221e-01
-1.36868680e+00 8.42154384e-01 1.60169870e-01 3.66543442e-01
-6.12807810e-01 1.46863377e-02 -1.14614284e+00 -2.19916672e-01
4.52053815e-01 -1.94621176e-01 1.35777354e+00 -1.35300267e+00
-1.55784333e+00 3.49827260e-01 -6.94679916e-01 -1.04333870e-01
-1.77755624e-01 -4.11003560e-01 -7.88849354e-01 3.60622436e-01
-1.43380746e-01 5.75747550e-01 7.74416447e-01 -1.33132672e+00
-7.45425284e-01 -2.55357802e-01 6.59788549e-01 5.56562960e-01
-8.73823524e-01 4.27583218e-01 -7.91364193e-01 -4.40626919e-01
-9.95329842e-02 -1.09397054e+00 -1.25437558e-01 -4.23568904e-01
-2.78049499e-01 -5.55826165e-02 1.23257077e+00 -6.05302393e-01
1.27921236e+00 -2.37445068e+00 1.04442984e-01 1.22050634e-02
8.73072594e-02 9.77824852e-02 -1.10117115e-01 5.70526421e-01
9.45108756e-02 8.82468969e-02 3.77064764e-01 -6.04200721e-01
2.44805142e-02 2.83279419e-01 -3.23262542e-01 4.96678352e-01
1.21333785e-01 7.37414718e-01 -8.47468317e-01 -5.37202835e-01
2.64489174e-01 6.75719917e-01 -1.33574799e-01 1.77843183e-01
8.15001875e-03 5.23292542e-01 -6.81024313e-01 7.71164417e-01
2.63229370e-01 -2.04749927e-01 2.91234225e-01 -4.83565032e-01
-7.57353753e-02 -6.41884282e-03 -8.72165203e-01 1.59890962e+00
-4.50609684e-01 1.24025595e+00 3.96893732e-02 -9.29722369e-01
6.41400695e-01 9.43470836e-01 4.34256881e-01 -5.69794834e-01
4.97130722e-01 -1.91591561e-01 -1.10407993e-01 -1.05115354e+00
6.90120995e-01 6.49309456e-02 -2.60836214e-01 3.56695890e-01
1.12629198e-01 4.46449965e-01 1.17011942e-01 1.67701125e-01
7.53291428e-01 2.01495290e-01 2.21449271e-01 2.12538585e-01
5.18602133e-01 -3.83851469e-01 3.12768251e-01 5.30184627e-01
-4.51884359e-01 3.04531097e-01 5.16086817e-01 -3.80296558e-01
-4.87639338e-01 -5.19732714e-01 1.99370123e-02 1.42344642e+00
4.90749329e-01 -3.02776664e-01 -7.66944230e-01 -7.20863402e-01
-3.44142169e-01 4.81795102e-01 -8.19303095e-01 -5.63902818e-02
-3.50810885e-01 -5.52629709e-01 4.04982835e-01 8.52849007e-01
3.92866224e-01 -1.06150186e+00 -7.74144590e-01 -1.65155362e-02
-5.67200541e-01 -1.57784104e+00 -2.43680775e-01 -1.30791953e-02
-5.37968695e-01 -8.35565507e-01 -4.31362718e-01 -7.62300670e-01
7.34263539e-01 3.84817004e-01 9.16179121e-01 -2.68156618e-01
4.95887548e-01 7.97309637e-01 -7.26601422e-01 -6.98066354e-01
-1.10371634e-01 -2.11872101e-01 2.12881029e-01 4.03304309e-01
7.95067906e-01 -3.95085588e-02 -4.26827729e-01 5.05541861e-01
-8.59106898e-01 -1.84173912e-01 4.65649188e-01 7.34522700e-01
4.41457301e-01 2.27149621e-01 6.15981102e-01 -6.12425148e-01
6.81638360e-01 -6.13288760e-01 5.91620170e-02 3.13008368e-01
2.80682612e-02 -5.08604705e-01 5.44738054e-01 -5.37336886e-01
-1.32021236e+00 1.15137957e-01 7.44124055e-02 -4.68780935e-01
-5.81906736e-01 8.14139545e-01 -2.06261352e-01 -1.90845385e-01
6.78304732e-02 1.09003678e-01 -3.65089148e-01 -5.61502005e-04
2.74403483e-01 1.04460788e+00 1.96376890e-01 -2.27408320e-01
1.11457091e-02 6.25553370e-01 -2.81008303e-01 -1.18892777e+00
-9.97897446e-01 -6.55918241e-01 -8.06216061e-01 -8.81492674e-01
9.64050949e-01 -1.10416007e+00 -9.08149838e-01 3.81838739e-01
-1.17904806e+00 1.19916633e-01 3.23465705e-01 7.39212871e-01
-2.77155966e-01 2.82664418e-01 -8.81487906e-01 -9.69495177e-01
-6.92680851e-02 -1.30292547e+00 1.18054473e+00 2.89530069e-01
-5.33629954e-01 -1.16457570e+00 -1.12092271e-01 6.74471915e-01
1.69545591e-01 2.64607519e-01 3.84316713e-01 -7.51696646e-01
-1.81253225e-01 -4.90199506e-01 -3.64007130e-02 4.64427173e-01
2.24313244e-01 2.58911878e-01 -1.17869699e+00 -2.43177146e-01
3.27332020e-01 -3.47448707e-01 5.19373894e-01 4.70929623e-01
6.02994561e-01 -3.09618235e-01 -3.66475642e-01 -7.08064908e-05
1.01462686e+00 8.05478036e-01 4.37250465e-01 2.47164696e-01
7.94340670e-01 1.04373085e+00 7.37842262e-01 3.87793243e-01
6.57503307e-01 4.41459447e-01 5.20471692e-01 -3.12937707e-01
3.83732289e-01 -4.16990072e-02 6.26281738e-01 1.19773877e+00
-3.62955555e-02 -7.08234429e-01 -7.01997519e-01 4.65020776e-01
-2.11423826e+00 -9.09804523e-01 -2.26982325e-01 1.63817370e+00
3.31466943e-01 -8.24191887e-03 -1.09358169e-01 -1.31356031e-01
7.35317290e-01 3.19258004e-01 -4.86489356e-01 -5.45217693e-01
-1.17840670e-01 -3.31629962e-01 2.74658144e-01 2.49897823e-01
-1.52449656e+00 9.58157659e-01 7.18797731e+00 2.67998993e-01
-1.41520202e+00 8.97948667e-02 5.60413241e-01 -3.50788355e-01
9.15675014e-02 -4.22521144e-01 -6.05210721e-01 4.14249808e-01
9.88115489e-01 3.30935001e-01 -5.42985871e-02 6.70567930e-01
4.94242817e-01 -2.81194806e-01 -1.01661909e+00 7.50037730e-01
8.03533971e-01 -8.47001314e-01 -5.15982136e-02 -2.05293983e-01
7.89358675e-01 3.46070901e-02 1.41081840e-01 1.17306843e-01
-4.02842090e-02 -6.88701332e-01 7.82027066e-01 6.40124381e-01
4.14316803e-01 -8.21600139e-01 1.32969511e+00 7.60648772e-02
-1.08054018e+00 -4.55689467e-02 2.01219376e-02 -1.80040136e-01
4.40802991e-01 -1.17087541e-02 -6.02362812e-01 4.47929591e-01
1.01069236e+00 9.95731056e-01 -3.15807790e-01 3.76443416e-01
-2.17720523e-01 7.08456278e-01 3.17349583e-02 -3.83948505e-01
3.25713754e-01 -1.04909785e-01 3.56301814e-01 1.53497791e+00
6.15973063e-02 2.56759018e-01 2.39702106e-01 8.46438110e-02
-3.11422870e-02 1.04886815e-01 -7.71470964e-01 -1.39814377e-01
1.48624182e-01 1.21116793e+00 -7.86530912e-01 -3.82466376e-01
-8.82893443e-01 8.52499187e-01 -1.67357195e-02 7.29091346e-01
-9.51808572e-01 -3.09063405e-01 5.12324750e-01 -6.68082058e-01
5.19818485e-01 -2.10114032e-01 -2.62667760e-02 -1.17754006e+00
1.80857211e-01 -7.89924979e-01 1.81662530e-01 -1.03893912e+00
-1.05112433e+00 8.65609527e-01 -1.01620413e-01 -1.42960238e+00
-2.20534667e-01 -7.61625707e-01 -5.06467760e-01 3.87487531e-01
-1.29344916e+00 -1.34703553e+00 -1.87982365e-01 6.42038405e-01
6.28009677e-01 -1.56262740e-01 9.01230097e-01 4.03279930e-01
-9.36958671e-01 4.52026427e-01 3.49049531e-02 4.61586624e-01
8.04568231e-01 -8.94962609e-01 -9.86694992e-02 8.78748894e-01
1.21116035e-01 8.03665042e-01 8.13426018e-01 -4.45538998e-01
-1.54118836e+00 -8.40177059e-01 9.55050826e-01 -4.79123771e-01
9.70099628e-01 -3.02351147e-01 -5.23611009e-01 9.57965791e-01
6.31494224e-01 -3.63463223e-01 1.29076898e+00 1.17485903e-01
-1.58307344e-01 -9.90612730e-02 -8.42696071e-01 6.81295156e-01
4.96419370e-01 -8.86570334e-01 -4.78727162e-01 -7.44299078e-03
7.84386992e-01 -4.60434526e-01 -8.69649947e-01 3.91149610e-01
8.63364339e-01 -8.92323792e-01 6.55807614e-01 -6.02041245e-01
6.62642300e-01 -2.61308223e-01 -5.01744866e-01 -1.14919019e+00
8.19071531e-02 -1.85881585e-01 -3.37197006e-01 1.27046907e+00
4.91100907e-01 -3.39293301e-01 6.30988002e-01 9.58275080e-01
-1.90910269e-02 -6.37931824e-01 -6.85920298e-01 -1.87318832e-01
-5.25187314e-01 -6.15673900e-01 2.91868240e-01 1.23704684e+00
5.34505188e-01 5.13936222e-01 -6.94821000e-01 3.65681231e-01
1.42137259e-01 -1.12216905e-01 5.97685575e-01 -5.35154402e-01
4.49967384e-02 -3.14607948e-01 -5.98001897e-01 -1.15513527e+00
4.45814699e-01 -6.12736344e-02 1.13237172e-01 -1.47559881e+00
5.25497235e-02 1.92953244e-01 -5.42954445e-01 5.23972094e-01
6.54622540e-02 3.62057269e-01 2.43164077e-01 -2.90042982e-02
-1.07891142e+00 5.59556127e-01 1.00397146e+00 -2.35327944e-01
-1.54800445e-01 -9.78750810e-02 -6.59733891e-01 9.88502264e-01
7.57921576e-01 -2.05889776e-01 -4.29177731e-01 -7.37471402e-01
1.98995396e-01 1.29868239e-01 8.16668943e-03 -6.75967276e-01
4.97379065e-01 -1.27365857e-01 2.98206568e-01 -6.55422747e-01
8.07035804e-01 -1.07648778e+00 -1.12830281e-01 -1.24813773e-01
-5.46594441e-01 2.79287577e-01 3.86037916e-01 6.57343447e-01
-8.02560806e-01 -2.11837977e-01 -2.94256806e-02 1.24656567e-02
-9.82215226e-01 1.45741869e-02 -7.97972977e-01 -4.85529840e-01
9.84525502e-01 -1.60602003e-01 -2.86853433e-01 -9.05523539e-01
-6.48557603e-01 3.17285180e-01 6.96788654e-02 6.81494474e-01
7.37906516e-01 -1.16365564e+00 -3.58196586e-01 -2.06197053e-01
2.24207625e-01 -3.27338219e-01 3.84221911e-01 8.43972981e-01
-1.61017045e-01 5.10983884e-01 5.12578199e-03 -5.22782862e-01
-1.65047967e+00 3.26112241e-01 2.12892696e-01 3.65972340e-01
-4.49024774e-02 8.67324293e-01 3.29940051e-01 -1.29257336e-01
5.28066218e-01 -1.17730409e-01 -9.19588685e-01 4.80873078e-01
4.97479469e-01 1.87526822e-01 -1.21285453e-01 -1.26667643e+00
-4.12422180e-01 5.36504149e-01 -9.57932249e-02 -1.23934932e-01
9.84833002e-01 -4.93942529e-01 -1.78005040e-01 1.03651261e+00
1.44834888e+00 -9.39776599e-02 -1.04795957e+00 -3.71926844e-01
-5.58666587e-02 -2.50440687e-01 1.26013681e-01 -6.67278588e-01
-1.19505095e+00 7.27815807e-01 3.45250338e-01 3.99226487e-01
1.24530053e+00 8.55379701e-02 6.71081424e-01 6.16327941e-01
2.39802346e-01 -1.29845691e+00 1.30742192e-01 5.35782754e-01
6.26094162e-01 -1.54634154e+00 -7.01560676e-02 -1.31559387e-01
-1.22852027e+00 1.14071941e+00 6.53534234e-01 2.16484308e-01
7.43081391e-01 2.09437028e-01 5.54882109e-01 -3.50868464e-01
-7.86147118e-01 -6.47242740e-02 3.72835845e-01 3.47429395e-01
8.04467618e-01 -4.76290099e-03 -1.18885804e-02 5.55560768e-01
1.57604679e-01 -1.55531511e-01 5.73248625e-01 1.16598916e+00
-1.45164594e-01 -6.77705824e-01 -2.52745211e-01 2.51534849e-01
-5.50783694e-01 -1.49459779e-01 -5.33756554e-01 6.20218873e-01
4.08184305e-02 1.46897960e+00 7.62091652e-02 -7.65634656e-01
4.77064252e-01 -9.14296731e-02 -5.14975674e-02 -3.66132587e-01
-4.50877488e-01 3.72162789e-01 3.23454618e-01 -6.61267757e-01
-1.21879148e+00 -7.91526437e-01 -1.14661860e+00 -9.03562456e-02
-4.68517274e-01 1.64697871e-01 8.44807923e-01 1.22390735e+00
4.18360621e-01 6.86132789e-01 8.28962684e-01 -8.29532087e-01
1.46566510e-01 -8.44893038e-01 -3.99779797e-01 2.72735983e-01
6.51730478e-01 -5.51312208e-01 -2.85073251e-01 4.98075366e-01]
|
[13.155423164367676, 5.25435733795166]
|
2041978e-3477-4abf-a985-e8516e4b5362
|
190600654
|
1906.00654
| null |
https://arxiv.org/abs/1906.00654v1
|
https://arxiv.org/pdf/1906.00654v1.pdf
|
Continual Learning of New Sound Classes using Generative Replay
|
Continual learning consists in incrementally training a model on a sequence of datasets and testing on the union of all datasets. In this paper, we examine continual learning for the problem of sound classification, in which we wish to refine already trained models to learn new sound classes. In practice one does not want to maintain all past training data and retrain from scratch, but naively updating a model with new data(sets) results in a degradation of already learned tasks, which is referred to as "catastrophic forgetting." We develop a generative replay procedure for generating training audio spectrogram data, in place of keeping older training datasets. We show that by incrementally refining a classifier with generative replay a generator that is 4% of the size of all previous training data matches the performance of refining the classifier keeping 20% of all previous training data. We thus conclude that we can extend a trained sound classifier to learn new classes without having to keep previously used datasets.
|
['Cem Subakan', 'Zhepei Wang', 'Paris Smaragdis', 'Efthymios Tzinis', 'Laurent Charlin']
|
2019-06-03
| null | null | null | null |
['sound-classification']
|
['audio']
|
[ 7.59167969e-01 2.91109294e-01 3.59214395e-01 -2.57682681e-01
-8.57057750e-01 -6.73455715e-01 4.05004740e-01 1.67499930e-01
-5.67833066e-01 1.05947340e+00 -8.18546638e-02 -8.19640085e-02
1.38455275e-02 -1.01331341e+00 -9.65109825e-01 -7.35646486e-01
-1.17129982e-01 5.81144452e-01 4.74182725e-01 3.29543068e-03
1.34350628e-01 3.72771442e-01 -2.20614600e+00 4.91626889e-01
5.63406944e-01 2.98708230e-01 3.69033605e-01 1.27858794e+00
6.05814010e-02 6.52191222e-01 -1.05656135e+00 -2.45763138e-01
1.26977146e-01 -8.40998590e-01 -1.08578110e+00 2.98311710e-01
4.45262700e-01 -6.81986660e-03 -8.37027952e-02 9.07401741e-01
2.79188275e-01 4.28590178e-01 2.26077154e-01 -1.38046873e+00
-5.20396590e-01 1.03876936e+00 -4.60367203e-02 2.14928880e-01
2.28922829e-01 9.69577357e-02 6.71624541e-01 -7.35065520e-01
3.45950216e-01 7.64361501e-01 9.90424275e-01 9.94229555e-01
-1.50181830e+00 -7.19556987e-01 9.80251431e-02 -2.05025405e-01
-1.29697299e+00 -7.20566809e-01 8.03075135e-01 -3.70315939e-01
8.04214537e-01 5.05026519e-01 9.68547106e-01 8.90060246e-01
2.34056845e-01 3.10867071e-01 8.94426167e-01 -8.79331946e-01
4.82828647e-01 2.86077917e-01 1.77272812e-01 6.48874938e-01
2.97657728e-01 1.13939039e-01 -6.49999321e-01 -3.48207206e-01
3.43425751e-01 -7.95771629e-02 -2.04571337e-01 -8.28134194e-02
-8.98390472e-01 4.64639455e-01 4.40730564e-02 4.10338730e-01
-2.05387697e-01 2.52513707e-01 1.88611209e-01 6.68379247e-01
6.66861057e-01 6.39298916e-01 -5.85505307e-01 -2.18053415e-01
-1.00365663e+00 1.91185653e-01 7.58013129e-01 9.38265026e-01
1.09363711e+00 4.01026130e-01 4.65127707e-01 9.26486433e-01
-3.81365806e-01 3.62647086e-01 9.06727791e-01 -9.54110920e-01
-9.74329747e-03 3.12362939e-01 -9.68661904e-02 -1.70945063e-01
-5.66428229e-02 -6.04617834e-01 -7.17201233e-01 1.96817324e-01
7.88671076e-02 -5.71496665e-01 -1.15242290e+00 2.02976656e+00
1.83569908e-01 7.21381724e-01 2.57243365e-01 -2.10439324e-01
2.14360118e-01 7.32445359e-01 -5.09185158e-03 -6.34837866e-01
4.63345706e-01 -5.76829314e-01 -2.33251527e-01 -2.19661117e-01
5.00410676e-01 -5.59703350e-01 1.16177440e+00 9.33084130e-01
-1.10265851e+00 -1.26007688e+00 -1.31670344e+00 3.80647451e-01
-2.44175673e-01 -2.07589999e-01 4.43362653e-01 7.56136417e-01
-1.21022332e+00 8.71356130e-01 -1.01021576e+00 -1.02267638e-01
1.50634363e-01 3.97674680e-01 -2.34673008e-01 1.79490130e-02
-9.52936292e-01 4.58194971e-01 7.11097896e-01 -2.00805157e-01
-1.24560833e+00 -8.92886400e-01 -5.02686799e-01 -1.14106268e-01
1.74214914e-01 -6.61417484e-01 1.68571067e+00 -9.34012592e-01
-1.10098326e+00 2.09722579e-01 -1.47811234e-01 -5.00940382e-01
2.42005348e-01 -4.31539893e-01 -6.35439396e-01 -3.74283195e-01
-4.39165439e-03 7.12820530e-01 1.07079268e+00 -1.38468671e+00
-9.86169398e-01 -1.70340165e-01 -1.66662022e-01 -7.52625391e-02
-4.81791347e-01 -6.06600940e-01 -1.63685493e-02 -5.89131057e-01
1.93940550e-01 -1.19713295e+00 -9.90094543e-02 -6.19003832e-01
4.14465880e-03 -9.74000841e-02 9.02255237e-01 -2.03785658e-01
1.37933969e+00 -2.26057982e+00 7.46719763e-02 1.45136550e-01
4.66756187e-02 8.16518813e-02 -1.09572545e-01 2.61671394e-01
-3.44872385e-01 9.84673500e-02 -5.80850065e-01 -4.66048300e-01
-5.91767967e-01 6.39255524e-01 -7.93016851e-01 -2.49665609e-04
1.25251114e-01 4.25482929e-01 -1.03550661e+00 -2.41371214e-01
-3.50484818e-01 1.61903128e-01 -9.70738471e-01 1.88990965e-01
-4.78572309e-01 2.93068916e-01 1.12199023e-01 8.30628444e-04
3.80112588e-01 7.64458254e-02 1.57766640e-01 5.41474223e-01
-4.87092659e-02 2.62102246e-01 -1.38561893e+00 1.85357606e+00
-5.79623938e-01 5.33727288e-01 -5.07560372e-01 -7.51732707e-01
8.35037053e-01 5.34795165e-01 3.64353567e-01 -3.82192917e-02
-3.57001632e-01 2.23124579e-01 6.46211877e-02 -2.90595710e-01
4.22905594e-01 -6.97438002e-01 -1.67670488e-01 8.92420113e-01
4.32035804e-01 -5.00639379e-01 1.39110714e-01 1.97942406e-02
1.24563563e+00 7.37820268e-02 1.06549216e-02 1.56007573e-01
2.65691578e-01 -5.03379665e-02 5.78595281e-01 1.01563239e+00
3.42888504e-01 4.05792624e-01 1.34144485e-01 -3.98369312e-01
-1.27801704e+00 -1.16616046e+00 -1.42298732e-02 1.40685022e+00
-5.56200743e-01 -5.40519714e-01 -9.60390210e-01 -8.68359685e-01
-6.20819777e-02 1.10947859e+00 -8.26449633e-01 -5.66738844e-01
-8.96446526e-01 -8.55986774e-01 6.39248669e-01 5.08067489e-01
3.88784915e-01 -1.27577853e+00 -6.06023729e-01 5.37102997e-01
1.35101691e-01 -2.69284278e-01 -3.37602407e-01 7.24405229e-01
-1.17667735e+00 -9.07717645e-01 -2.88745493e-01 -8.71280372e-01
6.40893996e-01 7.56214038e-02 1.03232062e+00 2.79167295e-01
-3.04246664e-01 3.30131143e-01 -3.52566689e-01 -6.01070344e-01
-1.13445330e+00 4.05520082e-01 3.56092870e-01 -1.22273788e-01
4.21817340e-02 -8.43545973e-01 1.24426417e-01 -1.90349519e-01
-1.22240925e+00 8.17056000e-02 3.72135639e-01 9.39718306e-01
6.04193449e-01 7.69330561e-01 9.15190399e-01 -1.38182032e+00
6.28829062e-01 -4.96294618e-01 -2.49589205e-01 4.26097006e-01
-7.47276127e-01 3.60973656e-01 6.45009875e-01 -7.28657305e-01
-1.12999010e+00 2.88573205e-01 -3.39422733e-01 -2.78674126e-01
-2.99848020e-01 2.06040278e-01 3.88351642e-02 3.78929526e-01
1.04625130e+00 4.04549807e-01 -1.24139749e-01 -7.07101941e-01
4.36625004e-01 6.60036981e-01 8.22652102e-01 -7.87177861e-01
7.57968962e-01 -6.39216751e-02 -3.01978201e-01 -6.89199150e-01
-9.10897851e-01 -6.99939802e-02 -9.24597859e-01 -2.77392328e-01
3.31331521e-01 -5.02684414e-01 -1.28970593e-01 6.38296485e-01
-8.99237633e-01 -4.89522308e-01 -1.10619640e+00 3.43405962e-01
-5.76022446e-01 3.06710023e-02 -2.38688916e-01 -7.81542480e-01
-1.22351922e-01 -6.77132607e-01 7.59205282e-01 -9.45666432e-02
-3.86494607e-01 -8.29646051e-01 8.03640485e-01 -2.99852252e-01
1.53273180e-01 2.21423656e-02 1.09322548e+00 -7.89207280e-01
-6.50545359e-02 -3.68519455e-01 9.25111949e-01 8.41762602e-01
4.98820722e-01 -2.04001442e-01 -1.34649217e+00 -5.25839567e-01
3.43262345e-01 -3.77088130e-01 9.20496821e-01 -2.58952737e-01
1.29454899e+00 -4.63383734e-01 -2.33779266e-01 2.29319364e-01
1.20062423e+00 7.14013636e-01 4.26254749e-01 -1.44205391e-01
4.45080310e-01 2.49420479e-01 3.14507753e-01 3.73398125e-01
-2.80706704e-01 6.37260973e-02 -1.93288296e-01 3.24695647e-01
-3.00471514e-01 -6.35295451e-01 2.27142408e-01 9.72661018e-01
-8.22284967e-02 -1.48053974e-01 -9.71207738e-01 5.52483082e-01
-1.40919685e+00 -1.22080183e+00 3.91523153e-01 2.53165627e+00
1.45215404e+00 4.72034693e-01 9.46996808e-02 8.96713436e-01
7.44126976e-01 -3.73051703e-01 -5.57036638e-01 -1.90225333e-01
3.29396933e-01 9.82614040e-01 -5.24338894e-02 8.73266876e-01
-8.76961052e-01 8.48871529e-01 7.08577061e+00 5.49318373e-01
-1.06065559e+00 2.72107959e-01 4.47749168e-01 -2.32621491e-01
-4.28027809e-01 2.71757215e-01 -9.24394011e-01 2.92197585e-01
1.39675653e+00 -3.33951056e-01 5.96637011e-01 8.81384850e-01
-4.25220907e-01 -2.26369962e-01 -1.45285034e+00 4.31657165e-01
2.55604655e-01 -1.14377356e+00 2.78613806e-01 -2.20761865e-01
1.04881835e+00 -2.34526455e-01 8.34066980e-03 6.61820948e-01
6.66646838e-01 -5.81704021e-01 6.83011532e-01 7.83285201e-01
8.90211165e-01 -9.67037499e-01 3.60149145e-01 7.58232951e-01
-9.82220888e-01 -3.11934263e-01 -2.01158091e-01 2.30560247e-02
-9.52306539e-02 5.39208472e-01 -1.28383434e+00 2.52388149e-01
5.20371556e-01 2.82103211e-01 -9.32802916e-01 1.18031657e+00
1.09734178e-01 1.07996321e+00 -2.76554257e-01 1.82031080e-01
-2.75079966e-01 3.88305902e-01 3.24215680e-01 1.00293851e+00
6.58518434e-01 4.31217067e-02 1.54247373e-01 4.08647329e-01
-9.58342850e-02 -3.18603992e-01 -5.87384820e-01 1.30086124e-01
6.15597665e-01 6.59773707e-01 -6.25758052e-01 -8.09769928e-01
-8.76081437e-02 9.05583262e-01 1.61772609e-01 1.25344083e-01
-5.26576281e-01 -5.21555066e-01 1.25302494e-01 2.12984860e-01
2.23448426e-01 -1.73814014e-01 -3.99993695e-02 -9.88940120e-01
-2.94661075e-01 -7.52955675e-01 4.78814274e-01 -9.31942523e-01
-1.02674675e+00 8.61474812e-01 2.26200759e-01 -1.12305748e+00
-7.02160478e-01 1.13559909e-01 -5.53996086e-01 6.83145642e-01
-7.67508984e-01 -7.78641224e-01 -1.86161086e-01 5.11306345e-01
1.00816500e+00 -4.41369981e-01 1.10894418e+00 -2.91348156e-02
-1.55198455e-01 5.15805900e-01 2.53161113e-03 -1.53827965e-01
5.42125762e-01 -1.33602691e+00 5.41631639e-01 7.82908082e-01
8.55516791e-01 7.28962362e-01 8.35624278e-01 -9.03696954e-01
-8.25237215e-01 -1.25190127e+00 9.96377349e-01 -5.79312801e-01
4.50082570e-01 -4.83751655e-01 -1.28151870e+00 1.00545382e+00
8.95390809e-02 -1.59171507e-01 8.74528825e-01 1.84847534e-01
-1.79266408e-01 -5.33108652e-01 -9.03977871e-01 1.10970266e-01
1.03843212e+00 -5.37216008e-01 -8.94082189e-01 1.49956465e-01
7.97548413e-01 -1.63097441e-01 -4.92368430e-01 3.38817447e-01
4.34589177e-01 -7.11223662e-01 4.96890098e-01 -8.50751221e-01
-1.50350645e-01 -2.50891536e-01 2.69934684e-02 -1.65414989e+00
-3.16836387e-01 -6.82440639e-01 -9.49445441e-02 1.23007679e+00
5.16309857e-01 -4.68678564e-01 9.14912105e-01 2.51359075e-01
-4.98748809e-01 -3.44851971e-01 -9.10808802e-01 -6.71146810e-01
3.10794026e-01 -7.61754572e-01 9.27057147e-01 7.78880656e-01
-1.14366271e-01 4.87657160e-01 -2.42833018e-01 1.56244829e-01
4.67493951e-01 9.25037311e-04 7.84910083e-01 -1.49974024e+00
-7.18416512e-01 1.10733837e-01 -1.48697495e-01 -7.68968701e-01
4.35650721e-03 -9.20668840e-01 3.74004453e-01 -8.99906218e-01
2.51573324e-01 -7.88675427e-01 -5.02995729e-01 8.53203058e-01
-1.53602809e-01 2.68763214e-01 -1.70007437e-01 3.03372949e-01
-2.98811764e-01 2.39989236e-01 1.03014135e+00 1.03990883e-02
-5.75281143e-01 5.90009511e-01 -7.26579249e-01 5.34723580e-01
8.48424435e-01 -8.89668763e-01 -1.02017200e+00 -1.84376836e-01
3.40554297e-01 5.64672612e-02 2.93303579e-01 -1.65670180e+00
3.81752878e-01 -7.82936215e-02 7.45958626e-01 -4.72111017e-01
1.82662725e-01 -6.05724931e-01 7.05140412e-01 5.86408734e-01
-7.21039236e-01 2.21184224e-01 3.93610358e-01 6.90021932e-01
1.72842462e-02 -6.76523089e-01 8.03303242e-01 -3.21538925e-01
-5.03762126e-01 1.56766083e-02 -5.51988959e-01 1.22911006e-01
9.42420959e-01 -1.78046331e-01 2.35422999e-02 -4.67544682e-02
-1.57142246e+00 -4.19027358e-01 3.03675503e-01 2.86077142e-01
6.05311871e-01 -1.11885345e+00 -5.59278309e-01 8.01184893e-01
-1.69969201e-01 -1.60385035e-02 1.33225068e-01 -9.49558467e-02
-1.58109456e-01 1.29108697e-01 -1.54727444e-01 -3.97206128e-01
-1.31793368e+00 6.69852734e-01 2.54031599e-01 1.32514775e-01
-5.69672465e-01 1.05050719e+00 -2.56857485e-01 -2.43351832e-01
2.56720394e-01 -4.34564948e-01 7.89751932e-02 7.34883919e-02
4.06071723e-01 2.12039515e-01 3.84403199e-01 1.01815276e-01
2.29958296e-01 7.77413324e-02 -3.24947774e-01 -6.32952929e-01
1.57383943e+00 2.86520332e-01 -1.52112711e-02 1.09723878e+00
9.16300118e-01 2.46399760e-01 -1.31865799e+00 -2.19672531e-01
-1.34698927e-01 -3.33170146e-01 -2.32845873e-01 -8.31271410e-01
-7.83668339e-01 7.86867738e-01 5.24275541e-01 5.24523616e-01
1.25503147e+00 9.56268609e-03 5.68183899e-01 8.57789695e-01
6.26141310e-01 -1.09137285e+00 3.92149240e-01 4.71672684e-01
8.43457103e-01 -4.35133845e-01 -1.07342645e-03 7.51470849e-02
-4.22135442e-01 1.04692519e+00 6.94747746e-01 -1.69247031e-01
9.44372475e-01 5.53324163e-01 -4.12774444e-01 1.96194008e-01
-1.26433206e+00 6.87140375e-02 4.14223485e-02 5.69099188e-01
2.36381575e-01 -1.31385237e-01 2.58276254e-01 4.94952530e-01
-9.61548030e-01 2.01594532e-01 5.55941582e-01 1.36345029e+00
-9.08242226e-01 -1.48071289e+00 -1.66432962e-01 6.50808215e-01
-4.90870103e-02 -6.84569404e-02 -3.91457647e-01 7.91735113e-01
9.53247130e-01 6.43073440e-01 3.90614301e-01 -7.14174390e-01
3.12373132e-01 9.30090606e-01 7.64702559e-01 -1.29774225e+00
-4.76195365e-01 1.58551097e-01 7.36469179e-02 3.54702771e-01
-2.79864162e-01 -1.05998981e+00 -1.16010725e+00 -1.64566100e-01
-4.62532580e-01 7.25352407e-01 3.88135523e-01 7.39605427e-01
2.86066309e-02 6.28338873e-01 8.09302688e-01 -4.22080845e-01
-6.02462113e-01 -1.12357616e+00 -5.54630637e-01 6.72841072e-03
7.19996631e-01 -4.56571460e-01 -5.01515090e-01 7.52142072e-01]
|
[9.971027374267578, 3.453063726425171]
|
67c86202-455b-4afa-a444-c024bb3edb1a
|
a-transition-based-system-for-universal
| null | null |
https://aclanthology.org/K17-3020
|
https://aclanthology.org/K17-3020.pdf
|
A Transition-based System for Universal Dependency Parsing
|
This paper describes the system for our participation in the CoNLL 2017 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies. In this work, we design a system based on UDPipe1 for universal dependency parsing, where multilingual transition-based models are trained for different treebanks. Our system directly takes raw texts as input, performing several intermediate steps like tokenizing and tagging, and finally generates the corresponding dependency trees. For the special surprise languages for this task, we adopt a delexicalized strategy and predict basing on transfer learning from other related languages. In the final evaluation of the shared task, our system achieves a result of 66.53{\%} in macro-averaged LAS F1-score.
|
['Hao Wang', 'Zhisong Zhang', 'Hai Zhao']
|
2017-08-01
| null | null | null |
conll-2017-8
|
['transition-based-dependency-parsing']
|
['natural-language-processing']
|
[-2.35163733e-01 9.15662050e-02 -2.05190629e-01 -6.52833164e-01
-1.52637446e+00 -8.16013098e-01 3.76244009e-01 1.25420123e-01
-7.09801972e-01 1.08976698e+00 3.30705911e-01 -8.38543177e-01
6.72524393e-01 -4.99184668e-01 -7.41745234e-01 -2.64849961e-01
-2.78650187e-02 6.19068384e-01 3.60455334e-01 -3.06779951e-01
-2.01731831e-01 2.09839165e-01 -5.77265024e-01 6.09665811e-01
1.01339817e+00 3.13033104e-01 3.30264598e-01 1.05973935e+00
-2.96391904e-01 8.90320241e-01 -7.49152243e-01 -9.23727930e-01
-8.12352523e-02 -4.42709744e-01 -1.30418539e+00 -4.54508662e-01
3.22164387e-01 -5.48811853e-02 2.17262015e-01 9.19246852e-01
3.69468391e-01 -2.05404729e-01 4.98545438e-01 -5.18225193e-01
-4.14676368e-01 1.45392430e+00 -2.39182234e-01 1.38157919e-01
5.29615164e-01 -8.69137570e-02 1.34395421e+00 -7.90361583e-01
6.96422637e-01 1.45844364e+00 4.35441583e-01 6.56363368e-01
-1.03403354e+00 -7.49968588e-01 2.25939259e-01 -1.44135594e-01
-1.08471942e+00 -5.33186316e-01 1.57629475e-01 -3.82428139e-01
1.53736913e+00 -1.55666545e-02 7.07058683e-02 1.17990959e+00
4.52898294e-01 7.23016918e-01 1.11396909e+00 -8.80154192e-01
-2.60762721e-01 6.56914562e-02 4.28940117e-01 6.49390996e-01
-1.60790935e-01 -2.90684193e-01 -4.26709354e-01 2.38101020e-01
3.23643327e-01 -9.49802756e-01 1.36249289e-01 6.02268279e-01
-1.18038225e+00 7.19250977e-01 1.31390214e-01 4.58550096e-01
-4.47400566e-03 -1.25527769e-01 5.88776052e-01 4.56355780e-01
7.96016634e-01 1.37241751e-01 -1.04077327e+00 -5.89259714e-02
-4.98034298e-01 -7.01317489e-02 1.09992778e+00 1.20440912e+00
6.05548501e-01 -2.30692014e-01 -9.90118980e-02 9.58021224e-01
2.39914849e-01 4.98008221e-01 3.15361768e-01 -5.88711977e-01
1.12032938e+00 2.04749808e-01 -5.55650145e-02 9.82321948e-02
-3.57079476e-01 7.55020604e-03 -3.44585270e-01 -1.56442940e-01
6.57452703e-01 -8.06322813e-01 -1.01145470e+00 1.70271802e+00
2.25238189e-01 -2.23765925e-01 7.27923274e-01 3.25738639e-01
8.19322288e-01 8.51919174e-01 7.69575536e-01 -1.43059462e-01
1.42676985e+00 -1.16465092e+00 -6.66851819e-01 -1.58547297e-01
1.20058060e+00 -1.22147036e+00 1.06628704e+00 3.45262468e-01
-1.07103205e+00 -6.19617224e-01 -9.00024235e-01 -3.05863202e-01
-3.44628692e-01 3.90870541e-01 3.11767966e-01 7.15913951e-01
-1.10150003e+00 5.37659466e-01 -8.71979058e-01 -5.21268845e-01
-3.36748898e-01 2.69274592e-01 -6.22862935e-01 -1.61848858e-01
-1.44779134e+00 1.03651047e+00 7.49945045e-01 3.36147361e-02
-6.09533548e-01 -2.92894274e-01 -1.07864940e+00 -2.91979611e-01
-8.53127912e-02 -2.08028600e-01 1.54089761e+00 -6.53762519e-01
-1.81731021e+00 1.05219793e+00 -2.53207207e-01 -4.26411331e-01
2.87067354e-01 -7.86530733e-01 -5.71859956e-01 -4.06474352e-01
3.33625436e-01 4.91729826e-01 2.31062397e-01 -8.02020967e-01
-1.03529668e+00 -2.02942602e-02 1.27757415e-01 1.15813494e-01
4.95165959e-02 8.62470686e-01 -6.40074670e-01 -4.57963169e-01
-2.33008832e-01 -1.12035644e+00 -3.45779061e-01 -1.23611569e+00
-6.33820593e-01 -6.52251244e-01 4.35724884e-01 -1.31488585e+00
1.23969579e+00 -1.98749065e+00 5.23642190e-02 -1.96479008e-01
-5.78371286e-01 2.87374049e-01 -3.02032620e-01 4.49593604e-01
-1.43321589e-01 4.61417019e-01 -3.53604168e-01 -7.12413490e-01
-2.21350178e-01 4.26444352e-01 -2.33600363e-01 6.54497370e-02
5.54793298e-01 8.17980349e-01 -9.91018295e-01 -6.77386761e-01
-4.43428531e-02 3.51037353e-01 -4.34166521e-01 5.90150356e-01
-2.71529913e-01 8.77901435e-01 -2.66212314e-01 5.27049124e-01
4.90321338e-01 5.02440274e-01 7.59132028e-01 1.31680258e-02
-6.37537003e-01 9.18290794e-01 -7.18343377e-01 2.00508428e+00
-1.06598628e+00 3.32034767e-01 9.10260656e-04 -7.02908337e-01
9.06374335e-01 6.04994297e-01 -9.48264971e-02 -5.16648173e-01
5.51261082e-02 6.24220967e-01 1.19883180e-01 -3.60127449e-01
4.31454778e-01 2.84717176e-02 -8.36462319e-01 3.37563068e-01
6.84783041e-01 -3.21707726e-01 4.87564355e-01 3.12817574e-01
1.00965893e+00 6.61615312e-01 5.39944470e-01 -4.55594569e-01
7.18679726e-01 5.79088693e-03 5.86264253e-01 3.33346277e-01
-8.23736042e-02 4.21329558e-01 5.97462237e-01 -2.78031826e-01
-8.85565162e-01 -1.18854916e+00 -1.22250192e-01 1.43229103e+00
-4.81893629e-01 -5.16164780e-01 -9.20526743e-01 -1.28646147e+00
-5.22234380e-01 8.60428870e-01 -3.11560631e-01 4.39885765e-01
-1.22969711e+00 -6.88099444e-01 8.71578395e-01 4.76144522e-01
2.29677334e-01 -1.47869647e+00 8.02951977e-02 5.94403386e-01
-5.34950018e-01 -1.73346925e+00 -3.94927204e-01 6.50469422e-01
-5.79199135e-01 -9.37702835e-01 -4.49651986e-01 -1.26468480e+00
2.29876786e-01 -6.14669919e-01 1.31333470e+00 -2.18156919e-01
1.90820754e-01 -2.16981500e-01 -6.93553507e-01 -3.90169799e-01
-8.79568458e-01 6.06823683e-01 -1.16548128e-01 -3.58369857e-01
2.80471295e-01 -2.03007177e-01 5.15022278e-02 -1.03595965e-01
-3.36769015e-01 -9.20283794e-02 6.20391965e-01 6.11426413e-01
5.91972470e-01 -6.80741489e-01 5.42098463e-01 -1.34778094e+00
5.41997433e-01 -4.55627233e-01 -5.09559453e-01 4.38726604e-01
-1.52175322e-01 2.45346278e-01 9.06676531e-01 -4.14745398e-02
-1.57971764e+00 3.38555276e-01 -8.00085962e-01 4.71220255e-01
-3.17152858e-01 4.37967062e-01 -5.71130693e-01 4.40432340e-01
3.71057391e-01 -1.41570196e-01 -8.17567408e-01 -8.30036640e-01
5.47020376e-01 8.47966254e-01 7.74545074e-01 -1.01950371e+00
3.66375327e-01 -3.37366611e-01 -4.37213629e-01 -7.07816780e-01
-1.30159140e+00 -2.35634133e-01 -1.22762263e+00 2.22188771e-01
1.33705604e+00 -1.24643338e+00 -8.22862014e-02 5.19356191e-01
-1.77083433e+00 -8.05656731e-01 -2.99423598e-02 6.33277476e-01
-2.53223658e-01 2.36566827e-01 -1.17871380e+00 -3.85863662e-01
-5.94505727e-01 -1.14443076e+00 8.82064104e-01 1.12102024e-01
-2.86263913e-01 -1.22906458e+00 5.90599000e-01 1.47288546e-01
-4.75820713e-03 2.41932105e-02 9.59570944e-01 -9.00231540e-01
2.68327203e-02 6.97626844e-02 -1.50396317e-01 7.12830782e-01
1.00164704e-01 1.95783809e-01 -8.51730883e-01 -7.64684528e-02
-6.45419478e-01 -5.67722857e-01 6.72448754e-01 2.07682475e-01
4.40476596e-01 4.15538885e-02 -1.11401290e-01 5.93744397e-01
1.20112562e+00 5.74363954e-02 3.20571303e-01 2.06231862e-01
8.10353994e-01 8.37653816e-01 8.11175764e-01 5.71215302e-02
7.17794895e-01 3.89252871e-01 1.48550078e-01 -1.25129089e-01
-2.91363716e-01 -3.40540528e-01 1.02262497e+00 1.42894292e+00
-2.31906325e-02 -3.71064663e-01 -1.01390219e+00 8.00539613e-01
-1.64847529e+00 -2.60649323e-01 -5.93632936e-01 1.93595994e+00
1.25562501e+00 2.90943652e-01 -1.71667755e-01 -5.00720918e-01
8.70497823e-01 -3.44315767e-02 9.41902474e-02 -1.13435853e+00
-1.28684625e-01 9.73434329e-01 4.89015549e-01 9.92425740e-01
-1.22705388e+00 2.02673745e+00 6.39217043e+00 5.72961092e-01
-9.80117798e-01 5.05371392e-01 4.91746008e-01 4.17780906e-01
-2.75446773e-02 4.16047186e-01 -1.30700660e+00 8.76183212e-02
1.61971688e+00 8.01245216e-03 -5.45791611e-02 5.15724599e-01
-5.95847704e-02 -9.77264121e-02 -1.08014405e+00 1.46097109e-01
-3.48021239e-01 -6.85929656e-01 2.84409653e-02 -3.82268250e-01
6.59790397e-01 6.06273293e-01 -4.39002067e-01 6.72640681e-01
1.21618617e+00 -9.01384711e-01 7.24884212e-01 2.24546678e-02
1.12019169e+00 -9.12720263e-01 8.11682940e-01 3.29740733e-01
-1.51300085e+00 4.72472578e-01 -2.95043498e-01 -3.78860570e-02
5.28838933e-01 5.83705842e-01 -8.64875913e-01 9.18330491e-01
6.60107076e-01 6.45140409e-01 -4.25259233e-01 3.79452825e-01
-1.11697423e+00 1.21660900e+00 -2.11028069e-01 1.59147844e-01
2.67373711e-01 -8.65202621e-02 2.76306927e-01 2.17806602e+00
2.86353707e-01 -2.06223473e-01 4.29771423e-01 -1.72640942e-02
-2.23413542e-01 6.61343098e-01 -3.26453328e-01 2.51549482e-01
4.27465826e-01 1.40892804e+00 -4.96817172e-01 -5.27695715e-01
-5.42382836e-01 1.25739110e+00 8.96428406e-01 1.79291025e-01
-7.57212341e-01 -6.32442117e-01 3.48707438e-01 -4.07222718e-01
2.01128915e-01 -4.80681568e-01 -3.22733745e-02 -1.23099947e+00
-1.68841660e-01 -6.15261972e-01 6.08250022e-01 -2.78017670e-01
-1.04034173e+00 1.15010333e+00 -1.27716675e-01 -7.41772830e-01
-4.49646592e-01 -6.60289466e-01 -8.40062678e-01 1.27520180e+00
-1.79760253e+00 -1.59924400e+00 4.57213789e-01 6.84174895e-01
7.48712778e-01 -7.16730282e-02 1.20149863e+00 3.30054820e-01
-7.94465899e-01 7.40727365e-01 -2.43842945e-01 5.53211629e-01
1.31316447e+00 -1.60135055e+00 1.04972124e+00 1.26349580e+00
1.85208663e-01 3.78726095e-01 5.13713539e-01 -6.90770388e-01
-6.70187056e-01 -1.39252663e+00 1.82479429e+00 -3.94093037e-01
1.10309267e+00 -4.96592045e-01 -6.67847812e-01 1.20221758e+00
6.70570254e-01 -5.19201607e-02 6.64776802e-01 4.39558715e-01
-4.61458236e-01 1.77427769e-01 -7.54504621e-01 3.50369930e-01
9.43445742e-01 -5.54025233e-01 -4.94787604e-01 3.37201744e-01
1.11872351e+00 -4.85602468e-01 -1.27474666e+00 3.16231072e-01
4.80345756e-01 -5.64392507e-01 3.95014912e-01 -7.46041238e-01
5.57821810e-01 9.34914798e-02 -3.34398866e-01 -1.69972813e+00
-1.25042498e-01 -6.23772562e-01 4.41815019e-01 1.67655146e+00
1.14193177e+00 -5.66041052e-01 1.72564134e-01 7.09896684e-02
-4.35851127e-01 -2.82817364e-01 -9.78662848e-01 -6.48258865e-01
6.80399299e-01 -5.88382900e-01 2.01381072e-01 9.54976201e-01
1.19537607e-01 8.45714867e-01 -2.47342929e-01 3.60362411e-01
5.31742156e-01 -2.16410771e-01 7.22912908e-01 -1.03052545e+00
-3.74398977e-01 1.49917588e-01 3.15465033e-01 -8.72484148e-01
6.87537551e-01 -1.07257688e+00 3.20694268e-01 -1.40827024e+00
-1.29727304e-01 -6.70797527e-01 -2.71714538e-01 6.97999239e-01
-3.46579134e-01 1.98780596e-01 3.51347566e-01 5.12033105e-02
-5.88343620e-01 -3.69849540e-02 8.73528838e-01 1.08003132e-01
-2.66434789e-01 3.65590528e-02 -5.26988089e-01 7.46819198e-01
1.02195120e+00 -7.49779999e-01 3.34919393e-01 -8.78743589e-01
4.60392935e-03 3.20542246e-01 -4.03481156e-01 -8.64405215e-01
-1.73064053e-01 -1.01446979e-01 -2.69622672e-02 -4.33006227e-01
-1.27987042e-01 -3.33641142e-01 -2.86039084e-01 5.90802252e-01
-1.36105835e-01 3.36158246e-01 2.78065234e-01 -2.28278294e-01
-3.60667050e-01 -4.23767000e-01 6.79574609e-01 -2.22668275e-01
-6.12378120e-01 1.25076190e-01 -4.26071674e-01 2.73061037e-01
7.68963397e-01 6.48339212e-01 -2.62118995e-01 1.54057771e-01
-9.94990230e-01 2.52856255e-01 -5.23367263e-02 4.49852049e-01
-1.10244013e-01 -8.42004776e-01 -1.29505920e+00 2.60039587e-02
-2.37502120e-02 1.63185541e-02 -1.21546552e-01 4.32971090e-01
-6.08872294e-01 5.65690994e-01 -3.72143053e-02 -2.46099502e-01
-1.25676906e+00 3.93527672e-02 1.43722817e-01 -1.22161758e+00
-2.66108751e-01 1.06338096e+00 1.78057328e-02 -9.55798149e-01
-1.67229965e-01 -4.15262967e-01 -4.15145457e-01 7.83735514e-03
3.62981200e-01 -4.97061983e-02 3.71369004e-01 -8.56879056e-01
-4.44433749e-01 3.73993218e-01 -3.32371771e-01 -5.03303945e-01
1.12678766e+00 6.27985178e-03 -3.24289083e-01 4.87745404e-01
1.15792906e+00 7.59853244e-01 -1.00796473e+00 -2.61557490e-01
4.80972081e-01 3.49524200e-01 -4.08820868e-01 -1.03608000e+00
-6.12969398e-01 9.57663178e-01 7.78252110e-02 -4.40505780e-02
6.78968072e-01 3.67706954e-01 9.33053792e-01 3.54503244e-01
3.25702637e-01 -1.12717545e+00 -4.61944133e-01 1.42549407e+00
6.27119839e-01 -1.22765088e+00 -3.48222792e-01 -5.41197538e-01
-6.27678931e-01 1.11248899e+00 7.57400393e-01 -1.96656749e-01
5.91177762e-01 6.21249080e-01 5.36915720e-01 3.19363117e-01
-7.82435775e-01 -3.48949224e-01 -7.10161179e-02 5.98617435e-01
1.25862443e+00 6.28593624e-01 -8.99705529e-01 8.81462991e-01
-5.12698948e-01 -5.01602292e-01 4.50809628e-01 7.86720216e-01
-3.94517362e-01 -1.90619099e+00 -1.26308396e-01 -1.86379939e-01
-1.14778268e+00 -6.47700131e-01 -3.16083074e-01 8.39991033e-01
1.28428936e-01 1.04387188e+00 1.46813774e-02 -3.62808377e-01
5.38488448e-01 3.36928874e-01 4.98983562e-01 -1.14666629e+00
-9.19549644e-01 7.90890157e-02 7.24114776e-01 -4.23449337e-01
-3.33048493e-01 -7.37520099e-01 -1.52111757e+00 1.41541272e-01
-1.46677405e-01 3.79298896e-01 8.41008127e-01 1.13578880e+00
-1.48799509e-01 5.51214635e-01 5.97772300e-01 -7.67963350e-01
-2.20288903e-01 -1.33205342e+00 -2.39431053e-01 1.23916343e-01
3.49024534e-02 5.94591089e-02 -7.53216371e-02 2.62263030e-01]
|
[10.477179527282715, 9.928075790405273]
|
9474ffb7-4556-4bfc-af86-f05eac618905
|
single-sequence-prediction-over-reasoning
|
2307.00335
| null |
https://arxiv.org/abs/2307.00335v1
|
https://arxiv.org/pdf/2307.00335v1.pdf
|
Single Sequence Prediction over Reasoning Graphs for Multi-hop QA
|
Recent generative approaches for multi-hop question answering (QA) utilize the fusion-in-decoder method~\cite{izacard-grave-2021-leveraging} to generate a single sequence output which includes both a final answer and a reasoning path taken to arrive at that answer, such as passage titles and key facts from those passages. While such models can lead to better interpretability and high quantitative scores, they often have difficulty accurately identifying the passages corresponding to key entities in the context, resulting in incorrect passage hops and a lack of faithfulness in the reasoning path. To address this, we propose a single-sequence prediction method over a local reasoning graph (\model)\footnote{Code/Models will be released at \url{https://github.com/gowtham1997/SeqGraph}} that integrates a graph structure connecting key entities in each context passage to relevant subsequent passages for each question. We use a graph neural network to encode this graph structure and fuse the resulting representations into the entity representations of the model. Our experiments show significant improvements in answer exact-match/F1 scores and faithfulness of grounding in the reasoning path on the HotpotQA dataset and achieve state-of-the-art numbers on the Musique dataset with only up to a 4\% increase in model parameters.
|
['Junjie Hu', 'Makesh Sreedhar', 'Gowtham Ramesh']
|
2023-07-01
| null | null | null | null |
['multi-hop-question-answering', 'question-answering']
|
['knowledge-base', 'natural-language-processing']
|
[ 1.82148427e-01 5.79211533e-01 6.48955777e-02 -4.70497251e-01
-1.53211093e+00 -8.54827404e-01 4.29892510e-01 5.95373154e-01
-1.27006575e-01 7.83676326e-01 6.68145537e-01 -5.65644741e-01
1.73674617e-02 -1.12361610e+00 -1.10242748e+00 7.08541125e-02
5.79405427e-01 9.27143455e-01 5.11618495e-01 -7.66378105e-01
1.41038597e-01 -3.87239307e-01 -1.11582172e+00 8.20858002e-01
1.09004343e+00 6.36456490e-01 7.99559131e-02 1.18695354e+00
-5.03131568e-01 1.28830445e+00 -7.48050451e-01 -1.22164047e+00
-2.01328948e-01 -9.53543723e-01 -1.35289180e+00 -6.01020277e-01
5.68340838e-01 -2.42724344e-01 -6.28835857e-01 9.10114646e-01
4.34335679e-01 2.91669428e-01 5.67185581e-01 -9.50124860e-01
-1.02665353e+00 1.04380131e+00 -9.47484002e-02 6.52604878e-01
1.04000270e+00 1.07904457e-01 1.74693573e+00 -7.63499558e-01
7.64586270e-01 1.05155480e+00 6.37599826e-01 5.46520114e-01
-9.59284008e-01 -2.89778709e-01 -2.53431555e-02 5.54537714e-01
-1.03910053e+00 -2.92186350e-01 4.78646040e-01 -8.92891511e-02
1.39271796e+00 4.28425223e-01 4.23651874e-01 9.66209173e-01
2.32767075e-01 8.03943217e-01 4.62690383e-01 -2.97604591e-01
-1.11036576e-01 -3.19323927e-01 4.63673025e-01 1.17879963e+00
1.65014230e-02 -5.39870977e-01 -5.70857644e-01 -1.61408290e-01
3.92844945e-01 -3.99790436e-01 -3.61081511e-01 3.29548329e-01
-1.06676161e+00 9.83261526e-01 7.47628272e-01 3.73576134e-02
-3.97384137e-01 3.26826632e-01 1.86925620e-01 4.17966455e-01
7.29444399e-02 6.05944633e-01 -3.58875245e-01 -2.44358122e-01
-6.63681746e-01 8.48974764e-01 1.14828336e+00 1.05102146e+00
7.62834191e-01 -2.96369225e-01 -6.95787430e-01 6.81277096e-01
2.61093915e-01 4.46447879e-01 1.54692650e-01 -1.10810232e+00
1.16977775e+00 9.23421860e-01 5.56485318e-02 -1.06802475e+00
-2.52137810e-01 -3.69032055e-01 -4.80064064e-01 -6.24895930e-01
4.24937516e-01 -1.59633085e-01 -7.54904151e-01 1.78525162e+00
3.37051511e-01 6.30276802e-04 2.00491890e-01 7.42362618e-01
1.36616588e+00 8.92827630e-01 1.09677628e-01 3.06053549e-01
1.45420790e+00 -1.32333708e+00 -6.29293740e-01 -4.14620161e-01
7.63418555e-01 -7.85044372e-01 1.25584114e+00 -1.02942392e-01
-1.32784808e+00 -6.19715035e-01 -9.23837245e-01 -6.90681934e-01
-1.12234652e-01 -1.72419280e-01 1.64679348e-01 2.19886318e-01
-9.95488167e-01 3.73339772e-01 -4.20048654e-01 -2.19861299e-01
4.86016124e-02 8.43346119e-04 4.75205630e-02 -4.30278182e-01
-1.68975413e+00 1.00391006e+00 4.88147557e-01 -1.11154117e-01
-6.36367917e-01 -8.60465825e-01 -9.68095541e-01 1.56832382e-01
5.02676666e-01 -1.30164063e+00 1.49947679e+00 -3.06497961e-01
-1.03052294e+00 5.99893272e-01 -4.86928135e-01 -6.16018951e-01
2.39724368e-01 -3.97393405e-01 -5.64942181e-01 3.17880988e-01
3.57076287e-01 6.79280162e-01 3.29522163e-01 -1.01777542e+00
-6.25742733e-01 3.41762267e-02 5.93502760e-01 2.80982703e-01
2.35170752e-01 -8.20734203e-02 -7.38780499e-01 -5.89937806e-01
7.49539807e-02 -8.89278114e-01 -2.44852044e-02 -5.81621051e-01
-6.82430327e-01 -4.06950146e-01 1.92650974e-01 -1.34491181e+00
1.68978190e+00 -1.58444357e+00 3.36731941e-01 -6.37555793e-02
2.21792519e-01 4.31335270e-02 -2.58665949e-01 8.13172400e-01
2.15596721e-01 2.43722335e-01 -2.73404896e-01 -9.03283283e-02
2.53852904e-01 3.08655739e-01 -5.85529029e-01 -2.15001971e-01
3.79866898e-01 1.47902203e+00 -1.07603633e+00 -4.77607399e-01
-3.22075099e-01 2.87335783e-01 -9.08706307e-01 3.53794158e-01
-7.99806476e-01 2.18225121e-01 -4.93334681e-01 4.31761712e-01
4.41155992e-02 -7.85358727e-01 6.17520288e-02 -1.49688944e-01
6.50589466e-01 9.67703998e-01 -9.05425251e-01 1.76181567e+00
-3.77718449e-01 3.39071453e-01 -4.80398357e-01 -4.26798612e-01
6.37422085e-01 4.66854036e-01 -3.40593308e-02 -7.34540999e-01
-6.61804378e-02 2.67092109e-01 -6.09254092e-02 -6.89682782e-01
8.97936404e-01 1.29577562e-01 -3.05331260e-01 4.19579417e-01
3.34394813e-01 -4.41559404e-01 5.74140131e-01 8.35847497e-01
1.34730661e+00 2.14225769e-01 3.55012231e-02 1.90883473e-01
4.94544655e-01 4.38612849e-01 3.26764762e-01 7.77999222e-01
2.77142644e-01 6.85912728e-01 6.06882334e-01 -2.60903895e-01
-1.07313931e+00 -1.08207083e+00 5.31671405e-01 1.02891302e+00
5.46733811e-02 -8.73615146e-01 -7.68583119e-01 -8.79797280e-01
-1.31084383e-01 1.44199133e+00 -5.39106369e-01 -3.22791010e-01
-9.33199227e-01 -1.90091401e-01 9.75460231e-01 6.28739059e-01
4.43799406e-01 -1.08026648e+00 -2.60509968e-01 2.74592906e-01
-1.16922390e+00 -1.19206309e+00 -5.25491416e-01 -1.77435711e-01
-5.56569874e-01 -1.09245014e+00 -5.00152409e-01 -7.14321494e-01
4.65004086e-01 -2.06177756e-01 1.74192333e+00 3.72268885e-01
1.85623363e-01 2.61265218e-01 -5.19423127e-01 -1.70625523e-01
-7.80396819e-01 2.17681438e-01 -5.61733603e-01 -3.75032485e-01
2.90689319e-01 -2.41822943e-01 -6.87222242e-01 6.41662702e-02
-8.58020425e-01 2.34089941e-01 2.39796326e-01 7.89258540e-01
6.97529137e-01 -2.16283545e-01 6.62568808e-01 -1.03424692e+00
9.20067191e-01 -7.51240849e-01 -2.33115315e-01 6.54064655e-01
-4.00897682e-01 3.42176378e-01 7.12099910e-01 9.37990025e-02
-1.10202050e+00 -6.41800702e-01 -6.66962445e-01 -9.46398750e-02
1.80418417e-01 9.25274253e-01 1.95941567e-01 4.94983464e-01
9.15187180e-01 1.64331734e-01 -3.76341730e-01 -2.49168679e-01
8.35946262e-01 2.23969817e-01 6.70323253e-01 -7.90713608e-01
7.01189220e-01 -1.47080138e-01 -2.54728168e-01 -1.92278847e-01
-1.41446412e+00 -4.17275995e-01 -3.38089645e-01 -1.64965063e-01
9.76811290e-01 -8.73056173e-01 -4.69087243e-01 -1.31411150e-01
-1.33063734e+00 -2.38199458e-01 -3.52429599e-01 7.22190440e-02
-5.54708600e-01 3.96481037e-01 -8.70416045e-01 -4.09879029e-01
-6.01237237e-01 -9.74356949e-01 8.51839960e-01 4.55247968e-01
-6.30761743e-01 -1.10601580e+00 1.94270387e-01 1.11725116e+00
1.73274353e-01 5.85671850e-02 1.57087076e+00 -9.01674569e-01
-7.26032674e-01 -2.59133369e-01 -9.27771181e-02 2.19060585e-01
-2.25894764e-01 -1.68696225e-01 -6.74425900e-01 -3.88483480e-02
-2.87605613e-01 -4.91258383e-01 8.31167161e-01 -1.37638971e-01
6.88706160e-01 -6.31471515e-01 -5.95746525e-02 1.71437725e-01
1.40643907e+00 -8.21609795e-02 7.18695045e-01 2.89771724e-02
7.96644330e-01 4.29788888e-01 3.59220713e-01 7.46424571e-02
9.31888759e-01 5.73174179e-01 2.49504477e-01 3.28279197e-01
-6.51356995e-01 -8.89701426e-01 3.11501026e-01 1.18171251e+00
9.39656515e-03 -6.59780204e-01 -9.55866814e-01 8.37665558e-01
-1.85917997e+00 -1.07179415e+00 -5.15747130e-01 1.84232891e+00
9.15206432e-01 1.44825622e-01 -3.90428901e-02 -1.99602231e-01
4.53635335e-01 1.26430064e-01 -3.76902789e-01 -4.07638550e-01
-1.49889544e-01 4.36056018e-01 3.63476612e-02 8.81310403e-01
-5.62416971e-01 1.01873565e+00 5.81815004e+00 7.29473829e-01
-4.26926941e-01 1.78870872e-01 4.15443391e-01 1.21204898e-01
-9.53071475e-01 1.95589870e-01 -1.01532471e+00 3.12062919e-01
1.30640686e+00 -3.28789353e-01 3.94447923e-01 3.91310394e-01
-2.89480686e-01 -2.50805598e-02 -9.88258719e-01 4.64515954e-01
2.34953284e-01 -1.71527255e+00 5.44401824e-01 -4.56492782e-01
7.18374431e-01 6.18812740e-02 -2.13139743e-01 6.55518711e-01
5.73648751e-01 -9.60362077e-01 7.68004894e-01 6.95430338e-01
5.86251974e-01 -7.08268404e-01 8.15464377e-01 3.42604667e-01
-1.14504206e+00 -1.21967802e-02 -2.59492338e-01 4.43965085e-02
5.04459858e-01 3.13858896e-01 -1.07506204e+00 1.06399584e+00
5.79504788e-01 1.67822152e-01 -6.54400468e-01 8.10613692e-01
-9.25820470e-01 8.72618496e-01 -5.20356484e-02 -4.58264351e-01
3.22391868e-01 6.26905784e-02 6.70755148e-01 1.17379200e+00
3.12112600e-01 3.48873883e-01 -1.02084450e-01 1.02941108e+00
-4.34765041e-01 4.86681499e-02 -3.19660634e-01 -2.11523354e-01
4.07972932e-01 8.80274594e-01 -4.37927693e-01 -5.25979221e-01
-4.99044240e-01 1.10007179e+00 6.26201868e-01 5.10716617e-01
-1.07587016e+00 -4.88148123e-01 3.76733631e-01 -7.75086507e-02
4.11954880e-01 -4.70183752e-02 -2.11784944e-01 -1.28670514e+00
2.20688075e-01 -1.14442182e+00 8.99880052e-01 -1.10344017e+00
-1.18806183e+00 6.96702838e-01 -4.79528755e-02 -6.97953224e-01
-5.14290869e-01 -2.41337866e-01 -5.44584930e-01 1.16131151e+00
-1.31632543e+00 -1.21829736e+00 -2.67395228e-02 6.00977063e-01
7.88873017e-01 1.00037821e-01 9.76995468e-01 2.80987233e-01
-3.46804231e-01 6.68382704e-01 -3.71925592e-01 3.77078831e-01
4.07587528e-01 -1.40416396e+00 9.16206241e-01 1.07285273e+00
6.92200780e-01 7.14776456e-01 7.62024105e-01 -8.39479089e-01
-1.25942135e+00 -9.22867954e-01 1.57724595e+00 -1.02872908e+00
7.40495682e-01 -1.01611048e-01 -1.10000229e+00 9.79959488e-01
4.28190142e-01 -4.80941504e-01 6.62507653e-01 2.09691077e-01
-5.06929338e-01 3.46218377e-01 -8.20673108e-01 7.81632781e-01
9.14876044e-01 -8.14981997e-01 -1.21050680e+00 3.96443039e-01
1.23644710e+00 -7.93224633e-01 -9.11886334e-01 3.27046871e-01
1.43750951e-01 -7.50312388e-01 9.14881349e-01 -9.44588363e-01
9.08873200e-01 -4.19219226e-01 -2.70630538e-01 -1.16095126e+00
-4.27725583e-01 -6.05053127e-01 -5.88706851e-01 1.26207209e+00
1.18968797e+00 -2.32660770e-01 5.91194808e-01 5.06607056e-01
-3.26912344e-01 -1.00865376e+00 -8.64644408e-01 -2.48779520e-01
1.21542297e-01 -4.45492387e-01 6.59567475e-01 6.07390225e-01
8.26709047e-02 9.76036608e-01 -1.79226436e-02 3.52338582e-01
3.36473346e-01 2.29964986e-01 5.42259514e-01 -6.77238643e-01
-4.57838774e-01 -2.58675337e-01 2.52151694e-02 -1.35436344e+00
-5.13819344e-02 -1.21854603e+00 5.97570799e-02 -2.40495420e+00
1.91469435e-02 -1.56675056e-01 -1.09986417e-01 4.06630367e-01
-7.28666782e-01 3.30295563e-02 1.18059434e-01 5.18416762e-02
-8.32147658e-01 4.39672947e-01 1.31759059e+00 -2.22079486e-01
1.24568582e-01 -6.69079646e-02 -9.30956662e-01 4.04046208e-01
5.32691061e-01 -4.41918701e-01 -6.11956298e-01 -9.42930460e-01
8.71130586e-01 4.93484378e-01 3.79667521e-01 -1.00230706e+00
4.89281148e-01 8.71997401e-02 2.32486784e-01 -5.75986087e-01
4.58907992e-01 -2.62259454e-01 1.32827535e-01 2.93847382e-01
-6.54816866e-01 5.63195288e-01 1.69303998e-01 7.15677440e-01
-3.15941125e-01 -2.88158685e-01 2.82237768e-01 -3.37610990e-01
-7.00885832e-01 1.54794296e-02 -1.02541065e-02 7.33476639e-01
4.75504994e-01 1.48215368e-01 -8.53049815e-01 -7.68546700e-01
-6.20719790e-01 4.84644800e-01 1.54769182e-01 5.58740020e-01
6.43379629e-01 -1.14099395e+00 -1.06061280e+00 -2.98533201e-01
1.75147220e-01 6.24761991e-02 5.97403288e-01 6.68331861e-01
-6.85056627e-01 5.67270279e-01 1.78290933e-01 -1.78416163e-01
-1.03546250e+00 3.61023068e-01 3.12329948e-01 -6.64531469e-01
-5.96024692e-01 1.36414278e+00 -2.54264504e-01 -6.54527009e-01
-1.48937836e-01 -4.77637917e-01 -2.58921146e-01 -1.11061335e-01
3.97269249e-01 2.73951262e-01 2.03352451e-01 -5.61014831e-01
-1.58953935e-01 2.98959315e-01 -8.85424167e-02 -2.12657735e-01
9.93891597e-01 -1.06509730e-01 -1.83626771e-01 3.33297700e-01
9.83518541e-01 1.59610957e-01 -8.87720585e-01 -3.76856595e-01
1.03885591e-01 -5.00527248e-02 -4.18014377e-01 -1.18702722e+00
-6.01415098e-01 6.87553167e-01 -2.20602185e-01 2.90783614e-01
7.70188510e-01 3.61489087e-01 1.33155239e+00 5.69717467e-01
2.29317009e-01 -6.98927581e-01 2.57129103e-01 9.02107298e-01
1.02784956e+00 -9.55144465e-01 -2.75465190e-01 -4.32770222e-01
-9.46987629e-01 8.22516739e-01 5.91998577e-01 9.09755230e-02
2.04449892e-01 -2.43447348e-01 1.25428170e-01 -4.59468633e-01
-9.91326153e-01 -2.23790959e-01 6.14418745e-01 2.55369037e-01
4.50545579e-01 -3.36991400e-02 -1.87134460e-01 6.72406673e-01
-6.99223578e-01 -1.88400537e-01 4.01016623e-01 8.05165052e-01
-3.80024344e-01 -9.97731984e-01 8.86426046e-02 5.05778790e-01
-6.13292217e-01 -4.90919977e-01 -4.87678647e-01 5.14268756e-01
-1.54591709e-01 1.32934284e+00 -1.09522715e-01 -5.37075579e-01
5.77833951e-01 5.29996634e-01 4.71843392e-01 -7.21352398e-01
-8.75244677e-01 -5.30448556e-01 7.69606411e-01 -4.16263670e-01
-6.05573021e-02 -3.49314094e-01 -1.64768708e+00 -4.94398654e-01
-1.32225975e-01 5.39913356e-01 1.30245343e-01 1.05485523e+00
5.23672223e-01 8.34095597e-01 -2.59843934e-03 2.00403243e-01
-3.19531471e-01 -9.26596224e-01 1.03938609e-01 5.37387729e-01
1.34597942e-01 -1.09515451e-01 -5.89841083e-02 1.72431037e-01]
|
[11.0382661819458, 7.969152927398682]
|
768de6e3-6b38-4807-bb40-cd85928a4937
|
deep-speaker-vectors-for-semi-text
|
1505.06427
| null |
http://arxiv.org/abs/1505.06427v1
|
http://arxiv.org/pdf/1505.06427v1.pdf
|
Deep Speaker Vectors for Semi Text-independent Speaker Verification
|
Recent research shows that deep neural networks (DNNs) can be used to extract
deep speaker vectors (d-vectors) that preserve speaker characteristics and can
be used in speaker verification. This new method has been tested on
text-dependent speaker verification tasks, and improvement was reported when
combined with the conventional i-vector method.
This paper extends the d-vector approach to semi text-independent speaker
verification tasks, i.e., the text of the speech is in a limited set of short
phrases. We explore various settings of the DNN structure used for d-vector
extraction, and present a phone-dependent training which employs the posterior
features obtained from an ASR system. The experimental results show that it is
possible to apply d-vectors on semi text-independent speaker recognition, and
the phone-dependent training improves system performance.
|
['Thomas Fang Zheng', 'Lantian Li', 'Zhiyong Zhang', 'Dong Wang']
|
2015-05-24
| null | null | null | null |
['text-independent-speaker-recognition', 'text-independent-speaker-verification', 'text-dependent-speaker-verification']
|
['speech', 'speech', 'speech']
|
[ 1.69976249e-01 -2.61840552e-01 -3.01753134e-01 -8.81014884e-01
-7.16836631e-01 -6.12498999e-01 5.65717697e-01 -4.37208802e-01
-3.87771755e-01 5.14704466e-01 6.64414346e-01 -5.93037844e-01
2.08502546e-01 -1.83865100e-01 -3.59620363e-01 -1.02094233e+00
1.10919122e-02 2.47410297e-01 -2.71750957e-01 -3.12490612e-01
1.20950997e-01 9.33546185e-01 -1.58788097e+00 9.27990600e-02
3.65632147e-01 7.39735126e-01 2.39971001e-02 8.70244503e-01
-2.78224975e-01 3.21936756e-01 -1.09919488e+00 -6.62834570e-02
1.19243875e-01 -4.25286800e-01 -7.06710458e-01 -1.26658127e-01
5.77495694e-01 -5.92654526e-01 -5.30052423e-01 1.02092850e+00
1.02467179e+00 2.05669284e-01 7.31324673e-01 -1.09285259e+00
-7.18038023e-01 1.10768306e+00 -9.69440788e-02 5.16480088e-01
1.88649863e-01 -3.61595571e-01 7.74094343e-01 -1.07548177e+00
2.57683724e-01 1.38987255e+00 7.13918686e-01 1.08693588e+00
-1.13541698e+00 -8.98472726e-01 -1.29341502e-02 2.94367880e-01
-1.63709116e+00 -1.17174721e+00 1.09657443e+00 -1.57476261e-01
1.26350641e+00 3.31619412e-01 4.42724936e-02 1.66325486e+00
-4.38065380e-02 1.10439229e+00 6.48312271e-01 -6.36114657e-01
1.90439105e-01 4.24912870e-01 5.66623867e-01 3.12262595e-01
-1.23699009e-01 4.66432124e-01 -8.82143319e-01 -2.03313045e-02
4.10411835e-01 -3.45658243e-01 -4.27899212e-01 -4.67430316e-02
-1.14763486e+00 1.04376936e+00 -9.58590955e-02 8.12918901e-01
-3.28589380e-01 -2.23054066e-01 7.21708953e-01 5.38942397e-01
4.58202899e-01 -1.32518858e-01 -4.31907713e-01 -2.74175197e-01
-1.44238305e+00 -1.45816311e-01 9.24275637e-01 8.45323086e-01
3.22015077e-01 1.02328420e+00 -3.11025172e-01 1.12711883e+00
4.06960219e-01 9.22608256e-01 1.05922866e+00 -3.24136049e-01
4.12721723e-01 -1.72034081e-03 -1.41149521e-01 -4.43481684e-01
-7.15555996e-02 -4.31280762e-01 -9.28402126e-01 6.82414174e-02
9.88963917e-02 -4.50542003e-01 -1.26298571e+00 1.91112077e+00
3.51787312e-03 5.14812581e-02 4.87562448e-01 6.00627244e-01
1.17745376e+00 8.91638696e-01 -3.34273040e-01 -1.74622566e-01
1.02907228e+00 -7.22960532e-01 -1.16201735e+00 -1.63823247e-01
4.99922812e-01 -7.41600692e-01 6.77584827e-01 2.62427747e-01
-7.49784827e-01 -7.36100256e-01 -1.25111103e+00 1.84708774e-01
-4.69028801e-01 1.59044236e-01 -1.19787175e-03 1.28700197e+00
-1.50105834e+00 2.07652539e-01 -6.94753110e-01 -2.46847630e-01
1.89583078e-01 6.19579792e-01 -5.24106324e-01 1.37755394e-01
-1.40273476e+00 9.36271548e-01 3.56756628e-01 3.23912680e-01
-1.15578556e+00 -2.29341418e-01 -1.19148588e+00 2.36793965e-01
-2.06289247e-01 -1.36156723e-01 1.27680492e+00 -9.88204241e-01
-1.89420211e+00 6.49217188e-01 -8.74533713e-01 -5.71607113e-01
6.96175173e-02 8.21821243e-02 -7.88367331e-01 9.35223475e-02
-3.06069314e-01 3.83089155e-01 1.36059189e+00 -1.06863189e+00
-3.65246505e-01 -4.57879424e-01 -6.56833410e-01 1.04698807e-01
-7.61532307e-01 4.61113036e-01 -1.08899817e-01 -5.34326255e-01
1.00777514e-01 -8.05868626e-01 2.43331745e-01 -4.31478083e-01
-8.17166626e-01 -4.73904103e-01 1.12109852e+00 -1.05091548e+00
1.11349547e+00 -2.41290402e+00 1.91216692e-01 2.55439520e-01
-1.06004536e-01 7.22744286e-01 -2.52199411e-01 2.45474458e-01
-4.04924154e-01 1.55217484e-01 -1.75348312e-01 -9.81182218e-01
1.90211833e-01 3.20428073e-01 -4.40761626e-01 5.44898331e-01
-1.22799918e-01 7.58847713e-01 3.91165204e-02 -5.04171014e-01
2.90418535e-01 8.39945018e-01 9.71935615e-02 1.08780995e-01
4.02784526e-01 5.30836359e-02 9.65196639e-02 5.83141625e-01
8.81443560e-01 4.85817134e-01 -8.43110606e-02 6.54317662e-02
-3.47071588e-02 6.47026122e-01 -1.04223502e+00 1.36839807e+00
-3.09110880e-01 1.24469006e+00 5.85277736e-01 -9.86148715e-01
1.13704503e+00 8.73504460e-01 2.22336724e-02 -2.28511557e-01
3.53152186e-01 6.76452834e-03 6.74705654e-02 -3.44106525e-01
4.74813044e-01 -5.08442223e-01 1.64326414e-01 5.27885139e-01
4.03258234e-01 1.29053704e-02 -1.50528088e-01 -1.36334375e-01
4.95476395e-01 -4.81404305e-01 5.38451448e-02 -2.42441684e-01
7.94644773e-01 -4.55822885e-01 4.12123859e-01 7.17384994e-01
-4.67250109e-01 5.29682875e-01 -1.35369822e-02 -4.56744060e-02
-8.85703206e-01 -9.48667407e-01 -4.43598688e-01 1.14482725e+00
-2.80462891e-01 1.34042665e-01 -6.69171214e-01 -6.91166401e-01
-4.25830558e-02 1.09273434e+00 -5.63020229e-01 -7.89511353e-02
-6.63088024e-01 -5.54248929e-01 1.08870828e+00 7.91664481e-01
2.23397553e-01 -9.86789584e-01 3.29762518e-01 2.14428946e-01
-1.18624732e-01 -9.16912556e-01 -7.91946709e-01 5.23720562e-01
-8.38230848e-01 -2.34916687e-01 -1.09575510e+00 -1.31042397e+00
3.22937787e-01 2.76378483e-01 5.73770642e-01 -3.02412242e-01
4.45651561e-01 1.95396543e-01 -1.89857692e-01 -4.77306217e-01
-1.00189841e+00 1.95287138e-01 6.49304569e-01 1.85863793e-01
8.16608429e-01 -5.12619436e-01 8.75538960e-02 2.86053985e-01
-7.17126548e-01 -6.68427587e-01 2.49958768e-01 9.57652330e-01
1.07136920e-01 1.83542043e-01 8.70368421e-01 -1.88821182e-01
7.80158281e-01 -1.40580535e-02 -3.04334581e-01 1.50020197e-01
-4.57733393e-01 3.16434920e-01 4.57880646e-01 -7.21776664e-01
-1.14199841e+00 8.25318918e-02 -8.72830331e-01 -6.13027275e-01
-4.91083324e-01 4.80769157e-01 -5.98162115e-01 -1.30388305e-01
6.06867075e-01 8.83132398e-01 3.62490654e-01 -6.02895439e-01
1.33185193e-01 1.50679886e+00 3.82520765e-01 1.16650015e-01
7.46265411e-01 2.04248875e-01 -6.37209594e-01 -1.49766338e+00
-2.59667277e-01 -6.73256099e-01 -7.75113165e-01 7.05222860e-02
6.36250973e-01 -8.76071930e-01 -5.84998965e-01 7.31435955e-01
-1.18451250e+00 -5.14062755e-02 -1.85377136e-01 9.25336838e-01
-1.43992350e-01 6.74001813e-01 -6.41935825e-01 -1.02734351e+00
-6.57242894e-01 -1.21227801e+00 9.53672886e-01 1.75034463e-01
-1.46393165e-01 -1.14134359e+00 1.68437228e-01 1.48671418e-01
6.71343744e-01 -6.82722569e-01 7.14558184e-01 -1.47752857e+00
4.84765656e-02 -4.95288491e-01 2.60967940e-01 1.14061999e+00
2.88961738e-01 -6.96918964e-02 -1.60596025e+00 -4.33642060e-01
4.28684413e-01 4.09691669e-02 8.92252982e-01 6.30397260e-01
9.02380705e-01 -4.35978740e-01 -3.39277655e-01 5.93879998e-01
9.51780975e-01 3.38851988e-01 4.82248455e-01 -5.53024784e-02
5.40362835e-01 3.99830759e-01 -1.80921584e-01 -9.16662440e-02
5.24537005e-02 6.23990238e-01 -1.27267921e-02 -2.49165334e-02
-2.51738638e-01 -1.02964595e-01 7.87707925e-01 1.16656899e+00
4.38619882e-01 -7.11285532e-01 -7.53878176e-01 8.40998530e-01
-1.21610808e+00 -1.12341392e+00 2.05698803e-01 2.07138062e+00
6.81248069e-01 3.88564616e-02 1.40684769e-01 6.42697453e-01
1.15359259e+00 3.32357734e-01 -6.25174999e-01 -7.20121205e-01
-3.82172823e-01 2.40892544e-01 3.83573532e-01 7.57166743e-01
-1.01721156e+00 7.30804920e-01 7.29716253e+00 8.20152938e-01
-1.60957444e+00 2.05443785e-01 1.24250166e-02 -9.08828825e-02
-2.22386762e-01 -6.49417758e-01 -1.18684876e+00 1.67809904e-01
1.57742667e+00 -1.90579236e-01 -2.63083894e-02 8.91757965e-01
1.99817061e-01 5.10120094e-01 -1.29128921e+00 1.23628330e+00
6.42883778e-01 -1.09472930e+00 1.14580885e-01 1.59016877e-01
3.11247289e-01 2.50216305e-01 1.99288279e-01 4.08282548e-01
-4.92231036e-03 -9.77812290e-01 6.95216119e-01 -5.84319197e-02
8.46619010e-01 -8.55726063e-01 1.13428295e+00 3.46881270e-01
-9.72404122e-01 -1.19704083e-02 -2.52445757e-01 3.50881547e-01
3.55570704e-01 1.79083511e-01 -1.35116673e+00 3.07463735e-01
4.15597230e-01 6.48556232e-01 -2.24900797e-01 7.67539501e-01
-8.50079060e-02 1.12875271e+00 -3.59632552e-01 -4.12882954e-01
1.11956321e-01 2.50148624e-01 9.49297428e-01 1.53036058e+00
3.10891956e-01 -4.57641423e-01 -3.88661712e-01 6.46463096e-01
-8.83082971e-02 3.82603519e-02 -6.77922308e-01 -2.24062487e-01
4.54721004e-01 7.08658040e-01 -2.82336146e-01 -4.51060623e-01
-8.71195830e-03 1.05533719e+00 -1.08756386e-01 6.40453398e-01
-4.75525409e-01 -8.03636253e-01 8.99160683e-01 -3.66235793e-01
8.19338322e-01 -3.52734506e-01 -3.00021589e-01 -1.09614968e+00
1.27639771e-02 -9.39262033e-01 -4.95445728e-02 -5.03690481e-01
-1.28896987e+00 1.08958995e+00 -6.49506003e-02 -9.43541944e-01
-7.30199456e-01 -6.15736723e-01 -6.60740972e-01 1.32872772e+00
-1.56131637e+00 -9.37987208e-01 4.32641804e-01 7.65696764e-01
6.35288060e-01 -6.47738338e-01 1.16228390e+00 3.02322149e-01
-6.65067136e-01 1.12739658e+00 5.84129035e-01 6.61508441e-01
5.13828933e-01 -1.07047009e+00 7.06853986e-01 1.06815791e+00
4.88528848e-01 7.75865853e-01 7.00661600e-01 -2.83476025e-01
-1.25298965e+00 -7.08725631e-01 1.29813480e+00 -3.68317246e-01
1.55199274e-01 -7.29779840e-01 -9.31577981e-01 7.83213913e-01
3.62130582e-01 -4.60766017e-01 1.07056940e+00 1.80947185e-01
-3.85548353e-01 -2.37014174e-01 -1.27723253e+00 2.13028267e-01
5.38168609e-01 -1.13168740e+00 -1.17685759e+00 9.91639644e-02
7.05140710e-01 -1.03352800e-01 -4.02825296e-01 3.82297225e-02
4.88283873e-01 -7.01160431e-01 1.11645675e+00 -5.83922744e-01
-3.87556583e-01 -8.75738263e-02 -3.81577641e-01 -1.42508054e+00
-8.71242434e-02 -4.40929055e-01 -1.16366014e-01 1.48029268e+00
6.64964318e-01 -8.14802647e-01 6.49064064e-01 2.99059868e-01
-2.44062394e-01 7.99311623e-02 -1.60279822e+00 -1.01316857e+00
7.61817098e-02 -4.52843249e-01 7.87355423e-01 8.74059796e-01
-1.40154332e-01 5.41737139e-01 -4.53579962e-01 4.31163073e-01
5.38449764e-01 -2.78438665e-02 3.67946178e-01 -1.14947879e+00
1.44661993e-01 -4.01259094e-01 -5.45911610e-01 -1.37815499e+00
8.21544349e-01 -9.50485229e-01 4.42683399e-01 -1.28895116e+00
-2.88168222e-01 1.41971791e-02 -3.21481019e-01 4.56769317e-01
7.74735585e-02 -1.37696881e-02 -3.61080430e-02 -7.08104745e-02
2.23870099e-01 8.21207345e-01 5.59793174e-01 -5.53200424e-01
-4.33659971e-01 5.70957482e-01 -5.23552716e-01 2.64967054e-01
1.01883960e+00 -4.84530866e-01 -1.75599888e-01 -2.54852831e-01
-7.21616805e-01 9.83074307e-02 -1.42250657e-01 -7.74773657e-01
1.98392898e-01 3.29718679e-01 3.59939605e-01 -1.04697478e+00
7.65881658e-01 -7.32324004e-01 -3.62980157e-01 3.45234275e-01
-5.42347789e-01 -3.11105788e-01 5.11143565e-01 3.48678082e-01
-5.17814398e-01 -5.20313084e-01 7.91021526e-01 3.53218496e-01
-4.74152505e-01 5.21488301e-02 -9.52337623e-01 -4.98405010e-01
4.36761498e-01 -4.06053603e-01 1.76422045e-01 -6.48819804e-01
-6.82138324e-01 -3.83653462e-01 -6.19154647e-02 5.37607193e-01
8.93961966e-01 -1.28273201e+00 -1.01464427e+00 7.00411201e-01
-5.42756729e-02 -5.01343668e-01 1.18812338e-01 2.23637402e-01
-1.35101229e-01 7.84728706e-01 -4.26770337e-02 -8.55365634e-01
-1.85718942e+00 5.33156753e-01 5.98686457e-01 3.66461039e-01
-4.33336973e-01 1.19813943e+00 -1.44995987e-01 -5.61710477e-01
8.49763215e-01 -3.71590972e-01 -2.51549125e-01 1.53864086e-01
9.23066199e-01 -2.64278203e-02 3.19780529e-01 -1.17510521e+00
-6.36159003e-01 2.09575295e-01 -3.86611134e-01 -6.35664880e-01
1.42579973e+00 -4.00148630e-01 7.76552260e-02 6.25168562e-01
1.47151387e+00 1.15473427e-01 -6.69852495e-01 -4.89761174e-01
-2.46927649e-01 1.58781558e-02 5.71533740e-01 -5.03632486e-01
-1.04925323e+00 1.27498782e+00 8.10653210e-01 2.75172889e-01
8.86809409e-01 -1.20860942e-01 7.82358050e-01 5.84163129e-01
-1.60575792e-01 -8.81219208e-01 -5.61399341e-01 6.08431697e-01
8.45607221e-01 -1.17618489e+00 -4.31958795e-01 2.25300565e-01
-5.56815684e-01 1.28371906e+00 1.39766425e-01 3.89820606e-01
1.02076828e+00 3.40481162e-01 5.11605084e-01 9.76077691e-02
-3.67206156e-01 1.62872467e-02 4.13176000e-01 8.49596024e-01
3.09085429e-01 1.83207676e-01 3.58560205e-01 3.92072082e-01
-4.40911114e-01 -3.22208166e-01 5.02053380e-01 9.54795480e-01
-4.11838084e-01 -1.10768616e+00 -7.41902232e-01 2.30268389e-01
-3.65547270e-01 -3.75450790e-01 -6.24972761e-01 4.03934658e-01
-3.99388075e-01 1.26273286e+00 -2.66088471e-02 -6.28418326e-01
8.00682679e-02 7.87297010e-01 7.05336854e-02 -3.92733246e-01
-6.45536542e-01 7.52945617e-02 1.85438037e-01 -1.22039996e-01
-4.82361645e-01 -7.17596829e-01 -9.48411644e-01 -4.37190384e-01
-8.64344120e-01 3.59457970e-01 1.45145941e+00 9.85452294e-01
2.40643233e-01 2.85063654e-01 8.44240308e-01 -8.57057273e-01
-8.45488787e-01 -1.39411294e+00 -8.31698298e-01 -7.75966644e-02
1.28170085e+00 -2.86284655e-01 -8.31287265e-01 9.05842781e-02]
|
[14.338537216186523, 6.089420795440674]
|
f9903160-bcaf-4202-90b5-0fa7f346d8b0
|
elastic-weight-removal-for-faithful-and
|
2303.17574
| null |
https://arxiv.org/abs/2303.17574v1
|
https://arxiv.org/pdf/2303.17574v1.pdf
|
Elastic Weight Removal for Faithful and Abstractive Dialogue Generation
|
Ideally, dialogue systems should generate responses that are faithful to the knowledge contained in relevant documents. However, many models generate hallucinated responses instead that contradict it or contain unverifiable information. To mitigate such undesirable behaviour, it has been proposed to fine-tune a `negative expert' on negative examples and subtract its parameters from those of a pre-trained model. However, intuitively, this does not take into account that some parameters are more responsible than others in causing hallucinations. Thus, we propose to weigh their individual importance via (an approximation of) the Fisher Information matrix, which measures the uncertainty of their estimate. We call this method Elastic Weight Removal (EWR). We evaluate our method -- using different variants of Flan-T5 as a backbone language model -- on multiple datasets for information-seeking dialogue generation and compare our method with state-of-the-art techniques for faithfulness, such as CTRL, Quark, DExperts, and Noisy Channel reranking. Extensive automatic and human evaluation shows that EWR systematically increases faithfulness at minor costs in terms of other metrics. However, we notice that only discouraging hallucinations may increase extractiveness, i.e. shallow copy-pasting of document spans, which can be undesirable. Hence, as a second main contribution, we show that our method can be extended to simultaneously discourage hallucinations and extractive responses. We publicly release the code for reproducing EWR and all baselines.
|
['Edoardo M. Ponti', 'Iryna Gurevych', 'Mrinmaya Sachan', 'Nouha Dziri', 'Nico Daheim']
|
2023-03-30
| null | null | null | null |
['dialogue-generation', 'dialogue-generation']
|
['natural-language-processing', 'speech']
|
[ 2.21711904e-01 6.82267368e-01 1.03973448e-01 -2.83122689e-01
-8.23413730e-01 -6.60416186e-01 8.76570463e-01 1.88755646e-01
-5.46612620e-01 9.72626567e-01 9.41043198e-01 -1.48542494e-01
-9.31320339e-02 -5.45815527e-01 -2.53265709e-01 -6.16363347e-01
3.38971943e-01 6.94312513e-01 8.84006023e-02 -7.90130973e-01
5.51617563e-01 2.18873382e-01 -9.71138179e-01 5.31698227e-01
8.55377734e-01 6.17243826e-01 -1.86198741e-01 6.57047570e-01
-1.16352551e-02 1.17244864e+00 -9.26568747e-01 -9.50445354e-01
-1.20830573e-01 -7.00575948e-01 -1.14370263e+00 8.84880126e-02
8.62949789e-02 -4.33667660e-01 -1.66920736e-01 9.91207063e-01
6.81689382e-01 3.45904291e-01 9.25296605e-01 -1.08133912e+00
-7.74468422e-01 9.86581922e-01 -3.92700136e-01 -1.61348507e-02
6.30303025e-01 2.67156392e-01 1.14190698e+00 -8.40793908e-01
5.33414900e-01 1.38552797e+00 6.36970818e-01 9.11115706e-01
-1.52073145e+00 -3.25468808e-01 -1.12273775e-01 3.11220344e-03
-9.06054020e-01 -7.96595693e-01 6.84026539e-01 -2.27577940e-01
1.12509573e+00 6.72983527e-01 4.86107290e-01 1.58790386e+00
-1.44807205e-01 9.78754699e-01 1.19064546e+00 -3.87485653e-01
2.10774839e-01 5.83036423e-01 7.48018324e-02 4.92870539e-01
7.41622448e-02 -1.27083331e-01 -7.98124552e-01 -6.80264711e-01
2.79726714e-01 -5.16958058e-01 -4.80087936e-01 -1.94962159e-01
-1.10643411e+00 1.19054949e+00 1.82667747e-01 1.86556309e-01
-4.36675429e-01 -6.77095130e-02 3.27415287e-01 5.55556655e-01
4.85353589e-01 9.41969335e-01 -2.36223847e-01 -2.96552241e-01
-7.25077987e-01 4.93173897e-01 1.21231496e+00 6.13903403e-01
4.33343828e-01 -3.35702933e-02 -6.07728183e-01 1.19120085e+00
1.56606257e-01 2.08039552e-01 8.01423013e-01 -1.14374506e+00
1.80186883e-01 3.27946544e-01 2.59515643e-01 -1.18032873e+00
-3.88106734e-01 -2.27770537e-01 -1.00226700e+00 -1.22317076e-01
4.85337734e-01 -2.60167241e-01 -5.17581224e-01 1.90137875e+00
-9.88201983e-03 -5.04917622e-01 2.86779463e-01 1.05990267e+00
5.99031687e-01 5.04129589e-01 -1.95188329e-01 -4.11464810e-01
1.15778852e+00 -9.16355550e-01 -7.89071679e-01 -3.90439242e-01
5.14036536e-01 -9.46159303e-01 1.25528145e+00 7.80441344e-01
-1.33361983e+00 2.05311505e-03 -8.03355396e-01 3.79968323e-02
-8.75244290e-03 -2.42677018e-01 5.99502504e-01 6.39997721e-01
-9.41240072e-01 7.22460985e-01 -2.39160478e-01 -3.02394181e-01
-3.74888331e-02 8.38382766e-02 -2.94830084e-01 3.28288704e-01
-1.66744101e+00 1.41539288e+00 2.48490751e-01 -7.73446634e-02
-5.45015395e-01 -4.22411352e-01 -5.28207183e-01 6.06870875e-02
5.03012955e-01 -7.12866783e-01 1.54188406e+00 -8.21504235e-01
-1.71004605e+00 6.30661964e-01 1.37746811e-01 -5.26703477e-01
7.59055197e-01 -3.63821685e-01 -1.09283626e-01 5.14172204e-02
-1.83632180e-01 7.34284878e-01 8.91139448e-01 -1.33314788e+00
1.52036119e-02 -3.28178965e-02 1.54906496e-01 2.69936532e-01
-6.16182268e-01 -7.13202357e-02 -2.03171909e-01 -6.66628420e-01
-1.42252147e-01 -1.03875780e+00 -2.98381865e-01 -2.77346730e-01
-8.42900634e-01 -2.15548411e-01 1.52084887e-01 -5.65682650e-01
1.40151727e+00 -1.87550080e+00 3.59503210e-01 3.85240585e-01
3.16268593e-01 2.81106532e-01 -2.34210506e-01 7.54516304e-01
1.14411497e-02 3.00813228e-01 -2.73005873e-01 -2.03454331e-01
2.52160996e-01 1.14359036e-01 -5.47535419e-01 2.16024533e-01
1.72679082e-01 7.22735047e-01 -9.82966602e-01 -3.27015579e-01
4.36364375e-02 4.27790344e-01 -8.84385705e-01 1.96847349e-01
-2.93527305e-01 1.02146499e-01 -1.04278244e-01 9.81786251e-02
2.97046423e-01 -3.10216308e-01 2.34348059e-01 -9.04484391e-02
2.79711604e-01 5.58042586e-01 -9.72319007e-01 1.25479865e+00
-4.88870323e-01 4.31454718e-01 -1.93995357e-01 -6.02659106e-01
7.46275067e-01 3.25319022e-01 1.19176291e-01 -5.60826421e-01
-3.93640175e-02 1.58831984e-01 1.07106686e-01 -3.92619133e-01
9.38922763e-01 -3.85331303e-01 -2.27898955e-01 7.80681014e-01
-4.63532060e-02 -4.45588678e-01 1.50333449e-01 7.94471681e-01
1.07131898e+00 -3.47546130e-01 4.81461525e-01 -7.61950985e-02
4.66644883e-01 -2.26988733e-01 5.90932742e-02 1.06498384e+00
1.20983683e-02 6.31415308e-01 9.42564309e-01 4.43274975e-02
-9.20561850e-01 -9.02944386e-01 6.69983104e-02 1.12965250e+00
-2.97897935e-01 -6.34038091e-01 -8.76206517e-01 -7.35850573e-01
-5.27492985e-02 1.34112716e+00 -5.35418808e-01 -5.25341809e-01
-2.32405961e-01 -9.01508510e-01 8.45950842e-01 1.51857570e-01
1.03279822e-01 -1.24755168e+00 -5.28189242e-01 1.69817731e-01
-6.39758110e-01 -6.34019077e-01 -5.15538096e-01 1.41644850e-01
-3.60618472e-01 -7.51359224e-01 -7.53697693e-01 1.15969516e-02
4.78445977e-01 1.43334478e-01 1.20278084e+00 2.51022726e-01
2.76090443e-01 2.49020085e-01 -4.18062598e-01 -1.17968485e-01
-8.42607737e-01 -3.06699984e-02 1.37088522e-01 -5.67576252e-02
2.29274660e-01 -4.14083511e-01 -5.06032109e-01 1.57077521e-01
-1.05381501e+00 2.05693822e-02 4.98495609e-01 1.15286803e+00
-1.03209287e-01 -3.43993276e-01 7.18249977e-01 -1.21618819e+00
1.69258034e+00 -4.55921263e-01 3.63713205e-02 2.90497422e-01
-7.78431356e-01 5.06962121e-01 6.36926889e-01 -5.70186794e-01
-1.16990709e+00 -3.49969596e-01 -2.83472866e-01 -8.64983127e-02
1.98465496e-01 5.42811513e-01 2.05266163e-01 2.05017239e-01
1.09250498e+00 1.90730333e-01 5.25090052e-03 -2.28292555e-01
7.45539367e-01 9.58248913e-01 3.46139073e-01 -5.48843861e-01
5.81718028e-01 1.81687206e-01 -4.88496810e-01 -7.55023181e-01
-1.13342285e+00 -4.06847417e-01 -3.53311636e-02 -2.03625649e-01
4.00507659e-01 -5.66899538e-01 -8.04827511e-01 1.91353932e-01
-1.37038410e+00 -1.76825583e-01 -2.88605601e-01 2.50523746e-01
-7.15609789e-01 6.72538400e-01 -7.64688969e-01 -1.11438823e+00
-4.82900232e-01 -8.43512595e-01 6.77393556e-01 -3.43601070e-02
-1.07853460e+00 -9.32226241e-01 2.55637199e-01 4.89309162e-01
7.74595320e-01 -1.27666906e-01 1.09079373e+00 -1.20766938e+00
1.60022601e-01 -2.05308441e-02 -1.19139798e-01 4.29396868e-01
-1.25559658e-01 -1.88753158e-01 -1.19093752e+00 -1.32357851e-01
2.45193943e-01 -9.72912669e-01 1.04610288e+00 -1.50317758e-01
7.42316544e-01 -1.08505404e+00 2.28904948e-01 -1.41881078e-01
8.10568631e-01 -3.06179971e-01 7.67511785e-01 1.21169955e-01
3.40623736e-01 1.03014827e+00 4.25286680e-01 9.26668167e-01
1.87610641e-01 7.06746936e-01 2.25813046e-01 8.13996047e-02
1.86924875e-01 -3.06939930e-01 6.37848198e-01 8.32458675e-01
-3.63044487e-03 -4.29949224e-01 -5.64401209e-01 3.38698089e-01
-1.85296369e+00 -1.16697896e+00 -1.30549639e-01 2.17838454e+00
1.53324223e+00 3.07839990e-01 2.32514367e-01 8.62498134e-02
3.32246751e-01 2.68648148e-01 -3.41156453e-01 -7.00028181e-01
-1.55553177e-01 1.14974212e-02 1.32195666e-01 9.48073566e-01
-5.02434134e-01 9.54208136e-01 6.01509333e+00 8.20465088e-01
-8.10018003e-01 -6.52589127e-02 5.04870415e-01 -3.02323252e-01
-8.54510009e-01 -8.98354847e-05 -2.84363717e-01 2.55148739e-01
9.94103611e-01 -2.96465576e-01 6.54401720e-01 6.83288753e-01
1.62260920e-01 -1.22330926e-01 -1.04782867e+00 5.61964571e-01
4.08255607e-01 -1.02790785e+00 2.26345688e-01 -1.39904752e-01
5.51938176e-01 -3.49128455e-01 1.66014638e-02 4.40174341e-01
6.47233486e-01 -9.28813934e-01 6.94949865e-01 6.92709327e-01
2.73573041e-01 -8.44385028e-01 7.37281084e-01 5.94062805e-01
-2.89062522e-02 1.75653279e-01 -3.95287842e-01 1.10629745e-01
2.33184278e-01 6.31522477e-01 -1.11737287e+00 2.07343131e-01
2.75464803e-01 1.22081235e-01 -5.78221023e-01 5.18606305e-01
-3.04940999e-01 4.88008529e-01 -1.45710006e-01 -2.66932249e-01
2.50417799e-01 2.32762122e-03 6.75252974e-01 1.30495214e+00
1.45455971e-01 1.52006596e-01 -1.84866846e-01 9.05739844e-01
-8.86753127e-02 1.84541091e-01 -5.80640018e-01 -2.13701576e-01
4.19061542e-01 1.32320023e+00 -2.24017143e-01 -4.23920631e-01
-1.57471210e-01 1.14694929e+00 4.41613019e-01 1.11838497e-01
-5.66180468e-01 -2.35958010e-01 2.67897457e-01 1.60445031e-02
-1.89212486e-01 2.59277910e-01 -1.82080537e-01 -1.18012452e+00
-2.38448039e-01 -1.38308585e+00 2.62518942e-01 -9.34440732e-01
-1.54868352e+00 7.26742983e-01 1.48926182e-02 -8.84255111e-01
-8.31164777e-01 -2.35618204e-01 -2.59372413e-01 7.99369156e-01
-1.21637976e+00 -5.87889791e-01 1.31552443e-01 5.69633663e-01
4.82470185e-01 -6.00008778e-02 1.05584228e+00 -2.13566780e-01
-2.74712890e-01 5.75348735e-01 -2.46695608e-01 -2.12065503e-01
1.12759387e+00 -1.45379567e+00 2.16065675e-01 4.90633339e-01
2.13913262e-01 1.00449347e+00 1.27701974e+00 -5.89098930e-01
-1.02404141e+00 -6.61786735e-01 1.28668046e+00 -5.68287790e-01
9.47332740e-01 -1.83003217e-01 -1.12635291e+00 3.19682330e-01
6.75415039e-01 -7.16328025e-01 7.76141107e-01 2.01818928e-01
-5.14134943e-01 3.17453533e-01 -1.05508423e+00 9.93504345e-01
7.05866218e-01 -4.09476876e-01 -8.79600883e-01 7.11178362e-01
6.22261405e-01 -1.20204605e-01 -5.93715012e-01 7.56802708e-02
4.57093805e-01 -1.36510718e+00 8.25816870e-01 -7.16254294e-01
5.91873705e-01 1.70871735e-01 -8.07955638e-02 -1.74838424e+00
-3.04682523e-01 -1.11178684e+00 -1.85397133e-01 1.13669729e+00
6.31918430e-01 -4.13358778e-01 5.92816770e-01 9.20229912e-01
8.71613696e-02 -5.78000605e-01 -6.55933380e-01 -4.87920314e-01
7.64207840e-02 -3.41003686e-01 2.20632508e-01 1.01471722e+00
7.21221924e-01 8.85924101e-01 -1.01483572e+00 -3.49150509e-01
3.63225728e-01 -3.10657024e-01 6.94844067e-01 -1.03432345e+00
-5.73787510e-01 -4.91850942e-01 3.11845034e-01 -9.54107344e-01
2.98812091e-01 -6.89213812e-01 2.15179235e-01 -1.30102539e+00
3.16794455e-01 -7.56474361e-02 -7.05116913e-02 5.90501726e-01
-2.55448371e-01 2.33806089e-01 3.33509773e-01 2.21392378e-01
-5.40414453e-01 6.59829259e-01 1.23011839e+00 -1.64942741e-02
-2.57200867e-01 1.11396704e-02 -1.14998341e+00 7.79570103e-01
9.06820834e-01 -5.10137498e-01 -5.27325273e-01 -1.36166019e-02
7.19079077e-01 3.66941571e-01 3.99131864e-01 -4.99612391e-01
3.06059062e-01 -2.44244084e-01 1.00622125e-01 -1.57507554e-01
5.72246134e-01 -4.24079776e-01 -1.00458115e-01 4.41888034e-01
-1.01099610e+00 -7.04907114e-04 -1.28535584e-01 5.04072905e-01
-5.27896686e-03 -4.68112260e-01 6.73023224e-01 -2.96609432e-01
-8.89186114e-02 -4.22124147e-01 -7.38735557e-01 3.14463526e-01
3.38690102e-01 1.93659946e-01 -5.39540231e-01 -1.09626937e+00
-4.66086060e-01 3.19117345e-02 3.71127397e-01 3.82431030e-01
8.12368691e-01 -1.17675900e+00 -9.33662474e-01 -1.66524369e-02
2.25597948e-01 -7.22473204e-01 1.64709270e-01 1.00669992e+00
-5.71331643e-02 4.01894301e-01 3.56891342e-02 -5.76550066e-02
-1.17472470e+00 4.14082915e-01 1.81107000e-01 -4.02935743e-01
-3.08768302e-01 8.06503236e-01 -8.44611600e-02 -5.53286076e-01
2.29795128e-01 1.03061438e-01 -3.86283010e-01 2.98018932e-01
5.30441523e-01 2.39664838e-01 1.47097692e-01 -4.08898294e-01
-2.50898033e-01 -1.59315109e-01 -4.17292088e-01 -7.20140636e-01
1.10754514e+00 -9.03101414e-02 -1.85817435e-01 3.66497010e-01
8.12404692e-01 5.25173128e-01 -8.33437979e-01 -2.35700086e-01
1.21377915e-01 -3.55997324e-01 -2.05956310e-01 -1.21209276e+00
-6.39811754e-01 7.18856871e-01 -1.56741425e-01 6.70795023e-01
8.94019663e-01 -1.73746765e-01 5.89527726e-01 7.76670635e-01
9.76989046e-02 -1.28934741e+00 4.58149850e-01 5.81443489e-01
1.32353008e+00 -1.10954726e+00 -6.62523881e-03 -1.58213958e-01
-1.44764662e+00 9.36000705e-01 4.99146432e-01 1.30696908e-01
7.70349503e-02 1.07172184e-01 2.35802084e-01 -1.14545956e-01
-1.35497034e+00 1.62765551e-02 2.69499421e-01 3.55810881e-01
7.50081897e-01 -7.60023519e-02 -4.56686527e-01 6.10895932e-01
-5.43956041e-01 -2.87702888e-01 8.82022083e-01 4.96507674e-01
-4.40211296e-01 -1.04624820e+00 -3.37327510e-01 4.49413508e-01
-4.33073074e-01 -4.53336328e-01 -1.21951413e+00 5.58778346e-01
-4.93031561e-01 1.32299542e+00 -4.47209716e-01 -6.22342348e-01
3.35077882e-01 1.63724676e-01 3.30306917e-01 -5.26259899e-01
-9.61101830e-01 1.30463362e-01 5.97583532e-01 -4.39655095e-01
-2.63936937e-01 -4.26039070e-01 -9.72067773e-01 -5.80323100e-01
-5.07653415e-01 3.40821773e-01 2.04229280e-01 8.39811921e-01
1.12872683e-01 2.55586296e-01 4.79268134e-01 -5.26015460e-01
-1.38771677e+00 -1.28268194e+00 -3.91822696e-01 6.73817575e-01
1.82082847e-01 -2.77328759e-01 -7.39143074e-01 -2.71972299e-01]
|
[12.480759620666504, 8.597088813781738]
|
dda4dd60-dfdb-4132-9221-eb305f6f942b
|
data-driven-approach-for-formality-sensitive
|
2306.14514
| null |
https://arxiv.org/abs/2306.14514v2
|
https://arxiv.org/pdf/2306.14514v2.pdf
|
Data-Driven Approach for Formality-Sensitive Machine Translation: Language-Specific Handling and Synthetic Data Generation
|
In this paper, we introduce a data-driven approach for Formality-Sensitive Machine Translation (FSMT) that caters to the unique linguistic properties of four target languages. Our methodology centers on two core strategies: 1) language-specific data handling, and 2) synthetic data generation using large-scale language models and empirical prompt engineering. This approach demonstrates a considerable improvement over the baseline, highlighting the effectiveness of data-centric techniques. Our prompt engineering strategy further improves performance by producing superior synthetic translation examples.
|
['Heuiseok Lim', 'Chanjun Park', 'Hyeonseok Moon', 'Seugnjun Lee']
|
2023-06-26
| null | null | null | null |
['synthetic-data-generation', 'synthetic-data-generation', 'machine-translation', 'prompt-engineering']
|
['medical', 'miscellaneous', 'natural-language-processing', 'natural-language-processing']
|
[ 3.91219884e-01 3.78197022e-02 -7.56507158e-01 -3.05104136e-01
-1.59123528e+00 -8.81242335e-01 1.34314561e+00 5.02771959e-02
-1.93190426e-01 1.11253667e+00 3.53394449e-01 -9.98658836e-01
2.07733095e-01 -4.41105098e-01 -6.94887698e-01 5.66968955e-02
3.09344292e-01 9.47848380e-01 1.41096860e-01 -6.52418494e-01
4.14392024e-01 5.34173250e-01 -1.31699455e+00 6.48863852e-01
1.15328813e+00 4.75209981e-01 -1.56427070e-01 5.36153078e-01
-5.31627536e-01 7.39811778e-01 -5.89566588e-01 -5.54212570e-01
3.55434626e-01 -4.83081222e-01 -1.07792544e+00 -6.38874620e-02
2.25469217e-01 2.08585292e-01 4.79860812e-01 8.08105171e-01
2.29862854e-01 -4.27510113e-01 4.48327422e-01 -1.46381998e+00
-6.72642350e-01 6.70686603e-01 -2.39073783e-01 1.31710157e-01
5.69889188e-01 3.36048305e-01 1.13987255e+00 -1.30308771e+00
8.83735657e-01 1.21024656e+00 3.84795696e-01 8.13429296e-01
-1.57169926e+00 -3.05568486e-01 -2.67676950e-01 -2.90163189e-01
-1.31010807e+00 -9.53290045e-01 7.20437706e-01 -4.40200657e-01
1.20886505e+00 5.57603359e-01 2.56328225e-01 1.22138619e+00
3.13275129e-01 7.32997775e-01 1.64649785e+00 -1.05907702e+00
1.07277080e-01 6.60539508e-01 -3.40155922e-02 4.28032815e-01
3.11279237e-01 4.66920704e-01 -6.80769682e-01 -5.34706414e-01
3.52426320e-01 -7.84729064e-01 2.01752275e-01 -3.87943774e-01
-1.63650537e+00 8.06744039e-01 -4.65438813e-01 5.16873240e-01
-1.44981295e-01 8.38060491e-03 7.76504934e-01 9.10594583e-01
5.64121604e-01 9.51062620e-01 -8.65914583e-01 -2.95306563e-01
-1.26446295e+00 4.83300000e-01 9.20860708e-01 1.53134656e+00
7.11009681e-01 9.09350663e-02 -3.09963852e-01 5.35856724e-01
2.61259437e-01 6.55301332e-01 5.09484231e-01 -7.48544991e-01
9.46510375e-01 8.12451899e-01 2.98808873e-01 -3.40667427e-01
7.89985284e-02 -1.75871402e-01 -1.34186089e-01 -2.34696940e-01
1.33034319e-01 -1.47628086e-02 -7.07963645e-01 1.84545815e+00
1.45355701e-01 -8.23596060e-01 6.37443960e-01 4.41781998e-01
3.70328903e-01 5.87452054e-01 1.36827663e-01 -6.18140757e-01
1.07926452e+00 -7.56014466e-01 -8.22438896e-01 -1.98678091e-01
1.13479948e+00 -1.02350366e+00 1.54567969e+00 3.98420654e-02
-1.29929650e+00 -1.88910902e-01 -7.82972395e-01 -1.98867902e-01
-4.17107791e-01 5.90042993e-02 7.07754433e-01 7.54628241e-01
-1.36150825e+00 1.71468601e-01 -5.71378231e-01 -4.48235452e-01
-3.71054113e-02 4.03132081e-01 -2.38960505e-01 3.22685450e-01
-1.31995749e+00 9.31870818e-01 4.26453799e-01 -5.39701045e-01
-6.04153872e-01 -8.11309218e-01 -7.90052831e-01 -5.33214152e-01
3.65759909e-01 -7.91740298e-01 1.54280150e+00 -9.94067550e-01
-1.58330202e+00 9.91400778e-01 -4.94042933e-01 -5.03968239e-01
6.71199858e-01 -9.59745571e-02 -6.03312731e-01 -3.43601078e-01
3.88418734e-01 5.15676200e-01 5.45985579e-01 -1.37078536e+00
-5.92427492e-01 8.36671591e-02 -2.33881921e-01 -6.08373918e-02
-2.71494150e-01 8.55200708e-01 -2.27598488e-01 -7.62162864e-01
-4.21701968e-01 -8.92046094e-01 -1.95171297e-01 -4.37647223e-01
-5.02972066e-01 -3.12718868e-01 5.01776457e-01 -5.26332915e-01
1.59687567e+00 -1.82067418e+00 2.80291587e-01 2.09860876e-01
-1.98461458e-01 3.23387891e-01 -4.53917086e-01 9.71407056e-01
2.16596443e-02 5.80688059e-01 -3.05597097e-01 -2.97588795e-01
1.53691947e-01 2.24982724e-01 -8.65188003e-01 -2.47298345e-01
6.54202998e-01 1.47704434e+00 -9.10331309e-01 -8.13596308e-01
-1.46076202e-01 -2.11294830e-01 -4.58198994e-01 2.00389296e-01
-6.83528781e-01 1.70110077e-01 -2.37580851e-01 8.68021727e-01
2.56767452e-01 1.77562788e-01 5.03345013e-01 6.49420992e-02
-4.59634244e-01 7.10338175e-01 -7.51857162e-01 1.74587631e+00
-5.04703641e-01 3.25199127e-01 -2.74169832e-01 -2.71833390e-01
1.21744394e+00 4.51848358e-01 1.90701738e-01 -8.57441962e-01
-1.87136024e-01 9.23786879e-01 -1.72290489e-01 -3.10096413e-01
7.67856240e-01 -3.92977238e-01 -3.20493251e-01 8.95799100e-01
-1.42804459e-01 -4.87600207e-01 5.43174148e-01 2.36912027e-01
7.12117910e-01 4.36091155e-01 4.70059335e-01 -7.63306975e-01
6.08149767e-01 6.37960196e-01 6.41751707e-01 4.43307191e-01
1.24165174e-02 2.81720310e-01 4.45321679e-01 -4.78608191e-01
-1.49484396e+00 -6.79828882e-01 2.25106224e-01 9.50154603e-01
-4.12537277e-01 -7.68960238e-01 -7.72955835e-01 -8.72796714e-01
-1.96779221e-01 1.17217994e+00 -4.08412099e-01 -1.31000653e-01
-9.10063505e-01 -5.49764931e-01 8.70912910e-01 3.49966794e-01
4.29423228e-02 -9.89048302e-01 -3.81493062e-01 1.84622616e-01
-4.92255032e-01 -1.04966474e+00 -6.01641536e-01 -1.68139711e-02
-1.01254666e+00 -6.53999031e-01 -1.07826792e-01 -5.98851144e-01
4.84406114e-01 5.69406673e-02 1.45901871e+00 -8.78873914e-02
2.19804615e-01 -4.48117126e-03 -3.40391606e-01 -5.15227377e-01
-1.34647477e+00 4.50970590e-01 3.19002122e-01 -1.31679788e-01
5.59943616e-01 -2.66200751e-01 1.80326208e-01 1.86969742e-01
-1.03237712e+00 1.85435981e-01 6.03941858e-01 6.75938189e-01
4.80037659e-01 -5.53259850e-01 5.34551501e-01 -9.53520834e-01
1.25339246e+00 -2.48522580e-01 -6.05821550e-01 8.78887236e-01
-9.88975823e-01 4.76425827e-01 9.13656831e-01 -3.89215976e-01
-8.74546051e-01 -1.36785209e-01 2.09523022e-01 -7.00556785e-02
6.33398211e-03 5.73901772e-01 -9.28958803e-02 8.19849893e-02
1.03304410e+00 5.08127093e-01 1.49854347e-01 -4.71828789e-01
4.41442043e-01 8.47349584e-01 1.46098897e-01 -1.25354135e+00
1.10571265e+00 -5.32346815e-02 -1.90024644e-01 -3.23554784e-01
-3.25864166e-01 4.03551087e-02 -5.81970692e-01 2.59919018e-01
1.59234285e-01 -6.77878320e-01 4.28878628e-02 3.70801464e-02
-1.34673762e+00 -3.76732469e-01 -6.34294808e-01 2.07916707e-01
-8.05467308e-01 2.31352866e-01 -3.99966896e-01 -7.88181007e-01
-6.95949256e-01 -1.25277448e+00 1.39153039e+00 -6.11991644e-01
-7.00546503e-01 -9.95489717e-01 4.89048898e-01 3.35695714e-01
5.31091630e-01 1.96833387e-01 1.25518119e+00 -8.26637387e-01
-5.58766901e-01 3.99575457e-02 -1.51091442e-01 1.38809741e-01
1.59175962e-01 4.07507688e-01 -4.67945069e-01 -3.35568368e-01
-2.99105436e-01 -6.24118388e-01 -8.31172056e-03 -3.42929512e-01
2.74779528e-01 -6.72045767e-01 -2.70817671e-02 3.09662879e-01
1.38395631e+00 -7.66191706e-02 2.64382750e-01 4.32348579e-01
4.72903073e-01 8.41946006e-01 7.18517721e-01 -8.67149141e-03
5.24542272e-01 9.42914546e-01 -1.97574839e-01 -2.84361273e-01
-2.50877440e-01 -4.59350795e-01 6.52169585e-01 1.04341435e+00
1.49305493e-01 -3.85065493e-03 -1.31634700e+00 8.23017418e-01
-1.81310475e+00 -6.22077942e-01 -3.20958436e-01 1.92153811e+00
1.41558087e+00 1.06696643e-01 2.57805884e-01 9.19218138e-02
3.69867474e-01 -9.89329740e-02 -2.14377150e-01 -1.29249215e+00
-3.26192051e-01 2.67603517e-01 4.09235656e-01 6.41947687e-01
-5.47738194e-01 1.36736381e+00 7.65158796e+00 8.92416298e-01
-1.06174254e+00 1.16714545e-01 1.73180923e-01 1.26913100e-01
-1.02737117e+00 3.04827213e-01 -9.36931014e-01 3.08866352e-01
1.30875206e+00 -8.10149729e-01 6.51410162e-01 5.61080694e-01
2.82562792e-01 3.77030671e-01 -1.41054714e+00 4.64657784e-01
6.04734682e-02 -1.68864644e+00 6.97162390e-01 2.35864297e-01
6.95354402e-01 -2.96288956e-04 -1.36866689e-01 2.42039189e-01
4.92878973e-01 -6.78486884e-01 1.12200463e+00 2.83395648e-01
1.39619422e+00 -6.82626605e-01 4.20826286e-01 5.38148165e-01
-8.08121622e-01 1.08417995e-01 7.19404668e-02 -2.21322682e-02
-5.01138568e-02 4.45590466e-01 -1.06672966e+00 1.03784192e+00
1.04233518e-01 3.73106420e-01 -4.88142014e-01 5.02218381e-02
-1.78491414e-01 7.18427658e-01 -1.05449371e-01 -1.24325328e-01
2.54318863e-01 -5.29441498e-02 7.16541529e-01 1.50398242e+00
6.62195459e-02 -4.82664555e-01 1.10980786e-01 9.15083826e-01
-8.06198493e-02 3.96039814e-01 -1.02429056e+00 -4.79922175e-01
5.10793269e-01 9.64593589e-01 -1.81319684e-01 -4.29806501e-01
-3.30062807e-01 6.26623154e-01 4.96432006e-01 2.87487447e-01
-3.48195344e-01 -1.33484215e-01 4.76523817e-01 6.00897446e-02
-8.32939297e-02 -5.44625938e-01 -6.14324450e-01 -1.37359881e+00
3.50013435e-01 -1.77688682e+00 3.76734406e-01 -3.95288169e-01
-1.02682281e+00 1.00884998e+00 1.40304193e-01 -1.19848061e+00
-7.75977254e-01 -3.68407547e-01 -2.20772177e-01 9.71873045e-01
-1.48701227e+00 -1.42889869e+00 6.20298922e-01 3.75438333e-01
6.44917250e-01 -4.49714661e-01 1.11408067e+00 1.85812771e-01
-2.86215901e-01 8.83203924e-01 -2.38441229e-02 -1.93790406e-01
7.24162996e-01 -1.16026998e+00 1.07457411e+00 1.18178189e+00
2.00551108e-01 1.15392172e+00 7.07654238e-01 -7.62218475e-01
-1.80811632e+00 -1.23030436e+00 2.04625583e+00 -1.03763533e+00
9.79962587e-01 -7.84205139e-01 -6.36423111e-01 7.52169847e-01
2.51925975e-01 -5.13914227e-01 7.41055906e-01 -3.37861129e-03
-5.40153801e-01 -1.31769866e-01 -1.13881993e+00 9.04598594e-01
1.02280784e+00 -6.98544562e-01 -8.74208093e-01 4.85969573e-01
9.16974962e-01 -1.39886707e-01 -7.80307770e-01 5.03130198e-01
3.22688460e-01 -3.10792923e-01 5.21926820e-01 -1.23135781e+00
6.06865942e-01 -4.01269972e-01 -5.17872095e-01 -1.25216556e+00
2.75094546e-02 -1.38549066e+00 -2.42729962e-01 1.23420215e+00
9.68420029e-01 -6.60994828e-01 2.71982044e-01 7.70413160e-01
1.02126837e-01 -7.47772336e-01 -8.48102212e-01 -1.16336715e+00
3.53976667e-01 -5.50166547e-01 1.18579924e+00 1.11565304e+00
6.50157481e-02 5.83914459e-01 -3.47273678e-01 -3.49977434e-01
4.23779428e-01 3.72213572e-01 1.05587304e+00 -7.69024432e-01
-5.58182560e-02 -3.83999765e-01 2.18199287e-02 -6.36746705e-01
1.09487548e-01 -1.08380902e+00 -1.53727010e-01 -1.29377770e+00
2.02595055e-01 -6.29629552e-01 2.08261982e-02 5.68649411e-01
-7.72315785e-02 3.28795940e-01 3.73446345e-02 4.24766332e-01
-4.03443128e-01 3.23799074e-01 1.01543152e+00 9.17297304e-02
-2.59645581e-01 -2.94674426e-01 -1.04863727e+00 3.36018167e-02
8.77158821e-01 -5.22978246e-01 -4.35939372e-01 -7.29549706e-01
4.51289594e-01 -8.20628405e-02 2.07301416e-02 -4.98439759e-01
2.49362784e-03 -8.58731627e-01 -4.76645052e-01 -3.21755201e-01
-2.03610525e-01 -5.83155036e-01 3.30164164e-01 5.48591137e-01
-6.42572403e-01 5.38400054e-01 2.93094724e-01 -7.84199983e-02
-2.59368300e-01 1.57474548e-01 5.01289129e-01 -1.28602192e-01
-3.66602302e-01 5.90170473e-02 -3.51305515e-01 4.14053172e-01
8.24783027e-01 -2.67307967e-01 -3.51380646e-01 7.99169391e-02
3.73390578e-02 4.75852862e-02 7.41080344e-01 7.25390732e-01
2.46419653e-01 -1.56949735e+00 -1.14030313e+00 4.14199889e-01
6.56061530e-01 -5.96865773e-01 -8.49145770e-01 8.41517210e-01
-2.96694010e-01 8.75851989e-01 -7.36937374e-02 -4.70044196e-01
-1.08933687e+00 7.04002678e-01 2.23920912e-01 -5.16588807e-01
-1.84392139e-01 3.76133174e-01 -3.54246736e-01 -9.76457000e-01
-2.82067090e-01 -3.12462032e-01 3.97148430e-01 -3.66308689e-01
3.84984970e-01 1.70549899e-01 4.22038227e-01 -6.49828374e-01
-6.30626261e-01 2.98723280e-01 -2.74343789e-01 -6.87439799e-01
9.59633529e-01 5.23155183e-02 -3.32996160e-01 6.24423862e-01
8.80159914e-01 3.49742353e-01 -5.59610009e-01 -5.24789751e-01
6.80425704e-01 -3.69281918e-01 -2.44368777e-01 -1.20757329e+00
-2.49589667e-01 5.47356725e-01 -1.05934627e-01 1.44076467e-01
9.76735532e-01 -1.12652116e-01 8.44223917e-01 4.34033364e-01
6.68691218e-01 -1.23213696e+00 -3.61972630e-01 6.44056439e-01
1.00296068e+00 -1.01287353e+00 -1.93720162e-01 -4.29996908e-01
-6.09272599e-01 9.93272066e-01 2.87593931e-01 3.96175504e-01
-9.51278210e-02 5.06260395e-01 5.46790123e-01 -9.07285791e-03
-1.32230115e+00 5.01563447e-03 3.94301862e-01 5.62715828e-01
5.35255790e-01 2.96606570e-01 -9.21297908e-01 2.20195770e-01
-2.62161106e-01 2.69984215e-01 2.82451779e-01 1.25837207e+00
-6.06850721e-02 -1.98535454e+00 -3.75895798e-01 2.14614540e-01
-3.36487174e-01 -5.53173542e-01 -1.23064840e+00 9.67013240e-01
-2.69248366e-01 1.09171081e+00 -3.59537452e-01 -3.61342251e-01
5.05990446e-01 4.14247334e-01 4.70878452e-01 -7.67235458e-01
-9.18918908e-01 -1.25393957e-01 6.36156321e-01 -5.47416806e-01
-3.58461112e-01 -8.84624362e-01 -8.98860514e-01 -4.39176261e-01
-1.89713478e-01 6.03526056e-01 8.56917322e-01 1.00076509e+00
7.45057642e-01 -1.06776066e-01 1.08422875e+00 2.78514065e-02
-8.55541289e-01 -7.29903758e-01 1.62242815e-01 3.98671985e-01
2.76563048e-01 -1.29975066e-01 -1.84732914e-01 2.68692285e-01]
|
[11.57086181640625, 10.289828300476074]
|
e600d9cf-17c8-4684-a3bd-daf868560d7b
|
structured-face-hallucination
| null | null |
http://openaccess.thecvf.com/content_cvpr_2013/html/Yang_Structured_Face_Hallucination_2013_CVPR_paper.html
|
http://openaccess.thecvf.com/content_cvpr_2013/papers/Yang_Structured_Face_Hallucination_2013_CVPR_paper.pdf
|
Structured Face Hallucination
|
The goal of face hallucination is to generate highresolution images with fidelity from low-resolution ones. In contrast to existing methods based on patch similarity or holistic constraints in the image space, we propose to exploit local image structures for face hallucination. Each face image is represented in terms of facial components, contours and smooth regions. The image structure is maintained via matching gradients in the reconstructed highresolution output. For facial components, we align input images to generate accurate exemplars and transfer the high-frequency details for preserving structural consistency. For contours, we learn statistical priors to generate salient structures in the high-resolution images. A patch matching method is utilized on the smooth regions where the image gradients are preserved. Experimental results demonstrate that the proposed algorithm generates hallucinated face images with favorable quality and adaptability.
|
['Ming-Hsuan Yang', 'Chih-Yuan Yang', 'Sifei Liu']
|
2013-06-01
| null | null | null |
cvpr-2013-6
|
['face-hallucination', 'patch-matching']
|
['computer-vision', 'computer-vision']
|
[ 4.24246699e-01 3.15868407e-01 -1.62171796e-01 -3.43836635e-01
-6.48256481e-01 1.61199681e-02 4.95261610e-01 -4.89986658e-01
2.30075687e-01 7.36291766e-01 5.60742736e-01 7.59178638e-01
-1.04911327e-01 -1.00629544e+00 -6.48696244e-01 -8.03148150e-01
1.50395483e-01 -2.17440605e-01 -6.97894916e-02 -1.58343270e-01
2.80538321e-01 7.81126022e-01 -1.82036471e+00 6.69790804e-01
7.69442856e-01 7.39183128e-01 3.20648342e-01 2.68421412e-01
1.09111536e-02 4.66699690e-01 -2.76462644e-01 -2.14391083e-01
2.77884007e-01 -7.48561084e-01 -4.66259241e-01 7.74251282e-01
6.82367802e-01 -1.68508142e-01 -2.81701505e-01 1.44988513e+00
3.49135488e-01 5.34438007e-02 6.81744695e-01 -8.48245800e-01
-1.19695854e+00 -1.34589132e-02 -1.13131404e+00 3.91865671e-02
6.56692505e-01 -1.97079673e-01 4.22343314e-01 -1.51735306e+00
8.64125729e-01 1.71796381e+00 5.32169700e-01 5.28969407e-01
-1.67614174e+00 -4.02847439e-01 -3.23949760e-04 4.25902233e-02
-1.70079732e+00 -8.60689461e-01 1.23441386e+00 -1.66045964e-01
3.82381260e-01 1.86882332e-01 5.07493496e-01 7.35273540e-01
5.42707503e-01 2.11732298e-01 1.21804905e+00 -4.66081709e-01
1.51231095e-01 2.89605618e-01 -5.03278077e-01 9.10865128e-01
7.01984614e-02 2.79015720e-01 -7.56330490e-01 -2.77649939e-01
1.36491203e+00 8.32308531e-02 -6.50972664e-01 -2.83488184e-01
-9.59071517e-01 6.65665686e-01 4.85436022e-01 5.34338593e-01
-8.35050702e-01 -3.29708189e-01 -3.45262736e-01 4.05030623e-02
4.90756899e-01 1.76090389e-01 2.88024485e-01 6.65992260e-01
-1.13471973e+00 2.16523573e-01 2.62914747e-01 9.78222072e-01
1.06329942e+00 5.88743806e-01 -2.31144950e-01 1.17083776e+00
3.37390721e-01 5.55558383e-01 3.92856389e-01 -1.30936146e+00
-1.66585848e-01 2.15311006e-01 4.58025895e-02 -1.38642275e+00
2.57796109e-01 -3.29293907e-01 -1.07087171e+00 5.52221358e-01
-2.18944833e-01 3.22202682e-01 -8.43916416e-01 1.68248045e+00
3.85160506e-01 4.08837020e-01 1.45967066e-01 1.12930846e+00
7.51049042e-01 8.14560890e-01 -2.97020040e-02 -7.11275101e-01
1.38789499e+00 -6.07273102e-01 -1.22738016e+00 -2.09665304e-04
-6.11352444e-01 -1.00138307e+00 7.33105481e-01 2.48125400e-02
-1.78944755e+00 -1.01289880e+00 -9.77099121e-01 -6.89208657e-02
1.68645784e-01 -1.48470402e-02 -7.86223263e-02 3.29498976e-01
-1.25973725e+00 6.75646901e-01 -3.23100805e-01 -1.19728968e-01
6.05186880e-01 -1.83357974e-03 -6.67449832e-01 -1.20214149e-01
-9.40279782e-01 7.74532318e-01 1.21179819e-01 4.30736952e-02
-9.15508926e-01 -8.28796387e-01 -1.23271728e+00 1.65256649e-01
-2.59973943e-01 -7.07193851e-01 6.52878404e-01 -1.23484707e+00
-1.45632589e+00 9.46264744e-01 -5.97739220e-01 1.70421094e-01
4.90686297e-02 3.53828192e-01 -5.52847028e-01 6.76564038e-01
1.42751276e-01 7.20143676e-01 1.61372149e+00 -1.88423538e+00
-3.57913971e-01 -5.10834694e-01 -7.16795087e-01 3.57747018e-01
-2.11875558e-01 7.35607296e-02 -1.84924141e-01 -1.01598465e+00
5.11941254e-01 -2.53629208e-01 -1.04221866e-01 2.53884792e-01
-7.19633549e-02 4.38383192e-01 1.09646642e+00 -9.79088664e-01
8.22172344e-01 -2.34117794e+00 2.78877169e-01 3.13410461e-01
2.46665046e-01 -2.21143097e-01 -2.14148521e-01 1.62293106e-01
-2.46557429e-01 -6.77048415e-02 -4.79865938e-01 -3.11777592e-01
-4.23545539e-01 4.62745735e-03 -4.06059355e-01 6.57283425e-01
5.72247326e-01 8.05970013e-01 -6.75031483e-01 -7.79935241e-01
2.08436638e-01 1.17317200e+00 -6.08036935e-01 4.19900417e-01
1.94490120e-01 6.46197438e-01 -4.99252945e-01 7.89514899e-01
1.24003506e+00 -3.13804075e-02 7.67091438e-02 -4.64177996e-01
-5.73920198e-02 -3.92359912e-01 -1.00731039e+00 1.62544250e+00
-2.33160734e-01 2.34077021e-01 5.80026984e-01 -7.79180110e-01
1.31545317e+00 5.11432648e-01 3.68096828e-01 -7.33659089e-01
-7.54492730e-02 1.04536116e-03 -5.56781411e-01 -4.37770844e-01
3.62054884e-01 -6.77252173e-01 4.72826004e-01 7.98838586e-02
9.92464572e-02 -3.34477365e-01 -3.46683383e-01 -1.43428728e-01
1.29688904e-01 1.86828583e-01 3.71517986e-01 -5.59820712e-01
7.35831738e-01 -4.73471642e-01 5.75481057e-01 7.69475922e-02
1.80816829e-01 1.08346462e+00 -1.45668656e-01 -3.00929934e-01
-1.28927708e+00 -1.54619348e+00 -5.76990306e-01 5.82152128e-01
3.40645313e-01 -8.50461274e-02 -8.52712154e-01 1.25036657e-01
-3.20094407e-01 3.76960069e-01 -7.06770122e-01 -2.61299789e-01
-6.05754375e-01 -2.38842875e-01 2.80710272e-02 2.24701077e-01
8.02294135e-01 -1.26475871e+00 -3.70402545e-01 2.56620437e-01
-2.27582023e-01 -9.44249749e-01 -7.78355300e-01 -7.27994561e-01
-8.87719750e-01 -6.35571182e-01 -1.03581452e+00 -1.31660342e+00
1.11676371e+00 5.23694813e-01 9.51498985e-01 -4.73061837e-02
-6.83629632e-01 2.22966075e-01 2.97849812e-02 1.32693976e-01
-3.09631824e-01 -9.31490242e-01 6.19472675e-02 6.69452369e-01
-2.56640851e-01 -1.04239523e+00 -5.56888700e-01 2.36269087e-01
-9.72808659e-01 2.13136390e-01 5.65857530e-01 1.05603337e+00
1.12341857e+00 3.14772457e-01 6.00015044e-01 -5.55285931e-01
5.46353519e-01 -3.57713312e-01 -2.79041737e-01 1.45508483e-01
-3.50920856e-01 -9.83323306e-02 6.30449295e-01 -5.50704420e-01
-1.66819155e+00 -1.93569604e-02 1.19577333e-01 -8.11968625e-01
-2.15852261e-01 7.23415017e-02 -4.09330547e-01 -2.35194728e-01
7.03715503e-01 7.43133068e-01 4.03568089e-01 -3.91991436e-01
3.81540596e-01 3.21470827e-01 9.22374129e-01 -7.49710262e-01
9.36742604e-01 7.94126034e-01 2.19629072e-02 -1.20115471e+00
-5.55234730e-01 9.38933492e-02 -6.80908799e-01 -3.53847176e-01
9.63687301e-01 -1.01376414e+00 -4.06708628e-01 6.09207973e-02
-9.38542187e-01 3.47201377e-01 -4.95298326e-01 4.26789910e-01
-9.44382310e-01 4.13069546e-01 -5.31554341e-01 -8.36760879e-01
-1.71081454e-01 -9.70671177e-01 1.14699495e+00 5.60516953e-01
9.10530686e-02 -8.67091179e-01 -7.41187483e-02 8.15918893e-02
5.93460917e-01 5.50243139e-01 9.10003662e-01 5.20896733e-01
-6.91304862e-01 2.46807963e-01 -1.06829688e-01 1.80126786e-01
4.19928461e-01 3.08824889e-02 -1.14274919e+00 -4.49507952e-01
3.94725144e-01 -1.56558171e-01 4.52695072e-01 5.37149310e-01
1.06767619e+00 -6.41505837e-01 -2.69002438e-01 6.96581960e-01
1.42361259e+00 1.30483568e-01 8.87368202e-01 -2.91555107e-01
3.67739528e-01 1.08608985e+00 4.39636141e-01 4.46495354e-01
-7.23351017e-02 7.67208636e-01 4.77312878e-02 -4.99186814e-01
-4.30793107e-01 -6.35805130e-01 3.31900120e-01 3.77157837e-01
-1.76764280e-01 4.90532577e-01 -1.90237850e-01 6.59126639e-01
-1.65703225e+00 -1.43131006e+00 4.07524645e-01 2.07266808e+00
9.25695240e-01 -5.43198705e-01 -2.88193762e-01 -3.39635611e-02
1.21868157e+00 2.44928077e-01 -2.90859014e-01 -2.37233073e-01
-4.16364372e-01 2.71174759e-01 -3.43242288e-01 8.74385834e-01
-5.63441455e-01 9.21253204e-01 6.75542212e+00 8.31843197e-01
-8.65181744e-01 -4.27621752e-02 8.24669003e-01 -1.32450042e-02
-6.09515727e-01 -9.24317166e-02 -4.82922792e-01 1.94631889e-01
4.47307885e-01 -4.96877283e-01 5.23716331e-01 6.37029171e-01
4.18724716e-01 1.97420210e-01 -6.80689812e-01 1.15382004e+00
3.23376924e-01 -1.50615144e+00 5.30259788e-01 -4.60960194e-02
1.12295306e+00 -1.04304790e+00 5.52245498e-01 -3.02270502e-01
-8.73662159e-02 -1.46767449e+00 6.74741864e-01 1.00450921e+00
1.04067338e+00 -1.13211131e+00 1.18386067e-01 1.26237094e-01
-1.34577394e+00 1.65501952e-01 -8.58913600e-01 3.17089498e-01
1.94304079e-01 4.41179663e-01 -3.52281153e-01 5.52082360e-01
7.49804497e-01 7.52175689e-01 -6.91533014e-02 5.72088838e-01
-4.65077125e-02 -7.52587151e-03 2.18425363e-01 7.71846950e-01
-2.50162274e-01 -7.51558423e-01 6.38386846e-01 9.22636807e-01
5.81953645e-01 5.00475407e-01 1.29350364e-01 1.41769826e+00
-4.26130183e-02 3.04718107e-01 -9.85432982e-01 2.65591234e-01
5.15969574e-01 1.41960430e+00 -2.60274678e-01 -3.13854396e-01
-3.71276826e-01 1.11148739e+00 1.31096289e-01 6.53648078e-01
-4.86658990e-01 -6.34766892e-02 6.52459502e-01 5.31636715e-01
1.89456224e-01 4.69314046e-02 -2.11399764e-01 -1.01167524e+00
1.67709868e-02 -8.87755215e-01 8.93457755e-02 -9.93505597e-01
-1.27497053e+00 9.15554643e-01 -3.66343255e-03 -1.24487090e+00
-2.57652104e-01 7.28016272e-02 -7.09958911e-01 1.29895902e+00
-1.60478854e+00 -1.37895942e+00 -4.75283056e-01 1.16942024e+00
5.67285240e-01 -3.21544528e-01 8.56911838e-01 -4.23922874e-02
-4.45710681e-03 4.61038262e-01 -1.55069873e-01 -1.96307138e-01
5.61971903e-01 -6.62709892e-01 -1.35305956e-01 7.38410175e-01
1.74596459e-02 7.49182582e-01 6.48779988e-01 -6.82542801e-01
-1.12730384e+00 -1.17270684e+00 7.56147981e-01 1.55247390e-01
-4.00579683e-02 -1.32248625e-01 -1.30223727e+00 4.33224261e-01
5.21440506e-01 2.20356792e-01 5.60135007e-01 -5.43255210e-01
-4.41974819e-01 7.31082186e-02 -1.77571523e+00 7.14197814e-01
8.52396786e-01 -6.21252477e-01 -6.91133142e-01 1.02800027e-01
3.31293255e-01 -6.11065514e-02 -1.19222999e+00 4.19135630e-01
4.84914780e-01 -1.15051210e+00 1.17596304e+00 -3.92570317e-01
5.25872409e-01 -5.80773473e-01 -2.68827289e-01 -1.29933453e+00
-8.33239853e-01 -6.16105676e-01 2.47286960e-01 1.18377948e+00
9.44452360e-02 -4.50494617e-01 5.56180775e-01 1.55054435e-01
6.62594568e-03 -4.73045647e-01 -7.64988720e-01 -3.70317191e-01
-1.37168825e-01 5.77955246e-01 5.91419339e-01 1.10637033e+00
-6.34827614e-02 1.32320255e-01 -5.33059895e-01 4.22965050e-01
1.19586349e+00 5.66693604e-01 3.28428984e-01 -1.06273627e+00
-6.18222542e-02 -3.03537309e-01 -2.94117212e-01 -4.88406956e-01
4.85798359e-01 -7.99450397e-01 -2.94649191e-02 -1.24439561e+00
4.72125769e-01 1.36735618e-01 2.01269314e-02 1.24810122e-01
6.85225148e-03 7.61158586e-01 2.13219973e-04 3.92870098e-01
1.06482580e-01 1.03528881e+00 1.85920334e+00 1.10736445e-01
-8.16869363e-02 -3.50693136e-01 -8.47874999e-01 7.88155437e-01
6.56129003e-01 -8.41351040e-03 -5.05680740e-01 -1.11724960e-03
-5.79804420e-01 5.35762787e-01 3.14586014e-01 -9.11907375e-01
6.64333552e-02 -4.24806416e-01 1.09297180e+00 -4.13714886e-01
8.16715598e-01 -7.36521959e-01 4.47523594e-01 2.35791922e-01
-3.76562268e-01 -1.14327870e-01 1.88529361e-02 4.98581827e-01
-6.50598347e-01 1.63808763e-01 1.79125595e+00 -1.90642565e-01
-3.72824520e-01 5.67649722e-01 -2.63889134e-01 -3.29944611e-01
1.17385423e+00 -5.34779727e-01 -1.06745124e-01 -5.35192847e-01
-9.20318723e-01 -3.46936196e-01 7.20027804e-01 3.77983242e-01
1.50992513e+00 -1.80990314e+00 -1.04002166e+00 8.78282905e-01
2.78478046e-03 -2.77587086e-01 5.22933662e-01 3.23836952e-01
-3.05462122e-01 -2.77700275e-03 -1.05305231e+00 -4.69792306e-01
-1.26251268e+00 6.94374204e-01 4.66093630e-01 4.54709470e-01
-9.24880981e-01 5.21022797e-01 7.59092093e-01 2.08302718e-02
-1.60272539e-01 3.31367671e-01 -4.55969483e-01 -2.58951247e-01
1.01834273e+00 -1.99858323e-02 -4.05956984e-01 -1.14781475e+00
-1.10587001e-01 1.16531682e+00 -1.82179622e-02 -2.87651122e-01
1.35074079e+00 -2.73784101e-01 -4.82328415e-01 4.10461500e-02
1.15165341e+00 3.25275123e-01 -1.40162921e+00 -3.54426146e-01
-4.18543816e-01 -1.26635039e+00 1.00365378e-01 -3.00301701e-01
-1.19646692e+00 6.83839440e-01 5.82122028e-01 -1.18655249e-01
1.54121697e+00 -2.28369907e-02 2.46641800e-01 -3.11856270e-01
3.83130133e-01 -6.73742652e-01 3.89347434e-01 -2.59808898e-02
1.45300817e+00 -8.98273468e-01 2.42217276e-02 -7.34601617e-01
-5.26551664e-01 1.00207138e+00 6.36600852e-01 -4.86432016e-01
6.48585737e-01 3.55367035e-01 -1.63448602e-01 -2.48591661e-01
-6.73519731e-01 8.78994018e-02 3.32786769e-01 9.53879356e-01
4.42679882e-01 -2.24952936e-01 -1.85490921e-01 2.76962757e-01
-1.73583403e-01 -3.37581635e-01 4.32705343e-01 5.58728218e-01
-8.56438279e-01 -6.67623460e-01 -9.02026534e-01 -9.14786160e-02
-3.37562799e-01 -1.41525446e-02 -1.33002192e-01 4.77326572e-01
-3.77923064e-02 9.20506775e-01 1.98769599e-01 -1.53776724e-02
3.34362209e-01 -8.06738585e-02 8.39515865e-01 -6.20198369e-01
-6.27504289e-02 6.27087891e-01 -4.23937023e-01 -6.36121631e-01
-3.76859933e-01 -3.05462748e-01 -1.31343997e+00 -3.14355969e-01
1.90385342e-01 1.90175802e-01 3.94507229e-01 2.45877743e-01
3.67300332e-01 3.96317542e-01 1.00275421e+00 -1.11980093e+00
-1.52796403e-01 -7.70125508e-01 -1.12159991e+00 4.31685627e-01
4.18668121e-01 -6.15974545e-01 -2.22966284e-01 6.01310134e-01]
|
[12.810325622558594, -0.07678718864917755]
|
1799463b-5019-4413-91ed-2e7482e9d011
|
action-spotting-using-dense-detection-anchors
|
2206.07846
| null |
https://arxiv.org/abs/2206.07846v2
|
https://arxiv.org/pdf/2206.07846v2.pdf
|
Action Spotting using Dense Detection Anchors Revisited: Submission to the SoccerNet Challenge 2022
|
This brief technical report describes our submission to the Action Spotting SoccerNet Challenge 2022. The challenge was part of the CVPR 2022 ActivityNet Workshop. Our submission was based on a recently proposed method which focuses on increasing temporal precision via a densely sampled set of detection anchors. Due to its emphasis on temporal precision, this approach had shown significant improvements in the tight average-mAP metric. Tight average-mAP was used as the evaluation criterion for the challenge, and is defined using small temporal evaluation tolerances, thus being more sensitive to small temporal errors. In order to further improve results, here we introduce small changes in the pre- and post-processing steps, and also combine different input feature types via late fusion. These changes brought improvements that helped us achieve the first place in the challenge and also led to a new state-of-the-art on SoccerNet's test set when using the dataset's standard experimental protocol. This report briefly reviews the action spotting method based on dense detection anchors, then focuses on the modifications introduced for the challenge. We also describe the experimental protocols and training procedures we used, and finally present our results.
|
['Avijit Shah', 'João V. B. Soares']
|
2022-06-15
| null | null | null | null |
['action-spotting']
|
['computer-vision']
|
[-2.75810678e-02 -2.68676013e-01 -1.41826287e-01 -3.62147212e-01
-8.95827293e-01 -3.59028876e-01 5.66658676e-01 1.65104568e-01
-1.04979324e+00 9.07521963e-01 2.04953834e-01 4.10044938e-01
-2.80280650e-01 -4.49569553e-01 -4.91318911e-01 -4.08720195e-01
-7.12465465e-01 5.42869449e-01 1.15326881e+00 -7.04396844e-01
4.03548092e-01 1.55448884e-01 -1.85745513e+00 5.91248333e-01
1.87695473e-01 9.29519594e-01 5.73363490e-02 8.14755857e-01
4.95237827e-01 6.25969589e-01 -1.01099730e+00 -1.44214839e-01
4.93578702e-01 -2.43347615e-01 -5.68357944e-01 -5.77264667e-01
5.64390421e-01 4.72119302e-02 -3.15752298e-01 7.48447835e-01
8.96499693e-01 7.47082889e-01 -4.27735001e-02 -1.38531852e+00
3.88822347e-01 7.35290647e-01 -4.30639505e-01 8.13176155e-01
6.89851403e-01 2.11499080e-01 9.97499406e-01 -7.02674150e-01
7.78462231e-01 9.70101297e-01 1.25249279e+00 5.39397299e-01
-1.07692564e+00 -6.90708399e-01 3.36789191e-01 7.85036743e-01
-1.53038466e+00 -3.77537519e-01 3.45728427e-01 -2.58419961e-01
1.15543377e+00 3.35218430e-01 9.68717933e-01 1.24750841e+00
1.33149698e-01 1.05423343e+00 1.07040799e+00 -2.54793763e-01
2.33147830e-01 -1.36540011e-01 2.34559029e-01 1.92595601e-01
-6.97438270e-02 6.21564209e-01 -8.13078046e-01 8.63487199e-02
3.56363803e-01 -6.15848184e-01 -1.48073569e-01 -4.12986070e-01
-1.54289556e+00 8.35363209e-01 4.61551785e-01 5.33388078e-01
-3.18072945e-01 2.87295759e-01 8.81674230e-01 2.69712001e-01
4.26567614e-01 5.09017527e-01 -4.33338732e-01 -7.88692236e-01
-1.15886748e+00 7.76441753e-01 7.95178115e-01 5.71677804e-01
1.33006737e-01 -2.38437936e-01 -6.70449972e-01 8.05536926e-01
3.78214605e-02 -7.83547163e-02 3.19722116e-01 -1.00650001e+00
5.18874764e-01 9.19929743e-02 9.26399305e-02 -1.23111928e+00
-6.16753161e-01 -4.88914967e-01 -1.73685297e-01 3.15875977e-01
4.23712432e-01 -3.71029601e-03 -9.33951139e-01 1.68185449e+00
2.80292779e-01 3.52736861e-01 -3.86807233e-01 1.32874155e+00
6.46825492e-01 3.44235390e-01 1.51416466e-01 5.79164624e-02
1.21385252e+00 -1.14417470e+00 -6.87063098e-01 -1.93683226e-02
7.75634527e-01 -7.79255688e-01 8.65738511e-01 6.86169088e-01
-8.98407519e-01 -8.11838388e-01 -1.62375796e+00 3.51284057e-01
-5.57814956e-01 -1.35344863e-01 6.32720828e-01 7.30960250e-01
-1.17448711e+00 8.89714837e-01 -9.63368177e-01 -6.61515057e-01
-2.10307818e-02 3.90702337e-01 -3.40905935e-01 3.32025774e-02
-1.56965554e+00 1.10511851e+00 6.73731148e-01 5.40000312e-02
-8.89383554e-01 -7.59084284e-01 -7.28632927e-01 -4.18538600e-01
7.25785255e-01 -3.03363025e-01 1.45652092e+00 -3.57677907e-01
-1.62500572e+00 8.03008437e-01 3.13471079e-01 -1.08601785e+00
1.04287899e+00 -4.70376194e-01 -6.30231977e-01 4.84010093e-02
3.04662913e-01 9.89703000e-01 1.56458989e-01 -7.53107607e-01
-1.03231776e+00 -2.10799556e-02 1.26420304e-01 4.02010083e-01
2.18691722e-01 2.86252022e-01 -7.75482178e-01 -6.48413658e-01
7.41373226e-02 -8.18955958e-01 -5.08878529e-01 -4.13952887e-01
6.12601545e-03 -2.52666384e-01 7.78264582e-01 -3.17812949e-01
1.23573744e+00 -2.11608267e+00 -1.54738545e-01 2.91013509e-01
1.84562936e-01 3.04922491e-01 -7.25351796e-02 4.14735228e-01
-1.87894069e-02 -1.37430280e-01 9.04043671e-03 -4.72501189e-01
5.84873259e-02 2.77408868e-01 -4.28027771e-02 5.95328689e-01
-1.90721408e-01 6.23314440e-01 -1.18461347e+00 -4.82416928e-01
5.55831015e-01 3.53936493e-01 -5.26554823e-01 -1.85897902e-01
-2.08430886e-02 4.00045991e-01 -9.55453068e-02 3.77376050e-01
3.86712879e-01 3.78169119e-01 -2.70315707e-01 -2.50805736e-01
-3.47322285e-01 3.50677907e-01 -1.53398478e+00 2.08882642e+00
2.43551880e-02 7.62807965e-01 -6.76302090e-02 -1.05302382e+00
8.18674445e-01 2.48753995e-01 8.68952096e-01 -9.00150359e-01
8.56624022e-02 3.00018132e-01 9.43960473e-02 -9.75716785e-02
9.97796834e-01 5.85104227e-02 -2.83662111e-01 8.10936373e-03
1.19789064e-01 -7.07397535e-02 9.20444846e-01 3.89448345e-01
1.22877550e+00 4.63704288e-01 1.51980013e-01 -5.36806643e-01
4.45762873e-01 1.38132140e-01 7.70281672e-01 1.08485663e+00
-7.39352524e-01 8.87551248e-01 3.29546124e-01 -7.08287477e-01
-7.26795375e-01 -8.58761370e-01 -1.32988364e-01 1.00575423e+00
3.80575836e-01 -8.97233665e-01 -3.79758596e-01 -9.06493902e-01
-1.56904042e-01 5.66072643e-01 -9.25333738e-01 -8.85468498e-02
-8.52460206e-01 -7.87718475e-01 9.43906426e-01 6.68974161e-01
7.23957539e-01 -9.80547428e-01 -7.13028848e-01 4.44888711e-01
-4.50838953e-01 -1.22420442e+00 -4.52278465e-01 2.48539463e-01
-5.64541578e-01 -1.25928843e+00 -6.97928131e-01 -3.34037900e-01
-2.64781918e-02 -1.97079908e-02 1.21317792e+00 -3.50441746e-02
-4.18466508e-01 3.21798086e-01 -6.71234369e-01 -4.45600241e-01
-1.26272708e-01 2.50627786e-01 4.33410078e-01 -3.22829545e-01
4.89753246e-01 -3.90869081e-01 -6.36934578e-01 8.04087698e-01
-5.25336564e-01 -3.04641843e-01 2.90103942e-01 6.38001740e-01
5.62827468e-01 -2.90527672e-01 3.24554116e-01 -4.70659137e-01
5.80280304e-01 -2.31118023e-01 -6.25558496e-01 -2.77270615e-01
-4.83021289e-01 -1.70271397e-01 -1.95819549e-02 -3.53345990e-01
-5.87284207e-01 -3.91186960e-03 -4.73768115e-01 -3.34883332e-01
1.57617237e-02 2.75620490e-01 3.65591258e-01 -3.23999733e-01
9.85248089e-01 -1.45228282e-01 -7.75401667e-02 -5.37653387e-01
1.04472294e-01 2.49625787e-01 6.00376308e-01 -7.41252661e-01
3.07628214e-01 3.82303625e-01 -8.20292830e-02 -6.16243660e-01
-7.68418789e-01 -8.91960680e-01 -7.48115957e-01 -6.52487218e-01
6.51888490e-01 -7.92457044e-01 -8.46120954e-01 4.03338104e-01
-9.49911535e-01 -3.72275949e-01 -5.97543597e-01 9.10814166e-01
-6.78896666e-01 2.82454163e-01 -4.86534506e-01 -4.00106996e-01
-9.89778861e-02 -1.00973856e+00 7.38731802e-01 4.57735099e-02
-3.81836265e-01 -8.85715604e-01 8.07961464e-01 1.32288009e-01
5.56409776e-01 5.51314473e-01 -2.51737237e-01 -8.65310371e-01
-1.83845356e-01 -4.66136426e-01 -8.76479670e-02 5.40853739e-01
-4.11172807e-01 -3.27257246e-01 -8.21114838e-01 -4.70979124e-01
-3.64204258e-01 -1.65040836e-01 1.11816430e+00 3.43806893e-01
7.48489559e-01 4.47021097e-01 -4.90968049e-01 3.88444930e-01
1.10890090e+00 3.01671606e-02 7.49589264e-01 8.16057146e-01
1.48842841e-01 4.19306248e-01 1.13759744e+00 3.04998577e-01
1.25110298e-01 1.31832993e+00 2.86968827e-01 1.55062109e-01
-4.38909471e-01 -2.41316557e-01 4.27130342e-01 6.09556437e-01
-4.02799398e-01 -1.36868194e-01 -8.21646333e-01 8.32671106e-01
-2.12618446e+00 -1.04032218e+00 -3.27989370e-01 2.25996399e+00
5.63172817e-01 5.02503455e-01 5.29163539e-01 3.73599291e-01
7.09101021e-01 3.24940860e-01 -2.60205306e-02 -1.70582190e-01
-1.96642175e-01 3.00289005e-01 6.48300111e-01 4.94933367e-01
-1.32465684e+00 9.11612928e-01 7.53801680e+00 1.02632916e+00
-8.95933926e-01 2.78609961e-01 -4.59967628e-02 -4.02502596e-01
7.53386438e-01 -1.55286834e-01 -1.19668448e+00 4.37109321e-01
9.65045094e-01 -1.30334035e-01 -6.59243837e-02 7.92058527e-01
2.06463739e-01 -5.04256248e-01 -1.09037149e+00 8.22156847e-01
1.77774370e-01 -1.29109406e+00 -4.00786519e-01 -2.06012383e-01
5.28786182e-01 5.63633144e-01 -2.66671062e-01 7.49359488e-01
3.65505368e-01 -7.04942644e-01 9.11307693e-01 3.28581542e-01
5.13175547e-01 -6.62576139e-01 1.13543987e+00 6.00254014e-02
-1.44027746e+00 6.07554764e-02 -1.50523320e-01 -3.23235095e-01
3.47496450e-01 4.58294392e-01 -7.42089748e-01 6.39181793e-01
1.04442060e+00 9.49063599e-01 -5.79531133e-01 1.73663044e+00
-2.55613625e-01 4.36621547e-01 -8.41764569e-01 -2.17864960e-01
3.94863456e-01 5.14642954e-01 1.12151408e+00 1.60963142e+00
-8.33148416e-03 -1.23214856e-01 6.31645322e-01 3.87133151e-01
4.23661053e-01 -3.49283069e-02 -2.57530928e-01 5.47785103e-01
1.50932834e-01 1.10978580e+00 -9.26184416e-01 -3.78778189e-01
9.58412513e-02 5.82358420e-01 -1.16995014e-01 4.29746509e-02
-1.15355587e+00 -6.67522132e-01 7.56333888e-01 3.95970017e-01
3.12999815e-01 -3.05721551e-01 -1.63119838e-01 -9.15588439e-01
3.38210282e-03 -7.64021993e-01 7.75114655e-01 -4.78071183e-01
-8.37085187e-01 6.44388199e-01 5.81220150e-01 -1.38620496e+00
-2.13326216e-01 -3.95741999e-01 -4.39933747e-01 4.90796238e-01
-9.95715559e-01 -6.54389560e-01 -2.01434806e-01 5.78554332e-01
6.18854880e-01 -2.85085607e-02 6.11808240e-01 7.28676856e-01
-3.15158516e-01 7.92338550e-01 -3.66818130e-01 5.51167950e-02
9.56132054e-01 -9.99946058e-01 4.60105717e-01 9.33054328e-01
2.41876304e-01 2.03451097e-01 1.14687479e+00 -5.69411516e-01
-5.74629724e-01 -7.34147251e-01 7.07420707e-01 -6.00391865e-01
7.29876995e-01 -4.98182774e-01 -4.33259189e-01 6.49652243e-01
1.57434672e-01 6.88066846e-03 2.29923978e-01 2.90123969e-01
-6.00327887e-02 -1.04576401e-01 -1.17252219e+00 3.46667618e-01
1.19067848e+00 6.86350763e-02 -7.79518783e-01 3.03700358e-01
6.30579293e-01 -8.93214464e-01 -6.68558180e-01 8.58296096e-01
5.49726844e-01 -1.04341531e+00 8.92814517e-01 -3.57787192e-01
-7.48619661e-02 -5.94397485e-01 -2.42144927e-01 -1.30138099e+00
-2.41964802e-01 -5.54205716e-01 5.02545722e-02 7.31276929e-01
3.68596077e-01 -3.23555291e-01 1.00956368e+00 -8.84104073e-02
-3.56519341e-01 -5.27631104e-01 -1.43941247e+00 -1.12953234e+00
-4.80181217e-01 -1.07276547e+00 1.25459746e-01 5.57071626e-01
1.12910561e-01 -1.72086488e-02 -3.84854823e-01 -1.99331671e-01
7.72113800e-01 -5.94098508e-01 9.10173774e-01 -1.02114999e+00
-3.68349820e-01 -3.33456337e-01 -1.18058038e+00 -1.13333166e+00
-5.65774858e-01 -5.76051474e-01 3.98172945e-01 -1.06681919e+00
-1.32628068e-01 -2.43463621e-01 -6.39707088e-01 4.94917214e-01
8.55780989e-02 9.29238617e-01 4.71285254e-01 5.58369234e-02
-1.23315144e+00 3.62705469e-01 8.76567364e-01 1.88093409e-02
-2.24004179e-01 2.72015065e-01 -2.02883616e-01 5.59537292e-01
6.63314760e-01 -6.10351503e-01 -3.43123786e-02 -1.57985419e-01
2.24201187e-01 -4.35563624e-01 4.56644684e-01 -1.76576746e+00
4.58465755e-01 2.32980654e-01 1.26242936e-01 -9.62009549e-01
6.40668809e-01 -3.75667095e-01 -1.43191233e-01 7.39972234e-01
-2.78492987e-01 1.26931578e-01 7.02897668e-01 4.65489715e-01
-4.63693470e-01 6.79197460e-02 8.79694581e-01 -6.98762983e-02
-1.30268621e+00 -3.34798805e-02 -3.85767996e-01 1.06634699e-01
1.33327878e+00 -4.49081600e-01 1.31096244e-01 -1.60944566e-01
-1.27194715e+00 6.20824099e-01 7.57284695e-03 8.71873260e-01
4.19149846e-01 -1.36857891e+00 -9.28218186e-01 4.05435041e-02
2.86514312e-01 -4.67548132e-01 3.77237111e-01 1.30250597e+00
-6.40136778e-01 4.99105990e-01 -4.13066387e-01 -1.02106798e+00
-1.46428263e+00 1.38559207e-01 5.58577061e-01 -6.80649638e-01
-7.53297806e-01 1.02496445e+00 -4.61429358e-01 -5.04012346e-01
6.29880250e-01 -3.01714987e-01 -3.24226141e-01 1.62980072e-02
8.21129262e-01 5.20617545e-01 3.97125959e-01 -4.55895334e-01
-6.56591594e-01 4.26831990e-01 -1.96066767e-01 -4.59047019e-01
1.15783405e+00 -2.31319834e-02 3.92833471e-01 5.34450293e-01
8.59209895e-01 -1.78541154e-01 -1.12041259e+00 -1.64988250e-01
2.58096308e-01 -5.71855903e-01 8.07518810e-02 -1.02637565e+00
-9.61823761e-01 5.06558001e-01 8.66936862e-01 -7.73483468e-03
7.50755370e-01 -1.85070783e-01 6.56052470e-01 2.28320479e-01
7.81067789e-01 -1.45105243e+00 -4.40220721e-02 8.07749450e-01
8.71717989e-01 -1.05082726e+00 1.62230685e-01 -2.41748989e-01
-3.99321407e-01 1.01876175e+00 8.14769447e-01 -3.29145461e-01
5.47249913e-01 4.36670423e-01 3.86118256e-02 -3.14901680e-01
-7.10572839e-01 -3.03057164e-01 2.02133015e-01 6.81874573e-01
2.19661847e-01 -5.67238629e-02 -7.96698570e-01 4.70167011e-01
-3.23871911e-01 2.73843765e-01 3.15689474e-01 1.04512513e+00
-4.46252167e-01 -1.03901601e+00 -4.65396017e-01 6.38172552e-02
-4.44485486e-01 1.28575727e-01 -2.64187962e-01 1.20226669e+00
4.18044895e-01 8.96244466e-01 -1.88456811e-02 -9.53428388e-01
9.26042020e-01 -2.48013675e-01 4.90504622e-01 -5.15162945e-01
-8.36333990e-01 -3.21649432e-01 4.00204152e-01 -1.18940318e+00
-5.04783392e-01 -7.23649323e-01 -1.04466224e+00 -5.70344448e-01
-1.64293364e-01 5.43535590e-01 7.65486538e-01 8.25502932e-01
2.49944225e-01 7.45016754e-01 8.76782686e-02 -1.25110698e+00
-3.74197930e-01 -1.22319114e+00 -4.84991550e-01 1.94817171e-01
-1.03174090e-01 -9.60298717e-01 -3.67539972e-01 -3.61502081e-01]
|
[8.039730072021484, 0.1878150999546051]
|
5576b60b-460b-4a6d-9f23-f4bc82f0601b
|
salsanext-fast-semantic-segmentation-of-lidar
|
2003.03653
| null |
https://arxiv.org/abs/2003.03653v3
|
https://arxiv.org/pdf/2003.03653v3.pdf
|
SalsaNext: Fast, Uncertainty-aware Semantic Segmentation of LiDAR Point Clouds for Autonomous Driving
|
In this paper, we introduce SalsaNext for the uncertainty-aware semantic segmentation of a full 3D LiDAR point cloud in real-time. SalsaNext is the next version of SalsaNet [1] which has an encoder-decoder architecture where the encoder unit has a set of ResNet blocks and the decoder part combines upsampled features from the residual blocks. In contrast to SalsaNet, we introduce a new context module, replace the ResNet encoder blocks with a new residual dilated convolution stack with gradually increasing receptive fields and add the pixel-shuffle layer in the decoder. Additionally, we switch from stride convolution to average pooling and also apply central dropout treatment. To directly optimize the Jaccard index, we further combine the weighted cross-entropy loss with Lovasz-Softmax loss [2]. We finally inject a Bayesian treatment to compute the epistemic and aleatoric uncertainties for each point in the cloud. We provide a thorough quantitative evaluation on the Semantic-KITTI dataset [3], which demonstrates that the proposed SalsaNext outperforms other state-of-the-art semantic segmentation networks and ranks first on the Semantic-KITTI leaderboard. We also release our source code https://github.com/TiagoCortinhal/SalsaNext.
|
['George Tzelepis', 'Eren Erdal Aksoy', 'Tiago Cortinhal']
|
2020-03-07
| null | null | null | null |
['robust-3d-semantic-segmentation']
|
['computer-vision']
|
[ 2.01795787e-01 3.18948925e-01 1.18740931e-01 -7.51796722e-01
-9.01437998e-01 -3.30393940e-01 4.81725812e-01 -2.08750248e-01
-5.51241755e-01 7.16301799e-01 -2.73134500e-01 6.12068959e-02
-1.52539760e-01 -8.93694937e-01 -1.12316203e+00 -5.52122056e-01
6.63632378e-02 7.80158818e-01 5.80905497e-01 1.36234090e-01
1.96018279e-01 5.77657938e-01 -1.56555283e+00 7.22754821e-02
1.09000254e+00 1.47816277e+00 3.08433205e-01 5.22354484e-01
-1.22161359e-01 8.31622481e-01 -2.98380703e-01 -6.25017107e-01
5.62588215e-01 1.80090562e-01 -8.63542438e-01 -3.08025777e-01
8.25883210e-01 -4.01233017e-01 -5.65637469e-01 1.20959425e+00
5.01056731e-01 2.15684444e-01 5.39080799e-01 -1.29865515e+00
-5.72817445e-01 8.29632163e-01 -5.93779504e-01 -1.57350469e-02
-3.72868538e-01 1.87406629e-01 1.03893971e+00 -1.08528244e+00
4.80308443e-01 1.35432363e+00 8.09983075e-01 2.49668717e-01
-1.02509630e+00 -9.28334832e-01 1.62254453e-01 2.97575414e-01
-1.56312871e+00 4.34994046e-03 5.77209115e-01 -4.76053834e-01
9.93474305e-01 -8.94248709e-02 5.87888896e-01 8.23119044e-01
2.09295496e-01 8.38849127e-01 9.03135657e-01 2.40105703e-01
3.56311917e-01 -3.20417136e-01 6.19037710e-02 6.34368777e-01
2.78879497e-02 2.89292544e-01 -4.09911424e-01 1.91335678e-01
4.90659535e-01 1.00251712e-01 3.38682644e-02 -3.18873525e-01
-8.35462928e-01 8.36473644e-01 1.21426499e+00 -2.46914327e-01
-3.09134424e-01 6.25069261e-01 1.75635248e-01 -1.65880308e-01
7.38278568e-01 4.35155965e-02 -5.58468521e-01 1.74849972e-01
-1.32334197e+00 3.88022035e-01 5.31769872e-01 1.19861877e+00
9.99047399e-01 -6.35823458e-02 -2.25800797e-01 7.16996670e-01
5.11919260e-01 6.05967045e-01 -4.78256233e-02 -1.35783160e+00
3.72856617e-01 2.08929464e-01 -2.84941703e-01 -1.81918174e-01
-2.20783159e-01 -6.24555111e-01 -5.41412711e-01 4.32761788e-01
-8.33677500e-03 4.23799232e-02 -1.53897429e+00 1.51348448e+00
2.31072620e-01 6.38040245e-01 -1.24868512e-01 9.45005059e-01
8.65540981e-01 4.53863800e-01 9.31106806e-02 4.48067665e-01
1.22599947e+00 -1.12118268e+00 -2.36734852e-01 -5.28942525e-01
-7.31318295e-02 -5.55410683e-01 6.70064330e-01 1.17444135e-01
-1.16428518e+00 -5.00496447e-01 -1.29658747e+00 -5.12644053e-01
-5.61741531e-01 -8.24458450e-02 5.09467840e-01 1.48756891e-01
-9.81195211e-01 1.12534642e+00 -1.08545494e+00 1.19256862e-01
9.15707886e-01 4.89152551e-01 -9.65566433e-04 -8.17294419e-02
-1.22128558e+00 7.91761756e-01 5.34963012e-01 3.22919160e-01
-8.73563826e-01 -8.92757356e-01 -1.12277031e+00 -5.77094173e-03
5.27457416e-01 -9.08431590e-01 1.51288486e+00 -6.38512135e-01
-1.55807447e+00 9.65076745e-01 -1.01407252e-01 -7.27204263e-01
6.09031439e-01 -2.56524742e-01 1.08739860e-01 4.36436199e-02
3.05683285e-01 1.47041905e+00 7.48396993e-01 -1.30392373e+00
-6.54468715e-01 -3.69461507e-01 3.40123451e-03 1.10820547e-01
5.58954298e-01 -3.03029895e-01 -7.01403439e-01 -5.26121557e-01
3.32671314e-01 -9.30382013e-01 -1.80874258e-01 1.91710889e-02
-5.20519137e-01 -6.79503977e-02 7.72373736e-01 -6.86600506e-01
7.28841066e-01 -2.18219662e+00 1.35383844e-01 2.80017585e-01
3.00155848e-01 4.57468517e-02 -8.35819095e-02 -1.22344092e-01
-1.00338804e-02 1.20693959e-01 -1.12625253e+00 -7.81122744e-01
2.35019162e-01 4.13130671e-01 -2.58769751e-01 3.60971868e-01
3.83585215e-01 1.14369881e+00 -7.57397950e-01 -4.34991807e-01
5.13744295e-01 8.30737293e-01 -6.07066393e-01 -8.26933235e-02
-5.94583988e-01 4.26144361e-01 -3.54006380e-01 7.21141636e-01
1.34013999e+00 -1.55248553e-01 -5.25168478e-01 7.28540029e-03
-1.82931781e-01 6.13974333e-01 -9.67142105e-01 2.04997182e+00
-5.04069448e-01 5.57949662e-01 1.79015130e-01 -5.82132459e-01
7.27252066e-01 -2.90716499e-01 4.93516564e-01 -5.66506982e-01
3.42818916e-01 3.61030847e-01 -2.58581430e-01 -1.40823983e-02
6.86305285e-01 -1.12840824e-01 -3.89419049e-02 -1.17838159e-01
3.98485154e-01 -8.02833736e-01 -3.10940649e-02 1.77096084e-01
9.31734622e-01 5.06831765e-01 -3.20300102e-01 -1.38678879e-01
3.76792431e-01 -1.11067273e-01 6.67648613e-01 5.79426169e-01
-1.18318632e-01 1.12538481e+00 3.17671686e-01 2.46926099e-02
-8.17735553e-01 -1.43675542e+00 -5.40174723e-01 6.78108871e-01
3.20028454e-01 -2.63554633e-01 -9.58220422e-01 -9.26385939e-01
5.01829326e-01 9.78054523e-01 -5.88449478e-01 8.83589983e-02
-2.76471645e-01 -5.50719678e-01 5.65003455e-01 6.49680734e-01
1.04413712e+00 -1.01702142e+00 -4.39510465e-01 -1.09611653e-01
7.76601583e-02 -1.25770152e+00 -3.97768438e-01 5.38082838e-01
-7.66613424e-01 -8.89343679e-01 -3.79411966e-01 -4.95002717e-01
3.69865745e-01 -1.80332363e-01 1.20666623e+00 -2.89696008e-01
-2.83703297e-01 3.25752571e-02 -5.64142428e-02 -5.44879377e-01
7.23942220e-02 4.14741963e-01 -4.90553141e-01 -4.10590738e-01
4.68622237e-01 -6.59159780e-01 -7.16350853e-01 6.66054338e-02
-9.06070888e-01 1.07257716e-01 3.37117106e-01 5.73071837e-01
1.05640924e+00 -1.78463295e-01 1.87638089e-01 -7.58801937e-01
2.53851768e-02 -5.27788162e-01 -1.01737154e+00 -2.05045536e-01
-5.17780602e-01 1.70962438e-01 1.54397637e-01 3.77340943e-01
-1.05435741e+00 2.03887269e-01 -7.15604901e-01 -7.65225172e-01
-1.60995254e-03 1.51793674e-01 -3.01984459e-01 -8.02197456e-02
2.32751265e-01 -1.74407735e-01 -3.57632399e-01 -4.37989831e-01
5.76021135e-01 7.34253526e-01 6.99305594e-01 -4.85787034e-01
7.91411042e-01 5.66562176e-01 1.97098963e-02 -5.30359685e-01
-1.04805887e+00 -2.47100055e-01 -7.20007420e-01 -1.48995429e-01
1.16516757e+00 -1.13555169e+00 -8.03460658e-01 6.37457728e-01
-1.26693249e+00 -5.99283278e-01 -8.67265463e-01 3.18712056e-01
-6.63077295e-01 1.56212941e-01 -4.68731284e-01 -4.58364069e-01
-3.92584890e-01 -1.37231433e+00 1.55244708e+00 3.98343414e-01
4.36632335e-01 -6.14445627e-01 -2.11782351e-01 5.08414745e-01
3.57945919e-01 1.99768156e-01 4.60570574e-01 -3.69679540e-01
-1.14471364e+00 1.29207164e-01 -5.92845559e-01 8.04295361e-01
-4.06365544e-01 2.11414113e-03 -1.48783207e+00 -1.87953059e-02
-8.39606896e-02 -2.82881469e-01 1.55403852e+00 5.89235365e-01
1.70374274e+00 1.71924323e-01 -6.72003254e-02 1.17730618e+00
1.50384724e+00 -2.29244623e-02 7.96258807e-01 1.71203285e-01
7.33923733e-01 3.30623299e-01 3.91680509e-01 1.89213827e-01
7.29088783e-01 3.31428379e-01 9.75624621e-01 1.83347270e-01
-3.74353111e-01 -4.55487132e-01 1.77812830e-01 6.18228436e-01
-3.06210816e-02 -3.55346620e-01 -8.30444932e-01 4.66127396e-01
-1.90575778e+00 -6.25803351e-01 -4.47353385e-02 2.05682683e+00
5.12346268e-01 2.16576174e-01 -5.09706795e-01 -2.79693305e-01
6.26244724e-01 4.44742113e-01 -9.50711548e-01 -2.35322312e-01
-1.38698712e-01 7.70300448e-01 1.07666528e+00 8.75027418e-01
-1.17429578e+00 1.36628795e+00 5.43576193e+00 9.82055306e-01
-9.32648540e-01 4.45075631e-01 5.70174336e-01 -2.13946402e-01
-3.28614801e-01 1.61771089e-01 -7.52084076e-01 6.34543478e-01
8.11562121e-01 2.80922532e-01 5.02644479e-01 7.97148943e-01
-1.54574186e-01 -4.45723623e-01 -9.24513459e-01 7.53154397e-01
-1.55119196e-01 -1.29251337e+00 -3.48433048e-01 4.31708992e-02
7.71203279e-01 1.11553144e+00 8.01712051e-02 2.44319782e-01
6.20569885e-01 -1.05866539e+00 9.78220105e-01 5.65482140e-01
8.55188072e-01 -8.28457832e-01 7.56955266e-01 7.38106519e-02
-1.20865870e+00 7.38520324e-02 -4.53250587e-01 2.47822016e-01
2.73897976e-01 9.10086811e-01 -5.68152905e-01 7.20652044e-01
9.21949148e-01 7.74583399e-01 -4.36135948e-01 1.10539067e+00
-6.55346870e-01 3.89415026e-01 -6.84784949e-01 5.42052865e-01
3.42390358e-01 -3.47662896e-01 5.67157686e-01 1.10678411e+00
4.28690612e-01 -9.70004797e-02 -9.34767649e-02 1.28712475e+00
-6.29917562e-01 -4.43823099e-01 -3.47641170e-01 3.18247765e-01
4.47555661e-01 1.10503483e+00 -6.79578781e-01 -2.99833536e-01
-1.68367922e-01 1.21937859e+00 1.52527779e-01 3.10342997e-01
-9.71782207e-01 -4.78371590e-01 9.60157037e-01 1.08551294e-01
7.04161346e-01 -1.77570388e-01 -6.87508285e-01 -9.95200098e-01
9.85371135e-03 -1.35730654e-01 1.45130649e-01 -1.17243743e+00
-1.30126953e+00 5.53007066e-01 1.21487439e-01 -8.44194829e-01
-2.75664888e-02 -6.81425214e-01 -4.34423566e-01 1.10122943e+00
-1.95654988e+00 -1.34275961e+00 -4.30949509e-01 2.96797544e-01
5.25303006e-01 2.00714916e-01 2.21976176e-01 3.83383691e-01
-4.36617672e-01 2.57749051e-01 1.25682335e-02 -4.33120737e-03
4.83521193e-01 -1.44280982e+00 8.59107494e-01 7.20347047e-01
-3.02998960e-01 1.99085236e-01 2.88341820e-01 -8.57903242e-01
-6.74327731e-01 -1.51724184e+00 7.16174841e-01 -3.63294482e-01
6.06509447e-01 -6.26834989e-01 -9.22022164e-01 9.61506665e-01
3.39331299e-01 2.47640207e-01 -5.97243384e-02 -3.44984680e-01
-1.97969481e-01 -1.52848303e-01 -1.39571953e+00 2.65859306e-01
1.36399734e+00 -5.65270245e-01 -5.91217339e-01 1.86103135e-01
1.09676838e+00 -8.61940324e-01 -7.91810989e-01 8.36718440e-01
3.16254228e-01 -1.12991285e+00 1.00365269e+00 9.02076736e-02
5.91757059e-01 -5.53874135e-01 -3.04344326e-01 -1.21195364e+00
-8.98247808e-02 -2.77898759e-01 7.16095418e-02 1.03210223e+00
3.92293125e-01 -7.41364002e-01 8.70515347e-01 3.25656056e-01
-6.90141201e-01 -7.96233296e-01 -1.29491389e+00 -6.65347576e-01
3.30958068e-01 -8.77633035e-01 8.85724366e-01 3.82608920e-01
-6.96627080e-01 6.19412027e-02 1.37112170e-01 3.22044283e-01
8.78758252e-01 -1.89192280e-01 3.31007361e-01 -1.25508857e+00
-4.12246250e-02 -4.54830587e-01 -3.60520095e-01 -1.09022999e+00
4.59990084e-01 -1.31398261e+00 4.13031697e-01 -1.68523252e+00
4.10391539e-02 -4.35814619e-01 -2.54545450e-01 4.76766378e-01
9.56725478e-02 3.35819930e-01 3.30571294e-01 -3.69616635e-02
-4.32378113e-01 9.82609570e-01 1.14799190e+00 -4.18314517e-01
1.26731217e-01 7.48947412e-02 -4.37353909e-01 7.96807349e-01
7.62096584e-01 -6.60772920e-01 -2.98532635e-01 -5.80949545e-01
2.29120985e-01 -3.77793700e-01 7.39078283e-01 -1.28811932e+00
1.77977026e-01 1.10801235e-01 3.44089627e-01 -9.84166801e-01
5.24836302e-01 -7.78749287e-01 9.88717750e-02 2.16052562e-01
-4.15302590e-02 -2.33360022e-01 3.97360802e-01 3.65976423e-01
-2.22604349e-01 -2.97649711e-01 1.00376582e+00 -8.97814706e-02
-7.34248161e-01 7.38753080e-01 3.76283117e-02 7.71636888e-02
8.61570954e-01 -1.31135032e-01 -3.33666921e-01 8.26916248e-02
-7.51316488e-01 6.96981013e-01 6.57835484e-01 3.40803206e-01
7.10703015e-01 -1.05865574e+00 -5.75941145e-01 2.52958864e-01
-5.83983399e-02 9.70239937e-01 5.90585291e-01 6.85022533e-01
-7.43231237e-01 1.77220821e-01 -2.03606449e-02 -8.53517950e-01
-7.00429678e-01 6.70995638e-02 7.41174877e-01 -6.57760203e-02
-7.43082523e-01 1.21299839e+00 2.55580902e-01 -8.47930372e-01
2.43961498e-01 -6.44031227e-01 2.92483509e-01 1.08052799e-02
-1.08003907e-01 3.70125473e-01 2.36880496e-01 -5.69057286e-01
-5.15919149e-01 6.57813191e-01 1.24064364e-01 -2.10448116e-01
1.35849464e+00 1.46064246e-02 -1.67429164e-01 3.14091027e-01
1.07442176e+00 -2.94092238e-01 -1.84187007e+00 -2.34051228e-01
-1.57031983e-01 -3.42808992e-01 4.48356688e-01 -1.03222513e+00
-1.58300948e+00 1.20100200e+00 8.04105341e-01 -3.66040826e-01
1.03510702e+00 2.12734029e-01 8.95703554e-01 9.37348902e-02
1.94885135e-01 -1.19047940e+00 -3.38242203e-01 8.86789024e-01
9.21453416e-01 -1.27153480e+00 -5.63808121e-02 -4.28016931e-01
-5.97136974e-01 6.78110480e-01 6.16742730e-01 -4.09446746e-01
1.04593480e+00 4.59358037e-01 -2.02689499e-01 -3.46002191e-01
-3.50790530e-01 -6.28393412e-01 8.53254497e-02 3.24749917e-01
8.76966044e-02 1.90154254e-01 3.18084769e-02 6.57525003e-01
-4.85401839e-01 2.41916582e-01 2.59254038e-01 6.88201070e-01
-4.62115288e-01 -8.81505787e-01 -1.53060257e-01 4.50482935e-01
-2.02553108e-01 -2.55942762e-01 -2.63315409e-01 5.70138812e-01
6.73349738e-01 6.10538900e-01 4.56325382e-01 -4.21211511e-01
4.55711037e-01 2.25788087e-01 4.86391127e-01 -5.72969735e-01
-5.76481223e-01 -2.13934015e-02 -2.17514068e-01 -7.23319471e-01
-1.56590164e-01 -7.46956050e-01 -1.71496475e+00 -4.38519359e-01
-8.49262848e-02 -9.46333632e-02 1.03236318e+00 8.00305843e-01
5.34268796e-01 9.41962600e-01 2.92750806e-01 -1.17981446e+00
-2.99009949e-01 -1.02152574e+00 -5.66052258e-01 -8.09246525e-02
3.53024632e-01 -7.89400876e-01 -3.28952819e-01 -4.26081359e-01]
|
[8.243489265441895, -2.6819064617156982]
|
09c54ed6-39a4-407e-9f84-0c0c3973f037
|
cross-lingual-adaptation-for-type-inference
|
2107.00157
| null |
https://arxiv.org/abs/2107.00157v5
|
https://arxiv.org/pdf/2107.00157v5.pdf
|
Cross-Lingual Transfer Learning for Statistical Type Inference
|
Hitherto statistical type inference systems rely thoroughly on supervised learning approaches, which require laborious manual effort to collect and label large amounts of data. Most Turing-complete imperative languages share similar control- and data-flow structures, which make it possible to transfer knowledge learned from one language to another. In this paper, we propose a cross-lingual transfer learning framework, PLATO, for statistical type inference, which allows us to leverage prior knowledge learned from the labeled dataset of one language and transfer it to the others, e.g., Python to JavaScript, Java to JavaScript, etc. PLATO is powered by a novel kernelized attention mechanism to constrain the attention scope of the backbone Transformer model such that the model is forced to base its prediction on commonly shared features among languages. In addition, we propose the syntax enhancement that augments the learning on the feature overlap among language domains. Furthermore, PLATO can also be used to improve the performance of the conventional supervised learning-based type inference by introducing cross-lingual augmentation, which enables the model to learn more general features across multiple languages. We evaluated PLATO under two settings: 1) under the cross-domain scenario that the target language data is not labeled or labeled partially, the results show that PLATO outperforms the state-of-the-art domain transfer techniques by a large margin, e.g., it improves the Python to TypeScript baseline by +5.40%@EM, +5.40%@weighted-F1, and 2) under the conventional monolingual supervised learning based scenario, PLATO improves the Python baseline by +4.40%@EM, +3.20%@EM (parametric).
|
['Yang Liu', 'Yi Li', 'Zhengzi Xu', 'Haoliang Li', 'Xiaofei Xie', 'Zhiming Li']
|
2021-07-01
| null | null | null | null |
['fault-localization']
|
['computer-code']
|
[-2.29582153e-02 3.07651460e-02 -6.28643036e-01 -4.81366277e-01
-8.12454164e-01 -9.43975329e-01 6.90718532e-01 2.18199589e-03
-6.15065575e-01 7.27896512e-01 -3.07119470e-02 -5.53286254e-01
5.64494193e-01 -9.60229933e-01 -1.23454773e+00 -5.14657199e-01
1.91181257e-01 3.49439949e-01 2.62037754e-01 -4.43948470e-02
1.25762615e-02 -5.34132533e-02 -1.28391731e+00 5.18090963e-01
1.23326397e+00 8.15796852e-01 3.99508536e-01 2.93878853e-01
-5.95563471e-01 9.17296469e-01 -2.48948589e-01 -5.37937760e-01
-4.54486199e-02 2.03971099e-02 -9.49479878e-01 -3.10510248e-01
4.19071376e-01 -1.71981886e-01 -4.64925468e-02 1.08710766e+00
1.21725813e-01 -2.28410125e-01 5.72631180e-01 -1.32629240e+00
-7.01695025e-01 9.21717882e-01 -5.39801240e-01 -3.17439049e-01
2.43699566e-01 2.44060531e-01 1.13592744e+00 -8.33599985e-01
5.61154902e-01 1.37320220e+00 6.61300719e-01 6.15204275e-01
-1.40698051e+00 -7.82192171e-01 2.21606404e-01 1.73869208e-01
-1.12749684e+00 -1.95833117e-01 7.24942863e-01 -5.74412584e-01
1.02089918e+00 -9.41545293e-02 1.53466791e-01 1.08145154e+00
-1.82934716e-01 8.92815530e-01 1.37723947e+00 -6.46929443e-01
1.08219050e-01 5.79469681e-01 2.89215326e-01 8.19976091e-01
6.63714334e-02 -9.99453887e-02 -3.84100258e-01 -1.96023345e-01
3.95174652e-01 -3.90168726e-01 -3.32460329e-02 -3.24022382e-01
-1.14579117e+00 7.24903882e-01 2.87432969e-01 2.44088769e-01
4.73627113e-02 5.50005659e-02 6.93577528e-01 3.41324210e-01
4.13860321e-01 4.63692874e-01 -1.03921366e+00 -1.67405114e-01
-5.12848377e-01 1.73653543e-01 9.50301290e-01 1.17294824e+00
1.30219042e+00 -1.04674771e-01 5.37862927e-02 9.85316217e-01
3.06057602e-01 7.09146261e-01 4.92628813e-01 -8.73398125e-01
7.55337894e-01 8.64837229e-01 -6.88499659e-02 -3.12165588e-01
-2.17713974e-03 -5.58235683e-02 -4.81585145e-01 -1.86783373e-02
6.65338993e-01 -2.81555742e-01 -6.72145128e-01 2.08633542e+00
3.28292787e-01 -9.60115343e-02 1.36099592e-01 4.89325613e-01
5.62619388e-01 6.29931629e-01 1.84730858e-01 7.45493695e-02
1.42352867e+00 -8.54597807e-01 -3.58689457e-01 -3.72847736e-01
1.14067137e+00 -6.82248473e-01 1.69378614e+00 2.73353368e-01
-7.37131357e-01 -3.94195467e-01 -7.98204482e-01 -2.90986687e-01
-4.89874780e-01 -1.65155940e-02 7.09422708e-01 5.58309495e-01
-7.56664872e-01 3.03902626e-01 -9.31500912e-01 -2.64098704e-01
3.76804113e-01 2.06559226e-01 -3.52009892e-01 -3.54586512e-01
-1.14521599e+00 6.67657018e-01 7.23213375e-01 -3.58764082e-01
-7.92700291e-01 -1.08824360e+00 -9.48975682e-01 6.59074560e-02
5.60932994e-01 -5.59203744e-01 1.25402057e+00 -1.08620524e+00
-1.57485747e+00 9.73285139e-01 -2.88927257e-01 -2.15058908e-01
3.41410547e-01 -4.11686987e-01 -4.76366132e-02 -3.10300976e-01
1.72772601e-01 5.26073754e-01 3.41488987e-01 -1.26271164e+00
-6.07946754e-01 -2.84873188e-01 3.49987239e-01 3.32554430e-02
-4.51946408e-01 2.24154830e-01 -5.60029149e-01 -5.66858590e-01
-4.60635602e-01 -1.06433284e+00 1.92967609e-01 -6.98541254e-02
-2.77093858e-01 -5.36352277e-01 7.01095939e-01 -7.25144148e-01
9.06317532e-01 -2.19290948e+00 2.73306459e-01 2.43160669e-02
-5.09852208e-02 2.87353575e-01 -5.66454343e-02 1.42225340e-01
5.95509671e-02 2.98870206e-01 -4.06278580e-01 -3.06626201e-01
2.65431762e-01 4.35791135e-01 -2.94467598e-01 8.90657678e-02
3.21147025e-01 8.97004247e-01 -7.26371288e-01 -5.66089869e-01
-1.57524914e-01 1.53848156e-01 -1.06702042e+00 5.24345219e-01
-8.22091460e-01 5.07903576e-01 -3.30913901e-01 4.26949710e-01
6.40098751e-01 -2.61489600e-01 5.34022927e-01 -5.32032549e-02
-1.91541880e-01 7.56537497e-01 -9.22961831e-01 1.82598507e+00
-1.04177606e+00 2.61433631e-01 -7.92848989e-02 -1.06282568e+00
8.81659806e-01 1.67280152e-01 7.24443123e-02 -4.63450611e-01
-2.61493415e-01 4.19305652e-01 -1.32565843e-02 -4.42193538e-01
1.66147858e-01 -1.65515423e-01 -5.49792171e-01 5.38845897e-01
1.55723006e-01 1.61171909e-02 7.61908740e-02 1.49198666e-01
7.84052014e-01 5.79299092e-01 7.82063231e-02 -3.55451107e-01
7.06925154e-01 2.25374922e-02 8.76884758e-01 4.79164749e-01
3.42012912e-01 -1.07330479e-01 7.30337083e-01 -1.93762690e-01
-9.97567415e-01 -1.01916003e+00 -2.58967280e-02 1.52572227e+00
-9.42009762e-02 -3.80681336e-01 -8.76514256e-01 -9.89842653e-01
9.93262380e-02 8.49588752e-01 -2.84561127e-01 -1.52009919e-01
-7.97987103e-01 -6.19223833e-01 7.55977273e-01 7.57548451e-01
7.60655463e-01 -1.02457595e+00 -2.81384755e-02 -1.29136723e-02
-4.05975908e-01 -1.44929278e+00 -4.63351876e-01 1.56943113e-01
-7.32145488e-01 -8.60750675e-01 -3.97486120e-01 -1.00837338e+00
5.45571208e-01 -3.53115171e-01 1.08322310e+00 1.90971419e-02
2.19860345e-01 1.72452465e-01 -2.88755655e-01 -1.82587937e-01
-5.62651813e-01 4.87695128e-01 3.74893136e-02 -4.89607174e-03
4.90869641e-01 -5.72781861e-01 3.67341861e-02 1.68549106e-01
-7.12316990e-01 8.72410089e-02 5.27935505e-01 9.65155721e-01
3.57639611e-01 -2.38119677e-01 3.78019929e-01 -1.24392581e+00
1.02304205e-01 -5.18896639e-01 -8.25156152e-01 3.81656408e-01
-4.99766350e-01 4.61340010e-01 1.09268022e+00 -6.61676347e-01
-1.27920520e+00 -1.39667735e-01 5.76541275e-02 -2.78381974e-01
-1.57839745e-01 7.50494897e-01 -7.79605627e-01 1.48107320e-01
4.92407143e-01 3.58414531e-01 -7.91352913e-02 -7.80733228e-01
4.23508435e-01 8.96535695e-01 4.99907583e-01 -1.27177238e+00
9.56421077e-01 1.09403685e-01 -4.94351000e-01 -5.70609629e-01
-7.00510442e-01 -1.43928332e-02 -6.75453663e-01 3.66276115e-01
7.93132186e-01 -9.20642734e-01 -6.18833184e-01 8.02240908e-01
-1.16968632e+00 -9.57284570e-01 7.81458393e-02 3.96712482e-01
-5.22071540e-01 4.32204455e-01 -7.45628536e-01 -4.02108103e-01
-1.71369240e-01 -1.23616672e+00 9.02983487e-01 -1.19507730e-01
-7.62837380e-02 -1.19878221e+00 -4.69650403e-02 4.23885584e-01
3.30484390e-01 -1.71110794e-01 1.64781320e+00 -7.53649235e-01
-5.84470749e-01 1.70134693e-01 -3.60131919e-01 6.60383582e-01
1.50195450e-01 -1.14335328e-01 -9.28232908e-01 -2.70073295e-01
-2.78066188e-01 -5.80331922e-01 4.59893882e-01 -2.46778741e-01
1.27127457e+00 -6.30744457e-01 -2.72615552e-01 7.28494763e-01
1.36389470e+00 -1.06716715e-01 5.51936567e-01 4.09853786e-01
1.00476635e+00 6.10889792e-01 4.85437244e-01 -6.00884017e-03
8.99158835e-01 8.84387016e-01 1.01550475e-01 -3.10652554e-02
-1.52345719e-02 -4.23209697e-01 6.86459422e-01 9.79957223e-01
2.45486528e-01 2.19358623e-01 -1.15513349e+00 5.19137621e-01
-1.67042589e+00 -5.06494582e-01 1.26507074e-01 2.29310679e+00
1.51844168e+00 8.42205510e-02 1.89493150e-01 -1.61224023e-01
6.61250889e-01 -2.27076486e-01 -4.80242521e-01 -4.93604422e-01
1.81288704e-01 3.67610037e-01 3.86817396e-01 5.65144420e-01
-9.48683858e-01 1.34753633e+00 4.90988302e+00 8.06069434e-01
-1.44356728e+00 2.33308196e-01 2.62156188e-01 3.25824827e-01
-4.86796200e-01 2.59736240e-01 -9.34072614e-01 7.32973397e-01
8.50435913e-01 -2.76345938e-01 6.69874907e-01 9.67548966e-01
-1.45126611e-01 4.07561380e-03 -1.40856969e+00 5.81627190e-01
-2.21938387e-01 -9.31822062e-01 4.62532714e-02 1.28448024e-01
5.00548244e-01 1.43924683e-01 -7.68097639e-02 8.72261643e-01
6.73289955e-01 -8.29909027e-01 5.25261462e-01 1.39497817e-01
1.01333618e+00 -6.75106823e-01 6.41949177e-01 5.66680729e-01
-9.72751379e-01 1.53521314e-01 -1.96794525e-01 -1.66120939e-02
-1.73404217e-01 4.77840215e-01 -7.49471843e-01 5.53052425e-01
6.23671353e-01 6.05209231e-01 -4.27949429e-01 4.89085317e-01
-5.14573038e-01 8.72821987e-01 -2.79734731e-01 -3.67728174e-02
1.03239313e-01 -8.68977904e-02 1.96271524e-01 1.35631692e+00
1.66418906e-02 -3.61777484e-01 5.42642832e-01 1.17782295e+00
-3.55502367e-01 1.72910005e-01 -5.16151488e-01 -1.13705128e-01
5.45808911e-01 1.06009483e+00 7.54386261e-02 -6.06238246e-01
-7.66509414e-01 7.81452537e-01 6.29442513e-01 3.08197021e-01
-9.19096947e-01 -5.33458233e-01 5.10348797e-01 1.20865896e-01
4.49341863e-01 -2.73796827e-01 -2.66684413e-01 -1.53411031e+00
2.60249645e-01 -1.07366848e+00 3.08467031e-01 -5.50454319e-01
-1.30471826e+00 3.23713124e-01 1.98156059e-01 -7.27990806e-01
-3.24897438e-01 -7.68526852e-01 -4.76927400e-01 1.15653980e+00
-1.73518503e+00 -1.43145812e+00 -7.42925704e-02 6.53238356e-01
1.31606504e-01 -1.85359403e-01 9.55015719e-01 5.02112329e-01
-5.06737709e-01 1.07440066e+00 -5.54955937e-02 5.96370816e-01
8.98724318e-01 -1.40945160e+00 2.70158678e-01 7.32386827e-01
-1.52715355e-01 9.06320930e-01 2.84146458e-01 -4.49642837e-01
-1.65614808e+00 -1.39637780e+00 1.07766938e+00 -3.24894845e-01
9.93018627e-01 -7.01944947e-01 -1.38721299e+00 1.07413721e+00
3.35531384e-02 1.35679498e-01 6.52761877e-01 5.14356613e-01
-1.18242252e+00 -2.98024923e-01 -1.05945015e+00 5.62071264e-01
7.92685449e-01 -8.18432152e-01 -8.05789530e-01 2.24414155e-01
8.68138075e-01 -1.84355423e-01 -1.19407105e+00 3.20644051e-01
6.21021986e-01 -3.96973044e-01 5.48842311e-01 -7.08052337e-01
5.85767448e-01 -1.99858978e-01 -4.62605566e-01 -1.18942034e+00
-6.74740225e-02 -4.88930702e-01 -5.36603341e-03 1.70772851e+00
5.55243552e-01 -9.72758710e-01 5.08691370e-01 5.03955781e-01
-1.40531212e-01 -3.96656543e-01 -7.92160153e-01 -9.03843820e-01
7.09113002e-01 -5.26011705e-01 5.67231119e-01 1.22554135e+00
1.84935391e-01 5.36792278e-01 -5.57896420e-02 2.36846745e-01
5.09352446e-01 2.62917072e-01 1.05548739e+00 -9.74441588e-01
-7.92844057e-01 -2.90768415e-01 -6.53003380e-02 -1.07605100e+00
7.53041685e-01 -1.42060649e+00 -1.83769502e-02 -9.68630731e-01
3.35507542e-01 -9.58944917e-01 -1.44325569e-01 1.02359438e+00
-3.81775737e-01 -8.29848424e-02 2.01963276e-01 2.13281259e-01
-3.27700734e-01 4.45598871e-01 8.01661968e-01 -1.87921301e-01
-1.09161243e-01 -3.00303012e-01 -7.30136812e-01 7.51173913e-01
7.91336000e-01 -5.50010860e-01 -1.31130025e-01 -7.54113317e-01
1.77153260e-01 -1.76091895e-01 3.14216018e-01 -6.05895936e-01
-7.25775808e-02 -1.87104672e-01 -2.10953146e-01 -8.61603990e-02
-2.76340768e-02 -6.00194275e-01 -2.92660147e-01 3.75741333e-01
-2.36593053e-01 -2.07869723e-01 3.87055129e-01 1.55392468e-01
-4.31828648e-02 -2.62368739e-01 7.85694003e-01 -7.97112808e-02
-7.82976806e-01 7.55001381e-02 -1.36628419e-01 5.70632339e-01
6.86995625e-01 2.65487462e-01 -4.44572777e-01 3.29483226e-02
-3.49588096e-01 2.08643988e-01 8.41108441e-01 3.85329068e-01
-8.97023901e-02 -1.28970742e+00 -6.04265511e-01 2.62821943e-01
3.81058395e-01 1.18440658e-01 -1.99979022e-01 8.04670513e-01
-1.68227032e-01 2.92352945e-01 -6.97967857e-02 -8.09374988e-01
-1.09092855e+00 6.56138539e-01 2.05788404e-01 -4.12449867e-01
-3.29173952e-01 4.94939208e-01 4.56934810e-01 -1.14442611e+00
7.68047050e-02 -3.90723169e-01 2.28359953e-01 -3.07742149e-01
2.90132731e-01 1.37355579e-02 4.67530712e-02 -3.44374269e-01
-4.64639783e-01 6.19667411e-01 -3.40423912e-01 -4.82950024e-02
1.31289387e+00 3.46563429e-01 -4.52469647e-01 7.00559855e-01
1.33334470e+00 2.54718423e-01 -1.04549003e+00 -6.75068557e-01
1.69592604e-01 -1.25372663e-01 -2.11390197e-01 -1.03149366e+00
-8.46029639e-01 9.97270584e-01 1.24866225e-01 -2.11257279e-01
9.29557562e-01 1.60722286e-01 6.34525001e-01 4.59998190e-01
7.13441074e-01 -7.52621531e-01 -2.90862471e-02 8.72396469e-01
5.17884433e-01 -1.20755017e+00 -3.01911831e-01 -5.87426245e-01
-5.71461856e-01 8.16537857e-01 7.14602232e-01 -8.64282018e-04
3.46873075e-01 4.94970948e-01 -6.53459951e-02 3.15981269e-01
-8.80857468e-01 7.90025964e-02 1.72734365e-01 5.75770676e-01
7.34215617e-01 1.86739206e-01 -1.48743436e-01 8.02452087e-01
-2.21029118e-01 1.88153028e-01 2.68140107e-01 8.90370011e-01
-1.17097817e-01 -1.62144804e+00 -2.13535532e-01 2.18221739e-01
-5.36329269e-01 -4.31375504e-01 -1.43386409e-01 9.28602576e-01
1.57710060e-01 5.86031199e-01 1.87093038e-02 -2.67955869e-01
1.55295357e-01 4.60186899e-01 6.10929668e-01 -9.19356883e-01
-5.97694039e-01 -6.49867356e-02 1.85736716e-01 -3.20828766e-01
-2.49942884e-01 -6.65773153e-01 -1.34266698e+00 -3.42308074e-01
-2.60518104e-01 1.02904506e-01 5.20561576e-01 1.05057192e+00
3.62116277e-01 3.28201175e-01 5.78501940e-01 -4.21031207e-01
-6.92682326e-01 -1.03999162e+00 -1.16940007e-01 5.02396286e-01
5.20677939e-02 -6.25709116e-01 -3.41630161e-01 2.33332574e-01]
|
[10.986761093139648, 9.856405258178711]
|
6474ae08-5810-46b1-8fe9-5f0d1212d019
|
towards-automated-imbalanced-learning-with
|
2208.12433
| null |
https://arxiv.org/abs/2208.12433v1
|
https://arxiv.org/pdf/2208.12433v1.pdf
|
Towards Automated Imbalanced Learning with Deep Hierarchical Reinforcement Learning
|
Imbalanced learning is a fundamental challenge in data mining, where there is a disproportionate ratio of training samples in each class. Over-sampling is an effective technique to tackle imbalanced learning through generating synthetic samples for the minority class. While numerous over-sampling algorithms have been proposed, they heavily rely on heuristics, which could be sub-optimal since we may need different sampling strategies for different datasets and base classifiers, and they cannot directly optimize the performance metric. Motivated by this, we investigate developing a learning-based over-sampling algorithm to optimize the classification performance, which is a challenging task because of the huge and hierarchical decision space. At the high level, we need to decide how many synthetic samples to generate. At the low level, we need to determine where the synthetic samples should be located, which depends on the high-level decision since the optimal locations of the samples may differ for different numbers of samples. To address the challenges, we propose AutoSMOTE, an automated over-sampling algorithm that can jointly optimize different levels of decisions. Motivated by the success of SMOTE~\cite{chawla2002smote} and its extensions, we formulate the generation process as a Markov decision process (MDP) consisting of three levels of policies to generate synthetic samples within the SMOTE search space. Then we leverage deep hierarchical reinforcement learning to optimize the performance metric on the validation data. Extensive experiments on six real-world datasets demonstrate that AutoSMOTE significantly outperforms the state-of-the-art resampling algorithms. The code is at https://github.com/daochenzha/autosmote
|
['Xia Hu', 'Na Zou', 'Sirui Ding', 'Qiaoyu Tan', 'Kwei-Herng Lai', 'Daochen Zha']
|
2022-08-26
| null | null | null | null |
['hierarchical-reinforcement-learning']
|
['methodology']
|
[-8.67349878e-02 -1.04358926e-01 -4.67350274e-01 -2.86000401e-01
-9.60220993e-01 -1.18393570e-01 1.74343243e-01 1.57583937e-01
-2.43682444e-01 1.00504804e+00 -9.35706049e-02 -3.08540285e-01
-3.28149758e-02 -1.07492840e+00 -7.74066985e-01 -7.46483386e-01
1.86696067e-01 7.46767223e-01 2.64580268e-02 -1.61895782e-01
3.36203873e-01 4.13868055e-02 -1.70559561e+00 4.32246357e-01
1.28448820e+00 8.77780139e-01 -9.18456540e-02 5.21948338e-01
-1.55683801e-01 5.65252602e-01 -9.32135463e-01 -6.29944429e-02
3.12530607e-01 -5.52352905e-01 -6.56833470e-01 1.33775502e-01
1.17193073e-01 -2.52234668e-01 2.45875090e-01 9.89053249e-01
6.84777021e-01 -1.03911288e-01 5.69824934e-01 -1.53997171e+00
-2.31100366e-01 7.64354169e-01 -7.88483799e-01 5.28888293e-02
2.96844691e-02 3.45918566e-01 9.40656602e-01 -6.48400307e-01
2.64755815e-01 1.18030310e+00 4.75685537e-01 5.38803339e-01
-1.32344043e+00 -9.88077998e-01 2.58798063e-01 4.30208981e-01
-1.34668958e+00 -2.60560513e-01 7.09782720e-01 -4.01736915e-01
4.53592062e-01 2.63641119e-01 7.91313589e-01 9.32251930e-01
1.99306309e-02 1.14604247e+00 1.28904438e+00 -4.92307723e-01
7.64242411e-01 5.18216342e-02 -2.96826661e-02 3.91416281e-01
5.11750400e-01 -1.54404612e-02 -5.35268247e-01 -2.63691157e-01
3.13198566e-01 2.19714031e-01 -1.39768377e-01 -4.69982535e-01
-6.39906228e-01 1.04977453e+00 3.22663844e-01 -5.55046536e-02
-4.55156624e-01 -1.75277978e-01 3.93403590e-01 1.84870183e-01
3.99865210e-01 7.26296663e-01 -5.36970973e-01 -1.45516038e-01
-9.41604495e-01 7.21839964e-01 7.39641428e-01 5.64664483e-01
8.48410249e-01 -1.18193328e-01 -3.30547988e-01 9.92381334e-01
-3.10257655e-02 2.02465817e-01 6.53765798e-01 -6.36700630e-01
8.10693145e-01 8.28509152e-01 3.25640589e-01 -8.43278944e-01
-2.01658651e-01 -4.21936870e-01 -1.00545132e+00 2.50792623e-01
4.57314044e-01 -3.18496197e-01 -8.66205752e-01 1.52396035e+00
7.15369046e-01 7.80731440e-02 -1.02811441e-01 9.58064437e-01
2.36362323e-01 5.66417992e-01 -1.51106983e-01 -3.12973082e-01
1.19337153e+00 -9.52207804e-01 -5.53706944e-01 -3.34726036e-01
7.10341632e-01 -4.14458394e-01 1.25582039e+00 6.04501426e-01
-9.96089041e-01 -4.95970577e-01 -1.32711220e+00 4.20682788e-01
-2.05437556e-01 8.88109356e-02 3.87071818e-01 6.75781369e-01
-4.52902466e-01 6.32135391e-01 -7.76403189e-01 2.14030966e-01
7.49200046e-01 2.26549223e-01 2.03125581e-01 -4.14939046e-01
-1.27791893e+00 4.39495027e-01 4.96498585e-01 5.76925166e-02
-8.27889621e-01 -7.27734625e-01 -6.02106333e-01 -1.45534454e-02
7.30266988e-01 -6.51469707e-01 1.18808508e+00 -9.16015625e-01
-1.37438655e+00 4.51240957e-01 -1.44953996e-01 -4.74863976e-01
9.34878349e-01 -1.87989384e-01 -7.66260698e-02 -2.08774537e-01
1.32868752e-01 4.10906702e-01 8.43107998e-01 -1.19141757e+00
-8.75620127e-01 -4.73134071e-01 -4.49924543e-02 2.64277756e-01
-2.96525270e-01 -4.58925247e-01 -1.74723610e-01 -6.81626916e-01
-3.09286229e-02 -8.74982715e-01 -5.49840093e-01 -4.59232032e-01
-6.03823185e-01 -2.85495728e-01 6.93892002e-01 -3.01335365e-01
1.42415738e+00 -1.94487250e+00 -5.35931140e-02 1.34684622e-01
2.40324169e-01 3.82497460e-01 5.89550883e-02 2.28820920e-01
6.66853264e-02 4.52433974e-01 -2.49961004e-01 -2.57319033e-01
-5.45511656e-02 2.28494452e-03 -2.72784412e-01 2.55692840e-01
2.40506262e-01 5.09012878e-01 -1.02800047e+00 -2.05494270e-01
4.99052294e-02 2.08287194e-01 -7.32360363e-01 4.07455593e-01
-4.43596482e-01 2.26125032e-01 -5.71687341e-01 7.04967856e-01
8.43224168e-01 -3.34468812e-01 2.93270677e-01 -2.66507436e-02
2.39897177e-01 2.35438213e-01 -1.51196563e+00 1.00467360e+00
-3.97508115e-01 1.19307846e-01 -1.97768405e-01 -1.28956354e+00
9.81432557e-01 7.14079244e-03 4.36671853e-01 -7.45712817e-01
6.91026896e-02 4.30887669e-01 3.04935314e-02 -4.87843633e-01
3.71878326e-01 -3.02643061e-01 -1.26030862e-01 5.80381870e-01
-5.07609069e-01 -1.51091889e-01 3.97456974e-01 -2.34074265e-01
1.15554702e+00 -6.65442199e-02 2.85213828e-01 -3.51466127e-02
3.53306860e-01 -2.66630296e-02 1.05373847e+00 8.59740913e-01
-2.53620595e-01 6.43354118e-01 8.46166134e-01 -6.49493873e-01
-1.00130320e+00 -7.23366559e-01 -6.45553097e-02 9.53067303e-01
1.98321342e-01 -3.26608866e-01 -8.48732412e-01 -8.42099547e-01
4.30489667e-02 7.41509616e-01 -7.39162683e-01 -3.23772877e-01
-3.90164793e-01 -1.27410448e+00 1.99565724e-01 2.11239204e-01
5.10336041e-01 -1.20318627e+00 -8.44811618e-01 3.07319462e-01
-2.76438683e-01 -8.87895823e-01 -2.50540078e-01 2.45100871e-01
-8.10643494e-01 -1.27777028e+00 -4.52765405e-01 -3.21406007e-01
7.33613253e-01 1.03733227e-01 1.31696129e+00 1.74490079e-01
-4.31654036e-01 -4.47535753e-01 -4.86414999e-01 -7.58986294e-01
-4.29754823e-01 5.12470603e-01 2.96249939e-03 1.06283821e-01
5.04157543e-01 -4.59427148e-01 -8.57147753e-01 5.23326755e-01
-9.72445011e-01 2.75469571e-01 4.79173332e-01 1.04123116e+00
8.36284280e-01 4.14189667e-01 9.72661138e-01 -1.30637372e+00
6.94233835e-01 -6.50029361e-01 -5.97474992e-01 2.30398566e-01
-6.34766519e-01 1.55193508e-01 9.14603591e-01 -4.75266606e-01
-5.21994889e-01 -2.50791967e-01 -9.69135854e-03 -6.09303117e-01
1.56387128e-03 3.02649736e-01 -3.85832071e-01 5.97784519e-01
7.41656363e-01 1.56160071e-01 1.19037822e-01 -2.59517729e-01
-6.04237244e-02 8.02264988e-01 -1.55179486e-01 -7.58152008e-01
4.52023596e-01 2.59330809e-01 -2.75813103e-01 -4.14373875e-01
-1.06950188e+00 -1.53008610e-01 -2.40018532e-01 1.45770507e-02
3.80089939e-01 -9.37204182e-01 -6.01157904e-01 5.81006229e-01
-5.77845454e-01 -6.39725566e-01 -3.12415242e-01 4.25157957e-02
-6.68306887e-01 -1.42451540e-01 -2.04168886e-01 -7.11045444e-01
-3.86003852e-01 -1.56488144e+00 9.28863883e-01 3.51095498e-01
-2.20241293e-01 -5.90279341e-01 -6.82025328e-02 6.99239671e-01
2.27382451e-01 4.93465543e-01 9.23542440e-01 -5.65560460e-01
-4.87681150e-01 -2.31628790e-01 4.93170545e-02 2.84541637e-01
3.48916829e-01 -3.49840410e-02 -6.79761827e-01 -5.48104465e-01
-1.92548349e-01 -6.67690217e-01 6.22279882e-01 3.17915261e-01
1.85469174e+00 -5.06889462e-01 -2.28085518e-01 6.27574921e-01
1.29731107e+00 1.38731301e-01 5.57568610e-01 6.32065296e-01
5.08557796e-01 4.57526296e-01 1.00387251e+00 7.97954202e-01
5.31570256e-01 6.81910038e-01 4.77170348e-01 3.50570083e-02
2.11110532e-01 -3.15235138e-01 1.30729437e-01 3.97445023e-01
3.41596305e-01 -1.03653774e-01 -9.30458367e-01 5.30423045e-01
-1.90511990e+00 -8.58369589e-01 2.54945606e-01 2.38516855e+00
1.28456736e+00 2.50615388e-01 4.43801403e-01 4.95853961e-01
7.69319654e-01 1.16421811e-01 -1.04022539e+00 -3.30258965e-01
2.42901430e-01 7.90440217e-02 2.72559047e-01 3.30328643e-01
-9.89769757e-01 7.43375480e-01 5.48036385e+00 1.11743188e+00
-1.28842187e+00 -2.06542134e-01 1.20865881e+00 -5.10456145e-01
-2.88497597e-01 -7.26008266e-02 -1.10047567e+00 9.31347311e-01
7.52106428e-01 -2.23281398e-01 4.75098491e-01 7.19733536e-01
3.50667030e-01 -3.11508447e-01 -1.03221583e+00 8.07927728e-01
-1.86366856e-01 -1.25142920e+00 3.31139006e-02 9.93810222e-02
9.64116216e-01 -9.25460085e-02 -1.08104639e-01 5.99302471e-01
5.44250667e-01 -1.10754025e+00 6.25046194e-01 1.95771679e-01
5.46903431e-01 -1.15524423e+00 6.81760252e-01 7.38343060e-01
-7.21713841e-01 -3.83430839e-01 -4.59116101e-01 -1.39697799e-02
-1.90197125e-01 1.19618118e+00 -1.01287735e+00 3.98902059e-01
7.20568836e-01 3.78612995e-01 -5.27298927e-01 1.02650714e+00
-2.63938457e-01 7.74061620e-01 -2.07757145e-01 -1.09404087e-01
-1.52444661e-01 -1.66734964e-01 2.19654337e-01 7.82116890e-01
2.97388881e-01 -1.10904716e-01 4.22491878e-01 6.56125665e-01
-2.35719666e-01 1.49756894e-01 -2.27175713e-01 2.59326845e-01
8.98299038e-01 9.46709991e-01 -5.74730694e-01 -3.18062574e-01
1.16170846e-01 4.04668182e-01 4.80356574e-01 2.32890382e-01
-8.36871684e-01 -4.18556213e-01 6.73534036e-01 3.91217411e-01
3.20627302e-01 3.47376883e-01 -6.40835762e-01 -1.09958649e+00
2.68601447e-01 -1.54007733e+00 4.56414431e-01 -3.13518286e-01
-1.32954586e+00 3.57722491e-01 -1.05375394e-01 -1.50593960e+00
-2.68267393e-01 -1.27131358e-01 -5.37207603e-01 8.72891307e-01
-1.47204340e+00 -4.74238276e-01 -4.02096748e-01 7.01666772e-02
6.79188967e-01 -1.03837542e-01 4.65702683e-01 1.37222469e-01
-1.01750314e+00 7.90192664e-01 1.25758991e-01 -5.93426228e-02
6.78249538e-01 -1.10501301e+00 2.05781609e-01 5.39511323e-01
-2.86011517e-01 2.21164674e-01 7.68401027e-01 -4.65975583e-01
-1.13755202e+00 -1.22668362e+00 4.04384047e-01 -1.17073059e-01
2.26925328e-01 -3.98763388e-01 -1.03332627e+00 1.49014726e-01
-2.89150864e-01 2.27730051e-01 8.06381702e-01 1.57565996e-01
-9.83629301e-02 -4.24249142e-01 -1.27747130e+00 6.36372387e-01
8.09562385e-01 2.34534461e-02 -3.06135684e-01 2.43143901e-01
4.92282420e-01 -4.96865749e-01 -7.49637604e-01 7.15068758e-01
4.05788809e-01 -1.31048417e+00 7.82189548e-01 -5.01770139e-01
7.57313609e-01 -3.59040707e-01 -5.93428984e-02 -1.64961135e+00
3.46051194e-02 -2.45741889e-01 -3.13876450e-01 1.24265528e+00
3.42249781e-01 -7.34763980e-01 1.05111957e+00 4.95799392e-01
3.95496577e-01 -1.53414679e+00 -8.21555197e-01 -5.51376820e-01
2.23764569e-01 -2.80786097e-01 1.08692408e+00 8.21221709e-01
-2.40018308e-01 2.11174801e-01 -3.99401754e-01 3.57619487e-03
6.68203533e-01 4.06020343e-01 1.08130610e+00 -1.05068839e+00
-4.54641134e-01 -5.29702842e-01 1.32810920e-02 -7.54048586e-01
-6.28488958e-02 -5.38543582e-01 1.68685704e-01 -1.24344969e+00
2.00524300e-01 -9.37412083e-01 -2.57542431e-01 6.50424421e-01
-7.95015097e-01 1.07776955e-01 1.91991314e-01 3.03119183e-01
-5.43777347e-01 8.88144732e-01 1.24957681e+00 -1.76024050e-01
-4.67902362e-01 3.36955220e-01 -1.06500030e+00 5.02098203e-01
9.84369278e-01 -6.64569497e-01 -4.70141053e-01 -1.52549401e-01
2.96752244e-01 2.10446119e-01 -1.49903502e-02 -1.09439671e+00
-1.17061064e-01 -5.77630997e-01 4.83636945e-01 -6.35255635e-01
-9.60906744e-02 -4.25154984e-01 5.35678752e-02 6.46298170e-01
-3.88506651e-01 7.55960718e-02 4.15224535e-03 4.15549368e-01
-2.60623962e-01 -1.94472298e-01 9.91971672e-01 -1.91158324e-01
-9.92645696e-02 3.93752038e-01 -1.27703995e-01 4.88592446e-01
1.20800674e+00 -1.46538898e-01 -1.92886487e-01 -1.09397098e-01
-2.60066718e-01 7.14063585e-01 5.14306009e-01 5.00903964e-01
3.90914351e-01 -1.19545269e+00 -7.65522420e-01 3.00757289e-01
8.35058242e-02 7.73847580e-01 2.47674301e-01 5.36889255e-01
-3.81672710e-01 -5.54914549e-02 -1.34277105e-01 -5.74582517e-01
-1.00536668e+00 4.29178149e-01 4.41744387e-01 -8.00933480e-01
-2.20504507e-01 5.56443870e-01 -1.71217937e-02 -5.96216619e-01
7.77983293e-02 -1.03276037e-01 -1.17429920e-01 3.27953815e-01
6.07852876e-01 5.37306011e-01 2.87356853e-01 7.09353909e-02
-1.30625546e-01 1.66122258e-01 -3.53669882e-01 2.13557661e-01
1.45308876e+00 1.47436291e-01 5.59328832e-02 4.85332429e-01
9.20808733e-01 -1.24885105e-01 -1.42501962e+00 -1.34272590e-01
-1.58783928e-01 -7.46845961e-01 -1.47908464e-01 -7.18602180e-01
-1.19567537e+00 7.27915585e-01 3.86667877e-01 3.57515752e-01
1.26870418e+00 -5.01197517e-01 7.37945259e-01 6.52564764e-02
4.46871102e-01 -1.30245304e+00 2.85873652e-01 2.75811106e-01
6.77583933e-01 -1.39572775e+00 1.80123374e-01 -1.97262272e-01
-6.02272987e-01 8.23069453e-01 1.10145831e+00 -1.28948674e-01
4.40795571e-01 3.22611064e-01 2.09661931e-01 1.54032111e-01
-1.01520884e+00 1.75952569e-01 -2.04604045e-01 2.78811842e-01
1.80596262e-01 3.38071644e-01 -4.95250553e-01 6.48449123e-01
-4.92257923e-01 2.24162146e-01 5.53463399e-01 9.39449251e-01
-2.67537355e-01 -1.36983609e+00 -6.24974668e-01 7.81773686e-01
-3.92072409e-01 3.03139061e-01 -6.23782910e-03 4.48926419e-01
3.26392442e-01 8.67300868e-01 5.76978736e-02 -4.61514801e-01
3.95077795e-01 -8.26965347e-02 9.17196870e-02 -5.16778767e-01
-4.53686535e-01 -1.78982750e-01 -1.51890650e-01 -4.02718574e-01
-3.31820734e-02 -7.65557051e-01 -1.11162770e+00 -2.49704629e-01
-2.21588135e-01 3.46636951e-01 4.09361362e-01 9.37929094e-01
5.36672354e-01 7.45363057e-01 1.11610115e+00 -8.02372396e-01
-9.23346162e-01 -7.90070772e-01 -3.96089077e-01 4.45789129e-01
3.72438341e-01 -7.01079845e-01 -3.17795038e-01 -4.22764301e-01]
|
[8.991134643554688, 4.033498764038086]
|
12ce1907-c614-412a-b84b-f5f6121d6f49
|
a-kolmogorov-complexity-approach-to
| null | null |
https://openreview.net/forum?id=Bke7MANKvS
|
https://openreview.net/pdf?id=Bke7MANKvS
|
A Kolmogorov Complexity Approach to Generalization in Deep Learning
|
Deep artificial neural networks can achieve an extremely small difference between training and test accuracies on identically distributed training and test sets, which is a standard measure of generalization. However, the training and test sets may not be sufficiently representative of the empirical sample set, which consists of real-world input samples. When samples are drawn from an underrepresented or unrepresented subset during inference, the gap between the training and inference accuracies can be significant. To address this problem, we first reformulate a classification algorithm as a procedure for searching for a source code that maps input features to classes. We then derive a necessary and sufficient condition for generalization using a universal cognitive similarity metric, namely information distance, based on Kolmogorov complexity. Using this condition, we formulate an optimization problem to learn a more general classification function. To achieve this end, we extend the input features by concatenating encodings of them, and then train the classifier on the extended features. As an illustration of this idea, we focus on image classification, where we use channel codes on the input features as a systematic way to improve the degree to which the training and test sets are representative of the empirical sample set. To showcase our theoretical findings, considering that corrupted or perturbed input features belong to the empirical sample set, but typically not to the training and test sets, we demonstrate through extensive systematic experiments that, as a result of learning a more general classification function, a model trained on encoded input features is significantly more robust to common corruptions, e.g., Gaussian and shot noise, as well as adversarial perturbations, e.g., those found via projected gradient descent, than the model trained on uncoded input features.
|
['Brian Kingsbury', 'Kush R. Varshney', 'Hazar Yueksel']
|
2019-09-25
| null | null | null | null |
['classification']
|
['methodology']
|
[ 6.41955078e-01 -9.42695048e-03 9.61329639e-02 -3.96966875e-01
-6.87435746e-01 -7.56506860e-01 4.29042816e-01 1.20848298e-01
-5.59209287e-01 8.94259930e-01 -2.65381038e-01 -2.71469116e-01
-2.03486681e-01 -1.12457299e+00 -1.07341409e+00 -9.40184832e-01
-3.44862118e-02 8.98539945e-02 -5.13353609e-02 1.89746767e-01
2.31685996e-01 5.87811351e-01 -1.72757876e+00 1.38283819e-01
8.96202385e-01 1.04715395e+00 -1.42772987e-01 6.19293094e-01
2.04617143e-01 3.04334641e-01 -9.87577081e-01 -5.26525259e-01
3.43370736e-01 -7.55479574e-01 -7.74952710e-01 1.40458122e-01
2.84732223e-01 -1.29807040e-01 -3.55833232e-01 1.59562850e+00
3.52990746e-01 1.95568934e-01 1.01475835e+00 -1.27933848e+00
-9.60133016e-01 5.36296785e-01 1.93435580e-01 -9.94882639e-03
3.17359298e-01 6.79107606e-02 7.33777761e-01 -6.58462524e-01
3.82315874e-01 8.96397650e-01 5.52968085e-01 6.45595551e-01
-1.46199465e+00 -6.43328726e-01 -1.83985293e-01 1.75832734e-01
-1.45054412e+00 -3.15231681e-01 6.58933282e-01 -3.32123071e-01
1.92919046e-01 3.58507365e-01 2.99369007e-01 1.31416857e+00
2.01173142e-01 4.61621940e-01 1.19331539e+00 -5.88035583e-01
7.05647528e-01 5.77046812e-01 2.19818026e-01 4.62446004e-01
5.32567978e-01 3.05410236e-01 -6.52621835e-02 -2.60400563e-01
1.78386793e-01 -3.21476832e-02 -8.16899776e-01 -3.02540779e-01
-7.20784843e-01 8.96044910e-01 4.49179620e-01 4.37645972e-01
-1.92358136e-01 -7.30640255e-03 2.42827684e-01 6.86490119e-01
1.00079127e-01 5.35268903e-01 -3.04943591e-01 6.26965091e-02
-5.85327625e-01 1.70435756e-02 9.24075365e-01 6.09815300e-01
1.00792384e+00 -9.11118910e-02 -8.38868171e-02 5.60865104e-01
-1.27129480e-01 6.66036844e-01 7.68545270e-01 -5.75693130e-01
3.03950548e-01 2.85750628e-01 -2.74844579e-02 -1.08427811e+00
-7.49317929e-02 -6.13513827e-01 -8.95280182e-01 2.53773391e-01
7.12179244e-01 -2.07401991e-01 -7.75390089e-01 2.17581749e+00
-2.10617743e-02 6.82683289e-02 4.80137706e-01 6.06359303e-01
3.85149509e-01 4.45330054e-01 -3.08679551e-01 -1.89984694e-01
8.67039919e-01 -1.97836250e-01 -3.63017201e-01 -1.24004535e-01
8.11294675e-01 -2.70009786e-01 1.26055264e+00 3.54591817e-01
-7.59302139e-01 -5.72949946e-01 -1.37158835e+00 4.30539101e-01
-5.54938078e-01 -2.10822579e-02 2.00298518e-01 1.19218731e+00
-6.79220200e-01 9.25603032e-01 -4.59732652e-01 -1.53487414e-01
5.21792889e-01 1.76603407e-01 -5.94522178e-01 -3.20565939e-01
-1.25854790e+00 6.84372187e-01 5.41186750e-01 -4.80683148e-02
-7.89506018e-01 -4.31115091e-01 -7.52665102e-01 2.59935200e-01
1.03253283e-01 -3.26087207e-01 8.79407823e-01 -1.26131320e+00
-1.28720593e+00 6.75361276e-01 2.67618243e-02 -5.42227983e-01
4.92413700e-01 3.04948509e-01 -4.64978009e-01 5.90747483e-02
-2.19035372e-01 1.03844687e-01 9.81047034e-01 -1.43608046e+00
-3.41871351e-01 -4.34461653e-01 -5.91119577e-04 -2.21982107e-01
-6.45163655e-01 -3.86148274e-01 9.64549929e-02 -5.34437597e-01
7.40689831e-03 -8.54509234e-01 9.21262279e-02 -8.18996653e-02
-6.11312211e-01 2.15714857e-01 5.89073718e-01 -3.45352978e-01
9.35281157e-01 -2.43339062e+00 2.51528826e-02 7.30214894e-01
-1.36773407e-01 2.78919756e-01 -1.89206019e-01 1.26899928e-01
-2.94509679e-01 2.80403554e-01 -6.20273829e-01 -8.22228342e-02
-4.65062167e-03 3.68312746e-01 -4.42883104e-01 6.89113259e-01
2.66116738e-01 5.72676837e-01 -9.04959023e-01 1.34893537e-01
-1.89624161e-01 2.98106581e-01 -5.20461619e-01 3.59765515e-02
7.95295462e-02 4.56952542e-01 -3.49873394e-01 1.77959263e-01
6.14700556e-01 -7.77929202e-02 -1.27967566e-01 9.12834555e-02
4.98682678e-01 2.17302367e-02 -1.11550951e+00 1.12803233e+00
-5.28052449e-01 7.00343668e-01 -4.84826654e-01 -1.61367965e+00
1.08147597e+00 1.59641519e-01 -2.96925120e-02 -4.57065582e-01
4.24208790e-01 3.50465417e-01 3.23349714e-01 -4.55197513e-01
8.41206089e-02 -3.39204252e-01 -2.69685119e-01 5.61526537e-01
1.85978442e-01 9.33146477e-02 -1.14970967e-01 -1.66559398e-01
1.19540358e+00 -5.30249536e-01 2.33131289e-01 -3.17897797e-02
5.73064446e-01 -4.93550390e-01 3.91036808e-01 1.16904032e+00
1.19517194e-02 5.54587245e-01 6.72699392e-01 -9.06267464e-02
-9.83352721e-01 -1.20902050e+00 -5.46387970e-01 3.85065675e-01
-3.59195583e-02 -6.00469857e-02 -9.26411152e-01 -1.01616001e+00
9.03963521e-02 7.86567748e-01 -8.47391665e-01 -9.39120770e-01
-1.22287255e-02 -7.61091173e-01 7.70636201e-01 3.88597071e-01
6.17083073e-01 -8.33745658e-01 -6.68647647e-01 -1.12454295e-01
8.89877602e-02 -9.15737450e-01 -2.90510133e-02 4.02209044e-01
-4.78505880e-01 -1.11096442e+00 -5.78574538e-01 -6.16253495e-01
9.05411065e-01 -1.88499629e-01 8.03155303e-01 1.16115794e-01
-1.00138314e-01 3.52380186e-01 -4.40272927e-01 -3.17872107e-01
-9.08048272e-01 -2.81319797e-01 2.30485052e-01 3.43130350e-01
2.97783047e-01 -6.66495442e-01 -2.26877689e-01 2.27821738e-01
-1.24722230e+00 -4.49414909e-01 4.90890235e-01 9.94895339e-01
3.30828696e-01 5.07736027e-01 6.98029399e-01 -6.85947776e-01
6.69745922e-01 -5.59074700e-01 -4.68524069e-01 2.83615261e-01
-3.57249856e-01 3.45885009e-01 1.20751429e+00 -6.86134398e-01
-5.55625021e-01 -9.99074951e-02 -5.40084727e-02 -2.72443265e-01
-4.34198767e-01 6.89622760e-01 -4.83783573e-01 -2.75912404e-01
9.79246259e-01 6.48558259e-01 3.01348232e-02 -1.79238468e-01
1.66274339e-01 1.00587308e+00 6.30911469e-01 -5.79771519e-01
9.10752475e-01 3.08588117e-01 8.03562105e-02 -7.07786024e-01
-6.50135636e-01 1.97267398e-01 -5.18731177e-01 -5.83969466e-02
4.25894886e-01 -2.45919749e-01 -5.41312397e-01 4.21221137e-01
-8.97843421e-01 -2.58561403e-01 -5.60175300e-01 6.20274067e-01
-5.98180354e-01 3.42887878e-01 -1.84573978e-01 -7.75296986e-01
1.95697159e-01 -1.15201390e+00 4.61437225e-01 -4.65183798e-03
-7.37942681e-02 -9.65021133e-01 -1.38717070e-01 -2.48551011e-01
2.00607434e-01 3.68445784e-01 1.27022862e+00 -1.28953016e+00
-2.50726402e-01 -8.29581499e-01 5.59517592e-02 1.02567673e+00
2.64243960e-01 -1.46510929e-01 -1.02036476e+00 -4.36252028e-01
3.55118603e-01 -3.16867411e-01 8.60245585e-01 5.17244637e-02
1.51624727e+00 -4.77879614e-01 -1.34242162e-01 6.65655375e-01
1.40707278e+00 3.31720799e-01 7.11629987e-01 2.23335445e-01
1.93850055e-01 3.97318900e-01 1.82944328e-01 3.30933243e-01
-2.65358746e-01 4.73731250e-01 3.87834400e-01 4.09101814e-01
1.40686423e-01 -2.58767545e-01 3.72148097e-01 4.36655879e-01
2.93521971e-01 -2.32715726e-01 -7.07768738e-01 3.41612309e-01
-1.29969025e+00 -1.01195335e+00 3.51087540e-01 2.92059708e+00
8.91330004e-01 2.69172817e-01 -1.02425121e-01 7.78317392e-01
8.31439734e-01 -3.00274402e-01 -6.35559976e-01 -4.06772882e-01
-1.42168224e-01 4.31245714e-01 4.10854042e-01 2.44916588e-01
-1.00635374e+00 3.00360262e-01 5.91669941e+00 8.10801506e-01
-1.32441521e+00 -1.07233472e-01 7.45829344e-01 9.19648260e-02
-3.75937074e-01 -1.79430291e-01 -4.10307854e-01 8.41675997e-01
1.11237574e+00 -3.90286237e-01 5.84174931e-01 7.33785748e-01
-2.58266270e-01 9.80549231e-02 -1.48053384e+00 8.19760442e-01
1.65784344e-01 -1.03871524e+00 7.19029009e-02 2.19294876e-01
7.81043887e-01 -3.50680381e-01 3.47793311e-01 4.14575964e-01
1.07047148e-01 -1.26486897e+00 7.25805163e-01 4.07081187e-01
7.48184860e-01 -8.14122796e-01 9.35669422e-01 5.83493114e-01
-5.24815202e-01 -3.55115980e-01 -5.53640246e-01 -6.73813671e-02
-4.21884328e-01 6.38014376e-01 -4.04638648e-01 3.87763768e-01
2.54016876e-01 3.87963325e-01 -4.90005553e-01 1.19341803e+00
-2.61760414e-01 6.02187991e-01 -2.34296918e-01 -1.26669690e-01
2.51471880e-03 4.46036085e-02 3.84618104e-01 8.77617836e-01
6.18895590e-01 -5.72415143e-02 -2.83719916e-02 9.23422992e-01
-4.49146986e-01 -1.00017533e-01 -8.51207495e-01 2.40677632e-02
7.15544641e-01 7.42451131e-01 -5.53376317e-01 -3.46876681e-01
-3.68543267e-01 1.07507288e+00 3.64412516e-01 6.39812529e-01
-8.52104843e-01 -8.90596747e-01 6.00829244e-01 -3.55172366e-01
3.74216139e-01 8.21187869e-02 -2.68602580e-01 -1.13176119e+00
2.83752263e-01 -9.19025362e-01 8.42503160e-02 -3.54228884e-01
-1.30277765e+00 6.03068709e-01 -1.42609984e-01 -1.38522398e+00
-2.33402461e-01 -6.94995642e-01 -6.35268152e-01 9.07725990e-01
-1.05358505e+00 -4.05364633e-01 -2.40958109e-01 6.79232180e-01
-1.82047516e-01 -1.66048571e-01 1.07444763e+00 8.96426886e-02
-4.95270818e-01 1.09056270e+00 5.79037547e-01 4.69436407e-01
3.11573297e-01 -1.03169370e+00 -2.60661822e-02 9.70677018e-01
3.38630021e-01 5.15927255e-01 6.68414354e-01 -2.23184824e-01
-1.09913623e+00 -1.25017071e+00 4.76610482e-01 -1.88433528e-01
5.61385453e-01 -3.15290064e-01 -1.15144384e+00 6.71073854e-01
-4.84348148e-01 3.82371217e-01 6.99025691e-01 -1.26130462e-01
-6.14349008e-01 -5.40837012e-02 -1.56659067e+00 4.90167081e-01
9.51098323e-01 -7.86269009e-01 -6.17060363e-01 1.99471489e-01
4.05643076e-01 -1.12711571e-01 -8.36865604e-01 3.01014721e-01
4.52077806e-01 -9.43611860e-01 7.36260474e-01 -7.43193865e-01
5.74720383e-01 -8.75640735e-02 -6.04301631e-01 -1.68243504e+00
-1.18296742e-01 -2.43475348e-01 8.86237398e-02 1.11521995e+00
4.00096774e-01 -9.47816551e-01 7.06212282e-01 4.38774347e-01
1.35414764e-01 -7.28117645e-01 -1.23126948e+00 -1.18524528e+00
3.72649074e-01 -6.77611351e-01 7.21218705e-01 8.57768774e-01
1.09523889e-02 2.67761610e-02 -2.45676860e-02 3.86479765e-01
5.52249014e-01 5.87131381e-02 5.97835302e-01 -1.21996379e+00
-5.11853814e-01 -4.44911689e-01 -1.00293708e+00 -7.73775637e-01
5.21906257e-01 -1.10563231e+00 1.69242546e-01 -8.30388308e-01
1.04984201e-01 -5.62591553e-01 -4.59837675e-01 2.61245549e-01
-8.52650926e-02 3.33379179e-01 1.69282675e-01 -2.08516158e-02
7.71286711e-02 5.33192635e-01 1.02004707e+00 -1.93118408e-01
-3.99246551e-02 3.63458842e-01 -8.01002026e-01 6.68151557e-01
8.29538763e-01 -5.89585483e-01 -4.04652596e-01 -1.12136476e-01
7.54742920e-02 -5.33605516e-02 7.03517497e-01 -1.36831498e+00
-4.43873964e-02 9.28000733e-02 5.08874536e-01 1.44590497e-01
1.82227835e-01 -8.88476968e-01 -3.86953689e-02 6.75768137e-01
-7.63225436e-01 -5.61135471e-01 -1.35988705e-02 6.63960636e-01
-1.01611488e-01 -8.24691176e-01 7.93644547e-01 1.90606609e-01
-2.52165109e-01 1.49063081e-01 -4.60075766e-01 4.13932167e-02
9.54305232e-01 -3.93521905e-01 -3.01607937e-01 -4.44251984e-01
-7.68027544e-01 -3.87430161e-01 5.41141689e-01 8.92101750e-02
7.20914841e-01 -1.36145210e+00 -6.00723922e-01 5.73222399e-01
2.17723444e-01 -3.40039045e-01 -1.49499282e-01 6.28350377e-01
3.45859826e-02 2.54091173e-01 -1.37438431e-01 -4.02297169e-01
-1.02027309e+00 7.64410734e-01 4.34456170e-01 1.47248268e-01
-3.74941736e-01 7.32707381e-01 1.29881212e-02 -3.41620505e-01
3.10486436e-01 -5.46156883e-01 1.35137632e-01 -2.49310970e-01
6.54483736e-01 2.36022435e-02 1.98399037e-01 -4.58231002e-01
-2.06090525e-01 4.87659216e-01 2.02548817e-01 2.22946592e-02
9.00926471e-01 7.27175698e-02 1.46084368e-01 5.21578908e-01
1.89252293e+00 2.83082407e-02 -1.00485408e+00 -3.50936919e-01
-1.13818593e-01 -6.42644703e-01 -1.69576585e-01 -5.49667001e-01
-1.04325616e+00 8.90900135e-01 7.63179779e-01 7.13969707e-01
1.24904108e+00 -5.90924099e-02 4.06615406e-01 7.41312385e-01
4.88306880e-01 -6.22684717e-01 -2.09611785e-02 4.66632336e-01
7.17056394e-01 -1.00090289e+00 -4.59912866e-01 -1.32519186e-01
-2.05626041e-01 1.16401625e+00 1.36005878e-01 -3.18458050e-01
6.53973758e-01 1.00759625e-01 -2.87776381e-01 3.23627651e-01
-4.29408401e-01 -9.23231989e-02 1.20208107e-01 8.29161704e-01
-2.06455495e-03 7.60232806e-02 -1.86033711e-01 7.58972347e-01
-4.55062628e-01 5.91119826e-02 6.29171610e-01 6.70168936e-01
-5.48895121e-01 -8.61936092e-01 -5.21164298e-01 5.65416455e-01
-3.61340880e-01 -3.23911496e-02 -2.36694202e-01 6.04045510e-01
1.52179360e-01 8.66326511e-01 1.75420016e-01 -7.40647197e-01
2.11634561e-01 1.87120304e-01 5.88076413e-01 -5.09106398e-01
-8.03749338e-02 -8.59319210e-01 -1.83809519e-01 -1.95318699e-01
-5.75038455e-02 -4.06476676e-01 -1.05198371e+00 -2.49854788e-01
-3.76878977e-01 2.44724914e-01 6.20885849e-01 1.23569155e+00
4.50643711e-02 5.35942733e-01 9.27893043e-01 -7.13396966e-01
-1.26105726e+00 -7.49730527e-01 -6.78236425e-01 7.94929504e-01
4.22866821e-01 -5.68426013e-01 -8.21466029e-01 -1.06362022e-01]
|
[5.718148708343506, 7.708550453186035]
|
04c3a8e6-ec64-4b98-8cdf-535a79a9218b
|
few-shot-text-independent-speaker
|
2008.11088
| null |
https://arxiv.org/abs/2008.11088v1
|
https://arxiv.org/pdf/2008.11088v1.pdf
|
Few Shot Text-Independent speaker verification using 3D-CNN
|
Facial recognition system is one of the major successes of Artificial intelligence and has been used a lot over the last years. But, images are not the only biometric present: audio is another possible biometric that can be used as an alternative to the existing recognition systems. However, the text-independent audio data is not always available for tasks like speaker verification and also no work has been done in the past for text-independent speaker verification assuming very little training data. Therefore, In this paper, we have proposed a novel method to verify the identity of the claimed speaker using very few training data. To achieve this we are using a Siamese neural network with center loss and speaker bias loss. Experiments conducted on the VoxCeleb1 dataset show that the proposed model accuracy even on training with very few data is near to the state of the art model on text-independent speaker verification
|
['Prateek Mishra']
|
2020-08-25
| null | null | null | null |
['text-independent-speaker-verification']
|
['speech']
|
[ 1.04989350e-01 -1.57429531e-01 1.96464099e-02 -6.75538063e-01
-6.12557530e-01 -1.33743942e-01 5.14230430e-01 -1.51570201e-01
-4.41359997e-01 7.56964087e-01 6.05957806e-02 2.49602832e-02
-1.35263875e-01 -2.14905411e-01 -4.73031431e-01 -9.78523850e-01
2.62224853e-01 6.01323307e-01 1.16258226e-02 -1.58220738e-01
3.38203877e-01 5.96248865e-01 -1.89996374e+00 4.09683064e-02
6.85440123e-01 9.51983452e-01 -2.71562576e-01 4.23125356e-01
-1.17303707e-01 3.75044078e-01 -5.81490695e-01 -4.47177559e-01
3.67055178e-01 -6.12173259e-01 -5.83670974e-01 2.54155863e-02
4.45382327e-01 -1.71856537e-01 -1.46915726e-02 9.91807163e-01
8.85113060e-01 2.91908272e-02 6.07847929e-01 -1.46326137e+00
-2.88683176e-01 5.47232687e-01 -6.95680261e-01 1.55529737e-01
3.40522647e-01 -3.38472992e-01 5.32342732e-01 -9.43606496e-01
2.85705298e-01 1.37384629e+00 5.35391212e-01 6.43651903e-01
-8.91792953e-01 -1.03612673e+00 -2.41793931e-01 5.85831523e-01
-1.88197982e+00 -8.64009380e-01 1.00757110e+00 -1.06123507e-01
4.50987279e-01 9.40448865e-02 3.70588332e-01 8.79930258e-01
-2.51535118e-01 7.48493671e-01 1.31732869e+00 -6.09804094e-01
5.13806604e-02 7.56774664e-01 1.39029130e-01 6.07694089e-01
-4.00149189e-02 -9.66405794e-02 -8.50638866e-01 2.41651814e-02
3.06873858e-01 -2.95817912e-01 -2.55103499e-01 -1.90386817e-01
-8.46510351e-01 6.68994904e-01 2.75594387e-02 6.50900424e-01
-5.33664599e-02 -1.29905269e-01 5.34547687e-01 4.28490788e-01
3.90064567e-01 -2.81656146e-01 -1.69760156e-02 -1.78952366e-01
-1.42553234e+00 -3.98242623e-02 7.75097311e-01 4.77976412e-01
3.94450366e-01 3.07500511e-01 3.53634328e-01 8.40981960e-01
4.21621799e-01 6.72107577e-01 7.04403520e-01 -3.12305480e-01
2.21706942e-01 6.35663092e-01 -1.10488959e-01 -8.23417604e-01
-9.78437141e-02 -2.81508178e-01 -9.04558480e-01 3.21268737e-01
6.34892821e-01 -6.19869903e-02 -8.61054420e-01 1.45742583e+00
3.46133977e-01 3.33790153e-01 4.99594100e-02 8.73017073e-01
8.14212143e-01 5.35552919e-01 -2.29749143e-01 -1.53691456e-01
1.19887555e+00 -5.62993884e-01 -8.17895830e-01 2.12490499e-01
4.95328046e-02 -1.18770421e+00 6.13411307e-01 7.36952662e-01
-8.26872289e-01 -5.60185015e-01 -1.11789501e+00 2.60997236e-01
-4.70158607e-01 1.91398442e-01 4.10761893e-01 1.51735365e+00
-1.02494693e+00 2.25940138e-01 -5.45928240e-01 -6.90431595e-01
3.60771149e-01 5.39534748e-01 -8.38877141e-01 -1.24324679e-01
-1.06528401e+00 1.02062201e+00 1.92282975e-01 2.42939278e-01
-7.39147067e-01 -1.69815436e-01 -5.71215451e-01 3.39847431e-03
2.01522872e-01 -1.87364653e-01 7.54109263e-01 -1.34996152e+00
-1.59572470e+00 8.64372730e-01 -5.52063406e-01 -5.75636089e-01
5.94535291e-01 1.86022773e-01 -6.87147439e-01 4.50410210e-02
-3.66554052e-01 5.27606905e-01 1.19130969e+00 -8.37756276e-01
-4.19870347e-01 -6.66921437e-01 -4.98635948e-01 -2.86728726e-03
-3.96435022e-01 4.50992525e-01 6.79370090e-02 -3.28540415e-01
2.28159919e-01 -8.65778446e-01 6.01087213e-01 2.42499765e-02
-3.33228469e-01 -2.92775899e-01 1.11667585e+00 -8.23638856e-01
7.69435406e-01 -2.34931350e+00 -8.90768096e-02 3.10154647e-01
-5.42068243e-01 4.29831713e-01 9.39093009e-02 3.51389170e-01
-2.89595962e-01 3.76229361e-02 -1.65023327e-01 -5.71967006e-01
-1.04421124e-01 1.47777889e-02 -1.16822109e-01 7.13718891e-01
7.77212437e-03 2.57158279e-01 -1.68511331e-01 -6.09860599e-01
1.91487148e-01 8.85422349e-01 -1.76696151e-01 -2.29147077e-01
3.90030950e-01 5.10000110e-01 5.03130117e-03 9.37338769e-01
9.64469969e-01 3.53431910e-01 -1.94346920e-01 -7.81721696e-02
-8.28429125e-03 -1.97773352e-01 -1.67166889e+00 1.42370331e+00
-2.15637550e-01 7.42394149e-01 3.79869372e-01 -1.27185643e+00
1.01986456e+00 8.81217062e-01 4.00999159e-01 -2.55417138e-01
3.22392076e-01 4.13685828e-01 3.05368721e-01 -5.15229940e-01
9.86402482e-02 -6.24341011e-01 4.14056122e-01 3.33740443e-01
2.59722859e-01 1.83599725e-01 5.76061308e-02 -2.30805352e-01
3.40560079e-01 -1.10059530e-01 8.35212842e-02 1.55564388e-02
1.11545146e+00 -3.90893787e-01 5.66003442e-01 2.88855970e-01
-5.60387075e-01 6.92104638e-01 9.23192725e-02 -1.77393556e-01
-8.47301900e-01 -6.82233155e-01 -4.87239242e-01 6.17554188e-01
-1.82487428e-01 1.87071726e-01 -8.41403782e-01 -4.85051960e-01
1.24944165e-01 4.23006117e-01 -5.09635448e-01 2.69064188e-01
-2.52087265e-01 -5.76790810e-01 1.16064942e+00 2.07912207e-01
8.69919777e-01 -8.19446921e-01 -9.29522365e-02 4.05177101e-03
-6.04933612e-02 -9.36864793e-01 -3.85629803e-01 -1.36300877e-01
-7.56687164e-01 -9.37292755e-01 -1.02837205e+00 -7.23614037e-01
6.10707700e-01 1.01781003e-01 3.68361771e-01 -1.30403787e-02
-3.65212023e-01 2.27848038e-01 -1.79729030e-01 -8.26670706e-01
-4.96896267e-01 -2.06748452e-02 3.95749629e-01 7.31359065e-01
6.62583947e-01 -4.11207616e-01 -2.63128847e-01 4.79206979e-01
-7.41217494e-01 -4.78045046e-01 5.15478969e-01 9.56410110e-01
1.31638750e-01 1.26885936e-01 1.06832063e+00 -5.47927439e-01
4.42080766e-01 -1.55935604e-02 -5.13695836e-01 2.15286821e-01
-5.24551809e-01 2.69115996e-02 4.13344681e-01 -4.10222769e-01
-1.17808497e+00 2.49348313e-01 -3.92057329e-01 -2.20922351e-01
-6.31060362e-01 3.98953646e-01 -4.81679946e-01 -5.11418879e-01
3.91668916e-01 5.12036920e-01 4.58400518e-01 -6.06508851e-01
-2.07403347e-01 1.13966024e+00 2.26656348e-01 -1.54185817e-01
7.01786816e-01 5.46667814e-01 2.25965276e-01 -1.29462707e+00
-2.50375897e-01 -8.53771806e-01 -6.02527797e-01 -1.99556649e-01
6.54492915e-01 -6.08328998e-01 -9.75660205e-01 8.52904379e-01
-9.55074608e-01 6.10973179e-01 8.73824582e-02 6.92787647e-01
-2.26631433e-01 5.21802545e-01 -5.35611399e-02 -1.42789447e+00
-5.33963144e-01 -1.12119365e+00 7.01269567e-01 3.02404016e-01
7.75124505e-02 -6.98724210e-01 -4.42916937e-02 5.97020030e-01
7.12066293e-01 -9.46137533e-02 5.20261407e-01 -9.53930497e-01
-4.27501917e-01 -7.45739043e-01 -8.30152109e-02 6.72619343e-01
4.35598522e-01 1.29539758e-01 -1.43263865e+00 -3.07410389e-01
3.08278948e-01 -3.06169033e-01 7.60326385e-01 2.37793028e-01
7.43314922e-01 -1.04665466e-01 -1.75857678e-01 3.20045948e-01
1.27132642e+00 1.97927937e-01 7.19559669e-01 -4.97435220e-02
2.13021725e-01 6.09597683e-01 4.01234180e-01 2.28961959e-01
6.93221688e-02 6.50477350e-01 1.90321639e-01 1.10397838e-01
-8.35674703e-02 -1.59418374e-01 3.76977891e-01 6.40134931e-01
-1.67368785e-01 -7.84998015e-02 -8.00545037e-01 6.60539746e-01
-1.42313087e+00 -1.40547621e+00 1.26345426e-01 2.55422449e+00
4.24887449e-01 -1.22633047e-01 3.23085725e-01 8.93148363e-01
8.95608783e-01 -1.13465115e-01 -2.60728449e-01 -2.44372487e-01
-2.25709900e-01 2.48824775e-01 2.64875025e-01 3.80200803e-01
-1.11677456e+00 5.89101672e-01 5.81635952e+00 7.24420607e-01
-1.71745133e+00 2.01082811e-01 3.15834254e-01 -2.01647785e-02
3.28162491e-01 -7.07114190e-02 -9.66228724e-01 5.19680142e-01
8.84049833e-01 -6.46764711e-02 2.38935679e-01 7.62502611e-01
3.33150715e-01 -2.46794701e-01 -8.62766266e-01 1.37908018e+00
7.14835048e-01 -7.85567522e-01 4.32580076e-02 -2.31085941e-02
4.22641426e-01 -2.54322410e-01 1.42737016e-01 -6.17895275e-02
-5.06374955e-01 -1.19465363e+00 4.66072917e-01 3.80027652e-01
4.72800732e-01 -1.00685132e+00 1.18579137e+00 5.68807721e-01
-9.45884645e-01 4.89585884e-02 -3.93516332e-01 2.62426376e-01
2.27585673e-01 2.82420337e-01 -1.16842830e+00 5.13686061e-01
6.13556623e-01 2.90191859e-01 -3.82017344e-01 1.43739307e+00
6.75469190e-02 8.64201128e-01 -6.14878118e-01 -2.33598024e-01
-9.26841721e-02 1.92665476e-02 6.60596669e-01 1.04871356e+00
6.00944042e-01 -2.98731953e-01 -2.69306779e-01 4.20401007e-01
-6.54819608e-02 6.29710436e-01 -7.41444945e-01 5.75398840e-03
1.98911160e-01 1.05978858e+00 -5.64043641e-01 -3.29586677e-02
-3.44643712e-01 9.04311240e-01 -2.81620800e-01 1.78112254e-01
-4.88021553e-01 -5.22584081e-01 2.75839984e-01 2.79058129e-01
2.31437460e-01 -1.22935399e-01 -1.50046134e-02 -9.01011288e-01
4.11897786e-02 -1.00395775e+00 3.05893302e-01 -4.22827542e-01
-1.17726529e+00 3.89408648e-01 -1.67079523e-01 -1.28003991e+00
-2.49674290e-01 -5.51656485e-01 -5.31361759e-01 1.15673470e+00
-1.60102630e+00 -1.40799975e+00 -2.22783163e-01 7.10470557e-01
4.17620748e-01 -6.71190262e-01 9.43282485e-01 5.10836065e-01
-3.21307600e-01 9.54544485e-01 2.14235052e-01 3.63988340e-01
1.11246479e+00 -7.49789774e-01 -2.05457792e-01 7.26510167e-01
4.09931511e-01 8.29782784e-01 8.20627272e-01 -3.73665333e-01
-1.36659813e+00 -3.14794779e-01 9.79333222e-01 -4.28178348e-02
2.70790756e-01 -1.62730217e-01 -9.95262980e-01 3.48501295e-01
4.43128824e-01 -1.00547411e-01 9.67279077e-01 4.51915860e-02
-2.66970128e-01 -6.61030769e-01 -1.60316551e+00 -3.36545184e-02
3.23779821e-01 -6.13986492e-01 -4.81018811e-01 4.11081873e-03
-2.29631573e-01 -4.60919961e-02 -5.50325692e-01 1.89270958e-01
7.58493483e-01 -1.00503242e+00 7.35063553e-01 -2.13810906e-01
-3.34454417e-01 -5.42137742e-01 -1.27577335e-01 -1.02458906e+00
3.42019737e-01 -3.00408989e-01 4.48992252e-01 1.71181083e+00
4.29542899e-01 -8.80067110e-01 1.08573568e+00 5.39412439e-01
4.61888254e-01 -1.03615344e-01 -1.55283749e+00 -9.22261596e-01
-1.13896027e-01 -2.18731821e-01 5.61668694e-01 9.49671745e-01
-2.67996006e-02 7.98787400e-02 -6.21619284e-01 7.65563101e-02
9.23860550e-01 -1.32084697e-01 7.34915614e-01 -1.42404997e+00
2.36225133e-05 -3.11905503e-01 -1.05186033e+00 -5.61777174e-01
2.48645023e-01 -7.99816191e-01 -1.58407554e-01 -1.15582645e+00
2.28028193e-01 -2.58681983e-01 -3.53060335e-01 3.85872424e-01
1.71954766e-01 3.46873373e-01 6.67492375e-02 7.48571241e-03
8.60335603e-02 3.92644703e-01 8.76691401e-01 -3.90632838e-01
3.77779864e-02 3.40058148e-01 -4.11942661e-01 5.64925194e-01
7.37214625e-01 -4.19773698e-01 -2.96633571e-01 1.91199649e-02
-2.78949529e-01 -1.01109594e-01 2.31857389e-01 -1.25523412e+00
5.59913576e-01 2.20503882e-01 4.10964876e-01 -7.17075527e-01
6.02990746e-01 -1.04341555e+00 2.38760293e-01 2.64905959e-01
-8.56823027e-02 -1.88578755e-01 1.25557065e-01 4.77293968e-01
-7.78593779e-01 -4.71454412e-01 9.98315096e-01 3.00284987e-03
-3.94724548e-01 5.12486100e-02 -1.62611559e-01 -3.45171213e-01
9.55407023e-01 -5.43133259e-01 4.29030731e-02 -5.95460713e-01
-5.96439421e-01 -2.30109140e-01 5.00715315e-01 4.43278968e-01
6.80343807e-01 -1.02909994e+00 -9.78062630e-01 3.48194152e-01
-4.04302739e-02 -6.72425210e-01 2.68583387e-01 8.11237097e-01
-3.86446923e-01 6.45804107e-01 -4.87812400e-01 -6.34268701e-01
-2.00310087e+00 4.77126718e-01 3.55910003e-01 3.73640865e-01
-8.88972878e-02 8.46965671e-01 -4.84489828e-01 -3.07357043e-01
4.72777456e-01 1.30307272e-01 -4.28102165e-01 2.26407081e-01
7.69177616e-01 3.23009431e-01 3.33652824e-01 -1.16956520e+00
-5.78493774e-01 7.59957433e-01 -6.76750913e-02 -4.55750912e-01
1.19882762e+00 -1.01732239e-02 -1.77082196e-01 5.96396148e-01
1.17472565e+00 1.45117298e-01 -5.20573914e-01 -1.03709556e-01
-7.82116950e-02 -6.44625247e-01 2.25954223e-02 -6.07762754e-01
-1.09465444e+00 1.41935444e+00 1.12234175e+00 5.75069971e-02
9.36438322e-01 -5.00990093e-01 2.83837587e-01 4.08644050e-01
4.88766998e-01 -1.22178543e+00 -2.58845270e-01 2.15728670e-01
7.97620595e-01 -1.55816650e+00 -8.69111493e-02 -1.87175974e-01
-5.36963820e-01 1.01305652e+00 2.03224704e-01 2.00568318e-01
8.98253798e-01 1.32391393e-01 2.95815289e-01 3.14282894e-01
-3.01562518e-01 -1.12044223e-01 3.63858849e-01 6.44846559e-01
6.94802344e-01 -2.09134221e-01 -4.13196713e-01 2.99979985e-01
2.47287937e-02 2.85403669e-01 2.92715997e-01 8.42399180e-01
-3.57262909e-01 -1.42349398e+00 -7.83726990e-01 3.15682381e-01
-8.97014499e-01 1.59607112e-01 -4.69938278e-01 6.62852108e-01
4.00125273e-02 1.11323345e+00 -4.67286021e-01 -2.19307616e-01
6.72908202e-02 7.04565763e-01 5.37555397e-01 -2.32715234e-01
-5.75191498e-01 -1.42777026e-01 -8.61404613e-02 -5.34487031e-02
-7.62309074e-01 -7.32489109e-01 -1.01776016e+00 -4.41076487e-01
-6.50815547e-01 3.42323631e-01 1.40698159e+00 9.66277182e-01
8.60234201e-02 -5.55186458e-02 5.45675993e-01 -5.83337724e-01
-5.97155988e-01 -1.09540498e+00 -7.94172347e-01 3.92771184e-01
3.25637192e-01 -6.82693243e-01 -5.26962101e-01 7.26251826e-02]
|
[13.304975509643555, 1.1581902503967285]
|
367ccfbd-e21f-4f31-a098-21585051ee79
|
semi-weakly-supervised-learning-of-complex
| null | null |
http://openaccess.thecvf.com//content/CVPR2022/html/Shen_Semi-Weakly-Supervised_Learning_of_Complex_Actions_From_Instructional_Task_Videos_CVPR_2022_paper.html
|
http://openaccess.thecvf.com//content/CVPR2022/papers/Shen_Semi-Weakly-Supervised_Learning_of_Complex_Actions_From_Instructional_Task_Videos_CVPR_2022_paper.pdf
|
Semi-Weakly-Supervised Learning of Complex Actions From Instructional Task Videos
|
We address the problem of action segmentation in instructional task videos with a small number of weakly-labeled training videos and a large number of unlabeled videos, which we refer to as Semi-Weakly-Supervised Learning (SWSL) of actions. We propose a general SWSL framework that can efficiently learn from both types of videos and can leverage any of the existing weakly-supervised action segmentation methods. Our key observation is that the distance between the transcript of an unlabeled video and those of the weakly-labeled videos from the same task is small yet often nonzero. Therefore, we develop a Soft Restricted Edit (SRE) loss to encourage small variations between the predicted transcripts of unlabeled videos and ground-truth transcripts of the weakly-labeled videos of the same task. To compute the SRE loss, we develop a flexible transcript prediction (FTP) method that uses the output of the action classifier to find both the length of the transcript and the sequence of actions occurring in an unlabeled video. We propose an efficient learning scheme in which we alternate between minimizing our proposed loss and generating pseudo-transcripts for unlabeled videos. By experiments on two benchmark datasets, we demonstrate that our approach can significantly improve the performance by using unlabeled videos, especially when the number of weakly-labeled videos is small.
|
['Ehsan Elhamifar', 'YuHan Shen']
|
2022-01-01
| null | null | null |
cvpr-2022-1
|
['action-segmentation']
|
['computer-vision']
|
[ 6.42354071e-01 1.00920543e-01 -5.61634123e-01 -4.86098588e-01
-1.08330142e+00 -8.37331295e-01 2.08583280e-01 -4.52639937e-01
-2.34772518e-01 6.13337696e-01 2.18239322e-01 -2.32919946e-01
4.04146105e-01 -1.55018255e-01 -1.16462290e+00 -8.76514494e-01
3.27583961e-02 1.27788782e-01 2.89108038e-01 4.59712327e-01
7.45457858e-02 -1.06559604e-01 -1.49494278e+00 6.53904140e-01
9.11117911e-01 1.06730962e+00 2.43314922e-01 6.82556510e-01
-2.26570144e-02 1.29453850e+00 -3.22855860e-01 -2.49314934e-01
4.70416278e-01 -9.70938623e-01 -1.16968942e+00 9.75098252e-01
7.15988874e-01 -3.43585849e-01 -2.10016578e-01 1.09654486e+00
-5.97349880e-03 4.86499637e-01 4.52426583e-01 -1.23824036e+00
-2.78036624e-01 7.85117090e-01 -6.82542086e-01 9.48497951e-02
5.55547476e-01 2.16436967e-01 8.65569234e-01 -9.07083154e-01
9.75305200e-01 1.11080515e+00 4.26005036e-01 5.25786757e-01
-9.91425991e-01 -5.69539964e-01 6.06880367e-01 1.82942361e-01
-1.14959824e+00 -3.32480311e-01 6.94526196e-01 -5.71758211e-01
4.10084397e-01 6.72524944e-02 6.61179125e-01 1.12415862e+00
-1.93548724e-01 1.25641501e+00 9.91917193e-01 -3.83418441e-01
2.77278453e-01 -8.69052708e-02 -2.05590557e-02 1.03962290e+00
-3.64049703e-01 -3.66927385e-01 -4.96129304e-01 -1.49351712e-02
5.24085999e-01 8.40315148e-02 -4.57241565e-01 -5.17304122e-01
-1.24150085e+00 5.87766469e-01 -2.39133954e-01 1.45823702e-01
-1.35408789e-01 6.65460005e-02 5.47236025e-01 4.72112447e-01
5.59365988e-01 1.30231231e-02 -5.69199800e-01 -3.82861853e-01
-8.72883320e-01 -3.08667179e-02 7.60098398e-01 1.22872686e+00
7.29138672e-01 -1.39479607e-01 -2.68580109e-01 5.61861396e-01
8.27536061e-02 5.44230901e-02 4.07651037e-01 -1.43321621e+00
6.25823855e-01 5.06331205e-01 2.38494083e-01 -3.79218996e-01
3.63220543e-01 1.52481154e-01 -1.79253995e-01 -3.00354362e-01
6.87264860e-01 -3.64785016e-01 -9.84965503e-01 1.73544145e+00
6.05529547e-01 8.23520720e-01 6.87631546e-03 7.20874250e-01
5.51175833e-01 7.82320559e-01 1.93521325e-02 -6.78000391e-01
6.56001627e-01 -1.51950264e+00 -6.75058603e-01 -1.81977227e-01
1.00401545e+00 -7.32644916e-01 1.17239892e+00 2.54515558e-01
-1.15831649e+00 -6.17912352e-01 -6.90930665e-01 -1.02214515e-01
2.29363397e-01 4.16522682e-01 2.15804815e-01 2.74046987e-01
-7.68273473e-01 6.37350976e-01 -1.02844787e+00 -2.10749567e-01
4.49085295e-01 2.09377125e-01 -3.79427105e-01 -4.12177056e-01
-6.74104095e-01 2.12233528e-01 3.74349207e-01 -5.97070791e-02
-1.38202465e+00 -6.55057192e-01 -1.05756748e+00 -9.84112546e-02
9.35945570e-01 -1.26132190e-01 1.40888369e+00 -1.89337790e+00
-1.51805210e+00 8.77715588e-01 -2.74205089e-01 -2.88434505e-01
6.97414935e-01 -3.38741481e-01 1.17974654e-01 4.84602720e-01
2.23466992e-01 7.36102402e-01 1.13985920e+00 -1.06540442e+00
-9.41345513e-01 -2.12284967e-01 1.93717778e-01 2.85912067e-01
-3.24241042e-01 -7.57401809e-02 -8.65540206e-01 -6.24539852e-01
-1.02417044e-01 -1.39215064e+00 -1.83539733e-01 1.05385400e-01
-3.52200031e-01 -4.66398478e-01 9.15776432e-01 -5.69928825e-01
8.97964716e-01 -2.40497708e+00 3.47017139e-01 -1.89754099e-01
-1.85223848e-01 2.79448688e-01 -4.62602824e-01 1.98918775e-01
-1.65983140e-01 -1.97020844e-01 -3.62280220e-01 -4.54258770e-01
-3.79458904e-01 4.88274366e-01 -3.72595757e-01 5.73417664e-01
1.36780918e-01 7.55158961e-01 -1.32522774e+00 -6.88146770e-01
-5.70480488e-02 5.39646186e-02 -5.75789571e-01 6.24672711e-01
-4.42515910e-01 8.55438352e-01 -5.50727248e-01 6.25117064e-01
3.20738345e-01 -3.38001251e-01 4.00930136e-01 5.71176559e-02
2.70242617e-02 5.95459379e-02 -1.09282458e+00 1.98639631e+00
-2.81978667e-01 4.32338268e-01 5.56200817e-02 -1.28022301e+00
4.49448735e-01 5.02022386e-01 6.68038607e-01 -2.26390129e-03
8.11979845e-02 4.59006280e-02 -3.29755723e-01 -9.45255578e-01
-1.02030644e-02 -1.30376786e-01 1.26364172e-01 7.88849711e-01
4.85139549e-01 1.38947917e-02 5.46240807e-01 2.81299323e-01
1.07201707e+00 7.52177775e-01 2.07465485e-01 1.78497687e-01
6.20684743e-01 -1.45795956e-01 8.56237769e-01 6.96735680e-01
-5.01698256e-01 4.23756540e-01 7.23225594e-01 -2.62939095e-01
-6.90279424e-01 -6.90749347e-01 2.77862459e-01 1.55488539e+00
1.03153490e-01 -3.24976742e-01 -1.09329295e+00 -1.49777365e+00
-3.98272693e-01 2.91204780e-01 -5.54223239e-01 -2.89076772e-02
-5.10669768e-01 -2.74800770e-02 5.76447733e-02 7.34827816e-01
1.70699462e-01 -1.06072247e+00 -2.45564818e-01 -9.64398310e-02
-3.67119670e-01 -1.53836024e+00 -1.00790977e+00 2.25271627e-01
-9.40861762e-01 -1.31666613e+00 -6.25523508e-01 -1.29496706e+00
1.06837106e+00 5.11349976e-01 7.91840792e-01 7.48205557e-02
2.19636321e-01 5.65419793e-01 -6.44575417e-01 9.85086188e-02
-3.62431854e-01 -1.83429077e-01 2.08508056e-02 4.11459118e-01
1.64748847e-01 -2.18371123e-01 -3.20996374e-01 4.33214128e-01
-1.09159720e+00 6.91894740e-02 4.19897407e-01 8.35918665e-01
9.62388873e-01 -6.47892058e-02 3.07624251e-01 -1.19829786e+00
-9.36571509e-02 -4.56721067e-01 -4.26924020e-01 4.18397337e-01
-1.04743041e-01 -1.15037123e-02 9.20711577e-01 -7.72517264e-01
-1.24736464e+00 7.67753482e-01 1.10623829e-01 -7.98404157e-01
-2.51310229e-01 3.85764658e-01 -2.69542664e-01 -8.63535777e-02
4.94342297e-02 2.15511590e-01 -1.57842338e-01 -1.95524573e-01
4.07224059e-01 6.87800229e-01 4.35604155e-01 -5.89951217e-01
6.25666440e-01 5.86336553e-01 -3.38554233e-01 -6.96574271e-01
-1.46373677e+00 -7.01558053e-01 -9.16234195e-01 -4.03310955e-01
8.17929268e-01 -1.09688997e+00 -5.73760867e-01 4.01663840e-01
-8.56944144e-01 -7.55083561e-01 -4.15407777e-01 6.60283983e-01
-9.50923026e-01 7.29931891e-01 -7.03242183e-01 -6.24095500e-01
1.50901705e-01 -1.41217506e+00 1.05235779e+00 -3.33500374e-03
-1.31837517e-01 -8.61958563e-01 9.34778973e-02 8.77162158e-01
-6.21353388e-01 2.04683259e-01 7.40109921e-01 -7.05415249e-01
-5.99364877e-01 -2.42530227e-01 1.09557532e-01 8.21557164e-01
1.93829253e-01 1.66644812e-01 -6.31112695e-01 -4.20066893e-01
1.25539392e-01 -9.95398879e-01 8.81852388e-01 2.97186762e-01
1.44002652e+00 -4.10070300e-01 -1.68452770e-01 5.48632383e-01
9.82122540e-01 2.03197971e-01 2.53644437e-01 -1.63622871e-01
9.72666919e-01 8.37090731e-01 1.21228611e+00 2.62185186e-01
5.62709309e-02 4.68399584e-01 1.96489051e-01 2.43992090e-01
1.26857772e-01 -5.43201804e-01 1.04338968e+00 1.10602915e+00
-6.19082823e-02 -2.71209896e-01 -4.75672573e-01 6.18445873e-01
-2.09649587e+00 -9.56287503e-01 8.51903185e-02 2.13536572e+00
1.12676275e+00 1.21489801e-01 1.34490952e-01 -5.71106113e-02
6.87932432e-01 3.23291004e-01 -7.80958772e-01 -2.60909528e-01
3.90846014e-01 7.65169710e-02 3.24811786e-01 3.70075673e-01
-1.45459294e+00 9.53000009e-01 5.91517353e+00 8.47909272e-01
-8.82036507e-01 1.37163088e-01 7.13887632e-01 -3.52856785e-01
1.85049337e-03 -7.14996606e-02 -5.46747625e-01 5.84142148e-01
1.02542257e+00 7.83640207e-05 2.87618697e-01 1.18042457e+00
5.37236452e-01 -1.05664752e-01 -1.49234402e+00 6.76027715e-01
4.01416481e-01 -1.09491205e+00 1.24784037e-01 -2.82067984e-01
1.40347362e+00 -2.23785445e-01 -2.77333781e-02 3.74240249e-01
3.15603465e-01 -5.98042488e-01 7.49436021e-01 4.78364713e-03
7.53891468e-01 -6.83788002e-01 5.62407017e-01 4.64333802e-01
-1.24242198e+00 -2.87032098e-01 -1.71295375e-01 4.36746888e-03
1.08049333e-01 1.47822767e-01 -4.90512788e-01 1.17368773e-01
5.45409560e-01 1.25615835e+00 -2.67001569e-01 6.91882789e-01
-4.98241007e-01 1.02235472e+00 4.87295538e-02 3.06906313e-01
6.61053121e-01 -4.60735768e-01 2.54572779e-01 1.01620817e+00
1.77618727e-01 2.71365911e-01 8.61354768e-01 3.38593572e-01
-4.41730559e-01 2.01169521e-01 -5.43860793e-01 -2.80676484e-01
1.38759717e-01 1.11682129e+00 -8.35240304e-01 -4.60360378e-01
-8.31726670e-01 1.17093480e+00 3.52344215e-01 5.22892535e-01
-7.92165458e-01 8.65852535e-02 4.08049494e-01 -4.84303981e-02
6.16821468e-01 9.32343528e-02 2.47318432e-01 -1.23277426e+00
2.16994002e-01 -9.25243378e-01 5.11652648e-01 -6.97102845e-01
-1.01094556e+00 1.76193953e-01 -1.76892444e-01 -1.59696090e+00
-2.08506435e-01 -3.24546039e-01 -5.77115953e-01 1.17465682e-01
-1.26850748e+00 -9.79905128e-01 -1.19139045e-01 6.25403106e-01
1.24970496e+00 1.69077009e-01 2.95396686e-01 1.91194266e-01
-7.09253609e-01 4.28107381e-01 2.39051729e-01 3.46785158e-01
7.67807364e-01 -1.21740544e+00 -1.06808823e-02 9.49189901e-01
3.71474504e-01 2.19434410e-01 4.54757988e-01 -5.97000301e-01
-1.29931033e+00 -1.48401117e+00 6.02052450e-01 -2.08057895e-01
7.17881083e-01 -3.09510261e-01 -8.75426054e-01 1.29349446e+00
1.52420193e-01 3.64518642e-01 9.89520133e-01 -4.50024813e-01
-1.17100500e-01 1.67979151e-01 -8.44279110e-01 5.48414111e-01
1.23665631e+00 -5.45999706e-01 -5.50429165e-01 7.99359620e-01
8.68769228e-01 -5.61834395e-01 -6.01976514e-01 2.25132078e-01
4.45858181e-01 -7.30238199e-01 5.91939390e-01 -9.35244083e-01
7.90318787e-01 -2.21254200e-01 9.12355408e-02 -1.09944034e+00
1.79590449e-01 -8.41346979e-01 -2.18327135e-01 1.12523711e+00
2.31663316e-01 1.37353778e-01 1.10209215e+00 5.68506896e-01
-2.96105772e-01 -8.60139549e-01 -6.95624590e-01 -8.71153831e-01
-2.68329382e-01 -3.20510685e-01 -1.97227284e-01 9.79793131e-01
1.89903826e-01 1.42191544e-01 -6.47622585e-01 -5.28002456e-02
5.40872455e-01 4.51878369e-01 6.78779960e-01 -6.91775978e-01
-3.34017128e-01 2.74522215e-01 -2.12911054e-01 -1.50950217e+00
8.73981118e-01 -7.61604786e-01 6.00467622e-01 -9.26765501e-01
5.71089387e-01 1.26505360e-01 -2.28719860e-01 5.66136658e-01
-3.21515292e-01 3.15417767e-01 -2.13349387e-02 1.41345695e-01
-1.19938803e+00 5.61291099e-01 1.48366988e+00 -1.90211535e-01
-3.07821214e-01 8.80750865e-02 -1.05838917e-01 1.13707340e+00
2.86058962e-01 -7.42435992e-01 -6.34123623e-01 -3.04336786e-01
-3.14015239e-01 2.48035058e-01 -1.04120553e-01 -6.27044141e-01
-4.58735190e-02 -4.30115670e-01 1.24737047e-01 -3.67066324e-01
-5.46327680e-02 -7.63027549e-01 -2.71558642e-01 3.61238718e-01
-9.54497635e-01 -4.02631730e-01 -2.58474708e-01 1.00336456e+00
-4.15489435e-01 -5.22734761e-01 6.78205073e-01 -2.86595553e-01
-8.21257234e-01 4.87068295e-01 -4.66530919e-01 3.18490475e-01
1.63021290e+00 -2.60267437e-01 1.23902299e-01 -4.51971382e-01
-8.76661956e-01 4.65414375e-01 4.96144414e-01 4.07261789e-01
5.35627127e-01 -1.12429249e+00 -3.90358150e-01 -1.08295940e-02
2.44781286e-01 9.04692262e-02 1.70454457e-01 9.64949131e-01
-3.39657426e-01 1.97797015e-01 2.05982998e-02 -6.19515657e-01
-1.62447858e+00 6.53987288e-01 -5.15670404e-02 -2.63062179e-01
-4.69440997e-01 1.03767478e+00 2.95322835e-01 -2.92361706e-01
5.98588228e-01 -4.24306780e-01 -6.22171536e-02 9.14754272e-02
5.55662632e-01 3.18935364e-01 -2.58992344e-01 -7.93400824e-01
-3.22477557e-02 3.88649851e-01 -2.77489752e-01 1.58472374e-01
1.36271608e+00 -1.73629463e-01 8.09702575e-02 6.93563044e-01
1.50856459e+00 -9.48422924e-02 -2.03792310e+00 -2.80315250e-01
2.74964739e-02 -5.76552808e-01 -2.71984100e-01 -2.51600772e-01
-1.36983931e+00 6.67011797e-01 2.95550466e-01 -2.15390906e-01
1.12894440e+00 1.45035431e-01 1.10481250e+00 4.79117036e-01
3.51131082e-01 -1.27028549e+00 6.17783427e-01 3.72632116e-01
3.79843831e-01 -1.21478879e+00 -2.65450567e-01 -5.73599756e-01
-1.10373044e+00 8.90471816e-01 7.05411792e-01 -3.82361710e-02
3.63289267e-01 1.85269699e-01 7.37569109e-02 1.91378340e-01
-7.88615286e-01 -1.76240161e-01 8.22785422e-02 4.35942739e-01
4.26750064e-01 -2.35795885e-01 -3.64980787e-01 3.41982812e-01
5.44613302e-01 4.36393589e-01 6.11980438e-01 1.26766682e+00
-4.07363981e-01 -1.12677777e+00 -8.74092355e-02 3.92668068e-01
-5.97232401e-01 1.59382164e-01 -6.49200380e-01 2.72014827e-01
2.52722442e-01 9.08610046e-01 3.32771726e-02 -3.47628444e-01
4.91095968e-02 3.16354603e-01 4.76302356e-01 -1.09819210e+00
-3.38677347e-01 1.85565174e-01 -1.52656287e-01 -8.74929011e-01
-9.73616362e-01 -8.59835982e-01 -1.27145875e+00 2.47958079e-01
-2.95343220e-01 3.48977506e-01 -5.02213985e-02 1.25805175e+00
5.70340119e-02 9.13580060e-02 1.02572048e+00 -7.30629563e-01
-5.66889942e-01 -7.57600307e-01 -7.04540253e-01 7.85672605e-01
2.90450335e-01 -5.12941718e-01 -3.91025484e-01 9.42827404e-01]
|
[8.603788375854492, 0.6318374872207642]
|
16baf432-aa1d-4bc7-9a48-50133c7940a3
|
self-supervised-learning-for-fine-grained
|
2105.08788
| null |
https://arxiv.org/abs/2105.08788v1
|
https://arxiv.org/pdf/2105.08788v1.pdf
|
Self-Supervised Learning for Fine-Grained Visual Categorization
|
Recent research in self-supervised learning (SSL) has shown its capability in learning useful semantic representations from images for classification tasks. Through our work, we study the usefulness of SSL for Fine-Grained Visual Categorization (FGVC). FGVC aims to distinguish objects of visually similar sub categories within a general category. The small inter-class, but large intra-class variations within the dataset makes it a challenging task. The limited availability of annotated labels for such a fine-grained data encourages the need for SSL, where additional supervision can boost learning without the cost of extra annotations. Our baseline achieves $86.36\%$ top-1 classification accuracy on CUB-200-2011 dataset by utilizing random crop augmentation during training and center crop augmentation during testing. In this work, we explore the usefulness of various pretext tasks, specifically, rotation, pretext invariant representation learning (PIRL), and deconstruction and construction learning (DCL) for FGVC. Rotation as an auxiliary task promotes the model to learn global features, and diverts it from focusing on the subtle details. PIRL that uses jigsaw patches attempts to focus on discriminative local regions, but struggles to accurately localize them. DCL helps in learning local discriminating features and outperforms the baseline by achieving $87.41\%$ top-1 accuracy. The deconstruction learning forces the model to focus on local object parts, while reconstruction learning helps in learning the correlation between the parts. We perform extensive experiments to reason our findings. Our code is available at https://github.com/mmaaz60/ssl_for_fgvc.
|
['Dhanalaxmi Gaddam', 'Hanoona Abdul Rasheed', 'Muhammad Maaz']
|
2021-05-18
| null | null | null | null |
['fine-grained-visual-categorization']
|
['computer-vision']
|
[ 3.02854359e-01 -1.41571641e-01 -4.59312052e-01 -3.89892757e-01
-8.13637674e-01 -8.16643059e-01 6.66901350e-01 1.48637965e-01
-6.29438236e-02 3.39958072e-01 2.10219204e-01 -3.24878693e-01
8.39922056e-02 -6.18201852e-01 -8.70709598e-01 -7.46156156e-01
1.66950785e-02 -6.53563738e-02 2.71753371e-01 1.26491822e-02
1.87416047e-01 5.34348369e-01 -1.67006862e+00 7.73004830e-01
6.77447855e-01 1.16122830e+00 5.32585025e-01 3.89722645e-01
-2.14487642e-01 9.12810564e-01 -5.28277278e-01 -5.26309758e-02
3.16148460e-01 -3.03487003e-01 -9.69423175e-01 1.91842377e-01
9.05921996e-01 -1.33332670e-01 -1.68545600e-02 8.39011729e-01
2.91466266e-01 -1.84798590e-03 9.50733602e-01 -1.42598867e+00
-7.87888229e-01 3.78888369e-01 -9.52224135e-01 1.92550659e-01
5.19989096e-02 1.53579935e-01 1.09736395e+00 -1.15065563e+00
5.25073409e-01 1.28854275e+00 6.90623820e-01 4.69625533e-01
-1.37026882e+00 -9.03816044e-01 5.52276969e-01 1.15352236e-01
-1.42320669e+00 -4.14022744e-01 7.97441542e-01 -6.47934079e-01
7.42957532e-01 1.36308104e-01 1.86607167e-01 1.16215026e+00
-1.03242226e-01 1.03022540e+00 1.43418527e+00 -5.65727770e-01
1.39251679e-01 8.00975263e-02 3.78702760e-01 7.80016899e-01
2.07317382e-01 2.26503283e-01 -3.82695675e-01 1.65414080e-01
9.53457177e-01 2.39995390e-01 -2.17335463e-01 -6.55191302e-01
-8.81720960e-01 8.54469180e-01 1.03056228e+00 2.31117815e-01
2.69956496e-02 1.05929606e-01 3.45852226e-01 3.11990440e-01
4.49049056e-01 6.12300634e-01 -7.40522683e-01 4.16611254e-01
-8.01559210e-01 -4.08172496e-02 3.41574818e-01 1.02857840e+00
9.21972632e-01 3.68270325e-03 -3.53695661e-01 1.08692408e+00
8.07881355e-02 3.55140060e-01 5.19614458e-01 -7.84252524e-01
3.15470248e-01 7.54142761e-01 -2.60056734e-01 -9.45392907e-01
-3.54946971e-01 -6.51743948e-01 -8.75931263e-01 3.07040393e-01
4.26188171e-01 1.75488666e-01 -1.40955532e+00 1.82261789e+00
1.99563224e-02 -5.01995273e-02 -1.14221714e-01 7.95511782e-01
1.03805184e+00 5.35982728e-01 3.71386915e-01 2.27613732e-01
1.25498438e+00 -1.26072383e+00 -1.81439370e-01 -5.63859522e-01
7.49022484e-01 -8.07729840e-01 1.51662922e+00 2.24168867e-01
-5.99828184e-01 -8.41748893e-01 -9.53325212e-01 -1.27982318e-01
-5.96045792e-01 5.11725426e-01 8.15157473e-01 2.71361172e-01
-9.25073802e-01 4.21390444e-01 -6.69164419e-01 -4.19834405e-01
8.65167558e-01 3.33589944e-03 -5.13509750e-01 -5.30751348e-01
-6.41641617e-01 6.64970100e-01 3.30277205e-01 -2.30598703e-01
-1.08583319e+00 -6.65716767e-01 -9.64720607e-01 7.09529594e-02
3.57763737e-01 -2.03185260e-01 9.77237165e-01 -1.16846812e+00
-8.93684089e-01 9.95072663e-01 -6.66162819e-02 -1.95869878e-01
1.75609171e-01 -1.46343052e-01 -1.68682918e-01 2.37572774e-01
3.94318581e-01 1.17354536e+00 1.05313015e+00 -1.59923863e+00
-6.57163441e-01 -3.94883245e-01 6.64859861e-02 1.88378077e-02
-2.45841205e-01 -2.41207540e-01 -3.89900446e-01 -1.04074323e+00
1.97126418e-01 -9.59633827e-01 1.29106700e-01 9.28950757e-02
-3.26650023e-01 -4.13554847e-01 9.29670572e-01 -6.16572559e-01
8.26717556e-01 -2.33727384e+00 -2.30555862e-01 -8.89127851e-02
5.56912199e-02 1.82562962e-01 -6.01202488e-01 1.37205511e-01
-3.39656740e-01 2.95429945e-01 -1.31164059e-01 -1.90191388e-01
-1.98327452e-01 -1.61236338e-02 -3.70882064e-01 1.90437436e-01
7.27449119e-01 1.10145009e+00 -6.56217217e-01 -3.24771285e-01
1.61068588e-01 3.21709186e-01 -4.28297222e-01 1.60813749e-01
-1.18063092e-01 2.07324535e-01 -3.18194300e-01 1.05056190e+00
7.94693828e-01 -5.54959059e-01 5.98132238e-02 -4.81876284e-01
8.93723369e-02 1.46965906e-02 -1.00022852e+00 1.62661302e+00
-4.73052263e-01 5.67791700e-01 -6.09025769e-02 -1.20620108e+00
9.44499195e-01 -2.26907492e-01 -1.42840464e-02 -8.70672286e-01
1.12399952e-02 -2.65301745e-02 -1.45806611e-01 -2.41766363e-01
2.28916705e-01 1.46946967e-01 -1.73303485e-01 3.02983224e-01
3.31444293e-01 -6.66120499e-02 8.69162306e-02 2.69142479e-01
9.61100698e-01 3.78350586e-01 3.99113059e-01 -3.87309551e-01
1.78552046e-01 1.10115752e-01 3.81332725e-01 6.57287061e-01
-2.88926154e-01 7.51800001e-01 2.85204113e-01 -2.90588796e-01
-7.36234605e-01 -1.14899099e+00 -2.16235146e-01 1.60911632e+00
3.35610121e-01 -3.13941330e-01 -3.64977479e-01 -1.14496017e+00
2.33775154e-01 5.89056015e-01 -8.53317142e-01 -2.54259825e-01
-4.09925342e-01 -6.15674496e-01 2.46519119e-01 9.76653695e-01
7.11898923e-01 -1.06305277e+00 -4.15479362e-01 -2.91457117e-01
-2.16404989e-01 -1.05738115e+00 -3.06631953e-01 7.12212980e-01
-8.75637889e-01 -1.19109941e+00 -5.84143221e-01 -1.13384438e+00
8.38174880e-01 8.03125381e-01 1.20734870e+00 1.08482994e-01
-5.39164364e-01 2.08806455e-01 -6.50590420e-01 -4.22388554e-01
-2.53725704e-02 2.06024915e-01 -3.41965228e-01 -2.39255816e-01
3.83360088e-01 -1.82600573e-01 -4.77233529e-01 5.85992634e-01
-7.30123639e-01 1.92711309e-01 8.19865525e-01 1.11443126e+00
6.69040620e-01 -9.71290097e-02 5.85488677e-01 -8.81808937e-01
1.33157849e-01 -4.49510396e-01 -2.85561562e-01 2.69376159e-01
-5.06736040e-01 1.96476743e-01 5.31397700e-01 -5.40683389e-01
-8.40202510e-01 1.75947279e-01 2.74106264e-01 -4.34747905e-01
-4.66989756e-01 2.14722857e-01 -1.35051101e-01 -1.19442046e-01
8.56750309e-01 4.61847782e-02 -1.65641129e-01 -6.54260635e-01
3.94286424e-01 5.90427220e-01 4.21904981e-01 -6.26305342e-01
8.17414880e-01 3.54269564e-01 -3.71770471e-01 -6.23773396e-01
-1.23782074e+00 -6.18808508e-01 -7.46315002e-01 1.23580150e-01
7.25920260e-01 -1.23194778e+00 -1.95923850e-01 2.71432012e-01
-5.14764905e-01 -7.07474768e-01 -3.67882073e-01 1.00538194e-01
-3.37235630e-01 3.30191478e-03 -3.77186477e-01 -4.25669402e-01
-1.00191981e-01 -1.06303227e+00 1.24605501e+00 2.90020168e-01
-9.60911140e-02 -7.28648126e-01 -4.15258586e-01 4.77741539e-01
4.00777847e-01 2.35876128e-01 9.25450563e-01 -4.41588998e-01
-5.49967468e-01 7.65589438e-03 -7.59801686e-01 5.50327182e-01
3.53089005e-01 -2.63734400e-01 -1.28026831e+00 -6.07950628e-01
-5.06741464e-01 -9.37810957e-01 1.43522525e+00 3.47162992e-01
1.59059715e+00 -1.46720260e-01 -5.30993342e-01 7.89404631e-01
1.48872650e+00 -1.69461016e-02 4.10015762e-01 4.05457169e-01
8.29089761e-01 5.94873428e-01 8.14282358e-01 6.84186444e-02
3.20351809e-01 5.93051374e-01 4.99572814e-01 -4.13772225e-01
-7.15940773e-01 -3.32488596e-01 2.40293056e-01 2.71067202e-01
-6.09572139e-03 1.88984368e-02 -8.30272615e-01 5.65185189e-01
-1.52421081e+00 -5.84820688e-01 2.88729846e-01 1.87041676e+00
8.92128229e-01 2.34230272e-02 1.41112106e-02 9.13994983e-02
6.07465088e-01 2.26807550e-01 -6.07529640e-01 -1.88441128e-01
-1.94972634e-01 3.65420341e-01 6.04200184e-01 1.94678247e-01
-1.58458495e+00 1.32798350e+00 5.09670973e+00 1.10481906e+00
-1.39028680e+00 4.49437415e-03 9.03760791e-01 2.08204225e-01
1.53061137e-01 -8.96703899e-02 -8.23959649e-01 2.15142548e-01
4.32415158e-01 5.04941702e-01 2.31084168e-01 9.87591743e-01
-1.89566359e-01 -1.88668892e-01 -9.70598161e-01 9.82058764e-01
2.21428931e-01 -1.19589317e+00 1.81272939e-01 -2.94455409e-01
8.82419825e-01 1.82967279e-02 2.03382403e-01 4.84983265e-01
4.02899861e-01 -1.28651047e+00 6.83996677e-01 1.11233674e-01
1.20081615e+00 -6.37587070e-01 6.36524796e-01 1.67534560e-01
-1.36549342e+00 -4.11639571e-01 -4.99229282e-01 -2.75551490e-02
-5.15610874e-01 2.72854149e-01 -8.96923304e-01 3.98646414e-01
1.02187634e+00 8.80787492e-01 -1.16016698e+00 7.79220641e-01
-4.19298798e-01 6.94602966e-01 6.76950887e-02 3.00421596e-01
2.17785135e-01 2.17767552e-01 -2.74772756e-02 1.35956538e+00
5.62541336e-02 -3.81398574e-02 5.72700500e-01 6.40143692e-01
-1.12235323e-01 -6.59185695e-03 -4.58181590e-01 -1.01330064e-01
3.78005445e-01 1.31175494e+00 -9.39508379e-01 -1.63286179e-01
-3.11297238e-01 1.02399194e+00 5.97829103e-01 3.60300273e-01
-6.33056462e-01 -3.63828629e-01 6.60470307e-01 2.66746134e-01
5.91616809e-01 -1.45983413e-01 -4.43978459e-01 -1.17714596e+00
-1.38220727e-01 -1.03873336e+00 6.26506090e-01 -7.78348267e-01
-1.41152787e+00 5.69777310e-01 -5.16711436e-02 -1.20863056e+00
5.21381125e-02 -7.89269805e-01 -3.42820436e-01 7.05395103e-01
-1.72908163e+00 -1.73821819e+00 -7.58751452e-01 6.84318364e-01
8.40239823e-01 -2.81952620e-01 8.44526350e-01 -5.63235395e-02
-4.65049177e-01 9.36965346e-01 -3.89690697e-02 3.05986702e-01
1.03453612e+00 -1.22386932e+00 1.28703386e-01 7.70114422e-01
2.17857063e-01 5.22091985e-01 2.68582672e-01 -5.11630893e-01
-9.52651918e-01 -1.47630346e+00 5.32137573e-01 -4.01817560e-01
4.33954924e-01 -5.13255119e-01 -8.85212600e-01 5.64694703e-01
-6.82407692e-02 3.97187859e-01 4.97798383e-01 2.05594346e-01
-1.06187642e+00 -1.14631638e-01 -1.05973101e+00 3.14624399e-01
1.18304574e+00 -6.60107076e-01 -4.25042719e-01 2.91678667e-01
7.66029418e-01 -5.86984605e-02 -5.33613265e-01 5.40426850e-01
4.27983850e-01 -7.26180792e-01 1.20906365e+00 -5.04141033e-01
6.52898610e-01 -3.24766368e-01 -4.18793410e-01 -1.32859731e+00
-7.04756677e-01 1.36156857e-01 3.97371911e-02 1.33790624e+00
3.07537556e-01 -4.22326386e-01 6.41667962e-01 -4.64915074e-02
-9.55474675e-02 -6.51282370e-01 -2.83215880e-01 -9.14519370e-01
2.95576930e-01 -3.35537076e-01 3.55412096e-01 1.14809108e+00
-3.88754815e-01 4.03511941e-01 -8.82794932e-02 2.42963746e-01
5.66212237e-01 5.36620915e-01 7.27464974e-01 -1.20393801e+00
-2.15282589e-01 -3.88411522e-01 -3.20052594e-01 -9.17373240e-01
1.79231167e-01 -1.17337775e+00 -6.87996149e-02 -1.50446153e+00
4.85696852e-01 -8.42451990e-01 -3.83301407e-01 1.05586743e+00
-2.54184455e-01 7.41143227e-01 4.84946221e-01 2.75428534e-01
-5.89495957e-01 3.28406215e-01 1.30408013e+00 -4.75958169e-01
-7.92806298e-02 -1.31940097e-01 -1.12501347e+00 7.36612380e-01
9.19148862e-01 -2.79182196e-01 -4.14252043e-01 -3.99290591e-01
-4.39246744e-01 -5.08236110e-01 6.30345166e-01 -9.66529012e-01
-1.83581889e-01 -9.80413780e-02 9.91038978e-01 -5.97999513e-01
1.57022744e-01 -5.99672318e-01 -3.80822152e-01 3.91928077e-01
-4.84367490e-01 -2.08158731e-01 4.08015579e-01 3.95385563e-01
-2.75654584e-01 -1.64883137e-02 1.05267656e+00 -1.64113596e-01
-1.18372881e+00 1.76566735e-01 4.73736487e-02 2.52032489e-01
8.63599360e-01 -1.29319206e-01 -5.70317864e-01 -2.10316330e-01
-7.46630013e-01 1.44683808e-01 5.85652530e-01 6.29285991e-01
5.19350529e-01 -1.18246400e+00 -5.82443297e-01 3.57899249e-01
6.09769821e-01 6.28611594e-02 2.22277790e-01 3.68857682e-01
-2.92416930e-01 3.36271256e-01 -4.93219882e-01 -8.16626132e-01
-1.36490738e+00 7.96052039e-01 7.91722164e-02 -1.82000116e-01
-4.50445235e-01 1.08380485e+00 8.08869600e-01 -4.87406075e-01
3.40698779e-01 -4.07422125e-01 -3.51157635e-01 2.11884513e-01
5.43752968e-01 1.83656380e-01 3.76319187e-03 -5.61931849e-01
-3.72721314e-01 6.80531323e-01 -2.97668725e-01 4.55090135e-01
1.36602020e+00 -9.36515853e-02 1.69969186e-01 1.93411350e-01
1.45543230e+00 -1.84839696e-01 -1.52742386e+00 -4.44633096e-01
5.91239370e-02 -5.20698369e-01 1.84590459e-01 -1.25535738e+00
-1.06270540e+00 1.00479925e+00 9.27365422e-01 -6.33031279e-02
1.37385511e+00 3.58344793e-01 2.48900622e-01 1.91413343e-01
4.91650224e-01 -8.69680822e-01 4.94750559e-01 6.61499560e-01
1.11317909e+00 -1.68587720e+00 1.25728371e-02 -6.35311365e-01
-7.67001271e-01 9.67556298e-01 9.18228447e-01 -2.72992522e-01
5.50506353e-01 2.62151688e-01 1.65332004e-01 -6.71371177e-04
-5.09470642e-01 -5.53468943e-01 5.58597565e-01 8.57519329e-01
4.24059033e-01 2.75180936e-01 1.82548776e-01 5.66089869e-01
-1.34368718e-01 -2.98931628e-01 1.79837823e-01 9.69399095e-01
-4.74049181e-01 -9.13392544e-01 -3.42180103e-01 5.62145829e-01
-1.90309048e-01 -1.98569641e-01 -5.67145288e-01 9.16054189e-01
4.01318014e-01 8.61994147e-01 2.21045405e-01 -3.86838019e-01
1.76520392e-01 4.72041108e-02 5.18755794e-01 -7.23307252e-01
-4.03778762e-01 1.49084091e-01 -1.20884679e-01 -7.33808041e-01
-3.06478858e-01 -5.87867320e-01 -1.09978223e+00 1.34821355e-01
-2.64505774e-01 -2.03646690e-01 4.11447406e-01 6.85416698e-01
3.45148504e-01 5.83391607e-01 7.19134033e-01 -9.27358747e-01
-4.73247945e-01 -1.00008833e+00 -4.12607938e-01 5.82067072e-01
4.35191661e-01 -9.71589744e-01 -4.31464583e-01 1.75010249e-01]
|
[9.607216835021973, 2.0105345249176025]
|
90e26be0-a44c-4ddf-8c35-a381c4e40240
|
x-llm-bootstrapping-advanced-large-language
|
2305.0416
| null |
https://arxiv.org/abs/2305.04160v3
|
https://arxiv.org/pdf/2305.04160v3.pdf
|
X-LLM: Bootstrapping Advanced Large Language Models by Treating Multi-Modalities as Foreign Languages
|
Large language models (LLMs) have demonstrated remarkable language abilities. GPT-4, based on advanced LLMs, exhibits extraordinary multimodal capabilities beyond previous visual language models. We attribute this to the use of more advanced LLMs compared with previous multimodal models. Unfortunately, the model architecture and training strategies of GPT-4 are unknown. To endow LLMs with multimodal capabilities, we propose X-LLM, which converts Multi-modalities (images, speech, videos) into foreign languages using X2L interfaces and inputs them into a large Language model (ChatGLM). Specifically, X-LLM aligns multiple frozen single-modal encoders and a frozen LLM using X2L interfaces, where ``X'' denotes multi-modalities such as image, speech, and videos, and ``L'' denotes languages. X-LLM's training consists of three stages: (1) Converting Multimodal Information: The first stage trains each X2L interface to align with its respective single-modal encoder separately to convert multimodal information into languages. (2) Aligning X2L representations with the LLM: single-modal encoders are aligned with the LLM through X2L interfaces independently. (3) Integrating multiple modalities: all single-modal encoders are aligned with the LLM through X2L interfaces to integrate multimodal capabilities into the LLM. Our experiments show that X-LLM demonstrates impressive multimodel chat abilities, sometimes exhibiting the behaviors of multimodal GPT-4 on unseen images/instructions, and yields a 84.5\% relative score compared with GPT-4 on a synthetic multimodal instruction-following dataset. And we also conduct quantitative tests on using LLM for ASR and multimodal ASR, hoping to promote the era of LLM-based speech recognition.
|
['Bo Xu', 'Shuang Xu', 'Jing Shi', 'Qingyang Zhang', 'Haozhi Zhao', 'Minglun Han', 'Feilong Chen']
|
2023-05-07
| null | null | null | null |
['instruction-following']
|
['natural-language-processing']
|
[ 2.14239150e-01 2.47248366e-01 -3.56664121e-01 -2.20737115e-01
-1.26535094e+00 -6.40727639e-01 6.53380632e-01 -4.62922633e-01
-3.98659259e-01 4.14447784e-01 1.56585351e-01 -6.91683471e-01
5.33131003e-01 -4.31494623e-01 -1.24145675e+00 -4.77512211e-01
8.86997730e-02 5.58873177e-01 -1.33828282e-01 -2.30834022e-01
-1.02048643e-01 8.05542469e-02 -1.40715134e+00 7.93689489e-01
7.03956068e-01 9.30689573e-01 5.26642740e-01 1.07673037e+00
-5.18642724e-01 1.01755917e+00 -4.55856115e-01 -5.80894172e-01
-2.21075654e-01 -4.82388645e-01 -7.85558641e-01 2.33455405e-01
6.04689956e-01 -3.15501779e-01 -6.17268026e-01 8.88943672e-01
3.38846147e-01 -8.98124371e-03 6.17423832e-01 -1.54657793e+00
-1.17854464e+00 6.10039175e-01 -5.70701480e-01 -5.05853713e-01
7.45561540e-01 5.11363506e-01 7.94291556e-01 -1.21898687e+00
5.93516648e-01 1.83900917e+00 2.75389433e-01 9.59411800e-01
-1.21118402e+00 -6.75306320e-01 1.39794976e-01 -1.67560503e-01
-1.27773070e+00 -6.96218252e-01 3.22679013e-01 -2.47649729e-01
1.15956140e+00 1.52525008e-01 9.39399526e-02 1.39789367e+00
2.64251411e-01 1.29859185e+00 1.22225273e+00 -5.64101934e-01
-8.19713995e-02 4.27048504e-01 4.81641106e-02 1.04294586e+00
-6.02057993e-01 -7.35899433e-02 -1.00087559e+00 2.56885856e-01
6.60622656e-01 -1.92741618e-01 -3.10935378e-01 -2.87248399e-02
-1.47132218e+00 6.57129407e-01 3.40291858e-01 2.02265847e-02
-2.13544983e-02 3.53668720e-01 2.21063778e-01 6.10212803e-01
-4.70421873e-02 2.93678194e-01 -3.59021612e-02 -2.87275553e-01
-6.63594604e-01 -5.05434453e-01 5.10742009e-01 1.49115252e+00
7.99323320e-01 7.37973973e-02 -3.92921343e-02 1.11101937e+00
5.96424162e-01 9.27043736e-01 6.44757807e-01 -1.04974878e+00
9.61231589e-01 6.53811932e-01 -2.97825605e-01 -3.64068210e-01
-1.26580924e-01 1.14979386e-01 -8.54199409e-01 1.52359620e-01
3.35507393e-02 -1.37184024e-01 -1.30784774e+00 1.89038575e+00
-3.19817036e-01 -2.91121542e-01 7.41635740e-01 6.74220800e-01
1.38508081e+00 1.24702799e+00 3.69778544e-01 1.32810354e-01
1.30509257e+00 -1.24785328e+00 -6.15179121e-01 -3.64304543e-01
6.65119410e-01 -8.93434286e-01 1.49155343e+00 1.48027360e-01
-1.43768990e+00 -9.65411007e-01 -7.73546100e-01 -2.66903311e-01
-4.52412069e-01 6.06608450e-01 2.59920686e-01 3.35708827e-01
-1.59077668e+00 -2.08399981e-01 -5.52086532e-01 -5.49371243e-01
6.13300800e-02 5.23286045e-01 -9.35208321e-01 -2.60535896e-01
-1.05968428e+00 8.42319071e-01 3.55748147e-01 1.45083353e-01
-1.27527070e+00 -5.36972918e-02 -1.48823059e+00 2.85095647e-02
1.96520478e-01 -7.78496325e-01 1.36515784e+00 -1.01720226e+00
-1.61436605e+00 9.61679637e-01 -6.82125628e-01 -2.86411270e-02
2.27938965e-01 9.51669645e-03 -6.14835441e-01 4.03910965e-01
5.36836386e-02 1.53372264e+00 8.71711195e-01 -1.77777410e+00
-5.71503043e-01 5.52435629e-02 6.43597320e-02 3.97118568e-01
-3.24916601e-01 -1.25160769e-01 -1.11208248e+00 -2.88791418e-01
-1.81326345e-02 -9.72690821e-01 2.04994142e-01 -3.08197111e-01
-6.76528990e-01 -7.13764504e-02 8.45081806e-01 -6.66046798e-01
1.05655265e+00 -2.35824895e+00 5.29505789e-01 -1.29207615e-02
2.15945259e-01 2.15898126e-01 -9.83293056e-01 7.06376970e-01
-2.12143660e-02 8.83131176e-02 -7.90278986e-02 -1.04210782e+00
4.22713608e-01 4.99569207e-01 -5.48451722e-01 -1.30641237e-01
3.76744777e-01 1.34001982e+00 -5.71992517e-01 -5.80144525e-01
3.63244981e-01 5.17406821e-01 -3.53030473e-01 5.36992490e-01
-1.96458876e-01 3.17060500e-01 5.61116673e-02 9.62484539e-01
4.31694806e-01 -4.29798782e-01 1.26110584e-01 -2.97792494e-01
-4.27995361e-02 -6.08177707e-02 -5.86733282e-01 1.71019173e+00
-7.21211493e-01 8.27816069e-01 3.39449346e-01 -5.90763867e-01
7.26909816e-01 5.68349421e-01 -3.18696313e-02 -9.66665268e-01
-6.06642105e-02 2.55459398e-01 -3.68429601e-01 -6.88728452e-01
6.56671464e-01 -1.06505513e-01 -4.72874492e-01 2.98195958e-01
7.22690940e-01 -3.12367529e-01 1.60699666e-01 6.25917852e-01
6.55603290e-01 1.73095688e-01 -2.22111538e-01 4.04625356e-01
6.20837510e-01 -2.15358436e-01 -6.51836023e-02 9.24195647e-01
1.88262705e-02 4.71790165e-01 5.04697621e-01 2.05401435e-01
-7.70163298e-01 -1.47593141e+00 2.40612224e-01 1.44441032e+00
2.46082440e-01 -3.92499804e-01 -6.02905691e-01 -7.45999157e-01
-1.37089819e-01 8.82644534e-01 -4.02914494e-01 -2.47393623e-01
-3.55769396e-01 -3.20505798e-01 8.97098184e-01 6.20819986e-01
5.30662775e-01 -1.02290738e+00 -5.73188476e-02 -1.12242766e-01
-3.57339114e-01 -1.35318983e+00 -7.31046200e-01 3.69985849e-01
-4.80551898e-01 -6.92183077e-01 -9.07000840e-01 -1.24820077e+00
8.53194833e-01 1.40325278e-01 1.10965395e+00 -9.86740217e-02
2.92303741e-01 9.37824249e-01 -2.52263337e-01 1.74723580e-01
-9.03398633e-01 -4.34180312e-02 -3.69018763e-02 -8.95021930e-02
1.13582663e-01 -5.80381826e-02 -1.04383782e-01 4.10410166e-01
-1.03240490e+00 5.23677707e-01 1.00505006e+00 9.86642599e-01
4.59198296e-01 -3.60180557e-01 4.38993126e-01 -4.97993320e-01
6.09206498e-01 -3.13563496e-01 -3.44976842e-01 6.63406909e-01
1.45761846e-02 1.40844369e-02 5.05555928e-01 -7.93858290e-01
-1.06469595e+00 2.79383883e-02 -3.35928768e-01 -6.70791268e-01
-2.86694020e-01 5.75535297e-01 -4.73662257e-01 -7.62363449e-02
1.08548664e-02 7.60204554e-01 6.43834025e-02 -2.30970934e-01
7.64907360e-01 9.81334984e-01 9.20972407e-01 -6.95279598e-01
5.68918824e-01 -6.31581768e-02 -4.05110389e-01 -9.21949863e-01
-3.17393124e-01 -2.02761292e-01 -4.68413353e-01 -2.39329889e-01
1.17877424e+00 -1.29813409e+00 -1.06404567e+00 5.39115965e-01
-1.09935498e+00 -6.76210582e-01 8.69671106e-02 5.25514483e-01
-7.30034292e-01 2.72570789e-01 -1.14211452e+00 -8.55797470e-01
3.75542380e-02 -1.87013888e+00 1.62512589e+00 3.46812993e-01
-2.75104612e-01 -1.14366710e+00 -2.19350949e-01 8.98248553e-01
1.76446632e-01 -5.10690928e-01 1.15064025e+00 -4.27145958e-01
-5.69574058e-01 5.46210520e-02 -4.07224834e-01 4.60459411e-01
-5.78391328e-02 -1.60405170e-02 -1.02607751e+00 -4.36865091e-01
-4.38479155e-01 -9.46913242e-01 8.56459141e-01 4.58281264e-02
6.78671479e-01 -2.78490454e-01 -3.31898034e-01 6.27232492e-01
1.03443789e+00 5.74022174e-01 6.16933465e-01 6.72207624e-02
8.62341404e-01 4.99868393e-01 2.93340892e-01 -2.81983316e-01
8.38756800e-01 4.56459075e-01 4.21741158e-01 -4.87088114e-01
-3.42891783e-01 -7.12885022e-01 1.15353370e+00 1.40865588e+00
3.98372173e-01 -5.85376143e-01 -9.07009900e-01 3.80769908e-01
-1.73862588e+00 -6.84896410e-01 1.36920348e-01 1.86010742e+00
8.74470413e-01 -9.06961411e-02 -1.82370663e-01 -5.85880458e-01
4.58585232e-01 4.41920348e-02 -4.32648689e-01 -6.19473338e-01
-4.89139408e-01 -1.30336732e-01 4.32790890e-02 8.64734650e-01
-9.55564678e-01 1.17040229e+00 6.36905098e+00 9.33833778e-01
-1.22699308e+00 -6.20965241e-03 6.31252646e-01 -4.98197041e-02
-6.35002136e-01 -1.69851646e-01 -9.44721580e-01 4.69521612e-01
1.08719671e+00 2.75891811e-01 5.01214027e-01 4.04826820e-01
-2.12997437e-01 -1.20444223e-01 -1.30072498e+00 1.19063795e+00
4.43074673e-01 -1.15054989e+00 6.73229277e-01 8.85572582e-02
6.54022932e-01 -4.66442481e-02 4.54660922e-01 9.23203349e-01
4.03214544e-01 -1.47548425e+00 9.02979374e-01 6.22874558e-01
1.25142252e+00 -5.02295971e-01 5.07920146e-01 4.55224186e-01
-1.32731974e+00 -3.47132050e-02 7.83672854e-02 3.55632365e-01
3.88954014e-01 -3.27038080e-01 -7.39398479e-01 4.90041733e-01
5.19308507e-01 5.11262774e-01 -5.79556763e-01 3.32913458e-01
-2.26050854e-01 4.84984726e-01 -7.37883300e-02 2.00295180e-01
5.15641510e-01 -2.12986380e-01 2.77171791e-01 1.52552354e+00
4.42567796e-01 -2.48243600e-01 2.95518398e-01 8.22697639e-01
-3.98549974e-01 -7.24066794e-02 -8.34604502e-01 -5.04742146e-01
2.66911060e-01 1.03463674e+00 -1.50545880e-01 -6.68589294e-01
-7.70176470e-01 1.17084014e+00 2.40955874e-01 8.45554650e-01
-7.56142497e-01 -2.01855764e-01 4.76076484e-01 -2.27601081e-01
-1.54411225e-02 -3.51526707e-01 1.90903991e-01 -1.35752988e+00
-3.08844924e-01 -1.21439815e+00 3.33600402e-01 -1.46393502e+00
-1.22292292e+00 9.72620547e-01 -9.97799486e-02 -1.17175961e+00
-3.66514802e-01 -9.13825512e-01 -2.72361487e-01 1.06860685e+00
-1.40288830e+00 -1.66608799e+00 2.46222364e-03 8.96230698e-01
8.20246994e-01 -5.03550887e-01 8.23341787e-01 2.50354856e-01
-6.04611933e-01 8.91212463e-01 -1.74198300e-01 3.66771430e-01
8.84258151e-01 -1.20380914e+00 1.78830072e-01 6.59546375e-01
1.92250788e-01 8.04637134e-01 2.22289383e-01 -5.15959263e-01
-1.81745422e+00 -1.02081180e+00 9.78588343e-01 -5.61798573e-01
7.20147908e-01 -4.85797197e-01 -8.74064803e-01 1.07460892e+00
7.92207778e-01 -4.85683143e-01 8.70829105e-01 -3.34367156e-01
-3.78116041e-01 2.19931781e-01 -7.04932272e-01 7.66203821e-01
6.30141735e-01 -1.22684658e+00 -6.31857336e-01 1.31695896e-01
8.91013563e-01 -5.23725033e-01 -9.58852291e-01 3.87662917e-01
3.86416823e-01 -6.20071292e-01 9.20058250e-01 -5.15512824e-01
6.45420074e-01 -3.99068773e-01 -6.92408204e-01 -1.27086365e+00
1.67071253e-01 -6.16540849e-01 -6.57213330e-02 1.32156110e+00
5.98761380e-01 -6.43474460e-01 2.57157147e-01 4.11151767e-01
-3.75531375e-01 -5.21557570e-01 -1.00425899e+00 -5.22975206e-01
1.15993232e-01 -5.34624219e-01 2.27555364e-01 7.38403320e-01
1.94875628e-01 5.97333789e-01 -5.00313580e-01 2.53989488e-01
3.37616146e-01 1.81430697e-01 9.69998360e-01 -2.64370978e-01
-4.19430643e-01 -5.15016794e-01 -3.47523810e-03 -1.71222067e+00
4.87878770e-01 -1.03909433e+00 2.19705895e-01 -1.39840364e+00
4.77749527e-01 -1.08226150e-01 -1.00639999e-01 9.29971278e-01
-3.80947664e-02 4.82248694e-01 4.66754466e-01 2.30640545e-01
-7.86604226e-01 7.05350399e-01 1.49085414e+00 -5.66176713e-01
-3.90537418e-02 -4.40340281e-01 -5.33477962e-01 5.75640917e-01
2.63943940e-01 1.81584850e-01 -6.20227098e-01 -7.43524194e-01
5.82443923e-02 6.80083096e-01 4.27444667e-01 -7.78986275e-01
3.59775007e-01 -3.61764356e-02 4.68459636e-01 -5.71021855e-01
7.97002017e-01 -6.10723257e-01 2.95129810e-02 5.61341047e-02
-5.30832648e-01 2.62312770e-01 4.94182914e-01 3.94964702e-02
-8.52296591e-01 1.08939537e-03 4.34774339e-01 -9.36695337e-02
-1.11158872e+00 -1.11878011e-02 -6.02864146e-01 -3.00550610e-01
7.19510555e-01 -5.50542623e-02 -7.34209418e-01 -8.00529122e-01
-1.08271253e+00 6.35530472e-01 6.85746789e-01 6.53969467e-01
9.58257139e-01 -1.45559728e+00 -3.38632166e-01 5.49861372e-01
2.87802309e-01 -2.21644357e-01 3.85665268e-01 9.18599486e-01
-2.03300580e-01 5.39243162e-01 -9.38703492e-03 -9.96894062e-01
-1.40360773e+00 5.93113065e-01 2.80681998e-01 -1.77725125e-02
-1.45192400e-01 8.40476334e-01 8.03557634e-01 -8.25159550e-01
4.41235721e-01 -2.62384713e-01 -4.09608707e-03 -3.00138369e-02
3.60957980e-01 -2.32791066e-01 -4.93246406e-01 -1.19499779e+00
-2.94421792e-01 8.07030678e-01 8.37139860e-02 -5.61986923e-01
8.91739726e-01 -4.95154381e-01 -1.16432168e-01 6.75700307e-01
1.52958345e+00 -1.37568600e-02 -1.11912513e+00 -7.35849217e-02
-4.26284820e-01 2.30714027e-02 -3.59237343e-01 -9.50604975e-01
-8.94102693e-01 1.30808032e+00 3.62313569e-01 -2.26126820e-01
1.26019752e+00 3.34068894e-01 8.27488661e-01 3.24355781e-01
3.39559168e-01 -8.28427374e-01 4.30832118e-01 7.25053728e-01
8.74547601e-01 -1.44003582e+00 -7.51180053e-01 -7.93458149e-03
-1.12807298e+00 1.07029319e+00 8.57553303e-01 5.88504851e-01
6.63995147e-02 4.60824549e-01 6.02044463e-01 -1.13918968e-02
-1.17263818e+00 -1.93979964e-01 5.18253028e-01 4.38302100e-01
4.88671064e-01 1.09152146e-01 6.15754962e-01 4.32008594e-01
-9.76593420e-02 -6.64683059e-02 1.57198921e-01 8.26796293e-01
-2.06257209e-01 -1.13990378e+00 -5.80599189e-01 -3.75634953e-02
1.51441738e-01 -2.51772255e-01 -3.59939694e-01 8.72396648e-01
1.45496011e-01 1.16584480e+00 6.98673874e-02 -7.37860560e-01
2.71100789e-01 3.48069519e-01 3.78301740e-01 -5.25207996e-01
-4.26960588e-01 3.62829775e-01 1.01157069e-01 -5.73758602e-01
-1.52268425e-01 -2.41003320e-01 -1.33894587e+00 -1.71179056e-01
1.53989583e-01 1.27421558e-01 5.38397610e-01 7.87514269e-01
3.03057224e-01 4.69540626e-01 3.17196935e-01 -9.22105312e-01
-1.91118106e-01 -8.06424379e-01 -1.24209970e-01 4.72541660e-01
5.79938591e-01 -2.83758014e-01 -2.72332013e-01 2.82922804e-01]
|
[10.986824035644531, 1.5183112621307373]
|
71ec62ad-167d-4c02-88d2-a8aebb2e3503
|
enhancement-of-underwater-images-with
|
1906.08673
| null |
http://arxiv.org/abs/1906.08673v1
|
http://arxiv.org/pdf/1906.08673v1.pdf
|
Enhancement of Underwater Images with Statistical Model of Background Light and Optimization of Transmission Map
|
Underwater images often have severe quality degradation and distortion due to
light absorption and scattering in the water medium. A hazed image formation
model is widely used to restore the image quality. It depends on two optical
parameters: the background light and the transmission map. Underwater images
can also be enhanced by color and contrast correction from the perspective of
image processing. In this paper, we propose an effective underwater image
enhancement method for underwater images in composition of underwater image
restoration and color correction. Firstly, a manually annotated background
lights (MABLs) database is developed. With reference to the relationship
between MABLs and the histogram distributions of various underwater images,
robust statistical models of BLs estimation are provided. Next, the TM of R
channel is roughly estimated based on the new underwater dark channel prior via
the statistic of clear and high resolution underwater images, then a scene
depth map based on the underwater light attenuation prior and an adjusted
reversed saturation map are applied to compensate and modify the coarse TM of R
channel. Next, TMs of G-B channels are estimated based on the difference of
attenuation ratios between R channel and G-B channels. Finally, to improve the
color and contrast of the restored image with a natural appearance, a variation
of white balance is introduced as post-processing. In order to guide the
priority of underwater image enhancement, sufficient evaluations are conducted
to discuss the impacts of the key parameters including BL and TM, and the
importance of the color correction. Comparisons with other state-of-the-art
methods demonstrate that our proposed underwater image enhancement method can
achieve higher accuracy of estimated BLs, less computation time, more superior
performance, and more valuable information retention.
|
[]
|
2019-06-19
| null | null | null | null |
['underwater-image-restoration']
|
['computer-vision']
|
[ 3.86941761e-01 -4.44496989e-01 8.57742965e-01 -3.33104312e-01
-2.72978246e-01 -1.20937623e-01 -2.77640019e-02 -3.56862657e-02
-9.01381135e-01 6.71912909e-01 1.71693772e-01 8.01474601e-02
-5.22240140e-02 -9.60487306e-01 -3.62700015e-01 -1.32143736e+00
3.79288793e-02 -4.54532892e-01 4.94254470e-01 -4.78094399e-01
4.64762866e-01 2.41754428e-01 -1.64564133e+00 -2.20560148e-01
1.28148913e+00 8.60051870e-01 6.96366549e-01 6.06790900e-01
-3.46331596e-01 5.60767055e-01 -5.80711663e-01 -7.82755911e-02
3.30342263e-01 -6.18251622e-01 -1.36136254e-02 1.89563647e-01
2.24593833e-01 -8.38757336e-01 -3.89240891e-01 1.57019556e+00
7.89676666e-01 3.56818706e-01 5.26336670e-01 -6.10231161e-01
-4.09272671e-01 2.35407189e-01 -5.92411876e-01 2.73995548e-01
-1.63292110e-01 1.58911914e-01 2.91870058e-01 -8.88145566e-01
1.65399283e-01 1.01972079e+00 3.87346566e-01 2.99037665e-01
-6.60689116e-01 -7.72584975e-01 1.63919050e-02 5.31458139e-01
-1.16046345e+00 -3.11346769e-01 6.39212430e-01 -2.17851788e-01
1.49061039e-01 2.49765322e-01 7.91633189e-01 -1.79879904e-01
4.60729927e-01 2.07829699e-01 1.32805812e+00 -5.20982683e-01
1.60411701e-01 -5.72662018e-02 -2.40414366e-02 5.52514493e-01
4.56082106e-01 8.59020352e-02 -2.69379675e-01 3.17118615e-01
5.40101767e-01 1.16999924e-01 -8.05398285e-01 2.26287425e-01
-4.35285389e-01 3.00687551e-01 3.89317095e-01 4.08007987e-02
-1.39359757e-01 -1.71966866e-01 2.33630333e-02 1.11141115e-01
2.63174534e-01 1.55151904e-01 -3.19957882e-01 1.88653007e-01
-6.42862022e-01 -2.09106952e-01 3.45559329e-01 5.84716499e-01
1.25291848e+00 2.96334088e-01 5.93044572e-02 9.57201958e-01
7.30134606e-01 1.07344532e+00 2.61452883e-01 -9.54893291e-01
1.68049827e-01 3.69708538e-01 2.71495789e-01 -7.54476905e-01
-2.35645637e-01 -2.18899008e-02 -7.58179545e-01 5.28570831e-01
7.23276436e-02 -1.72078028e-01 -1.19966185e+00 1.07690871e+00
2.36801028e-01 5.69487475e-02 5.28882802e-01 9.41233039e-01
8.30737770e-01 1.02063274e+00 -2.48662144e-01 -6.28158987e-01
1.06596696e+00 -4.91262019e-01 -1.24812436e+00 -2.11207286e-01
1.11308172e-01 -8.30693126e-01 8.14521372e-01 2.76914120e-01
-1.11368978e+00 -4.62410301e-01 -1.25426161e+00 3.15418020e-02
-3.89403105e-02 1.96060732e-01 1.48385644e-01 6.82182372e-01
-9.60314870e-01 4.18167382e-01 -7.41714537e-01 1.36719167e-01
-1.33237541e-01 4.73352298e-02 -1.41128479e-03 -5.55592537e-01
-1.16388547e+00 9.40281332e-01 3.04921478e-01 6.91679955e-01
-9.58896160e-01 -5.66451252e-01 -9.92701888e-01 -1.89037681e-01
-1.23251572e-01 -6.04634471e-02 8.64912391e-01 -8.13103616e-01
-1.60966837e+00 3.25398892e-01 -1.05803818e-01 1.57775253e-01
2.84789294e-01 -2.15184093e-01 -5.16387463e-01 4.99671340e-01
1.87633757e-03 -9.52875428e-03 6.93416476e-01 -1.65147662e+00
-9.73346889e-01 -4.15121555e-01 -9.74617675e-02 6.20862126e-01
-8.28876719e-02 -5.55335432e-02 -7.04406619e-01 -1.76437512e-01
4.66281950e-01 -4.11859185e-01 -2.92852521e-01 2.82308996e-01
-2.46486198e-02 5.48275352e-01 8.95484984e-01 -1.13972497e+00
9.88755822e-01 -2.52105999e+00 -7.38521740e-02 2.27591977e-01
-7.91940019e-02 2.73100257e-01 -1.00998409e-01 1.83198303e-01
3.60133767e-01 -2.81594604e-01 -5.95289469e-01 -1.12820650e-02
-4.46638823e-01 4.12988454e-01 1.02300912e-01 8.37270558e-01
-3.12326849e-01 -1.24425618e-02 -8.56005490e-01 -5.84227145e-01
4.26458448e-01 5.87872028e-01 -2.44343609e-01 5.35298407e-01
3.06347668e-01 4.73286033e-01 -3.79868001e-02 6.20580018e-01
1.33305418e+00 6.31028056e-01 1.07447557e-01 -6.25562012e-01
-8.05753648e-01 -3.54150414e-01 -1.25296950e+00 8.24390233e-01
-4.74052995e-01 6.29403293e-01 6.86580598e-01 -5.23193896e-01
1.27685452e+00 4.53945510e-02 2.32851714e-01 -9.31072056e-01
2.06930816e-01 3.33108395e-01 -1.16056532e-01 -9.30635393e-01
6.72525406e-01 -4.56511438e-01 6.86924338e-01 -8.64700302e-02
-4.42445606e-01 -2.95274019e-01 2.62375355e-01 -4.40301709e-02
3.12664121e-01 4.53127967e-03 -1.11639336e-01 -4.33778137e-01
5.86708367e-01 -4.03697163e-01 7.99450338e-01 3.95996571e-01
-7.38282278e-02 5.19189239e-01 -1.14334777e-01 -2.40086690e-01
-9.27125990e-01 -7.87963510e-01 -4.16684568e-01 7.86629140e-01
1.15507340e+00 2.91906148e-01 -6.07637465e-01 3.47244322e-01
-3.42489839e-01 5.65492153e-01 -4.53248978e-01 -2.34460503e-01
-5.06075144e-01 -1.23346424e+00 2.08801568e-01 5.81996851e-02
1.09996688e+00 -8.48097920e-01 -2.96836257e-01 1.70565665e-01
-2.22371280e-01 -9.46684659e-01 -1.41326100e-01 -4.90508005e-02
-9.81225193e-01 -1.10126913e+00 -7.95736134e-01 -8.70770454e-01
9.94314134e-01 7.03613639e-01 4.81909931e-01 5.90255320e-01
-8.76772404e-02 1.41104311e-01 -6.18219972e-01 -4.12020326e-01
-3.84951144e-01 -9.43224370e-01 -1.43231437e-01 2.51360357e-01
-1.36237442e-01 -3.61525059e-01 -9.88157451e-01 4.45379227e-01
-1.24650478e+00 -8.04734044e-03 7.05885887e-01 6.14217043e-01
2.54414350e-01 4.68560010e-01 -2.10529029e-01 -3.68478477e-01
5.63826077e-02 -2.05679551e-01 -9.55586016e-01 7.21098855e-02
-6.69621527e-01 -2.08108082e-01 2.18523338e-01 -7.89291635e-02
-1.63173437e+00 -3.28075975e-01 -1.03348657e-01 1.46624207e-01
8.04365203e-02 5.50054729e-01 -4.33313966e-01 -3.59361976e-01
1.73229501e-01 6.53614223e-01 2.91716009e-01 -5.19875884e-01
-2.02534482e-01 7.64055610e-01 4.53882009e-01 -1.26713455e-01
1.01312852e+00 8.02877188e-01 -6.50928020e-02 -1.31156409e+00
-4.17814136e-01 -5.00972033e-01 -3.20771128e-01 -5.92400551e-01
9.83367205e-01 -1.18877912e+00 -6.23013973e-01 1.17711318e+00
-8.10237467e-01 -5.33347189e-01 3.15375507e-01 7.94277310e-01
3.16180617e-01 1.00734746e+00 -1.05173302e+00 -1.08843195e+00
-3.55118304e-01 -1.26968586e+00 5.93730927e-01 7.74525940e-01
9.89771247e-01 -9.29248989e-01 -1.37296140e-01 2.01803818e-01
5.48808873e-01 -1.48848683e-01 6.48441553e-01 5.40415823e-01
-6.63582444e-01 1.71057597e-01 -5.23272932e-01 7.64711320e-01
3.68863702e-01 3.55045795e-01 -7.33424962e-01 -3.80447894e-01
-8.55695903e-02 2.78855920e-01 9.88813937e-01 6.25754058e-01
6.41449332e-01 -1.04595691e-01 6.88464046e-02 1.02150941e+00
1.88945794e+00 5.37651896e-01 1.42527044e+00 8.51122320e-01
4.98941332e-01 7.10418522e-01 9.99392211e-01 6.00001693e-01
4.33025211e-01 3.03053051e-01 8.36751282e-01 -5.38320005e-01
-1.14137501e-01 2.80898303e-01 5.37295461e-01 8.51062596e-01
-5.03082812e-01 -1.05264597e-01 -4.39152151e-01 5.02432823e-01
-1.12087846e+00 -8.06252837e-01 -5.80366492e-01 2.20945597e+00
9.10800099e-01 -2.61819631e-01 -6.29417181e-01 9.46742073e-02
9.62398827e-01 -8.60518515e-02 -2.29958043e-01 1.00228272e-01
-4.22237784e-01 -5.31860813e-02 1.00011396e+00 8.87458384e-01
-6.31463230e-01 6.27744913e-01 5.74527693e+00 4.39165533e-01
-1.07171452e+00 -1.55959696e-01 1.69782609e-01 4.84098226e-01
-6.10931754e-01 7.28779212e-02 -8.88300657e-01 7.70679653e-01
2.38697648e-01 2.47074097e-01 3.41614127e-01 3.34008127e-01
6.75993085e-01 -5.71971536e-01 -2.10737847e-02 8.40035737e-01
7.03377202e-02 -7.14477003e-01 8.36197212e-02 7.74886832e-02
8.22043836e-01 -1.14443831e-01 -2.08062008e-01 -4.02363241e-02
1.14780076e-01 -3.60971153e-01 7.35516191e-01 7.09498286e-01
8.43121350e-01 -4.79466856e-01 1.35811079e+00 -7.18106031e-02
-1.10829961e+00 -1.84233353e-01 -7.57474422e-01 -2.34773353e-01
3.55076313e-01 6.51173472e-01 -2.46548206e-01 4.31183785e-01
1.04395533e+00 4.72923964e-01 -3.76592875e-01 1.41948295e+00
-5.64342260e-01 4.81751770e-01 -2.45434508e-01 1.42401785e-01
-4.92856912e-02 -9.11897779e-01 3.43463838e-01 1.14004207e+00
5.44046640e-01 7.05691516e-01 -1.00219019e-01 4.46876198e-01
2.81131178e-01 1.70236140e-01 6.91645443e-02 4.49661821e-01
4.66014475e-01 1.10856366e+00 -4.09573466e-01 -3.55701655e-01
-4.68647271e-01 7.31818855e-01 -5.72783530e-01 6.81134939e-01
-4.90989923e-01 -6.08747840e-01 6.31537497e-01 -3.19248997e-02
1.42335698e-01 -2.70990670e-01 -5.13710193e-02 -9.15945649e-01
-2.32402086e-01 -5.36970913e-01 7.35561401e-02 -1.09627867e+00
-9.43561733e-01 3.49873573e-01 -3.00451647e-02 -1.43240380e+00
7.61010945e-01 -6.47363186e-01 -7.56799519e-01 9.88809288e-01
-2.28559041e+00 -6.46008492e-01 -1.00569201e+00 5.11076927e-01
3.48640114e-01 2.52086580e-01 4.15567309e-01 5.23290575e-01
-5.91596842e-01 9.65045542e-02 6.98930800e-01 1.12331271e-01
6.35519981e-01 -1.00756907e+00 -5.24275422e-01 1.37441695e+00
-8.91402900e-01 1.96197376e-01 1.03891695e+00 -7.17573762e-01
-1.21822977e+00 -8.26165199e-01 1.29591420e-01 5.31035006e-01
3.56765300e-01 3.69373024e-01 -1.09722996e+00 2.78174669e-01
1.37892902e-01 -1.48956820e-01 4.79669422e-01 -7.40994751e-01
3.05996478e-01 -5.44595540e-01 -1.06962764e+00 5.45177042e-01
2.57127613e-01 -9.61895511e-02 -2.35987529e-01 -1.12547420e-01
4.93228257e-01 -5.07648289e-01 -7.63397396e-01 6.61278903e-01
5.25764346e-01 -1.07495761e+00 7.26538122e-01 4.57572520e-01
3.29391599e-01 -9.85824466e-01 -3.80417615e-01 -1.26064575e+00
-1.12029947e-01 -7.66808242e-02 7.90704787e-01 1.16315937e+00
2.19364241e-01 -6.12138331e-01 3.10232073e-01 4.31523591e-01
-6.62230790e-01 6.96864650e-02 -5.75196862e-01 -2.39331707e-01
-4.80057865e-01 6.49414733e-02 1.79553762e-01 6.30278885e-01
-3.62659186e-01 -2.39207558e-02 -5.39317191e-01 1.04648149e+00
1.11431611e+00 3.44607607e-02 8.29186261e-01 -1.11309588e+00
1.12438582e-01 3.02300565e-02 -2.58562505e-01 -1.00799453e+00
-4.26513940e-01 5.08839823e-02 6.83907688e-01 -1.94843900e+00
2.71246582e-01 -1.72850192e-01 -2.65389502e-01 7.24949837e-02
-6.32812619e-01 6.46269202e-01 -4.75736149e-02 3.91897410e-01
-1.70722902e-01 9.26639974e-01 1.53010714e+00 -1.69760674e-01
-3.32804680e-01 -1.01669326e-01 -2.53166288e-01 6.13948703e-01
7.21304893e-01 -1.34456426e-01 5.45114279e-02 -7.22418547e-01
8.03280026e-02 6.75037801e-02 7.67111629e-02 -8.59182298e-01
2.37961099e-01 -2.93164164e-01 3.00621718e-01 -4.39568073e-01
3.62833500e-01 -9.52867985e-01 -2.13261768e-02 6.41552985e-01
2.84446865e-01 -3.83312404e-01 1.20736465e-01 5.40890574e-01
-5.05730391e-01 -6.17714226e-01 1.23597944e+00 -2.39179626e-01
-1.18618011e+00 1.28297895e-01 -6.39798284e-01 -4.60906237e-01
6.77467465e-01 -6.30844116e-01 -5.85620403e-01 -5.01401603e-01
-1.72812968e-01 1.56026959e-01 6.22804821e-01 -1.96333408e-01
1.09787488e+00 -5.34821212e-01 -8.82639110e-01 2.46945947e-01
8.13399255e-02 -1.15594994e-02 7.74655223e-01 9.14757669e-01
-1.40173733e+00 -7.19086528e-01 -3.60697687e-01 -3.45727444e-01
-1.37815797e+00 -6.09203465e-02 6.05699658e-01 3.73805791e-01
-6.53600633e-01 6.84725285e-01 4.01332587e-01 -5.94681017e-02
1.69701141e-03 -1.19258299e-01 -5.50329030e-01 -1.90507576e-01
9.54040349e-01 7.35507786e-01 -2.53244102e-01 -7.46737778e-01
-5.35251899e-03 9.89845514e-01 3.64247143e-01 1.60586543e-03
1.41574311e+00 -9.75618303e-01 -6.68214023e-01 1.50001824e-01
8.41857791e-01 4.08589453e-01 -1.63458455e+00 4.07941341e-02
-6.94900513e-01 -7.76480854e-01 4.89222020e-01 -5.55949986e-01
-1.30926490e+00 8.45797837e-01 9.04342651e-01 2.94249561e-02
1.50549424e+00 -4.11959469e-01 6.55523658e-01 -2.89019244e-03
2.04914674e-01 -1.07115567e+00 -8.83851666e-03 3.60920608e-01
4.36455190e-01 -1.17153597e+00 4.08441663e-01 -5.47428191e-01
-4.57955658e-01 1.37713253e+00 6.48838341e-01 2.63415426e-01
4.31451410e-01 4.78572309e-01 7.55982935e-01 6.29554968e-03
1.75068989e-01 -3.77040923e-01 -2.24181995e-01 5.43127716e-01
2.19888613e-02 -3.66048694e-01 -4.41434562e-01 3.59242149e-02
2.17944548e-01 -3.89127195e-01 1.01799357e+00 5.82550347e-01
-1.17290998e+00 -5.82751453e-01 -7.24847376e-01 -1.66855007e-03
-6.14688456e-01 -2.94106126e-01 5.83883107e-01 5.13504386e-01
2.66848087e-01 1.23950696e+00 -5.84057271e-02 -2.00183436e-01
2.79856473e-01 -6.78162813e-01 1.47304714e-01 -2.83043891e-01
2.78756529e-01 2.88654268e-01 -1.86293542e-01 -1.18844900e-02
-9.76807594e-01 -3.04548502e-01 -1.52817822e+00 -8.75821412e-02
-6.25701904e-01 6.29074574e-01 9.33600962e-01 7.33334303e-01
-3.62054527e-01 3.61404181e-01 7.58786261e-01 -1.08652747e+00
5.87743986e-03 -1.16139877e+00 -1.40180779e+00 3.04437608e-01
4.74191189e-01 -5.14331698e-01 -1.12663543e+00 1.77548423e-01]
|
[10.704217910766602, -3.4148592948913574]
|
e1f4816f-d740-41ba-bf36-3c334eef5e9c
|
greener-yet-powerful-taming-large-code
|
2303.05378
| null |
https://arxiv.org/abs/2303.05378v1
|
https://arxiv.org/pdf/2303.05378v1.pdf
|
Greener yet Powerful: Taming Large Code Generation Models with Quantization
|
ML-powered code generation aims to assist developers to write code in a more productive manner, by intelligently generating code blocks based on natural language prompts. Recently, large pretrained deep learning models have substantially pushed the boundary of code generation and achieved impressive performance. Despite their great power, the huge number of model parameters poses a significant threat to adapting them in a regular software development environment, where a developer might use a standard laptop or mid-size server to develop her code. Such large models incur significant resource usage (in terms of memory, latency, and dollars) as well as carbon footprint. Model compression is a promising approach to address these challenges. Several techniques are proposed to compress large pretrained models typically used for vision or textual data. Out of many available compression techniques, we identified that quantization is mostly applicable for code generation task as it does not require significant retraining cost. As quantization represents model parameters with lower-bit integer (e.g., int8), the model size and runtime latency would both benefit from such int representation. We extensively study the impact of quantized model on code generation tasks across different dimension: (i) resource usage and carbon footprint, (ii) accuracy, and (iii) robustness. To this end, through systematic experiments we find a recipe of quantization technique that could run even a $6$B model in a regular laptop without significant accuracy or robustness degradation. We further found the recipe is readily applicable to code summarization task as well.
|
['Bing Xiang', 'Parminder Bhatia', 'Murali Krishna Ramanathan', 'Mingyue Shang', 'Ben Athiwaratkun', 'Qing Sun', 'Yuchen Tian', 'Zijian Wang', 'Varun Kumar', 'Xiaopeng Li', 'Haifeng Qian', 'Baishakhi Ray', 'Shiqi Wang', 'Wasi Ahmad', 'Sujan Gonugondla', 'Xiaokai Wei']
|
2023-03-09
| null | null | null | null |
['model-compression']
|
['methodology']
|
[ 3.57120663e-01 1.20723464e-01 -3.66279751e-01 -7.72526935e-02
-8.15726280e-01 -5.77507615e-01 3.49483341e-01 1.21657118e-01
-3.47490609e-02 4.24813062e-01 -8.65581408e-02 -5.38474619e-01
1.05494373e-01 -1.00947249e+00 -1.05442035e+00 -3.50280702e-01
1.41406074e-01 -1.27030522e-01 -2.50061929e-01 -7.07308725e-02
5.40499926e-01 1.72356382e-01 -1.57318830e+00 1.80104345e-01
1.20517957e+00 9.41918492e-01 5.87747514e-01 6.45573378e-01
-2.95831561e-01 8.38710606e-01 -6.95361793e-01 -5.62606990e-01
4.17217433e-01 -3.71756226e-01 -7.45230317e-01 -2.73477644e-01
2.63023883e-01 -6.66271627e-01 -2.21560940e-01 1.10091448e+00
4.85013455e-01 -3.47699463e-01 5.52580059e-01 -1.17366028e+00
-8.59926820e-01 8.33165228e-01 -6.69432878e-01 -2.04033658e-01
1.83033407e-01 3.85403723e-01 7.39767432e-01 -6.10242367e-01
2.63284355e-01 8.74696314e-01 7.72739649e-01 5.78447998e-01
-1.19352186e+00 -8.58505309e-01 -1.83772221e-01 -1.31145149e-01
-1.19292307e+00 -5.93968034e-01 5.20239055e-01 -4.90737855e-01
1.27096331e+00 2.93772995e-01 7.14544892e-01 1.22066844e+00
3.81076515e-01 3.67370248e-01 6.66076362e-01 -4.61517245e-01
4.95692849e-01 1.53057232e-01 -2.99185365e-01 7.87695467e-01
7.15631604e-01 -2.86782444e-01 -6.04504526e-01 -2.99435854e-01
6.62389278e-01 4.98365648e-02 -7.01034218e-02 1.41297709e-02
-1.00920463e+00 7.64798403e-01 1.92256898e-01 5.23037314e-02
-1.94588855e-01 8.90353858e-01 3.91991377e-01 2.17723936e-01
2.60479450e-01 9.15808678e-01 -6.78664088e-01 -7.24297464e-01
-1.20133388e+00 3.23472887e-01 6.96147680e-01 1.37864840e+00
7.30761945e-01 2.70572066e-01 -1.63987249e-01 8.07805538e-01
2.38846630e-01 5.26097655e-01 7.68887281e-01 -1.13555467e+00
7.52164423e-01 7.03844428e-01 -8.42749178e-02 -9.66894567e-01
-3.35382223e-02 -2.91190952e-01 -9.76224065e-01 -7.64529258e-02
3.65866795e-02 -4.43858534e-01 -6.15599692e-01 1.69237518e+00
-8.85305405e-02 -1.11631639e-01 -1.69399709e-01 1.86663270e-01
3.02169114e-01 7.96424687e-01 -2.16440737e-01 -2.49994606e-01
1.19542909e+00 -1.00677025e+00 -4.50611174e-01 -4.12906319e-01
9.27707970e-01 -7.04008877e-01 1.47345853e+00 4.54416454e-01
-1.24075019e+00 -4.34817612e-01 -1.29253018e+00 -1.11723736e-01
-2.47708157e-01 2.06216350e-01 9.82049525e-01 9.32920933e-01
-1.19849217e+00 8.44025493e-01 -8.38820577e-01 -1.36902496e-01
6.62790358e-01 2.14949876e-01 3.67689021e-02 -1.03246793e-01
-8.02522242e-01 5.13068914e-01 1.74955338e-01 -2.70323604e-01
-9.07753110e-01 -9.47863102e-01 -6.46237910e-01 4.85065728e-01
9.45827588e-02 -8.22757065e-01 1.30405056e+00 -7.76425242e-01
-1.46567750e+00 2.60104567e-01 7.41750118e-04 -5.34759760e-01
2.65731186e-01 -3.40175807e-01 7.07477927e-02 -1.89768508e-01
-1.24589048e-01 8.19191575e-01 9.38626945e-01 -8.50866139e-01
-3.15081298e-01 -1.29566994e-02 1.82545140e-01 -1.93189234e-01
-1.20128608e+00 -9.00478587e-02 -2.13002950e-01 -8.44540715e-01
-2.37905294e-01 -9.79250848e-01 -2.16851458e-01 3.71520370e-02
-2.24017933e-01 -9.16233286e-02 6.16165936e-01 -7.85398543e-01
1.71354425e+00 -1.84922826e+00 -1.94009975e-01 -2.22757220e-01
1.69282570e-01 2.45753050e-01 -1.76682130e-01 3.98553401e-01
1.23834252e-01 8.17791343e-01 -2.22788244e-01 -1.13730514e-02
8.98202509e-02 -2.56373316e-01 -3.65511656e-01 -1.56434365e-02
3.59796107e-01 9.42748845e-01 -7.27201939e-01 -3.10902566e-01
-3.29865724e-01 3.61015230e-01 -1.05330503e+00 2.43690923e-01
-4.50702131e-01 -2.55806029e-01 -4.12883133e-01 6.58926308e-01
4.19485569e-01 -4.68165845e-01 5.75304441e-02 9.47920382e-02
-1.02754004e-01 4.51491058e-01 -6.77803576e-01 1.91086352e+00
-9.73321378e-01 8.45357776e-01 -2.32958198e-01 -8.34640861e-01
8.21064889e-01 1.49731174e-01 2.97079146e-01 -7.45912254e-01
-2.79315531e-01 1.79204985e-01 3.02778762e-02 -6.13988757e-01
5.83662987e-01 3.25803787e-01 -1.43345803e-01 8.95900428e-01
-3.49972725e-01 -3.71911615e-01 2.65849918e-01 1.29353195e-01
1.57257569e+00 2.34970301e-01 2.14113817e-01 -2.01412261e-01
-1.00216128e-01 -2.28848923e-02 3.95442158e-01 5.67810893e-01
1.48884386e-01 4.29443270e-01 7.25312471e-01 -3.22043836e-01
-1.33708835e+00 -5.01681566e-01 1.60462856e-01 1.11671960e+00
-3.45750958e-01 -9.18953180e-01 -1.24200559e+00 -3.79557282e-01
-3.06249022e-01 8.41008425e-01 -2.69939274e-01 -6.13245904e-01
-6.39220357e-01 -7.27230430e-01 7.70435929e-01 4.70498681e-01
5.56483030e-01 -7.11504102e-01 -1.01799774e+00 2.00651392e-01
-1.54848754e-01 -6.56425834e-01 -6.91247344e-01 1.35552272e-01
-1.01714599e+00 -5.85444450e-01 -6.38388336e-01 -5.40858984e-01
7.58506835e-01 3.56207967e-01 1.15757012e+00 3.06231886e-01
-4.09861594e-01 3.09057273e-02 -4.03374374e-01 -7.05426991e-01
-5.85597992e-01 2.58522272e-01 9.80286896e-02 -6.69531167e-01
-2.60059745e-03 -7.25703418e-01 -7.89330184e-01 -1.63132772e-01
-9.59816098e-01 3.82400453e-01 8.53061080e-01 6.51580989e-01
3.84829938e-01 2.76757181e-01 6.31455421e-01 -5.81095159e-01
8.94597411e-01 -5.82015753e-01 -7.51414359e-01 3.95799190e-01
-1.16155279e+00 3.02610904e-01 9.49418902e-01 -4.51190621e-01
-8.09629321e-01 1.44199571e-02 1.24371340e-02 -1.24029845e-01
2.98833400e-01 5.97176552e-01 3.60882166e-03 2.25376356e-02
8.43825519e-01 3.53676260e-01 -1.63670525e-01 -3.84193033e-01
3.06169033e-01 8.87113988e-01 5.19946665e-02 -7.19170570e-01
8.23309839e-01 -1.53929621e-01 -1.68089867e-01 -4.86980885e-01
-1.91904560e-01 2.94093758e-01 -2.12925017e-01 9.72856432e-02
6.82429731e-01 -9.30608034e-01 -4.74165857e-01 1.89982578e-01
-1.33033311e+00 -4.25943494e-01 -9.99578610e-02 4.78732213e-03
-5.08386374e-01 2.40468293e-01 -4.62638438e-01 -6.03367805e-01
-7.56731927e-01 -1.31815231e+00 9.72961903e-01 1.40478760e-01
-2.67179519e-01 -6.13928199e-01 -1.00385927e-01 4.36491251e-01
7.95878887e-01 3.02832782e-01 1.40190840e+00 -6.93379268e-02
-7.23400116e-01 -1.84977263e-01 -2.50386775e-01 5.14085293e-01
1.78948805e-01 2.43437082e-01 -8.50078106e-01 -3.99698764e-01
4.09078076e-02 -4.25875276e-01 5.18067420e-01 8.80128369e-02
1.85543203e+00 -8.32857907e-01 -1.22482263e-01 6.47132874e-01
1.43550050e+00 3.20933580e-01 4.87722427e-01 1.08977765e-01
7.66109765e-01 2.97362179e-01 4.72564280e-01 6.70144439e-01
2.72474408e-01 5.60912192e-01 4.85332817e-01 1.68305919e-01
-3.95499617e-02 -3.78155470e-01 7.81111419e-01 1.15542305e+00
-2.68518459e-03 -2.60327578e-01 -1.31172228e+00 4.48176414e-01
-1.40972948e+00 -7.17254937e-01 3.00597638e-01 2.38297558e+00
1.25228584e+00 1.87239110e-01 -8.88679922e-02 1.18895993e-01
3.41358334e-01 -2.24689901e-01 -8.06095958e-01 -6.80120885e-01
3.84994417e-01 2.85116732e-01 5.94482243e-01 -2.42693260e-01
-5.16971767e-01 6.35463595e-01 6.42406559e+00 1.09955132e+00
-1.17225683e+00 1.73351973e-01 1.01446080e+00 -4.94072646e-01
-5.51612258e-01 -1.22242365e-02 -6.98097169e-01 7.83943713e-01
1.59997010e+00 -6.21943712e-01 8.25923800e-01 1.29136717e+00
3.35183889e-01 7.19171539e-02 -1.50437355e+00 1.16189039e+00
-1.58104613e-01 -1.76123595e+00 3.23052347e-01 3.54003608e-01
9.86247957e-01 -1.64062873e-01 2.34858245e-01 4.77203965e-01
1.41560674e-01 -1.40300727e+00 8.34699094e-01 2.12901622e-01
1.23997855e+00 -7.21064687e-01 4.42770481e-01 5.11567652e-01
-1.01385725e+00 -3.03074896e-01 -6.05537355e-01 -2.12131739e-01
-3.01059872e-01 5.35647869e-01 -9.38374162e-01 7.03005791e-02
5.40711582e-01 3.69741470e-01 -9.06074762e-01 6.92771256e-01
2.92656779e-01 8.50113690e-01 -1.24498121e-01 -1.36552244e-01
9.80865490e-03 2.19838638e-02 -1.01711825e-01 1.23971975e+00
9.17598724e-01 -7.97314122e-02 -3.77845764e-01 1.28011787e+00
-4.97058213e-01 -7.68260984e-03 -6.92352176e-01 -5.67582309e-01
8.16731989e-01 1.03955936e+00 -5.48864365e-01 -3.15864354e-01
-4.20677006e-01 8.47598851e-01 2.45974019e-01 4.67357747e-02
-1.14512110e+00 -5.78909993e-01 6.58660293e-01 1.94410890e-01
1.48798227e-01 -1.57765538e-01 -6.40183330e-01 -9.42556679e-01
2.87714094e-01 -1.14607084e+00 -4.17865008e-01 -6.96754098e-01
-6.40973806e-01 2.90746510e-01 -8.29251185e-02 -1.26561451e+00
-3.77661914e-01 -2.77341902e-01 -5.49595416e-01 6.87511086e-01
-1.12156677e+00 -6.95347607e-01 -4.06908184e-01 2.22821496e-02
8.24940920e-01 -3.34881216e-01 7.39045322e-01 3.20976526e-01
-8.08549404e-01 1.02643979e+00 2.20554367e-01 -2.59876758e-01
4.16981667e-01 -1.02518523e+00 8.78297329e-01 9.51467693e-01
-1.02203526e-02 1.13571405e+00 4.77378309e-01 -5.65727115e-01
-1.89935482e+00 -1.32775974e+00 5.77301085e-01 -3.97877395e-01
3.80511850e-01 -5.64046144e-01 -5.04618645e-01 2.46010080e-01
1.03200540e-01 -3.37427527e-01 6.32644594e-01 -2.89561361e-01
-2.71024406e-01 -2.14720473e-01 -1.09104276e+00 7.18945503e-01
1.10115647e+00 -4.89542842e-01 1.80631965e-01 5.99863231e-01
1.37499881e+00 -2.73645341e-01 -1.05250263e+00 1.19369619e-01
5.96769333e-01 -8.80306840e-01 8.05824578e-01 -3.41831118e-01
1.29363847e+00 5.08954339e-02 -1.95372164e-01 -1.09591222e+00
-1.95292085e-01 -8.34836483e-01 -5.87847829e-01 1.43133712e+00
5.52647233e-01 -1.91937864e-01 6.53907716e-01 1.15418744e+00
-1.74322039e-01 -1.21349096e+00 -5.60873687e-01 -7.18983650e-01
2.24640593e-01 -3.94764811e-01 1.03680778e+00 7.34096527e-01
5.86140566e-02 8.23250860e-02 -3.15147877e-01 -3.29689384e-01
3.01823020e-01 -9.15188566e-02 7.79328823e-01 -8.10467482e-01
-5.47192991e-01 -5.93420863e-01 -4.56454791e-02 -1.02235496e+00
-6.21806495e-02 -8.71839285e-01 -6.22738861e-02 -1.41712630e+00
4.53836262e-01 -5.54546714e-01 1.79899521e-02 5.74798048e-01
-1.16032489e-01 -5.93877621e-02 2.72941828e-01 2.04217821e-01
-3.34780902e-01 3.64924073e-01 8.70098054e-01 -2.22911015e-01
-2.10586429e-01 -1.50703371e-01 -1.20651042e+00 5.53353965e-01
8.91453981e-01 -5.58807135e-01 -7.48255074e-01 -9.49190438e-01
8.90188873e-01 2.00850442e-01 1.22532342e-02 -1.08540845e+00
1.18199766e-01 -2.37968221e-01 1.74464732e-01 -3.65357995e-02
-1.39254272e-01 -6.63500190e-01 3.40268075e-01 7.16640651e-01
-4.66315806e-01 2.75398880e-01 1.64557531e-01 4.49479103e-01
8.37890059e-02 -4.28268880e-01 7.10055590e-01 -2.56850868e-01
-2.28372008e-01 1.97272271e-01 -1.81690931e-01 -6.76911399e-02
9.29992557e-01 -3.60716403e-01 -5.32894313e-01 -1.67692274e-01
1.85349926e-01 -7.21087726e-03 7.85829782e-01 5.33082366e-01
5.50940156e-01 -1.25157201e+00 -4.67469782e-01 1.42103031e-01
1.00505147e-02 1.41278759e-01 1.33190434e-02 4.52376276e-01
-8.98208976e-01 5.46394587e-01 -1.28734633e-01 -2.57957518e-01
-9.17995214e-01 5.09242535e-01 9.58808661e-02 -3.02472413e-01
-2.89394140e-01 8.91710222e-01 9.52918455e-02 -2.26217527e-02
2.14779437e-01 -6.73487306e-01 2.60887563e-01 -1.58016592e-01
6.61223471e-01 6.53256416e-01 2.57328421e-01 8.23254362e-02
-5.57581820e-02 3.59780014e-01 -7.50877708e-02 2.78379142e-01
1.46550512e+00 1.02568381e-01 -3.02432418e-01 1.55673817e-01
1.12227821e+00 -1.25915408e-01 -1.37453902e+00 3.01571250e-01
-1.56497881e-01 -4.08455342e-01 1.69428885e-01 -6.58748329e-01
-1.07249117e+00 9.56909060e-01 6.38778925e-01 3.07983488e-01
1.14657176e+00 -4.44177508e-01 1.09415305e+00 5.85618556e-01
6.35704517e-01 -1.17962158e+00 3.89483720e-01 1.86753199e-01
9.10111725e-01 -1.12453496e+00 1.87049016e-01 4.47459705e-02
-3.15771013e-01 1.16483855e+00 8.45788360e-01 3.43595535e-01
2.58845508e-01 5.08891702e-01 -5.70413888e-01 6.28668889e-02
-1.17404640e+00 6.20770872e-01 -9.97961313e-02 5.98875284e-01
8.38723958e-01 3.10213920e-02 -1.74579158e-01 6.54667795e-01
-5.50536036e-01 -4.15564254e-02 9.02714968e-01 7.71259367e-01
-4.82333750e-01 -1.04190230e+00 -2.74573237e-01 9.68172908e-01
-4.67052996e-01 -5.19966662e-01 -1.35636091e-01 1.70675144e-01
2.65342951e-01 9.33826625e-01 -1.16619349e-01 -5.86826503e-01
4.59112488e-02 5.12053221e-02 1.99434280e-01 -7.41412759e-01
-6.12365186e-01 -2.61716396e-01 -1.41599134e-01 -6.86598897e-01
1.11878425e-01 -4.56782490e-01 -1.07945490e+00 -6.21123910e-01
-4.06471431e-01 -2.39285052e-01 9.32106674e-01 4.51472372e-01
8.18704665e-01 6.29650950e-01 6.19071245e-01 -8.49177897e-01
-7.42275000e-01 -9.90530968e-01 -2.00279832e-01 -9.02159661e-02
9.23093781e-02 -2.33972728e-01 -1.30716473e-01 3.82430315e-01]
|
[7.880749225616455, 7.6932454109191895]
|
7aa942f9-83a3-452d-91b5-bb9be20be2ed
|
assurance-monitoring-of-cyber-physical
|
2001.05014
| null |
https://arxiv.org/abs/2001.05014v2
|
https://arxiv.org/pdf/2001.05014v2.pdf
|
Assurance Monitoring of Cyber-Physical Systems with Machine Learning Components
|
Machine learning components such as deep neural networks are used extensively in Cyber-Physical Systems (CPS). However, they may introduce new types of hazards that can have disastrous consequences and need to be addressed for engineering trustworthy systems. Although deep neural networks offer advanced capabilities, they must be complemented by engineering methods and practices that allow effective integration in CPS. In this paper, we investigate how to use the conformal prediction framework for assurance monitoring of CPS with machine learning components. In order to handle high-dimensional inputs in real-time, we compute nonconformity scores using embedding representations of the learned models. By leveraging conformal prediction, the approach provides well-calibrated confidence and can allow monitoring that ensures a bounded small error rate while limiting the number of inputs for which an accurate prediction cannot be made. Empirical evaluation results using the German Traffic Sign Recognition Benchmark and a robot navigation dataset demonstrate that the error rates are well-calibrated while the number of alarms is small. The method is computationally efficient, and therefore, the approach is promising for assurance monitoring of CPS.
|
['Xenofon Koutsoukos', 'Dimitrios Boursinos']
|
2020-01-14
| null | null | null | null |
['traffic-sign-recognition']
|
['computer-vision']
|
[-1.38917133e-01 4.80910182e-01 -1.59944482e-02 -4.24114227e-01
-4.79293644e-01 -3.43096614e-01 5.02338707e-01 4.61232476e-02
-2.12496921e-01 6.69966102e-01 -5.34181476e-01 -6.63999200e-01
-3.21910888e-01 -1.04581964e+00 -9.64979768e-01 -7.64919400e-01
-2.10804701e-01 3.38481188e-01 4.61119741e-01 -4.83758859e-02
-1.90669559e-02 8.93494248e-01 -1.60669172e+00 -1.08671710e-02
7.27661967e-01 1.67424250e+00 -5.01578629e-01 3.81747037e-01
3.35839033e-01 8.70872974e-01 -6.90973997e-01 -5.64592838e-01
3.42789620e-01 1.89732015e-01 -3.54836583e-01 -5.57722092e-01
1.83761880e-01 -5.06434917e-01 -2.96757191e-01 1.25278151e+00
5.46197779e-03 -2.66712084e-02 7.85791814e-01 -1.88083661e+00
-2.62587637e-01 1.50872201e-01 1.04301102e-01 -3.44560057e-01
-6.99653774e-02 2.81329960e-01 7.86680043e-01 -5.30058682e-01
1.16472924e-02 9.64475691e-01 8.91752481e-01 4.25135732e-01
-9.29433763e-01 -8.51666391e-01 1.31687671e-01 4.81282026e-01
-9.52516198e-01 -7.59818181e-02 6.86874270e-01 -4.65420812e-01
1.01235044e+00 1.29783764e-01 5.22111654e-01 1.34475327e+00
8.37875962e-01 3.83271754e-01 6.40602767e-01 -1.97489724e-01
7.01896608e-01 3.29555631e-01 3.10835302e-01 6.87099278e-01
7.62805283e-01 8.40970814e-01 -1.54130235e-01 -2.63217807e-01
3.43400687e-01 5.83010614e-02 4.27289158e-02 -5.84518433e-01
-8.67261827e-01 9.36419785e-01 5.82436264e-01 2.51563609e-01
-5.18637538e-01 4.78594780e-01 4.97478873e-01 2.84730554e-01
1.52271375e-01 5.14349401e-01 -5.53788066e-01 -1.23807423e-01
-3.21749896e-01 6.55284559e-04 8.19642663e-01 9.80417252e-01
3.80232632e-01 4.18172151e-01 5.20723723e-02 1.50891498e-01
2.72580355e-01 6.39556766e-01 -6.25783056e-02 -9.57508266e-01
1.70226082e-01 5.48153162e-01 4.02081639e-01 -1.20013309e+00
-5.31122029e-01 -3.44340801e-01 -8.68578315e-01 9.11317229e-01
2.21968502e-01 -7.88420886e-02 -7.04671979e-01 1.47305441e+00
3.50456238e-01 2.25569755e-01 1.82414547e-01 5.90354919e-01
1.55843437e-01 4.46993381e-01 -2.63613723e-02 2.35188872e-01
1.14229453e+00 -6.14555776e-01 -6.71642900e-01 -1.47951409e-01
6.75796270e-01 -6.44871220e-02 8.81734490e-01 9.11294103e-01
-5.47390282e-01 -5.13546646e-01 -1.63705516e+00 5.87920964e-01
-5.96663773e-01 -4.99652624e-02 4.12838846e-01 1.14295816e+00
-6.87902272e-01 8.56084466e-01 -9.85040963e-01 4.58305478e-02
3.31215084e-01 3.32756966e-01 -3.23748469e-01 7.67409280e-02
-1.32305622e+00 1.30432725e+00 5.77108562e-01 2.60689557e-01
-1.10029519e+00 -5.25385857e-01 -8.70161831e-01 2.13978007e-01
2.41340742e-01 -2.23012045e-01 1.21890414e+00 -3.57248098e-01
-1.49027944e+00 -1.85588840e-02 7.24405885e-01 -8.24644148e-01
6.53606653e-01 -3.85974705e-01 -9.21569943e-01 5.57082519e-02
-3.10100943e-01 3.69286448e-01 9.04223979e-01 -1.08188033e+00
-9.26063597e-01 -7.11273588e-03 1.50924414e-01 -7.82046974e-01
-6.80538416e-01 -8.52315649e-02 3.95699531e-01 -2.50046537e-03
-3.72325182e-02 -1.14336407e+00 -1.42870054e-01 2.31604278e-01
-2.78207660e-01 3.94023620e-02 9.13210452e-01 -7.17299283e-01
6.58749521e-01 -2.18354917e+00 -4.28126425e-01 5.67971826e-01
8.39114264e-02 2.85370469e-01 -5.72921261e-02 1.42878696e-01
-8.18791464e-02 -1.54936314e-01 -2.42293656e-01 2.44661774e-02
4.01631504e-01 2.70825684e-01 -5.50118089e-01 6.65895104e-01
6.84734583e-01 6.48137093e-01 -7.98975289e-01 1.04509760e-02
4.63603944e-01 5.86256146e-01 -3.78136426e-01 2.84853935e-01
-1.69150114e-01 2.27715746e-01 -3.33472908e-01 6.23112202e-01
5.88117182e-01 3.58946179e-03 -1.56433821e-01 -1.85493290e-01
3.50808680e-01 9.69623104e-02 -1.17260969e+00 7.94421673e-01
-5.23362637e-01 6.98818147e-01 -2.28242844e-01 -1.06735766e+00
1.10982049e+00 2.80663878e-01 3.29972148e-01 -7.17474043e-01
4.79014963e-01 2.41712153e-01 -6.87262118e-02 -3.25281233e-01
3.12705129e-01 -5.61428145e-02 -3.97072315e-01 3.48075449e-01
2.28750482e-02 -1.55508950e-01 -2.64076263e-01 -2.31884390e-01
1.53848326e+00 -1.32827088e-01 -1.63378455e-02 -7.52421329e-03
4.24745798e-01 -5.51307946e-02 7.45326400e-01 5.18215418e-01
-3.07050914e-01 -6.78937882e-02 5.61162949e-01 -7.61842549e-01
-9.71346080e-01 -1.05698287e+00 -6.39809966e-02 4.60894495e-01
6.10347763e-02 4.65164669e-02 -6.95028424e-01 -9.36759889e-01
1.11866921e-01 1.27751839e+00 -5.30617833e-01 -9.83520210e-01
-4.14436191e-01 -3.16049844e-01 8.36220622e-01 1.05883360e+00
2.51186579e-01 -7.92357445e-01 -1.01087248e+00 3.49240243e-01
2.36713976e-01 -1.45638442e+00 2.97959834e-01 3.46747041e-01
-6.12229049e-01 -1.26546216e+00 -9.18689966e-02 -3.35426629e-01
6.93829596e-01 -2.40365818e-01 8.84269357e-01 -2.59748865e-02
-1.07732497e-01 5.10194004e-01 -3.06344301e-01 -8.26749384e-01
-8.06861639e-01 -4.40131426e-01 6.38298631e-01 -1.20684570e-02
4.79275495e-01 -2.97523439e-01 -2.52053142e-01 5.89056313e-01
-7.82771289e-01 -5.58790445e-01 5.55059195e-01 1.08160639e+00
1.60929561e-01 4.31389064e-01 6.85220718e-01 -1.76284790e-01
6.85290754e-01 -2.46379524e-01 -1.33592939e+00 3.75440150e-01
-9.96031284e-01 1.32596314e-01 6.82668686e-01 -6.92978024e-01
-8.36881578e-01 -1.03271790e-01 2.38589291e-02 -5.33420682e-01
-2.76498407e-01 4.33095306e-01 -2.01376796e-01 -4.25165504e-01
7.53962576e-01 -1.70487717e-01 2.74562240e-02 -6.10937774e-02
1.69659361e-01 4.71498430e-01 4.11396056e-01 -4.95629579e-01
1.03460741e+00 3.23085368e-01 4.26913977e-01 -4.71331418e-01
-4.15172458e-01 1.94182187e-01 -4.80112642e-01 -5.97563684e-01
6.79592967e-01 -5.73750436e-01 -9.77290571e-01 3.37599128e-01
-1.28278041e+00 -1.38994977e-01 -3.26182485e-01 5.67886651e-01
-4.86708462e-01 3.72393459e-01 -5.58175862e-01 -1.26533854e+00
-2.32531115e-01 -1.11881638e+00 7.80654132e-01 -1.51513264e-01
-2.86437184e-01 -6.85913026e-01 -8.64671394e-02 -6.51763305e-02
4.45753932e-01 5.23554921e-01 1.01507735e+00 -9.54563081e-01
-6.40188336e-01 -9.69290733e-01 -5.37742674e-02 1.02257001e+00
-5.53474687e-02 1.15389384e-01 -1.29885137e+00 -2.88485795e-01
2.66983688e-01 -2.18947574e-01 3.55439126e-01 8.36350098e-02
1.08569670e+00 -4.11801070e-01 -2.28021145e-01 1.21294349e-01
9.34063911e-01 5.33673167e-01 6.16852462e-01 4.46354508e-01
2.10376412e-01 8.52515817e-01 9.63168502e-01 3.65003854e-01
-1.03619710e-01 3.52622271e-01 9.39793408e-01 2.01795667e-01
5.84579170e-01 -9.34052095e-02 5.11025071e-01 5.48573554e-01
2.10581645e-01 -1.19135547e-02 -1.05799770e+00 2.94746190e-01
-1.85804617e+00 -6.99708581e-01 2.32776571e-02 2.28660560e+00
2.36892909e-01 7.23656654e-01 -1.02712967e-01 6.78423882e-01
4.49720860e-01 -4.75290626e-01 -6.23771012e-01 -4.64644551e-01
2.59667695e-01 6.98529333e-02 5.51253319e-01 1.49439871e-01
-9.82173860e-01 3.96659642e-01 6.14989853e+00 4.55687523e-01
-9.33378994e-01 3.00827086e-01 4.59377468e-01 9.48248208e-02
3.15017328e-02 -2.14058310e-01 -6.54718399e-01 4.61142927e-01
1.65631878e+00 2.53283590e-01 -6.53545335e-02 1.34168959e+00
-2.16285959e-01 1.43334925e-01 -1.42546070e+00 6.19585395e-01
-2.70990670e-01 -1.18341732e+00 -4.84337807e-01 3.91281843e-02
2.32849687e-01 1.84278581e-02 2.15860128e-01 6.47722006e-01
4.52080488e-01 -8.86828423e-01 8.78896534e-01 5.90517223e-01
5.26151240e-01 -1.14576483e+00 1.12341940e+00 4.52239275e-01
-8.58835876e-01 -5.87510705e-01 -4.19148654e-01 7.33108819e-02
2.69928843e-01 7.97623277e-01 -8.79659355e-01 4.78133976e-01
8.30410123e-01 1.29709244e-01 -2.30257198e-01 1.04803419e+00
-4.88164395e-01 4.32837784e-01 -3.36616814e-01 -3.93117309e-01
1.48489803e-01 -2.86056306e-02 3.77724409e-01 7.70384431e-01
5.95063210e-01 -4.61696506e-01 -9.07942280e-02 1.07242358e+00
4.08540249e-01 -6.10219240e-01 -9.25365269e-01 -1.19123831e-01
4.54273254e-01 1.04336846e+00 -4.34639454e-01 -1.33470178e-01
-2.47607797e-01 5.10896742e-01 1.02896370e-01 2.68945903e-01
-1.42643440e+00 -4.02968705e-01 8.23373795e-01 -1.22207657e-01
2.10437775e-01 -2.71731734e-01 -3.59954417e-01 -6.58796728e-01
3.07522535e-01 -6.72255099e-01 1.56145632e-01 -6.28712475e-01
-1.42014515e+00 8.62703085e-01 -2.31793970e-02 -1.76451743e+00
-5.50471425e-01 -1.20973074e+00 -4.60923404e-01 3.44229728e-01
-1.44556117e+00 -1.05943787e+00 -2.85625249e-01 2.37143651e-01
-1.82723522e-01 -4.71731842e-01 1.01792109e+00 2.26257890e-01
-4.98492986e-01 6.37936115e-01 -4.58882637e-02 1.31760940e-01
4.21536952e-01 -8.32908511e-01 4.45302755e-01 9.54825699e-01
-9.05319229e-02 1.42196223e-01 8.54206979e-01 -7.04433262e-01
-1.14168704e+00 -1.31855536e+00 4.60779071e-01 -5.08845985e-01
9.22503054e-01 -3.89586866e-01 -7.82314003e-01 5.48186004e-01
-1.98025569e-01 1.60171360e-01 5.63186765e-01 4.82212417e-02
-6.77402675e-01 -3.23162019e-01 -1.58949518e+00 5.41468680e-01
5.60961783e-01 -6.20950162e-01 -5.20111799e-01 3.31539661e-01
1.04865324e+00 -1.09370872e-01 -1.05123591e+00 7.38535106e-01
4.19105679e-01 -7.66626000e-01 9.08836424e-01 -8.26373935e-01
2.77194679e-02 -2.77809858e-01 -2.19559044e-01 -1.23951113e+00
-2.84510255e-01 -3.04920226e-01 -6.30766034e-01 8.93071055e-01
6.02645993e-01 -1.20102060e+00 5.96211672e-01 1.23234320e+00
-2.86724210e-01 -5.19220293e-01 -1.39507318e+00 -1.56373036e+00
9.02508721e-02 -1.03449619e+00 8.78314078e-01 6.98725760e-01
2.20234860e-02 -2.90267885e-01 -3.38359088e-01 6.75593555e-01
8.52684736e-01 -4.57744420e-01 6.23434126e-01 -1.60358775e+00
-1.70930594e-01 -3.63936633e-01 -9.00144279e-01 -1.40723765e-01
4.47885662e-01 -3.04929137e-01 2.91030228e-01 -9.79403675e-01
-4.18458521e-01 -4.30640668e-01 -7.28753924e-01 5.67352355e-01
4.78862792e-01 -2.74363548e-01 -1.12605676e-01 -3.40262681e-01
-2.43975610e-01 7.55786061e-01 3.51815790e-01 -6.62854373e-01
2.46146098e-01 1.57087952e-01 -1.66836992e-01 8.09326649e-01
9.41490233e-01 -6.33955777e-01 -2.90643275e-01 3.16321366e-02
2.32838288e-01 -1.50580421e-01 6.81707621e-01 -1.53794980e+00
3.52577269e-01 -9.13207531e-02 2.92629093e-01 -6.01318300e-01
6.80654526e-01 -1.60932410e+00 2.23597661e-01 9.16657150e-01
-6.41285181e-02 2.86298580e-02 4.30496871e-01 8.81192267e-01
-1.62738472e-01 -3.97481263e-01 7.94838250e-01 5.59142292e-01
-4.26099539e-01 1.85154721e-01 -6.13443732e-01 -6.50146723e-01
1.43192470e+00 -1.52074024e-01 -3.37564558e-01 -3.55753809e-01
-4.19093579e-01 7.71049708e-02 1.64620340e-01 6.26319468e-01
1.01750660e+00 -1.67106843e+00 -2.20618308e-01 3.78303945e-01
3.90664577e-01 -2.87807703e-01 1.04209661e-01 5.70015073e-01
-2.88126647e-01 6.19491696e-01 -3.72907549e-01 -7.46860683e-01
-9.24173534e-01 8.47585499e-01 4.37157243e-01 -7.79108405e-02
-5.72981000e-01 4.68453407e-01 -6.99236393e-02 -6.06007516e-01
6.42071426e-01 -8.27251971e-01 -5.41471094e-02 -2.87512153e-01
7.68831789e-01 4.21838015e-01 4.30707783e-01 -2.05151960e-01
-4.02081996e-01 1.05503723e-02 1.65458709e-01 1.79771502e-02
1.43744791e+00 5.46088934e-01 2.08089769e-01 2.94928581e-01
8.28126848e-01 -5.85133493e-01 -1.44598472e+00 9.07975510e-02
4.84527886e-01 -3.54165256e-01 1.70795083e-01 -7.77584612e-01
-1.21424341e+00 1.02618814e+00 1.05958748e+00 4.20098990e-01
9.23376501e-01 -4.76466954e-01 7.55956054e-01 9.02129829e-01
9.70139921e-01 -1.31656098e+00 2.52242208e-01 5.11395395e-01
9.51998115e-01 -1.08652318e+00 -4.24289823e-01 -1.77146718e-01
-4.49704528e-01 1.16425681e+00 7.43388534e-01 -1.83096021e-01
8.77377450e-01 5.64676225e-01 -2.37653390e-01 -1.01571165e-01
-6.74388945e-01 4.19073522e-01 2.23435298e-01 9.59958375e-01
-3.84666204e-01 2.43200392e-01 3.81586462e-01 9.19365585e-01
5.72894961e-02 -1.59515142e-01 5.84596097e-01 1.00563025e+00
-4.03722316e-01 -8.71181428e-01 -6.62869871e-01 3.51321846e-01
5.06294854e-02 2.58107692e-01 -8.48950520e-02 6.79371715e-01
-1.65045440e-01 1.09030402e+00 -1.59177616e-01 -1.04857349e+00
6.36337399e-01 1.03267312e-01 2.58141875e-01 -9.80659276e-02
-2.86301702e-01 -5.68208218e-01 5.27627409e-01 -9.90008473e-01
1.04309700e-01 -5.26907384e-01 -1.15192878e+00 -2.25592002e-01
-4.85130429e-01 -1.19928181e-01 1.01469219e+00 9.68869388e-01
4.84654218e-01 7.57836819e-01 4.61906672e-01 -6.64079666e-01
-1.26841617e+00 -8.87556791e-01 -4.76538867e-01 -9.09347646e-03
2.75979936e-01 -1.18559146e+00 -5.40129721e-01 -4.73625034e-01]
|
[5.505715847015381, 7.337414741516113]
|
f5ef0979-ac3a-40f7-93c3-7896de25d4cd
|
benchmarking-common-uncertainty-estimation
|
2301.01054
| null |
https://arxiv.org/abs/2301.01054v2
|
https://arxiv.org/pdf/2301.01054v2.pdf
|
Benchmarking common uncertainty estimation methods with histopathological images under domain shift and label noise
|
In the past years, deep learning has seen an increase in usage in the domain of histopathological applications. However, while these approaches have shown great potential, in high-risk environments deep learning models need to be able to judge their uncertainty and be able to reject inputs when there is a significant chance of misclassification. In this work, we conduct a rigorous evaluation of the most commonly used uncertainty and robustness methods for the classification of Whole Slide Images, with a focus on the task of selective classification, where the model should reject the classification in situations in which it is uncertain. We conduct our experiments on tile-level under the aspects of domain shift and label noise, as well as on slide-level. In our experiments, we compare Deep Ensembles, Monte-Carlo Dropout, Stochastic Variational Inference, Test-Time Data Augmentation as well as ensembles of the latter approaches. We observe that ensembles of methods generally lead to better uncertainty estimates as well as an increased robustness towards domain shifts and label noise, while contrary to results from classical computer vision benchmarks no systematic gain of the other methods can be shown. Across methods, a rejection of the most uncertain samples reliably leads to a significant increase in classification accuracy on both in-distribution as well as out-of-distribution data. Furthermore, we conduct experiments comparing these methods under varying conditions of label noise. Lastly, we publish our code framework to facilitate further research on uncertainty estimation on histopathological data.
|
['Titus J. Brinker', 'Tabea-Clara Bucher', 'Alexander Kurz', 'Hendrik A. Mehrtens']
|
2023-01-03
| null | null | null | null |
['whole-slide-images']
|
['computer-vision']
|
[ 3.77319276e-01 1.71928346e-01 2.96545289e-02 -4.65413541e-01
-1.37418211e+00 -4.56576139e-01 6.44655347e-01 4.70274925e-01
-6.19587541e-01 1.15433979e+00 -3.74963023e-02 -3.53681713e-01
-3.49455178e-01 -7.44779825e-01 -7.00196445e-01 -1.21931458e+00
9.47341621e-02 8.29039335e-01 3.22224021e-01 3.26485336e-01
1.32454425e-01 4.52960223e-01 -1.42019105e+00 2.46137455e-01
7.39378691e-01 9.40100491e-01 -4.44213986e-01 5.80595195e-01
9.92379151e-03 3.09067398e-01 -6.61974430e-01 -5.37086666e-01
9.88602173e-04 -2.22704291e-01 -7.92165637e-01 -3.98786440e-02
3.30283165e-01 -1.28504172e-01 1.72817290e-01 1.23542011e+00
7.70770431e-01 -1.22801766e-01 1.09429348e+00 -1.00006247e+00
3.04753911e-02 6.42494678e-01 -5.79909444e-01 1.33513451e-01
-1.19973524e-02 3.82210165e-01 5.92124820e-01 -5.77100933e-01
6.79564178e-01 9.94769037e-01 8.12891006e-01 4.42613482e-01
-1.64919794e+00 -4.85185742e-01 9.74975247e-03 6.41973829e-03
-1.32978618e+00 -2.89006144e-01 3.40742499e-01 -5.01738429e-01
5.16909838e-01 3.37428421e-01 1.04880258e-01 1.42827415e+00
5.90369046e-01 4.98694986e-01 1.32832432e+00 -3.73443455e-01
7.38908410e-01 2.72134602e-01 1.41737297e-01 2.32323065e-01
4.41614538e-01 3.09360474e-01 -2.24054739e-01 -3.64486426e-01
4.01510119e-01 -1.30128533e-01 -1.08653478e-01 -2.93137491e-01
-9.94716763e-01 7.21897244e-01 1.76850393e-01 2.72597581e-01
-6.95513412e-02 4.18441236e-01 5.33815503e-01 1.45139843e-01
8.06460381e-01 1.84518874e-01 -2.46547684e-01 9.76380631e-02
-1.19519639e+00 2.34762609e-01 7.61841238e-01 4.40118939e-01
3.65042806e-01 -1.97653025e-01 -5.66108644e-01 6.93828940e-01
1.71274111e-01 2.80340582e-01 3.33824247e-01 -7.18746960e-01
-3.53313349e-02 2.52345979e-01 3.79137062e-02 -5.36525548e-01
-5.20034432e-01 -6.88330531e-01 -1.13417268e+00 4.67041880e-01
6.84933841e-01 -9.09637138e-02 -1.21616101e+00 1.79804826e+00
3.02073658e-01 5.24536908e-01 -1.64649952e-02 5.31638622e-01
4.38391238e-01 1.10989384e-01 1.54818326e-01 -2.76022434e-01
1.34465539e+00 -1.72499016e-01 -7.41771638e-01 2.18725935e-01
7.11469471e-01 -5.89921653e-01 6.19677901e-01 5.89480400e-01
-7.99921453e-01 -1.48429781e-01 -9.82406795e-01 2.31076390e-01
-3.51682007e-01 -1.19713664e-01 3.74590993e-01 9.10553217e-01
-1.06331563e+00 6.21717811e-01 -1.02315521e+00 -3.35336506e-01
6.50136173e-01 4.79426056e-01 -3.09632063e-01 -5.33088446e-02
-1.13193381e+00 8.52788925e-01 1.75285950e-01 2.57536620e-01
-9.96707499e-01 -7.99817622e-01 -7.26160586e-01 -4.96537201e-02
3.65079194e-01 -6.12078965e-01 1.07191634e+00 -5.92499852e-01
-1.10611272e+00 1.00876868e+00 -4.95544784e-02 -7.74665236e-01
1.10068226e+00 1.42340258e-01 -1.62159666e-01 -3.38735610e-01
-1.24338426e-01 4.85945225e-01 5.07946074e-01 -1.17599988e+00
-6.43153667e-01 -4.61698472e-01 -1.89511180e-01 -3.17565560e-01
1.78987488e-01 -3.06285024e-01 -5.54305017e-02 -5.04728317e-01
1.24865957e-02 -1.08544004e+00 -3.94800395e-01 -2.96781529e-02
-6.25632584e-01 -1.38238251e-01 3.35633606e-01 -2.65000910e-01
1.07130337e+00 -2.08020520e+00 -1.42979011e-01 4.16234165e-01
-4.17458154e-02 -7.87205920e-02 1.89398870e-01 1.55971915e-01
8.73275995e-02 4.01341498e-01 -6.23676360e-01 -4.49079037e-01
6.33269921e-03 3.15706283e-01 -2.07637995e-01 8.02907407e-01
3.84942412e-01 4.15899456e-01 -7.46331096e-01 -6.46548808e-01
3.44733112e-02 4.32110071e-01 -5.39736569e-01 -2.28410468e-01
-3.19088906e-01 4.17514801e-01 -2.13608295e-01 6.39116645e-01
8.10152769e-01 -3.01493883e-01 1.75131708e-01 9.80229229e-02
1.53939396e-01 -6.25787824e-02 -1.30278492e+00 1.31656849e+00
-4.78350908e-01 5.09231925e-01 -9.69695151e-02 -8.79249156e-01
4.68902171e-01 3.15961719e-01 1.37315139e-01 -1.22314855e-01
1.70008928e-01 4.59505200e-01 1.86662942e-01 -2.15081200e-01
2.45273307e-01 -5.01445830e-01 -8.51795003e-02 2.98823148e-01
-1.57547016e-02 -2.57103801e-01 3.07933688e-02 3.96885350e-02
1.19752550e+00 -7.52565935e-02 1.17218472e-01 -4.48445976e-01
5.38682759e-01 -7.16853589e-02 4.91876572e-01 9.93173182e-01
-2.55440563e-01 8.02754462e-01 9.20962214e-01 -6.85387105e-02
-8.73568594e-01 -1.30517900e+00 -8.49762440e-01 6.17096782e-01
-2.97723431e-02 2.59816676e-01 -7.33272970e-01 -8.92050207e-01
1.39645427e-01 8.86733651e-01 -9.55814064e-01 -2.95806918e-02
-5.36869839e-02 -1.41378903e+00 6.64999723e-01 3.70300770e-01
1.59925714e-01 -7.40118682e-01 -3.85095894e-01 -2.91504115e-02
1.27405018e-01 -8.26439857e-01 1.00367174e-01 4.02865857e-01
-8.74963224e-01 -1.05530918e+00 -7.46995926e-01 -3.11326534e-01
7.11606085e-01 -5.56016505e-01 9.89450037e-01 1.01506645e-02
-4.15198147e-01 1.86736479e-01 -1.01888843e-01 -4.50904220e-01
-6.25085294e-01 3.37490179e-02 -4.68177423e-02 -4.22168821e-02
2.82984257e-01 -3.32593590e-01 -6.97396100e-01 3.26750159e-01
-1.28546441e+00 -4.04616207e-01 6.59774780e-01 1.12993014e+00
8.25278044e-01 9.42759514e-02 6.07001126e-01 -1.22556770e+00
4.76872087e-01 -6.04981780e-01 -5.41051984e-01 3.33927840e-01
-7.16464996e-01 3.10960710e-01 2.10075572e-01 -1.66095033e-01
-9.47994649e-01 -1.15135960e-01 -3.78396183e-01 -5.52968681e-03
-3.49510163e-01 5.56817591e-01 1.97945923e-01 1.66137069e-01
8.23166609e-01 -5.62550798e-02 9.01806056e-02 -2.19133377e-01
-1.05258524e-01 7.32810318e-01 2.53574699e-01 -5.51920176e-01
3.47884238e-01 8.23768556e-01 3.52328956e-01 -5.64066291e-01
-7.29890049e-01 -2.15642005e-01 -3.21387529e-01 -1.45503744e-01
7.39148080e-01 -6.31934643e-01 -6.02185905e-01 6.19941115e-01
-8.34776580e-01 -3.42012107e-01 -2.52154857e-01 4.14689273e-01
-5.51817119e-01 3.87900025e-01 -5.88488460e-01 -8.05200458e-01
-5.85059784e-02 -1.71490431e+00 1.22790515e+00 5.75841703e-02
-3.35064292e-01 -1.15526831e+00 1.94497973e-01 7.06085637e-02
3.49248648e-01 4.88142073e-01 1.03646314e+00 -1.07106102e+00
-3.71876091e-01 -3.86868954e-01 2.64790989e-02 3.27432692e-01
-3.35479639e-02 3.22298467e-01 -1.23926163e+00 -3.57024461e-01
-1.67586654e-01 -3.21834117e-01 1.14321434e+00 7.40377486e-01
1.31452692e+00 2.36926973e-01 -5.51189542e-01 4.07192320e-01
1.57055116e+00 -1.68684721e-01 7.88082719e-01 2.35332340e-01
-8.90058056e-02 7.18896925e-01 5.57054996e-01 3.06976438e-01
-1.42570570e-01 6.25926197e-01 5.51699877e-01 1.31715611e-02
1.96794588e-02 9.42786038e-02 -2.73186490e-02 -2.68602706e-02
2.32011169e-01 -5.25117755e-01 -1.10940433e+00 6.59970224e-01
-1.64436305e+00 -7.84917235e-01 -1.75078362e-01 2.57556868e+00
9.58674729e-01 3.80926520e-01 -1.40089035e-01 2.10079744e-01
8.25211704e-01 -1.51330769e-01 -5.51226914e-01 -3.67195010e-01
-7.62265399e-02 1.52150944e-01 5.06066561e-01 5.85384011e-01
-1.00649118e+00 4.02929425e-01 6.62580776e+00 1.30208254e+00
-1.19959688e+00 2.87908595e-02 1.45858240e+00 -1.01254895e-01
-4.55433488e-01 -1.25578687e-01 -7.38349617e-01 6.87841773e-01
9.46193874e-01 1.65125027e-01 -2.19710782e-01 5.11053801e-01
1.37272738e-02 -7.08131373e-01 -1.35324931e+00 5.72048783e-01
-1.97369114e-01 -1.30866110e+00 -2.54506409e-01 2.14554101e-01
9.17159140e-01 -8.88685212e-02 3.17010820e-01 2.63400346e-01
4.93647933e-01 -1.28524923e+00 4.72419113e-01 6.46962285e-01
8.44158530e-01 -9.09626305e-01 1.46320260e+00 3.64888459e-01
-2.33706877e-01 3.70377213e-01 -2.06547037e-01 2.54704833e-01
-1.45543501e-01 1.26571476e+00 -9.54598963e-01 6.18068457e-01
5.59164822e-01 1.21522881e-01 -3.53710949e-01 1.11920214e+00
-1.25733435e-01 6.80820763e-01 -5.59496760e-01 -1.34840772e-01
1.55930787e-01 2.13408709e-01 3.76097202e-01 1.35855353e+00
3.56280059e-01 -1.77605122e-01 -2.66158789e-01 8.65027130e-01
1.28415093e-01 -1.68396309e-01 -2.03047454e-01 3.87996078e-01
3.73442739e-01 1.11680734e+00 -1.04774940e+00 -3.15295130e-01
5.10395803e-02 6.42489731e-01 9.16771740e-02 3.11110198e-01
-8.41065705e-01 -6.69221282e-02 4.75550115e-01 1.15213640e-01
9.91357863e-02 1.92728892e-01 -6.12981915e-01 -8.42085004e-01
-1.10930935e-01 -6.43234015e-01 5.38257778e-01 -4.29199845e-01
-1.49433732e+00 4.64442879e-01 4.49140333e-02 -1.00689292e+00
-3.93770397e-01 -7.35484064e-01 -4.95552421e-01 8.66246581e-01
-1.27057135e+00 -6.34973586e-01 8.23197514e-02 7.15250447e-02
2.06588879e-01 3.12916964e-01 7.73925185e-01 1.11722805e-01
-4.04180825e-01 6.94006622e-01 5.58287978e-01 -7.33519867e-02
8.61481369e-01 -1.35092676e+00 -1.37915760e-01 6.18323565e-01
-1.50413081e-01 4.56283212e-01 1.15289557e+00 -6.96350873e-01
-7.63674617e-01 -9.12070692e-01 5.70872307e-01 -5.97880602e-01
6.24259055e-01 -3.50120515e-01 -9.15761054e-01 4.13175762e-01
-1.61948144e-01 2.14496434e-01 8.10415685e-01 2.58880734e-01
-7.56179467e-02 -6.09312691e-02 -1.61793053e+00 5.58031380e-01
5.97308218e-01 -2.57379800e-01 -2.78319716e-01 3.70809376e-01
2.53707141e-01 -4.89944190e-01 -9.51712906e-01 7.83936381e-01
5.32875657e-01 -1.18686199e+00 6.25503778e-01 -3.91994715e-01
5.73792577e-01 -2.43102968e-01 -1.13464437e-01 -1.33640468e+00
-9.62607283e-03 -1.30798712e-01 3.80773038e-01 1.23857617e+00
7.61623204e-01 -6.39635265e-01 9.83060062e-01 5.84348381e-01
6.51081800e-02 -9.19896126e-01 -1.48393941e+00 -6.91108823e-01
4.19994086e-01 -7.43260205e-01 3.63355935e-01 6.27543092e-01
-1.25284940e-01 -3.23417187e-01 -9.83562693e-02 2.22153455e-01
7.31870413e-01 -1.35631055e-01 4.29149300e-01 -1.10220802e+00
-2.16412142e-01 -5.74714780e-01 -7.54370093e-01 -2.98867136e-01
1.48850650e-01 -8.18347275e-01 2.96850175e-01 -1.49880207e+00
4.75365400e-01 -4.90128368e-01 -5.62967837e-01 1.20197080e-01
-2.02083215e-01 4.02932197e-01 -3.99162203e-01 -2.26376522e-02
-4.49324876e-01 2.02633083e-01 8.49676490e-01 -1.90671638e-01
1.73252404e-01 3.38163882e-01 -5.02278686e-01 5.78944087e-01
6.35190189e-01 -6.95350647e-01 -1.67841837e-01 -8.52843225e-02
3.72349322e-01 5.10055535e-02 6.35523677e-01 -1.08562255e+00
1.78265885e-01 -8.23018793e-03 4.88903046e-01 -5.60898900e-01
8.94294381e-02 -7.48904109e-01 3.72262657e-01 7.31472313e-01
-5.52093923e-01 -4.45205599e-01 3.69043589e-01 8.43338728e-01
-1.69240341e-01 -3.17223787e-01 9.42767501e-01 1.92712471e-02
-1.96642295e-01 4.12973091e-02 -5.63546658e-01 -7.05789775e-02
1.00233400e+00 -2.78329581e-01 -3.20753098e-01 -2.21008599e-01
-9.94785964e-01 2.86955506e-01 4.79210407e-01 -8.23138431e-02
3.34314764e-01 -9.51095819e-01 -9.94631469e-01 5.49760535e-02
1.87702119e-01 1.46877423e-01 7.15832710e-01 1.13385475e+00
-4.75087434e-01 2.54673660e-01 1.87497810e-01 -1.12267923e+00
-1.06073821e+00 4.21948880e-01 5.21296144e-01 -6.48448050e-01
-6.78889453e-02 1.00571609e+00 2.27338523e-01 -4.32677031e-01
3.25751632e-01 -6.37515724e-01 1.80127528e-02 6.52573407e-02
3.15570831e-01 3.66281271e-01 4.28183466e-01 -8.61734599e-02
-4.23457026e-01 2.03151181e-01 -2.18348026e-01 -2.88141012e-01
1.12347066e+00 1.10729620e-01 -1.33025616e-01 7.13131785e-01
7.46948600e-01 -1.99984945e-02 -1.23246205e+00 3.74465026e-02
1.72831967e-01 -2.42716566e-01 9.78627652e-02 -1.18862653e+00
-7.79130518e-01 9.55889761e-01 8.38937700e-01 2.28487819e-01
8.96050453e-01 -3.84600721e-02 1.05181240e-01 -8.55913460e-02
4.31192696e-01 -9.84748006e-01 -4.28429961e-01 2.82055765e-01
6.12816870e-01 -1.35323262e+00 2.67502129e-01 -3.15730929e-01
-3.48942608e-01 8.46177876e-01 2.81156093e-01 -3.37123834e-02
8.52179468e-01 5.63141942e-01 1.89973623e-03 1.63622070e-02
-9.99669075e-01 1.21072531e-02 1.90383241e-01 5.81688881e-01
5.19427955e-01 4.75960746e-02 -3.87558520e-01 4.70459849e-01
1.17886402e-01 1.62214786e-01 6.20243490e-01 5.94558775e-01
-2.40851671e-01 -8.84308219e-01 -3.76792580e-01 6.67458236e-01
-8.52632940e-01 -3.26214321e-02 -1.22298449e-01 9.35481608e-01
1.30820781e-01 6.78965211e-01 2.72462308e-01 -3.57681736e-02
1.10536218e-01 1.15617350e-01 4.75647837e-01 -5.59651256e-01
-6.03227139e-01 2.08180193e-02 8.49239603e-02 -2.11665064e-01
-3.32977295e-01 -7.82223821e-01 -1.17910612e+00 -1.76227823e-01
-5.14444470e-01 4.97186668e-02 7.51433372e-01 9.95424807e-01
9.05927345e-02 6.56566739e-01 8.37445632e-02 -5.15345991e-01
-9.49918091e-01 -1.01237381e+00 -7.78434277e-01 2.18077153e-01
3.14382136e-01 -7.09104300e-01 -9.13909793e-01 -2.45442614e-01]
|
[14.801787376403809, -2.623075008392334]
|
a7b47a80-29ae-4f66-a5d9-ecfc245fda39
|
modelling-aspects-of-planar-multi-mode
|
1807.02077
| null |
http://arxiv.org/abs/1807.02077v2
|
http://arxiv.org/pdf/1807.02077v2.pdf
|
Modelling Aspects of Planar Multi-Mode Antennas for Direction-of-Arrival Estimation
|
Multi-mode antennas are an alternative to classical antenna arrays, and hence
a promising emerging sensor technology for a vast variety of applications in
the areas of array signal processing and digital communications. An unsolved
problem is to describe the radiation pattern of multi-mode antennas in closed
analytic form based on calibration measurements or on electromagnetic field
(EMF) simulation data. As a solution, we investigate two modeling methods: One
is based on the array interpolation technique (AIT), the other one on wavefield
modeling (WM). Both methods are able to accurately interpolate quantized EMF
data of a given multi-mode antenna, in our case a planar four-port antenna
developed for the 6-8.5 GHz range. Since the modeling methods inherently depend
on parameter sets, we investigate the influence of the parameter choice on the
accuracy of both models. Furthermore, we evaluate the impact of modeling errors
for coherent maximum-likelihood direction-of-arrival (DoA) estimation given
different model parameters. Numerical results are presented for a single
polarization component. Simulations reveal that the estimation bias introduced
by model errors is subject to the chosen model parameters. Finally, we provide
optimized sets of AIT and WM parameters for the multi-mode antenna under
investigation. With these parameter sets, EMF data samples can be reproduced in
interpolated form with high angular resolution.
|
[]
|
2019-05-29
| null | null | null | null |
['direction-of-arrival-estimation']
|
['audio']
|
[ 2.63438344e-01 -2.38701090e-01 4.90422696e-01 -1.45403385e-01
-6.94181025e-01 -3.87270242e-01 3.98000777e-01 1.72850206e-01
-2.61021614e-01 5.71612000e-01 -2.87396815e-02 -2.42731199e-01
-8.06402743e-01 -9.17854726e-01 -4.49312270e-01 -1.07251060e+00
-2.61742234e-01 3.04453343e-01 -1.61585242e-01 3.26688699e-02
-2.38573607e-02 6.79847717e-01 -1.46031272e+00 -2.38378823e-01
8.40299547e-01 1.21094966e+00 6.65731654e-02 5.75470626e-01
1.75098002e-01 2.05610111e-01 -5.48948050e-01 -3.58037166e-02
2.52300590e-01 -2.78118968e-01 5.52404970e-02 -1.04232393e-01
-1.31175548e-01 -5.59194833e-02 2.27251604e-01 9.57655668e-01
5.96511602e-01 -2.02586174e-01 8.71614814e-01 -8.89740884e-01
3.32149684e-01 1.72807068e-01 -1.53096393e-01 -2.24650413e-01
1.89014286e-01 -1.25407144e-01 2.22037837e-01 -6.91269279e-01
3.05535346e-01 6.81365371e-01 7.17740059e-01 -1.95005089e-01
-1.16823578e+00 -4.38517749e-01 -7.67862916e-01 7.69275948e-02
-1.66516685e+00 -3.66194725e-01 9.25127625e-01 -6.39407158e-01
2.22705036e-01 2.97106445e-01 4.95645970e-01 6.62356615e-01
4.50077504e-01 -2.72968888e-01 1.25905430e+00 -9.00885284e-01
6.81259274e-01 2.16042891e-01 1.47588260e-03 3.68282974e-01
6.28467977e-01 2.45356902e-01 -1.18620589e-01 -4.77498233e-01
6.69310927e-01 -5.29046297e-01 -4.21156853e-01 -5.70316494e-01
-1.09799349e+00 6.62876189e-01 1.45936802e-01 6.45554960e-01
-7.40439236e-01 1.42442271e-01 -3.23957890e-01 9.32914019e-02
2.50873357e-01 5.71152329e-01 -1.64628997e-01 2.08468303e-01
-9.21288788e-01 1.35305062e-01 8.18231642e-01 4.55684066e-01
9.02310252e-01 2.90525407e-01 8.50506350e-02 8.29236865e-01
7.71126807e-01 1.20226753e+00 1.04231708e-01 -5.30650675e-01
4.69279513e-02 5.80374002e-02 4.27138954e-01 -1.12373769e+00
-7.37731993e-01 -1.14288199e+00 -1.04180837e+00 5.75252250e-02
6.30164146e-01 -6.41499996e-01 -4.64097381e-01 1.51597238e+00
3.65532130e-01 6.33789003e-02 1.43104836e-01 7.07248032e-01
4.40483689e-01 7.95060515e-01 -2.12924570e-01 -5.56979001e-01
1.34414482e+00 -7.37512261e-02 -6.49314463e-01 -3.32930386e-01
5.11146307e-01 -8.49227130e-01 9.48522761e-02 6.85029805e-01
-8.76019239e-01 -3.36986095e-01 -1.17959142e+00 1.01769340e+00
6.74168952e-03 2.96100378e-01 1.77054912e-01 8.53073180e-01
-7.47771204e-01 1.70457140e-01 -5.23179412e-01 -1.75663441e-01
-1.06382035e-01 -1.07592093e-02 -1.06784612e-01 -4.70264368e-02
-1.01668727e+00 8.23058069e-01 3.80942971e-02 3.13880861e-01
-3.65999714e-02 -8.07404160e-01 -5.24282694e-01 7.53189549e-02
-3.06934923e-01 -7.92864501e-01 8.80362689e-01 -7.26151109e-01
-1.48487604e+00 3.85614336e-01 -1.81405023e-01 -4.54663098e-01
1.75330684e-01 2.22919762e-01 -1.04792762e+00 3.72137040e-01
7.95352459e-02 -1.17740571e-01 7.34721720e-01 -1.50166845e+00
-2.17300922e-01 -2.49100462e-01 -6.87863946e-01 -2.73989379e-01
-1.35366753e-01 -3.42724711e-01 2.13063300e-01 -4.01786953e-01
5.67621946e-01 -8.05325389e-01 -4.00099635e-01 -1.06106199e-01
-3.56148109e-02 4.36933428e-01 2.49896958e-01 -4.93976802e-01
1.05941689e+00 -1.95200181e+00 -5.86473532e-02 8.25349629e-01
-3.03417712e-01 3.48005705e-02 -4.68137525e-02 7.70016968e-01
-5.00004292e-02 -4.14502144e-01 -2.53552526e-01 1.39824361e-01
-2.64554828e-01 -1.44873470e-01 2.38759425e-02 6.63459659e-01
-1.10012069e-01 3.41513425e-01 -4.17345196e-01 9.10845622e-02
4.08700019e-01 5.77130139e-01 -5.51059067e-01 1.84789836e-01
-3.56900506e-02 8.35102737e-01 -6.75374091e-01 3.98549169e-01
9.94030833e-01 -2.01553673e-01 2.58835971e-01 -5.96896946e-01
-5.20500362e-01 -2.54748881e-01 -1.61739993e+00 1.20693874e+00
-9.10940349e-01 2.37168342e-01 4.74860609e-01 -1.11240017e+00
1.30612350e+00 4.10718888e-01 7.01851726e-01 -8.87857914e-01
3.29976141e-01 6.93340778e-01 2.61632442e-01 -5.51623285e-01
7.07123056e-02 -3.17869514e-01 5.32873869e-02 2.54035950e-01
-7.88881928e-02 -1.27917819e-03 3.31640653e-02 -4.17349666e-01
9.36236620e-01 -3.36110383e-01 5.34934938e-01 -7.11971581e-01
6.75833642e-01 -8.31688046e-02 3.19331288e-01 5.60657620e-01
3.92335683e-01 5.00569522e-01 2.96333842e-02 -2.14643791e-01
-8.02534819e-01 -8.10494721e-01 -6.57387316e-01 7.74248457e-03
2.42537722e-01 -2.24618521e-02 -5.04959106e-01 4.24305916e-01
-6.06109872e-02 8.79553139e-01 -9.96663421e-02 4.33938056e-02
-5.86468697e-01 -1.39156163e+00 2.08452448e-01 -1.06626883e-01
5.83147824e-01 -3.69645208e-01 -9.94348943e-01 4.68266547e-01
-2.33580485e-01 -1.25437450e+00 6.59929752e-01 2.13790029e-01
-6.43179953e-01 -8.70758414e-01 -6.68078899e-01 -1.84894174e-01
4.89625722e-01 -7.57109281e-03 1.00391078e+00 -2.44623229e-01
-6.73005432e-02 8.23169589e-01 -3.57049942e-01 -3.77793252e-01
-6.11188591e-01 -3.14632833e-01 2.18063127e-02 6.60831332e-01
-1.65040419e-01 -8.97660792e-01 -6.07871711e-01 6.18614554e-01
-7.75097787e-01 -5.35074845e-02 8.61464858e-01 4.78057414e-01
3.50724518e-01 1.38780907e-01 7.47198820e-01 -3.95870596e-01
2.76010007e-01 -7.84405768e-01 -1.03369808e+00 7.14662671e-02
-4.95361269e-01 1.35193318e-01 6.84197187e-01 -1.48045748e-01
-1.01027620e+00 -1.72257330e-02 -5.56422114e-01 1.50544569e-01
-3.55011344e-01 8.10448945e-01 -3.98346186e-01 -5.14879525e-01
5.45838773e-01 1.45698145e-01 -1.07776187e-01 -4.69717652e-01
-1.98752359e-01 5.46984375e-01 5.14612794e-01 -4.52555239e-01
9.10084426e-01 3.99385840e-01 7.75072277e-01 -1.48427331e+00
-2.65200168e-01 -4.19744581e-01 -1.45111799e-01 -5.09388924e-01
6.60207868e-01 -8.11037183e-01 -4.68294501e-01 5.73832095e-01
-1.12087536e+00 -2.54418198e-02 2.25178614e-01 9.85987008e-01
-2.79488504e-01 2.65303671e-01 1.77715287e-01 -9.96898115e-01
-2.10607395e-01 -9.05836403e-01 8.13263416e-01 9.10484120e-02
1.72613990e-02 -1.09271479e+00 1.86535046e-01 3.26897055e-02
8.77372146e-01 5.50310791e-01 9.92630422e-01 -2.21630648e-01
-7.14452326e-01 -3.70732546e-01 2.75734484e-01 4.81165014e-03
-4.81079072e-02 -5.43415606e-01 -7.65428424e-01 -3.08255553e-01
4.04805183e-01 4.48777229e-01 3.41206074e-01 9.40070868e-01
7.44691551e-01 -1.76834360e-01 -5.16809285e-01 7.14153290e-01
1.98681509e+00 1.87858149e-01 7.60612726e-01 2.31456175e-01
5.28612100e-02 2.52052665e-01 5.75993419e-01 6.84260726e-01
-5.60561866e-02 9.79023099e-01 5.85076630e-01 1.94617912e-01
1.51801497e-01 2.89204806e-01 -2.04656437e-01 5.78912735e-01
6.93183541e-02 -5.76318502e-01 -8.40259194e-01 3.19981396e-01
-1.37082696e+00 -6.68464303e-01 -6.26553535e-01 2.51623201e+00
8.89339745e-02 -8.92265067e-02 -2.03886271e-01 3.71523649e-01
4.28070217e-01 -5.30780517e-02 -1.26117924e-02 -7.33044073e-02
-3.06561083e-01 3.35655659e-01 7.96621442e-01 5.40160477e-01
-7.57190824e-01 -3.33252907e-01 5.03630638e+00 4.40844625e-01
-1.25663149e+00 -1.69398412e-02 7.32770562e-02 4.71062273e-01
-6.49391234e-01 -1.11278687e-02 -5.97719073e-01 5.79486310e-01
1.04607415e+00 5.45661673e-02 6.36723489e-02 2.06756204e-01
2.95382977e-01 -6.26485169e-01 -5.15042663e-01 1.22403538e+00
-1.73332274e-01 -1.20746279e+00 -1.29857570e-01 1.39975116e-01
4.95830148e-01 -4.02769059e-01 -8.77423733e-02 -3.37954044e-01
-6.63476229e-01 -6.66337550e-01 6.48519278e-01 8.57799113e-01
5.88896215e-01 -3.98409814e-01 7.63520062e-01 5.74841440e-01
-9.78012025e-01 -1.46650285e-01 -1.20364400e-02 4.59520705e-02
4.60298061e-01 1.41415048e+00 -7.41535604e-01 9.32198763e-01
3.76066476e-01 -3.53795663e-02 -2.01086104e-01 1.38165092e+00
1.75978109e-01 8.48893642e-01 -6.83018684e-01 -2.61423320e-01
-7.95588568e-02 -4.58348006e-01 8.82252872e-01 8.54691148e-01
1.12925839e+00 1.48305804e-01 -3.47313225e-01 8.68748069e-01
4.99185115e-01 1.20950520e-01 -3.76387864e-01 3.87943476e-01
6.69521213e-01 1.28697956e+00 -5.01168609e-01 2.90187240e-01
-3.55614364e-01 1.83712319e-01 -6.03166699e-01 6.10348165e-01
-4.96714324e-01 -1.26981407e-01 4.51576233e-01 5.85770249e-01
4.42549080e-01 -5.77193558e-01 -3.06849599e-01 -6.45820618e-01
1.56489193e-01 -3.78732413e-01 -3.40055185e-03 -5.93688607e-01
-7.73365796e-01 5.41512072e-01 2.28435114e-01 -1.46587813e+00
-6.24801278e-01 -6.45245671e-01 -4.93019313e-01 8.99386168e-01
-1.44646382e+00 -9.26922202e-01 -3.46015424e-01 9.23114344e-02
-3.75268698e-01 -5.67489490e-02 1.02112556e+00 6.17504537e-01
-2.94036344e-02 1.02945715e-01 6.26228213e-01 -2.45069504e-01
2.80929536e-01 -5.89376926e-01 -2.62031108e-01 8.02295327e-01
-2.58206338e-01 3.23707849e-01 1.19254720e+00 -2.76661992e-01
-1.53978240e+00 -1.02366042e+00 4.91831273e-01 2.06582084e-01
2.76578546e-01 -1.65607899e-01 -6.83014154e-01 1.77608982e-01
1.84003443e-01 2.35306293e-01 6.81698024e-01 -3.35093111e-01
2.35411584e-01 -3.54138225e-01 -1.24810922e+00 1.89738393e-01
3.17762136e-01 2.06765518e-01 -1.22033156e-01 4.85140570e-02
-2.02307433e-01 -3.79581377e-02 -1.16281569e+00 1.01832664e+00
5.33518612e-01 -1.05137670e+00 1.00628221e+00 2.98725843e-01
-1.58173680e-01 -5.71740270e-01 -5.48376799e-01 -1.58568370e+00
-3.45921218e-01 -5.18455505e-01 2.69352913e-01 1.03107941e+00
2.43532032e-01 -9.66954589e-01 3.52762461e-01 -1.12100951e-01
6.22491464e-02 -6.34633660e-01 -1.10395551e+00 -6.95343614e-01
-1.25629112e-01 -3.76422346e-01 4.44787413e-01 5.57146192e-01
-4.72427189e-01 2.76671678e-01 -2.07106605e-01 9.73162711e-01
1.01055324e+00 3.24243754e-01 6.59824729e-01 -1.47075033e+00
-5.54418266e-01 -2.89725559e-03 -3.50883812e-01 -9.35691953e-01
-2.86789477e-01 -5.69039047e-01 -7.41208270e-02 -1.44370127e+00
-6.08509302e-01 -8.89690757e-01 8.33968371e-02 -3.50978732e-01
4.55967903e-01 2.76884109e-01 -2.96751261e-01 -1.36059448e-01
8.73841643e-02 3.71471882e-01 6.50379062e-01 2.07269713e-01
4.80437018e-02 4.03379083e-01 -7.39475414e-02 6.27541721e-01
6.63514793e-01 -3.06684315e-01 -1.69922277e-01 -3.69912028e-01
5.28145492e-01 5.71806729e-01 7.71611929e-01 -1.62850916e+00
3.89447451e-01 1.59742191e-01 4.52971160e-01 -4.83444542e-01
3.27019960e-01 -1.06618059e+00 9.52741683e-01 5.03711462e-01
3.21671218e-01 -3.49673539e-01 3.44153970e-01 3.55823994e-01
-2.05251113e-01 -4.37711179e-01 1.06650853e+00 1.70831785e-01
-5.86127818e-01 -1.69019952e-01 -8.18774939e-01 -2.47747883e-01
8.56908083e-01 -6.53887168e-02 -1.06320582e-01 -5.29350877e-01
-4.43850160e-01 -2.46008098e-01 8.65564793e-02 -1.49780303e-01
2.43453532e-01 -1.36468756e+00 -7.77374148e-01 5.18935204e-01
1.39370516e-01 -3.77044797e-01 3.25513810e-01 9.42152441e-01
-5.46392024e-01 5.72549760e-01 -1.49664536e-01 -8.78294528e-01
-8.51338863e-01 5.06848022e-02 7.76026487e-01 7.96488374e-02
8.12742785e-02 1.71116278e-01 -1.76700920e-01 -3.32068980e-01
-2.18138933e-01 -2.51869559e-01 -1.68784648e-01 -1.61731258e-01
2.62812793e-01 3.29720765e-01 4.12650049e-01 -6.31803751e-01
-4.62034941e-01 1.14092731e+00 9.29472506e-01 -2.18056664e-01
1.12799537e+00 -2.02093020e-01 -1.04038805e-01 2.80333579e-01
1.15575850e+00 4.66830879e-01 -5.27988374e-01 -1.57485884e-02
-8.39813650e-02 -2.43837997e-01 2.28200570e-01 -8.27092528e-01
-6.42173886e-01 4.84433532e-01 9.19267297e-01 4.41758335e-01
1.26803243e+00 -8.06130320e-02 2.88324386e-01 2.68776655e-01
7.85783648e-01 -6.51766956e-01 -4.02313381e-01 2.80602686e-02
7.55635321e-01 -6.84812129e-01 5.73162548e-03 -6.11464500e-01
3.20459783e-01 1.19224167e+00 2.10077297e-02 1.73916854e-02
1.06228411e+00 5.79372108e-01 1.92186892e-01 1.89269725e-02
-2.48196006e-01 1.01270229e-01 1.31409973e-01 4.48157370e-01
3.55752707e-01 7.64651969e-02 -5.58476567e-01 6.78656161e-01
-4.15112078e-02 6.81677982e-02 5.31401157e-01 5.65722525e-01
-6.05613828e-01 -1.10373890e+00 -9.30003643e-01 4.36846018e-01
-2.83160329e-01 2.43034244e-01 5.57198584e-01 5.69802642e-01
9.54537913e-02 1.08092034e+00 1.26097158e-01 -4.66477908e-02
5.44359088e-01 -6.32486567e-02 6.04152858e-01 3.03190369e-02
-8.99671167e-02 1.27919018e-01 1.42607182e-01 -1.38273656e-01
-4.74777848e-01 -6.80483520e-01 -9.13017511e-01 1.70955762e-01
-3.40059549e-01 4.75207984e-01 1.00830078e+00 8.11560035e-01
3.82036984e-01 4.48157996e-01 8.17056537e-01 -5.85526824e-01
-6.12962604e-01 -8.07810068e-01 -9.23118532e-01 -1.07203253e-01
2.18059331e-01 -8.18156004e-01 -5.03538311e-01 -4.39625740e-01]
|
[6.497955799102783, 1.3432073593139648]
|
2f41baf5-8a98-4431-ab1b-ff71f9024265
|
medsegdiff-medical-image-segmentation-with
|
2211.00611
| null |
https://arxiv.org/abs/2211.00611v5
|
https://arxiv.org/pdf/2211.00611v5.pdf
|
MedSegDiff: Medical Image Segmentation with Diffusion Probabilistic Model
|
Diffusion probabilistic model (DPM) recently becomes one of the hottest topic in computer vision. Its image generation application such as Imagen, Latent Diffusion Models and Stable Diffusion have shown impressive generation capabilities, which aroused extensive discussion in the community. Many recent studies also found it is useful in many other vision tasks, like image deblurring, super-resolution and anomaly detection. Inspired by the success of DPM, we propose the first DPM based model toward general medical image segmentation tasks, which we named MedSegDiff. In order to enhance the step-wise regional attention in DPM for the medical image segmentation, we propose dynamic conditional encoding, which establishes the state-adaptive conditions for each sampling step. We further propose Feature Frequency Parser (FF-Parser), to eliminate the negative effect of high-frequency noise component in this process. We verify MedSegDiff on three medical segmentation tasks with different image modalities, which are optic cup segmentation over fundus images, brain tumor segmentation over MRI images and thyroid nodule segmentation over ultrasound images. The experimental results show that MedSegDiff outperforms state-of-the-art (SOTA) methods with considerable performance gap, indicating the generalization and effectiveness of the proposed model. Our code is released at https://github.com/WuJunde/MedSegDiff.
|
['Huiying Liu', 'Haoyi Xiong', 'Yu Zhang', 'Huihui Fang', 'Rao Fu', 'Yanwu Xu', 'Yehui Yang', 'Junde Wu']
|
2022-11-01
| null | null | null | null |
['tumor-segmentation', 'optic-cup-segmentation', 'brain-tumor-segmentation']
|
['computer-vision', 'medical', 'medical']
|
[ 3.30077261e-01 3.05036247e-01 -7.45197833e-02 -2.19968930e-01
-7.19295681e-01 -4.41155247e-02 7.62321055e-01 -1.55922115e-01
-4.33543861e-01 2.03913093e-01 6.23450816e-01 -1.28927812e-01
-1.51519716e-01 -4.56257641e-01 -2.37295389e-01 -1.06189466e+00
5.02646118e-02 2.16212809e-01 7.59757996e-01 2.02340424e-01
4.35204655e-01 1.12529904e-01 -1.19568241e+00 3.46800357e-01
1.36220717e+00 6.47784650e-01 5.46900332e-01 5.95607519e-01
-3.78346533e-01 7.82448888e-01 -4.54788953e-01 -3.50319356e-01
6.12752065e-02 -5.31316936e-01 -9.31634009e-01 4.40737545e-01
5.64423613e-02 -2.86291122e-01 -4.57428992e-01 1.61088371e+00
7.14294970e-01 3.22129726e-02 7.58577883e-01 -9.79526758e-01
-1.22607529e+00 8.28738153e-01 -1.10407603e+00 6.93878114e-01
-1.20585531e-01 2.87387967e-01 4.11621660e-01 -6.98947012e-01
7.09585369e-01 1.47551274e+00 4.24482048e-01 6.86030269e-01
-6.90330446e-01 -3.76552850e-01 1.29864901e-01 3.27776760e-01
-1.05941749e+00 -2.17732936e-01 5.80210090e-01 -4.97267872e-01
7.32097805e-01 1.08988903e-01 3.77519608e-01 9.41086233e-01
7.10011244e-01 1.29242194e+00 1.28879726e+00 -4.20668811e-01
4.20280695e-02 -1.28970057e-01 4.53602701e-01 8.44921768e-01
1.18925758e-01 -1.28215417e-01 -3.09181362e-01 -7.77316839e-02
8.23505521e-01 -7.35540986e-02 -3.78614515e-01 6.14417195e-02
-1.25977004e+00 9.09050286e-01 4.11823124e-01 5.75638533e-01
-6.08790338e-01 8.05950607e-04 2.06659719e-01 -1.46384746e-01
5.85532546e-01 -9.18629095e-02 -9.71836317e-03 -3.99313681e-02
-8.74962270e-01 -1.28647685e-02 2.73562908e-01 8.44733536e-01
7.72863030e-02 -3.72403674e-02 -6.48747146e-01 1.07461011e+00
6.27677977e-01 4.73442614e-01 1.37247694e+00 -8.93059492e-01
1.41398311e-01 4.53280121e-01 -2.45854020e-01 -8.95963788e-01
-4.34298813e-01 -4.52125192e-01 -1.11685514e+00 -2.02296395e-02
1.17992640e-01 -1.68900296e-01 -1.56241596e+00 1.42909908e+00
4.70444173e-01 4.11357850e-01 2.23234631e-02 9.56589520e-01
9.36711490e-01 8.07714581e-01 1.55839443e-01 -4.28527147e-01
1.52235222e+00 -1.34515250e+00 -8.91309738e-01 -2.56740510e-01
5.67499697e-01 -9.23641860e-01 7.31621325e-01 4.30219144e-01
-1.05022323e+00 -4.59688872e-01 -7.57555008e-01 -3.57694328e-02
-3.45741771e-02 5.30398563e-02 7.52799988e-01 7.42826939e-01
-1.06956136e+00 3.34434956e-01 -1.25633323e+00 -4.37065125e-01
7.22439110e-01 1.90809101e-01 -8.06276277e-02 -3.17972839e-01
-1.12072706e+00 7.64938772e-01 2.95108259e-01 2.49061182e-01
-8.72076333e-01 -4.69796300e-01 -5.86624801e-01 -3.21995765e-01
1.85733408e-01 -8.08323503e-01 1.12783611e+00 -7.96494901e-01
-1.40002310e+00 8.85056317e-01 -1.85177103e-01 -6.78964734e-01
5.05739927e-01 -9.87758562e-02 -4.20579582e-01 2.98698157e-01
2.90994376e-01 1.05310559e+00 1.06169963e+00 -7.47301996e-01
-6.93839610e-01 -5.04558146e-01 -3.31195235e-01 3.29120934e-01
-7.49122724e-02 1.22478738e-01 -8.54361653e-01 -7.27550626e-01
4.53712910e-01 -9.15905416e-01 -5.31358421e-01 -3.27565610e-01
-6.70598447e-01 -3.22821766e-01 6.66070282e-01 -8.24305952e-01
1.23484039e+00 -2.25776792e+00 1.91129148e-01 -3.35762538e-02
3.90877515e-01 2.38815993e-01 -1.46232069e-01 -2.05185652e-01
3.85014489e-02 1.25359997e-01 -6.82743311e-01 -3.02040637e-01
-2.10573331e-01 2.21449837e-01 1.01240933e-01 4.89703774e-01
7.22417384e-02 1.07214379e+00 -7.65061200e-01 -7.89171219e-01
1.20675758e-01 4.47621703e-01 -4.87174809e-01 -1.54083595e-01
-9.48084295e-02 7.17430055e-01 -6.54560030e-01 7.76582479e-01
8.49074781e-01 -3.98457438e-01 -2.01587141e-01 -1.69010550e-01
-6.02693409e-02 -2.95188904e-01 -1.07642114e+00 1.96996343e+00
4.24264967e-02 4.98367548e-01 -8.87256488e-02 -7.95438170e-01
5.75586855e-01 1.28321052e-01 6.44112527e-01 -7.07012951e-01
1.93250641e-01 1.74029514e-01 2.22489193e-01 -9.85004783e-01
4.72874731e-01 2.48047188e-02 1.84746340e-01 1.63677454e-01
1.84055194e-01 1.89452805e-02 2.41180673e-01 4.35047776e-01
9.58466470e-01 -5.49973026e-02 1.27015710e-01 -4.33180243e-01
6.12806618e-01 1.32477200e-02 6.22822702e-01 8.84666681e-01
-5.93537748e-01 8.62198412e-01 4.57165211e-01 8.26195180e-02
-6.85931146e-01 -8.35604429e-01 -3.71199936e-01 6.68511868e-01
3.01580966e-01 -7.45933950e-02 -1.13301015e+00 -6.86993301e-01
-3.49255562e-01 5.62935829e-01 -6.97347701e-01 -1.19381912e-01
-3.85253638e-01 -1.58571374e+00 4.97059792e-01 2.39069164e-01
9.73959029e-01 -1.15533912e+00 -2.61286795e-01 2.23760292e-01
-2.63870448e-01 -9.09397185e-01 -7.02613652e-01 -4.05730307e-01
-1.00225103e+00 -8.37786198e-01 -1.43126857e+00 -9.14697349e-01
7.64912784e-01 2.94198453e-01 6.79108381e-01 -1.50610209e-01
-5.85188031e-01 4.30900156e-01 -3.49043787e-01 -2.52320498e-01
-5.19285202e-01 4.77550961e-02 -2.24663079e-01 1.84554353e-01
4.70245302e-01 -4.25263584e-01 -1.05536544e+00 2.28357181e-01
-1.14818418e+00 1.35160208e-01 1.08156061e+00 8.14199507e-01
7.89866388e-01 1.21598534e-01 3.61407518e-01 -1.25793052e+00
9.34280157e-01 -6.15536928e-01 -3.90200853e-01 1.17389202e-01
-7.53577173e-01 -1.29582673e-01 -1.56914830e-01 -5.20135105e-01
-1.31779802e+00 -1.53056726e-01 -4.19199765e-01 -2.41397157e-01
-2.22565860e-01 5.79476357e-01 6.13056831e-02 1.19520143e-01
6.46899760e-01 5.65382659e-01 1.10538803e-01 -4.15032268e-01
5.07166266e-01 8.08202267e-01 5.37427068e-01 -1.15231641e-01
2.20122069e-01 5.86969137e-01 -2.58557141e-01 -8.81036341e-01
-6.09208941e-01 -4.86441880e-01 -3.73255879e-01 -8.00901353e-02
1.25305390e+00 -5.87442458e-01 -2.02549785e-01 9.76725042e-01
-1.04271352e+00 -7.72203803e-02 -1.40207812e-01 7.04607308e-01
-3.57195765e-01 7.17777967e-01 -8.79839838e-01 -5.80557704e-01
-4.80702817e-01 -1.70918572e+00 8.39512587e-01 5.84642708e-01
1.57896087e-01 -9.79820192e-01 -9.62678120e-02 4.08610612e-01
5.37098408e-01 2.85577239e-03 8.37314844e-01 -4.49782878e-01
-7.06050277e-01 9.64712948e-02 -2.75134951e-01 3.66374165e-01
1.35148838e-01 -2.56304204e-01 -8.42120767e-01 -1.54785300e-02
4.04487342e-01 1.50017217e-01 1.20734441e+00 1.15466201e+00
1.08576393e+00 3.79281584e-03 -6.38425827e-01 6.44044399e-01
1.12453520e+00 2.31200725e-01 9.20103371e-01 1.72836199e-01
7.31334746e-01 4.74877477e-01 4.36360240e-01 3.56910238e-03
3.37071031e-01 2.99196959e-01 2.25801408e-01 -1.68702126e-01
-7.00536430e-01 8.34378526e-02 3.11914116e-01 1.20944309e+00
1.02292821e-01 -3.29968303e-01 -9.32446122e-01 7.14440107e-01
-1.75792444e+00 -6.27756059e-01 -4.79856372e-01 1.73308396e+00
7.62408257e-01 1.97891027e-01 -2.58516282e-01 -3.13466728e-01
8.00617456e-01 9.39568505e-02 -6.84583902e-01 -8.36296007e-03
-1.05773330e-01 -1.25090837e-01 5.85686028e-01 5.89381039e-01
-1.12482464e+00 9.78321671e-01 5.73357487e+00 1.26329958e+00
-9.36515629e-01 5.69590569e-01 9.24363554e-01 3.09725642e-01
-2.58761108e-01 -2.27932170e-01 -8.18250895e-01 6.82306528e-01
6.10538840e-01 -3.12459245e-02 -9.40403789e-02 5.59088290e-01
2.45512724e-01 -4.70382959e-01 -4.59634542e-01 1.01868665e+00
1.58462331e-01 -1.15439820e+00 2.30334923e-01 4.94685024e-02
8.71881902e-01 4.25242066e-01 4.06068742e-01 1.49763897e-01
2.63723671e-01 -7.22542346e-01 2.00287715e-01 7.47509897e-01
3.88044357e-01 -3.59099597e-01 8.46992373e-01 3.17865849e-01
-6.95007801e-01 5.69580086e-02 -5.37207425e-01 6.89556181e-01
2.72212327e-01 9.49546099e-01 -5.79157710e-01 4.72465008e-01
5.94934165e-01 7.65107572e-01 -6.69126034e-01 1.45242763e+00
-2.56197095e-01 8.20293486e-01 -1.77748322e-01 3.11336547e-01
3.64256978e-01 -2.69328326e-01 8.84912670e-01 1.20118332e+00
3.92659634e-01 1.76263064e-01 -9.20798108e-02 8.27336192e-01
9.97483209e-02 1.62942484e-01 -2.54641622e-01 1.55602880e-02
1.37737721e-01 1.23701644e+00 -1.08571410e+00 -3.57770026e-01
-3.72806698e-01 1.21979725e+00 -2.99551636e-01 3.92200947e-01
-8.78091335e-01 -1.46246597e-01 8.25585723e-02 -1.69152260e-01
2.36411363e-01 -6.28862204e-03 -2.01837540e-01 -1.11340845e+00
-2.62629688e-01 -6.65817022e-01 5.88573098e-01 -8.29719007e-01
-1.29725730e+00 7.97801435e-01 9.54902545e-03 -1.06517172e+00
2.51935329e-02 -3.17104697e-01 -3.97032827e-01 8.20394576e-01
-1.61159515e+00 -1.10314465e+00 -1.97939143e-01 5.38501859e-01
9.79784250e-01 -1.46570697e-01 3.54018360e-01 3.31156731e-01
-6.38488293e-01 3.34140807e-01 1.59831122e-01 4.25887033e-02
6.45523250e-01 -1.09511626e+00 3.77155244e-01 1.14359975e+00
-1.03430785e-01 5.67981601e-01 4.70208228e-01 -8.82717907e-01
-1.08592963e+00 -1.08683014e+00 4.89484876e-01 -2.77540833e-01
4.93902802e-01 2.01511383e-01 -1.02028811e+00 4.26187515e-01
5.22991955e-01 -1.21503517e-01 3.36389244e-01 -4.97233063e-01
2.95452654e-01 2.03269988e-01 -1.27762783e+00 5.25972843e-01
8.66786599e-01 1.58303961e-01 -6.28592253e-01 4.80951101e-01
8.19579482e-01 -6.29719436e-01 -6.72661006e-01 2.92734116e-01
1.28146723e-01 -9.73421216e-01 7.58468747e-01 -6.61868826e-02
5.01130223e-01 -2.41701514e-01 1.37441739e-01 -1.16168189e+00
-5.16366601e-01 -6.23124659e-01 -3.01558990e-02 1.30087614e+00
3.52294654e-01 -8.49024057e-01 6.74824834e-01 6.04372323e-01
-4.01045233e-01 -7.75697052e-01 -8.37478995e-01 -2.85686255e-01
1.42359853e-01 -3.33841890e-01 2.26856738e-01 9.40411448e-01
-4.11396235e-01 1.06217243e-01 -1.23427190e-01 3.24547112e-01
6.80853248e-01 -1.17524348e-01 1.65319547e-01 -8.32622766e-01
-4.02470231e-01 -7.27613866e-01 -3.69073540e-01 -1.40412486e+00
-3.84457231e-01 -1.03221524e+00 -1.43490275e-02 -1.88243890e+00
3.98548990e-01 -2.73075908e-01 -3.19996685e-01 2.49202505e-01
-4.84543294e-01 2.90566862e-01 -1.12090468e-01 5.18471658e-01
-5.71279824e-01 4.50487942e-01 1.71371305e+00 -3.55580539e-01
-2.17876807e-01 1.62014827e-01 -7.83761621e-01 8.35685790e-01
8.10885072e-01 -6.08162224e-01 -5.45067668e-01 -6.64818764e-01
-3.14715981e-01 -2.27483492e-02 8.26874375e-02 -8.30661058e-01
4.16704297e-01 1.00380130e-01 2.93975025e-01 -5.88462234e-01
5.42833880e-02 -4.12075043e-01 -2.48966619e-01 6.52477562e-01
-2.78800905e-01 -9.26937088e-02 1.46291271e-01 6.37703717e-01
-3.32070857e-01 -5.32043993e-01 8.72272134e-01 -3.16608906e-01
-1.06219733e+00 5.02047777e-01 -5.89799166e-01 4.79637347e-02
9.22870636e-01 -1.52242169e-01 -3.20121944e-01 -7.75809363e-02
-9.60706472e-01 3.04811269e-01 -7.60911182e-02 4.38198477e-01
7.24782169e-01 -1.00136805e+00 -9.85023320e-01 1.98196009e-01
-1.91200137e-01 1.11430287e-01 7.92514265e-01 1.44629312e+00
-4.90668178e-01 2.75101900e-01 8.50856379e-02 -9.34616625e-01
-1.10202229e+00 5.34804106e-01 3.75280917e-01 -2.22985506e-01
-9.65662777e-01 1.23156202e+00 5.21168292e-01 3.58986482e-02
6.97384849e-02 -5.98388016e-01 -4.92777914e-01 -1.15388714e-01
6.09537721e-01 2.65600473e-01 -9.21359882e-02 -5.71237087e-01
-1.78123027e-01 5.08390963e-01 -5.81069291e-01 -2.62190074e-01
9.71935689e-01 -4.43010062e-01 -3.43237638e-01 8.42963904e-02
8.32007110e-01 -6.49218932e-02 -1.20602143e+00 -3.34035397e-01
5.25643723e-03 -3.45979363e-01 4.92154390e-01 -8.70716691e-01
-1.35192919e+00 8.39783192e-01 1.14327633e+00 2.79563576e-01
1.28233302e+00 1.64468378e-01 9.72156465e-01 -3.60440850e-01
8.43832865e-02 -7.91942418e-01 -1.14454202e-01 2.17894420e-01
6.88740909e-01 -1.23748219e+00 -9.30570960e-02 -4.89921182e-01
-9.20071185e-01 7.63740838e-01 3.64363879e-01 -5.66893108e-02
7.53727496e-01 1.76415280e-01 1.59127757e-01 -1.83875442e-01
-4.70527530e-01 -1.88261718e-01 3.87441158e-01 7.11602509e-01
2.95935214e-01 -3.57811525e-02 -6.90753520e-01 5.36135316e-01
1.65205449e-01 1.63499340e-01 5.33169210e-01 7.99592674e-01
-5.64818978e-01 -1.07725692e+00 -3.25594872e-01 6.53046191e-01
-7.69301474e-01 -2.60713547e-01 8.56143385e-02 4.82205093e-01
2.64291614e-01 7.27520525e-01 -1.26963660e-01 5.44521287e-02
6.13763183e-02 -3.53938416e-02 4.98031855e-01 -4.49272424e-01
-2.19819933e-01 4.17091429e-01 -3.12193364e-01 -3.94538552e-01
-6.70967221e-01 -7.13916361e-01 -1.30046177e+00 2.24548429e-01
-3.48263681e-01 1.56005681e-01 6.17651701e-01 8.39233398e-01
4.56028789e-01 8.18029523e-01 2.22093537e-01 -4.93160099e-01
-3.77898842e-01 -1.39656532e+00 -5.47932267e-01 3.65473896e-01
1.63754418e-01 -5.94910443e-01 -7.89270774e-02 2.74430335e-01]
|
[14.549259185791016, -2.2755320072174072]
|
3f9539bc-9b04-4760-927d-6b45182f7ee3
|
clip-vip-adapting-pre-trained-image-text
|
2209.0643
| null |
https://arxiv.org/abs/2209.06430v4
|
https://arxiv.org/pdf/2209.06430v4.pdf
|
CLIP-ViP: Adapting Pre-trained Image-Text Model to Video-Language Representation Alignment
|
The pre-trained image-text models, like CLIP, have demonstrated the strong power of vision-language representation learned from a large scale of web-collected image-text data. In light of the well-learned visual features, some existing works transfer image representation to video domain and achieve good results. However, how to utilize image-language pre-trained model (e.g., CLIP) for video-language pre-training (post-pretraining) is still under explored. In this paper, we investigate two questions: 1) what are the factors hindering post-pretraining CLIP to further improve the performance on video-language tasks? and 2) how to mitigate the impact of these factors? Through a series of comparative experiments and analyses, we find that the data scale and domain gap between language sources have great impacts. Motivated by these, we propose a Omnisource Cross-modal Learning method equipped with a Video Proxy mechanism on the basis of CLIP, namely CLIP-ViP. Extensive results show that our approach improves the performance of CLIP on video-text retrieval by a large margin. Our model also achieves SOTA results on a variety of datasets, including MSR-VTT, DiDeMo, LSMDC, and ActivityNet. We will release our code and pre-trained CLIP-ViP models at https://github.com/microsoft/XPretrain/tree/main/CLIP-ViP.
|
['Jiebo Luo', 'Houqiang Li', 'Ruihua Song', 'Jianlong Fu', 'Bei Liu', 'Yuchong Sun', 'Hongwei Xue']
|
2022-09-14
| null | null | null | null |
['video-text-retrieval']
|
['computer-vision']
|
[-7.50812739e-02 -7.41729438e-01 -5.53250968e-01 -2.37749889e-01
-8.55813622e-01 -4.67037708e-01 7.42663383e-01 -4.40467149e-01
-6.08803034e-01 3.90690535e-01 4.61116344e-01 -1.51076302e-01
1.87419131e-01 -4.86248523e-01 -9.90875840e-01 -4.58726078e-01
2.30568215e-01 1.66534372e-02 3.02007645e-01 -2.07610324e-01
2.64423490e-01 -7.55793378e-02 -1.51419103e+00 7.43890584e-01
6.06763363e-01 9.33730125e-01 5.00973463e-01 4.84604180e-01
-3.69103253e-01 1.08783650e+00 -3.99900049e-01 -3.66226196e-01
3.83932054e-01 -3.63674521e-01 -5.80582857e-01 1.10086299e-01
5.03464818e-01 -6.65378809e-01 -7.84296393e-01 1.01772046e+00
5.25923312e-01 -8.92730728e-02 5.84059477e-01 -1.39481604e+00
-1.10423493e+00 4.90594983e-01 -8.27479541e-01 3.13328177e-01
4.19401169e-01 2.69403577e-01 7.88082004e-01 -1.22070122e+00
7.46405363e-01 1.27659476e+00 3.27953637e-01 4.01510715e-01
-6.64089680e-01 -9.18160379e-01 2.03028262e-01 4.16862190e-01
-1.68576252e+00 -3.73589396e-01 7.52165496e-01 -5.82087100e-01
6.83455288e-01 1.06951669e-02 4.64914918e-01 1.64301455e+00
1.27201572e-01 1.08195245e+00 1.07482803e+00 -4.33901042e-01
-2.17965573e-01 4.26845640e-01 -6.48632273e-02 6.82950675e-01
-5.39283417e-02 -1.06173959e-02 -8.20440769e-01 1.20145939e-01
8.13370466e-01 1.75175041e-01 -4.05176759e-01 -3.01165551e-01
-1.28675497e+00 7.65483618e-01 3.37604433e-01 3.71389091e-01
-1.80894941e-01 2.18346655e-01 7.15359330e-01 3.81770790e-01
4.02567923e-01 -1.57031596e-01 -4.19643581e-01 -2.33983412e-01
-1.09881794e+00 -6.67468905e-02 5.23701906e-01 1.19742882e+00
7.80757487e-01 8.29860047e-02 -1.91777468e-01 9.54231739e-01
3.93792868e-01 8.92681897e-01 7.43985236e-01 -7.16536880e-01
7.29391158e-01 4.73162591e-01 -3.23643267e-01 -1.03506470e+00
1.28237322e-01 -7.40601569e-02 -6.46647990e-01 -2.25868538e-01
1.52618527e-01 -1.34724498e-01 -9.04610753e-01 1.43876350e+00
-4.41904068e-02 3.14538479e-01 8.77056792e-02 1.12908745e+00
1.17195725e+00 9.68508780e-01 2.30017990e-01 -1.21588610e-01
1.35487604e+00 -1.34010673e+00 -6.11739993e-01 -2.60991633e-01
5.82461894e-01 -9.53087211e-01 1.43751431e+00 2.74621755e-01
-8.47655118e-01 -6.45128131e-01 -8.27401936e-01 -1.27719611e-01
-4.78738755e-01 4.58258808e-01 3.77420753e-01 2.64255822e-01
-9.96420741e-01 8.32717642e-02 -5.46432316e-01 -7.81964302e-01
2.82078773e-01 -1.64466389e-02 -5.07457018e-01 -3.21405202e-01
-1.24098754e+00 5.33832729e-01 4.97656524e-01 -1.59240931e-01
-1.17796290e+00 -4.88961250e-01 -5.54999828e-01 -1.58288985e-01
6.61157548e-01 -5.20381391e-01 9.24265862e-01 -1.50801349e+00
-1.18542397e+00 1.02699113e+00 -1.24414742e-01 -4.09788668e-01
5.75964510e-01 -4.79586542e-01 -5.09281158e-01 5.92236757e-01
1.42126247e-01 9.70732391e-01 1.12834859e+00 -1.30983007e+00
-5.19921839e-01 -2.40708232e-01 2.43970647e-01 3.77589166e-01
-7.91173637e-01 3.48746538e-01 -1.36271310e+00 -6.95350170e-01
-4.32615161e-01 -9.53973651e-01 8.33868384e-02 1.27053618e-01
-5.59849106e-02 -1.83190539e-01 9.91619706e-01 -7.47155845e-01
1.29721820e+00 -2.41176176e+00 -3.81744877e-02 -2.04379812e-01
3.55278794e-03 4.17951107e-01 -5.02685547e-01 6.73869491e-01
2.14873180e-02 2.38097206e-01 2.04334468e-01 -1.83789693e-02
-2.28196234e-01 1.60580173e-01 -4.85147655e-01 3.24828744e-01
3.47559042e-02 1.00690222e+00 -6.86412454e-01 -9.92824733e-01
4.16176260e-01 5.46119452e-01 -5.53788900e-01 3.47087085e-01
-1.88868046e-01 1.84841186e-01 -6.56690657e-01 8.28791320e-01
5.97266138e-01 -3.13257277e-01 -6.71944618e-02 -5.09519160e-01
-1.03446648e-01 -2.28304699e-01 -9.75499988e-01 1.99301183e+00
-2.60304302e-01 8.51686239e-01 5.58393486e-02 -1.03927457e+00
7.01444566e-01 2.26097524e-01 6.15106881e-01 -1.03145123e+00
2.63442218e-01 1.30360778e-02 -4.31138247e-01 -1.02048993e+00
6.03241682e-01 2.99932063e-01 8.65926966e-02 5.78860082e-02
2.04157934e-01 2.88771629e-01 2.21493527e-01 6.06351554e-01
8.94013166e-01 3.15593600e-01 9.04840231e-02 -5.63036986e-02
6.86783552e-01 1.08469330e-01 3.39207768e-01 7.47392058e-01
-4.80149031e-01 8.87738407e-01 2.80255944e-01 -8.58176053e-02
-8.91511798e-01 -8.71914387e-01 -2.75663193e-02 1.46370542e+00
3.29172313e-01 -7.82095134e-01 -6.22108042e-01 -6.70368195e-01
-1.03925459e-01 3.27102363e-01 -3.53676319e-01 -2.73050386e-02
-4.20933068e-01 -3.63225430e-01 6.56965494e-01 5.26205242e-01
7.02593982e-01 -1.07484591e+00 -1.68891698e-01 -1.07234471e-01
-4.13668633e-01 -1.57733536e+00 -6.67084217e-01 -2.69145668e-01
-6.67631686e-01 -9.19152915e-01 -8.95561039e-01 -9.75285530e-01
5.36476433e-01 8.81950974e-01 9.15070832e-01 2.52357006e-01
-1.11102484e-01 9.95310485e-01 -8.19915235e-01 -2.77848005e-01
-9.06240568e-02 -2.66221128e-02 -7.93129653e-02 -1.52004594e-02
5.37243068e-01 -3.09390783e-01 -6.85651720e-01 3.83755654e-01
-1.19178033e+00 1.00227118e-01 7.01071501e-01 6.86853409e-01
5.43390810e-01 -1.54616758e-01 1.98200300e-01 -5.61207354e-01
6.02519572e-01 -7.62840271e-01 -4.22884256e-01 4.72074538e-01
-4.06568319e-01 -2.75574565e-01 4.37859893e-01 -7.18197346e-01
-9.35352325e-01 -4.37091775e-02 6.57088608e-02 -1.14039254e+00
-5.78041673e-02 7.89529026e-01 -3.17621864e-02 4.45639603e-02
5.26074886e-01 5.46320915e-01 1.67682115e-02 -3.23027134e-01
4.10259634e-01 8.37104023e-01 4.94917363e-01 -7.22182930e-01
8.71750176e-01 5.98541856e-01 -6.33534610e-01 -1.04497123e+00
-6.78627491e-01 -7.15144932e-01 -3.73365849e-01 -3.76597852e-01
1.00294733e+00 -1.65457678e+00 -2.68157065e-01 3.38046938e-01
-1.00531483e+00 -2.82977164e-01 2.79584825e-01 5.85075021e-01
-3.07885259e-01 6.63311005e-01 -5.56231678e-01 -4.63205129e-01
-3.88687164e-01 -1.17355382e+00 1.02578175e+00 1.84944287e-01
3.79218549e-01 -8.12270820e-01 -8.21071863e-02 7.51401246e-01
4.09919083e-01 -3.07504207e-01 4.17900592e-01 -5.58967948e-01
-8.00830841e-01 7.55536705e-02 -5.80813229e-01 5.66221416e-01
-2.42165804e-01 2.08415881e-01 -8.58650267e-01 -4.48140979e-01
-1.09434195e-01 -6.10187769e-01 9.96900618e-01 3.19033921e-01
1.37828374e+00 -2.02970117e-01 -1.99849442e-01 7.65441298e-01
1.60569334e+00 4.13522124e-03 7.81143010e-01 5.44577956e-01
8.24820876e-01 2.93513149e-01 8.11741292e-01 4.18924749e-01
4.81023818e-01 5.79867661e-01 3.07209373e-01 -4.76717763e-02
-2.81333357e-01 -5.01445532e-01 8.90552878e-01 1.11691463e+00
-7.49617368e-02 -4.15957391e-01 -8.40089977e-01 6.51332796e-01
-1.94038486e+00 -9.72039342e-01 9.28452983e-02 1.77558494e+00
6.81128681e-01 1.99230965e-02 -2.21077688e-02 -4.12818611e-01
5.36353230e-01 4.33575064e-01 -3.45323563e-01 8.75794590e-02
-2.79226363e-01 -2.98442453e-01 5.81027269e-01 7.49345049e-02
-1.13066113e+00 1.28953397e+00 5.46123075e+00 1.23308206e+00
-1.47343361e+00 2.72091448e-01 3.15505356e-01 -2.14140147e-01
-6.27616942e-02 2.33428944e-02 -8.87747645e-01 6.48267388e-01
8.73386204e-01 -3.12076569e-01 4.57847953e-01 8.84418488e-01
2.70514876e-01 -2.71432847e-02 -9.45954084e-01 1.37630916e+00
6.12792671e-01 -1.33644843e+00 3.62869233e-01 -7.69949928e-02
6.19303346e-01 5.10884941e-01 1.05644554e-01 7.35939622e-01
-8.47021416e-02 -6.75685346e-01 7.66796947e-01 2.45160267e-01
8.89837027e-01 -2.51303315e-01 4.74641442e-01 2.81613350e-01
-1.21212912e+00 -3.33479308e-02 -4.98046964e-01 2.48555452e-01
4.32503708e-02 3.94579083e-01 -5.01814067e-01 6.37233317e-01
9.50240076e-01 1.06646407e+00 -8.60059857e-01 8.46985221e-01
-1.25600651e-01 8.65530193e-01 -1.94001362e-01 2.46843174e-01
3.51289541e-01 -1.89041287e-01 3.86312872e-01 1.51093876e+00
2.78286964e-01 -8.57787132e-02 3.61679584e-01 6.33345127e-01
-1.90489918e-01 4.44130719e-01 -7.93430984e-01 -2.98861921e-01
3.66206795e-01 1.24179637e+00 -4.08615321e-01 -3.58483404e-01
-9.83405828e-01 8.93387794e-01 1.36245057e-01 6.66951597e-01
-1.18718469e+00 2.21309327e-02 3.77143651e-01 1.82921425e-01
4.14258450e-01 -3.21292609e-01 2.79673129e-01 -1.55411768e+00
1.40443549e-01 -1.15630949e+00 5.25427639e-01 -1.15987217e+00
-1.33111000e+00 4.11307395e-01 2.51685947e-01 -1.39803219e+00
9.25575793e-02 -6.09673440e-01 -4.24886703e-01 2.60249346e-01
-1.78933918e+00 -1.37246096e+00 -3.90623450e-01 1.24828255e+00
9.88780439e-01 -4.36710030e-01 2.30210021e-01 7.19699502e-01
-5.98267138e-01 6.42719448e-01 2.38767609e-01 5.35079539e-01
1.17193520e+00 -5.20159364e-01 -2.53981501e-01 7.51414478e-01
4.74919587e-01 4.76649731e-01 3.71826231e-01 -6.01311684e-01
-1.87335408e+00 -1.07472861e+00 4.09421355e-01 -4.35749859e-01
8.68575871e-01 -3.09449583e-01 -8.65219474e-01 7.78341293e-01
6.29473984e-01 5.81974871e-02 4.27695841e-01 -2.24210158e-01
-5.61569214e-01 -2.32265919e-01 -6.39810801e-01 6.65807009e-01
1.00408554e+00 -7.55199969e-01 -6.44846141e-01 4.49964345e-01
8.48434448e-01 -1.73939750e-01 -6.80644214e-01 2.97297329e-01
4.91553485e-01 -8.19143295e-01 1.19063294e+00 -3.72646868e-01
8.11313272e-01 -2.15168044e-01 -4.57454562e-01 -7.79854476e-01
-8.00602660e-02 -2.22296968e-01 8.40131119e-02 1.48382223e+00
1.30564347e-01 -2.94946283e-01 5.68200886e-01 1.73117265e-01
1.22781403e-01 -4.60537702e-01 -6.86889172e-01 -8.05277884e-01
8.73801634e-02 -5.70021272e-01 6.59900829e-02 1.03537524e+00
-3.15434664e-01 3.35447907e-01 -6.87298834e-01 5.80864027e-02
4.21399027e-01 -1.23198636e-01 1.03815842e+00 -6.02316856e-01
-1.84648469e-01 -1.10072859e-01 -2.41529733e-01 -1.18634689e+00
3.14258665e-01 -9.00516868e-01 -1.65027022e-01 -1.37453103e+00
6.49506986e-01 -3.28791812e-02 -4.71873134e-01 5.17785788e-01
-5.57767898e-02 2.94455469e-01 6.41503572e-01 6.17718458e-01
-1.14339340e+00 5.85481703e-01 1.28568149e+00 -2.78921217e-01
6.17071390e-02 -5.85969150e-01 -5.84552765e-01 7.74536967e-01
6.67893589e-01 -4.62032139e-01 -5.11547983e-01 -8.16712797e-01
4.67657149e-02 5.04110828e-02 3.44544977e-01 -1.00384879e+00
3.18798900e-01 -3.33285332e-01 2.42350027e-01 -6.67663395e-01
3.56768489e-01 -9.37175333e-01 -9.90103930e-02 1.31905213e-01
-4.24213022e-01 1.75984934e-01 2.19807625e-01 6.06017172e-01
-5.37633896e-01 -1.74766272e-01 3.96030307e-01 -2.19327673e-01
-1.25165176e+00 2.57585078e-01 -3.45455080e-01 2.89968699e-01
9.82929468e-01 -7.81738609e-02 -5.42959869e-01 -5.54741144e-01
-1.53719917e-01 4.30943608e-01 4.37377870e-01 1.00909460e+00
6.25342727e-01 -1.27836597e+00 -6.97540700e-01 1.67301297e-02
4.70108539e-01 -4.86402452e-01 3.98169965e-01 9.13357496e-01
-4.94849473e-01 5.67042232e-01 -1.86065078e-01 -7.76437342e-01
-1.41148818e+00 7.11175203e-01 -2.33372562e-02 -8.04788992e-02
-5.29321969e-01 6.10919178e-01 5.24248183e-01 -1.55114487e-01
5.10034740e-01 -1.03669576e-01 -7.58145899e-02 5.22468798e-02
5.79673350e-01 -5.32290190e-02 -2.60418504e-01 -7.67168403e-01
-3.70102614e-01 7.45793462e-01 -2.94914305e-01 4.65164930e-02
1.15455270e+00 -3.38707387e-01 7.38750249e-02 2.43107393e-01
1.47414768e+00 -1.01771839e-01 -1.20418429e+00 -4.65820819e-01
-3.34731013e-01 -5.91421664e-01 1.10185713e-01 -6.71869934e-01
-1.42413890e+00 8.96216035e-01 7.23398447e-01 -2.91259229e-01
1.20938587e+00 9.89885256e-02 7.73503304e-01 4.64297950e-01
2.94912487e-01 -1.22308111e+00 5.34722865e-01 5.05140007e-01
9.87125516e-01 -1.63171089e+00 1.58057913e-01 -2.22033083e-01
-1.03105342e+00 9.04868424e-01 8.87209535e-01 -1.10049732e-01
7.16058850e-01 6.53263927e-02 3.25042635e-01 -6.35101125e-02
-8.99075568e-01 -2.92599469e-01 2.51319259e-01 4.16589826e-01
4.04670238e-01 -1.57534301e-01 -4.31638777e-01 4.19005275e-01
1.59388736e-01 3.94879431e-01 3.79030436e-01 9.96393919e-01
-3.52988988e-01 -9.88920867e-01 -4.72166300e-01 1.69279009e-01
-5.51552057e-01 -2.10340142e-01 -3.89715225e-01 9.02313828e-01
2.90217213e-02 8.26076627e-01 -1.61786556e-01 -5.80790639e-01
1.69151857e-01 -1.08744591e-01 2.17046097e-01 -3.37998569e-01
-4.25706685e-01 3.82133186e-01 -1.11028634e-01 -6.42398715e-01
-7.10550547e-01 -3.23417634e-01 -1.03766906e+00 -2.99882680e-01
-2.50532508e-01 -6.40484467e-02 6.64075196e-01 7.13467836e-01
4.57624048e-01 2.16607168e-01 4.27681476e-01 -6.34341657e-01
-3.77453268e-01 -7.66712129e-01 -3.97334248e-01 6.57069325e-01
-2.75811572e-02 -5.07281005e-01 -5.07266700e-01 4.33163285e-01]
|
[10.319777488708496, 0.9855776429176331]
|
c15e4e83-123c-4428-9ffe-e064cf22ed76
|
volatility-inspired-s-lstm-cell
|
2205.07022
| null |
https://arxiv.org/abs/2205.07022v1
|
https://arxiv.org/pdf/2205.07022v1.pdf
|
Volatility-inspired $σ$-LSTM cell
|
Volatility models of price fluctuations are well studied in the econometrics literature, with more than 50 years of theoretical and empirical findings. The recent advancements in neural networks (NN) in the deep learning field have naturally offered novel econometric modeling tools. However, there is still a lack of explainability and stylized knowledge about volatility modeling with neural networks; the use of stylized facts could help improve the performance of the NN for the volatility prediction task. In this paper, we investigate how the knowledge about the "physics" of the volatility process can be used as an inductive bias to design or constrain a cell state of long short-term memory (LSTM) for volatility forecasting. We introduce a new type of $\sigma$-LSTM cell with a stochastic processing layer, design its learning mechanism and show good out-of-sample forecasting performance.
|
['Nino Antulov-Fantulin', 'German Rodikov']
|
2022-05-14
| null | null | null | null |
['econometrics']
|
['miscellaneous']
|
[-4.71826524e-01 -1.44780055e-01 -1.18381344e-01 -4.44094688e-01
-4.10877280e-02 -2.74492055e-01 8.88583362e-01 -1.56640068e-01
-1.78313389e-01 6.85256600e-01 2.15547964e-01 -7.07203567e-01
-1.70888096e-01 -1.00922775e+00 -6.54974401e-01 -7.84202933e-01
-1.46912649e-01 4.16455835e-01 -2.37138793e-01 -2.62322098e-01
2.12215379e-01 3.98265481e-01 -1.15470421e+00 2.00682491e-01
7.93976903e-01 1.61200142e+00 4.75551859e-02 7.21361265e-02
-7.97757447e-01 1.13514984e+00 -5.17007291e-01 -6.62132978e-01
3.20922166e-01 -1.15876287e-01 -3.22895855e-01 -5.71905971e-01
-1.84183180e-01 1.75311677e-02 -4.19889659e-01 1.17688644e+00
2.55907238e-01 1.11201078e-01 7.25979090e-01 -9.12011445e-01
-1.07871640e+00 1.14677799e+00 -4.03440088e-01 5.40677428e-01
-5.12482464e-01 4.00081545e-01 7.96541512e-01 -4.28095520e-01
3.40167969e-01 1.33916736e+00 6.97093070e-01 4.57072914e-01
-1.23303378e+00 -8.79609585e-01 5.66769898e-01 -9.08063203e-02
-9.89152253e-01 -1.06995739e-01 1.00044942e+00 -6.89515293e-01
1.05367398e+00 -2.49049604e-01 8.72379243e-01 1.55905163e+00
8.48220289e-01 5.73083639e-01 1.30496597e+00 1.80720016e-02
3.61879736e-01 1.67140946e-01 1.52490512e-01 3.75084579e-01
3.97990555e-01 6.35147512e-01 -2.85116762e-01 1.75406132e-02
1.01358366e+00 1.03658676e-01 2.77827412e-01 -4.25545499e-02
-6.33664966e-01 1.11605728e+00 5.20646572e-01 5.35543561e-01
-5.92248321e-01 5.70835829e-01 6.11001194e-01 6.09176278e-01
9.62418675e-01 5.31774640e-01 -6.57661438e-01 -2.77125657e-01
-8.91741574e-01 2.81731695e-01 8.11798990e-01 3.82539749e-01
4.23357189e-01 1.14030457e+00 -1.15907371e-01 5.65905333e-01
2.81433225e-01 7.74678409e-01 9.62219596e-01 -5.86696982e-01
3.52346957e-01 2.17010945e-01 -1.07705571e-01 -1.06861103e+00
-3.45637292e-01 -5.53088009e-01 -1.20913303e+00 2.15883583e-01
3.01713467e-01 -5.34295082e-01 -8.89549434e-01 1.83377922e+00
-2.91843951e-01 3.19241643e-01 1.97120667e-01 4.32820439e-01
4.84534860e-01 7.71132171e-01 2.51511902e-01 -1.63561061e-01
9.66650128e-01 -6.10501587e-01 -7.93751359e-01 -2.38734066e-01
3.78455102e-01 -2.59917438e-01 7.10507572e-01 3.05071861e-01
-8.77175629e-01 -6.46143079e-01 -6.38694525e-01 2.47767910e-01
-8.72232497e-01 -5.61744630e-01 1.09934807e+00 7.70761251e-01
-9.30399001e-01 8.66792202e-01 -6.31971478e-01 2.18386143e-01
2.46349618e-01 4.03047949e-01 2.34086588e-01 6.03647113e-01
-1.56575930e+00 1.10533285e+00 3.21475327e-01 4.44939226e-01
-6.02357984e-01 -5.45651972e-01 -6.91962838e-01 3.70095998e-01
1.31550163e-01 -9.00040567e-01 1.27682376e+00 -1.26787853e+00
-1.86904645e+00 3.87959987e-01 -5.39109334e-02 -9.63752210e-01
3.95671457e-01 4.84752320e-02 -5.96884072e-01 -5.05623460e-01
-3.93869787e-01 3.93154681e-01 1.15151083e+00 -8.36207926e-01
-5.09232163e-01 -2.38689750e-01 -2.53801644e-01 -4.65621412e-01
-1.21138208e-01 -7.67393336e-02 2.12304309e-01 -1.12928843e+00
-2.76221670e-02 -7.28628755e-01 -4.53011215e-01 -6.93089902e-01
-7.14740902e-02 -3.02510500e-01 4.31189418e-01 -4.23080713e-01
1.32849419e+00 -1.94544375e+00 -4.08160239e-01 5.27484059e-01
-7.47622736e-03 3.22071105e-01 -8.18782151e-02 2.51654476e-01
-2.91387916e-01 3.73765826e-01 -1.11649401e-01 -2.96024352e-01
3.19798470e-01 2.10778043e-01 -1.31582534e+00 -3.09007615e-02
9.57924500e-02 1.40553892e+00 -6.21544242e-01 2.20396131e-01
3.89426917e-01 5.14975011e-01 -1.56290635e-01 3.96013260e-04
-5.29052258e-01 3.16684932e-01 -4.94292736e-01 3.69656444e-01
5.80133319e-01 -1.51008489e-02 6.65409537e-03 4.68357176e-01
-1.35023922e-01 4.87785429e-01 -1.07443357e+00 8.99579406e-01
-5.49208820e-01 7.49672830e-01 -2.37441882e-01 -9.94678915e-01
1.20747042e+00 2.29582146e-01 3.01487714e-01 -1.01448381e+00
3.60307127e-01 3.40145767e-01 2.21335098e-01 -2.66795218e-01
2.66453981e-01 -5.19844472e-01 -1.62748337e-01 4.45010424e-01
-1.64823577e-01 -1.02688149e-01 -4.31779288e-02 -4.47650462e-01
5.42868197e-01 -2.08344743e-01 -1.07774183e-01 -5.23064077e-01
3.18725199e-01 -5.10538936e-01 5.76808929e-01 9.03326273e-01
1.00155346e-01 6.52500521e-03 7.67512143e-01 -9.22287047e-01
-8.66741419e-01 -9.56747174e-01 -3.79363656e-01 7.86745906e-01
-5.02166629e-01 -2.67419219e-03 -6.73651934e-01 -2.70254165e-02
3.50214273e-01 8.40706110e-01 -8.03760588e-01 -1.20746560e-01
-5.29124677e-01 -9.75953281e-01 6.11138344e-01 8.51144314e-01
4.21430796e-01 -1.57698965e+00 -3.44925612e-01 4.35319424e-01
5.10986209e-01 -8.94342840e-01 4.20181789e-02 6.50083780e-01
-1.28477466e+00 -4.75559860e-01 -5.98648906e-01 -4.36413974e-01
1.46487787e-01 -3.48476231e-01 1.04608798e+00 -3.40684325e-01
4.26963627e-01 6.72134161e-02 1.82933956e-01 -8.17736089e-01
-2.25418553e-01 2.00681955e-01 3.41464907e-01 2.22357493e-02
6.23543024e-01 -7.85846949e-01 -2.10914031e-01 -3.06670934e-01
-8.37915480e-01 -4.10159767e-01 3.54791135e-01 7.34133005e-01
3.54560822e-01 3.01037133e-01 6.66465759e-01 -9.12403822e-01
1.25967181e+00 -5.53979814e-01 -1.24135292e+00 -1.05778575e-01
-1.13676262e+00 6.35111272e-01 8.02803934e-01 -4.70976174e-01
-1.26460290e+00 -6.69464588e-01 -1.53098881e-01 -5.78430414e-01
1.19489068e-02 9.45795655e-01 1.58343852e-01 1.53512955e-01
-7.55443573e-02 1.19361944e-01 -3.50678474e-01 -5.93221366e-01
2.00149685e-01 1.44739792e-01 2.90841579e-01 -6.41680658e-01
6.81207240e-01 2.69460469e-01 1.05507530e-01 -5.26791513e-01
-6.74094260e-01 3.88235807e-01 -2.86000401e-01 1.70071602e-01
6.44211888e-01 -9.50267732e-01 -9.62187111e-01 7.83820033e-01
-1.08712041e+00 -6.47775888e-01 -4.88270283e-01 5.05649328e-01
-3.86999696e-01 -1.55334666e-01 -8.51252913e-01 -1.25121510e+00
-4.48318660e-01 -1.10840034e+00 5.33400238e-01 2.45580852e-01
-2.25509003e-01 -1.58823621e+00 1.95974410e-01 -3.95699471e-01
8.62006724e-01 8.59128609e-02 1.25177598e+00 -6.47729099e-01
-6.15694284e-01 -1.89808220e-01 -3.49189527e-02 4.61252511e-01
-2.42816389e-01 2.24859908e-01 -1.27602994e+00 3.40014249e-01
5.01409292e-01 1.22981042e-01 1.45524824e+00 1.18432820e+00
1.49701989e+00 -2.48139054e-01 -7.31236562e-02 9.49288368e-01
1.22090006e+00 5.00370145e-01 6.51172400e-01 4.91174668e-01
4.70252186e-01 7.30298400e-01 -1.17031664e-01 4.15738136e-01
3.19939107e-01 6.23860583e-02 1.68962866e-01 1.97916880e-01
6.10708475e-01 -4.17372197e-01 6.31241202e-01 1.11049974e+00
-2.10450187e-01 -2.43135855e-01 -9.23932791e-01 1.76289022e-01
-1.80769598e+00 -1.37535870e+00 8.00971966e-03 1.90495765e+00
5.16048193e-01 6.26680434e-01 -1.84205860e-01 -2.29520887e-01
5.11421621e-01 6.61662400e-01 -5.55911958e-01 -1.07966936e+00
-3.83782417e-01 3.53072643e-01 5.03722608e-01 5.90793371e-01
-1.10203326e+00 1.17259252e+00 7.26754713e+00 9.31622446e-01
-1.56227994e+00 -2.67167509e-01 1.09504306e+00 5.40697724e-02
-6.79022014e-01 -2.42132872e-01 -9.82948542e-01 8.49999189e-01
1.45912087e+00 -2.27550730e-01 3.96377295e-01 8.67338121e-01
4.96792465e-01 -2.04676036e-02 -8.91015887e-01 9.88731384e-01
-4.48657811e-01 -1.60103273e+00 3.76973659e-01 2.37599477e-01
9.66734052e-01 1.80546388e-01 8.54137480e-01 7.37726033e-01
5.96527100e-01 -1.27185047e+00 5.99689066e-01 1.07881916e+00
2.98599511e-01 -9.44821358e-01 1.03250873e+00 3.10665637e-01
-1.18374586e+00 -4.09049273e-01 -7.01186836e-01 -4.76590365e-01
1.29878536e-01 1.00616074e+00 -1.59319058e-01 4.93955761e-02
7.05643833e-01 4.27920938e-01 -4.08437848e-01 5.45426190e-01
-1.33276016e-01 7.44256735e-01 -2.61507869e-01 -2.01341316e-01
6.82009161e-01 -7.43674636e-01 2.85075337e-01 1.14683020e+00
3.34992081e-01 -6.31137341e-02 -2.31277570e-01 1.27596200e+00
-1.66660279e-01 -1.51777551e-01 -8.29776347e-01 -5.55916965e-01
1.67102382e-01 7.32923090e-01 -6.46422625e-01 -4.66772377e-01
-2.51510352e-01 3.66535753e-01 1.86172336e-01 6.03543937e-01
-6.87759638e-01 -2.10868027e-02 8.75994861e-01 3.06067746e-02
5.38131356e-01 -2.74763018e-01 -9.02603507e-01 -1.34700191e+00
4.70503531e-02 -7.70521283e-01 1.44597754e-01 -5.58051109e-01
-1.69282961e+00 6.88178778e-01 -9.45232064e-02 -5.74855268e-01
-7.69954920e-01 -1.02801692e+00 -1.05191791e+00 1.22248304e+00
-1.69437265e+00 -6.61445796e-01 2.27027893e-01 2.58282602e-01
3.49116832e-01 -7.16228783e-01 7.59957135e-01 -2.57859200e-01
-6.27018213e-01 2.18590796e-01 8.99253309e-01 2.64988422e-01
2.59172231e-01 -1.31629288e+00 1.01222968e+00 5.07384658e-01
1.08100012e-01 9.24553812e-01 7.48353839e-01 -8.15759897e-01
-1.15553224e+00 -8.99766147e-01 8.68610322e-01 -5.73235512e-01
1.25235891e+00 -4.83296692e-01 -1.03935051e+00 7.88972199e-01
2.85157502e-01 -2.13115066e-01 3.78567725e-01 3.55281711e-01
-3.40391278e-01 -4.18800801e-01 -9.22878981e-01 6.45320773e-01
5.91319919e-01 -7.76440620e-01 -7.18341827e-01 -3.00326645e-01
5.89054525e-01 -3.63322347e-02 -6.89778507e-01 3.32817733e-01
6.61481619e-01 -9.26666796e-01 7.61420131e-01 -7.39734471e-01
4.16510880e-01 3.85745376e-01 -7.17776790e-02 -1.52750516e+00
-3.50668758e-01 -7.81982839e-01 -4.79877830e-01 1.12115633e+00
3.60037208e-01 -1.17602253e+00 6.37249172e-01 1.03449118e+00
4.31179069e-02 -8.40374410e-01 -9.09174919e-01 -1.02762628e+00
8.13099802e-01 -7.72302151e-01 8.38076293e-01 7.69730926e-01
-2.11881682e-01 8.23296979e-02 -2.41820917e-01 -3.22438657e-01
3.37207764e-01 1.42979428e-01 5.24706364e-01 -1.70105481e+00
-9.96567160e-02 -1.11976588e+00 -1.32424027e-01 -9.07478750e-01
7.15172708e-01 -6.46837652e-01 -4.53395218e-01 -1.03293085e+00
-1.40971124e-01 -1.92931592e-01 -7.57883847e-01 -5.74698076e-02
5.95152527e-02 -3.40485245e-01 1.72657430e-01 -2.66636685e-02
7.48945326e-02 7.99633384e-01 8.05096090e-01 -2.42962420e-01
-2.14312091e-01 5.65548986e-02 -5.96005917e-01 1.00366688e+00
1.01873815e+00 -4.97271299e-01 -3.15399170e-01 -4.73048776e-01
4.59880024e-01 -7.09333420e-02 1.94469675e-01 -8.71875167e-01
-1.01114698e-02 -2.95470536e-01 7.71524847e-01 -5.52707613e-01
2.93024242e-01 -6.87727928e-01 7.19960555e-02 5.59322298e-01
-3.99027437e-01 7.17309356e-01 4.35396075e-01 5.42717576e-01
-4.04438555e-01 -3.32959116e-01 6.31103873e-01 -4.06269699e-01
-6.31587207e-01 5.93923450e-01 -6.96662486e-01 7.97156319e-02
5.34244657e-01 -3.40231694e-02 -8.88091773e-02 -5.12582183e-01
-6.66879296e-01 1.04140162e-01 9.00679827e-02 5.95488966e-01
3.25630665e-01 -1.20885408e+00 -2.76440948e-01 3.52530271e-01
-5.91811657e-01 -4.58612323e-01 1.71426833e-01 4.97382402e-01
-2.34886840e-01 1.05033755e+00 -1.19607560e-01 -1.33166984e-01
-2.04999477e-01 8.87865305e-01 5.71272433e-01 -5.72849691e-01
-5.54023802e-01 6.60902977e-01 3.35323125e-01 -1.74562395e-01
3.14312160e-01 -8.55166793e-01 -2.15572983e-01 4.27241564e-01
3.77101630e-01 3.13259393e-01 -1.83322012e-01 -3.37014854e-01
3.76069508e-02 4.33984458e-01 1.35608718e-01 -3.29549581e-01
1.68665171e+00 -5.27530834e-02 -7.57250488e-02 1.19707465e+00
8.62263620e-01 -1.60476238e-01 -1.08154130e+00 -1.81690678e-01
5.58598995e-01 -7.96074793e-02 2.38277301e-01 -5.83191991e-01
-1.31483030e+00 1.41049349e+00 5.24111211e-01 5.34246325e-01
5.60983717e-01 -4.77806270e-01 9.18108284e-01 7.73572862e-01
2.09423915e-01 -1.50480187e+00 -2.36126944e-01 1.11037004e+00
7.47304440e-01 -1.35859859e+00 -5.34521699e-01 4.80983704e-01
-8.33204687e-01 1.32566643e+00 3.93064708e-01 -5.00848293e-01
1.13185596e+00 4.74202305e-01 2.39052862e-01 -2.55076796e-01
-1.02788925e+00 -1.38585716e-01 1.37300700e-01 4.63994771e-01
5.43709934e-01 2.85028785e-01 1.15057766e-01 1.07392013e+00
-5.27643859e-01 -1.41238883e-01 2.92939425e-01 2.03387976e-01
-4.16777045e-01 -8.35901320e-01 -2.55363852e-01 6.33883178e-01
-6.74000740e-01 -4.80450928e-01 -2.52581626e-01 7.75882661e-01
-2.87912926e-03 5.78830540e-01 4.69617933e-01 -1.51273370e-01
2.97339410e-02 5.30104518e-01 1.07449390e-01 -3.22724342e-01
-8.99616003e-01 1.81238517e-01 -7.76866749e-02 -2.52908528e-01
2.20151562e-02 -7.36203730e-01 -8.29285741e-01 -6.70451224e-01
2.59105824e-02 2.45302022e-01 5.43417633e-01 1.08275688e+00
1.64426953e-01 5.38199961e-01 6.24199927e-01 -7.17210710e-01
-8.90976489e-01 -1.01547468e+00 -9.96012151e-01 1.16913848e-01
3.33095551e-01 -7.59528935e-01 -4.29275513e-01 -2.16133446e-01]
|
[4.584366321563721, 4.156107425689697]
|
28b52fde-bf93-4f5b-925e-b10f24471886
|
explainable-artificial-intelligence-toward
|
2302.06613
| null |
https://arxiv.org/abs/2302.06613v1
|
https://arxiv.org/pdf/2302.06613v1.pdf
|
Explainable artificial intelligence toward usable and trustworthy computer-aided early diagnosis of multiple sclerosis from Optical Coherence Tomography
|
Background: Several studies indicate that the anterior visual pathway provides information about the dynamics of axonal degeneration in Multiple Sclerosis (MS). Current research in the field is focused on the quest for the most discriminative features among patients and controls and the development of machine learning models that yield computer-aided solutions widely usable in clinical practice. However, most studies are conducted with small samples and the models are used as black boxes. Clinicians should not trust machine learning decisions unless they come with comprehensive and easily understandable explanations. Materials and methods: A total of 216 eyes from 111 healthy controls and 100 eyes from 59 patients with relapsing-remitting MS were enrolled. The feature set was obtained from the thickness of the ganglion cell layer (GCL) and the retinal nerve fiber layer (RNFL). Measurements were acquired by the novel Posterior Pole protocol from Spectralis Optical Coherence Tomography (OCT) device. We compared two black-box methods (gradient boosting and random forests) with a glass-box method (explainable boosting machine). Explainability was studied using SHAP for the black-box methods and the scores of the glass-box method. Results: The best-performing models were obtained for the GCL layer. Explainability pointed out to the temporal location of the GCL layer that is usually broken or thinning in MS and the relationship between low thickness values and high probability of MS, which is coherent with clinical knowledge. Conclusions: The insights on how to use explainability shown in this work represent a first important step toward a trustworthy computer-aided solution for the diagnosis of MS with OCT.
|
['Elena Garcia-Martin', 'Elvira Mayordomo', 'Beatriz Cordon', 'Elisa Vilades', 'Ubaldo Ramon-Julvez', 'Monica Hernandez']
|
2023-02-13
| null | null | null | null |
['clinical-knowledge']
|
['miscellaneous']
|
[-3.82230873e-03 -1.45057470e-01 -4.62884635e-01 -2.95275122e-01
-2.31321320e-01 -2.06601679e-01 1.99731678e-01 -5.43214455e-02
-5.13489962e-01 1.09439754e+00 2.42469355e-01 -5.56358397e-01
-4.55378324e-01 -3.52527112e-01 -3.08885247e-01 -7.67493665e-01
-2.70555794e-01 7.31642544e-01 -2.58220732e-02 1.88380294e-02
4.44997817e-01 6.48412704e-01 -1.54377377e+00 8.34901810e-01
1.28059649e+00 8.76350343e-01 5.96371353e-01 5.80333710e-01
5.78581775e-03 3.21578413e-01 1.87505838e-02 -1.05850607e-01
1.43422008e-01 -4.10705984e-01 -7.57508695e-01 2.08218284e-02
3.82511824e-01 -4.61418092e-01 3.66690248e-01 7.72066057e-01
6.37004912e-01 -5.18432856e-01 7.55501151e-01 -8.57575834e-01
-5.47030449e-01 1.99971318e-01 -7.10617244e-01 5.16829073e-01
-1.46215782e-01 3.55594367e-01 4.46238965e-01 -6.58738136e-01
8.69509161e-01 1.03231645e+00 5.26398540e-01 6.25266314e-01
-1.36808228e+00 -5.83204091e-01 9.17581022e-02 1.09262013e+00
-9.51216578e-01 -1.72293648e-01 -9.35437232e-02 -1.10503113e+00
1.11206055e+00 5.42100370e-01 1.36718869e+00 5.16942263e-01
5.38099647e-01 1.47360802e-01 2.01715922e+00 -4.81293589e-01
1.14011474e-01 2.01768298e-02 5.69558024e-01 5.18993497e-01
4.60872352e-01 4.53473747e-01 -3.21360677e-01 -2.54456609e-01
8.00900042e-01 -1.87788218e-01 -5.00749290e-01 -1.37760848e-01
-1.05268741e+00 7.11123943e-01 5.73310077e-01 3.42376411e-01
-6.63263559e-01 -2.42922440e-01 9.59617570e-02 1.36908457e-01
3.84173423e-01 2.31516346e-01 -3.51729453e-01 1.46687135e-01
-1.03597450e+00 1.14137419e-01 9.67376903e-02 2.34026704e-02
4.00668681e-01 -2.32831731e-01 -1.81287125e-01 6.88458204e-01
4.10651863e-01 5.45300126e-01 3.38218838e-01 -8.91828835e-01
9.13972855e-02 8.55214894e-01 5.21931425e-02 -4.42384630e-01
-7.06628740e-01 -7.73852646e-01 -7.18181252e-01 1.05909967e+00
5.57713032e-01 -1.15640327e-01 -9.78019297e-01 1.15542448e+00
6.83975667e-02 -2.17058674e-01 -6.18491828e-01 1.34419012e+00
2.37598151e-01 -5.89130707e-02 -9.56882350e-03 -4.41056013e-01
1.46782684e+00 -7.63634503e-01 -4.52806681e-01 -3.15113097e-01
7.08838642e-01 -7.34708130e-01 9.54711735e-01 7.53952801e-01
-9.80259180e-01 -2.29321852e-01 -6.87581182e-01 7.65012428e-02
1.46638468e-01 5.88433087e-01 5.90485990e-01 5.59030771e-01
-1.27654624e+00 4.97214466e-01 -8.96692455e-01 -6.35856986e-01
4.79036242e-01 5.31569660e-01 -5.42994559e-01 -4.76448387e-02
-4.36809123e-01 1.48617923e+00 -7.60819390e-02 3.62342536e-01
-1.51552901e-01 -4.94593590e-01 1.68359533e-01 -3.82745385e-01
-3.77435505e-01 -1.36868453e+00 6.72431827e-01 -1.06333590e+00
-8.75385582e-01 1.14327133e+00 -6.81146860e-01 -5.13624668e-01
4.04010028e-01 -3.66961241e-01 -2.57043600e-01 3.83726686e-01
1.60860345e-02 3.93824965e-01 6.41387939e-01 -1.10302162e+00
-5.34122765e-01 -1.15377533e+00 -4.50391710e-01 -3.53417218e-01
3.03053528e-01 3.05395067e-01 6.57133818e-01 -2.51866549e-01
6.03540361e-01 -9.48302090e-01 -3.22032154e-01 2.90547669e-01
-5.28831244e-01 1.15521461e-01 3.14452469e-01 -1.17864561e+00
1.17136037e+00 -1.49606371e+00 -1.39367476e-01 1.61457464e-01
7.31943369e-01 3.87604773e-01 6.25968650e-02 1.68606400e-01
-5.81028104e-01 3.41518819e-01 1.82705075e-01 2.49757305e-01
-4.44411039e-01 -2.44886696e-01 -3.04576997e-02 6.04777098e-01
-8.81875120e-03 7.26010025e-01 -3.52576792e-01 -2.07945570e-01
3.07526320e-01 4.49052095e-01 -5.56421578e-01 -2.60350883e-01
1.02110736e-01 8.48361492e-01 -2.41278350e-01 5.30305088e-01
7.03407109e-01 -3.34425956e-01 1.46514684e-01 -2.94926614e-01
-3.97380650e-01 1.31474808e-01 -7.76975453e-01 1.10144019e+00
-1.29367933e-02 7.85907805e-01 -1.29039213e-01 -6.13713086e-01
7.88709342e-01 1.35163531e-01 9.27622020e-02 -3.85158032e-01
2.17388973e-01 6.25810087e-01 7.20497549e-01 -7.47570038e-01
-4.33364004e-01 -3.99023294e-01 1.30378747e+00 3.43150944e-01
-5.39052546e-01 3.15908551e-01 1.09904945e-01 -2.06304297e-01
7.97820389e-01 1.95950978e-02 6.25443935e-01 -1.08508967e-01
4.37147498e-01 1.95877880e-01 4.38297302e-01 5.35786033e-01
-1.86463729e-01 7.78164804e-01 6.51717842e-01 -6.93002701e-01
-9.83746707e-01 -1.00420666e+00 -5.67413628e-01 6.58037812e-02
-4.22234118e-01 -3.78567837e-02 -5.99320173e-01 -1.39692158e-01
7.32982624e-03 5.87394834e-01 -7.10566998e-01 2.17538983e-01
-2.11268798e-01 -8.12179863e-01 -3.57975483e-01 2.05807760e-01
2.62999207e-01 -5.12767196e-01 -5.34388125e-01 6.99501559e-02
-3.00027907e-01 -6.51381373e-01 2.98428476e-01 -4.26533341e-01
-1.47142351e+00 -1.39623713e+00 -8.98350358e-01 -3.41697395e-01
7.86728978e-01 2.35738963e-01 7.61442959e-01 1.96126565e-01
-6.46722853e-01 -1.42238155e-01 -2.70670623e-01 -4.45602238e-01
-1.47156835e-01 -4.59189087e-01 1.60440296e-01 -7.38510117e-03
6.02490187e-01 -8.36616457e-01 -1.15524340e+00 2.73674339e-01
-2.07047328e-01 3.71177107e-01 9.93133903e-01 5.17012000e-01
7.29990542e-01 -5.58925629e-01 2.29263112e-01 -4.75215018e-01
3.42250675e-01 -4.97209430e-01 -1.68983117e-01 2.72118539e-01
-1.14647925e+00 -2.88759112e-01 -1.59432366e-01 -3.09111029e-02
-5.26333988e-01 -5.05814075e-01 1.89626709e-01 6.95432425e-02
-2.74062306e-01 2.72152781e-01 2.54649043e-01 -2.68952578e-01
9.57345247e-01 -1.16667661e-05 6.77574635e-01 -7.11258054e-01
8.93943459e-02 1.02185023e+00 8.41434672e-02 -1.23029537e-01
9.40869600e-02 8.48893344e-01 1.56935796e-01 -7.55235195e-01
-4.71696138e-01 -4.77166623e-01 -5.32563627e-01 -5.23653150e-01
7.66150296e-01 -4.13747638e-01 -6.75422251e-01 4.20478731e-01
-1.18715286e+00 -1.08949214e-01 4.48083282e-02 1.26163256e+00
-7.24634707e-01 2.65365660e-01 -2.88830906e-01 -7.93202817e-01
-6.79556668e-01 -1.23484802e+00 5.96825957e-01 4.52316962e-02
-5.31359673e-01 -5.52769423e-01 2.11653918e-01 8.00514400e-01
4.15353358e-01 1.60473153e-01 1.65415752e+00 -2.99692392e-01
-6.17096305e-01 -5.36296293e-02 -6.45356119e-01 3.81540745e-01
-4.75271344e-02 2.45958582e-01 -7.29138494e-01 -1.10776663e-01
4.53976728e-02 1.36953026e-01 9.38966691e-01 1.25789809e+00
5.91175973e-01 -3.72061506e-02 -3.85864764e-01 3.15889657e-01
1.67198443e+00 1.76309243e-01 1.04470944e+00 8.33714306e-01
5.29948175e-02 9.63968933e-01 2.99381942e-01 2.44832456e-01
1.54325619e-01 7.73544014e-01 4.91107881e-01 -1.89738944e-01
-5.86946428e-01 4.76841658e-01 2.41172295e-02 3.96217287e-01
-9.31827962e-01 4.22542155e-01 -1.12363207e+00 4.03250068e-01
-1.68809581e+00 -8.03768158e-01 -1.00677454e+00 2.24728227e+00
4.23853219e-01 -8.24164152e-02 3.42959970e-01 3.96172144e-02
8.62601101e-01 -6.53654873e-01 -4.55616981e-01 -3.91036600e-01
-3.23003680e-01 1.95689693e-01 1.25751480e-01 3.78223717e-01
-2.81872481e-01 3.79766315e-01 6.63181591e+00 3.12510043e-01
-1.39140439e+00 2.23871917e-01 3.18294942e-01 -5.90968609e-01
-1.83638081e-01 2.86054134e-01 -6.87049508e-01 4.65135008e-01
6.84586585e-01 1.65362597e-01 3.94549638e-01 1.86324134e-01
1.00323117e+00 -5.58163643e-01 -5.79393327e-01 8.01640570e-01
-5.35239041e-01 -1.53053558e+00 1.04480118e-01 3.20882171e-01
3.92164499e-01 3.61217946e-01 1.02894478e-01 -4.84625489e-01
-5.28807640e-01 -1.01665068e+00 4.16085005e-01 1.26154423e+00
7.46960223e-01 -1.65462732e-01 9.42363024e-01 8.48736428e-03
-2.42540330e-01 -1.66268542e-01 -1.89743295e-01 -4.07779425e-01
1.27192423e-01 1.10196638e+00 -1.18772328e+00 1.46957546e-01
8.04740548e-01 4.81209129e-01 -3.82184684e-01 1.67790329e+00
-3.86991441e-01 7.51028180e-01 3.59819196e-02 2.15939030e-01
-4.94003929e-02 -4.44215298e-01 1.02550948e+00 5.65415025e-01
3.16149265e-01 -5.10923527e-02 -6.18935347e-01 1.17974579e+00
1.07547176e+00 3.98216963e-01 -1.49772286e-01 -1.28057972e-01
9.09856632e-02 9.70567822e-01 -5.67459941e-01 2.72591650e-01
-5.80891013e-01 3.18661094e-01 8.87804553e-02 5.45900345e-01
-1.80012509e-02 2.31349200e-01 5.75529635e-01 9.29742217e-01
-1.11012958e-01 -8.01045224e-02 -9.43845153e-01 -7.72163570e-01
2.26019055e-01 -8.33149254e-01 3.58103551e-02 -1.28033102e+00
-1.17861235e+00 7.23205626e-01 -3.81810129e-01 -1.15719783e+00
-2.00686261e-01 -8.23421240e-01 -3.84987175e-01 1.41924644e+00
-1.33706415e+00 -1.29702508e+00 -2.14821815e-01 3.40654641e-01
-1.08853057e-01 -4.06041563e-01 9.39265609e-01 -1.75070077e-01
-4.00097817e-01 -1.90362900e-01 2.52595663e-01 -4.34298158e-01
6.21995091e-01 -1.08410680e+00 -6.13032468e-02 4.94882494e-01
-4.70703661e-01 9.23680902e-01 7.57767320e-01 -9.34536338e-01
-3.55981648e-01 -4.45241392e-01 1.05417979e+00 -1.30628571e-01
6.36610985e-01 4.68474180e-01 -7.19201744e-01 4.25555736e-01
1.52212288e-02 -3.90974581e-01 1.07029676e+00 4.33326304e-01
-1.38747454e-01 -9.19577554e-02 -1.07468736e+00 5.75274706e-01
6.75779045e-01 -1.92112461e-01 -6.65094137e-01 3.31108868e-01
-3.59997042e-02 1.25740245e-01 -7.11947680e-01 5.07805228e-01
1.06567335e+00 -1.54564714e+00 6.58364713e-01 -1.13851130e+00
3.19862276e-01 -2.81829506e-01 -6.12994023e-02 -1.06835687e+00
-3.28904361e-01 -2.93845892e-01 1.28296092e-01 4.79963422e-01
4.37762320e-01 -9.43095148e-01 9.10797298e-01 5.88366449e-01
-7.21171498e-03 -1.15468645e+00 -8.82147253e-01 -5.70472836e-01
1.42957598e-01 -2.65264690e-01 1.13306247e-01 4.19674277e-01
-7.01748356e-02 -4.94566076e-02 1.52560070e-01 5.82858436e-02
7.22042203e-01 2.96762675e-01 2.63583958e-01 -1.48805726e+00
-1.68391779e-01 -4.62293804e-01 -9.32000518e-01 -1.24200664e-01
-2.87148714e-01 -1.05230629e+00 -9.14424479e-01 -2.13057137e+00
6.69562578e-01 -5.53507388e-01 -4.18782309e-02 2.57398367e-01
-1.81539319e-02 1.14130683e-01 1.42945230e-01 5.52783966e-01
5.03277183e-01 -9.97720063e-02 1.66741478e+00 1.25563800e-01
-2.14011118e-01 4.09176379e-01 -8.38606179e-01 9.77238417e-01
8.53948653e-01 -4.73922908e-01 -3.33321869e-01 -3.98182660e-01
4.39980090e-01 -4.94925603e-02 1.10621107e+00 -8.23793590e-01
1.34076979e-02 -2.30337799e-01 3.74036908e-01 -6.54891014e-01
3.12630743e-01 -5.19333482e-01 2.25961939e-01 7.93590307e-01
2.49313697e-01 -3.71353477e-01 -8.39760527e-02 2.73172677e-01
8.24044570e-02 -2.67925054e-01 9.02046442e-01 -1.33679643e-01
-4.67406929e-01 8.73958517e-04 -4.98773813e-01 -3.69571000e-01
9.51812744e-01 -7.76911616e-01 -7.10384905e-01 -3.32544655e-01
-1.51620805e+00 -2.47060806e-01 4.07115757e-01 1.97785683e-02
5.97675800e-01 -8.70922327e-01 -1.02177930e+00 -1.94392446e-02
-1.58632532e-01 -7.48405576e-01 5.68308115e-01 1.85704875e+00
-7.55839109e-01 8.10870409e-01 -8.66933942e-01 -8.42198908e-01
-1.50156617e+00 6.65855557e-02 6.66379273e-01 1.15874838e-02
-5.14199138e-01 8.51453006e-01 3.39617766e-02 1.13916829e-01
-1.11815110e-01 -2.70558208e-01 -3.85454208e-01 1.00893430e-01
8.79085720e-01 9.58341777e-01 3.23063552e-01 -2.80033052e-01
-2.91609615e-01 8.47828627e-01 -2.74434000e-01 -8.57482925e-02
1.33910096e+00 -4.74368989e-01 -8.40452969e-01 3.70546490e-01
5.62273145e-01 -5.92436865e-02 -5.82116663e-01 1.36699051e-01
-2.56791711e-01 -5.02651930e-01 3.96041185e-01 -1.48175335e+00
-7.50469923e-01 1.10106421e+00 1.47553647e+00 -1.00399546e-01
1.09754133e+00 -1.07231587e-01 3.76938552e-01 -9.56618562e-02
7.39029408e-01 -6.15183413e-01 -7.06018984e-01 -1.80550829e-01
1.27529538e+00 -8.07853103e-01 1.54647142e-01 -5.56098163e-01
-4.40550804e-01 1.30680680e+00 2.15552643e-01 3.96726392e-02
5.99930644e-01 -2.83649087e-01 2.91514754e-01 -2.31284156e-01
-5.76137245e-01 -5.22501230e-01 5.05075812e-01 1.01858354e+00
3.95707518e-01 3.39721680e-01 -1.20915008e+00 9.26956475e-01
-2.86690205e-01 4.84617591e-01 4.80977237e-01 3.51853818e-01
-8.66858840e-01 -1.34441102e+00 -5.16212106e-01 1.09964490e+00
-2.55011022e-01 -1.98745430e-01 -6.59895778e-01 9.04929519e-01
5.40542781e-01 9.63183045e-01 -9.96678416e-03 -2.58053720e-01
5.93653508e-02 1.20426804e-01 8.11906934e-01 -5.86917639e-01
-3.98996532e-01 6.14461601e-02 3.35735887e-01 -5.56121111e-01
-4.50982541e-01 -8.60647142e-01 -1.28611183e+00 -3.45278949e-01
-3.43171597e-01 -1.40039384e-01 9.48296607e-01 1.18343639e+00
5.08176625e-01 1.48857921e-01 1.95307612e-01 -3.53982776e-01
-1.99842423e-01 -1.06904650e+00 -1.08962476e+00 -2.62594461e-01
4.07292873e-01 -8.13984215e-01 -3.49324048e-01 -3.92978787e-02]
|
[15.801244735717773, -3.9593403339385986]
|
78f1a42f-bd9f-44c5-aedc-77b33a4ba4b3
|
modelling-stance-detection-as-textual
|
2212.06543
| null |
https://arxiv.org/abs/2212.06543v1
|
https://arxiv.org/pdf/2212.06543v1.pdf
|
Modelling Stance Detection as Textual Entailment Recognition and Leveraging Measurement Knowledge from Social Sciences
|
Stance detection (SD) can be considered a special case of textual entailment recognition (TER), a generic natural language task. Modelling SD as TER may offer benefits like more training data and a more general learning scheme. In this paper, we present an initial empirical analysis of this approach. We apply it to a difficult but relevant test case where no existing labelled SD dataset is available, because this is where modelling SD as TER may be especially helpful. We also leverage measurement knowledge from social sciences to improve model performance. We discuss our findings and suggest future research directions.
|
['Ayoub Bagheri', 'Anastasia Giachanou', 'Qixiang Fang']
|
2022-12-13
| null | null | null | null |
['stance-detection']
|
['natural-language-processing']
|
[ 4.41625237e-01 6.32356465e-01 -6.92340016e-01 -5.64243972e-01
-1.06298125e+00 -5.57521641e-01 9.96290267e-01 4.66513515e-01
-6.88908458e-01 9.13846016e-01 7.04821229e-01 -8.04080427e-01
9.82786529e-03 -5.60767114e-01 -5.50960183e-01 -1.97544456e-01
1.07880034e-01 4.74179864e-01 1.38482690e-01 -2.26593032e-01
5.33799410e-01 1.78065836e-01 -1.27111804e+00 5.91312706e-01
7.89020181e-01 6.02891982e-01 -7.94046596e-02 3.87175649e-01
1.41293049e-01 1.36437595e+00 -7.86311567e-01 -8.66060853e-01
1.83400940e-02 -1.97903618e-01 -1.25346887e+00 1.08912706e-01
2.07297817e-01 -1.91038907e-01 2.23281801e-01 5.38338125e-01
3.00233454e-01 2.31028404e-02 9.69815850e-01 -1.07551467e+00
-3.59461606e-01 1.11263001e+00 -4.31020916e-01 4.11314189e-01
7.66212821e-01 -9.12723131e-03 1.61720848e+00 -6.65461719e-01
8.55841994e-01 1.33298862e+00 6.61671698e-01 4.90016453e-02
-1.12063313e+00 -3.72697115e-01 1.92043260e-01 3.80358934e-01
-6.19866788e-01 -4.62604612e-01 7.65585840e-01 -5.17763317e-01
1.02194583e+00 3.35574150e-01 2.39661470e-01 1.55813599e+00
8.03506747e-02 1.15694320e+00 1.69110739e+00 -6.83974743e-01
2.62994990e-02 1.97312847e-01 5.45778573e-01 3.43172342e-01
5.18296301e-01 -1.46949366e-01 -4.04291898e-01 -3.38887453e-01
1.21400081e-01 -3.37438315e-01 4.60295975e-02 2.21458897e-01
-1.13424134e+00 1.15817213e+00 1.76667258e-01 2.92597950e-01
-4.78125811e-01 -1.53030694e-01 6.17960393e-01 8.25048685e-01
8.21544886e-01 7.27945864e-01 -5.03167033e-01 -4.70584661e-01
-1.04565108e+00 4.79858726e-01 1.33106661e+00 5.41199505e-01
3.51335555e-01 -7.05224872e-02 -7.05198050e-02 8.45111072e-01
3.96638632e-01 1.36261597e-01 5.08597493e-01 -1.07805586e+00
8.53551865e-01 6.62430286e-01 9.59587991e-02 -8.65144610e-01
-3.82783502e-01 -2.37303331e-01 -2.15505317e-01 -2.46557087e-01
5.40106177e-01 -3.22842211e-01 -3.93261731e-01 1.36736476e+00
8.48372057e-02 -9.94859785e-02 1.29935443e-01 7.05697775e-01
8.52821290e-01 4.16404486e-01 1.84493568e-02 -4.25149739e-01
1.49722874e+00 -5.67354262e-01 -6.91698492e-01 -7.97745705e-01
1.31159246e+00 -9.25953448e-01 1.16532457e+00 4.18743104e-01
-1.03401911e+00 3.09801232e-02 -9.25388694e-01 -7.27703944e-02
-2.53248334e-01 -3.89090367e-02 6.73176169e-01 5.61356008e-01
-5.69604635e-01 6.36855066e-01 -5.62457681e-01 -3.61391842e-01
2.06399947e-01 -1.10101901e-01 5.19358777e-02 -2.09258452e-01
-1.56403530e+00 1.48025930e+00 1.69196963e-01 -9.43920463e-02
-3.32429677e-01 -2.70511359e-01 -1.20487511e+00 -5.46703815e-01
8.18493247e-01 -2.41439983e-01 1.64938962e+00 -6.58103585e-01
-1.17416799e+00 1.24454808e+00 -2.66702384e-01 -6.43233597e-01
6.55886054e-01 -1.16188183e-01 -2.10499808e-01 -1.75456226e-01
4.96946812e-01 -1.93463594e-01 5.68355799e-01 -9.32526827e-01
-3.22718143e-01 -2.74201632e-01 3.16977382e-01 1.45030245e-01
-8.14949945e-02 6.46274030e-01 4.08856601e-01 -8.53433907e-01
-2.75985133e-02 -9.65433538e-01 -1.37360170e-01 -6.74770594e-01
-3.88580978e-01 -5.26497781e-01 4.93646115e-01 -8.61452699e-01
1.71329236e+00 -1.44889808e+00 1.64662436e-01 2.49357328e-01
2.13058963e-01 2.03399494e-01 2.17944205e-01 8.98949146e-01
-4.00670990e-02 3.28337759e-01 -3.29466939e-01 -4.26250100e-01
2.49968201e-01 4.53118563e-01 -8.73382539e-02 5.62029541e-01
3.66272449e-01 1.02812886e+00 -6.70538545e-01 -5.93091607e-01
-2.73919642e-01 -1.35638788e-01 -4.08671468e-01 -2.05679521e-01
-1.45222411e-01 -3.57576869e-02 -3.35653484e-01 6.86620951e-01
1.86623067e-01 -1.91731632e-01 4.76154089e-01 4.24087308e-02
-1.86274156e-01 1.09706247e+00 -1.03065908e+00 7.74016023e-01
-3.73260438e-01 9.68462348e-01 -1.80406854e-01 -1.44759941e+00
7.67486453e-01 2.86332369e-01 1.68237984e-01 -6.57091737e-01
2.19650134e-01 2.39099741e-01 4.14653897e-01 -7.89248407e-01
3.86048228e-01 -4.62225527e-01 -3.45697314e-01 1.07175422e+00
-5.36927283e-01 -1.25200406e-01 4.28645432e-01 2.16310114e-01
9.93920743e-01 -1.69687271e-01 8.79114687e-01 -2.07010165e-01
2.90690541e-01 9.35688689e-02 4.81614500e-01 6.59911871e-01
-1.79478750e-01 1.92629069e-01 7.94919312e-01 -4.74978387e-02
-1.06678998e+00 -4.18451697e-01 -4.00894523e-01 1.09304726e+00
-6.26426697e-01 -7.53624678e-01 -4.45176870e-01 -8.75892639e-01
1.24408118e-02 9.17010188e-01 -6.15145802e-01 5.80106229e-02
-8.11618567e-01 -8.05530787e-01 6.37802362e-01 6.30842507e-01
1.38143092e-01 -1.22976530e+00 -6.78362370e-01 7.81736448e-02
-4.86851037e-01 -1.22063863e+00 1.88816153e-02 4.10092056e-01
-7.22595453e-01 -1.23511887e+00 -2.86401719e-01 -6.92593753e-01
4.15195264e-02 2.06226841e-01 1.30525327e+00 1.49015233e-01
2.01833889e-01 4.85489778e-02 -7.02858210e-01 -7.25745142e-01
-9.06667471e-01 1.84740022e-01 6.38079345e-02 -3.89657736e-01
7.28795826e-01 -2.47732446e-01 -6.11922704e-02 7.71148428e-02
-1.06670082e+00 -6.95580170e-02 3.98451626e-01 8.19221318e-01
-4.07603718e-02 -2.55135864e-01 8.51840854e-01 -1.33366656e+00
1.24160874e+00 -9.14890468e-01 -1.25636429e-01 -1.25885785e-01
-7.87899971e-01 -1.09357402e-01 2.20548302e-01 -3.63425672e-01
-9.18720961e-01 -7.31854916e-01 -2.84672707e-01 5.55528820e-01
-1.66567668e-01 1.59128392e+00 2.45090812e-01 3.00698429e-01
8.58334482e-01 -2.91055709e-01 2.49434099e-01 -3.81274939e-01
5.07927164e-02 1.17416382e+00 -2.13640362e-01 -5.84544599e-01
6.54416382e-01 1.36125877e-01 -2.13168576e-01 -9.27741289e-01
-1.36932302e+00 -6.23262405e-01 -6.05687022e-01 -9.71161723e-02
2.90302336e-01 -8.66089225e-01 -4.29253459e-01 -1.30193248e-01
-9.28890169e-01 -6.37847304e-01 7.70964772e-02 5.41862786e-01
-6.58073246e-01 5.49364626e-01 -8.84609878e-01 -7.99416602e-01
-8.15110430e-02 -8.76341760e-01 8.85194778e-01 -3.83897126e-01
-1.10158515e+00 -1.43492436e+00 2.92948596e-02 8.69889140e-01
6.44405186e-02 5.19093692e-01 8.26471627e-01 -1.10019767e+00
1.28136083e-01 -3.51395071e-01 6.34983033e-02 4.57475215e-01
2.17781588e-01 1.13386035e-01 -7.20247447e-01 -1.92360535e-01
4.52597141e-01 -8.05908799e-01 7.40225315e-01 2.65289038e-01
5.04966915e-01 -6.95748389e-01 5.30377869e-03 -2.23346502e-01
1.10494947e+00 -2.41707712e-01 3.30938190e-01 9.21156764e-01
3.83968979e-01 1.10088623e+00 1.16720927e+00 6.22778893e-01
8.59686673e-01 6.90617323e-01 -6.02540635e-02 -4.44872957e-03
5.74492365e-02 3.86994556e-02 6.28878653e-01 9.83893096e-01
-6.99763522e-02 -1.04114071e-01 -1.35374868e+00 4.87274557e-01
-1.93947530e+00 -1.18197834e+00 -5.11903405e-01 1.72103512e+00
1.27490008e+00 5.98694563e-01 7.12611616e-01 7.48367548e-01
2.56119400e-01 2.35890478e-01 6.29137382e-02 -8.76248837e-01
-1.49758890e-01 2.95062602e-01 2.05996588e-01 5.26789725e-01
-9.78329182e-01 7.66654134e-01 6.99115276e+00 5.15431404e-01
-7.48917282e-01 1.06902502e-01 3.49924028e-01 4.22194190e-02
-3.33959848e-01 1.23676829e-01 -5.98706067e-01 3.94584328e-01
1.24022341e+00 -2.16191858e-01 -1.14798108e-02 5.86001337e-01
3.90723079e-01 -2.70311683e-01 -1.29551125e+00 4.23339367e-01
2.17268899e-01 -1.15945125e+00 -3.54180545e-01 2.00028077e-01
5.10778785e-01 1.06454566e-01 -1.61272541e-01 3.83894086e-01
4.88850594e-01 -9.10147607e-01 7.06382811e-01 3.64394188e-02
4.34702724e-01 -5.89594364e-01 1.04055035e+00 8.07487011e-01
-6.69155180e-01 -3.20557594e-01 -1.51796073e-01 -7.50868320e-01
6.95992410e-02 4.16261911e-01 -1.12752032e+00 4.49678570e-01
4.02934939e-01 1.07839811e+00 -7.12667227e-01 7.60648370e-01
-4.79481459e-01 1.24228680e+00 -2.02632383e-01 -2.00750709e-01
4.80473608e-01 4.49315971e-03 5.63632488e-01 1.53453398e+00
-1.59745246e-01 7.18236296e-03 2.05709711e-01 5.76497495e-01
-3.29197804e-03 1.42114922e-01 -1.05344176e+00 -1.19431071e-01
4.69312608e-01 9.37206388e-01 -5.61862051e-01 -4.63401169e-01
-7.12192297e-01 4.25959259e-01 4.74963248e-01 1.06503405e-01
-6.73584759e-01 4.46831323e-02 4.10392672e-01 1.90040722e-01
1.07203275e-01 -1.76759034e-01 -4.29225683e-01 -1.33602273e+00
1.16191939e-01 -1.33456385e+00 5.80436885e-01 -3.59394580e-01
-1.59400022e+00 9.51589122e-02 3.25071454e-01 -1.19092703e+00
-7.11715162e-01 -7.60496080e-01 -5.29456258e-01 5.17508388e-01
-1.76236463e+00 -1.02897453e+00 1.46840200e-01 1.49824291e-01
6.79403424e-01 1.38998792e-01 6.14017248e-01 9.27916393e-02
-6.41897976e-01 5.52499413e-01 -1.56059757e-01 8.94490927e-02
9.02668238e-01 -1.32135379e+00 3.78031731e-01 6.93060577e-01
1.13606555e-02 7.40635335e-01 8.65528941e-01 -8.39635670e-01
-1.02102685e+00 -8.41226101e-01 1.55886149e+00 -6.71173215e-01
1.17489159e+00 -2.14851111e-01 -9.21027005e-01 1.17188478e+00
2.00692818e-01 -8.37227285e-01 1.13016498e+00 4.90725875e-01
-3.25982481e-01 4.86555219e-01 -9.58141267e-01 7.71183908e-01
8.71054888e-01 -7.31071651e-01 -1.33997679e+00 5.35084546e-01
3.89889598e-01 -3.52609485e-01 -1.15267861e+00 3.13546777e-01
3.30256015e-01 -6.72728062e-01 7.67527461e-01 -8.25130641e-01
9.71033037e-01 1.17465772e-01 -2.62592942e-01 -1.34673262e+00
-2.11544350e-01 -5.27843595e-01 -2.40331069e-01 1.24065018e+00
6.34812474e-01 -6.70871675e-01 5.16389132e-01 6.99475765e-01
-1.05457127e-01 -7.61004329e-01 -6.49801254e-01 -8.57280433e-01
5.52873254e-01 -7.69891381e-01 4.00464058e-01 1.36816335e+00
5.56381702e-01 6.13846779e-01 -2.82179981e-01 -2.54707694e-01
4.48438287e-01 2.28723839e-01 6.22988343e-01 -1.52433980e+00
-3.05172056e-01 -4.44815129e-01 -1.81090161e-01 -5.65690935e-01
5.77032626e-01 -1.20153356e+00 -2.03174382e-01 -1.70859206e+00
2.07666948e-01 -1.61173344e-01 2.24131271e-01 4.12074983e-01
-2.03376606e-01 1.33610114e-01 1.41400144e-01 8.91939625e-02
-3.84823948e-01 1.44928977e-01 9.61826622e-01 2.63207816e-02
1.53757766e-01 3.05347741e-01 -1.06595790e+00 8.39717627e-01
1.04711258e+00 -5.81301987e-01 -4.01813269e-01 -2.24260464e-01
4.47410107e-01 -1.42464554e-02 2.03823656e-01 -1.99219182e-01
-1.27752721e-02 -4.87742871e-01 -1.37720659e-01 -3.15319836e-01
1.38023019e-01 -5.69340587e-01 -3.27816099e-01 3.02246094e-01
-5.86353481e-01 2.57809520e-01 8.70238766e-02 1.95831761e-01
-2.51180440e-01 -7.14821577e-01 3.60305071e-01 -2.92602032e-01
-2.23490477e-01 -3.94226432e-01 -8.10799241e-01 2.95170158e-01
8.97367299e-01 -3.60817552e-01 -4.08627629e-01 -5.10828197e-01
-4.60955292e-01 1.45970702e-01 4.50290561e-01 2.46229067e-01
5.93570292e-01 -1.19079673e+00 -1.02326274e+00 -2.30058625e-01
2.82762587e-01 -4.50783163e-01 -4.86435533e-01 1.15426612e+00
-1.90401793e-01 5.06054997e-01 2.01951876e-01 -1.90842107e-01
-1.40520465e+00 2.88810015e-01 -8.99698064e-02 -4.15761411e-01
-4.81543660e-01 4.58410382e-01 -3.91066700e-01 -4.43601668e-01
8.81632268e-02 -3.78531158e-01 -6.15298748e-01 5.73897123e-01
3.91214043e-01 3.69677573e-01 1.13851100e-01 -7.27165103e-01
-3.52243483e-01 2.92808600e-02 -1.55166656e-01 -2.29522541e-01
1.59411037e+00 -2.33146861e-01 -3.28048915e-01 9.22625661e-01
1.08946681e+00 3.48373502e-02 -5.01833260e-01 -6.09858811e-01
8.26149404e-01 -3.24962616e-01 3.68058681e-03 -6.20926559e-01
-3.12656283e-01 5.39078355e-01 -3.39729518e-01 6.37475729e-01
6.42882347e-01 2.79822618e-01 4.17538077e-01 4.42295045e-01
2.34596282e-01 -1.22114944e+00 1.49772406e-01 8.65098357e-01
9.84602153e-01 -1.51204240e+00 3.98027778e-01 -3.59993935e-01
-8.90770137e-01 8.81982803e-01 3.25289011e-01 -2.71217562e-02
6.36017501e-01 2.99083531e-01 -7.25414753e-02 -4.75056529e-01
-1.04886651e+00 -2.62238055e-01 3.48464668e-01 3.48253608e-01
1.16497004e+00 9.39156860e-03 -9.03818905e-01 7.29148746e-01
-7.24027455e-01 -1.51095986e-01 9.76904154e-01 1.35732901e+00
-2.79321104e-01 -1.34176898e+00 -2.56194592e-01 9.72219467e-01
-8.57436955e-01 -3.75203699e-01 -8.25849473e-01 8.90990078e-01
-3.64753067e-01 1.25591111e+00 -3.30506086e-01 -3.62808645e-01
2.96159416e-01 3.16647857e-01 4.56093520e-01 -1.01874983e+00
-6.39206171e-01 -1.63064480e-01 9.54125166e-01 -3.10868114e-01
-8.35039735e-01 -1.16005933e+00 -9.11121249e-01 -8.14496517e-01
-3.72603774e-01 -9.26267058e-02 2.25970551e-01 1.44975770e+00
-1.12952851e-01 3.00582677e-01 5.28945923e-01 -4.62239146e-01
-8.63608658e-01 -1.20264304e+00 -3.49851131e-01 4.54831213e-01
3.97802174e-01 -7.45934665e-01 -3.36652547e-01 6.68541668e-03]
|
[9.16277027130127, 9.8939847946167]
|
36e88279-76c2-4068-bfd8-3691fdeccd78
|
diaasq-a-benchmark-of-conversational-aspect
|
2211.05705
| null |
https://arxiv.org/abs/2211.05705v4
|
https://arxiv.org/pdf/2211.05705v4.pdf
|
DiaASQ : A Benchmark of Conversational Aspect-based Sentiment Quadruple Analysis
|
The rapid development of aspect-based sentiment analysis (ABSA) within recent decades shows great potential for real-world society. The current ABSA works, however, are mostly limited to the scenario of a single text piece, leaving the study in dialogue contexts unexplored. To bridge the gap between fine-grained sentiment analysis and conversational opinion mining, in this work, we introduce a novel task of conversational aspect-based sentiment quadruple analysis, namely DiaASQ, aiming to detect the quadruple of target-aspect-opinion-sentiment in a dialogue. We manually construct a large-scale high-quality DiaASQ dataset in both Chinese and English languages. We deliberately develop a neural model to benchmark the task, which advances in effectively performing end-to-end quadruple prediction, and manages to incorporate rich dialogue-specific and discourse feature representations for better cross-utterance quadruple extraction. We hope the new benchmark will spur more advancements in the sentiment analysis community.
|
['Shengqiong Wu', 'Jinsong Zhang', 'Donghong Ji', 'Fei Li', 'Tat-Seng Chua', 'Lizi Liao', 'Yijiang Liu', 'Jingye Li', 'Yuhan Wu', 'Hao Fei', 'Bobo Li']
|
2022-11-10
| null | null | null | null |
['aspect-based-sentiment-analysis']
|
['natural-language-processing']
|
[ 2.84266621e-01 2.08381310e-01 -9.97055694e-02 -6.93055868e-01
-1.05302155e+00 -6.55270815e-01 1.07726908e+00 3.10224324e-01
-1.82391673e-01 5.97022533e-01 9.03814852e-01 -3.98500234e-01
3.89155984e-01 -7.46795475e-01 -2.60370970e-01 -3.90109658e-01
1.49344683e-01 5.82118213e-01 -4.30820026e-02 -1.05975068e+00
2.90394336e-01 -3.16498905e-01 -1.34662235e+00 8.74627531e-01
6.77484810e-01 1.07362425e+00 -2.89252013e-01 8.59782457e-01
-4.81692791e-01 9.60885108e-01 -1.00615203e+00 -9.30379212e-01
-3.06019872e-01 -6.46236718e-01 -1.11112857e+00 2.13478550e-01
2.48316098e-02 1.92336500e-01 2.44812816e-01 4.09961551e-01
6.66076303e-01 -3.41613963e-02 6.04299664e-01 -9.58186090e-01
-3.81831706e-01 9.02905345e-01 -3.74412417e-01 1.34422788e-02
7.48664439e-01 -1.25897869e-01 1.59988415e+00 -7.45698571e-01
6.44385695e-01 1.26301372e+00 6.36906385e-01 6.01352930e-01
-7.64449596e-01 -2.17544019e-01 3.18132788e-01 1.08038880e-01
-4.21164811e-01 -4.09668893e-01 9.59305644e-01 -3.51559550e-01
1.43131137e+00 4.54200029e-01 1.02360594e+00 1.44578052e+00
1.66245654e-01 1.37591767e+00 1.29344058e+00 -5.64636827e-01
1.94119047e-02 3.62605333e-01 4.93573755e-01 4.73700583e-01
-3.55560809e-01 -4.85137045e-01 -1.06037939e+00 -1.52085230e-01
-3.39166313e-01 -3.88463587e-01 -1.97105736e-01 -1.88675746e-01
-1.14973104e+00 1.23053420e+00 -2.23183826e-01 3.93671244e-01
-1.39153630e-01 -5.68637788e-01 1.08864510e+00 8.07750583e-01
9.33306992e-01 7.13384092e-01 -1.10954559e+00 -7.58652747e-01
-6.80070102e-01 5.40180743e-01 1.62196791e+00 6.94899619e-01
4.98655617e-01 -3.19562286e-01 -2.35992849e-01 1.17560291e+00
1.17910184e-01 3.96333486e-01 7.72717357e-01 -4.59465533e-01
7.26592958e-01 1.06175947e+00 -1.81179658e-01 -1.03569531e+00
-7.28453457e-01 -1.43196627e-01 -5.47210395e-01 -2.96952009e-01
3.55099499e-01 -5.74497521e-01 -2.19509184e-01 1.33580434e+00
5.52745819e-01 -6.99800730e-01 3.88362169e-01 5.92144370e-01
1.18761075e+00 6.62571907e-01 -3.04497927e-01 -3.21255684e-01
1.71767259e+00 -1.32608140e+00 -7.72761881e-01 -3.33292216e-01
9.91196156e-01 -9.87145305e-01 1.36608315e+00 4.24693972e-01
-9.24716651e-01 -2.28116944e-01 -1.06338775e+00 -3.95567082e-02
-6.79481804e-01 -8.06352049e-02 8.70056152e-01 8.13311160e-01
-6.51238441e-01 1.34357363e-01 -4.80682760e-01 -2.79167473e-01
1.71247452e-01 7.20808432e-02 -3.32863659e-01 2.23202735e-01
-1.41618836e+00 9.21163380e-01 -1.51204333e-01 4.49649878e-02
-3.40642273e-01 -5.91245115e-01 -1.13536429e+00 -3.01598668e-01
6.57510877e-01 -6.04097784e-01 1.62195778e+00 -1.22346687e+00
-1.81976867e+00 1.21782172e+00 -4.07559276e-01 -3.45128775e-01
2.06099465e-01 -3.98798376e-01 -4.64134783e-01 -3.89183939e-01
2.45247737e-01 1.57142207e-01 6.64559245e-01 -9.45262969e-01
-8.53642762e-01 -4.88934785e-01 4.02250707e-01 4.00890708e-01
-7.14638710e-01 3.54415983e-01 -2.40470931e-01 -4.43286210e-01
-3.30440819e-01 -9.37357068e-01 -1.92941055e-01 -7.74395049e-01
-6.63990557e-01 -6.83184624e-01 7.39971280e-01 -3.29940945e-01
1.21077585e+00 -1.78810096e+00 3.04308563e-01 -2.95296282e-01
4.24130708e-02 1.05162099e-01 -1.66942000e-01 7.97374189e-01
1.79076493e-01 -1.44563213e-01 -7.45857358e-02 -7.87050307e-01
1.57601357e-01 -2.82896608e-02 -3.95744592e-01 1.22988774e-02
3.86652321e-01 9.01313484e-01 -8.87776792e-01 -3.01107526e-01
-1.03457190e-01 2.22119048e-01 -5.94228625e-01 4.25087661e-01
-4.80620116e-01 3.67564380e-01 -7.41460621e-01 5.68928182e-01
2.55221099e-01 -1.73376158e-01 2.28695005e-01 -1.44186705e-01
-1.67620689e-01 8.85640204e-01 -4.31733072e-01 1.55895662e+00
-9.09984052e-01 8.12275767e-01 4.29701433e-02 -9.37773108e-01
1.01874733e+00 3.31678778e-01 4.52417433e-01 -6.34344101e-01
4.07277465e-01 7.88875297e-02 9.10410658e-02 -4.62822407e-01
1.06336963e+00 -2.45170206e-01 -8.14482033e-01 6.99774265e-01
1.73934713e-01 -6.27608597e-01 4.28192407e-01 2.66879976e-01
8.83531868e-01 -1.86034232e-01 6.16492927e-01 -1.38128325e-01
8.95660579e-01 4.06733483e-01 9.14763063e-02 3.92313242e-01
-2.51615137e-01 5.23004293e-01 1.04248273e+00 -4.60984588e-01
-7.45378375e-01 -2.12809920e-01 -1.24330744e-01 1.62488484e+00
-2.82273591e-01 -8.92873943e-01 -7.34279990e-01 -1.03700256e+00
-2.16806889e-01 5.96602559e-01 -7.48076379e-01 1.85496420e-01
-6.12429559e-01 -1.01368356e+00 3.35261583e-01 1.40175581e-01
3.64357293e-01 -1.30307007e+00 -2.78519958e-01 2.42264688e-01
-5.21242678e-01 -1.27263546e+00 -4.43282455e-01 1.72509044e-01
-3.14220577e-01 -1.13772619e+00 -4.80265409e-01 -6.80816412e-01
-1.42266586e-01 1.60002902e-01 1.76321232e+00 -2.85655618e-01
6.54399917e-02 4.17207688e-01 -9.68177319e-01 -8.08715820e-01
-7.70351410e-01 5.62285066e-01 -2.61222482e-01 -9.67804901e-03
9.15547132e-01 -2.32651070e-01 -4.25626904e-01 2.06054017e-01
-3.98196965e-01 -5.63808344e-02 1.19352810e-01 9.79142249e-01
1.47562012e-01 -2.66478777e-01 8.15184951e-01 -1.35735130e+00
1.25471127e+00 -4.28017735e-01 -1.88747272e-01 -1.63800567e-01
-4.37400401e-01 -2.63593197e-01 7.35900402e-01 -3.07399896e-03
-1.18111920e+00 -4.47192401e-01 -6.72210872e-01 7.16091156e-01
5.25189973e-02 1.03220689e+00 -1.56453535e-01 5.05137444e-01
5.01519680e-01 2.38188356e-01 -6.89416304e-02 -1.92104742e-01
5.69095254e-01 1.09913039e+00 8.81020539e-03 -3.37600321e-01
1.27758712e-01 2.99143255e-01 -4.49444264e-01 -1.09156322e+00
-1.50066078e+00 -7.39242554e-01 -4.82416034e-01 -3.42408687e-01
7.80477941e-01 -1.04523766e+00 -7.19092369e-01 5.44431746e-01
-1.12609160e+00 -7.62581676e-02 -2.88281858e-01 -1.86422095e-02
-5.84975183e-01 3.38208109e-01 -5.73459804e-01 -8.36670697e-01
-7.54021406e-01 -1.08590734e+00 1.23842561e+00 1.25013605e-01
-8.37794602e-01 -1.08954883e+00 5.94157159e-01 1.15017641e+00
2.96908706e-01 -6.35399073e-02 7.92403817e-01 -1.18916976e+00
-2.78865676e-02 -4.22800750e-01 1.88980088e-01 5.24110675e-01
1.19266108e-01 2.52244752e-02 -1.07409251e+00 -7.00661168e-02
2.22943142e-01 -8.55225623e-01 6.85922444e-01 7.94938654e-02
5.34348607e-01 -2.44087681e-01 5.87491319e-02 4.08575200e-02
7.05335021e-01 -3.59107628e-02 2.33646512e-01 7.04973519e-01
5.84908426e-01 9.66961503e-01 1.04341829e+00 5.54433107e-01
9.19748962e-01 6.63431287e-01 1.81281522e-01 7.99354389e-02
1.27400815e-01 1.01626217e-01 7.17040539e-01 1.72130096e+00
1.28267035e-02 -4.81490254e-01 -7.15879977e-01 8.79145801e-01
-1.78070164e+00 -7.57436633e-01 -1.53071031e-01 1.39776301e+00
1.20978844e+00 3.06467801e-01 3.77955765e-01 1.11102387e-01
1.13486536e-01 8.38670254e-01 -1.78795248e-01 -1.03960526e+00
-3.15411299e-01 -2.40664296e-02 -3.19383085e-01 2.84491450e-01
-1.39914942e+00 8.90725195e-01 5.64646578e+00 8.06778014e-01
-8.99910152e-01 1.83774292e-01 7.06313074e-01 1.84933376e-02
-4.73077923e-01 -2.20682040e-01 -1.05235958e+00 1.87108696e-01
1.19696259e+00 -5.62739931e-02 -1.12251356e-01 1.06986070e+00
2.02790424e-01 3.73464674e-02 -9.83578324e-01 5.85285485e-01
4.73776668e-01 -1.23730803e+00 -2.78232573e-03 -2.43442059e-01
8.86449456e-01 2.16149315e-01 -1.26126066e-01 7.28335023e-01
2.51307279e-01 -8.21424246e-01 3.23261678e-01 1.09408386e-01
2.19685346e-01 -8.76617908e-01 1.18060720e+00 2.82958448e-01
-8.83128345e-01 1.39718398e-01 -1.29345104e-01 -1.29240125e-01
3.81373197e-01 6.03525758e-01 -8.71548831e-01 4.62484479e-01
6.80037916e-01 1.02909303e+00 -3.03250223e-01 1.91166863e-01
-1.63757205e-01 6.92024350e-01 1.27730578e-01 -7.80965805e-01
5.05578220e-01 -3.70052397e-01 6.86701953e-01 1.60225689e+00
-2.02150926e-01 -1.62659109e-01 1.51269555e-01 2.91643530e-01
-1.22457668e-01 6.53858721e-01 -5.56219280e-01 -3.94207329e-01
-8.82535577e-02 1.58033550e+00 -3.48138630e-01 -2.61833161e-01
-8.55034590e-01 8.75404835e-01 3.27765167e-01 -1.83400393e-01
-3.13749969e-01 -2.58729339e-01 8.15356672e-01 -3.54921490e-01
4.54536736e-01 1.72034819e-02 -3.05035055e-01 -1.37873292e+00
3.05503141e-02 -1.69353747e+00 2.69075722e-01 -3.31903726e-01
-1.56592929e+00 9.77256119e-01 -4.43716228e-01 -9.59332585e-01
-7.06033826e-01 -8.41933727e-01 -8.18349242e-01 6.96602821e-01
-1.51745319e+00 -1.44829214e+00 7.51115680e-02 4.85380113e-01
1.22932553e+00 -4.47553575e-01 1.13799465e+00 1.15481399e-01
-4.30182248e-01 6.32045031e-01 1.57466810e-02 2.14386940e-01
9.66098905e-01 -1.43660998e+00 5.74983180e-01 2.18747392e-01
2.20790748e-02 5.25559783e-01 9.41808343e-01 -2.98066407e-01
-1.49346101e+00 -6.86151683e-01 1.39325714e+00 -1.01933813e+00
1.23446536e+00 -5.92389226e-01 -6.71117604e-01 4.70232934e-01
6.46246374e-01 -7.52133310e-01 1.11533272e+00 1.03207624e+00
-3.01770687e-01 5.43266796e-02 -6.31949723e-01 7.38208234e-01
4.78930056e-01 -8.28479230e-01 -8.21812153e-01 4.45554078e-01
9.43128288e-01 -3.33715022e-01 -1.08770502e+00 4.32426870e-01
6.67971909e-01 -1.16464734e+00 6.24862432e-01 -8.32211614e-01
9.10468638e-01 1.94044232e-01 -1.17932461e-01 -1.66448331e+00
5.23961365e-01 -6.41062617e-01 -1.34411246e-01 1.46845925e+00
9.57304239e-01 -4.16046679e-01 9.65580165e-01 2.72518843e-01
-2.82494307e-01 -1.16215491e+00 -7.49710619e-01 -1.13691822e-01
2.09238231e-01 -5.63676596e-01 5.48968315e-01 8.69644880e-01
7.51004934e-01 1.20326996e+00 -3.99833709e-01 -5.12399733e-01
-1.34514347e-02 6.55916512e-01 1.07161689e+00 -9.04365122e-01
-4.21599269e-01 -5.96657395e-01 -2.04140738e-01 -1.18132186e+00
1.59412965e-01 -5.25865018e-01 9.84818861e-02 -1.18337417e+00
2.47568026e-01 -1.07620969e-01 1.40018567e-01 7.95137063e-02
-3.79588366e-01 2.08968595e-01 -7.25692138e-03 -3.60503912e-01
-8.76600802e-01 9.67568159e-01 1.35414350e+00 -3.80179793e-01
-2.16396928e-01 4.80628252e-01 -1.06645155e+00 8.76355827e-01
6.66322649e-01 -1.99033916e-01 -3.22696090e-01 3.43080461e-02
7.75117755e-01 8.33140612e-02 -4.33336258e-01 -2.94939101e-01
-7.49442577e-02 1.10675953e-01 -2.27526769e-01 -8.04432631e-01
6.37446582e-01 -4.09975171e-01 -8.64637971e-01 -1.12017520e-01
-5.46355844e-01 8.42698291e-02 2.50577666e-02 4.14660573e-01
-7.46899009e-01 -2.33895004e-01 3.06945831e-01 -1.77578568e-01
-5.00507951e-01 -2.75875535e-02 -6.25489354e-01 4.52903748e-01
5.91010928e-01 1.40152931e-01 -4.14502412e-01 -7.87111044e-01
-5.52440524e-01 2.40194470e-01 9.79373753e-02 7.58832157e-01
2.12292194e-01 -8.03585351e-01 -8.73450935e-01 2.97954567e-02
5.39155185e-01 -1.62511215e-01 3.34214598e-01 8.03802967e-01
-1.64998900e-02 5.76566100e-01 2.81692117e-01 -4.92393672e-01
-1.62104774e+00 1.13946974e-01 1.18685812e-01 -7.45852828e-01
-2.82648474e-01 1.00271618e+00 -9.83284414e-02 -1.12227273e+00
6.37981063e-03 -4.77090999e-02 -9.17199433e-01 7.05863774e-01
6.64746940e-01 3.03595159e-02 2.46499225e-01 -9.49707329e-01
-1.39392167e-01 3.62270117e-01 -4.60890114e-01 -7.05934539e-02
1.37749505e+00 -4.29854989e-01 -3.75629455e-01 9.22020078e-01
1.29130542e+00 3.76691878e-01 -6.63905144e-01 -1.95822954e-01
6.00188784e-02 -1.59333155e-01 -1.34369507e-01 -9.37842429e-01
-6.68624341e-01 8.35198581e-01 -5.37593029e-02 9.73114491e-01
7.98659563e-01 1.63195819e-01 1.01101911e+00 6.96535826e-01
8.80052224e-02 -1.22044539e+00 2.45344475e-01 1.24576604e+00
1.00002503e+00 -1.66905355e+00 1.17395387e-03 -2.56158978e-01
-1.29691458e+00 1.06170583e+00 4.94189858e-01 1.08515248e-01
7.57488668e-01 8.64888802e-02 5.64327478e-01 -5.60362637e-01
-1.12778926e+00 -1.56580091e-01 3.52258205e-01 3.90667111e-01
9.62739825e-01 1.80488318e-01 -4.04554516e-01 9.91511345e-01
-9.51783299e-01 -4.42506790e-01 6.36580586e-01 8.03391337e-01
-3.22996706e-01 -1.30421865e+00 8.79443213e-02 5.92949033e-01
-7.96516538e-01 -3.91969085e-01 -8.04964662e-01 7.03067482e-01
-4.76731420e-01 1.35766423e+00 -1.56644434e-01 -3.67140830e-01
6.99519098e-01 1.50807887e-01 -5.53363608e-03 -7.28038251e-01
-1.17671561e+00 1.08200878e-01 1.03590739e+00 -5.26066840e-01
-8.50509584e-01 -8.21853697e-01 -7.70774126e-01 -4.11421925e-01
-4.40030545e-01 4.56680149e-01 8.25286865e-01 1.34075069e+00
2.81914264e-01 5.96001983e-01 9.26550150e-01 -6.10465944e-01
-4.51584250e-01 -1.32602990e+00 -5.03677726e-01 2.69936949e-01
4.48719501e-01 -1.25450999e-01 -4.02901053e-01 -2.59935051e-01]
|
[11.456389427185059, 6.965762138366699]
|
1922bdab-088b-47d1-818f-68e049417977
|
gn-transformer-fusing-ast-and-source-code
| null | null |
https://openreview.net/forum?id=XavM6v_q59q
|
https://openreview.net/pdf?id=XavM6v_q59q
|
GN-Transformer: Fusing AST and Source Code information in Graph Networks
|
As opposed to natural languages, source code understanding is influenced by grammar relations between tokens regardless of their identifier name. Considering graph representation of source code such as Abstract Syntax Tree (AST) and Control Flow Graph (CFG), can capture a token’s grammatical relationships that are not obvious from the source code. Most existing methods are late fusion and underperform when supplementing the source code text with a graph representation. We propose a novel method called GN-Transformer to fuse representations learned from graph and text modalities under the Graph Networks (GN) framework with attention mechanism. Our method learns the embedding on a constructed graph called Syntax-Code Graph (SCG). We perform experiments on the structure of SCG, an ablation study on the model design and the hyper-paramaters to conclude that the performance advantage is from the fusion method and not the specific details of the model. The proposed method achieved state of the art performance in two code summarization datasets and across three metrics.
|
['Barry Boehm', 'Iordanis Fostiropoulos', 'Junyan Cheng']
|
2021-01-01
| null | null | null | null |
['code-summarization']
|
['computer-code']
|
[ 3.16539109e-01 6.90571308e-01 -2.05193400e-01 -2.87527204e-01
-4.84548539e-01 -6.37599409e-01 6.54191971e-01 7.13761628e-01
2.72912145e-01 1.29671782e-01 8.35503340e-01 -6.20735943e-01
-6.41600266e-02 -6.84642136e-01 -8.90009403e-01 -1.02782600e-01
-2.88463086e-01 -3.81896079e-01 -1.12303942e-02 -2.03778699e-01
5.35592496e-01 -1.64441943e-01 -1.36933076e+00 3.66329998e-01
1.22917712e+00 5.24885714e-01 2.06111491e-01 7.10898519e-01
-8.66701663e-01 1.52141011e+00 -5.61586797e-01 -6.50603294e-01
-6.81521520e-02 -5.09072363e-01 -1.05144370e+00 -8.67104381e-02
7.61626065e-01 -7.35879540e-02 -3.83001775e-01 1.35185134e+00
3.09100449e-02 -3.93425286e-01 5.00129998e-01 -1.38382447e+00
-1.08656240e+00 1.45757675e+00 -7.08193004e-01 2.75681578e-02
5.98404288e-01 -1.48354217e-01 1.47756648e+00 -7.19496608e-01
7.91670978e-01 1.28808749e+00 9.47761893e-01 4.19113934e-01
-9.90394831e-01 -2.06117183e-01 4.92338270e-01 8.18582103e-02
-8.55429828e-01 -1.66390628e-01 9.69478905e-01 -7.86233783e-01
1.37342942e+00 -3.18313651e-02 5.00012934e-01 8.24055374e-01
6.76622868e-01 7.13753581e-01 4.25425321e-01 -3.25283587e-01
-1.15514733e-02 -1.58309802e-01 7.87940502e-01 1.28371727e+00
7.44104862e-01 -4.33737427e-01 -4.98186350e-01 -4.32917327e-01
3.68083566e-01 4.27453890e-02 -4.32637721e-01 -6.16409361e-01
-1.15757692e+00 7.75691628e-01 7.73367763e-01 3.70537728e-01
-1.35811180e-01 8.10998142e-01 7.48345435e-01 5.15107334e-01
2.77603298e-01 3.30622345e-01 -3.59659761e-01 -1.64739758e-01
-7.36022651e-01 -4.07012887e-02 1.05362833e+00 1.52455783e+00
1.02437985e+00 2.47782007e-01 -1.68101385e-01 5.42277992e-01
7.24942923e-01 1.93715304e-01 3.93553376e-01 -6.20394945e-01
8.64804924e-01 1.32926381e+00 -5.14882922e-01 -1.15591955e+00
-1.41275823e-01 -5.52013218e-01 -5.55945337e-01 -9.08232406e-02
1.62514284e-01 -1.19125620e-01 -8.55525315e-01 1.64923954e+00
-1.18954368e-01 -6.26453757e-02 1.20622464e-01 2.50215501e-01
1.37223065e+00 3.91371369e-01 2.41000392e-02 2.67071068e-01
1.31430638e+00 -1.09999275e+00 -7.51481593e-01 -4.89485949e-01
1.16392982e+00 -3.71159345e-01 7.34657824e-01 -1.59166351e-01
-7.54299283e-01 -3.78352374e-01 -1.24148095e+00 -3.70186806e-01
-3.60616624e-01 6.86957464e-02 7.28925765e-01 4.62446570e-01
-1.39992201e+00 6.29766583e-01 -6.93293393e-01 -6.47354305e-01
4.45482016e-01 1.52267337e-01 -5.87987840e-01 -1.04193456e-01
-7.58242846e-01 5.29842138e-01 6.10013664e-01 -1.68908685e-01
-9.62307572e-01 -8.29668522e-01 -1.42022061e+00 6.98727667e-01
4.64059114e-01 -8.25302422e-01 9.78484511e-01 -1.27717364e+00
-9.09559786e-01 7.53698468e-01 -1.20062269e-01 -4.24942285e-01
5.08676544e-02 -1.58580586e-01 -1.72809660e-01 2.07807515e-02
8.35914910e-02 8.62669200e-02 9.43169415e-01 -1.34850812e+00
-4.19334561e-01 -2.26273984e-01 6.41544700e-01 -1.56043440e-01
-2.38628134e-01 -5.68876155e-02 -1.47357300e-01 -5.05354166e-01
1.19506650e-01 -6.04983628e-01 5.03502376e-02 -3.48007888e-01
-7.89990723e-01 -3.82396877e-01 8.39811444e-01 -9.37231481e-01
1.67824090e+00 -2.28427577e+00 3.13159108e-01 5.10176877e-03
7.20200658e-01 4.62435633e-02 -3.95695537e-01 1.09055197e+00
-3.99074793e-01 6.67542100e-01 -5.07808924e-01 -8.10217038e-02
2.51038641e-01 5.86397536e-02 -4.85254467e-01 4.23840374e-01
1.88016757e-01 1.03077507e+00 -1.12282395e+00 -4.90300447e-01
-2.87807345e-01 1.94563046e-01 -9.55308378e-01 2.69248575e-01
-3.40427786e-01 -1.23576678e-01 -7.35108256e-01 6.40654385e-01
5.95192134e-01 -5.82964063e-01 4.07077283e-01 -2.76343584e-01
1.44690424e-01 3.25708061e-01 -8.35797071e-01 2.26500487e+00
-5.40792346e-01 5.96101403e-01 1.36877403e-01 -8.05447817e-01
9.19874430e-01 2.02355146e-01 1.48829266e-01 -1.79652929e-01
-5.62162604e-03 -5.94804585e-02 1.99543968e-01 -8.02222431e-01
5.63406229e-01 4.93882060e-01 -3.23286355e-01 5.10906696e-01
5.55623412e-01 -6.01257421e-02 3.64988148e-01 1.03174841e+00
1.62760222e+00 3.48707676e-01 5.61758518e-01 -6.05184674e-01
5.99068940e-01 -1.49827152e-01 4.46341038e-01 8.68957877e-01
1.16223305e-01 2.70003229e-01 1.37744415e+00 -1.65604934e-01
-7.50780523e-01 -8.01869571e-01 2.99820304e-01 1.04763353e+00
1.01271376e-01 -1.43765640e+00 -8.50001812e-01 -1.27450943e+00
9.82053578e-02 6.77964509e-01 -8.28510165e-01 -3.20146501e-01
-6.53100073e-01 -2.36706704e-01 6.85443223e-01 5.69111705e-01
3.81673336e-01 -6.63258731e-01 -3.67749661e-01 9.29016843e-02
-7.06028789e-02 -9.15444553e-01 -6.08393610e-01 -7.37197399e-02
-8.05912971e-01 -1.47454941e+00 -8.41600075e-02 -7.27383077e-01
9.59776044e-01 2.55598258e-02 1.44877923e+00 7.33933032e-01
2.04539984e-01 7.30308235e-01 -6.93668544e-01 -3.80083889e-01
-7.59363651e-01 2.20051274e-01 -7.83123434e-01 -3.57314982e-02
3.13201547e-02 -6.67949975e-01 -3.31286818e-01 -4.18942541e-01
-9.92796600e-01 2.39872098e-01 6.31125987e-01 6.70863926e-01
-6.92793801e-02 -4.03056055e-01 4.32208896e-01 -1.37324309e+00
8.23364258e-01 -8.17637444e-01 -4.09047723e-01 4.51125056e-01
-5.64872563e-01 5.94868898e-01 7.75283754e-01 2.06608206e-01
-1.10598826e+00 -1.94293424e-01 1.32502645e-01 -4.23288196e-02
1.91419587e-01 1.13377190e+00 -3.76595110e-02 -3.26446034e-02
6.53299809e-01 8.83734375e-02 -2.29313448e-01 -5.06865859e-01
5.65532386e-01 4.77449656e-01 2.65037030e-01 -8.03177416e-01
7.99917340e-01 2.63057023e-01 5.19469540e-05 -5.00516653e-01
-4.59898442e-01 -2.09861636e-01 -5.39669752e-01 3.07041942e-03
7.93776095e-01 -8.40139210e-01 -2.93860435e-01 2.10956901e-01
-1.54308307e+00 1.35470927e-01 -7.47231692e-02 1.47890180e-01
-2.83378035e-01 7.56545722e-01 -5.88397026e-01 -4.28707004e-01
-3.60938638e-01 -1.09258521e+00 1.15298605e+00 7.00180531e-02
3.58384073e-04 -1.28657234e+00 1.36710241e-01 3.98109742e-02
6.36682689e-01 6.01088643e-01 1.48279488e+00 -8.15384746e-01
-8.53218913e-01 -1.80528596e-01 -4.15626258e-01 2.12335780e-01
3.80963176e-01 3.23382139e-01 -8.92083287e-01 -4.68232602e-01
-1.74166530e-01 -3.64251174e-02 9.30133641e-01 6.70444444e-02
9.63773966e-01 -5.23701251e-01 -3.51194859e-01 6.64453089e-01
1.93221831e+00 -5.85013032e-02 4.98679847e-01 -7.66237900e-02
1.29592800e+00 4.96572942e-01 -8.66923407e-02 3.75794262e-01
5.77031434e-01 2.04370126e-01 9.33471382e-01 2.24091098e-01
-3.97945583e-01 -6.08783305e-01 7.71602035e-01 1.31728816e+00
9.36658457e-02 -3.63885820e-01 -1.22117686e+00 7.07291126e-01
-1.96928680e+00 -7.67547309e-01 -5.01830995e-01 1.77405643e+00
5.93794823e-01 -2.31452599e-01 -4.39255118e-01 -2.71004617e-01
7.38057554e-01 4.41193312e-01 -3.03490818e-01 -6.31932259e-01
1.06645226e-01 -1.30502686e-01 5.37236631e-01 5.83558142e-01
-6.22579634e-01 7.09503174e-01 5.98414326e+00 5.23686767e-01
-7.70989954e-01 7.81714320e-02 -8.21882114e-02 5.47781169e-01
-9.32150960e-01 6.97449863e-01 -4.62187022e-01 2.66947269e-01
8.82592380e-01 -7.25212812e-01 5.00132143e-01 8.78471076e-01
-3.01715344e-01 7.44204745e-02 -1.47232461e+00 7.73302615e-01
3.79505157e-01 -1.46214306e+00 3.39195341e-01 -1.22573689e-01
7.07076192e-01 1.62138924e-01 -4.85362470e-01 5.85038960e-01
6.14476979e-01 -8.65008473e-01 8.60467255e-01 4.33090776e-01
6.16583467e-01 -3.53377521e-01 7.47522056e-01 4.73211743e-02
-1.52323687e+00 -3.13587874e-01 -2.42744833e-01 3.88867967e-02
-1.76748812e-01 4.32653636e-01 -6.65667593e-01 1.31625926e+00
2.95855135e-01 1.40193963e+00 -1.22791409e+00 8.48528028e-01
-5.01014531e-01 7.69708335e-01 3.12639028e-01 -2.25669257e-02
3.33914220e-01 -3.60439047e-02 7.95743465e-01 1.40172863e+00
4.23599780e-01 -4.83153284e-01 -8.62584449e-03 1.39385951e+00
-4.41430151e-01 1.10861987e-01 -1.06334484e+00 -5.79394102e-01
3.13452870e-01 1.22622526e+00 -6.36919677e-01 -5.49095929e-01
-9.84907031e-01 4.57370877e-01 4.20421720e-01 6.23934269e-01
-7.87833631e-01 -7.99681306e-01 3.89518261e-01 -7.03631565e-02
3.60661745e-01 -2.94051971e-02 -6.15532100e-02 -1.48750293e+00
3.03908020e-01 -8.22380066e-01 5.35263360e-01 -7.36925125e-01
-8.73502791e-01 6.94848597e-01 2.53400773e-01 -1.07145464e+00
-2.03449428e-01 -3.57325733e-01 -8.32830191e-01 6.90883458e-01
-1.50798988e+00 -1.25590992e+00 -3.97815049e-01 3.97231489e-01
3.36890697e-01 -1.24728762e-01 5.67961633e-01 1.15341231e-01
-3.76592457e-01 4.58179742e-01 -1.60305083e-01 4.00686592e-01
1.60531908e-01 -1.52653861e+00 5.71081042e-01 1.40031099e+00
1.17067918e-01 1.00642955e+00 5.98110676e-01 -7.28782117e-01
-1.84986675e+00 -1.15685463e+00 9.13904369e-01 -4.68957394e-01
7.59145856e-01 -5.76175034e-01 -1.11167419e+00 1.16148186e+00
7.94560611e-01 6.85996115e-02 3.45531493e-01 4.94043380e-02
-7.73581982e-01 1.59566104e-01 -8.82065892e-01 2.46423706e-01
1.28248703e+00 -7.60218143e-01 -8.75597239e-01 9.72142965e-02
1.09121692e+00 -3.11548144e-01 -9.24492419e-01 1.43686965e-01
1.20999917e-01 -1.05948782e+00 4.30505037e-01 -6.90279484e-01
8.35325301e-01 -4.93708283e-01 -2.09725350e-01 -1.36306083e+00
-3.45564097e-01 -6.45489931e-01 -3.28957438e-01 1.49604046e+00
2.87357301e-01 -5.60999095e-01 3.33759725e-01 3.11035424e-01
-6.57694578e-01 -3.97026479e-01 -5.60254276e-01 -5.80982685e-01
4.06483002e-02 -2.17991143e-01 8.94619465e-01 1.17143142e+00
4.70546573e-01 4.00022984e-01 -6.33957386e-02 1.54934198e-01
5.48053861e-01 2.61597723e-01 8.01741600e-01 -1.17785311e+00
-2.97909558e-01 -4.27207172e-01 -6.70605123e-01 -6.45044088e-01
5.19921660e-01 -1.57148814e+00 -2.49235138e-01 -2.13339686e+00
3.82762939e-01 2.94138104e-01 -2.31285885e-01 7.45944262e-01
-2.27709934e-01 -8.17218482e-01 2.72885978e-01 -1.81199051e-03
-5.88398218e-01 5.52264810e-01 1.05328429e+00 -7.47727990e-01
1.67648494e-01 -6.14245415e-01 -9.89141583e-01 5.37188232e-01
4.85481232e-01 -6.45385683e-01 -8.80488217e-01 -8.39088202e-01
7.84406543e-01 2.98482001e-01 2.45429620e-01 -6.79793954e-01
3.19062859e-01 1.52821183e-01 -5.06361604e-01 -6.97858855e-02
-5.47634721e-01 -7.90513158e-01 2.00498238e-01 4.90346849e-01
-4.28126782e-01 2.52712816e-01 2.62532085e-01 7.71932900e-01
-4.49109942e-01 -5.10115683e-01 3.54884595e-01 -2.18058497e-01
-8.71823847e-01 1.68755084e-01 -2.65678436e-01 3.70400757e-01
6.69616699e-01 -1.54408544e-01 -8.18879128e-01 -2.73358077e-01
-3.05677444e-01 1.43263891e-01 6.94516242e-01 7.91803539e-01
5.86855352e-01 -1.30455327e+00 -7.62207568e-01 1.97361544e-01
4.96951014e-01 -1.28510192e-01 1.14264034e-01 7.76442409e-01
-6.46013916e-01 2.23103613e-01 -1.17054112e-01 -3.86736512e-01
-1.23910904e+00 6.33558273e-01 4.14217681e-01 -4.64090019e-01
-7.96288669e-01 5.80904961e-01 4.97545421e-01 -4.11282152e-01
8.04327875e-02 -9.65373158e-01 -3.86776328e-01 -1.18750080e-01
3.37141246e-01 1.74036771e-01 4.72104438e-02 -6.55950546e-01
-5.19949734e-01 7.06599355e-01 -7.18481541e-02 5.30564666e-01
1.20199680e+00 2.98323785e-03 -6.69917464e-01 3.64395976e-01
1.40668821e+00 2.20694795e-01 -9.20319557e-01 -2.84290820e-01
2.93643177e-01 -4.68784988e-01 -1.43051580e-01 -5.94727397e-01
-1.27544701e+00 7.74762988e-01 4.94810604e-02 4.01691586e-01
6.88273430e-01 9.40237343e-02 3.91097933e-01 3.53526503e-01
3.77784491e-01 -6.41600251e-01 2.52671707e-02 7.46251404e-01
1.03022695e+00 -9.91466939e-01 1.80576066e-03 -5.53304315e-01
-3.64075720e-01 1.43533838e+00 4.72606003e-01 -2.80256689e-01
6.28550053e-01 2.13495404e-01 -1.84146807e-01 -7.83704281e-01
-9.85094190e-01 -3.73594696e-03 3.04785490e-01 4.61251259e-01
7.74155617e-01 -2.78898388e-01 -3.00118923e-01 5.98890305e-01
5.23829572e-02 -1.55133680e-01 1.06436098e+00 1.25169957e+00
-2.21222192e-01 -1.02576923e+00 3.37546654e-02 7.54384100e-01
-4.48192716e-01 -4.04979020e-01 -5.37746072e-01 8.01271737e-01
-5.81614040e-02 9.99649227e-01 -1.64963350e-01 -4.84181851e-01
2.62025088e-01 6.19259849e-02 4.13569331e-01 -1.07073271e+00
-9.65391815e-01 -5.49541831e-01 1.62817940e-01 -8.30865204e-01
-5.20690739e-01 -3.87318432e-01 -1.40682387e+00 -1.92328095e-01
-3.74900699e-01 3.29125673e-01 5.41167736e-01 6.59897089e-01
7.60720432e-01 9.68732715e-01 2.66157180e-01 -3.45150024e-01
-2.95715600e-01 -9.63847458e-01 -5.89990914e-01 6.68387055e-01
5.75799406e-01 -3.79880339e-01 -4.90845352e-01 1.99948594e-01]
|
[7.540642738342285, 7.918572425842285]
|
53c04fc4-a5c5-4523-a67f-be7b01c02a46
|
what-is-wrong-with-scene-text-recognition
|
1904.01906
| null |
https://arxiv.org/abs/1904.01906v4
|
https://arxiv.org/pdf/1904.01906v4.pdf
|
What Is Wrong With Scene Text Recognition Model Comparisons? Dataset and Model Analysis
|
Many new proposals for scene text recognition (STR) models have been introduced in recent years. While each claim to have pushed the boundary of the technology, a holistic and fair comparison has been largely missing in the field due to the inconsistent choices of training and evaluation datasets. This paper addresses this difficulty with three major contributions. First, we examine the inconsistencies of training and evaluation datasets, and the performance gap results from inconsistencies. Second, we introduce a unified four-stage STR framework that most existing STR models fit into. Using this framework allows for the extensive evaluation of previously proposed STR modules and the discovery of previously unexplored module combinations. Third, we analyze the module-wise contributions to performance in terms of accuracy, speed, and memory demand, under one consistent set of training and evaluation datasets. Such analyses clean up the hindrance on the current comparisons to understand the performance gain of the existing modules.
|
['Seong Joon Oh', 'Sangdoo Yun', 'Junyeop Lee', 'Hwalsuk Lee', 'Jeonghun Baek', 'Geewook Kim', 'Dongyoon Han', 'Sungrae Park']
|
2019-04-03
|
what-is-wrong-with-scene-text-recognition-1
|
http://openaccess.thecvf.com/content_ICCV_2019/html/Baek_What_Is_Wrong_With_Scene_Text_Recognition_Model_Comparisons_Dataset_ICCV_2019_paper.html
|
http://openaccess.thecvf.com/content_ICCV_2019/papers/Baek_What_Is_Wrong_With_Scene_Text_Recognition_Model_Comparisons_Dataset_ICCV_2019_paper.pdf
|
iccv-2019-10
|
['image-matching']
|
['computer-vision']
|
[ 3.45582932e-01 -3.33019763e-01 -6.83179945e-02 -4.48742837e-01
-7.68472314e-01 -6.48727655e-01 8.20224822e-01 1.29497662e-01
-2.42116362e-01 2.70313919e-01 1.84610069e-01 -3.02181363e-01
-4.16952521e-01 -4.50712740e-01 -2.42605448e-01 -4.22155887e-01
2.74961948e-01 3.32376748e-01 3.60778719e-01 -3.04757338e-02
5.74835479e-01 4.70911622e-01 -1.88865447e+00 6.85754657e-01
6.97873116e-01 7.08068192e-01 1.58700451e-01 5.86649597e-01
-4.67568070e-01 8.15364778e-01 -5.02381086e-01 -5.11070788e-01
1.34950578e-01 -2.56219119e-01 -7.78091609e-01 3.46226692e-01
6.87170267e-01 -2.02198833e-01 -3.01027894e-01 5.74182451e-01
5.44553399e-01 -2.34447345e-01 4.79015976e-01 -1.16480577e+00
-9.25730392e-02 4.63625759e-01 -4.16908950e-01 -2.66414993e-02
3.60994160e-01 1.24531895e-01 8.83430064e-01 -1.05814815e+00
6.75619364e-01 1.02495182e+00 8.95608187e-01 2.64772624e-01
-1.13875437e+00 -1.96711943e-01 3.72792874e-03 1.90391779e-01
-1.43527615e+00 -7.27696538e-01 5.84554613e-01 -5.90213001e-01
1.18559945e+00 4.94939506e-01 3.57884049e-01 1.03241348e+00
-1.58624157e-01 9.72808301e-01 1.07023287e+00 -7.91575611e-01
2.61798166e-02 4.02249515e-01 5.17015219e-01 5.55728495e-01
6.15835130e-01 -3.12783837e-01 -7.20992684e-01 7.27133304e-02
3.97082061e-01 -4.47234631e-01 4.17698696e-02 -7.00002849e-01
-8.50794673e-01 4.44171131e-01 -2.73799598e-01 7.01975703e-01
3.15927923e-01 -2.34149039e-01 5.89731216e-01 1.50561929e-01
2.67175108e-01 3.38961273e-01 -4.99088705e-01 -2.19492823e-01
-1.20672333e+00 1.03495874e-01 8.52596879e-01 7.37477601e-01
6.32373571e-01 1.16420969e-01 -2.83682514e-02 8.57501209e-01
2.56254882e-01 1.18397139e-01 3.87994230e-01 -5.17167926e-01
7.07724035e-01 9.01795626e-01 -4.98855598e-02 -9.84171450e-01
-4.66786534e-01 -4.19837534e-01 -4.48185086e-01 5.74509427e-02
5.00081182e-01 1.23134591e-01 -6.00310266e-01 1.19766796e+00
4.45682779e-02 -3.62977386e-01 -1.35768935e-01 3.85985315e-01
8.72066140e-01 1.39269039e-01 3.90445292e-02 6.71312213e-02
1.20287454e+00 -9.03728902e-01 -4.46986437e-01 -2.36128569e-01
8.98664773e-01 -9.91828501e-01 8.22255135e-01 4.82937992e-01
-9.89482284e-01 -6.38659835e-01 -1.21313679e+00 -3.92359942e-02
-5.38432717e-01 5.08310854e-01 6.51453674e-01 1.18860257e+00
-9.60712373e-01 5.28985381e-01 -6.38247430e-01 -7.32334137e-01
3.01889628e-01 3.33414912e-01 -3.11318547e-01 7.93920626e-05
-5.92696488e-01 1.06390309e+00 3.55978638e-01 8.14693049e-02
-2.61257797e-01 -4.53230590e-01 -6.78188324e-01 -8.95062536e-02
3.45940292e-01 -5.66723704e-01 1.30748582e+00 -9.49573755e-01
-1.18348062e+00 9.08927083e-01 -1.91658705e-01 -2.77831018e-01
6.84012353e-01 -3.25156838e-01 -4.43891257e-01 -1.02297485e-01
-2.00862423e-01 4.27048415e-01 5.53498864e-01 -1.10843575e+00
-6.02099180e-01 -4.50375706e-01 -2.45081082e-01 3.95610631e-01
-4.47873950e-01 2.50930011e-01 -6.61587536e-01 -5.80760419e-01
2.11020172e-01 -4.85034913e-01 -1.10901156e-02 -2.18813330e-01
-1.40269339e-01 7.40348250e-02 7.84006059e-01 -4.12116081e-01
1.53527522e+00 -2.31480837e+00 -1.22085800e-02 -5.65553457e-02
-4.94287983e-02 2.37544551e-01 -9.19071808e-02 8.63089681e-01
-1.00452073e-01 2.72987485e-01 -1.16455734e-01 -5.80221236e-01
3.71517763e-02 1.12198107e-01 -4.76740688e-01 4.25219297e-01
1.25273764e-01 7.50225663e-01 -3.39375615e-01 -5.02698004e-01
5.06031692e-01 3.29846233e-01 -2.89808095e-01 -1.54554233e-01
-9.20614973e-02 -4.18593362e-02 -3.41272622e-01 7.52658427e-01
7.58152425e-01 -4.51486595e-02 3.41769427e-01 -1.81725711e-01
-5.92345417e-01 5.72622538e-01 -1.57580256e+00 1.61724043e+00
1.25523880e-01 8.29242468e-01 -1.95804730e-01 -9.18176949e-01
9.22207475e-01 2.56643027e-01 6.11975789e-01 -7.34427035e-01
1.78161532e-01 3.05986941e-01 -2.21402332e-01 -4.95715678e-01
9.36770797e-01 1.53272882e-01 -1.83280310e-04 3.91968399e-01
1.12026669e-01 9.97814611e-02 2.31469050e-01 -5.87843694e-02
1.03749394e+00 2.96972245e-01 3.69448006e-01 -1.66389212e-01
4.86665845e-01 2.50349551e-01 3.00484002e-01 7.73493052e-01
-2.47333944e-01 8.06087434e-01 3.48053426e-01 -5.30617595e-01
-1.18808305e+00 -9.53801274e-01 -4.78881806e-01 8.39861810e-01
-3.79737578e-02 -7.03728914e-01 -8.03344369e-01 -5.27783513e-01
-1.55619964e-01 4.33662742e-01 -3.76692593e-01 2.13335395e-01
-5.47553480e-01 -1.07328570e+00 8.42809796e-01 4.79275674e-01
6.66839659e-01 -6.95949376e-01 -9.53733325e-01 9.13572218e-03
-1.15259781e-01 -1.24834156e+00 3.03561747e-01 1.64224476e-01
-1.28113222e+00 -1.11106443e+00 -1.92841798e-01 -6.07426524e-01
2.92351484e-01 6.06650531e-01 1.05482435e+00 -8.27906560e-03
-4.32975173e-01 6.12301230e-01 -5.14644027e-01 -2.25405708e-01
-2.13245139e-01 2.14661062e-01 -2.14413956e-01 -1.66558355e-01
6.03768647e-01 -1.64823085e-01 -1.64296195e-01 4.06874776e-01
-8.91219735e-01 3.18354368e-01 5.79788685e-01 5.33463061e-01
2.21602842e-01 5.27764484e-02 3.19891989e-01 -8.64975393e-01
4.30256218e-01 -2.84206867e-01 -5.87379813e-01 4.87652242e-01
-8.37782621e-01 1.34787545e-01 1.41519397e-01 -6.20674752e-02
-1.13587379e+00 1.47826329e-01 -4.94863726e-02 1.78315967e-01
-5.25135040e-01 5.47263741e-01 -2.25647464e-01 -7.58931935e-02
6.42643929e-01 3.03430617e-01 -2.69450754e-01 -9.04726744e-01
1.99466363e-01 6.58460975e-01 3.98590863e-01 -6.07203841e-01
6.26044869e-01 4.17940617e-01 -2.87931710e-01 -1.19190395e+00
-4.87153322e-01 -6.53018951e-01 -9.69744205e-01 -1.45819277e-01
5.16503811e-01 -8.10754120e-01 -1.59399688e-01 6.91848218e-01
-9.96644735e-01 -8.61200690e-02 -2.46937603e-01 2.82895207e-01
-4.49996948e-01 9.05539453e-01 -3.12017739e-01 -9.10370708e-01
-2.72712171e-01 -1.05052662e+00 8.42966974e-01 5.10995127e-02
-3.23340595e-01 -8.13985527e-01 2.27535501e-01 5.72961211e-01
3.62171948e-01 -2.42360439e-02 7.31390238e-01 -5.25735021e-01
-6.74877763e-01 -3.78659219e-01 -3.03697109e-01 1.48189470e-01
-1.80027112e-02 3.46197397e-01 -1.36334217e+00 -3.02377433e-01
-7.38653019e-02 -4.38767001e-02 8.26959908e-01 1.15241937e-01
8.32578361e-01 7.21851438e-02 -3.47227752e-01 4.72247809e-01
1.51887274e+00 7.83986375e-02 9.26873922e-01 6.74389839e-01
4.06592816e-01 7.32136846e-01 3.26015860e-01 3.99203986e-01
3.19102347e-01 8.47192287e-01 8.92006531e-02 4.57421504e-02
-2.98278332e-01 -2.41013780e-01 5.52070260e-01 9.20978129e-01
1.70302466e-01 -2.20833585e-01 -1.09770751e+00 5.40745318e-01
-1.85195518e+00 -8.55205894e-01 -3.94670606e-01 2.37950134e+00
2.66259223e-01 2.36704975e-01 2.87466615e-01 5.20182610e-01
5.63609123e-01 7.02460408e-02 -9.75602269e-02 -3.52263451e-01
-5.57794333e-01 1.13012105e-01 3.05445701e-01 3.07539463e-01
-1.09750831e+00 7.26103365e-01 7.68739271e+00 7.50587881e-01
-9.84794438e-01 -1.83497742e-01 4.00101095e-01 -6.07138209e-04
-1.48678496e-01 2.43850961e-01 -9.33462381e-01 1.22421861e-01
8.97572637e-01 -1.03811421e-01 5.36292195e-02 7.55099416e-01
8.10222179e-02 -3.82050574e-01 -1.24819994e+00 1.02706432e+00
4.33518887e-01 -1.28312898e+00 1.29852280e-01 -1.02546163e-01
5.43497622e-01 2.03692168e-01 8.75615031e-02 2.07623243e-01
-1.11517712e-01 -9.15933371e-01 9.15243089e-01 3.87142986e-01
6.35752857e-01 -3.36599827e-01 5.66541076e-01 2.74610341e-01
-1.20904052e+00 -8.29486549e-02 -1.78435221e-01 -2.46786967e-01
-3.61726642e-01 3.90105486e-01 -8.72908354e-01 8.77745569e-01
5.98814905e-01 4.59001213e-01 -1.16533387e+00 1.25616026e+00
5.72268292e-02 5.63551307e-01 -2.27826729e-01 -6.21547969e-03
4.02856581e-02 -7.96308964e-02 4.67714429e-01 1.68971264e+00
1.85698956e-01 -5.09511530e-01 -7.10575655e-02 6.05051696e-01
2.42326483e-01 3.28211904e-01 -8.39526892e-01 1.01066470e-01
4.33213264e-01 1.14541316e+00 -7.11519957e-01 -2.15748638e-01
-7.13480473e-01 7.48989940e-01 2.75975555e-01 2.47811168e-01
-5.59128344e-01 -5.76513968e-02 3.51266682e-01 1.74468413e-01
1.93923131e-01 -4.16970134e-01 -9.58349943e-01 -1.25800002e+00
4.45341647e-01 -8.98280144e-01 6.51964068e-01 -7.11208940e-01
-9.09073710e-01 2.85437137e-01 1.15709662e-01 -1.18289506e+00
-1.29857343e-02 -7.36353755e-01 -3.99544597e-01 7.31016636e-01
-1.28782523e+00 -1.09816992e+00 -2.03780040e-01 1.83739975e-01
7.57560909e-01 -1.80670381e-01 7.88735449e-01 4.43609148e-01
-8.25233281e-01 8.32108319e-01 3.55466187e-01 7.24613369e-02
7.23554969e-01 -8.88318896e-01 5.43946803e-01 9.98826325e-01
3.72944206e-01 5.04697919e-01 5.26759505e-01 -3.37165356e-01
-1.47878456e+00 -6.55142009e-01 1.14496684e+00 -9.26148117e-01
5.06171107e-01 -3.88776362e-01 -9.25111830e-01 4.17140663e-01
-1.63354725e-02 -5.05086064e-01 6.30119741e-01 2.87215590e-01
-7.27935433e-01 -9.17503331e-03 -9.51886475e-01 6.73178613e-01
7.71626770e-01 -6.56483650e-01 -6.33936584e-01 -2.43245080e-01
3.12044136e-02 -1.47904664e-01 -5.42579055e-01 5.45440316e-01
8.06098938e-01 -1.25811327e+00 8.01186681e-01 -4.46346164e-01
2.35214725e-01 -3.32896888e-01 -4.51361239e-01 -4.55713242e-01
-2.18647301e-01 -3.02652627e-01 1.21642455e-01 1.35696423e+00
4.78513837e-01 -3.80211651e-01 8.46194983e-01 6.21243834e-01
-2.32044071e-01 -7.04705775e-01 -8.40009570e-01 -7.34115779e-01
6.69573098e-02 -8.31453025e-01 3.28772187e-01 1.00538838e+00
2.13826343e-01 4.74559605e-01 -2.35690832e-01 -1.57263339e-01
4.81145233e-01 1.34316355e-01 1.05983937e+00 -1.09717917e+00
-3.07078391e-01 -7.93216884e-01 -5.14763117e-01 -9.84678924e-01
-5.75165093e-01 -8.84330392e-01 -3.07837218e-01 -1.66985571e+00
6.35941088e-01 -3.60307992e-01 -9.63103175e-02 4.31724638e-01
1.09123345e-02 6.49530068e-02 3.89256030e-01 4.31476206e-01
-7.87751496e-01 1.88281700e-01 6.17571771e-01 7.17836097e-02
-6.73675016e-02 -2.96047181e-01 -6.88558221e-01 8.60603333e-01
9.47582841e-01 -1.43168107e-01 -2.48369858e-01 -7.43943453e-01
3.85289043e-01 -3.44201565e-01 3.57013881e-01 -1.23704565e+00
3.09458017e-01 -1.30330061e-03 4.42921281e-01 -8.64764094e-01
1.80094376e-01 -7.18947291e-01 3.13606799e-01 2.61311740e-01
-1.09041959e-01 1.22084662e-01 5.44945061e-01 2.09918007e-01
-1.26602665e-01 -3.73114228e-01 7.07536936e-01 9.55262110e-02
-1.03186369e+00 -2.86002845e-01 -4.37775940e-01 -1.71770096e-01
7.59150684e-01 -9.72856820e-01 -4.76180017e-01 1.65545523e-01
-3.13511521e-01 -1.15797140e-01 5.57693243e-01 6.05028093e-01
4.24060136e-01 -8.74232233e-01 -6.18143201e-01 2.32493132e-01
2.82619685e-01 -4.13262248e-01 2.94670701e-01 8.27852011e-01
-4.48137283e-01 6.34249985e-01 -2.65643001e-01 -6.17576957e-01
-1.51280487e+00 4.12693620e-01 3.81488025e-01 -4.69791085e-01
-5.22984743e-01 4.14480597e-01 1.01675084e-02 -3.58323336e-01
4.26560968e-01 -9.29223150e-02 -1.73561350e-01 1.24843851e-01
3.38503242e-01 6.58123791e-01 4.49800909e-01 -5.52749693e-01
-3.66630822e-01 6.45398438e-01 -9.92147997e-02 -1.56972587e-01
1.11497426e+00 -1.79050148e-01 3.55846360e-02 6.01495385e-01
7.31429040e-01 -6.95714056e-02 -6.82361245e-01 -1.24646634e-01
4.62066323e-01 -4.36130911e-01 -1.80235624e-01 -8.47094536e-01
-4.58972961e-01 9.45293367e-01 6.84755266e-01 6.64582327e-02
1.08825457e+00 -2.91941106e-01 2.14479700e-01 4.17810053e-01
2.43488714e-01 -1.29353404e+00 -1.02000833e-01 7.26469398e-01
5.42474806e-01 -9.46133912e-01 3.31033915e-01 -5.11206567e-01
-3.72916907e-01 1.11542070e+00 4.67580259e-01 3.64138752e-01
2.22361714e-01 4.41940844e-01 -8.26800168e-02 -1.27252415e-01
-8.47798347e-01 -1.70697525e-01 1.45446986e-01 4.63865995e-01
8.16884100e-01 -9.23433527e-02 -6.34430587e-01 5.04891753e-01
-5.26376255e-02 1.74431168e-02 3.73670280e-01 1.05555809e+00
-6.28852308e-01 -1.44064021e+00 -4.78610814e-01 2.98348129e-01
-3.15950304e-01 1.01783248e-02 -8.17977905e-01 9.61632431e-01
1.56367317e-01 9.40597475e-01 -1.72017485e-01 -5.76237321e-01
5.18498719e-01 4.04115140e-01 7.01097071e-01 -5.03502667e-01
-8.06347847e-01 7.36124963e-02 3.37017775e-01 -1.97498411e-01
-2.09330410e-01 -8.48706722e-01 -7.89707243e-01 -3.17799091e-01
-4.13689256e-01 -4.36845347e-02 8.03603947e-01 9.18371677e-01
4.10545677e-01 2.61014521e-01 4.16158199e-01 -5.45492887e-01
-6.17727995e-01 -7.84271061e-01 -3.69314373e-01 3.50353837e-01
-9.47335586e-02 -4.50039834e-01 -1.10089555e-01 1.70360267e-01]
|
[11.757062911987305, 2.4683685302734375]
|
a13ff1b9-2c8d-43cb-9f63-d8c8414d8292
|
high-fidelity-image-compression-with-score
|
2305.18231
| null |
https://arxiv.org/abs/2305.18231v1
|
https://arxiv.org/pdf/2305.18231v1.pdf
|
High-Fidelity Image Compression with Score-based Generative Models
|
Despite the tremendous success of diffusion generative models in text-to-image generation, replicating this success in the domain of image compression has proven difficult. In this paper, we demonstrate that diffusion can significantly improve perceptual quality at a given bit-rate, outperforming state-of-the-art approaches PO-ELIC and HiFiC as measured by FID score. This is achieved using a simple but theoretically motivated two-stage approach combining an autoencoder targeting MSE followed by a further score-based decoder. However, as we will show, implementation details matter and the optimal design decisions can differ greatly from typical text-to-image models.
|
['Lucas Theis', 'George Toderici', 'Luca Versari', 'Fabian Mentzer', 'Eirikur Agustsson', 'Emiel Hoogeboom']
|
2023-05-26
| null | null | null | null |
['image-compression']
|
['computer-vision']
|
[ 7.91500032e-01 2.49014050e-01 7.08528981e-02 -6.66345209e-02
-7.58884490e-01 -2.21941322e-01 7.91190386e-01 -2.05131978e-01
-1.76539317e-01 6.76544309e-01 3.46549779e-01 -5.05082428e-01
-1.90467075e-01 -6.19602680e-01 -6.20352566e-01 -8.33817899e-01
-1.29819751e-01 1.63786173e-01 7.24292323e-02 -7.90965632e-02
4.79695648e-01 2.05328956e-01 -1.37396288e+00 2.68829167e-01
6.01634741e-01 9.77367580e-01 3.98943752e-01 1.32114136e+00
9.61351395e-02 1.15191185e+00 -8.16582978e-01 -6.75610602e-01
2.10148603e-01 -1.14754760e+00 -4.81629580e-01 2.81531185e-01
2.43686527e-01 -5.82811117e-01 -5.50291181e-01 9.78325963e-01
7.96490788e-01 -2.51147628e-01 8.67401361e-01 -8.73052180e-01
-1.07023895e+00 5.71094334e-01 -4.18073535e-01 4.78292033e-02
2.99373895e-01 2.55052269e-01 9.04022574e-01 -6.70093775e-01
7.68029690e-01 1.18107986e+00 5.49324274e-01 6.10917270e-01
-1.45584536e+00 -3.64865303e-01 -5.96292675e-01 1.72110081e-01
-1.34700668e+00 -7.22693026e-01 6.38320446e-01 -2.44972929e-01
9.87168074e-01 1.36064500e-01 5.65171301e-01 9.49243605e-01
6.26828849e-01 6.26299620e-01 1.16048181e+00 -6.87805057e-01
3.54695529e-01 -4.19933274e-02 -8.05739522e-01 4.75679487e-01
1.42355144e-01 2.40486220e-01 -4.95930493e-01 1.60777599e-01
8.29839051e-01 -4.63731915e-01 -2.71952391e-01 -1.62177369e-01
-9.83793318e-01 8.36142540e-01 2.74730712e-01 3.60419989e-01
-5.87487698e-01 6.79772854e-01 1.42682418e-01 5.52979887e-01
3.51185054e-01 3.45973492e-01 9.23657119e-02 -4.55857605e-01
-1.73871064e+00 1.98809400e-01 7.80279219e-01 9.98826742e-01
2.48994619e-01 5.79933047e-01 -2.79211223e-01 7.58622229e-01
4.52495217e-01 5.30573487e-01 5.90177000e-01 -1.30797148e+00
2.02610373e-01 -2.87127018e-01 -1.47912160e-01 -8.91489923e-01
1.32744685e-01 -4.94690120e-01 -8.75157356e-01 4.91039693e-01
-2.08352804e-02 -3.18341821e-01 -1.05723357e+00 1.38613141e+00
-2.40382552e-01 -1.64226800e-01 1.65477902e-01 6.50665462e-01
3.03271472e-01 9.90680754e-01 -1.67344198e-01 -3.83818060e-01
1.00810480e+00 -9.15534019e-01 -9.45605934e-01 -1.93488032e-01
1.53775781e-01 -1.07496083e+00 5.85942209e-01 5.70155799e-01
-1.74394023e+00 -5.98857880e-01 -1.45459735e+00 -6.11923635e-02
1.11548856e-01 1.30087966e-02 4.15081263e-01 1.12130618e+00
-1.47205329e+00 7.76314437e-01 -5.02997160e-01 -2.79701740e-01
4.83853221e-01 1.60530567e-01 9.82512236e-02 -7.56310374e-02
-9.03555393e-01 8.44366074e-01 3.57235134e-01 -3.10789794e-01
-1.04359841e+00 -5.44474542e-01 -6.34005129e-01 2.00101152e-01
8.90775248e-02 -9.36470687e-01 1.39273751e+00 -8.55559826e-01
-1.90877783e+00 4.56360787e-01 -1.73202485e-01 -9.70301449e-01
7.14978218e-01 -1.07337236e-01 -3.82931679e-01 5.04814625e-01
-2.70795554e-01 1.03397989e+00 1.33591378e+00 -1.38319933e+00
-5.50500274e-01 2.97102779e-01 -1.33315891e-01 1.16951562e-01
-2.74969429e-01 -5.34182563e-02 -5.41385114e-01 -9.18352604e-01
-1.52240455e-01 -7.90171921e-01 -3.11546266e-01 8.54708627e-02
-2.29465783e-01 4.38744694e-01 8.90738189e-01 -8.29190135e-01
1.47176063e+00 -2.07176042e+00 6.04869910e-02 -3.06508504e-02
1.82487965e-01 3.08587998e-01 -1.68086797e-01 7.04661727e-01
-3.86685096e-02 2.50566989e-01 -3.90495479e-01 -4.61520791e-01
8.15886706e-02 -1.98474824e-02 -2.99586952e-01 3.17494839e-01
3.34286422e-01 9.53192830e-01 -6.17995381e-01 -5.28120041e-01
3.67341250e-01 6.72858119e-01 -8.50004792e-01 1.68220282e-01
-1.42319351e-01 4.99069318e-02 1.51977669e-02 3.65608454e-01
4.85542238e-01 -2.86445081e-01 1.31804064e-01 4.83821705e-02
6.92220032e-02 1.39998719e-01 -8.70268762e-01 1.59278214e+00
-5.00755429e-01 1.14411986e+00 2.36734245e-02 -6.22091830e-01
6.72773123e-01 3.83329809e-01 3.81685734e-01 -7.86304474e-01
8.95291865e-02 2.58345991e-01 2.26981118e-01 -1.66772947e-01
8.61828625e-01 -3.81071389e-01 3.62666190e-01 4.19118077e-01
2.75548369e-01 -6.51002944e-01 3.99360865e-01 3.60917419e-01
1.08988488e+00 8.21025297e-02 2.11434305e-01 -1.57954730e-02
1.99184373e-01 -1.11953065e-01 -1.11359432e-01 8.52469385e-01
-3.85875590e-02 1.01224709e+00 4.12235737e-01 3.03086728e-01
-1.78038466e+00 -9.49707627e-01 -2.25541070e-01 4.43431199e-01
-4.96512465e-02 -6.59432054e-01 -1.15130055e+00 -1.69398651e-01
-3.71187091e-01 9.76821125e-01 -4.13173139e-01 -8.39134678e-02
-3.42333138e-01 -7.95552194e-01 7.15145707e-01 1.15665741e-01
5.29244781e-01 -1.00042844e+00 -7.36246347e-01 4.72748518e-01
1.02772467e-01 -1.19696701e+00 -4.19885129e-01 1.88635364e-01
-8.39110136e-01 -3.44844937e-01 -1.34513557e+00 -4.63870734e-01
4.78366613e-01 1.68001264e-01 1.11075437e+00 4.26119193e-02
-2.96670318e-01 4.10352379e-01 -4.25028950e-01 -4.03383166e-01
-1.05177295e+00 -1.70524120e-01 -4.74938780e-01 -1.26831412e-01
-9.94290113e-02 -6.37340128e-01 -9.88868594e-01 -9.54528525e-02
-1.37870467e+00 2.81051219e-01 9.69747424e-01 7.27902114e-01
4.46638018e-01 3.22781473e-01 3.85391027e-01 -5.04364014e-01
9.84122396e-01 -3.47384632e-01 -3.50011975e-01 -1.38872951e-01
-1.10921824e+00 5.56377321e-02 6.37533784e-01 -2.31777132e-01
-8.50362718e-01 -1.29506215e-01 -4.50707674e-01 -2.42811948e-01
3.28129567e-02 4.11064804e-01 1.99612603e-01 -4.84100692e-02
6.55160487e-01 6.45518720e-01 8.05109367e-03 -9.92733166e-02
6.16258979e-01 7.48484313e-01 5.89789629e-01 -1.44323753e-02
9.24619615e-01 3.18009675e-01 1.97968155e-01 -8.40277314e-01
-1.43045157e-01 8.83605406e-02 -1.69024512e-01 -2.72013098e-01
9.32375371e-01 -9.60633576e-01 -2.19994202e-01 4.17563170e-01
-1.04878592e+00 -4.30391818e-01 -6.80622816e-01 2.29225308e-01
-1.07694340e+00 4.54614133e-01 -7.64912903e-01 -8.19083452e-01
-3.60490978e-01 -1.17183089e+00 9.17778969e-01 1.54330716e-01
-5.64746745e-02 -1.06671596e+00 -1.14411943e-01 1.10098347e-01
9.07803118e-01 -9.66514871e-02 6.99962556e-01 -3.24195921e-02
-7.39569187e-01 -1.99283078e-01 -2.88275599e-01 6.81362987e-01
-1.67764306e-01 2.70163771e-02 -8.41417968e-01 -3.34495425e-01
2.36406431e-01 -2.05140233e-01 9.13938642e-01 6.94481730e-01
9.15571451e-01 -3.35092157e-01 1.72948390e-02 6.51401699e-01
1.63268864e+00 2.85605490e-01 1.23426116e+00 2.34200791e-01
2.17384100e-01 1.67225569e-01 8.39551389e-02 5.25262952e-01
1.53402850e-01 6.07722223e-01 1.46439388e-01 -1.45310208e-01
-9.16192353e-01 -4.13666725e-01 4.72395539e-01 1.04549396e+00
8.72710422e-02 -7.13236749e-01 -5.02159417e-01 5.96827626e-01
-1.40941823e+00 -1.18531585e+00 -1.65025312e-02 1.97412395e+00
9.79449332e-01 3.13108027e-01 -1.15759648e-01 4.55795944e-01
4.13383186e-01 3.94100189e-01 -2.68354148e-01 -6.53545558e-01
-3.16076964e-01 2.32920021e-01 8.47705007e-01 5.41949809e-01
-6.02983057e-01 6.44420326e-01 8.12367630e+00 1.21619189e+00
-9.88420010e-01 1.71011105e-01 8.64043057e-01 -3.78963398e-03
-4.54963058e-01 8.83062929e-02 -4.88533676e-01 4.90638196e-01
1.56403875e+00 -3.69633138e-01 6.42797828e-01 6.25714958e-01
2.35256061e-01 -2.40242824e-01 -7.99692929e-01 1.03680634e+00
3.63261700e-01 -1.38331115e+00 2.50706315e-01 2.84293175e-01
8.80283296e-01 -1.86249077e-01 3.87537092e-01 -6.63499609e-02
1.83998406e-01 -1.06795573e+00 1.07063532e+00 5.01824975e-01
1.02750206e+00 -5.56384146e-01 3.07824165e-01 1.94630653e-01
-7.59701192e-01 -1.12953052e-01 -5.18822253e-01 1.87958702e-01
5.64520836e-01 8.29042435e-01 -6.80794477e-01 4.89945114e-01
3.43358546e-01 4.14078265e-01 -4.72848594e-01 1.19637346e+00
-1.75366610e-01 9.46689129e-01 -1.35266319e-01 7.17474371e-02
2.48460278e-01 8.08822215e-02 4.89364892e-01 1.51570857e+00
9.01006639e-01 -7.31210262e-02 -5.29112339e-01 1.01203346e+00
-8.08259547e-02 -2.40201391e-02 -5.60452104e-01 -2.62450457e-01
3.28947641e-02 8.46113265e-01 -6.15069926e-01 -4.79422748e-01
-1.67253822e-01 1.43518424e+00 -3.45409393e-01 4.63798344e-01
-8.65406036e-01 -5.63700974e-01 1.47521734e-01 2.69471109e-01
8.25445592e-01 -4.69125956e-01 -3.68655741e-01 -9.18546915e-01
-1.34692565e-01 -9.59943771e-01 -2.94974208e-01 -1.10572386e+00
-7.91132271e-01 5.08348942e-01 9.45524406e-03 -1.10944939e+00
-7.92382956e-01 -2.67710626e-01 -2.12411776e-01 9.01368499e-01
-1.57239068e+00 -8.56249154e-01 -5.40204123e-02 2.86356777e-01
7.75787830e-01 -2.48480812e-01 7.09086180e-01 2.72083819e-01
-9.23373154e-04 6.45537198e-01 4.70660180e-01 -1.90409526e-01
3.84529173e-01 -1.14241624e+00 7.94429302e-01 1.16574275e+00
3.61276954e-01 2.88658619e-01 1.07474303e+00 -5.78174770e-01
-1.62126386e+00 -7.54512131e-01 7.82413423e-01 6.76632673e-03
4.73355830e-01 -1.79663762e-01 -4.65492785e-01 3.11090380e-01
8.29523683e-01 -3.90277863e-01 3.81083101e-01 -6.12976611e-01
-1.77891715e-03 1.88395768e-01 -1.08518469e+00 9.42618430e-01
9.92870867e-01 -4.80435997e-01 -1.81768879e-01 1.26185730e-01
6.14068270e-01 -3.62163574e-01 -9.05344009e-01 1.21833473e-01
4.22607541e-01 -1.27986217e+00 9.71275687e-01 7.42336288e-02
1.00149226e+00 -2.19871923e-01 -3.28751206e-01 -1.36466193e+00
-4.44903702e-01 -1.16009009e+00 -4.02654201e-01 1.02914882e+00
4.05842394e-01 -1.25924423e-01 7.10034132e-01 2.14018121e-01
-2.17835885e-02 -5.44387758e-01 -8.53000104e-01 -8.47801924e-01
1.29186019e-01 -6.53086901e-01 3.10113817e-01 1.44980371e-01
-1.26096219e-01 3.81732404e-01 -7.96531379e-01 -2.79528201e-01
7.26162374e-01 -3.40583920e-01 6.80985272e-01 -5.10725141e-01
-7.27165580e-01 -8.59948516e-01 -5.33072174e-01 -1.61260605e+00
-5.28081238e-01 -5.51158905e-01 1.55799493e-01 -1.55528164e+00
1.45916328e-01 -1.17699862e-01 7.65554681e-02 -1.43070489e-01
-1.21527798e-01 5.40389538e-01 5.97536743e-01 2.46648118e-01
-3.59854609e-01 6.50784075e-01 1.36479402e+00 -1.68128148e-01
1.44761264e-01 -4.55451846e-01 -7.26855338e-01 1.85297459e-01
6.72793269e-01 -5.02478480e-01 -6.75035000e-01 -3.90333205e-01
2.56591082e-01 2.97670215e-01 2.15535522e-01 -1.58636272e+00
2.46948630e-01 1.10145159e-01 4.30058718e-01 -3.36577415e-01
4.52344567e-01 -7.56142676e-01 3.92804652e-01 6.59991503e-01
-4.22861427e-01 9.96717997e-03 1.04400165e-01 6.53980792e-01
-2.71342427e-01 -5.53556025e-01 8.09110582e-01 5.03943041e-02
-3.64759713e-01 -8.19086283e-03 -5.77490866e-01 -2.42480621e-01
9.23161328e-01 -4.01999265e-01 -2.84066886e-01 -1.15006459e+00
-3.63540739e-01 -5.50992310e-01 6.80467248e-01 1.97989941e-01
8.65853667e-01 -1.19468403e+00 -1.03284299e+00 2.08592460e-01
-2.82301575e-01 -5.75025856e-01 2.30635077e-01 6.44469500e-01
-7.54262388e-01 4.92239028e-01 -1.59569830e-01 -4.48861688e-01
-1.02126968e+00 4.96411413e-01 2.78834879e-01 -2.55493283e-01
-6.29786193e-01 6.46310091e-01 -5.21471836e-02 6.05604887e-01
2.13220032e-04 1.31699860e-01 2.91854560e-01 -3.52559716e-01
5.75981438e-01 1.50676459e-01 -2.62832046e-02 -5.46592832e-01
2.73267776e-01 3.84697944e-01 7.62478188e-02 -6.80466771e-01
1.15400004e+00 -4.28892940e-01 2.20738158e-01 1.53099865e-01
1.10089052e+00 -9.21262428e-03 -1.45883393e+00 1.60773784e-01
-4.22646910e-01 -5.97867787e-01 4.09828365e-01 -8.31490099e-01
-1.11142802e+00 9.93856370e-01 8.93700659e-01 4.73037541e-01
1.41590595e+00 -2.75458306e-01 9.79153752e-01 -5.11942990e-02
2.30576694e-01 -1.02909040e+00 1.76644474e-01 2.44948879e-01
9.85815167e-01 -8.08157980e-01 2.56923854e-01 -9.93263721e-02
-5.25554001e-01 1.00928712e+00 -2.96405673e-01 -8.04706737e-02
6.12060606e-01 5.42912066e-01 -6.98890984e-02 1.51062340e-01
-9.72335696e-01 -4.14051823e-02 1.56641856e-01 8.26863587e-01
6.39905274e-01 -2.03650981e-01 -5.54426193e-01 -1.39311627e-01
-3.70372176e-01 2.70231545e-01 8.04492891e-01 9.53061283e-01
-4.87091482e-01 -1.23804533e+00 -4.62709576e-01 4.19762105e-01
-8.60506892e-01 -5.07409692e-01 -2.37043179e-03 3.82575631e-01
-1.69161811e-01 1.11897397e+00 -1.06135324e-01 -4.77476031e-01
-8.66817161e-02 -1.04212880e-01 6.40173018e-01 -1.19634211e-01
-4.42227513e-01 3.88985515e-01 3.54697593e-02 -2.93988913e-01
-5.07290244e-01 -6.61989093e-01 -8.19704950e-01 -7.23476648e-01
-2.58450270e-01 -1.96271598e-01 1.01181197e+00 5.62749207e-01
5.03108203e-01 7.69130170e-01 5.36791503e-01 -9.09528434e-01
-5.63698113e-01 -1.03750610e+00 -5.29860139e-01 2.74667293e-01
3.22916955e-01 -5.30146696e-02 -2.67676592e-01 5.94668567e-01]
|
[11.293355941772461, -0.9024261236190796]
|
5724074f-7b66-4ecf-8e60-8d472409e766
|
unsupervised-counselor-dialogue-clustering
| null | null |
https://aclanthology.org/W18-5017
|
https://aclanthology.org/W18-5017.pdf
|
Unsupervised Counselor Dialogue Clustering for Positive Emotion Elicitation in Neural Dialogue System
|
Positive emotion elicitation seeks to improve user{'}s emotional state through dialogue system interaction, where a chat-based scenario is layered with an implicit goal to address user{'}s emotional needs. Standard neural dialogue system approaches still fall short in this situation as they tend to generate only short, generic responses. Learning from expert actions is critical, as these potentially differ from standard dialogue acts. In this paper, we propose using a hierarchical neural network for response generation that is conditioned on 1) expert{'}s action, 2) dialogue context, and 3) user emotion, encoded from user input. We construct a corpus of interactions between a counselor and 30 participants following a negative emotional exposure to learn expert actions and responses in a positive emotion elicitation scenario. Instead of relying on the expensive, labor intensive, and often ambiguous human annotations, we unsupervisedly cluster the expert{'}s responses and use the resulting labels to train the network. Our experiments and evaluation show that the proposed approach yields lower perplexity and generates a larger variety of responses.
|
['Koichiro Yoshino', 'Satoshi Nakamura', 'Sakriani Sakti', 'Nurul Lubis']
|
2018-07-01
| null | null | null |
ws-2018-7
|
['goal-oriented-dialogue-systems']
|
['natural-language-processing']
|
[ 5.97625136e-01 7.51542628e-01 1.45485491e-01 -8.23355317e-01
-6.88627958e-01 -4.66569185e-01 4.88327026e-01 1.26380980e-01
-5.11759758e-01 9.81361151e-01 4.40722883e-01 -1.59174368e-01
6.67161271e-02 -4.91608441e-01 3.43552977e-01 -4.78635103e-01
3.52815151e-01 7.33964086e-01 -3.03764880e-01 -6.49324596e-01
3.34810138e-01 1.81838617e-01 -1.09547746e+00 6.39770210e-01
8.78723264e-01 7.21970260e-01 -6.74584731e-02 1.10582173e+00
-1.92336291e-01 1.37592840e+00 -9.33796167e-01 -5.76532483e-01
-3.44287395e-01 -1.02743471e+00 -1.50829601e+00 1.52223662e-01
-3.82705063e-01 -4.55589056e-01 1.16979755e-01 8.03000033e-01
7.63403475e-01 5.10651708e-01 6.86517477e-01 -1.00699675e+00
-3.08860898e-01 6.63316607e-01 2.55280286e-02 -2.18863532e-01
9.01372313e-01 2.69200891e-01 9.97707784e-01 -5.92112124e-01
7.06910789e-01 1.22301173e+00 5.01428604e-01 1.43328941e+00
-1.32112753e+00 -2.47124150e-01 -5.40930666e-02 -1.06461547e-01
-6.74362421e-01 -5.91022551e-01 1.00817549e+00 -5.08902609e-01
9.85224724e-01 4.35935020e-01 5.69222808e-01 1.46948540e+00
-3.56685400e-01 8.99342477e-01 1.07523966e+00 -4.90392655e-01
3.45329404e-01 6.96141720e-01 2.93296486e-01 3.29744130e-01
-9.64080095e-01 -3.00177634e-01 -2.99517184e-01 -3.96272540e-01
3.41420114e-01 -5.46131372e-01 -2.55765051e-01 2.24923119e-01
-7.72629380e-01 1.12561560e+00 -7.69508928e-02 4.58180904e-01
-7.24925458e-01 -3.12700152e-01 7.24810958e-01 5.08028269e-01
3.96915942e-01 9.95135188e-01 -2.36616537e-01 -5.76169848e-01
-4.78385180e-01 2.39685968e-01 1.48004913e+00 5.29404581e-01
5.08910656e-01 1.50908902e-02 -4.30020809e-01 1.36588359e+00
1.68897454e-02 -1.16246544e-01 3.98282707e-01 -1.12190819e+00
1.53836936e-01 7.13673651e-01 4.72025186e-01 -8.74874830e-01
-6.10304177e-01 1.85591131e-01 -8.55343759e-01 5.21071292e-02
3.55298907e-01 -9.81644273e-01 -5.28947413e-01 1.85370040e+00
4.15928841e-01 -5.81937969e-01 5.23732364e-01 9.05507922e-01
9.90426004e-01 6.27630770e-01 2.62015522e-01 -5.06767929e-01
1.07015908e+00 -7.01611698e-01 -9.22837436e-01 -1.49139404e-01
8.67383599e-01 -6.45878017e-01 9.91146803e-01 5.12095451e-01
-1.34127283e+00 -2.40707979e-01 -4.88660723e-01 7.16982558e-02
-7.92738274e-02 2.26745635e-01 5.99630713e-01 6.36834264e-01
-1.11319137e+00 3.79840314e-01 -1.16202436e-01 -4.43939418e-01
-1.18105687e-01 6.01048768e-01 -6.27200902e-02 4.72819507e-01
-1.62868381e+00 1.10915756e+00 3.39811534e-01 2.72507161e-01
-3.87957841e-01 -2.59277493e-01 -8.15910161e-01 3.56711410e-02
3.43470573e-01 -3.83787334e-01 1.71007395e+00 -1.36541009e+00
-2.22977257e+00 7.74768472e-01 1.76720127e-01 -6.60286695e-02
4.79781240e-01 9.23855826e-02 -1.68395370e-01 2.93959707e-01
-3.40751797e-01 7.80165732e-01 2.83379614e-01 -1.13991570e+00
-4.75408077e-01 -1.00254416e-01 3.56406063e-01 5.43637395e-01
-3.06810111e-01 4.36041832e-01 -3.46049182e-02 -2.27625653e-01
-4.22902226e-01 -8.67011726e-01 -5.79881191e-01 -6.73620880e-01
-4.40746516e-01 -6.59304738e-01 3.96868050e-01 -6.60696328e-01
1.15173960e+00 -1.76541638e+00 2.50774711e-01 3.47465962e-01
2.99005896e-01 3.50301623e-01 3.88276875e-02 7.30374455e-01
-1.76460817e-01 1.06194578e-01 -1.56557977e-01 -2.35968649e-01
3.16385597e-01 9.47411358e-02 1.48658037e-01 -2.14707360e-01
3.88266772e-01 6.17935538e-01 -1.07937229e+00 -5.25451720e-01
2.23739818e-01 3.41852456e-01 -7.48733878e-01 9.21312630e-01
-3.23112309e-01 7.26085901e-01 -5.36520123e-01 1.97977543e-01
-7.46196955e-02 -7.16504678e-02 5.39942324e-01 2.11105525e-01
1.59116670e-01 3.12402427e-01 -8.70948970e-01 1.24702990e+00
-4.40144271e-01 3.53985876e-01 2.81532347e-01 -9.87063229e-01
1.27925193e+00 8.97018433e-01 5.96562326e-01 -4.95083332e-01
4.04890984e-01 -3.34559940e-02 2.42789939e-01 -9.92693365e-01
5.89860201e-01 -5.39292037e-01 -4.23874110e-01 9.09579813e-01
1.63523614e-01 -2.58493394e-01 1.10677779e-01 3.12324852e-01
1.19729567e+00 -4.15451564e-02 4.81602877e-01 2.30058819e-01
6.10999942e-01 -2.35853702e-01 4.57188725e-01 6.38048351e-01
-3.39595914e-01 2.00987667e-01 1.05836892e+00 -2.96290010e-01
-7.27700889e-01 -2.32767791e-01 4.06844318e-01 1.37678218e+00
-4.58013982e-01 -8.88006985e-02 -1.11664903e+00 -7.34997153e-01
-7.47504115e-01 9.41459358e-01 -6.53591156e-01 -2.25516498e-01
-4.68162477e-01 -4.86240625e-01 7.13268042e-01 2.30152428e-01
1.32575855e-01 -1.73008573e+00 -7.74888217e-01 5.70545018e-01
-6.16389275e-01 -7.85401464e-01 -1.47662729e-01 2.23094106e-01
-2.96602160e-01 -7.47445107e-01 -6.86710119e-01 -6.52369142e-01
6.28318489e-01 -5.01101255e-01 1.06512809e+00 1.33372173e-01
-2.89621688e-02 6.60080314e-01 -5.97642720e-01 -2.61291623e-01
-9.06467855e-01 8.40201974e-02 -1.10227026e-01 7.94067010e-02
4.70847398e-01 -4.49067265e-01 -4.27403957e-01 1.34698465e-01
-8.10937762e-01 3.13998640e-01 4.36311334e-01 1.14719474e+00
-3.33489716e-01 -3.85110140e-01 1.07487285e+00 -1.29955804e+00
1.58570552e+00 -5.87194741e-01 2.23891407e-01 3.05620760e-01
-4.11823660e-01 -1.42760381e-01 6.88875675e-01 -5.74264705e-01
-1.68886125e+00 2.56202936e-01 -5.79549968e-01 1.59975156e-01
-7.61295617e-01 4.02501941e-01 -2.42430419e-02 1.81373656e-01
9.43152130e-01 -1.48050025e-01 -1.15206800e-01 -5.54899760e-02
2.66390264e-01 1.10833180e+00 6.44945979e-01 -8.39952528e-01
6.80892393e-02 -4.16859299e-01 -6.91282094e-01 -7.15894401e-01
-5.80143094e-01 -3.64786774e-01 -4.87749994e-01 -8.55058968e-01
9.91673291e-01 -2.15215892e-01 -1.38257229e+00 1.17042668e-01
-1.44052875e+00 -6.29850745e-01 3.22216935e-02 4.46873426e-01
-7.89694369e-01 2.89083600e-01 -9.19784486e-01 -1.47375584e+00
-6.34036481e-01 -9.27911103e-01 7.41292894e-01 4.45165902e-01
-1.22569835e+00 -1.05061603e+00 2.50091434e-01 5.78355432e-01
2.31653050e-01 3.78636450e-01 1.04415810e+00 -1.22559619e+00
4.71868187e-01 -4.05792952e-01 5.50944880e-02 4.14430380e-01
1.91229805e-02 -1.65493742e-01 -9.63732183e-01 1.39192611e-01
1.40311807e-01 -1.11975121e+00 -1.02365993e-01 -2.61865437e-01
9.10777807e-01 -7.54737675e-01 3.21882337e-01 -2.60742277e-01
6.99081719e-01 7.00087786e-01 5.34040272e-01 -3.01625520e-01
2.34680772e-01 1.48173058e+00 6.47958755e-01 8.59614909e-01
3.63000244e-01 4.36322421e-01 1.26303717e-01 -3.14416319e-01
7.10193098e-01 7.45123103e-02 3.56819957e-01 6.44097924e-01
-2.26754382e-01 -3.23402941e-01 -8.60870600e-01 4.21079636e-01
-1.90145099e+00 -1.01387596e+00 2.20954716e-02 1.62100017e+00
1.48703718e+00 -1.51265800e-01 2.59309828e-01 1.13563137e-02
6.13361895e-01 -8.61969516e-02 -4.54383641e-01 -1.08500063e+00
3.79381865e-01 3.38857144e-01 -4.82654452e-01 8.10300767e-01
-5.76185346e-01 9.65366602e-01 5.53959608e+00 2.51273841e-01
-1.09036565e+00 -2.38132656e-01 8.91774297e-01 -1.35550633e-01
-2.84102798e-01 -3.09841633e-01 -2.82919139e-01 2.47730330e-01
1.09650075e+00 -1.79321859e-02 3.60042334e-01 6.98690593e-01
3.57292503e-01 -4.80091684e-02 -1.40528226e+00 6.69460297e-01
9.90332514e-02 -7.46662498e-01 -4.08663183e-01 -2.38937065e-01
3.40373605e-01 -7.81708598e-01 -2.89891720e-01 6.89324200e-01
7.14100778e-01 -1.06587481e+00 8.53578076e-02 7.04261661e-01
4.99705821e-01 -8.96916211e-01 9.35647786e-01 6.88070059e-01
-3.30303729e-01 1.37986362e-01 1.84743315e-01 -1.77170008e-01
2.72719920e-01 -1.15031965e-01 -1.44502509e+00 1.36775717e-01
2.95148760e-01 -2.39827596e-02 6.76736385e-02 2.07699403e-01
-3.35435092e-01 5.48584044e-01 -1.25688240e-02 -5.66539466e-01
3.94180745e-01 -2.95667410e-01 2.55751371e-01 1.62819397e+00
-8.87136087e-02 1.00504565e+00 3.48901182e-01 7.48600900e-01
-5.45604117e-02 4.10863757e-01 -5.87910056e-01 -3.77489030e-01
2.00282767e-01 1.53360701e+00 -4.72726554e-01 -4.92419899e-01
-3.15545946e-02 1.13373721e+00 2.97909617e-01 4.11473125e-01
-5.92913151e-01 -5.96375346e-01 2.62699515e-01 -4.67024028e-01
-3.62478346e-01 4.95390058e-01 -1.62994131e-01 -7.77549565e-01
-2.36681581e-01 -1.30494344e+00 3.68503988e-01 -8.45193386e-01
-1.17808294e+00 8.01224887e-01 -2.39132181e-01 -6.45564318e-01
-1.10624862e+00 -3.89995337e-01 -7.02903152e-01 1.04498243e+00
-5.70387602e-01 -7.72413552e-01 -1.36083350e-01 4.87841666e-01
6.82784319e-01 8.05928186e-02 1.32480323e+00 1.68541238e-01
-6.73128426e-01 6.07768357e-01 -7.09258735e-01 3.32899064e-01
8.49581063e-01 -1.48223925e+00 -3.47352624e-01 2.89876640e-01
-6.02638006e-01 5.67470968e-01 9.94020820e-01 -4.65104014e-01
-8.92316341e-01 -4.84301239e-01 1.16451550e+00 -2.82321274e-01
5.61514437e-01 -2.80331552e-01 -9.81989801e-01 4.59526390e-01
7.37393260e-01 -8.62796366e-01 1.34706354e+00 8.49734023e-02
2.71914780e-01 4.20366436e-01 -1.42328346e+00 8.95760357e-01
5.07191777e-01 -4.97762054e-01 -6.87648416e-01 4.46191162e-01
4.56381798e-01 -3.55672359e-01 -1.17276263e+00 1.99144229e-01
5.51581562e-01 -8.58030260e-01 4.91072536e-01 -1.12470365e+00
7.27623045e-01 3.46846998e-01 2.37050310e-01 -1.42695200e+00
-2.96506044e-02 -1.10230803e+00 2.90684521e-01 1.32869470e+00
5.62051058e-01 -3.03922504e-01 7.32451260e-01 1.63492465e+00
-7.71687776e-02 -8.77914011e-01 -3.80556405e-01 2.92672366e-01
-1.50610745e-01 -2.48393893e-01 3.01581025e-01 1.25796688e+00
9.18448567e-01 9.57471848e-01 -8.73326063e-01 -3.48592430e-01
-5.22040948e-02 -2.47784048e-01 8.26044679e-01 -1.32175744e+00
-3.45258147e-01 -4.58413690e-01 1.75509676e-01 -8.05149317e-01
4.01216656e-01 -4.05349374e-01 6.57085180e-01 -1.30565381e+00
-1.72445979e-02 -2.10536659e-01 -1.02821169e-02 6.47508681e-01
-2.60543764e-01 -7.40213543e-02 -4.41504456e-02 -3.38541865e-01
-6.79930806e-01 5.42439997e-01 9.44154561e-01 1.44233748e-01
-6.48644388e-01 8.62659737e-02 -7.23171949e-01 9.26212728e-01
8.95763397e-01 -2.26540327e-01 -5.48170984e-01 6.18872717e-02
3.63887697e-01 8.73579502e-01 -1.50252864e-01 -3.53363723e-01
3.12622517e-01 -5.46945572e-01 2.65583873e-01 -9.64705199e-02
4.84428942e-01 -5.89003623e-01 -1.61973789e-01 1.93097249e-01
-1.12534225e+00 -1.91388100e-01 -1.64817512e-01 5.75697906e-02
-1.75343081e-01 -6.26601934e-01 8.34412038e-01 -3.93001229e-01
-3.52525979e-01 -1.09130703e-01 -1.02646983e+00 -1.87674209e-01
9.20214295e-01 -1.40305161e-01 5.02270497e-02 -1.02533793e+00
-1.19960296e+00 4.42992449e-01 7.73492679e-02 2.25271791e-01
6.45953238e-01 -1.07533538e+00 -6.64315164e-01 -2.71507770e-01
-3.52008678e-02 -2.50332296e-01 4.03660327e-01 7.58714378e-01
-1.34680748e-01 2.20324278e-01 -3.06546122e-01 -2.48380825e-01
-1.57145274e+00 6.06444664e-02 4.70299453e-01 -4.77456331e-01
-1.98450238e-01 8.44115973e-01 -3.24278735e-02 -9.13769484e-01
4.26806211e-01 2.44564608e-01 -8.39519262e-01 4.56605643e-01
4.34306502e-01 2.23473370e-01 -2.48354256e-01 -4.94018793e-01
2.33663768e-01 -2.52570957e-01 -1.03026330e-01 -6.62823975e-01
1.24007118e+00 -7.17318326e-05 -1.30663872e-01 5.07209897e-01
8.21369529e-01 -3.02618235e-01 -8.13355923e-01 -2.27943629e-01
2.91127145e-01 -3.49684320e-02 -5.31071723e-01 -1.03759408e+00
-4.51498061e-01 8.60573471e-01 2.00663969e-01 5.87391675e-01
1.19050527e+00 -1.86762303e-01 7.14521408e-01 7.87073135e-01
1.50666460e-02 -1.71976650e+00 4.50156003e-01 6.94038570e-01
9.46492732e-01 -1.20640016e+00 -4.70053285e-01 -4.74206693e-02
-1.23059249e+00 1.30142367e+00 1.01950824e+00 3.47930789e-01
-1.00131243e-01 7.88765550e-02 4.52894837e-01 -2.92372525e-01
-1.17462397e+00 -1.13956621e-02 -1.75732486e-02 4.76638585e-01
8.79414380e-01 1.55041981e-02 -5.06802857e-01 7.02975690e-01
-2.82077819e-01 -7.07238466e-02 5.84029734e-01 8.06529284e-01
-3.55405927e-01 -1.15420151e+00 -1.83801129e-01 3.61218333e-01
-5.13057172e-01 1.74341887e-01 -1.41246688e+00 4.21423972e-01
-1.67780176e-01 1.32448137e+00 -3.30582023e-01 -4.72949207e-01
4.82831299e-01 6.01311803e-01 7.06921965e-02 -8.44174266e-01
-1.15523148e+00 -1.89919490e-02 8.26736867e-01 -4.12777543e-01
-4.46512401e-01 -4.17750776e-01 -1.35343647e+00 -1.07115433e-01
-1.08035065e-01 5.67267001e-01 3.99256796e-01 1.03658438e+00
-8.54570791e-02 4.00017351e-01 9.94357824e-01 -7.15811729e-01
-6.62559330e-01 -1.42878294e+00 -1.51877120e-01 7.53587246e-01
8.53561610e-02 -9.49643478e-02 -1.28084704e-01 -1.90137327e-02]
|
[13.104079246520996, 7.712787628173828]
|
0ed119cf-31a0-4f5e-9256-860a413ffc1f
|
bayesian-eye-tracking
|
2106.13387
| null |
https://arxiv.org/abs/2106.13387v1
|
https://arxiv.org/pdf/2106.13387v1.pdf
|
Bayesian Eye Tracking
|
Model-based eye tracking has been a dominant approach for eye gaze tracking because of its ability to generalize to different subjects, without the need of any training data and eye gaze annotations. Model-based eye tracking, however, is susceptible to eye feature detection errors, in particular for eye tracking in the wild. To address this issue, we propose a Bayesian framework for model-based eye tracking. The proposed system consists of a cascade-Bayesian Convolutional Neural Network (c-BCNN) to capture the probabilistic relationships between eye appearance and its landmarks, and a geometric eye model to estimate eye gaze from the eye landmarks. Given a testing eye image, the Bayesian framework can generate, through Bayesian inference, the eye gaze distribution without explicit landmark detection and model training, based on which it not only estimates the most likely eye gaze but also its uncertainty. Furthermore, with Bayesian inference instead of point-based inference, our model can not only generalize better to different sub-jects, head poses, and environments but also is robust to image noise and landmark detection errors. Finally, with the estimated gaze uncertainty, we can construct a cascade architecture that allows us to progressively improve gaze estimation accuracy. Compared to state-of-the-art model-based and learning-based methods, the proposed Bayesian framework demonstrates significant improvement in generalization capability across several benchmark datasets and in accuracy and robustness under challenging real-world conditions.
|
['Kang Wang', 'Qiang Ji']
|
2021-06-25
| null | null | null | null |
['gaze-estimation']
|
['computer-vision']
|
[-3.04452032e-01 -1.72076553e-01 1.12872552e-02 -5.00426292e-01
-1.73044443e-01 -1.71954557e-01 1.71127200e-01 -3.77287954e-01
-4.06872779e-01 4.82050776e-01 -2.83671945e-01 -1.39476791e-01
-1.06831439e-01 -2.18720689e-01 -8.44880879e-01 -5.68552434e-01
2.60778636e-01 7.55794719e-02 3.54884803e-01 1.52796730e-01
3.03838402e-01 4.36055899e-01 -2.15959072e+00 -4.82746065e-01
9.24611270e-01 1.23095787e+00 1.57036707e-01 6.15501702e-01
2.16742769e-01 4.00998265e-01 -5.56380630e-01 -2.05740735e-01
-1.66852772e-01 -5.68283349e-02 -2.71707594e-01 1.37386277e-01
9.92541134e-01 -4.71111268e-01 1.68818720e-02 1.05088353e+00
4.18066978e-01 2.45916292e-01 6.58069074e-01 -1.38760889e+00
-7.59268522e-01 -3.82479697e-01 -8.66327345e-01 2.18669549e-01
5.62487602e-01 4.93978918e-01 5.30942857e-01 -8.72698903e-01
1.62579834e-01 1.28013051e+00 6.63751781e-01 8.27873349e-01
-1.19513309e+00 -1.11880827e+00 3.75736535e-01 4.17742908e-01
-1.64354432e+00 -5.69652200e-01 5.10455012e-01 -5.55226743e-01
5.31392932e-01 -4.99261655e-02 6.30641639e-01 1.04705215e+00
3.18574369e-01 7.41220057e-01 1.23934209e+00 -5.03757000e-01
1.41591113e-03 1.76846862e-01 2.49619856e-01 7.71804035e-01
2.72005379e-01 5.06100297e-01 -8.83057714e-01 2.02637300e-01
3.44374806e-01 -2.33519375e-02 -5.28790891e-01 -1.92681879e-01
-7.32885897e-01 3.41715097e-01 8.08429897e-01 -3.48711818e-01
-2.87641108e-01 1.82962596e-01 -3.89470667e-01 -3.07218611e-01
5.01106679e-01 4.26061414e-02 -3.03239673e-01 1.23719750e-02
-1.12723720e+00 1.17582276e-01 6.06103957e-01 1.07102036e+00
8.80746186e-01 -2.61198461e-01 -2.73457199e-01 3.77522349e-01
1.21034193e+00 7.88021922e-01 1.60453811e-01 -4.16128397e-01
1.58580109e-01 6.28266096e-01 1.82430550e-01 -8.41505289e-01
-5.84320962e-01 -2.94445276e-01 -4.61886525e-01 4.97260898e-01
5.71149588e-01 -1.21407069e-01 -1.31278086e+00 1.82998121e+00
6.42724752e-01 5.69934070e-01 -4.48950559e-01 9.24232006e-01
8.85909736e-01 1.69209376e-01 3.39751281e-02 -1.87920675e-01
1.42956853e+00 -5.72722912e-01 -8.24750245e-01 -2.74836242e-01
3.10647786e-01 -7.41401553e-01 9.11133885e-01 3.44559431e-01
-8.01431537e-01 -6.20926797e-01 -8.99704814e-01 -8.03564861e-02
-1.56756103e-01 3.37189913e-01 3.37442547e-01 9.64012623e-01
-1.28066874e+00 -6.44935668e-02 -9.51191008e-01 -4.38124865e-01
6.10913992e-01 6.55954301e-01 -2.10146919e-01 5.83887212e-02
-7.24020362e-01 9.49397504e-01 2.31481642e-01 5.49751341e-01
-6.77632987e-01 -7.90873051e-01 -1.02212346e+00 8.74575451e-02
2.81731546e-01 -7.79120386e-01 1.16621935e+00 -6.90942526e-01
-1.85413074e+00 6.23726606e-01 -1.03060508e+00 -2.16716096e-01
1.78863660e-01 -3.85610282e-01 -4.90289986e-01 -1.26113698e-01
-3.75857204e-01 1.03781295e+00 1.47488499e+00 -1.10168433e+00
-8.52780044e-01 -5.89460611e-01 5.29343216e-03 -3.07561569e-02
-7.64413849e-02 3.76964271e-01 -6.95195317e-01 -5.66694997e-02
8.80998224e-02 -1.20144665e+00 4.61172998e-01 3.15334558e-01
-3.85963827e-01 -5.62370539e-01 8.14274430e-01 -5.04790246e-01
1.33307528e+00 -2.07954240e+00 -1.15879953e-01 1.39296785e-01
5.75813890e-01 1.72940910e-01 2.96108842e-01 -4.75197077e-01
1.18145190e-01 -1.15432709e-01 3.05875689e-01 -6.98715389e-01
-2.91099492e-03 -2.08709121e-01 -1.08374611e-01 5.92004955e-01
2.45750457e-01 9.06247616e-01 -5.85342169e-01 -5.27481377e-01
2.30730370e-01 7.73325324e-01 -5.35776675e-01 9.96908695e-02
-2.62774259e-01 7.45925248e-01 -9.83336940e-02 6.15065873e-01
7.39027917e-01 -5.00546396e-01 -3.64935994e-01 -2.16417387e-01
-7.68676698e-02 -5.77440895e-02 -1.06862056e+00 1.19123900e+00
-5.55551767e-01 1.09862483e+00 -3.29008609e-01 1.47735819e-01
8.07367504e-01 4.29516993e-02 -2.41926357e-01 -4.17548329e-01
4.58794415e-01 -1.68516099e-01 2.60280252e-01 -4.48573321e-01
4.09572512e-01 3.81189048e-01 4.28897560e-01 5.54033101e-01
2.93358624e-01 3.56739402e-01 -3.26461308e-02 -2.43009225e-01
3.37672561e-01 4.61586446e-01 2.22577434e-02 -4.16713506e-02
7.44038999e-01 -6.32638514e-01 5.00055432e-01 4.35390592e-01
-2.95475006e-01 5.36645293e-01 1.17971547e-01 -3.07959735e-01
-3.72618914e-01 -8.05499136e-01 -4.60293323e-01 1.13742495e+00
3.74215782e-01 -1.64107516e-01 -9.82795715e-01 -4.84428763e-01
-1.57261610e-01 7.76808679e-01 -7.80718505e-01 -2.31153607e-01
-1.16157316e-01 -6.38899505e-01 2.14144215e-01 3.27377915e-01
5.12298048e-01 -6.78789258e-01 -6.96027994e-01 -3.60819846e-01
1.80160794e-02 -1.01362264e+00 -8.03108633e-01 -4.55227613e-01
-6.41187668e-01 -1.33839357e+00 -5.89039505e-01 -3.01438838e-01
9.94122148e-01 1.35584652e-01 8.91219258e-01 2.21458033e-01
-1.55652523e-01 3.91019046e-01 5.01133017e-02 -9.81371224e-01
-1.09750904e-01 -8.79831314e-02 4.11521375e-01 5.78333855e-01
1.03466642e+00 -9.87273753e-02 -8.16212237e-01 6.06712759e-01
-4.24751818e-01 -2.65799105e-01 4.48083758e-01 8.00074041e-01
4.76939887e-01 -1.82288632e-01 2.65219808e-02 -5.09428680e-01
5.14113128e-01 -2.79242903e-01 -1.25039876e+00 4.53010410e-01
-8.66648257e-01 8.82418081e-02 -2.39844278e-01 -5.86349547e-01
-1.24183941e+00 -7.61455446e-02 1.99571416e-01 -6.36448920e-01
-3.54705691e-01 2.97619581e-01 -1.12828799e-01 -4.61084336e-01
8.78071487e-01 6.04691841e-02 2.21260339e-01 -5.32515764e-01
7.05417693e-02 8.62719297e-01 2.96598107e-01 -2.62163699e-01
5.49574435e-01 4.74299103e-01 3.19840133e-01 -7.33288407e-01
-1.17443454e+00 -3.93130362e-01 -9.67076004e-01 -4.08483386e-01
1.02230322e+00 -9.70637500e-01 -1.51636982e+00 9.00034487e-01
-1.18676090e+00 -3.82319801e-02 5.61390936e-01 8.14251959e-01
-2.92033553e-01 3.29507701e-02 -6.01300150e-02 -1.05761397e+00
-1.88864812e-01 -1.35852849e+00 1.13805115e+00 9.30793107e-01
-1.36451542e-01 -9.74971890e-01 -1.80034816e-01 2.37703159e-01
1.99633554e-01 -4.43639122e-02 4.73153830e-01 -2.60650814e-01
-9.41927552e-01 -4.79102075e-01 -2.86750227e-01 2.71320879e-01
1.57793209e-01 4.08435762e-01 -1.35650373e+00 -4.25084829e-01
-1.46266356e-01 5.14376611e-02 5.17499924e-01 7.76991963e-01
8.79321933e-01 3.40258610e-03 -6.57350719e-01 8.92871320e-01
9.76998985e-01 -2.20984116e-01 5.28890252e-01 1.22118644e-01
8.07037473e-01 5.26713252e-01 4.98467386e-01 1.86121091e-01
7.22252905e-01 8.36632669e-01 5.15788376e-01 1.46271318e-01
-2.59503245e-01 -1.79508463e-01 1.20002367e-01 1.07567251e-01
-1.27675995e-01 -1.90933287e-01 -1.01519120e+00 3.27450812e-01
-1.57056355e+00 -6.74814999e-01 -2.24540904e-01 2.45098591e+00
8.15095723e-01 1.03715062e-01 3.08801115e-01 -2.89013952e-01
8.42112720e-01 -3.46970052e-01 -8.81695211e-01 7.00016096e-02
3.51384163e-01 9.82286483e-02 4.40535784e-01 3.07558864e-01
-8.95738125e-01 8.74608457e-01 6.39473629e+00 2.32789502e-01
-1.42240834e+00 1.51962548e-01 2.20880970e-01 -3.18397343e-01
3.27059776e-01 -1.67756498e-01 -1.65417445e+00 8.42136264e-01
9.61469412e-01 1.37608051e-01 4.17285353e-01 6.79484010e-01
2.06878692e-01 -5.87072670e-01 -1.20418644e+00 1.20249271e+00
4.21976268e-01 -9.79663551e-01 -3.79249662e-01 3.37231457e-01
4.63352561e-01 1.85326204e-01 3.86832088e-01 3.85038070e-02
3.13119479e-02 -1.22329414e+00 7.56075740e-01 1.12102103e+00
8.79251719e-01 -6.30132556e-01 6.23902321e-01 5.00906944e-01
-7.48878539e-01 -9.12559927e-02 -1.79862559e-01 4.96274680e-02
-3.03011723e-02 1.28714191e-02 -8.67023647e-01 5.64653315e-02
1.07419538e+00 5.46927512e-01 -9.57101643e-01 1.60622847e+00
-6.79813504e-01 5.38652003e-01 -6.25008225e-01 -1.49287239e-01
-3.01704109e-01 1.76614657e-01 5.43507218e-01 4.96175379e-01
2.88042456e-01 -1.59633413e-01 -2.10739791e-01 1.20614672e+00
3.24153341e-02 -3.27034026e-01 -2.62456924e-01 6.03916943e-01
7.51113832e-01 1.08850288e+00 -3.19791108e-01 1.47399172e-01
-4.31239337e-01 4.50569123e-01 4.20669377e-01 6.74939513e-01
-7.63836086e-01 -3.15905690e-01 9.30915892e-01 2.42634401e-01
3.51925731e-01 -2.12270424e-01 -1.17792420e-01 -9.14187193e-01
8.56276006e-02 -5.82967162e-01 1.50438562e-01 -1.14594817e+00
-8.83226752e-01 7.40624905e-01 3.06263149e-01 -1.16749609e+00
-3.48686129e-01 -8.22931767e-01 -5.22220016e-01 1.56507480e+00
-1.76781869e+00 -1.38750350e+00 -5.81489265e-01 7.50053346e-01
6.78086802e-02 -1.73370972e-01 5.05807817e-01 -4.53283899e-02
-9.36016560e-01 1.11925244e+00 -2.29510263e-01 3.48392427e-02
9.41533327e-01 -9.77232695e-01 2.68680960e-01 9.93456841e-01
7.33219013e-02 9.48458076e-01 7.19053984e-01 -5.19348800e-01
-1.10741794e+00 -9.28542256e-01 6.54245377e-01 -1.11742306e+00
3.77903789e-01 -2.60142148e-01 -8.91489387e-01 7.41030097e-01
-4.26397547e-02 3.26411158e-01 6.23839259e-01 5.12256444e-01
-3.42025280e-01 -2.82575727e-01 -1.03798401e+00 6.04875147e-01
7.48841107e-01 -6.11703634e-01 -4.78509337e-01 1.94439944e-03
4.61031616e-01 -9.15436745e-01 -5.21966159e-01 2.92120814e-01
7.76067317e-01 -9.23274219e-01 6.45339072e-01 -4.55269068e-01
-3.31873566e-01 -6.65082872e-01 3.92383933e-01 -1.06478345e+00
-2.17654571e-01 -7.67885268e-01 -6.17873073e-01 1.10035157e+00
3.34551483e-01 -8.02667022e-01 5.71075976e-01 1.07309628e+00
2.42731139e-01 -5.38032293e-01 -8.26726615e-01 -5.46060383e-01
-3.68621975e-01 -5.44989347e-01 9.22188997e-01 1.15489133e-01
-5.07414162e-01 2.84540117e-01 -3.06335300e-01 8.30099225e-01
9.88360941e-01 -1.17086768e-01 1.10211849e+00 -1.71948004e+00
6.80482090e-02 -3.08900803e-01 -6.21473253e-01 -1.27184141e+00
4.79216456e-01 -3.04352134e-01 3.18534702e-01 -1.01830339e+00
-1.26136184e-01 -2.14746982e-01 -2.26517498e-01 2.98867375e-01
-5.54019392e-01 4.25032943e-01 6.27961829e-02 3.81721020e-01
-5.51467419e-01 3.26978207e-01 1.16512859e+00 1.53544769e-01
-4.83090341e-01 5.24610758e-01 -5.40609777e-01 8.18221152e-01
2.17476562e-01 -6.03125930e-01 -3.37686837e-01 -4.76214111e-01
4.41456765e-01 -3.96680206e-01 7.12000489e-01 -9.77529705e-01
8.19480717e-01 2.93721299e-04 6.55825377e-01 -7.90262461e-01
3.23494136e-01 -7.80056596e-01 -7.30678663e-02 -1.30022630e-01
1.30018219e-01 -2.67765969e-01 4.79764670e-01 9.14322257e-01
-1.95197146e-02 -2.54531056e-01 8.50317895e-01 4.52522039e-01
-7.02282846e-01 5.37437499e-01 -2.85117868e-02 -2.12762937e-01
1.00062084e+00 -6.95971370e-01 -1.75719067e-01 -2.49734059e-01
-7.52127528e-01 2.89140433e-01 6.20730042e-01 5.88801861e-01
4.20853704e-01 -9.69728947e-01 -3.76899749e-01 6.93074048e-01
4.65718061e-01 2.30541810e-01 2.19333410e-01 1.21585894e+00
-2.25592330e-01 5.74329376e-01 -3.27370875e-02 -1.30678892e+00
-1.38669395e+00 5.52144110e-01 5.79626203e-01 5.36626518e-01
-1.93281457e-01 1.04701579e+00 3.31843704e-01 -5.03726602e-02
5.40235639e-01 -5.29130042e-01 -4.56643760e-01 -1.07239515e-01
8.61067235e-01 2.22917855e-01 3.56301814e-02 -1.00231743e+00
-3.74894917e-01 8.49042654e-01 -1.24692082e-01 2.42283940e-01
8.01273525e-01 -6.51446223e-01 1.62816066e-02 2.75376678e-01
6.64329886e-01 -8.03751945e-02 -1.62102318e+00 -3.94168854e-01
-9.58899688e-03 -6.99973464e-01 5.41483045e-01 -7.06424356e-01
-7.99369514e-01 1.01575339e+00 9.04565752e-01 -1.94475129e-01
1.17507505e+00 -8.67795572e-02 3.37789029e-01 2.73484439e-01
3.08602750e-01 -7.32828915e-01 -2.39987522e-01 4.60216373e-01
6.09586000e-01 -1.66232121e+00 -3.37082334e-03 -2.29797855e-01
-3.64585489e-01 1.06848085e+00 8.37023318e-01 1.61897659e-01
1.28212833e+00 -2.30370894e-01 2.14378133e-01 -2.84282386e-01
-6.86365247e-01 -5.68232536e-01 1.12413204e+00 8.35727632e-01
1.74030423e-01 -2.55484104e-01 5.21236598e-01 3.38338375e-01
-1.25460535e-01 3.83417964e-01 1.71351194e-01 4.03709650e-01
-1.85887530e-01 -7.06967473e-01 -6.50015652e-01 2.98529446e-01
-3.54696393e-01 -2.53680915e-01 6.47355393e-02 7.87707269e-01
2.14252453e-02 1.06451547e+00 4.73091960e-01 -2.41253018e-01
1.57835618e-01 1.64835840e-01 6.63318753e-01 -7.39900827e-01
-8.57056826e-02 -9.86502022e-02 -4.96847034e-01 -5.40740490e-01
-5.36269426e-01 -6.46922290e-01 -7.29430199e-01 -2.74985582e-01
-1.03443837e+00 -3.97555858e-01 8.95251274e-01 1.16988003e+00
7.89378047e-01 3.33761066e-01 3.90470356e-01 -1.00436151e+00
-3.65192860e-01 -1.25094652e+00 -4.25642937e-01 -1.09513476e-02
7.59042323e-01 -1.27287889e+00 -3.77402008e-01 1.33680210e-01]
|
[14.136371612548828, 0.048909977078437805]
|
a9b3089f-16af-4cc4-9533-ef30e0bfac05
|
attention-based-occlusion-removal-for-hybrid
|
2112.01098
| null |
https://arxiv.org/abs/2112.01098v1
|
https://arxiv.org/pdf/2112.01098v1.pdf
|
Attention based Occlusion Removal for Hybrid Telepresence Systems
|
Traditionally, video conferencing is a widely adopted solution for telecommunication, but a lack of immersiveness comes inherently due to the 2D nature of facial representation. The integration of Virtual Reality (VR) in a communication/telepresence system through Head Mounted Displays (HMDs) promises to provide users a much better immersive experience. However, HMDs cause hindrance by blocking the facial appearance and expressions of the user. To overcome these issues, we propose a novel attention-enabled encoder-decoder architecture for HMD de-occlusion. We also propose to train our person-specific model using short videos (1-2 minutes) of the user, captured in varying appearances, and demonstrated generalization to unseen poses and appearances of the user. We report superior qualitative and quantitative results over state-of-the-art methods. We also present applications of this approach to hybrid video teleconferencing using existing animation and 3D face reconstruction pipelines.
|
['Avinash Sharma', 'Ashwath Shetty', 'Surabhi Gupta']
|
2021-12-02
| null | null | null | null |
['3d-face-reconstruction', 'face-reconstruction']
|
['computer-vision', 'computer-vision']
|
[ 1.69417396e-01 4.31788772e-01 4.25299346e-01 -4.44716871e-01
-6.87441826e-01 -3.11984122e-01 4.97468770e-01 -9.08233583e-01
-3.80577222e-02 5.02292752e-01 3.33324164e-01 4.10795212e-02
4.17921275e-01 -9.99185517e-02 -5.69048285e-01 -1.77634284e-01
-9.64346454e-02 1.85538217e-01 -1.17106795e-01 -4.46679264e-01
-1.30384386e-01 5.87788165e-01 -2.04012513e+00 4.58386838e-01
2.67310649e-01 8.98183942e-01 1.11839503e-01 9.76008832e-01
4.01767075e-01 4.53187853e-01 -5.42869151e-01 -5.50743401e-01
2.20906675e-01 -3.11288774e-01 -4.25217390e-01 5.70925236e-01
1.05789685e+00 -9.95990694e-01 -8.61696243e-01 6.62001371e-01
8.79268527e-01 1.82220504e-01 1.56501368e-01 -1.25349236e+00
-4.81547087e-01 -4.97241467e-01 -6.02922916e-01 -1.20740592e-01
1.32540643e+00 -3.18823308e-01 6.57173574e-01 -8.99151981e-01
1.10249102e+00 1.45413589e+00 4.12889570e-01 1.09412646e+00
-1.20617437e+00 -5.18409908e-01 3.08012575e-01 -7.50034377e-02
-1.62575603e+00 -1.09091175e+00 5.98714828e-01 -3.33084971e-01
1.09499454e+00 6.31802201e-01 9.59625721e-01 1.51266038e+00
5.20295203e-02 8.82449985e-01 8.20006013e-01 -2.07704008e-01
-1.57738701e-01 5.61321974e-01 -5.53867877e-01 6.70941532e-01
-6.05482578e-01 1.92690656e-01 -5.79140365e-01 4.61487807e-02
1.23446918e+00 -1.62455261e-01 -7.83656240e-01 -4.49372977e-01
-6.06631815e-01 3.13378036e-01 1.00760438e-01 -2.62428410e-02
-2.28206396e-01 3.75042409e-02 4.20004338e-01 4.35628921e-01
7.36432552e-01 -7.80540332e-02 -5.18066347e-01 -6.45941138e-01
-7.96073198e-01 1.45696163e-01 6.89093530e-01 1.21014392e+00
2.98247021e-02 -1.71893016e-02 1.66111931e-01 6.85843647e-01
3.09405655e-01 1.03100128e-01 -1.74998730e-01 -1.04123235e+00
2.09830418e-01 -4.84711789e-02 1.94614187e-01 -9.14019346e-01
-2.93531686e-01 6.86176168e-03 -4.08434868e-01 5.72665274e-01
-8.36597607e-02 -3.04557204e-01 -7.91356444e-01 1.70346820e+00
5.10697186e-01 5.62274098e-01 -4.44792688e-01 1.25210941e+00
1.09493077e+00 3.76321465e-01 -3.67322534e-01 -1.78641528e-01
1.21895015e+00 -6.44549787e-01 -1.03160334e+00 1.09682968e-02
3.20996791e-01 -7.55287230e-01 9.08496916e-01 6.55457616e-01
-1.22451246e+00 -4.79517579e-01 -7.17541516e-01 -2.79612780e-01
3.71849537e-03 6.69459999e-02 4.77747500e-01 1.15574467e+00
-1.54240811e+00 5.02321601e-01 -7.31960833e-01 -6.12846851e-01
2.38278657e-01 8.10973167e-01 -8.39918017e-01 3.03059779e-02
-8.06310594e-01 1.03194046e+00 -4.16412503e-01 3.23080420e-01
-5.24178803e-01 -4.63857055e-01 -1.03264344e+00 -1.62615236e-02
1.59871057e-01 -6.84899092e-01 1.29090357e+00 -1.22289515e+00
-1.96310604e+00 1.26814759e+00 -2.56180763e-01 2.66537338e-01
8.22198153e-01 -5.35733700e-01 -6.40313685e-01 2.38733977e-01
-6.05536997e-01 8.54864717e-01 1.02826309e+00 -1.32505262e+00
-3.56805384e-01 -3.73975784e-01 3.10339093e-01 4.80105996e-01
-1.47850290e-01 4.92798448e-01 -8.73027027e-01 -1.34332493e-01
-3.89219485e-02 -1.02511716e+00 1.83443666e-01 4.94693011e-01
-2.08390802e-01 4.76000756e-01 1.31701624e+00 -1.03591216e+00
7.97902703e-01 -2.35755610e+00 4.94886309e-01 -1.32600158e-01
3.75721633e-01 9.48905125e-02 -5.95726296e-02 1.29015937e-01
-3.33162129e-01 -3.16813916e-01 4.05042052e-01 -1.15437591e+00
5.77894785e-02 6.52953312e-02 5.33914007e-02 6.75501406e-01
5.90793565e-02 6.51881278e-01 -5.22523224e-01 -3.31811935e-01
7.31614351e-01 1.37349808e+00 -7.50409603e-01 4.62583244e-01
3.37170392e-01 9.18150544e-01 1.99784577e-01 6.42660499e-01
9.85345602e-01 9.22185183e-02 3.03311408e-01 7.66854957e-02
6.69295713e-02 1.70132831e-01 -1.14799893e+00 1.99389291e+00
-5.89941442e-01 1.09619987e+00 7.87810504e-01 -2.31973827e-01
6.01166368e-01 8.18400681e-01 4.20556992e-01 -7.70212948e-01
3.41909111e-01 -6.46654591e-02 -5.56027114e-01 -8.38206172e-01
7.71734357e-01 -2.46043846e-01 2.72853464e-01 -4.76008654e-02
-3.03348689e-03 -1.85666621e-01 -5.45128942e-01 2.14876041e-01
8.13904285e-01 6.87935591e-01 -1.26185283e-01 3.16742480e-01
1.62425280e-01 -6.87064588e-01 2.93064594e-01 7.56672323e-02
-3.43123674e-01 1.22677302e+00 3.87709677e-01 -2.40585506e-01
-1.02269363e+00 -9.46078718e-01 -2.60842502e-01 9.99924362e-01
1.71375442e-02 -5.92824221e-01 -6.71630800e-01 -3.48078609e-01
-1.99739441e-01 4.48422253e-01 -5.55722296e-01 1.69738635e-01
-3.49337637e-01 -1.07524052e-01 2.32757628e-01 4.57121044e-01
1.90212145e-01 -5.39131582e-01 -6.33022428e-01 -8.23396631e-03
-1.53492078e-01 -1.49352872e+00 -3.45973998e-01 -5.79537094e-01
-6.07190967e-01 -6.82156980e-01 -9.81212258e-01 -5.92275739e-01
7.23208725e-01 4.09883589e-01 1.11207533e+00 4.37036306e-02
-3.49701643e-01 8.33271325e-01 -2.42554739e-01 2.13172227e-01
-2.10815649e-02 -4.81734395e-01 2.32898280e-01 3.86114754e-02
1.43532187e-01 -6.55660927e-01 -6.69180036e-01 3.93721282e-01
-5.15803337e-01 3.75884235e-01 -8.24649781e-02 5.88475645e-01
2.12078150e-02 -2.41762489e-01 1.73398629e-02 -8.35234284e-01
2.92259395e-01 -1.39532968e-01 -3.71000826e-01 -1.95519105e-01
2.04177290e-01 -8.91814709e-01 2.39475444e-01 -4.28046942e-01
-1.33368492e+00 1.69658467e-01 -3.70642871e-01 -8.45133662e-01
-2.14450970e-01 -3.29204574e-02 -3.72200668e-01 -2.91647524e-01
1.57779977e-01 -3.85471851e-01 1.07634425e-01 -4.55901384e-01
2.84389883e-01 1.01004577e+00 5.09211838e-01 -1.39557675e-01
4.26044792e-01 6.47484243e-01 -2.68702686e-01 -1.35783672e+00
-5.03104478e-02 -3.35742205e-01 -7.63312042e-01 -6.50681853e-01
7.67190814e-01 -1.30649531e+00 -1.00526905e+00 2.79914379e-01
-1.28152955e+00 -7.59290159e-02 2.37704232e-01 4.71073359e-01
-6.17237210e-01 2.08767846e-01 -7.52840996e-01 -1.27160382e+00
1.82066008e-01 -1.10491097e+00 1.69377124e+00 4.52915877e-02
-5.68575025e-01 -8.38595390e-01 -1.79291202e-03 6.04664743e-01
2.88590491e-01 4.74167734e-01 2.47380301e-01 1.77548692e-01
-5.09567499e-01 -4.99411821e-01 -1.88453138e-01 9.76088196e-02
7.21033514e-02 2.25698203e-01 -1.59210598e+00 -4.94176418e-01
-2.41565630e-01 -2.85637885e-01 2.22068280e-01 4.39383388e-01
8.51444423e-01 -2.64950544e-02 -3.50648195e-01 7.06701756e-01
1.00526667e+00 1.17909454e-01 1.23883724e+00 -2.35367958e-02
8.56818438e-01 8.48940253e-01 5.49083114e-01 6.09599292e-01
3.03245425e-01 1.11406302e+00 5.04796267e-01 -4.47551847e-01
5.14987633e-02 -2.33598411e-01 4.04751033e-01 5.83307385e-01
-4.68378127e-01 -2.83907115e-01 -3.01121026e-01 6.74360245e-02
-1.44226551e+00 -9.39684391e-01 -1.98151380e-01 2.31981802e+00
1.28566816e-01 -1.92274585e-01 2.51106352e-01 -1.24570029e-02
6.99519932e-01 -6.77813590e-02 -1.16551273e-01 -5.75040758e-01
5.38379662e-02 1.53545514e-01 -5.86097464e-02 6.40024483e-01
-1.02740145e+00 7.67018557e-01 6.72980309e+00 2.01835811e-01
-1.23189759e+00 5.46664000e-02 3.74770671e-01 -6.19662046e-01
-2.63047367e-01 -3.36325586e-01 -2.97941059e-01 1.79273382e-01
8.85120213e-01 4.48133081e-01 4.32242811e-01 6.60615325e-01
3.37841421e-01 -1.13047361e-01 -1.32298172e+00 1.52292907e+00
4.32882875e-01 -8.93098414e-01 -6.92570865e-01 4.78889555e-01
4.56181347e-01 -2.07979307e-01 4.39859778e-01 2.48668596e-01
-4.94060189e-01 -1.10273314e+00 6.42038822e-01 3.53136510e-01
1.45366096e+00 -6.39299691e-01 4.15989876e-01 7.07271695e-02
-1.10701895e+00 1.23021238e-01 -1.07413456e-01 -2.43744686e-01
5.47294438e-01 -1.07803211e-01 -7.31142819e-01 2.40220830e-01
7.36953199e-01 5.69660604e-01 -9.92768183e-02 7.24942684e-01
1.55293554e-01 -3.62449110e-01 -3.19086909e-01 3.61886948e-01
-3.64498109e-01 -3.15122128e-01 6.52713835e-01 1.09649813e+00
3.52037847e-01 4.39342856e-01 -3.20749104e-01 3.86924714e-01
6.69475738e-03 -1.08507492e-01 -1.04124749e+00 8.57362971e-02
-5.08951880e-02 1.06086171e+00 -3.92834842e-01 -5.74674197e-02
-7.40500689e-01 1.89834607e+00 1.70854792e-01 3.93438727e-01
-9.31536973e-01 -2.41888180e-01 1.04933596e+00 3.66713405e-01
3.29719692e-01 -3.27514678e-01 4.00752127e-01 -1.59173179e+00
1.38884693e-01 -1.00693035e+00 -1.49976000e-01 -1.22487366e+00
-6.36579514e-01 8.37065220e-01 -3.16966414e-01 -1.53050494e+00
-3.41104984e-01 -6.22689903e-01 -1.57417908e-01 7.22928047e-01
-1.02832949e+00 -1.12385929e+00 -5.51743507e-01 3.94146442e-01
5.63797295e-01 1.37348309e-01 1.18023586e+00 7.70240247e-01
-6.49210036e-01 6.48018181e-01 -5.17188013e-02 -2.71037668e-01
8.68558764e-01 -9.01326299e-01 5.32194376e-01 3.72117460e-01
-1.85813606e-02 5.81618428e-01 9.64730203e-01 -2.72139549e-01
-1.73415947e+00 -4.37336236e-01 8.20046008e-01 -8.30458641e-01
1.17842726e-01 -8.71088564e-01 -6.05431914e-01 1.04591620e+00
2.93958455e-01 -1.15568958e-01 7.91038096e-01 2.47483149e-01
-6.96845949e-02 2.12874129e-01 -1.49018562e+00 7.93952286e-01
1.34201086e+00 -9.96390820e-01 -2.18235090e-01 1.27263159e-01
3.67008418e-01 -1.03468609e+00 -7.07203448e-01 2.71627214e-03
1.18927634e+00 -1.46407449e+00 8.58814061e-01 -4.69579756e-01
5.46294272e-01 5.82719967e-02 -3.06504637e-01 -1.22135341e+00
-1.79749563e-01 -1.18813932e+00 -6.31181374e-02 1.03538513e+00
2.12407038e-02 -4.20928448e-01 8.80317450e-01 1.28303218e+00
-3.35038677e-02 -5.91949224e-01 -1.06158805e+00 -2.80815244e-01
-6.21436656e-01 -4.83837724e-01 2.83165634e-01 9.13579822e-01
5.10277212e-01 3.29582244e-01 -9.67791080e-01 3.08629572e-01
2.35924870e-01 -4.40717161e-01 1.00745153e+00 -1.19595838e+00
-4.43853319e-01 1.53906614e-01 -9.11026776e-01 -1.26143563e+00
2.42073327e-01 -3.16075593e-01 -1.38813585e-01 -1.12374520e+00
-3.46152037e-02 6.74528629e-02 2.06306949e-01 -1.79057091e-01
2.70948648e-01 5.14792502e-01 2.47183174e-01 -2.06670985e-01
-3.83444846e-01 7.15117097e-01 1.28384256e+00 4.58856374e-01
-2.83579350e-01 2.88146250e-02 -3.51114720e-01 6.07332110e-01
1.69304922e-01 1.05478317e-02 -5.07764101e-01 -6.17562115e-01
2.18141854e-01 6.31854057e-01 4.68875527e-01 -6.24902368e-01
-9.37109664e-02 3.83634835e-01 6.78935587e-01 -2.83735782e-01
1.34607971e+00 -1.14688694e+00 5.25539994e-01 -4.54473607e-02
2.46270113e-02 -2.13162899e-02 3.85201156e-01 3.92371505e-01
-1.39880348e-02 4.03223187e-01 5.85352063e-01 8.01981091e-02
-4.57410842e-01 2.88724601e-01 -5.99597931e-01 -4.17067498e-01
7.91844308e-01 -5.64684510e-01 1.18886538e-01 -1.05443978e+00
-1.05048156e+00 1.00939423e-02 7.71956384e-01 5.86266339e-01
1.03389847e+00 -1.18975925e+00 -4.51363206e-01 6.53531194e-01
-1.69363707e-01 -3.68389636e-01 5.62074244e-01 8.82167399e-01
-6.10681653e-01 2.95621604e-01 -3.58077139e-01 -6.20723069e-01
-2.06026936e+00 2.18253344e-01 4.44824576e-01 6.44037366e-01
-8.87036800e-01 1.18554533e+00 3.29860538e-01 -3.76335472e-01
4.69159126e-01 1.35315418e-01 -1.81543842e-01 -6.31785169e-02
7.03548074e-01 4.05542582e-01 2.01637343e-01 -1.08239365e+00
-3.64951938e-01 5.89042366e-01 -5.59094138e-02 -5.46788454e-01
1.21176827e+00 -5.74530303e-01 3.14215660e-01 2.18773305e-01
1.46330225e+00 3.47994640e-02 -1.61650467e+00 2.90654689e-01
-6.99624419e-01 -1.37584174e+00 2.55877912e-01 -3.20744693e-01
-1.06393099e+00 8.94495547e-01 7.58125365e-01 -1.11198366e-01
1.15870655e+00 -1.55212045e-01 7.02390611e-01 3.10882344e-03
5.96986592e-01 -1.03857708e+00 -1.89079851e-01 1.60823956e-01
8.55489671e-01 -1.22871244e+00 -1.19430661e-01 -8.10509264e-01
-9.20301914e-01 9.29991663e-01 6.11435890e-01 8.49637091e-02
5.16241074e-01 3.93808573e-01 2.91083336e-01 -2.87260830e-01
-8.78644466e-01 4.81578782e-02 2.06247374e-01 8.27734590e-01
8.63964438e-01 -5.71677908e-02 3.60385627e-01 1.37324706e-01
-2.28657082e-01 -2.86054015e-02 5.42336047e-01 1.07850564e+00
1.06468141e-01 -8.17612886e-01 -3.29325378e-01 -8.95999745e-02
-3.88308585e-01 2.19127387e-01 -5.32616138e-01 9.34968948e-01
-4.56490368e-02 9.68245566e-01 4.61650312e-01 -6.81271136e-01
6.14558637e-01 -1.21527269e-01 1.09274936e+00 -5.67754507e-01
-4.87909466e-01 5.86594522e-01 5.71803808e-01 -8.68569911e-01
-2.18410417e-01 -6.25685096e-01 -8.55955243e-01 -6.57631397e-01
-4.11374658e-01 -3.98139745e-01 8.03777099e-01 6.29409850e-01
5.24442554e-01 3.75014395e-01 8.40901196e-01 -1.52560341e+00
7.74135515e-02 -7.50328302e-01 -9.16277945e-01 3.18555266e-01
6.83047891e-01 -6.25932992e-01 -4.12523806e-01 4.97403294e-02]
|
[12.919301986694336, -0.34482306241989136]
|
76191a67-85fc-4af3-a7d8-293c2b2167be
|
unsupervised-chinese-word-segmentation-with-1
| null | null |
https://aclanthology.org/2022.findings-acl.310
|
https://aclanthology.org/2022.findings-acl.310.pdf
|
Unsupervised Chinese Word Segmentation with BERT Oriented Probing and Transformation
|
Word Segmentation is a fundamental step for understanding Chinese language. Previous neural approaches for unsupervised Chinese Word Segmentation (CWS) only exploits shallow semantic information, which can miss important context. Large scale Pre-trained language models (PLM) have achieved great success in many areas because of its ability to capture the deep contextual semantic relation. In this paper, we propose to take advantage of the deep semantic information embedded in PLM (e.g., BERT) with a self-training manner, which iteratively probes and transforms the semantic information in PLM into explicit word segmentation ability. Extensive experiment results show that our proposed approach achieves state-of-the-art F1 score on two CWS benchmark datasets.
|
['Yanqiu Shao', 'Qi Su', 'Yuhan Song', 'Wei Li']
| null | null | null | null |
findings-acl-2022-5
|
['chinese-word-segmentation']
|
['natural-language-processing']
|
[ 2.76158482e-01 1.08286403e-01 -4.47788745e-01 -5.90058923e-01
-5.79651654e-01 -4.15840000e-01 2.12765068e-01 1.36285797e-01
-7.11342335e-01 4.19493884e-01 3.65970820e-01 -5.25539160e-01
4.98308718e-01 -7.34002948e-01 -4.37798440e-01 -3.63974184e-01
3.69876385e-01 3.91670495e-01 6.54670954e-01 -1.52477860e-01
4.01188940e-01 -1.40292376e-01 -7.35478997e-01 2.74011046e-01
1.45817149e+00 6.04148984e-01 8.16001713e-01 2.93605447e-01
-8.58072400e-01 5.06763697e-01 -6.11264288e-01 -1.84480205e-01
-2.77346373e-01 -6.55020714e-01 -1.21281183e+00 -2.67983489e-02
-2.98267186e-01 -1.48749694e-01 1.11860096e-01 1.34652984e+00
5.58651350e-02 2.25680217e-01 2.46212751e-01 -5.41917264e-01
-7.29050398e-01 1.09674525e+00 -5.85982740e-01 2.32863232e-01
-5.25574796e-02 2.27766000e-02 1.32332194e+00 -7.96481729e-01
5.34455001e-01 1.32377470e+00 4.40979958e-01 7.54174829e-01
-9.98383939e-01 -6.20484352e-01 8.36356878e-01 5.78866042e-02
-1.15921438e+00 1.27529085e-01 8.09857428e-01 2.26777475e-02
1.00754952e+00 -1.65032558e-02 7.01776862e-01 8.48158062e-01
-1.04947351e-01 1.64909112e+00 1.01929915e+00 -6.98234975e-01
2.82888025e-01 -1.77006274e-01 8.65241647e-01 4.38518912e-01
4.93435740e-01 -5.27421236e-01 -2.21813634e-01 2.76666075e-01
7.06272900e-01 -1.11134298e-01 -2.51183044e-02 3.08463573e-01
-7.60792613e-01 9.51947629e-01 2.59479135e-01 7.20254481e-01
-1.17754012e-01 1.46220520e-01 2.92783886e-01 -7.41824508e-02
5.59808493e-01 5.85643411e-01 -8.81966054e-01 -1.81648448e-01
-7.90580690e-01 -2.87977904e-01 5.56948900e-01 8.89897466e-01
7.57357419e-01 3.57873105e-02 -2.15304449e-01 8.11710179e-01
3.16352904e-01 1.31789908e-01 9.58557129e-01 -5.83113968e-01
4.31554466e-01 6.93022668e-01 -2.25927025e-01 -5.74418008e-01
-4.07399148e-01 -4.40850109e-01 -5.50149739e-01 -6.30344391e-01
2.35888362e-01 -4.44976538e-01 -1.43313920e+00 1.82646036e+00
3.97995207e-03 3.82981569e-01 2.46642366e-01 8.23958576e-01
7.84626663e-01 7.59232759e-01 6.91487372e-01 1.46379635e-01
1.39500189e+00 -1.22589374e+00 -7.83055663e-01 -9.12433982e-01
7.46724069e-01 -5.49632251e-01 1.41105700e+00 2.13733837e-01
-8.30940425e-01 -6.29249096e-01 -7.56992877e-01 -1.37812719e-01
-4.95203286e-01 1.01589359e-01 7.54130363e-01 7.58871436e-01
-8.55328202e-01 3.07854563e-01 -1.17368948e+00 -3.50080341e-01
6.92305684e-01 2.53997386e-01 1.58816814e-01 -3.00796121e-01
-1.51958919e+00 4.84420359e-01 9.55557644e-01 2.74466932e-01
-6.88711643e-01 -3.32285196e-01 -1.01877081e+00 1.91202983e-01
8.28540981e-01 -1.69439629e-01 1.25269902e+00 -1.15252352e+00
-1.48324752e+00 7.05967605e-01 -4.57575113e-01 -4.22783017e-01
6.08409196e-02 -7.51718163e-01 -2.02880934e-01 2.56458014e-01
1.75347835e-01 9.47934687e-01 4.71442163e-01 -1.26877069e+00
-6.32912159e-01 -2.56214291e-01 -4.67315204e-02 7.25612268e-02
-4.47614759e-01 2.81177551e-01 -9.87906516e-01 -8.72247577e-01
2.72947043e-01 -5.30294597e-01 -6.89450145e-01 -8.95669162e-01
-6.21152341e-01 -4.57512915e-01 7.24216163e-01 -8.24680328e-01
1.37199807e+00 -1.74214244e+00 -8.97703990e-02 6.04114868e-02
-7.33198151e-02 7.89876044e-01 -4.35552508e-01 1.97191641e-01
2.79686242e-01 5.47697604e-01 -3.82053047e-01 -4.90079969e-01
-9.35572907e-02 5.51407397e-01 -2.63386339e-01 -1.08122036e-01
4.52061623e-01 1.44177091e+00 -1.00763583e+00 -5.70546389e-01
-6.87965080e-02 1.65910736e-01 -5.82337797e-01 2.45423004e-01
-7.50541747e-01 5.32762468e-01 -9.03431833e-01 4.59237009e-01
6.62934959e-01 -3.51266921e-01 4.38876122e-01 3.14387590e-01
6.60954267e-02 5.90208709e-01 -7.82333672e-01 1.90577543e+00
-3.61145705e-01 4.26038235e-01 -7.30624571e-02 -1.27219093e+00
7.90211916e-01 1.02193020e-01 -2.62449831e-02 -8.00641477e-01
3.39178860e-01 1.62097737e-01 6.72110319e-02 -1.79179013e-01
5.97540259e-01 -1.24477521e-01 -2.82480419e-01 2.76342392e-01
2.26470202e-01 5.04756682e-02 1.91400379e-01 3.16334069e-01
5.97333491e-01 2.34982356e-01 2.45089069e-01 -5.54477155e-01
5.94185352e-01 1.22255675e-01 8.56987953e-01 7.19733417e-01
-1.64368555e-01 6.26749516e-01 6.93789065e-01 -8.96598175e-02
-5.40912628e-01 -8.49753976e-01 2.40429536e-01 1.11135817e+00
3.87051463e-01 -4.42095488e-01 -1.20633531e+00 -7.80071318e-01
-4.91374046e-01 9.73084450e-01 -4.91211891e-01 -2.63740003e-01
-9.52182591e-01 -8.78112972e-01 5.03500819e-01 9.67826784e-01
7.22475588e-01 -1.39858413e+00 -4.94248509e-01 4.23621446e-01
-2.99857199e-01 -1.54614604e+00 -4.24598068e-01 3.10615778e-01
-1.07727134e+00 -8.09634984e-01 -5.97372711e-01 -1.30011344e+00
9.51898336e-01 3.99988413e-01 1.08910370e+00 4.25723106e-01
-6.25960007e-02 9.80638992e-03 -6.99204147e-01 -3.16255689e-01
-1.35572746e-01 6.47215843e-01 -5.42081296e-01 -1.31518066e-01
8.59816849e-01 -2.55837012e-02 -5.78530431e-01 1.84076905e-01
-1.05789399e+00 3.59393090e-01 6.61914647e-01 7.67284453e-01
6.49651349e-01 1.81860059e-01 6.79403603e-01 -1.54202116e+00
6.20341063e-01 -3.34220588e-01 -5.61166823e-01 2.91150331e-01
-5.21411836e-01 7.83226416e-02 6.51796699e-01 -3.83528531e-01
-1.45295763e+00 -1.89470693e-01 -4.32365686e-01 3.01024914e-01
-4.63754505e-01 8.56131911e-01 -5.75189114e-01 4.83498096e-01
-3.13721225e-02 3.78204823e-01 -5.61801136e-01 -9.56039310e-01
4.61035609e-01 6.07197702e-01 5.91355860e-01 -9.93456662e-01
3.87875855e-01 3.18530411e-01 -6.26623809e-01 -7.83245146e-01
-1.33739769e+00 -8.02169919e-01 -9.23269391e-01 3.88014376e-01
1.36526811e+00 -9.39849019e-01 -1.57058448e-01 8.26283216e-01
-1.16325366e+00 -6.99834168e-01 -7.29906708e-02 2.56114990e-01
1.14528961e-01 5.18748343e-01 -9.08546925e-01 -6.91765964e-01
-4.06927377e-01 -9.67179298e-01 9.33609307e-01 7.74035037e-01
-1.87526941e-01 -1.19240117e+00 -3.18504393e-01 3.24368238e-01
3.27612817e-01 -1.30478308e-01 8.83246899e-01 -1.00107384e+00
-5.23477733e-01 -1.86313599e-01 -5.20952761e-01 4.66101170e-01
1.55003682e-01 -4.59875405e-01 -7.23590195e-01 -7.81569481e-02
-1.79188624e-01 -2.23372474e-01 1.42043149e+00 3.62691700e-01
1.36393619e+00 -1.58639073e-01 -4.12921607e-01 3.42856973e-01
1.58024168e+00 2.71906674e-01 5.41361153e-01 1.89451411e-01
1.00795841e+00 5.55137336e-01 6.93416655e-01 7.93780312e-02
4.30821329e-01 -4.07449454e-02 2.36865386e-01 -2.62642771e-01
5.38803190e-02 -4.69449461e-01 2.20506519e-01 1.09776115e+00
3.25753063e-01 -3.15396726e-01 -1.14126766e+00 9.97777998e-01
-1.84141493e+00 -1.33998975e-01 -2.44301874e-02 1.62403286e+00
1.20095146e+00 6.52216017e-01 -2.84806758e-01 -2.66211390e-01
6.92225337e-01 3.35354507e-01 -4.91945773e-01 -3.42336476e-01
-1.35160625e-01 6.28139377e-01 4.83336508e-01 6.81985199e-01
-1.05284858e+00 2.05639768e+00 5.44463825e+00 1.08974028e+00
-9.13307250e-01 1.40346259e-01 8.46771121e-01 6.53999090e-01
-6.07933402e-01 8.99221599e-02 -1.06306088e+00 3.27091634e-01
8.33103716e-01 2.64425665e-01 6.84672669e-02 7.83584237e-01
1.56726718e-01 -4.40636337e-01 -4.81136829e-01 2.85486490e-01
-1.00888796e-01 -1.15816462e+00 2.58882284e-01 -3.07311058e-01
1.03741348e+00 1.77460060e-01 -2.22924456e-01 3.65852624e-01
7.68270850e-01 -1.04962993e+00 5.86546063e-01 -3.70319076e-02
7.14968503e-01 -7.88399398e-01 9.36761498e-01 5.88473678e-01
-1.26857388e+00 2.10486963e-01 -4.57124412e-01 -1.53638020e-01
2.27991864e-01 4.81127024e-01 -4.99031872e-01 3.38768601e-01
5.14527261e-01 6.37651026e-01 -6.36015058e-01 6.76539183e-01
-1.00213754e+00 1.35559976e+00 -1.54087886e-01 -3.74249905e-01
8.89503241e-01 -3.08777064e-01 1.14156112e-01 1.55387735e+00
-2.21678570e-01 4.08768862e-01 4.14594501e-01 9.16918576e-01
-1.01086430e-01 1.80720031e-01 1.89718097e-01 -4.39580917e-01
3.26782525e-01 9.77704108e-01 -1.45386064e+00 -5.42785287e-01
-5.49142241e-01 1.11236918e+00 3.60145241e-01 4.90368009e-01
-5.04405439e-01 -4.89883840e-01 7.87224531e-01 -4.02237117e-01
6.68087542e-01 -5.91104388e-01 -6.82864428e-01 -1.23651695e+00
-2.40059912e-01 -5.01718581e-01 5.09957373e-01 -4.23887879e-01
-9.44745243e-01 4.47686136e-01 -1.72379777e-01 -3.93807083e-01
1.18938528e-01 -7.48469055e-01 -9.37479079e-01 9.73212838e-01
-1.83214676e+00 -1.33393204e+00 -1.29327610e-01 2.28897169e-01
1.03663993e+00 1.86388552e-01 6.95323467e-01 -8.25747252e-02
-8.11808288e-01 4.53054160e-01 -2.55768429e-02 7.70563304e-01
1.94750533e-01 -1.41665184e+00 1.03642988e+00 1.34466314e+00
3.28831047e-01 9.05276239e-01 4.45538551e-01 -9.05652761e-01
-9.26793039e-01 -1.18017793e+00 9.10606742e-01 -2.27405712e-01
4.81650621e-01 -6.16064966e-01 -1.11404991e+00 7.35869586e-01
3.39247257e-01 -2.91982681e-01 7.01332271e-01 1.67963043e-01
-2.53522664e-01 3.57306838e-01 -7.41782546e-01 6.92237139e-01
9.50176001e-01 -3.77007306e-01 -1.03056693e+00 -1.29132196e-01
1.40655088e+00 -2.78875917e-01 -1.03970371e-01 1.84675127e-01
1.58390820e-01 -4.69011277e-01 7.79388487e-01 -9.47100639e-01
3.95153046e-01 -1.78576574e-01 -9.71525069e-03 -1.24748707e+00
-5.73931001e-02 -5.98009765e-01 2.37454876e-01 1.28806651e+00
4.37564969e-01 -5.71759343e-01 9.37546968e-01 4.45234656e-01
-2.88289368e-01 -7.37640619e-01 -4.34838653e-01 -6.74632072e-01
4.01234537e-01 -7.98575044e-01 6.07919872e-01 9.16815281e-01
-6.44715279e-02 5.10529637e-01 4.08066995e-02 1.19577982e-01
2.73994118e-01 2.94994652e-01 2.30144292e-01 -1.10052204e+00
-1.65769994e-01 -5.70227206e-01 1.67542860e-01 -1.67438900e+00
5.33489227e-01 -6.94750011e-01 4.28945363e-01 -1.81659830e+00
2.95598298e-01 -3.57277185e-01 -6.30130887e-01 5.62590897e-01
-8.66698861e-01 2.14023009e-01 1.46881536e-01 -1.10171631e-01
-9.80038226e-01 3.90459865e-01 1.21262896e+00 4.40658033e-02
-3.65455300e-01 -1.56581670e-01 -1.03135026e+00 9.80157197e-01
1.00004053e+00 -3.65989834e-01 -2.69242734e-01 -9.01037037e-01
4.03207056e-02 -4.03482199e-01 -2.37232819e-01 -6.16556525e-01
1.47846062e-02 -4.43710089e-01 2.33715132e-01 -7.07581639e-01
-1.57677770e-01 -4.89159375e-01 -8.58012259e-01 5.71895897e-01
-3.29830527e-01 -3.81940067e-01 2.32456505e-01 4.89345729e-01
-4.57830191e-01 -4.75917727e-01 7.77621567e-01 -5.26438355e-01
-1.33495319e+00 1.58221081e-01 -4.34944153e-01 5.05584478e-01
5.75549543e-01 1.93719529e-02 -1.32562280e-01 -1.03474796e-01
-4.71607566e-01 5.92545748e-01 1.70165658e-01 6.13249362e-01
4.99144226e-01 -8.62997174e-01 -4.26400870e-01 1.67746052e-01
-1.22250751e-01 6.62664413e-01 -1.07232472e-02 3.93947303e-01
-5.46816111e-01 6.49622679e-01 1.84098363e-01 -3.43184114e-01
-8.18543077e-01 4.32518810e-01 -2.58864891e-02 -4.75383401e-01
-6.38334811e-01 1.19389296e+00 4.75098342e-01 -2.92361021e-01
1.02361426e-01 -8.02617669e-01 -4.15052205e-01 -2.12900154e-02
5.38933158e-01 -1.88833937e-01 -2.44247168e-01 -5.15078664e-01
-3.27208817e-01 6.53306782e-01 -2.33870074e-01 -8.38674009e-02
1.20532835e+00 -3.30228895e-01 -1.96521252e-01 2.06398457e-01
1.14445484e+00 2.88897101e-02 -1.17707396e+00 -5.97544551e-01
6.92316771e-01 -1.77148372e-01 1.12612262e-01 -8.30106318e-01
-1.09289443e+00 1.31751227e+00 -1.27766222e-01 -1.29574224e-01
1.20255971e+00 1.96704879e-01 1.54419804e+00 2.28386134e-01
1.72878847e-01 -1.46247661e+00 1.43905833e-01 9.75282013e-01
1.02853522e-01 -1.37181425e+00 -3.88888180e-01 -5.52258670e-01
-7.68859327e-01 1.00215507e+00 9.16688740e-01 -2.29050756e-01
6.17851853e-01 1.31062597e-01 1.98155150e-01 -2.18872055e-02
-3.61380905e-01 -6.30169928e-01 2.14128479e-01 4.23936903e-01
4.04573083e-01 2.36577332e-01 -6.38650537e-01 1.44763231e+00
9.34729576e-02 -3.95243138e-01 1.98924839e-01 9.67698097e-01
-9.40084934e-01 -1.33599138e+00 -5.89642972e-02 2.79381603e-01
-7.17487693e-01 -5.23434103e-01 -5.57851493e-01 5.93474090e-01
7.13913888e-02 8.65619481e-01 5.82366623e-02 -7.77680501e-02
-6.58486458e-03 2.16939569e-01 3.69377658e-02 -9.78598237e-01
-3.93082768e-01 4.28413987e-01 -1.35210082e-01 -5.12034535e-01
-2.45256945e-01 -6.26388848e-01 -2.07253933e+00 3.85823429e-01
-2.92145044e-01 3.70624840e-01 3.84836257e-01 1.39075530e+00
-2.05032937e-02 6.09824419e-01 2.84932643e-01 -2.33389735e-01
-8.35482255e-02 -9.86234009e-01 -3.27700168e-01 3.50507408e-01
1.26684889e-01 -1.70797497e-01 9.34038237e-02 -2.93661328e-03]
|
[9.955297470092773, 10.076038360595703]
|
843791fc-09ee-4203-b93a-e207b0aa79a1
|
what-makes-entities-similar-a-similarity
|
2306.02622
| null |
https://arxiv.org/abs/2306.02622v1
|
https://arxiv.org/pdf/2306.02622v1.pdf
|
What Makes Entities Similar? A Similarity Flooding Perspective for Multi-sourced Knowledge Graph Embeddings
|
Joint representation learning over multi-sourced knowledge graphs (KGs) yields transferable and expressive embeddings that improve downstream tasks. Entity alignment (EA) is a critical step in this process. Despite recent considerable research progress in embedding-based EA, how it works remains to be explored. In this paper, we provide a similarity flooding perspective to explain existing translation-based and aggregation-based EA models. We prove that the embedding learning process of these models actually seeks a fixpoint of pairwise similarities between entities. We also provide experimental evidence to support our theoretical analysis. We propose two simple but effective methods inspired by the fixpoint computation in similarity flooding, and demonstrate their effectiveness on benchmark datasets. Our work bridges the gap between recent embedding-based models and the conventional similarity flooding algorithm. It would improve our understanding of and increase our faith in embedding-based EA.
|
['Wei Hu', 'Weijun Ren', 'Qijin Chen', 'Xiaozhou Xu', 'Jiacheng Huang', 'Zequn Sun']
|
2023-06-05
| null | null | null | null |
['knowledge-graph-embeddings', 'entity-alignment', 'knowledge-graphs', 'knowledge-graph-embeddings', 'entity-alignment']
|
['graphs', 'knowledge-base', 'knowledge-base', 'methodology', 'natural-language-processing']
|
[-1.34294713e-02 2.38867074e-01 -6.10936165e-01 -2.13173106e-01
-6.64883316e-01 -7.20599115e-01 6.82916164e-01 6.17181599e-01
-3.70090365e-01 4.30545330e-01 6.38517737e-01 -6.23440444e-01
-5.96019208e-01 -1.17968500e+00 -6.81986392e-01 -3.58815968e-01
-5.59288681e-01 4.47364926e-01 3.93095911e-01 -4.37320381e-01
3.01063120e-01 3.66806239e-01 -1.00311828e+00 5.58581837e-02
8.06808889e-01 2.57181257e-01 -6.93884073e-03 5.84308922e-01
-3.41797918e-01 6.97585344e-01 -3.14044148e-01 -7.82550156e-01
2.71912158e-01 -3.49132776e-01 -1.15151155e+00 -6.33495092e-01
2.98233002e-01 -1.63151801e-01 -5.62134743e-01 9.49267089e-01
3.91468525e-01 5.97085990e-02 7.07006752e-01 -1.70256698e+00
-1.33345163e+00 1.10062373e+00 -4.28986222e-01 2.60593653e-01
8.65370259e-02 -2.77534008e-01 1.82726455e+00 -1.01046455e+00
9.17765021e-01 9.82165277e-01 1.04766560e+00 3.65336984e-01
-1.07074296e+00 -4.48096395e-01 -4.87622879e-02 4.95496094e-01
-1.37491405e+00 -2.08282962e-01 7.34657109e-01 -1.83370858e-01
1.25529754e+00 9.92843732e-02 7.94622362e-01 8.45354140e-01
2.04329304e-02 8.11430097e-01 6.16586864e-01 -6.85930967e-01
9.10480469e-02 2.01017991e-01 3.40264082e-01 9.66071665e-01
9.50391471e-01 4.74238358e-02 -7.67783880e-01 -3.77449065e-01
7.02110589e-01 -5.19783907e-02 -3.13009769e-01 -8.49300206e-01
-1.18912745e+00 1.25236928e+00 7.85112143e-01 5.20096004e-01
-8.93567353e-02 6.50570571e-01 3.08869869e-01 5.27738094e-01
5.11296272e-01 7.56918252e-01 -3.84115487e-01 -2.29294226e-01
-8.46261263e-01 1.59249991e-01 1.01461804e+00 9.60490763e-01
8.59825015e-01 -4.41037029e-01 1.95499361e-01 3.90125692e-01
4.74809140e-01 1.16540723e-01 3.06399226e-01 -8.30628872e-01
4.53231633e-01 7.86684275e-01 -1.34461790e-01 -1.32599902e+00
-1.33280113e-01 -3.90015543e-01 -3.55733037e-01 -2.31905729e-01
3.47319990e-01 1.28871426e-01 -2.48236999e-01 1.92983902e+00
2.31767684e-01 6.10448956e-01 3.48002374e-01 5.90429246e-01
4.66410160e-01 5.10408223e-01 -5.73668778e-02 2.67325640e-01
1.21547091e+00 -1.27924204e+00 -6.19792938e-01 -5.75265661e-02
1.43172169e+00 -5.38380861e-01 6.98830247e-01 -3.59189332e-01
-9.90070343e-01 1.08090322e-02 -1.10623765e+00 -3.27930570e-01
-8.24655354e-01 -2.48359412e-01 1.08788478e+00 6.71618104e-01
-1.41387761e+00 7.39127874e-01 -8.58617187e-01 -7.00413823e-01
4.09007877e-01 3.27434003e-01 -6.50222421e-01 -2.04134900e-02
-1.54573715e+00 1.14786911e+00 1.47616655e-01 -4.16294746e-02
-3.45010847e-01 -8.66957486e-01 -9.87543344e-01 9.44493786e-02
2.61063099e-01 -1.18960655e+00 8.42679322e-01 -3.99983585e-01
-9.77690041e-01 6.92620277e-01 -9.21404809e-02 -7.29304910e-01
1.70858473e-01 -2.60075152e-01 -1.08739346e-01 1.74350128e-01
1.11846216e-01 5.72863877e-01 3.60123992e-01 -1.11117625e+00
-3.91637802e-01 -2.05508217e-01 2.95696735e-01 1.74687922e-01
-8.40285718e-01 -1.92237142e-02 -1.43038213e-01 -4.28379267e-01
-8.76362845e-02 -8.15168619e-01 -1.74929917e-01 3.45685333e-01
-2.02866554e-01 -4.18351710e-01 5.00780761e-01 -2.45841324e-01
1.48687422e+00 -1.82517374e+00 4.79070365e-01 2.42387578e-01
6.20914280e-01 1.98806867e-01 -3.62372726e-01 1.16248190e+00
1.13903530e-01 6.48226678e-01 -2.78115630e-01 -4.38913435e-01
3.86257201e-01 3.54258895e-01 -5.71846068e-01 4.41081792e-01
3.79167944e-01 1.46292353e+00 -1.31334424e+00 -6.06357992e-01
-6.03664033e-02 5.81411839e-01 -6.90268517e-01 -4.33216197e-03
2.40290135e-01 -4.79968995e-01 -5.55336416e-01 6.46135569e-01
4.17546570e-01 -3.58209729e-01 3.84381115e-01 -3.82355869e-01
-2.08944306e-02 2.91173607e-01 -9.02681828e-01 1.97001231e+00
-2.97069162e-01 8.11076522e-01 -2.69272387e-01 -1.19029522e+00
6.99516892e-01 7.33814342e-03 4.17280972e-01 -3.15530688e-01
-1.39670417e-01 3.86015862e-01 -8.26172233e-02 -2.59728312e-01
8.92035365e-01 1.74203850e-02 -4.20635268e-02 7.66693652e-01
5.13835326e-02 1.38807490e-01 4.50124927e-02 7.99012363e-01
1.50512409e+00 1.93827942e-01 4.00079846e-01 -2.35023230e-01
8.78226236e-02 1.62460253e-01 2.13720515e-01 7.50167787e-01
-4.37757909e-01 2.18807176e-01 4.80501026e-01 -3.41962963e-01
-1.12389922e+00 -1.21883166e+00 -1.65223181e-02 9.21936572e-01
3.64678115e-01 -1.08813965e+00 -4.84473437e-01 -8.90938461e-01
3.64052504e-01 4.39790815e-01 -8.36573899e-01 -5.78146636e-01
-5.99517345e-01 -6.95316672e-01 9.51947808e-01 1.00378942e+00
1.05812028e-01 -4.40084189e-01 -3.90213370e-01 1.54259965e-01
-2.31149048e-01 -9.40543532e-01 -3.36767673e-01 2.16523454e-01
-9.69812810e-01 -1.17979765e+00 -4.65623856e-01 -8.82194042e-01
6.02151394e-01 5.21478236e-01 1.14117181e+00 2.78387457e-01
-1.83548212e-01 6.69123054e-01 -6.02643192e-01 -1.62360698e-01
-2.82924592e-01 3.99753809e-01 1.67398348e-01 -5.88835776e-01
8.02154899e-01 -8.61365199e-01 -5.60814142e-01 1.20717682e-01
-8.59662354e-01 -1.43265828e-01 6.96822405e-01 8.82711589e-01
2.97282010e-01 -2.84549296e-01 6.95575178e-01 -6.65303588e-01
9.71768141e-01 -8.57309163e-01 -1.83700845e-01 5.39862990e-01
-9.54623342e-01 5.04721463e-01 3.05254996e-01 -1.38577387e-01
-3.34302992e-01 -4.19785529e-01 1.66466177e-01 -3.94003898e-01
5.56528568e-01 7.50963151e-01 2.65887976e-01 -3.39296013e-01
6.03732824e-01 1.91458866e-01 -5.23818806e-02 -3.21937621e-01
1.01866710e+00 5.29620409e-01 3.21530432e-01 -8.99730682e-01
1.18908811e+00 5.36195815e-01 6.29362017e-02 -4.16487545e-01
-5.09915888e-01 -5.42995572e-01 -7.25816309e-01 2.63036549e-01
6.61216915e-01 -9.09906626e-01 -2.81126201e-01 -1.34523079e-01
-1.20758688e+00 -2.01752916e-01 -4.12334830e-01 4.37035531e-01
-4.34664458e-01 5.66155136e-01 -4.61227596e-01 -5.73706746e-01
-3.88270468e-01 -7.26140678e-01 1.07598162e+00 -2.34446060e-02
-3.38707328e-01 -1.63534033e+00 6.91750169e-01 9.77416486e-02
6.65975749e-01 8.64454806e-02 1.05007207e+00 -9.45998192e-01
-7.38668621e-01 -1.25742391e-01 -3.22409034e-01 -9.43389311e-02
1.70024827e-01 1.12700403e-01 -6.22821510e-01 -1.11997023e-01
-5.73233008e-01 -1.51312038e-01 9.93724287e-01 -9.04248729e-02
5.61878324e-01 -1.98697135e-01 -7.32883453e-01 5.98904073e-01
1.60058630e+00 -4.82982337e-01 5.41640103e-01 5.99778354e-01
6.32452130e-01 4.96731192e-01 3.48515034e-01 1.03383243e-01
8.05716634e-01 7.62284100e-01 2.96630353e-01 1.63535506e-01
-3.91995817e-01 -5.87507010e-01 4.34257150e-01 1.36364055e+00
-8.32184777e-02 -2.89156884e-01 -8.50587904e-01 1.16435361e+00
-1.98906362e+00 -1.10733724e+00 -1.00737475e-01 1.78493416e+00
9.76274788e-01 -2.16536924e-01 -8.08667839e-02 8.61109421e-02
6.17293537e-01 2.26475850e-01 -1.31447017e-01 -6.93711400e-01
-6.38964847e-02 4.08141732e-01 5.83372235e-01 6.17775977e-01
-8.36881578e-01 1.02616286e+00 6.37368202e+00 5.68513691e-01
-4.46867824e-01 2.08831832e-01 -2.06700057e-01 -6.12126011e-03
-8.71329963e-01 3.49926949e-01 -7.82303631e-01 1.32049337e-01
9.94006574e-01 -7.00524449e-01 4.08843368e-01 6.98759496e-01
-4.44055229e-01 3.08832794e-01 -1.40250933e+00 6.30715191e-01
3.29163186e-02 -1.72249663e+00 2.08141431e-01 1.79202199e-01
6.39835060e-01 2.03459263e-01 5.14048524e-02 3.26952249e-01
7.80148029e-01 -8.54278803e-01 3.24086905e-01 2.41942659e-01
5.52971840e-01 -6.44850731e-01 7.34641969e-01 -7.76428729e-02
-1.70092082e+00 1.29943965e-02 -5.61196446e-01 2.31480263e-02
2.47386947e-01 4.96175766e-01 -9.54620004e-01 1.10634875e+00
2.48150110e-01 9.51438427e-01 -5.47493994e-01 7.79910028e-01
-4.29949373e-01 6.98634684e-01 -4.03788716e-01 -1.62376016e-01
5.44359446e-01 -1.55190766e-01 5.72100461e-01 1.40031278e+00
3.79161954e-01 -3.08381766e-01 -1.60373881e-01 9.68856752e-01
-3.98840308e-01 1.42159939e-01 -1.08019698e+00 -3.69573712e-01
9.57542181e-01 1.19658828e+00 -5.24621427e-01 -1.29826054e-01
-5.34556150e-01 1.03593206e+00 8.39212596e-01 1.45166516e-01
-9.30817902e-01 -7.45167792e-01 9.92431581e-01 -1.49244651e-01
4.69075978e-01 -5.13816714e-01 -1.17083400e-01 -1.28534889e+00
2.83404384e-02 -3.17472786e-01 4.29204166e-01 -4.37418252e-01
-1.43782198e+00 2.30736464e-01 6.41176030e-02 -8.90297949e-01
-1.34655744e-01 -5.21033227e-01 -7.29516983e-01 5.32018840e-01
-1.98468685e+00 -1.24548900e+00 1.40677309e-02 3.74022454e-01
1.32834613e-01 9.77522228e-03 1.17463899e+00 2.92865455e-01
-4.14497256e-01 7.84935057e-01 3.53628159e-01 3.84631127e-01
7.60830224e-01 -1.49735534e+00 7.96244740e-01 8.15911591e-01
7.49378860e-01 1.17516267e+00 4.99711066e-01 -6.91640437e-01
-1.89907444e+00 -1.20127118e+00 1.31365883e+00 -9.43691015e-01
1.31167388e+00 -1.66283965e-01 -8.52704525e-01 9.97996628e-01
3.38914990e-01 2.57987708e-01 1.09584403e+00 4.56075072e-01
-8.30572844e-01 1.66302323e-01 -9.33988512e-01 6.51482582e-01
1.62244546e+00 -7.75650799e-01 -1.08983338e+00 2.30150029e-01
9.55798924e-01 7.58619234e-02 -1.17950475e+00 2.67466247e-01
6.91324770e-01 -5.62585473e-01 1.19603372e+00 -1.09901476e+00
5.04316270e-01 -2.06468701e-01 -3.95635784e-01 -1.52585900e+00
-3.38463366e-01 -6.42522693e-01 -5.42832136e-01 1.10785532e+00
4.24365461e-01 -9.43865359e-01 6.98831618e-01 3.08566213e-01
-6.25315011e-02 -9.51357245e-01 -8.77090871e-01 -1.22686946e+00
5.66609323e-01 -1.91006571e-01 6.02928817e-01 1.28227222e+00
5.63263178e-01 1.21394381e-01 -8.78609419e-02 2.68811017e-01
7.47319400e-01 2.01927587e-01 7.09349453e-01 -1.28022957e+00
-1.84161961e-01 -5.28298855e-01 -8.21223378e-01 -8.15914094e-01
3.64854187e-01 -1.45564461e+00 -6.03655696e-01 -1.96906757e+00
3.04403216e-01 -6.14833653e-01 -4.15182024e-01 6.87021196e-01
-1.92802832e-01 1.94208935e-01 -3.97623843e-03 2.58055657e-01
-7.23603964e-01 6.43867493e-01 6.50998771e-01 -9.69421044e-02
1.16868347e-01 -7.52066374e-01 -1.04023921e+00 3.79597098e-01
8.66903484e-01 -7.02945352e-01 -6.88101172e-01 -6.81298673e-01
8.00053596e-01 -7.10970581e-01 4.03491765e-01 -4.80367333e-01
6.64881170e-01 4.44081537e-02 -2.32084990e-01 -6.07543364e-02
1.02479517e-01 -7.64173925e-01 -1.34857714e-01 5.29597640e-01
-5.11700809e-01 2.88647413e-01 -1.80154741e-02 9.72515285e-01
-5.32207713e-02 -2.42094472e-01 9.30043012e-02 1.16630651e-01
-7.99268365e-01 2.56729394e-01 1.89499676e-01 2.90755630e-01
1.09007955e+00 -1.56241491e-01 -7.41363525e-01 -2.22790495e-01
-4.28809285e-01 2.28354529e-01 4.97536272e-01 4.11727250e-01
6.49036050e-01 -1.70117640e+00 -8.42874408e-01 -1.20900348e-01
3.76712084e-01 -3.32256883e-01 -3.87514681e-01 9.49429929e-01
-3.75437468e-01 4.22972322e-01 1.06975898e-01 -1.48207530e-01
-1.10465455e+00 5.08549154e-01 1.30861536e-01 -3.97708386e-01
-5.60716212e-01 8.46430421e-01 -2.09531531e-01 -4.55122769e-01
1.32695213e-02 -2.93771476e-01 7.09813163e-02 -1.76611710e-02
4.16122705e-01 5.39471149e-01 -2.54416108e-01 -3.61298233e-01
-6.15083098e-01 8.04145753e-01 -2.17045560e-01 -8.22778791e-02
1.52521110e+00 -4.93109263e-02 -2.32479602e-01 2.94749349e-01
1.56791079e+00 3.35491061e-01 -6.27641618e-01 -4.41463232e-01
2.69489408e-01 -4.42548990e-01 -1.94396591e-04 -4.33535099e-01
-7.24086761e-01 9.32466030e-01 1.51835993e-01 3.43887210e-01
7.11654484e-01 2.61125982e-01 9.80430901e-01 7.03443229e-01
4.71247017e-01 -8.16920996e-01 7.13894889e-02 2.99306542e-01
4.62950587e-01 -9.75941896e-01 1.26275435e-01 -3.88617665e-01
-3.09816748e-01 1.22896159e+00 2.63587683e-01 -2.10540876e-01
6.34188354e-01 3.67261410e-01 -4.09875333e-01 -5.30371547e-01
-9.57780778e-01 -3.82308155e-01 4.03674506e-02 6.96513951e-01
2.84648836e-01 -6.69037923e-02 -4.61169600e-01 6.10631585e-01
-1.42006710e-01 -1.36883464e-02 3.63455445e-01 1.08310306e+00
-6.10515773e-01 -1.50231850e+00 2.64754027e-01 -5.03514856e-02
-1.97242290e-01 -4.98229682e-01 -6.95306301e-01 1.00814164e+00
-1.57256827e-01 6.47160709e-01 -8.35409686e-02 -6.11251235e-01
2.39759181e-02 2.97643006e-01 5.82381666e-01 -5.37112653e-01
-4.93197888e-01 -8.17215741e-01 1.86548397e-01 -5.11345625e-01
-5.33918858e-01 -4.19837326e-01 -1.33976150e+00 -5.82034171e-01
-6.55762494e-01 4.40807939e-01 7.20827818e-01 6.34952545e-01
7.56345868e-01 2.38720357e-01 6.26449645e-01 -1.97610274e-01
-7.72063076e-01 -5.68351328e-01 -3.53711903e-01 2.00642094e-01
2.25419998e-02 -6.13506138e-01 -6.26449823e-01 -2.53433198e-01]
|
[8.759212493896484, 7.875870704650879]
|
3d76edcb-b7cf-4786-a259-bf6e6e9e1f42
|
do-saliency-models-detect-odd-one-out-targets
|
2005.06583
| null |
https://arxiv.org/abs/2005.06583v2
|
https://arxiv.org/pdf/2005.06583v2.pdf
|
Do Saliency Models Detect Odd-One-Out Targets? New Datasets and Evaluations
|
Recent advances in the field of saliency have concentrated on fixation prediction, with benchmarks reaching saturation. However, there is an extensive body of works in psychology and neuroscience that describe aspects of human visual attention that might not be adequately captured by current approaches. Here, we investigate singleton detection, which can be thought of as a canonical example of salience. We introduce two novel datasets, one with psychophysical patterns and one with natural odd-one-out stimuli. Using these datasets we demonstrate through extensive experimentation that nearly all saliency algorithms do not adequately respond to singleton targets in synthetic and natural images. Furthermore, we investigate the effect of training state-of-the-art CNN-based saliency models on these types of stimuli and conclude that the additional training data does not lead to a significant improvement of their ability to find odd-one-out targets. Datasets are available at http://data.nvision2.eecs.yorku.ca/P3O3/.
|
['Iuliia Kotseruba', 'John K. Tsotsos', 'Amir Rasouli', 'Calden Wloka']
|
2020-05-13
| null | null | null | null |
['odd-one-out']
|
['reasoning']
|
[ 5.01042485e-01 -1.84621066e-01 -7.16888830e-02 -8.91739279e-02
-3.33928406e-01 -2.26318553e-01 5.49531460e-01 2.75396317e-01
-4.80095714e-01 6.79693878e-01 5.20270057e-02 -2.96002239e-01
1.87867269e-01 -3.17627907e-01 -7.24111617e-01 -3.85496616e-01
-5.51035143e-02 -5.36705926e-02 8.23654175e-01 -3.44340652e-01
6.80672646e-01 3.02195728e-01 -2.11748075e+00 4.16842014e-01
7.41295338e-01 8.88927996e-01 4.81595725e-01 5.87911367e-01
3.29715341e-01 6.18976653e-01 -6.00645542e-01 -7.10787401e-02
3.72254342e-01 -5.36635160e-01 -6.73253179e-01 -1.65184334e-01
7.17399061e-01 -5.29190525e-02 -3.25882018e-01 1.04922402e+00
6.73376501e-01 1.84688628e-01 4.14143354e-01 -1.24675071e+00
-8.31408381e-01 3.47659916e-01 -6.42183304e-01 1.02969885e+00
4.00057346e-01 5.55384815e-01 9.75369513e-01 -1.05079198e+00
5.07749319e-01 1.14660561e+00 2.27055103e-01 4.37363178e-01
-1.37286592e+00 -4.82138187e-01 -1.03476727e-02 4.29929018e-01
-1.25385165e+00 -6.36927485e-01 8.70808423e-01 -1.91944420e-01
8.98705006e-01 4.53244746e-01 8.77780139e-01 9.53883410e-01
3.28760557e-02 8.66689384e-01 1.42821264e+00 -6.23434484e-01
1.46996289e-01 8.02253932e-02 -1.31394304e-02 1.87528223e-01
5.37341833e-01 3.71560544e-01 -7.37081409e-01 2.22738892e-01
1.01983249e+00 -2.29285732e-01 -4.35305446e-01 -5.40349841e-01
-1.38602865e+00 5.82488537e-01 8.44079316e-01 3.54477525e-01
-3.19045275e-01 3.39887813e-02 -1.23009877e-03 8.13719854e-02
2.46868402e-01 8.52741957e-01 4.47156429e-02 6.60862029e-02
-1.08559167e+00 4.46676910e-01 3.31474274e-01 9.95193660e-01
7.33656108e-01 3.25560093e-01 -4.07098860e-01 6.05790436e-01
-1.42200828e-01 6.16447628e-01 4.07737583e-01 -8.53496790e-01
-1.71882603e-02 3.85801703e-01 5.28890908e-01 -1.08814800e+00
-5.89943528e-01 -4.13859546e-01 -2.15689361e-01 4.57642734e-01
7.36379325e-01 7.24867284e-02 -9.21863496e-01 1.77464902e+00
-1.30155802e-01 6.80115968e-02 -3.09661061e-01 1.26491177e+00
7.95749187e-01 1.33280098e-01 1.59422666e-01 -9.32431296e-02
1.36104667e+00 -1.00837231e+00 -5.03816664e-01 -6.40727937e-01
1.21462144e-01 -8.72603178e-01 1.27598751e+00 2.62457669e-01
-1.43126357e+00 -7.14251757e-01 -1.08831263e+00 -1.59802824e-01
-4.32907522e-01 -1.31681010e-01 7.43166924e-01 4.25311804e-01
-1.22172868e+00 3.68095309e-01 -4.95561481e-01 -6.56393111e-01
6.98990643e-01 1.96579874e-01 2.45073959e-02 7.08067417e-02
-1.18488085e+00 1.26372230e+00 3.28103215e-01 -1.66373000e-01
-9.97663558e-01 -5.67121983e-01 -7.32941866e-01 2.21896693e-01
4.52535659e-01 -4.61122632e-01 1.39720094e+00 -1.39203191e+00
-6.31917655e-01 1.09784591e+00 -5.05615056e-01 -5.04782677e-01
1.17514931e-01 -2.27164347e-02 -5.54543734e-01 2.88002402e-01
1.45255208e-01 1.23488140e+00 8.16651225e-01 -1.32542467e+00
-6.64251268e-01 -2.09486216e-01 2.77471721e-01 2.30865791e-01
-4.53335159e-02 2.83638239e-01 -1.19009847e-02 -6.84823334e-01
-7.47533143e-02 -8.61411273e-01 -9.54937264e-02 -4.65731509e-02
-3.21085960e-01 -1.15212895e-01 3.29548031e-01 -8.29741433e-02
1.18540835e+00 -2.10058093e+00 -1.44501671e-01 -1.96367502e-01
1.40874118e-01 4.16239500e-01 -1.86182007e-01 3.05020332e-01
-2.27128908e-01 1.26091167e-01 -1.85623929e-01 -2.12720320e-01
-2.35533947e-03 -2.11169228e-01 -3.04719508e-01 4.67332453e-01
5.62253058e-01 1.10170853e+00 -1.05495334e+00 -3.15267324e-01
9.48149413e-02 2.39784867e-01 -3.67418081e-01 -1.56935632e-01
-1.65505514e-01 3.41886938e-01 8.85216147e-03 7.70502388e-01
6.00941420e-01 -3.77664417e-01 -8.19384232e-02 5.93884774e-02
-4.59779948e-01 4.35793281e-01 -8.41797173e-01 1.57368636e+00
2.19678357e-01 1.00425208e+00 -3.15990716e-01 -7.68334746e-01
6.63208723e-01 -4.76593245e-03 -4.80188839e-02 -1.13623703e+00
3.65443200e-01 3.24228674e-01 6.99337244e-01 -2.26728112e-01
8.21915627e-01 7.29897767e-02 3.03746134e-01 2.55586356e-01
-1.46016469e-02 -8.52284804e-02 6.14543378e-01 3.71837355e-02
8.51530254e-01 -1.46249458e-01 2.00603604e-01 -6.68012798e-01
2.45020613e-01 2.17392787e-01 3.30288321e-01 1.02317095e+00
-5.61946392e-01 8.62492442e-01 3.54156464e-01 -2.93479294e-01
-1.14870393e+00 -1.07258749e+00 -2.54315466e-01 1.22635567e+00
4.73686606e-01 -1.13615170e-01 -7.79598653e-01 -9.75305140e-02
-1.23200998e-01 8.67221892e-01 -9.05080497e-01 -2.19738469e-01
-2.46773511e-01 -5.82999945e-01 3.34483445e-01 4.83209580e-01
2.84575462e-01 -1.74207759e+00 -1.66906297e+00 -1.72568068e-01
-1.18083991e-01 -9.51586246e-01 -3.41091782e-01 1.55761957e-01
-7.18918622e-01 -1.12443852e+00 -9.04616535e-01 -8.21278036e-01
6.48318946e-01 8.85438442e-01 1.29114473e+00 3.61016959e-01
-4.76509959e-01 2.13219583e-01 -3.27353656e-01 -8.97008836e-01
5.93266822e-02 6.84855953e-02 9.84625518e-02 -1.87814832e-01
8.50503564e-01 -3.40524197e-01 -9.29644048e-01 3.50010067e-01
-9.68022406e-01 2.01550856e-01 6.12883627e-01 5.42996168e-01
3.72222662e-01 -5.61491609e-01 6.35814250e-01 -6.03203952e-01
5.16862512e-01 -4.42227453e-01 -4.66073960e-01 -8.21756274e-02
-3.87031674e-01 -1.28203139e-01 2.99710691e-01 -4.19519424e-01
-6.26366794e-01 -3.53795409e-01 3.92763168e-01 -5.07305205e-01
-4.65245813e-01 3.56739849e-01 3.49582195e-01 -1.48537874e-01
1.00783122e+00 2.30850413e-01 -3.02231610e-01 -7.61406496e-02
-3.81007716e-02 2.84476846e-01 5.27042806e-01 -1.77860320e-01
5.89843392e-01 5.57858586e-01 -6.50912384e-03 -8.52381647e-01
-9.84396935e-01 -3.59008074e-01 -5.95459700e-01 -2.16528371e-01
6.60439014e-01 -6.94922090e-01 -5.49204290e-01 5.03800988e-01
-8.87217462e-01 -4.85671073e-01 -3.72367263e-01 4.51751083e-01
-6.93535626e-01 3.33666764e-02 -9.29556638e-02 -8.22631061e-01
1.10092722e-02 -1.13660336e+00 7.39185333e-01 6.79855347e-01
-3.92943621e-01 -5.70420027e-01 -2.00683519e-01 9.58448928e-03
8.32948506e-01 1.13574930e-01 6.26689732e-01 -5.84832668e-01
-6.69195890e-01 1.88791782e-01 -3.40969384e-01 -8.20131674e-02
7.95026943e-02 -2.93497071e-02 -1.20775414e+00 -2.95549184e-01
-6.29945621e-02 -5.00652432e-01 1.08963668e+00 7.41601884e-01
9.13640261e-01 2.49846563e-01 -2.66485453e-01 3.69016945e-01
1.25061834e+00 1.55648604e-01 7.97096848e-01 4.71000046e-01
2.67876983e-01 6.49977446e-01 6.81953192e-01 2.66208619e-01
2.77898163e-01 6.38592064e-01 6.39059961e-01 -2.07120240e-01
-3.67974252e-01 -1.61778063e-01 -7.17629641e-02 9.30760130e-02
-1.88521892e-01 -2.33899832e-01 -1.01235569e+00 9.73872602e-01
-1.64423919e+00 -1.28253889e+00 -1.92820936e-01 2.37323737e+00
7.61789262e-01 4.65362251e-01 3.51981521e-01 1.63169831e-01
9.36382532e-01 3.08480293e-01 -6.99631989e-01 -4.84659940e-01
-4.71324295e-01 2.40186825e-01 3.39279473e-01 1.37267798e-01
-1.15561867e+00 1.02598584e+00 6.72907734e+00 4.96248752e-01
-1.25090408e+00 -1.57800123e-01 6.20305896e-01 -3.80425423e-01
-3.55834872e-01 1.02901146e-01 -5.58871031e-01 4.92393970e-01
8.71161640e-01 -2.95722574e-01 4.89223003e-01 4.70217168e-01
1.60670355e-01 -7.67926574e-01 -9.73885000e-01 8.12829614e-01
2.57149160e-01 -1.11226773e+00 -2.20690802e-01 -1.62503734e-01
7.53690779e-01 6.13102540e-02 5.69912910e-01 9.32688415e-02
-5.53163327e-02 -1.30086470e+00 9.06184554e-01 5.46483994e-01
7.16434538e-01 -4.17479485e-01 3.67356360e-01 2.17527837e-01
-6.75843239e-01 -7.60150477e-02 -6.06525242e-01 -4.71420437e-01
-6.86780959e-02 1.71802193e-01 -5.72573543e-01 4.92429435e-02
9.88360524e-01 6.07315004e-01 -1.13312590e+00 1.56595218e+00
-2.05331534e-01 4.60177839e-01 -1.84393749e-01 -3.91138583e-01
1.84013367e-01 2.99131423e-01 5.76452732e-01 1.05697441e+00
1.71850637e-01 9.01964903e-02 -1.54573128e-01 1.14937651e+00
7.04055429e-02 -1.67356476e-01 -5.79972982e-01 8.99308100e-02
6.34388268e-01 1.05992627e+00 -9.51865077e-01 -1.47289440e-01
-5.11604488e-01 6.27857029e-01 3.13377678e-01 4.49545175e-01
-8.49308014e-01 -3.81582022e-01 4.44096923e-01 2.76241422e-01
6.64391160e-01 -1.10994980e-01 -3.85281384e-01 -9.64137197e-01
-1.99069023e-01 -8.04762363e-01 -3.88400406e-02 -1.30516505e+00
-1.00656116e+00 4.61847812e-01 8.80554542e-02 -9.63344634e-01
2.83585135e-02 -5.23110032e-01 -6.48444951e-01 1.01593637e+00
-1.75382650e+00 -7.12140143e-01 -4.90519732e-01 5.51729441e-01
5.37428975e-01 1.86299071e-01 5.89742959e-01 -8.88995156e-02
-2.89527178e-01 4.86032844e-01 -2.18422264e-01 -1.07883357e-01
8.51809621e-01 -1.17341471e+00 5.94784379e-01 1.08954096e+00
2.01302707e-01 5.55472791e-01 1.10011232e+00 -4.24589008e-01
-9.84228492e-01 -5.71503878e-01 8.26038837e-01 -6.41451836e-01
4.77203906e-01 -3.47340882e-01 -9.41225767e-01 5.54560304e-01
5.70140600e-01 1.63088202e-01 4.12915558e-01 -2.36542039e-02
-2.05333665e-01 2.99365848e-01 -9.90776122e-01 9.26808596e-01
1.20661628e+00 -2.62667835e-01 -8.38396788e-01 3.67948189e-02
4.86615628e-01 -4.38671142e-01 -2.19149455e-01 4.75994110e-01
5.01970291e-01 -1.42071211e+00 9.65859532e-01 -6.37837768e-01
6.48804545e-01 -2.98037380e-01 -6.74029142e-02 -1.29141951e+00
-6.24274135e-01 -2.45849565e-01 -3.28793041e-02 7.67849684e-01
3.90397877e-01 -3.93573046e-01 4.77976114e-01 3.07285428e-01
-2.95397401e-01 -5.83413899e-01 -8.41854572e-01 -6.49393797e-01
-2.51188083e-03 -1.13337971e-01 3.02870482e-01 6.12886488e-01
5.34482300e-02 2.38411918e-01 -1.29258558e-01 -8.38962197e-02
5.32990515e-01 2.89089680e-01 5.50413251e-01 -1.15922832e+00
7.66417161e-02 -6.55929565e-01 -4.27962333e-01 -8.59237254e-01
-9.00226906e-02 -5.84524095e-01 1.48702681e-01 -1.40980780e+00
1.69457689e-01 -8.48681852e-02 -5.95817387e-01 4.01459843e-01
-3.61746371e-01 5.66223741e-01 4.11859244e-01 1.82311073e-01
-6.84853613e-01 3.78233582e-01 1.37108934e+00 1.14673637e-01
6.76726699e-02 -1.57701984e-01 -9.96365786e-01 5.89398682e-01
1.06143582e+00 -3.19289237e-01 -3.77357513e-01 -4.58778411e-01
-8.66726637e-02 -2.31878638e-01 7.68525243e-01 -1.38407147e+00
2.53960013e-01 -2.84938782e-01 5.91466725e-01 -3.84487778e-01
4.84750092e-01 -4.68247771e-01 -3.92722934e-01 4.64114249e-01
-4.61539596e-01 3.27287674e-01 7.07965612e-01 3.51240188e-01
4.01907377e-02 -3.74670029e-01 9.43588674e-01 -2.86505044e-01
-9.62523222e-01 -6.46045208e-02 -4.09637749e-01 4.55743790e-01
8.92873466e-01 -5.42897105e-01 -7.57228911e-01 -3.17282826e-01
-5.76519430e-01 1.54426187e-01 8.31615865e-01 5.93822598e-01
5.72841465e-01 -1.14256907e+00 -6.34569645e-01 2.51852661e-01
1.75531387e-01 -4.13586795e-01 1.97792619e-01 1.11101663e+00
-3.30944657e-01 7.46709883e-01 -7.58625746e-01 -6.01043046e-01
-1.06303477e+00 9.90275502e-01 1.84123427e-01 3.30246598e-01
-7.51675144e-02 8.42575908e-01 2.47339740e-01 2.63324499e-01
9.46039110e-02 -2.43282557e-01 -3.18458885e-01 -6.03026375e-02
5.84926367e-01 2.01536313e-01 -1.70443907e-01 -7.49597847e-01
-4.86991853e-01 2.37798229e-01 5.14264703e-02 -1.24243289e-01
9.96899486e-01 -9.09743905e-02 1.62742093e-01 7.88040340e-01
5.00006676e-01 -4.71329056e-02 -1.31177628e+00 -1.20075628e-01
-2.87681241e-02 -8.33422422e-01 -1.94431514e-01 -7.76298046e-01
-7.30777144e-01 1.09758842e+00 6.05650663e-01 5.09644151e-01
1.20418835e+00 -5.13355881e-02 3.10794920e-01 7.01308548e-02
3.54276627e-01 -8.42018127e-01 2.02853531e-01 6.00272477e-01
9.38407898e-01 -1.54466856e+00 -1.65975496e-01 -1.56355232e-01
-8.16311479e-01 6.89339221e-01 9.60582495e-01 -3.93200308e-01
3.03031117e-01 -6.65461794e-02 -1.03808410e-01 -2.30360419e-01
-7.87146270e-01 -8.37866187e-01 3.77884179e-01 7.20050097e-01
4.93402451e-01 -1.26562431e-01 -2.20713943e-01 2.05078751e-01
-2.57785767e-01 7.68209342e-03 8.68472993e-01 9.84398723e-01
-9.12297606e-01 -4.52804148e-01 -3.76652658e-01 5.83105803e-01
-4.74894434e-01 -6.08763516e-01 -4.11681503e-01 8.54762971e-01
-1.08547963e-01 9.92083311e-01 3.52154911e-01 -1.47885248e-01
3.44119310e-01 -3.80539219e-03 7.26364434e-01 -8.58199477e-01
-4.89181906e-01 8.24892987e-03 -1.83628783e-01 -3.54173750e-01
-6.44967675e-01 -8.58034968e-01 -1.05277765e+00 -2.91117996e-01
-1.90808505e-01 -5.54198166e-03 2.11248949e-01 5.80813408e-01
3.82394671e-01 6.56637847e-01 1.17090881e-01 -1.22037625e+00
-2.74611682e-01 -1.00149775e+00 -6.08236015e-01 4.96929079e-01
4.62599993e-01 -1.03079009e+00 -4.87094164e-01 -3.69725823e-02]
|
[10.04655647277832, 1.6622695922851562]
|
110b2150-da71-4ad5-a989-5ccff4f0571a
|
unsupervised-learning-of-object-landmarks-by
|
1705.02193
| null |
http://arxiv.org/abs/1705.02193v2
|
http://arxiv.org/pdf/1705.02193v2.pdf
|
Unsupervised learning of object landmarks by factorized spatial embeddings
|
Learning automatically the structure of object categories remains an
important open problem in computer vision. In this paper, we propose a novel
unsupervised approach that can discover and learn landmarks in object
categories, thus characterizing their structure. Our approach is based on
factorizing image deformations, as induced by a viewpoint change or an object
deformation, by learning a deep neural network that detects landmarks
consistently with such visual effects. Furthermore, we show that the learned
landmarks establish meaningful correspondences between different object
instances in a category without having to impose this requirement explicitly.
We assess the method qualitatively on a variety of object types, natural and
man-made. We also show that our unsupervised landmarks are highly predictive of
manually-annotated landmarks in face benchmark datasets, and can be used to
regress these with a high degree of accuracy.
|
['Hakan Bilen', 'Andrea Vedaldi', 'James Thewlis']
|
2017-05-05
|
unsupervised-learning-of-object-landmarks-by-1
|
http://openaccess.thecvf.com/content_iccv_2017/html/Thewlis_Unsupervised_Learning_of_ICCV_2017_paper.html
|
http://openaccess.thecvf.com/content_ICCV_2017/papers/Thewlis_Unsupervised_Learning_of_ICCV_2017_paper.pdf
|
iccv-2017-10
|
['unsupervised-facial-landmark-detection']
|
['computer-vision']
|
[ 3.00426246e-03 2.19608799e-01 -7.81302005e-02 -7.30641484e-01
-5.09501100e-01 -8.17857504e-01 9.37667966e-01 2.14635625e-01
-3.53922665e-01 2.32888147e-01 2.61082828e-01 2.52677917e-01
-1.47495344e-01 -6.13265514e-01 -1.01862407e+00 -6.79455578e-01
-4.29545864e-02 6.05272233e-01 2.14535952e-01 8.86435211e-02
3.92581195e-01 9.03819501e-01 -1.40895045e+00 -5.67195788e-02
3.79713476e-01 7.20792532e-01 -1.82147935e-01 4.52733517e-01
9.28582326e-02 4.00120527e-01 -3.61675680e-01 -2.70198017e-01
5.24831533e-01 -3.19500566e-01 -9.61528897e-01 4.55601066e-01
9.79642928e-01 -7.96689466e-02 -9.99008864e-02 1.15838349e+00
1.42387569e-01 2.12884411e-01 1.06286001e+00 -1.07080722e+00
-7.58471131e-01 3.06113064e-01 -5.31476557e-01 -2.53338180e-03
2.29546517e-01 -9.66638848e-02 1.20139897e+00 -1.11536431e+00
7.75463343e-01 1.28641510e+00 6.48816884e-01 5.18965781e-01
-1.74861860e+00 -1.90198004e-01 4.55323495e-02 7.45405108e-02
-1.37264359e+00 -6.23842001e-01 1.12346756e+00 -9.16781068e-01
4.50611621e-01 8.93513858e-02 3.51662338e-01 6.50872409e-01
5.45577034e-02 2.61490494e-01 1.01169646e+00 -5.61475694e-01
3.44148725e-01 2.37899050e-02 2.97470719e-01 9.49118257e-01
2.20996693e-01 1.74919635e-01 -2.64596254e-01 6.63256943e-02
9.16227818e-01 7.44820237e-02 4.50139381e-02 -8.82667482e-01
-1.26662445e+00 8.18595529e-01 8.74458253e-01 3.93849492e-01
-2.53466606e-01 3.75806630e-01 -1.79382581e-02 9.22311842e-02
3.93530637e-01 6.11397147e-01 -3.65099221e-01 4.30973113e-01
-8.16945195e-01 5.41094020e-02 6.19676471e-01 8.78799617e-01
1.17543972e+00 -2.18834858e-02 7.53865689e-02 7.03321397e-01
3.11150819e-01 3.94450903e-01 3.61784667e-01 -9.62220490e-01
-2.39782691e-01 6.27570868e-01 -1.06109865e-01 -1.21133244e+00
-5.84653199e-01 -8.83175805e-02 -6.62401438e-01 4.30470079e-01
6.34597957e-01 2.79324025e-01 -1.18957806e+00 1.96836531e+00
2.33740941e-01 1.91325441e-01 -2.62283534e-01 7.22158670e-01
6.17212176e-01 1.63460091e-01 1.00894906e-01 -9.98593774e-03
1.25498545e+00 -6.06199503e-01 -4.42178607e-01 -3.83163765e-02
3.48633438e-01 -6.77743196e-01 1.05473936e+00 1.47557259e-01
-9.52825725e-01 -6.82743073e-01 -7.47310400e-01 -2.31178388e-01
-3.96148235e-01 1.90804482e-01 6.35405838e-01 3.99937809e-01
-1.26954734e+00 7.12860882e-01 -1.01052833e+00 -6.07615113e-01
5.16473830e-01 6.52745247e-01 -6.27138436e-01 3.04486275e-01
-3.99330437e-01 8.96659493e-01 1.83927104e-01 1.38299400e-02
-1.04175138e+00 -5.75882375e-01 -9.42203403e-01 -2.29326878e-02
7.34960362e-02 -4.23843056e-01 9.94090974e-01 -1.25388408e+00
-1.33616841e+00 1.30074322e+00 -3.09998959e-01 -3.43232512e-01
3.78594160e-01 -6.23833425e-02 -4.96504642e-02 2.04897940e-01
1.02491572e-01 9.01815355e-01 1.01734376e+00 -1.53086722e+00
-3.84677321e-01 -5.85900784e-01 2.52918750e-01 -1.65198952e-01
-2.67927855e-01 4.68272343e-02 -2.10802987e-01 -6.12098277e-01
4.84054565e-01 -1.17382848e+00 -3.13970238e-01 2.58431882e-01
-5.05875707e-01 -3.80376577e-01 7.00313687e-01 -4.23762232e-01
3.97341639e-01 -1.90094554e+00 5.02737522e-01 4.67776299e-01
4.28152144e-01 -8.05762261e-02 -2.17144057e-01 9.72445831e-02
-1.14700563e-01 2.03835458e-01 -3.38514537e-01 -2.40421325e-01
6.09852746e-02 3.03730607e-01 -2.64072746e-01 7.73049653e-01
3.99649411e-01 1.03285456e+00 -8.39205027e-01 -2.97644019e-01
1.81414798e-01 4.50672418e-01 -6.96849883e-01 1.78153828e-01
-1.85002387e-01 7.42475092e-01 -2.67404288e-01 5.38778901e-01
5.14873266e-01 -4.82738279e-02 1.24744624e-01 -3.22262138e-01
1.40833586e-01 1.95792198e-01 -9.82499123e-01 1.62709069e+00
-3.77955049e-01 6.74154937e-01 -2.16807827e-01 -1.28528643e+00
8.84835124e-01 2.75621265e-02 5.52889347e-01 -3.64255637e-01
2.03663751e-01 3.86287421e-02 4.50562946e-02 -3.96206111e-01
1.38811693e-01 -1.67535871e-01 1.41438797e-01 5.02537251e-01
4.69887704e-01 -9.86057594e-02 1.28880247e-01 -5.08046895e-02
8.03123891e-01 -5.28550595e-02 2.20287487e-01 -6.10183716e-01
5.03976285e-01 -3.45405132e-01 5.11593044e-01 4.47678387e-01
-5.41023389e-02 6.04114532e-01 4.25330579e-01 -6.60654843e-01
-1.09032023e+00 -1.35395467e+00 -4.20689195e-01 1.04419196e+00
1.04813203e-01 -6.35808930e-02 -7.88865387e-01 -9.05117452e-01
1.88160956e-01 2.70692706e-01 -9.45481420e-01 -1.82921216e-01
-5.66129088e-01 -3.86730582e-01 3.29675525e-01 6.74249411e-01
2.32048720e-01 -1.08753622e+00 -4.25568610e-01 -1.52299330e-01
1.14578001e-01 -1.18084466e+00 -5.13567209e-01 5.99306077e-02
-7.99292803e-01 -1.35478818e+00 -3.22776198e-01 -1.08745313e+00
1.34420335e+00 1.87711388e-01 1.16112292e+00 2.62412310e-01
-4.45434868e-01 6.78268850e-01 -5.16454056e-02 -2.27554485e-01
-3.29792500e-01 -1.12586536e-01 3.92214537e-01 3.38715851e-01
3.91741574e-01 -7.75342047e-01 -3.16786498e-01 4.32950646e-01
-8.48134220e-01 -2.85764813e-01 4.00917828e-01 7.62305558e-01
6.71216309e-01 -2.89471537e-01 3.50214392e-01 -9.65886891e-01
1.05668902e-01 -2.24133506e-01 -9.11586106e-01 9.76127237e-02
-3.51122916e-01 3.74909222e-01 4.03059393e-01 -3.93674076e-01
-7.08951116e-01 6.41566098e-01 1.83153432e-02 -3.94001395e-01
-4.84494448e-01 3.23756307e-01 -3.98121536e-01 -5.18035114e-01
7.64990568e-01 2.36737281e-02 -6.30873628e-03 -6.05828404e-01
7.29620039e-01 1.06240630e-01 7.08756447e-01 -7.03647017e-01
1.46658170e+00 8.06901991e-01 3.21484357e-01 -8.56747448e-01
-8.07026923e-01 -4.11074311e-01 -1.56390357e+00 -6.68556318e-02
9.67056751e-01 -6.61098778e-01 -4.96945381e-01 1.73410892e-01
-1.14242494e+00 -3.91873926e-01 -3.94583970e-01 3.66781354e-01
-6.67325139e-01 2.72644728e-01 -3.94401819e-01 -2.20748335e-01
2.15101868e-01 -9.71563280e-01 1.12058675e+00 2.52592385e-01
-1.17064506e-01 -1.22205400e+00 1.54655769e-01 -6.24165386e-02
1.62917838e-01 4.08440411e-01 9.00868416e-01 -7.62948275e-01
-8.17558527e-01 -1.12656116e-01 -2.29079038e-01 2.54225016e-01
6.17282271e-01 2.39347711e-01 -1.00874472e+00 -3.58149797e-01
-1.22530118e-01 -1.51501983e-01 8.89367342e-01 3.35202992e-01
1.29260051e+00 -4.02948260e-01 -2.77469516e-01 9.17549849e-01
1.30325973e+00 -1.97558850e-02 3.77426773e-01 2.62563061e-02
1.05427182e+00 8.07376206e-01 1.22418955e-01 -5.03657199e-02
1.64575845e-01 8.94183397e-01 3.11934292e-01 -1.58956498e-01
-2.09653109e-01 -2.27528766e-01 1.01090729e-01 7.04461277e-01
-4.08121198e-01 3.49725991e-01 -9.06037807e-01 6.41358197e-01
-1.50791073e+00 -7.59019911e-01 8.61055776e-02 2.21288562e+00
7.36430705e-01 -7.08497912e-02 1.93674415e-01 -4.65690345e-02
6.79178596e-01 -9.99428704e-02 -3.33952934e-01 -2.14553595e-01
6.53073490e-02 3.43784362e-01 3.62551838e-01 5.88608861e-01
-1.36707342e+00 1.18629587e+00 7.09967613e+00 7.16847852e-02
-1.20910811e+00 -1.82009040e-04 3.98156106e-01 3.89161289e-01
-1.98372111e-01 5.96335940e-02 -5.54465592e-01 1.34108409e-01
6.15578830e-01 -2.12555274e-01 3.15091997e-01 9.67987835e-01
-6.89447150e-02 2.96519607e-01 -1.69424224e+00 7.32249975e-01
2.14340404e-01 -1.25486064e+00 1.03179410e-01 2.00056747e-01
8.13240707e-01 -3.17629158e-01 2.39320979e-01 -1.45276174e-01
4.62778181e-01 -1.08481550e+00 5.99977911e-01 4.73317862e-01
7.17259288e-01 -4.54533339e-01 2.97144026e-01 -7.77033642e-02
-1.00632560e+00 1.19413547e-01 -7.09178567e-01 -1.47868786e-02
-2.00350657e-01 3.42623562e-01 -8.54495108e-01 1.86056763e-01
4.33757514e-01 8.86303067e-01 -9.02355194e-01 1.14863491e+00
-4.73970503e-01 6.00602508e-01 -2.98263311e-01 3.60162020e-01
-2.53951177e-02 -1.80417001e-01 4.65955555e-01 1.12827301e+00
1.50142098e-02 2.91826222e-02 2.64309794e-01 1.01650465e+00
-3.75301808e-01 9.64260995e-02 -8.03639770e-01 2.41289929e-01
1.79022431e-01 1.41942859e+00 -1.09748662e+00 -1.81305721e-01
-2.97956914e-01 8.80677462e-01 5.54237783e-01 3.66325319e-01
-5.97689211e-01 -8.59307423e-02 7.47392118e-01 2.11536676e-01
2.25152791e-01 -4.12961245e-01 -1.62927076e-01 -1.26900685e+00
8.73307791e-03 -3.90985429e-01 -9.46571585e-03 -5.50273120e-01
-1.37325644e+00 5.23813903e-01 8.90134797e-02 -1.16193485e+00
-2.08887652e-01 -8.19473267e-01 -7.15199292e-01 4.68839109e-01
-1.38596010e+00 -1.29139507e+00 -5.09523451e-01 6.55939937e-01
3.35219860e-01 -2.11537406e-01 8.06348264e-01 1.64849371e-01
-3.40221196e-01 6.12061918e-01 6.71235695e-02 5.97839057e-01
6.94564819e-01 -1.52848506e+00 4.14702922e-01 9.30496871e-01
9.22470570e-01 8.75399530e-01 5.82587242e-01 -2.84494698e-01
-1.31340373e+00 -1.33552206e+00 8.03329349e-01 -8.41756225e-01
5.94784796e-01 -6.29119098e-01 -9.86384869e-01 1.13206005e+00
-4.98454906e-02 6.01100087e-01 4.69008803e-01 2.41750151e-01
-8.41139555e-01 -1.88302010e-01 -1.13182163e+00 3.68556559e-01
1.20332372e+00 -7.91873932e-01 -8.59689295e-01 5.78379929e-01
5.11659682e-01 -2.24400952e-01 -5.80097973e-01 2.75098860e-01
4.74634349e-01 -8.41542721e-01 1.18624115e+00 -9.91439223e-01
3.15773636e-01 -3.05700302e-01 -2.48378366e-02 -1.44528627e+00
-5.20629823e-01 -4.37587380e-01 3.28488201e-01 1.26432455e+00
1.65603802e-01 -6.12364650e-01 8.07372451e-01 5.48614800e-01
-7.37969056e-02 -8.35793763e-02 -8.97634149e-01 -8.87615800e-01
1.89762503e-01 4.88640592e-02 4.14045841e-01 1.07258582e+00
-1.64492369e-01 2.77642816e-01 -6.57906532e-02 2.66244978e-01
7.32075989e-01 1.42377838e-02 9.12477553e-01 -1.69565737e+00
-1.28314812e-02 -5.07525623e-01 -1.14012349e+00 -7.69006789e-01
6.34171486e-01 -1.17776382e+00 1.90039948e-01 -1.16482794e+00
2.66484678e-01 -4.44927722e-01 -4.36894715e-01 7.09349990e-01
5.34512326e-02 6.48857951e-01 1.21531762e-01 3.24189335e-01
-5.24369478e-01 3.67196083e-01 8.38769138e-01 -2.19835311e-01
-3.81095298e-02 -2.82870680e-02 -5.61460674e-01 1.16330516e+00
6.35203362e-01 -4.28533316e-01 -1.86094344e-01 -5.12305379e-01
-2.30533987e-01 -4.77464795e-01 8.23124111e-01 -9.45541561e-01
5.00124842e-02 -1.26574084e-01 5.08593559e-01 -1.75933853e-01
9.75102186e-02 -1.05555427e+00 -6.76388070e-02 2.89021492e-01
-5.37118018e-01 -3.42440233e-02 4.47657444e-02 4.92156684e-01
-1.27257630e-01 -1.54632613e-01 1.11735404e+00 6.60299137e-02
-7.24605799e-01 4.00453210e-01 1.03098109e-01 -6.33162335e-02
1.04459536e+00 -5.55409156e-02 -5.80553301e-02 -2.21092612e-01
-8.90408099e-01 -2.11980298e-01 6.70552433e-01 4.57729459e-01
5.25478601e-01 -1.53390002e+00 -5.48801899e-01 3.98298532e-01
1.91949904e-01 -9.39826444e-02 -4.14919168e-01 5.63607454e-01
-4.62311566e-01 2.92636633e-01 -4.59277123e-01 -9.55269575e-01
-1.25779426e+00 7.47819066e-01 4.54375833e-01 2.60001361e-01
-4.82410938e-01 7.73185372e-01 5.17631948e-01 -4.65180069e-01
2.21720174e-01 -5.12978017e-01 -1.81036353e-01 3.94352749e-02
2.83166885e-01 -1.36571288e-01 2.15600971e-02 -1.23337805e+00
-3.69928241e-01 1.13977265e+00 2.42603533e-02 9.57391132e-03
1.41204131e+00 -1.00499783e-02 -3.68547767e-01 4.10904616e-01
1.44571126e+00 1.40111372e-01 -1.28530228e+00 -5.00002325e-01
4.99286443e-01 -5.70097387e-01 -2.37685114e-01 -3.92929733e-01
-1.33474660e+00 8.92901599e-01 6.65071130e-01 4.46644574e-02
9.41419363e-01 3.57269675e-01 2.39403933e-01 5.48286855e-01
4.50332165e-01 -7.05751359e-01 2.61099517e-01 3.00825983e-01
1.00464892e+00 -1.44013834e+00 -3.04874033e-02 -4.39819127e-01
-1.89839602e-01 1.19759214e+00 4.28578913e-01 -5.74875653e-01
9.19600248e-01 -6.06669709e-02 1.38357460e-01 -3.46565783e-01
-2.17778400e-01 -4.68124688e-01 7.85913169e-01 7.90588260e-01
3.50042343e-01 9.00985301e-02 -2.21008718e-01 1.75832048e-01
-2.16608346e-01 -3.39778453e-01 4.31772888e-01 6.61623418e-01
-4.79879141e-01 -1.13129997e+00 -1.77032501e-01 2.11448163e-01
-4.71115202e-01 1.44329041e-01 -6.30425274e-01 8.30067217e-01
7.57371783e-02 5.17441571e-01 2.72399575e-01 -3.36977512e-01
3.26350033e-01 6.22829385e-02 7.00130701e-01 -7.59405553e-01
-1.73176497e-01 1.48230284e-01 -4.22001749e-01 -6.16812587e-01
-6.60399675e-01 -7.27999330e-01 -1.22335219e+00 1.11507297e-01
-1.36067018e-01 -3.41883004e-02 5.83005250e-01 9.57757831e-01
1.11682713e-01 2.34570935e-01 8.31778944e-01 -1.03537023e+00
-3.88371259e-01 -8.37106526e-01 -5.57644904e-01 9.16235149e-01
4.98454779e-01 -1.02019167e+00 -5.59634447e-01 7.10290194e-01]
|
[9.109687805175781, 2.414032220840454]
|
9b18d0a7-3947-43d0-8d5e-9b581c85ab3f
|
learning-agent-representations-for-ice-hockey
| null | null |
http://proceedings.neurips.cc/paper/2020/hash/d90e5b6628b4291225cba0bdc643c295-Abstract.html
|
http://proceedings.neurips.cc/paper/2020/file/d90e5b6628b4291225cba0bdc643c295-Paper.pdf
|
Learning Agent Representations for Ice Hockey
|
Team sports is a new application domain for agent modeling with high real-world impact. A fundamental challenge for modeling professional players is their large number (over 1K), which includes many bench players with sparse participation in a game season. The diversity and sparsity of player observations make it difficult to extend previous agent representation models to the sports domain. This paper develops a new approach for agent representations, based on a Markov game model, that is tailored towards applications in professional ice hockey. We introduce a novel player representation via player generation framework where a variational encoder embeds player information with latent variables. The encoder learns a context-specific shared prior to induce a shrinkage effect for the posterior player representations, allowing it to share statistical information across players with different participations. To model the play dynamics in sequential sports data, we design a Variational Recurrent Ladder Agent Encoder (VaRLAE). It learns a contextualized player representation with a hierarchy of latent variables that effectively prevents latent posterior collapse. We validate our player representations in major sports analytics tasks. Our experimental results, based on a large dataset that contains over 4.5M events, show state-of-the-art performance for our VarLAE on facilitating 1) identifying the acting player, 2) estimating expected goals, and 3) predicting the final score difference.
|
['Mehrsan Javan', 'Mike Rudd', 'Pascal Poupart', 'Oliver Schulte', 'Guiliang Liu']
|
2020-12-01
| null | null | null |
neurips-2020-12
|
['sports-analytics']
|
['computer-vision']
|
[-7.75554404e-02 3.01481932e-01 -5.49444675e-01 -1.75010059e-02
-9.81891155e-01 -3.29869062e-01 7.51013279e-01 -5.10434471e-02
-2.71210104e-01 5.06706715e-01 8.13971221e-01 3.61169845e-01
-1.71668917e-01 -1.05510485e+00 -8.19001615e-01 -5.54727137e-01
-3.37512612e-01 1.11197495e+00 2.76363790e-01 -6.77335620e-01
-2.32358593e-02 -3.15508127e-01 -1.72821665e+00 7.08304644e-01
5.53294778e-01 5.32636464e-01 1.44437283e-01 9.33708966e-01
1.98936701e-01 1.79236352e+00 -8.20903301e-01 -6.63788319e-01
4.35301930e-01 -5.51874816e-01 -7.80797720e-01 5.62813580e-02
-1.15212835e-01 -4.36011285e-01 -9.26767588e-01 4.54106003e-01
4.05162334e-01 4.94991064e-01 8.99799407e-01 -1.63228881e+00
-5.72243214e-01 1.40488112e+00 -7.17963696e-01 2.74030179e-01
2.58161843e-01 3.94801974e-01 1.40028703e+00 -2.90979207e-01
5.84368527e-01 1.26538908e+00 7.49140322e-01 5.42748213e-01
-1.09117496e+00 -6.95553660e-01 4.56863970e-01 5.13718903e-01
-1.21427393e+00 6.64822683e-02 7.78948247e-01 -6.77355111e-01
7.86768734e-01 7.45017752e-02 1.07887268e+00 1.65718091e+00
2.12100253e-01 1.40379238e+00 4.79823232e-01 1.28586479e-02
1.74260840e-01 -3.25118244e-01 2.88736582e-01 5.01154721e-01
-1.58477545e-01 2.35188276e-01 -1.36700511e+00 -3.15404266e-01
9.86430109e-01 2.85387665e-01 3.13831210e-01 -3.15929532e-01
-7.95578599e-01 1.32025671e+00 1.05896011e-01 -5.38136959e-01
-6.44557059e-01 6.91370487e-01 5.45103431e-01 5.75574413e-02
5.35355091e-01 4.26730514e-02 -1.67411670e-01 -7.75360227e-01
-7.86539376e-01 1.14235985e+00 7.35029101e-01 9.01634037e-01
2.61238188e-01 2.91992337e-01 -6.71997905e-01 8.26504707e-01
4.00579244e-01 1.41954988e-01 7.31387138e-01 -1.22222281e+00
5.48746526e-01 6.92629158e-01 -6.93209991e-02 -7.81182587e-01
-3.12694877e-01 -5.85586905e-01 -5.90694189e-01 2.39282608e-01
3.72614413e-01 -8.21049511e-02 -7.10321724e-01 1.95230103e+00
9.16149244e-02 6.92170560e-01 2.92255372e-01 8.07747424e-01
1.12600935e+00 7.08868206e-01 3.29471707e-01 1.69527352e-01
1.51393390e+00 -1.27576280e+00 -4.33795273e-01 -4.65333402e-01
3.61755252e-01 -5.26648620e-03 9.76916909e-01 4.25926864e-01
-1.36399531e+00 -5.86162329e-01 -7.44073570e-01 -1.49347216e-01
1.75882310e-01 -1.27493188e-01 8.82666409e-01 2.87979364e-01
-4.23129737e-01 4.88002211e-01 -1.17144656e+00 1.21820912e-01
3.53339463e-01 1.87666446e-01 4.04719748e-02 3.59373599e-01
-1.35981429e+00 4.32384938e-01 3.57389718e-01 -2.74786472e-01
-1.64290404e+00 -1.12176347e+00 -9.29483712e-01 3.19304377e-01
6.37706816e-01 -7.48457730e-01 1.48263788e+00 -5.36965311e-01
-1.85269845e+00 7.12480605e-01 2.75016785e-01 -8.68953168e-01
6.98936939e-01 -3.66741598e-01 1.63746983e-01 -2.14971706e-01
4.93068188e-01 3.30517381e-01 6.00785494e-01 -9.08771396e-01
-9.01180208e-01 -1.61938608e-01 2.79004693e-01 6.94651425e-01
-1.57199353e-01 1.08284841e-03 -6.40345156e-01 -7.81559646e-01
-2.91549116e-01 -9.46315527e-01 -4.96008277e-01 -6.91467404e-01
-6.74187019e-02 -6.60500228e-01 1.51829142e-02 -6.44821405e-01
1.35350990e+00 -1.90977752e+00 6.77702248e-01 -1.38851792e-01
5.05355537e-01 -3.29727024e-01 -1.27898782e-01 6.98114693e-01
1.91194326e-01 -2.32613549e-01 8.42722505e-02 -7.27205634e-01
3.13825369e-01 3.68292034e-01 -6.08122826e-01 2.20131963e-01
-1.34314939e-01 1.05699611e+00 -9.68044460e-01 -3.79725188e-01
-9.19312686e-02 1.39133543e-01 -1.12134242e+00 3.64342809e-01
-3.60879660e-01 1.54683515e-01 -5.15813649e-01 3.84708464e-01
1.26467854e-01 -1.25672132e-01 3.69784921e-01 4.71863031e-01
4.68114875e-02 4.45581734e-01 -1.50242615e+00 1.86768734e+00
9.20633301e-02 4.03902620e-01 -2.20964685e-01 -9.02656078e-01
7.61539698e-01 3.28962088e-01 8.10623407e-01 -3.37211549e-01
9.82563943e-03 -3.22360575e-01 -5.39994128e-02 -3.81787568e-01
1.08451903e+00 1.53870694e-02 -8.82387042e-01 5.72814941e-01
1.91713035e-01 2.38086611e-01 6.63449526e-01 6.68347716e-01
1.18314767e+00 3.71008068e-01 -1.74333695e-02 1.36352107e-01
-1.80863053e-01 1.83839336e-01 1.19670737e+00 1.16141486e+00
-3.39914970e-02 6.09686196e-01 1.06687760e+00 -6.95112705e-01
-9.07003462e-01 -1.18210948e+00 5.24077535e-01 1.64004636e+00
4.27247882e-02 -9.93174911e-01 -3.41197014e-01 -2.93841392e-01
6.62359446e-02 3.35797846e-01 -9.81555104e-01 -4.22652662e-01
-7.79686332e-01 -1.03341889e+00 8.14106703e-01 9.58506525e-01
1.23388432e-01 -1.20240366e+00 -5.70218801e-01 5.13550282e-01
-4.97594476e-01 -7.62962103e-01 -3.99310946e-01 7.93579817e-02
-5.03114343e-01 -1.19814897e+00 -5.19698679e-01 -5.30598879e-01
-1.44302949e-01 -1.61440641e-01 1.65268588e+00 -2.13461012e-01
-2.53849655e-01 4.68762755e-01 -4.59538847e-01 -6.51232064e-01
-4.50169772e-01 3.17810744e-01 2.62564182e-01 -1.23583332e-01
3.48236501e-01 -4.47276741e-01 -4.74446326e-01 -5.69174774e-02
-6.12760484e-01 2.61010379e-01 2.71925986e-01 9.25167084e-01
5.39653897e-01 -1.16910644e-01 3.66688818e-01 -8.94540668e-01
8.20197523e-01 -9.50627744e-01 -1.83525115e-01 -4.53259349e-02
-2.23436877e-01 -1.61648944e-01 1.43297836e-01 -8.82068038e-01
-9.24777150e-01 -1.70459867e-01 1.10959530e-01 -4.22279894e-01
3.25830877e-01 6.66056931e-01 1.66564628e-01 9.93635952e-01
7.89673209e-01 3.39375407e-01 1.02109835e-01 -4.07888353e-01
4.36413139e-01 3.40178162e-01 7.23342180e-01 -8.95484805e-01
5.65518737e-01 3.69274169e-01 -2.71990836e-01 -5.75347543e-01
-7.56138802e-01 -3.69072437e-01 -2.14354798e-01 -4.28881615e-01
8.04508746e-01 -1.61769414e+00 -1.37632895e+00 5.76962650e-01
-7.04964519e-01 -8.39664757e-01 -8.94660115e-01 6.16815269e-01
-1.01985812e+00 1.68296203e-01 -1.22206461e+00 -8.61905456e-01
-1.15676440e-01 -9.98585224e-01 9.49727952e-01 3.34583402e-01
-3.88448417e-01 -7.00955391e-01 6.29599154e-01 7.92132378e-01
-2.85598189e-01 1.79139480e-01 5.33891916e-01 -4.60510939e-01
-5.70669591e-01 -4.78714556e-02 5.14323711e-01 -1.33856878e-01
-4.77669209e-01 -2.44845510e-01 -4.30445820e-01 -2.18092918e-01
-6.02763653e-01 -6.36392593e-01 9.82596397e-01 7.83804119e-01
9.14783299e-01 6.52803332e-02 -1.82507768e-01 5.52985549e-01
8.29468846e-01 -2.69064009e-01 6.76166356e-01 4.83846605e-01
6.93759561e-01 4.69059676e-01 5.31980932e-01 1.11428237e+00
9.02142286e-01 8.63909721e-01 4.37389582e-01 2.10709125e-01
-4.80289152e-03 -8.89689922e-01 7.34511077e-01 6.35606229e-01
-6.79614782e-01 -3.94159496e-01 -6.51153684e-01 7.57721126e-01
-2.52020264e+00 -1.51342154e+00 -3.12336177e-01 1.65232193e+00
8.44747961e-01 9.76366252e-02 7.84399331e-01 -1.05138332e-01
2.99448937e-01 3.47107410e-01 -6.27089143e-01 -1.81658283e-01
2.83960137e-03 1.72906876e-01 4.53500897e-01 3.71587276e-01
-9.78608787e-01 1.25825536e+00 6.67375517e+00 1.12486637e+00
-1.62265927e-01 2.41344601e-01 3.08124661e-01 -7.42922843e-01
-1.76627681e-01 -1.12296909e-01 -1.09819984e+00 5.15420377e-01
8.49245012e-01 -3.57608378e-01 1.68519750e-01 9.66908216e-01
3.27564120e-01 3.27934623e-01 -8.98644209e-01 8.60057235e-01
1.68708131e-01 -1.55540729e+00 5.97174242e-02 3.72361600e-01
8.17919552e-01 -3.28171328e-02 2.10077569e-01 1.20770609e+00
1.28552830e+00 -1.15889251e+00 1.07412052e+00 7.72011936e-01
1.93927392e-01 -9.21866596e-01 4.57595825e-01 7.53892124e-01
-1.40506113e+00 -3.91925752e-01 -4.40678358e-01 -6.93391860e-01
3.90060067e-01 -3.52080405e-01 -4.69472021e-01 2.19526082e-01
9.55080211e-01 1.07481861e+00 -9.45017561e-02 6.10604525e-01
-2.54115582e-01 8.89261782e-01 -2.36069515e-01 2.33690172e-01
1.81487486e-01 -2.98727185e-01 6.69258595e-01 6.86345875e-01
7.50875995e-02 2.80238479e-01 7.01619148e-01 8.54071319e-01
-1.88573807e-01 -4.17741016e-03 -2.57004768e-01 -3.39511666e-03
4.59725082e-01 9.37194765e-01 -3.87414753e-01 -4.35701102e-01
-2.11375356e-01 8.49261940e-01 6.30378246e-01 1.52655035e-01
-1.23430634e+00 3.97801578e-01 1.29846561e+00 2.63900816e-01
2.88861901e-01 -4.38162535e-02 -3.67483944e-02 -1.18246806e+00
-2.88046420e-01 -1.09393573e+00 9.14267361e-01 -4.00030136e-01
-1.36811900e+00 3.50517511e-01 2.01548606e-01 -1.26443255e+00
-8.75501335e-01 -2.30379224e-01 -7.73339272e-01 7.65036821e-01
-1.01324391e+00 -1.47240102e+00 -6.27828240e-02 5.54430842e-01
1.04409957e+00 -7.71139562e-01 6.07221067e-01 1.06054127e-01
-7.45279789e-01 6.49351299e-01 3.96881402e-02 3.69420081e-01
3.07545215e-01 -1.27556467e+00 4.06471729e-01 5.05768776e-01
3.42371553e-01 2.71937281e-01 8.00257385e-01 -8.03800046e-01
-1.06083488e+00 -7.96450317e-01 3.18757832e-01 -8.50147605e-01
6.30625844e-01 -3.41425627e-01 -6.53411031e-01 1.21414518e+00
8.44363719e-02 -5.17093420e-01 1.21110570e+00 6.84316278e-01
-1.56912401e-01 3.39338779e-01 -3.40658307e-01 6.97306156e-01
9.88210499e-01 -1.62965253e-01 -6.91605210e-01 2.45543197e-01
5.53012908e-01 -7.63294816e-01 -8.86140168e-01 8.09941813e-02
6.54218674e-01 -6.72583997e-01 1.21179843e+00 -1.29289114e+00
9.94852602e-01 -4.00196202e-02 2.52128020e-02 -1.18905127e+00
-7.17080772e-01 -6.01661384e-01 -6.32365704e-01 1.07449961e+00
1.20351247e-01 1.87397912e-01 1.56270623e+00 8.20047140e-01
-2.39181563e-01 -8.38081717e-01 -7.32771099e-01 -5.52544355e-01
2.09324807e-01 -7.16332257e-01 6.99715674e-01 5.81944644e-01
1.95990503e-01 3.02320600e-01 -1.20144761e+00 2.43770659e-01
8.61036122e-01 8.86692777e-02 1.28568101e+00 -1.35387194e+00
-9.83880222e-01 -4.65606123e-01 -5.81684053e-01 -1.42994595e+00
2.50430107e-01 -9.83909726e-01 -9.52577740e-02 -1.44162858e+00
7.16836452e-01 -1.46113887e-01 -9.16780382e-02 3.49040449e-01
-3.11008960e-01 3.82214665e-01 4.03185189e-01 2.95332491e-01
-7.79872119e-01 9.82740939e-01 8.98920536e-01 -1.81660041e-01
-4.50214505e-01 2.38929704e-01 -9.07806933e-01 9.63796318e-01
5.36035836e-01 -7.56710052e-01 -4.74447727e-01 -4.18883771e-01
4.92418349e-01 2.90514857e-01 4.92734224e-01 -7.77219355e-01
2.08205163e-01 -4.36238259e-01 8.35777819e-02 -5.96987605e-01
8.36978912e-01 4.30076793e-02 2.43722215e-01 4.68325943e-01
-7.26992607e-01 -1.45703167e-01 -2.16791481e-01 1.00373590e+00
-2.46226072e-01 -3.29368591e-01 1.32954344e-01 -3.24787498e-01
-8.63553941e-01 6.79084420e-01 -8.14383507e-01 4.69110012e-01
1.07979941e+00 -1.06727615e-01 9.96203721e-03 -7.61434019e-01
-1.04725254e+00 7.63499260e-01 -1.54202342e-01 5.99837482e-01
4.34425116e-01 -1.48641312e+00 -1.46961510e+00 -3.75499651e-02
1.41222775e-01 -1.06001586e-01 8.55280757e-01 1.22486718e-01
-4.60887909e-01 -3.38485509e-01 -2.31480241e-01 -4.99804288e-01
-1.36793947e+00 1.96459487e-01 2.26488918e-01 -1.09751153e+00
-1.01576996e+00 1.08359361e+00 2.45641530e-01 -4.02175039e-01
2.60261714e-01 9.27629601e-03 -6.10609531e-01 2.49254748e-01
6.03541255e-01 5.51355362e-01 -6.32015407e-01 -6.20584786e-01
1.21290192e-01 -7.30120540e-02 -2.30297983e-01 -2.19387248e-01
1.75620866e+00 2.57168978e-01 5.08984804e-01 6.25440538e-01
3.48726898e-01 -2.41204798e-01 -1.85488737e+00 -3.65649790e-01
-4.68360841e-01 -4.04175967e-01 -8.22531506e-02 -1.82136357e-01
-1.10007942e+00 5.52037835e-01 2.87047535e-01 1.86304674e-01
3.70537490e-01 3.07832628e-01 8.72008979e-01 -1.09841563e-01
3.78836423e-01 -1.38478374e+00 5.11016369e-01 6.97788000e-01
8.02099526e-01 -1.05212140e+00 -2.43699066e-02 -1.56933784e-01
-1.40287185e+00 6.11857831e-01 7.78681219e-01 -5.15660465e-01
4.97061014e-01 2.24133477e-01 -5.61629012e-02 -5.15997410e-01
-1.23254383e+00 -4.09878463e-01 2.65408363e-02 6.02413416e-01
2.42657751e-01 4.28437561e-01 -2.04849288e-01 1.60166061e+00
-7.74166167e-01 -6.26282394e-02 8.18880439e-01 7.22389221e-01
-1.72801241e-01 -1.17996752e+00 -1.70863673e-01 4.35714722e-01
-3.96418631e-01 1.08432204e-01 -2.56205983e-02 7.03798950e-01
2.56051153e-01 8.11729908e-01 4.19986725e-01 -4.14821327e-01
4.42377388e-01 -1.62005965e-02 3.22946370e-01 -8.32014143e-01
-9.36093688e-01 -4.87906411e-02 -1.16512299e-01 -4.69102442e-01
-1.84798822e-01 -9.50667024e-01 -1.17922497e+00 -6.66981876e-01
1.96157545e-01 3.19276720e-01 3.50491963e-02 6.78938687e-01
1.41837925e-01 1.12577510e+00 1.76704213e-01 -8.97827268e-01
-6.07916057e-01 -1.14840674e+00 -1.10527110e+00 8.05382729e-01
-3.45134467e-01 -1.03989875e+00 1.19867198e-01 1.03917398e-01]
|
[6.661679744720459, 0.351650208234787]
|
c98b0fd1-677d-488c-815e-d9a543208cfb
|
generalization-bounds-for-set-to-set-matching
|
2302.12991
| null |
https://arxiv.org/abs/2302.12991v1
|
https://arxiv.org/pdf/2302.12991v1.pdf
|
Generalization Bounds for Set-to-Set Matching with Negative Sampling
|
The problem of matching two sets of multiple elements, namely set-to-set matching, has received a great deal of attention in recent years. In particular, it has been reported that good experimental results can be obtained by preparing a neural network as a matching function, especially in complex cases where, for example, each element of the set is an image. However, theoretical analysis of set-to-set matching with such black-box functions is lacking. This paper aims to perform a generalization error analysis in set-to-set matching to reveal the behavior of the model in that task.
|
['Masanari Kimura']
|
2023-02-25
| null | null | null | null |
['set-matching']
|
['computer-vision']
|
[ 6.97898090e-01 -1.24575049e-01 1.31770819e-01 -7.70602226e-01
-4.87910002e-01 -4.83553529e-01 3.96918356e-01 1.97620392e-01
-4.16637301e-01 4.84468341e-01 -4.84662235e-01 -9.72736180e-02
-5.53813696e-01 -8.81328940e-01 -7.90779769e-01 -5.86155593e-01
3.63951661e-02 6.30296826e-01 6.58072606e-02 -3.94424558e-01
2.44634166e-01 8.57542813e-01 -2.11025858e+00 3.48775417e-01
6.03905499e-01 1.08666503e+00 1.99796259e-01 3.64530087e-01
-3.66691686e-02 2.53186315e-01 -7.06738889e-01 -6.95407331e-01
7.17990518e-01 -5.74384987e-01 -6.24502540e-01 2.02260494e-01
1.02410817e+00 8.42259079e-02 -3.59669715e-01 1.47165513e+00
5.52490234e-01 2.76445508e-01 6.28582001e-01 -1.40262413e+00
-5.64464927e-01 5.94875097e-01 -2.49022543e-01 6.33576214e-02
4.62346643e-01 -6.83278665e-02 1.01841068e+00 -8.31204832e-01
3.06999207e-01 1.18339050e+00 6.38022244e-01 6.52600229e-01
-1.13288069e+00 -8.40729892e-01 -2.91465819e-01 1.54725388e-01
-1.55626643e+00 -4.25644457e-01 7.13650882e-01 -1.76639587e-01
7.88835943e-01 4.62222278e-01 5.82133353e-01 3.02572250e-01
1.03742279e-01 6.26893878e-01 1.09195900e+00 -8.09148848e-01
-4.21133041e-02 3.41162682e-01 3.37601781e-01 6.16932988e-01
5.41705489e-01 2.90364623e-01 -9.98995155e-02 -8.34513381e-02
8.03069711e-01 1.49269491e-01 -4.12519515e-01 -2.37981960e-01
-7.35295713e-01 6.06536984e-01 5.74203730e-01 7.89143980e-01
-2.00143203e-01 2.29913920e-01 2.02991232e-01 6.98257565e-01
2.51173466e-01 7.93022215e-01 5.20468205e-02 3.77263129e-01
-8.93020272e-01 3.24670851e-01 8.68860364e-01 1.13131225e+00
8.37644279e-01 -9.79552194e-02 1.59281865e-01 7.36012876e-01
-2.26079434e-01 4.29118812e-01 2.01024428e-01 -8.06033492e-01
4.98072147e-01 6.59469903e-01 -1.71288326e-01 -1.14972472e+00
-4.76770312e-01 -3.84716243e-01 -1.13536429e+00 3.68229091e-01
6.39226735e-01 2.33667925e-01 -5.06636918e-01 1.77628171e+00
2.57545989e-02 -5.05421236e-02 4.52393442e-02 8.65484238e-01
9.77792680e-01 2.79938251e-01 -4.82630044e-01 -1.05930492e-01
1.20292854e+00 -6.50546670e-01 -6.32115424e-01 -8.33422020e-02
5.47617316e-01 -6.89768374e-01 8.05494308e-01 7.14829415e-02
-1.33830178e+00 -8.99852991e-01 -1.17041707e+00 4.78396088e-01
-5.58266461e-01 -2.87052631e-01 4.55550313e-01 8.11975181e-01
-8.58202875e-01 9.86337423e-01 -2.04957485e-01 -4.75259930e-01
3.17260861e-01 8.02388132e-01 -5.55085123e-01 -1.90708354e-01
-1.28705800e+00 1.16017175e+00 7.56628335e-01 1.86712071e-01
-2.15052634e-01 -5.40468276e-01 -8.43419313e-01 3.74561161e-01
3.02270681e-01 -6.65794671e-01 1.04798222e+00 -1.18035614e+00
-9.74056184e-01 1.16898322e+00 -5.65763377e-02 -2.98415840e-01
3.75620395e-01 3.48978311e-01 -5.22547007e-01 -1.12888478e-01
-3.35380495e-01 5.17130375e-01 6.70102060e-01 -1.23749793e+00
-4.11558360e-01 -4.92157578e-01 2.49496430e-01 1.65941373e-01
-4.95242774e-01 3.51034492e-01 -2.11224303e-01 -5.62456131e-01
2.35858470e-01 -9.00453448e-01 -2.36790687e-01 1.58424795e-01
-1.40795052e-01 -3.23440552e-01 4.37463403e-01 -1.58195004e-01
1.32810855e+00 -2.06328249e+00 -1.35710433e-01 4.78345454e-01
2.58575052e-01 3.02688122e-01 -1.33498833e-01 5.57710886e-01
-6.48950219e-01 -1.12513289e-01 -4.06285405e-01 -2.29855791e-01
-1.09791823e-01 2.38186941e-01 -8.57307985e-02 7.33360231e-01
7.67382011e-02 7.38137126e-01 -4.62077975e-01 -5.26506543e-01
8.56237039e-02 4.80068056e-03 -4.33086395e-01 3.57296020e-01
2.24892855e-01 -1.51410192e-01 -7.29481205e-02 4.55681235e-01
7.72496879e-01 -3.88190389e-01 -8.85026902e-02 -1.52832761e-01
8.97716358e-02 -4.14998055e-01 -1.34956467e+00 1.15715480e+00
-1.30214235e-02 9.08159614e-01 -2.16938153e-01 -1.23999786e+00
1.20371222e+00 2.95593381e-01 6.70473933e-01 -7.66356766e-01
4.13319081e-01 3.26285303e-01 5.48957705e-01 -1.93307742e-01
6.14054441e-01 -4.28081721e-01 -1.72170535e-01 4.77245837e-01
-1.89275548e-01 -3.08170199e-01 3.93110543e-01 -2.41493031e-01
7.47738242e-01 -3.75829279e-01 6.12787306e-01 -2.33301893e-01
6.39061868e-01 -2.78704353e-02 1.44051969e-01 9.38896835e-01
-2.60751247e-02 9.05811846e-01 -1.07805036e-01 -3.69688153e-01
-1.21374643e+00 -7.02934861e-01 -2.41881698e-01 6.34860992e-01
4.43614632e-01 -6.88308477e-03 -9.85159516e-01 -1.38053879e-01
1.53894827e-01 5.79228342e-01 -5.66249788e-01 -3.72329384e-01
-9.96016324e-01 -4.95234311e-01 5.58804452e-01 5.92704177e-01
6.16235614e-01 -1.18930793e+00 -5.81325114e-01 3.22078437e-01
2.96840638e-01 -1.05180681e+00 -7.68351376e-01 2.48554170e-01
-9.78345633e-01 -1.28948724e+00 -5.99404156e-01 -1.24478042e+00
9.83830750e-01 4.37169373e-01 1.24013329e+00 4.62170929e-01
-2.17133686e-01 2.85394758e-01 -3.09171677e-02 -5.61629713e-01
-6.51008248e-01 -2.47290611e-01 1.73464239e-01 1.46470353e-01
4.43780780e-01 -3.60560536e-01 -2.10391685e-01 9.11234915e-01
-1.18506312e+00 -7.01536611e-02 4.19589937e-01 8.77070546e-01
3.22272032e-01 2.98285604e-01 5.98093390e-01 -7.28184760e-01
7.78427243e-01 -6.28775731e-02 -9.27998304e-01 3.23431522e-01
-5.03295064e-01 -7.61530595e-03 8.18808973e-01 -6.94316328e-01
-6.58969045e-01 1.33328006e-01 -7.16426894e-02 -5.74955881e-01
-1.01727314e-01 2.63300121e-01 -3.93176049e-01 -6.61140501e-01
7.48236477e-01 3.35938960e-01 2.68394083e-01 -2.83571929e-01
1.23062290e-01 7.82124579e-01 6.84431612e-01 -4.21641380e-01
9.29232955e-01 1.94234997e-01 2.83770949e-01 -5.45252204e-01
-7.18776166e-01 -6.96106315e-01 -5.95503390e-01 -4.37827498e-01
3.88204962e-01 -3.98370385e-01 -1.00122058e+00 5.83987296e-01
-1.20855510e+00 1.36328682e-01 -3.62424910e-01 3.70583922e-01
-8.27573657e-01 2.16025218e-01 -3.23142916e-01 -6.51353776e-01
-5.06730139e-01 -1.05995500e+00 4.71224666e-01 3.65496963e-01
-1.50649354e-01 -1.08535719e+00 -2.61427552e-01 -8.20150375e-02
3.04076284e-01 9.61995423e-02 8.81842196e-01 -1.26598465e+00
-6.43096685e-01 -6.16576374e-01 -2.28187904e-01 4.68878716e-01
1.35608524e-01 -2.71188051e-01 -8.15228879e-01 -4.52781111e-01
4.03299749e-01 -4.29256633e-02 5.59229553e-01 2.93847084e-01
1.27416682e+00 -2.16041729e-01 -3.41411591e-01 4.75935251e-01
1.61298537e+00 5.27160823e-01 6.79891169e-01 3.30209106e-01
2.78190225e-01 7.42425382e-01 6.06777012e-01 2.27300912e-01
-2.15799615e-01 7.62680829e-01 3.65589023e-01 -2.09355801e-01
-1.60330087e-01 6.51954487e-02 -3.89183104e-01 7.16506302e-01
1.55856997e-01 -4.43470150e-01 -7.13337958e-01 2.18975991e-01
-1.69443142e+00 -1.26114786e+00 -1.19972704e-02 2.48093414e+00
5.45720041e-01 2.85744250e-01 -2.40919050e-02 5.46313405e-01
1.25712228e+00 -3.18166651e-02 -4.44916397e-01 -4.58340682e-02
-4.75645185e-01 1.45268500e-01 3.93332273e-01 1.21650152e-01
-9.07051206e-01 5.08486211e-01 7.31234169e+00 9.63539422e-01
-9.37629879e-01 -2.32403472e-01 4.24071610e-01 3.08057845e-01
-2.00184062e-01 -8.90864134e-02 -8.58282447e-01 3.81840199e-01
5.02123594e-01 -3.58187079e-01 3.14228058e-01 6.07738376e-01
-2.57279932e-01 7.24164536e-03 -1.66412592e+00 1.26895249e+00
2.65661508e-01 -1.29683495e+00 9.91111100e-02 -2.40150709e-02
6.37664616e-01 -6.12982452e-01 1.43516153e-01 1.18004791e-01
-2.18103319e-01 -1.12529802e+00 3.09092641e-01 5.82897186e-01
6.67931736e-01 -6.98208749e-01 7.18789458e-01 6.13653481e-01
-1.12195194e+00 -5.99785075e-02 -6.01515472e-01 -1.79704189e-01
-8.00350532e-02 3.10665071e-01 -7.92709589e-01 5.68548441e-01
4.82517391e-01 3.83107752e-01 -1.90060839e-01 1.65685523e+00
5.33815026e-01 2.21523151e-01 -4.22015101e-01 -4.28272605e-01
9.72109810e-02 -3.91122431e-01 5.20121336e-01 9.56035256e-01
3.70023996e-01 1.96891636e-01 1.30997464e-01 1.07266080e+00
-2.56021857e-01 1.26702592e-01 -8.93325150e-01 1.46752223e-01
4.35040504e-01 8.67375076e-01 -6.61369681e-01 -3.71022224e-01
-2.53204226e-01 7.00570643e-01 4.26458985e-01 -1.79984458e-02
-6.73341990e-01 -6.65514648e-01 6.13694727e-01 1.84173837e-01
1.30524620e-01 1.33228332e-01 -2.88007766e-01 -6.02324307e-01
3.13167050e-02 -8.15296531e-01 5.45573473e-01 -8.11911762e-01
-1.34568250e+00 6.11930966e-01 4.06302780e-01 -1.52571523e+00
-2.77854800e-01 -6.86583698e-01 -5.82152545e-01 9.84554708e-01
-9.47398305e-01 -7.22223043e-01 -3.92317891e-01 5.47245204e-01
3.24770600e-01 -4.88528579e-01 7.06630349e-01 4.89463985e-01
-2.96775758e-01 1.11682165e+00 1.97894484e-01 2.05737650e-01
5.19842446e-01 -1.03918612e+00 4.03009295e-01 6.40504539e-01
2.68477887e-01 5.58863580e-01 9.93301749e-01 -2.25029320e-01
-1.40409851e+00 -8.49571288e-01 8.40223432e-01 -2.96275020e-01
3.21335405e-01 -1.60527155e-01 -1.24320829e+00 4.56689656e-01
1.36036649e-01 4.30934913e-02 3.56668562e-01 -3.82105321e-01
-2.30287462e-01 -3.56091768e-01 -1.46228540e+00 6.08224154e-01
1.21455324e+00 -3.86053056e-01 -7.09956527e-01 4.33711857e-01
4.78118896e-01 -7.09471524e-01 -1.07339323e+00 5.29991090e-01
5.34285367e-01 -1.03284633e+00 1.01456881e+00 -6.23778999e-01
1.84931740e-01 -1.82365432e-01 -2.60864228e-01 -1.29416049e+00
-3.96622121e-01 -3.67907256e-01 2.14329988e-01 9.38338041e-01
1.50114149e-01 -9.66359794e-01 9.96661961e-01 5.24904847e-01
-1.94994420e-01 -7.80075371e-01 -8.72206092e-01 -1.23580170e+00
-2.77893860e-02 -1.76552698e-01 7.82994866e-01 7.66066849e-01
-1.78674608e-01 2.05730900e-01 -3.74313802e-01 7.18003325e-03
6.27908707e-01 3.68344128e-01 6.48101628e-01 -1.49192786e+00
-2.41601661e-01 -8.14336479e-01 -7.49207258e-01 -1.11047530e+00
2.89597899e-01 -1.00445795e+00 8.21963325e-02 -1.23623812e+00
3.05879802e-01 -6.14289105e-01 -1.29181966e-01 2.25239024e-01
-1.19746111e-01 3.62588614e-01 4.86949652e-01 2.34740347e-01
-2.25485995e-01 2.84706056e-01 1.54216087e+00 -2.82004565e-01
1.03661820e-01 4.76872325e-01 -6.06001198e-01 5.08702695e-01
8.29814136e-01 -8.04891586e-01 -1.58973798e-01 -5.20301498e-02
5.84960021e-02 3.51491332e-01 1.62333444e-01 -1.09844840e+00
6.03094697e-01 -4.18735333e-02 1.99518397e-01 -6.63464844e-01
7.26805866e-01 -1.35703552e+00 6.08368993e-01 5.83163321e-01
-4.98882920e-01 3.89509320e-01 2.64632285e-01 2.34615892e-01
-5.12176335e-01 -9.71484005e-01 9.29285645e-01 -1.92009851e-01
-6.80065393e-01 3.23988557e-01 1.62010655e-01 -2.71845851e-02
1.04317415e+00 -9.56042111e-01 -1.85942337e-01 -3.66562784e-01
-4.13375765e-01 -3.40277655e-03 3.79556060e-01 2.00298563e-01
7.82807291e-01 -1.60609102e+00 -7.97476351e-01 2.74170786e-01
4.12842631e-01 -1.78568065e-01 1.00750513e-01 5.56403339e-01
-3.03381592e-01 3.63946795e-01 -3.28313410e-01 -5.34220517e-01
-1.76349747e+00 8.39831412e-01 6.60343945e-01 8.36613178e-02
-5.34815192e-01 7.13100195e-01 2.09314510e-01 -4.50002372e-01
2.92616844e-01 -2.19408311e-02 -2.40538687e-01 -1.28669605e-01
5.68274796e-01 2.83191621e-01 -2.12483685e-02 -7.11476624e-01
-3.35894644e-01 8.84426475e-01 1.18667096e-01 1.88680455e-01
1.11441863e+00 2.98141271e-01 -2.61001170e-01 2.70881385e-01
1.60095632e+00 -3.80570471e-01 -5.02367795e-01 -6.38147116e-01
-9.32146832e-02 -7.40234911e-01 -4.27405059e-01 -2.93929160e-01
-9.90013063e-01 5.45008838e-01 7.00445175e-01 4.59274739e-01
1.32831740e+00 -7.84398690e-02 4.97068435e-01 8.88483644e-01
4.97630477e-01 -8.78398895e-01 2.40469836e-02 4.54550236e-01
1.10893452e+00 -1.23501861e+00 -1.91430971e-01 -5.22059858e-01
-9.48658139e-02 1.33166683e+00 7.40332901e-01 -3.72029006e-01
6.57568514e-01 3.76967490e-01 -2.73196906e-01 2.50805803e-02
-4.04447138e-01 -2.45989531e-01 6.65259659e-01 4.95279104e-01
9.29945335e-02 2.13398468e-02 -7.66856372e-02 3.26802880e-02
-1.65858701e-01 -7.92499073e-03 3.90852988e-01 1.01652372e+00
-6.76099420e-01 -9.57902789e-01 -7.17424393e-01 6.97060406e-01
-1.88316613e-01 -1.79536879e-01 -2.12429211e-01 1.17038119e+00
-1.15955941e-01 7.42998242e-01 3.99908394e-01 -4.75331932e-01
7.25696206e-01 -2.41850436e-01 8.59667659e-01 -3.68449211e-01
-1.11337554e+00 -5.22231817e-01 9.54702199e-02 -3.03701252e-01
-5.35028875e-01 -5.61472714e-01 -1.16582704e+00 -4.92269963e-01
-6.35477424e-01 1.11239024e-01 4.57612872e-01 8.72605622e-01
-1.92676395e-01 2.81113505e-01 6.44554257e-01 -6.65963888e-01
-8.78358185e-01 -8.97021532e-01 -7.50034213e-01 8.54979753e-01
1.83393419e-01 -6.08801007e-01 -3.90641361e-01 -3.94674867e-01]
|
[9.730236053466797, 3.1024017333984375]
|
624db4ad-3789-4766-a4b0-0b1c810f7ad8
|
end-to-end-optimization-of-scene-layout-1
|
2007.11744
| null |
https://arxiv.org/abs/2007.11744v1
|
https://arxiv.org/pdf/2007.11744v1.pdf
|
End-to-End Optimization of Scene Layout
|
We propose an end-to-end variational generative model for scene layout synthesis conditioned on scene graphs. Unlike unconditional scene layout generation, we use scene graphs as an abstract but general representation to guide the synthesis of diverse scene layouts that satisfy relationships included in the scene graph. This gives rise to more flexible control over the synthesis process, allowing various forms of inputs such as scene layouts extracted from sentences or inferred from a single color image. Using our conditional layout synthesizer, we can generate various layouts that share the same structure of the input example. In addition to this conditional generation design, we also integrate a differentiable rendering module that enables layout refinement using only 2D projections of the scene. Given a depth and a semantics map, the differentiable rendering module enables optimizing over the synthesized layout to fit the given input in an analysis-by-synthesis fashion. Experiments suggest that our model achieves higher accuracy and diversity in conditional scene synthesis and allows exemplar-based scene generation from various input forms.
|
['Joshua B. Tenenbaum', 'Andrew Luo', 'Zhoutong Zhang', 'Jiajun Wu']
|
2020-07-23
|
end-to-end-optimization-of-scene-layout
|
http://openaccess.thecvf.com/content_CVPR_2020/html/Luo_End-to-End_Optimization_of_Scene_Layout_CVPR_2020_paper.html
|
http://openaccess.thecvf.com/content_CVPR_2020/papers/Luo_End-to-End_Optimization_of_Scene_Layout_CVPR_2020_paper.pdf
|
cvpr-2020-6
|
['scene-generation', 'indoor-scene-reconstruction', 'indoor-scene-synthesis']
|
['computer-vision', 'computer-vision', 'computer-vision']
|
[ 4.11804438e-01 2.54210800e-01 3.18437368e-01 -6.43019795e-01
-6.72977865e-01 -9.31000590e-01 7.61526108e-01 -3.96587364e-02
1.63307115e-01 5.14649808e-01 3.87835592e-01 -3.27823192e-01
2.00545732e-02 -1.10787439e+00 -7.77215540e-01 -3.58842403e-01
3.92466724e-01 4.09479648e-01 -5.19266948e-02 -1.01409398e-01
2.34700724e-01 6.97327256e-01 -1.54517794e+00 4.96133476e-01
9.32127833e-01 4.97742474e-01 5.79676390e-01 1.09341526e+00
-4.81794357e-01 6.39521539e-01 -6.78421497e-01 -3.26008320e-01
2.10395128e-01 -6.72728240e-01 -5.68555176e-01 7.14568913e-01
5.78126729e-01 -3.06219637e-01 -1.72014624e-01 8.13555539e-01
2.38681734e-01 2.57344544e-01 8.12735915e-01 -1.21707976e+00
-7.95031548e-01 4.91970658e-01 -3.30836117e-01 -6.50050282e-01
7.23162115e-01 4.54933524e-01 1.17332208e+00 -8.81637216e-01
9.66180384e-01 1.51503456e+00 -1.08707165e-02 4.68974590e-01
-1.77579558e+00 -1.85965464e-01 4.55224127e-01 -4.42803919e-01
-1.31105971e+00 -4.17216480e-01 1.10104311e+00 -4.72382843e-01
7.83423364e-01 5.26848733e-01 9.27468479e-01 9.46765006e-01
2.28931308e-02 7.48950422e-01 8.89708400e-01 -4.42670196e-01
5.19535601e-01 2.54109651e-01 -2.70296782e-01 7.21651495e-01
8.74190852e-02 -3.11126322e-01 -4.23973799e-01 -1.61968730e-02
1.15148032e+00 -2.26370215e-01 -1.84923068e-01 -8.91535699e-01
-1.21024060e+00 8.05852294e-01 5.47929585e-01 -2.63308138e-01
-1.49223819e-01 2.91852951e-01 2.10156124e-02 -4.10324894e-02
2.97105849e-01 8.87972713e-01 -1.99593052e-01 1.29071221e-01
-1.11906314e+00 6.33988917e-01 8.97221327e-01 1.42621827e+00
1.03839731e+00 2.41785407e-01 -4.48186666e-01 5.59744358e-01
4.20961469e-01 6.13900661e-01 -4.69577998e-01 -1.26024199e+00
4.09182668e-01 6.90793276e-01 6.00131527e-02 -8.61340642e-01
-1.50870845e-01 -3.68551522e-01 -6.50933027e-01 4.43797886e-01
2.52441227e-01 -5.73802739e-02 -1.18281198e+00 1.82363880e+00
3.53735507e-01 -1.17236324e-01 -4.71866392e-02 6.34555399e-01
7.96514630e-01 7.33083308e-01 -8.59190375e-02 3.38842794e-02
1.08947361e+00 -9.03580785e-01 -4.78261709e-01 -2.04810217e-01
5.48450887e-01 -6.10526025e-01 1.58191144e+00 3.45720291e-01
-1.27409148e+00 -5.29634953e-01 -9.69225109e-01 -4.81055260e-01
-2.97516346e-01 -1.15253003e-02 6.46706760e-01 5.83873749e-01
-1.33672333e+00 2.47584015e-01 -5.59466422e-01 -3.55158389e-01
3.28196198e-01 -5.98083921e-02 -3.66922989e-02 -1.23108506e-01
-5.32936752e-01 5.55871725e-01 2.90117443e-01 -8.09745211e-03
-9.52056050e-01 -9.40400720e-01 -1.19056833e+00 4.18599620e-02
3.34362864e-01 -1.37078178e+00 1.20993769e+00 -7.33678460e-01
-1.62070143e+00 7.54853964e-01 -3.25112283e-01 5.80867417e-02
4.89151835e-01 9.86512154e-02 1.05097666e-01 6.34039845e-03
1.84655532e-01 8.21762681e-01 7.42554009e-01 -1.71691990e+00
-2.21101850e-01 3.73997465e-02 4.69294339e-01 4.19616461e-01
4.09572929e-01 -3.57946396e-01 -7.46261835e-01 -3.92394602e-01
1.30917281e-01 -6.64254963e-01 -6.02494121e-01 3.45162004e-01
-1.02779686e+00 4.53487456e-01 7.66776562e-01 -3.00556391e-01
1.22317719e+00 -2.10209703e+00 5.60230792e-01 5.59512675e-01
3.72973718e-02 -5.07950723e-01 -1.61450505e-01 5.80310524e-01
-2.05285195e-02 2.94066936e-01 -4.81013089e-01 -8.35749745e-01
3.44844252e-01 2.65272170e-01 -5.19384861e-01 2.36488879e-02
4.08341080e-01 1.02181101e+00 -1.00074351e+00 -5.43448627e-01
5.75574279e-01 3.70110631e-01 -1.18320632e+00 5.90942562e-01
-9.38706279e-01 7.63278067e-01 -4.88495827e-01 2.71500349e-01
6.32789254e-01 -3.49968851e-01 3.52114379e-01 -2.65957057e-01
-1.18574418e-01 1.81145057e-01 -1.36568964e+00 2.41316080e+00
-5.71314514e-01 4.76929754e-01 2.27283817e-02 -3.28922749e-01
9.06424224e-01 -1.19006366e-01 6.32198676e-02 -3.76704693e-01
-1.07687660e-01 -2.83845961e-01 -3.55227709e-01 -2.12662563e-01
7.97904789e-01 -3.33137959e-02 -3.49285960e-01 5.70496976e-01
-3.90313864e-02 -1.04912364e+00 3.85701001e-01 7.09801078e-01
6.76828206e-01 7.91201413e-01 1.89456284e-01 -3.51666331e-01
1.63851470e-01 3.85461673e-02 1.13005996e-01 7.10194468e-01
6.03347719e-01 9.21088696e-01 8.00316274e-01 -1.36484131e-01
-1.34060228e+00 -1.48665452e+00 -1.26191853e-02 8.09098601e-01
-5.67728421e-03 -8.00464272e-01 -8.22478235e-01 -3.14875573e-01
-2.55091906e-01 1.11204064e+00 -6.37877941e-01 2.23546326e-01
-4.75729108e-01 -2.26439014e-01 1.33527413e-01 3.45160127e-01
2.29608580e-01 -1.02194643e+00 -8.28985453e-01 3.28349434e-02
5.69807924e-03 -1.02792394e+00 -5.04889846e-01 7.01650279e-04
-7.08955526e-01 -8.36456478e-01 -4.51647401e-01 -6.44980907e-01
1.07851517e+00 5.68298846e-02 1.39113379e+00 -5.67641631e-02
-3.76090050e-01 5.86927891e-01 -1.14127623e-04 -6.20877743e-02
-5.41351795e-01 -8.77226442e-02 -4.71468061e-01 -1.03104919e-01
-5.49130976e-01 -7.85731733e-01 -5.22328079e-01 -2.46668205e-01
-1.13838887e+00 9.69067216e-01 9.86261889e-02 4.60852534e-01
7.36178577e-01 -2.69112941e-02 -5.95955476e-02 -1.19182909e+00
6.80359066e-01 -2.86729068e-01 -7.86691964e-01 3.56323868e-01
-1.62256479e-01 4.51248586e-01 7.74175823e-01 -2.42833108e-01
-1.34277654e+00 1.85710356e-01 -1.34826461e-02 -5.67256391e-01
-4.24507141e-01 4.74075079e-01 -6.75759852e-01 5.82735062e-01
6.80634439e-01 1.41974375e-01 -3.62099320e-01 -3.63697916e-01
1.27035403e+00 -1.81120038e-02 6.06601715e-01 -1.04176199e+00
9.03659582e-01 3.73916447e-01 1.61985323e-01 -8.26255620e-01
-5.27715683e-01 2.72145331e-01 -9.31284130e-01 -8.12647045e-02
1.13760674e+00 -7.07447231e-01 -4.25372928e-01 1.57109022e-01
-1.29607129e+00 -6.93221331e-01 -6.10482693e-01 -1.90375000e-01
-7.78644562e-01 4.00986560e-02 -1.84508562e-01 -9.92350996e-01
1.62652090e-01 -1.37957120e+00 1.50478983e+00 1.08398676e-01
-4.98497128e-01 -1.03137732e+00 -1.11409031e-01 -3.05540949e-01
1.16658799e-01 7.73201466e-01 1.34759641e+00 1.21704444e-01
-1.34503007e+00 2.21466258e-01 -2.16394350e-01 -1.50044054e-01
3.98430496e-01 6.09588802e-01 -1.07090914e+00 -6.87375963e-02
-4.62408870e-01 3.35493349e-02 4.30205613e-01 5.04525185e-01
1.12424147e+00 -3.19325387e-01 -2.00935900e-01 1.10069978e+00
1.63783276e+00 2.37023205e-01 7.37723827e-01 1.06341792e-02
1.24180615e+00 7.51023769e-01 1.21491604e-01 4.80162352e-01
4.67642158e-01 4.48183388e-01 3.43995303e-01 -2.83114225e-01
-2.29754046e-01 -8.50444555e-01 4.28614095e-02 4.39377129e-01
5.45266628e-01 -4.92499352e-01 -7.25808144e-01 3.95003647e-01
-1.73848259e+00 -8.35838139e-01 -1.71605408e-01 2.11439514e+00
6.20057404e-01 8.93091336e-02 1.07422121e-01 -2.79155910e-01
3.24134022e-01 4.58850116e-01 -4.97075379e-01 -6.79884851e-01
-1.28866702e-01 2.55707711e-01 1.68918557e-02 8.63403678e-01
-6.16818726e-01 1.12475252e+00 7.21631002e+00 5.84094405e-01
-8.63044024e-01 -4.83265668e-01 6.83324158e-01 -1.49827942e-01
-1.37902069e+00 3.70849609e-01 -4.99000579e-01 1.43434793e-01
4.92731452e-01 -2.42304742e-01 7.46321797e-01 6.93235636e-01
2.97634274e-01 -1.89987630e-01 -1.50930524e+00 9.07105088e-01
-1.99899226e-01 -1.66788316e+00 5.43384314e-01 5.59402443e-02
8.79823327e-01 -6.80458665e-01 1.81547180e-01 3.89553644e-02
6.88932478e-01 -1.06079912e+00 1.04835844e+00 6.55896544e-01
1.15090096e+00 -6.97974741e-01 -3.07557106e-01 3.51929963e-01
-1.34244049e+00 3.84728134e-01 -7.78387412e-02 -5.19395322e-02
4.91888314e-01 4.42115247e-01 -9.00590420e-01 7.10631609e-01
1.32108778e-01 5.46421230e-01 -6.41248167e-01 7.21992970e-01
-4.98199314e-01 1.57023460e-01 -1.08890004e-01 1.55257005e-02
1.44394875e-01 -6.05923057e-01 5.00357389e-01 1.34426129e+00
2.69937187e-01 -1.10989794e-01 2.80546516e-01 1.66366160e+00
1.60703808e-01 1.39741525e-02 -1.11037397e+00 -1.71812461e-03
4.38725531e-01 1.19380224e+00 -8.71371686e-01 -4.34473395e-01
-1.33149743e-01 1.22761595e+00 2.98463434e-01 9.21600938e-01
-8.65788281e-01 -1.86932221e-01 7.13523686e-01 1.25789598e-01
1.15438834e-01 -5.20363986e-01 -7.98126221e-01 -1.19704235e+00
-2.87532389e-01 -7.22934961e-01 -2.69847270e-02 -1.45467174e+00
-9.27854896e-01 5.11953771e-01 4.39304978e-01 -1.00921690e+00
-4.36512858e-01 -4.03899103e-01 -7.44942367e-01 1.30530715e+00
-1.01536536e+00 -1.35736191e+00 -2.15293169e-01 5.18026233e-01
6.65363073e-01 1.62420064e-01 8.80939424e-01 -2.48469979e-01
-4.07514632e-01 2.37261936e-01 -3.99507463e-01 -1.60963878e-01
2.61709362e-01 -1.61627591e+00 8.39397192e-01 1.12412870e+00
3.33545834e-01 9.54148173e-01 8.48028898e-01 -7.57181585e-01
-1.51866448e+00 -1.08164036e+00 3.82488221e-01 -7.67069280e-01
3.76507699e-01 -7.86852360e-01 -6.80620909e-01 7.91804433e-01
4.65772629e-01 -4.64493096e-01 4.66144145e-01 1.12461887e-01
-4.18959260e-01 2.24394009e-01 -1.05462670e+00 1.24638534e+00
1.38716483e+00 -8.26144874e-01 -2.29762018e-01 5.11629023e-02
9.77482319e-01 -7.33738422e-01 -5.07075965e-01 -4.05890308e-02
4.65532005e-01 -9.19065237e-01 9.78739679e-01 -5.09928107e-01
7.45130658e-01 -7.30342984e-01 -4.19783026e-01 -1.41629660e+00
-6.16060376e-01 -7.21852601e-01 1.03911743e-01 1.15721333e+00
6.74829066e-01 -2.07770959e-01 6.68024242e-01 1.00791562e+00
-3.57194752e-01 -5.29085696e-01 -1.89864233e-01 -2.92121500e-01
-8.91498178e-02 -7.16478765e-01 9.63398516e-01 5.89551926e-01
-2.74610698e-01 5.23740768e-01 -1.60544455e-01 2.58954108e-01
7.59040117e-01 5.62784016e-01 1.26220071e+00 -9.00873899e-01
-6.55634463e-01 -5.04396677e-01 9.62183401e-02 -1.42511630e+00
1.23205796e-01 -9.88124132e-01 1.21433184e-01 -2.14491367e+00
3.94486673e-02 -4.39003438e-01 2.65551805e-01 1.09818906e-01
-9.89070088e-02 -9.27764401e-02 4.14472103e-01 -2.63757914e-01
-3.72883618e-01 5.40975869e-01 1.82462013e+00 -4.81886007e-02
-3.79314721e-01 -4.12403047e-01 -1.03257751e+00 5.73097706e-01
4.09454644e-01 -9.32336822e-02 -1.04780698e+00 -6.55729830e-01
3.74845266e-01 3.30315083e-01 3.90560776e-01 -5.41541338e-01
3.51248085e-02 -4.80338693e-01 5.56689918e-01 -7.05584109e-01
5.16899526e-01 -6.23509288e-01 7.00210035e-01 -1.18606456e-01
-4.51792121e-01 5.49182706e-02 2.76351303e-01 3.62597048e-01
2.33563095e-01 7.48742148e-02 5.51421344e-01 -3.45438749e-01
-5.72918534e-01 2.46297792e-01 -1.76509991e-01 1.26778940e-02
6.79436147e-01 -6.50667787e-01 -2.72079676e-01 -5.54570615e-01
-8.71638417e-01 1.36470661e-01 1.03306520e+00 2.62368470e-01
7.40175486e-01 -1.42175472e+00 -4.38940704e-01 5.38202703e-01
1.70791090e-01 5.51374972e-01 2.08057866e-01 7.84095228e-02
-6.36293828e-01 1.01573072e-01 -1.50201172e-01 -7.66342402e-01
-8.84507537e-01 6.01011276e-01 2.20021054e-01 -1.92716941e-02
-6.49808466e-01 8.79960120e-01 1.02326500e+00 -5.72335541e-01
-1.03611991e-01 -8.22564900e-01 3.38117540e-01 -2.25045577e-01
2.26666972e-01 -4.41196971e-02 -3.14743012e-01 -3.89280438e-01
-1.58533037e-01 5.06945431e-01 4.38149303e-01 -6.93823338e-01
9.76863325e-01 -1.69342250e-01 -4.88659833e-04 6.54031098e-01
1.03475440e+00 2.99376130e-01 -1.68750262e+00 1.75764844e-01
-3.74503583e-01 -6.56027198e-01 -2.65019029e-01 -7.71122456e-01
-7.46445000e-01 8.21306705e-01 -1.19611785e-01 1.98838964e-01
1.03846109e+00 -8.85833874e-02 1.88814506e-01 8.61943141e-02
5.23429573e-01 -6.40745521e-01 1.48558661e-01 3.51232380e-01
1.19460714e+00 -6.95496202e-01 2.85579786e-02 -6.29874825e-01
-7.72074342e-01 9.91012752e-01 5.48821867e-01 -2.81441242e-01
3.99469495e-01 5.26696384e-01 -6.78995475e-02 -2.84936368e-01
-7.26855278e-01 -2.40080103e-01 6.26242697e-01 7.28836238e-01
5.02853751e-01 2.04366326e-01 3.68417680e-01 -3.66372312e-03
-5.42174578e-01 -3.21091771e-01 4.89257574e-01 7.70889997e-01
-2.89903879e-01 -1.18815815e+00 -2.09668055e-01 2.83429265e-01
2.24106178e-01 -2.16862991e-01 -3.98251355e-01 6.98542356e-01
3.60963941e-02 7.51787364e-01 2.54902035e-01 -1.90462872e-01
4.83397365e-01 6.05649827e-03 8.20715368e-01 -1.09069180e+00
-2.87450641e-01 4.82884139e-01 3.24738085e-01 -6.70651793e-01
-1.52324531e-02 -5.23950636e-01 -1.18249130e+00 -1.88763678e-01
1.09749466e-01 -5.01065888e-03 5.36024868e-01 6.19643152e-01
3.40402037e-01 9.25199747e-01 5.25854051e-01 -1.04139745e+00
2.88851321e-01 -4.83952850e-01 -6.69515848e-01 4.65151638e-01
4.82106686e-01 -3.39319736e-01 5.03809229e-02 5.12724578e-01]
|
[11.1659574508667, -0.32497742772102356]
|
004f9660-b0a6-43c8-84ec-c4d4eb308771
|
using-causal-analysis-for-conceptual-deep
|
2107.06098
| null |
https://arxiv.org/abs/2107.06098v1
|
https://arxiv.org/pdf/2107.06098v1.pdf
|
Using Causal Analysis for Conceptual Deep Learning Explanation
|
Model explainability is essential for the creation of trustworthy Machine Learning models in healthcare. An ideal explanation resembles the decision-making process of a domain expert and is expressed using concepts or terminology that is meaningful to the clinicians. To provide such an explanation, we first associate the hidden units of the classifier to clinically relevant concepts. We take advantage of radiology reports accompanying the chest X-ray images to define concepts. We discover sparse associations between concepts and hidden units using a linear sparse logistic regression. To ensure that the identified units truly influence the classifier's outcome, we adopt tools from Causal Inference literature and, more specifically, mediation analysis through counterfactual interventions. Finally, we construct a low-depth decision tree to translate all the discovered concepts into a straightforward decision rule, expressed to the radiologist. We evaluated our approach on a large chest x-ray dataset, where our model produces a global explanation consistent with clinical knowledge.
|
['Kayhan Batmanghelich', 'Sofia Triantafillou', 'Stephen Wallace', 'Sumedha Singla']
|
2021-07-10
| null | null | null | null |
['clinical-knowledge']
|
['miscellaneous']
|
[ 6.65601015e-01 9.08288538e-01 -9.25079107e-01 -5.32894254e-01
-5.25096595e-01 -2.25426272e-01 3.46448481e-01 4.84478056e-01
4.77516614e-02 1.06651950e+00 8.02698493e-01 -1.00943494e+00
-4.99124736e-01 -5.89169323e-01 -8.39411855e-01 -4.23663229e-01
-5.38453385e-02 5.14410436e-01 -6.31235600e-01 4.08646941e-01
7.62672201e-02 1.68548152e-01 -9.52786505e-01 7.09759653e-01
9.46659982e-01 6.24707580e-01 -8.04693028e-02 4.40641195e-01
1.28624402e-02 1.32212913e+00 -1.90752074e-01 -3.61346245e-01
-1.56122521e-02 -8.83378923e-01 -7.56258368e-01 -1.83805257e-01
-2.02658430e-01 -2.52257317e-01 6.90745264e-02 6.34155214e-01
-3.71368974e-02 -2.46800020e-01 9.99525905e-01 -1.23649418e+00
-8.79189074e-01 1.29022312e+00 -2.25570276e-01 -9.26986113e-02
4.80503231e-01 -2.78241616e-02 9.03717279e-01 -6.92397475e-01
8.35921347e-01 1.16061902e+00 6.82509124e-01 6.82258546e-01
-1.43272781e+00 -9.22705114e-01 2.07917973e-01 1.35851651e-01
-9.41409528e-01 -2.46139601e-01 3.99157792e-01 -5.32506883e-01
7.14004815e-01 6.40559018e-01 7.18266785e-01 1.35110462e+00
8.41348350e-01 1.57315999e-01 1.17084396e+00 -6.25840962e-01
5.56427956e-01 2.14724839e-01 2.19270751e-01 6.38844371e-01
7.52942741e-01 3.88221443e-01 -5.68204045e-01 -6.19211137e-01
7.16439009e-01 4.35161412e-01 -2.38644689e-01 -1.50852986e-02
-1.32518435e+00 1.17613947e+00 5.50705910e-01 2.01227039e-01
-9.22014296e-01 4.30387974e-01 -1.34650692e-01 2.94023938e-02
1.98115468e-01 6.42556250e-01 -4.96595383e-01 3.49779576e-01
-8.21452260e-01 -8.21414776e-03 1.01941240e+00 4.91886109e-01
2.37104923e-01 -3.44831616e-01 -1.68747380e-01 2.03185380e-01
4.42882031e-01 3.35415393e-01 4.20744866e-01 -1.16995919e+00
6.53753653e-02 5.04404545e-01 8.16487987e-03 -7.88162887e-01
-2.91898102e-01 -5.46834350e-01 -8.58008564e-01 -4.78216484e-02
1.19182706e-01 -1.55451328e-01 -8.64317536e-01 1.63596809e+00
2.60079741e-01 2.02379405e-01 2.12298140e-01 8.82743776e-01
6.20587587e-01 9.58619416e-02 6.46151662e-01 -6.72829747e-01
1.51702714e+00 -4.68261868e-01 -1.11521745e+00 -9.48757827e-02
8.94489706e-01 -4.12320435e-01 7.03422427e-01 4.88728821e-01
-8.33897948e-01 -1.03964441e-01 -9.12544370e-01 6.28373921e-02
8.69377702e-02 -1.96037665e-01 8.30650508e-01 3.56955558e-01
-3.32264513e-01 4.75099295e-01 -6.97724462e-01 -1.26943707e-01
5.63989103e-01 3.15597296e-01 -3.39003265e-01 -3.28127563e-01
-1.31121278e+00 9.86817479e-01 1.90455139e-01 -1.86596319e-01
-8.16129327e-01 -1.26529562e+00 -8.06420624e-01 2.33920097e-01
6.05719507e-01 -1.58812332e+00 1.29422188e+00 -8.59847784e-01
-9.26074445e-01 5.29436648e-01 -4.42144394e-01 -5.95393062e-01
2.84022689e-01 -1.00818329e-01 -2.49372557e-01 1.50170520e-01
3.56671989e-01 4.30697858e-01 7.40533531e-01 -1.36010003e+00
-6.36238158e-01 -4.32602495e-01 -2.02288300e-01 -1.74279898e-01
1.58515528e-01 -2.09645271e-01 2.39588201e-01 -7.13111579e-01
2.74886012e-01 -9.31954265e-01 -7.48559892e-01 -1.10057034e-01
-7.22915053e-01 5.84560186e-02 2.49176532e-01 -7.22703576e-01
1.35064542e+00 -1.74590671e+00 -7.19941929e-02 5.82161486e-01
8.88992608e-01 -6.54847562e-01 4.33209330e-01 1.32934496e-01
-6.64269388e-01 3.94769013e-01 -4.02042538e-01 8.65921304e-02
-2.66972452e-01 5.23183584e-01 -4.61340606e-01 3.13527793e-01
1.95467114e-01 1.15164542e+00 -8.84762824e-01 -5.93153358e-01
1.40313119e-01 3.60668331e-01 -8.69937360e-01 1.77459478e-01
-2.50050366e-01 4.71318662e-01 -6.77455962e-01 4.82341498e-01
1.13561824e-01 -7.06684947e-01 6.82583451e-01 -3.66178490e-02
1.50930464e-01 4.98972118e-01 -6.63659871e-01 1.44553924e+00
-4.61103886e-01 1.42669290e-01 -4.27143604e-01 -9.39770222e-01
6.14401698e-01 5.78538775e-01 6.70466900e-01 -2.57362366e-01
3.69708464e-02 6.81393370e-02 2.23106503e-01 -6.47538185e-01
-2.28871480e-01 -9.16183412e-01 4.92638163e-02 7.74802327e-01
-3.06280732e-01 2.01240614e-01 -6.37761831e-01 3.46599430e-01
1.20673978e+00 -3.48050088e-01 1.05620062e+00 -2.13159710e-01
4.84796464e-02 3.97340178e-01 5.89838266e-01 8.99379134e-01
3.93330157e-01 5.73826909e-01 7.74861932e-01 -5.54414093e-01
-8.01137090e-01 -9.18024659e-01 -2.75891393e-01 5.07232189e-01
-4.04634029e-01 -2.76329666e-01 -5.53531051e-01 -7.67174482e-01
2.10958675e-01 1.39317179e+00 -1.22410417e+00 -3.53654981e-01
-1.79986313e-01 -5.78182220e-01 2.15374783e-01 5.95565200e-01
-2.18758285e-01 -7.58935928e-01 -9.40694869e-01 3.54809344e-01
-2.02050641e-01 -9.21160877e-01 -1.70312643e-01 3.08517694e-01
-1.01538205e+00 -1.36594200e+00 -1.88845366e-01 -8.37706253e-02
8.40700686e-01 -2.01983407e-01 1.15182650e+00 3.24346691e-01
-3.48003805e-01 3.29610109e-01 -2.54458725e-01 -6.75646961e-01
-7.90391147e-01 -3.69550407e-01 -2.74196137e-02 -2.63521403e-01
4.58260447e-01 -3.61907631e-01 -8.24796379e-01 5.89218028e-02
-8.05750251e-01 4.39918160e-01 9.64005113e-01 9.52090502e-01
6.72257841e-01 -2.11313233e-01 6.56516552e-01 -1.51724124e+00
7.07712293e-01 -1.05574393e+00 6.74896059e-04 3.52140367e-01
-1.25567770e+00 2.28406653e-01 3.41143966e-01 -4.56944674e-01
-1.08201551e+00 1.38671309e-01 2.74079949e-01 -3.34999681e-01
-2.19377980e-01 1.05514050e+00 9.39459056e-02 5.78048825e-01
9.68202889e-01 -2.88067043e-01 1.06201254e-01 -3.46579969e-01
8.08564782e-01 5.24370551e-01 4.57095146e-01 -4.53648627e-01
4.05026108e-01 5.65157056e-01 2.12259918e-01 -2.34493315e-01
-8.10794055e-01 -1.79902300e-01 -3.07796776e-01 1.53682396e-01
1.11606991e+00 -7.12136686e-01 -7.59006858e-01 -8.88224542e-01
-1.06700253e+00 9.07600950e-03 -5.59418678e-01 1.00689948e+00
-4.69476819e-01 -1.95848763e-01 -1.66752264e-01 -8.31625521e-01
-1.72569126e-01 -9.62945461e-01 8.23005140e-01 -2.09212631e-01
-1.11060929e+00 -1.07822490e+00 2.18652692e-02 3.14887643e-01
-4.49562445e-04 4.44672495e-01 1.56719232e+00 -8.82067442e-01
-4.95338976e-01 -2.71071136e-01 -1.36444792e-01 -2.67554492e-01
3.35415065e-01 -2.12641060e-01 -8.54545653e-01 3.80399257e-01
3.28768671e-01 1.30826816e-01 6.56460464e-01 1.02584338e+00
1.46679628e+00 -7.40654051e-01 -7.78222620e-01 3.28214586e-01
1.10968232e+00 2.57192343e-01 3.12381297e-01 -1.58433333e-01
5.79758406e-01 8.45869124e-01 4.63034242e-01 4.55003053e-01
4.58022505e-01 2.90626884e-01 4.93336432e-02 -4.81655091e-01
7.61160776e-02 -6.78867161e-01 -1.60207108e-01 3.02375585e-01
-2.90672660e-01 1.58738375e-01 -1.01449692e+00 2.80613422e-01
-1.86668491e+00 -6.88996494e-01 -3.21315914e-01 2.01361656e+00
7.77438223e-01 1.75745264e-02 -2.70736873e-01 -2.21881159e-02
3.13580632e-01 -6.54284418e-01 -5.54415107e-01 -4.51903492e-01
2.60890692e-01 2.33199790e-01 3.92215252e-01 5.97338557e-01
-3.10006142e-01 4.18955445e-01 7.37802124e+00 3.09088975e-01
-9.10544276e-01 2.81185865e-01 9.40044880e-01 -2.54090071e-01
-1.06769228e+00 3.68382245e-01 -9.30367112e-02 1.14253499e-01
1.06795073e+00 -4.99908298e-01 1.07427679e-01 6.61091685e-01
7.90419698e-01 -1.06888838e-01 -1.70956993e+00 5.19436836e-01
-2.48542607e-01 -1.74899435e+00 2.50941962e-01 3.15800071e-01
7.47837067e-01 -6.49086297e-01 8.16181004e-02 -5.69482185e-02
6.52164340e-01 -1.69315243e+00 3.97125781e-01 7.26810038e-01
9.90721047e-01 -2.88661987e-01 7.29798853e-01 1.90679431e-01
-4.81549054e-01 -3.70809883e-01 -1.03835307e-01 -2.25226045e-01
-1.57862920e-02 8.20616186e-01 -1.53795147e+00 6.29723072e-01
3.85531783e-01 4.68063980e-01 6.14526123e-02 2.71529734e-01
-5.32408476e-01 1.05268013e+00 5.12098484e-02 2.66074866e-01
-1.09634869e-01 1.06960930e-01 1.32571056e-01 9.94370759e-01
3.40151131e-01 8.37606013e-01 -2.34373987e-01 1.46339202e+00
-9.45286676e-02 2.42787197e-01 -9.15877461e-01 1.63261071e-02
3.97737503e-01 9.08636570e-01 -4.96056348e-01 -5.90933263e-01
-5.03778636e-01 4.78603065e-01 -1.19513385e-01 5.22494555e-01
-7.50187695e-01 5.29320717e-01 4.74535376e-01 4.09279644e-01
-1.92925274e-01 4.91902530e-01 -1.01927197e+00 -9.29526448e-01
-3.32413405e-01 -1.07544613e+00 7.39568233e-01 -1.01401329e+00
-1.23975718e+00 2.88206846e-01 3.04192692e-01 -9.20438528e-01
-6.72522247e-01 -3.00172865e-01 -4.47794795e-01 1.04671276e+00
-1.11909509e+00 -1.20078981e+00 -1.02158755e-01 4.01373923e-01
2.49151379e-01 1.10219689e-02 1.13366127e+00 -2.41200820e-01
-3.80717278e-01 2.87334383e-01 -4.52522755e-01 -2.07483098e-01
5.86615086e-01 -1.24259162e+00 -6.98854215e-03 4.65846539e-01
6.18214831e-02 1.26066363e+00 9.08091307e-01 -1.17556572e+00
-1.10668814e+00 -9.45579886e-01 1.22721493e+00 -6.64803386e-01
5.84726572e-01 7.10968971e-02 -9.79527414e-01 9.96774793e-01
-7.93049932e-02 -3.09541762e-01 1.37987924e+00 4.47455436e-01
-2.50383347e-01 2.95714319e-01 -1.09195673e+00 5.90183794e-01
9.09586668e-01 -3.72256428e-01 -1.15143406e+00 2.71227390e-01
9.17235255e-01 -3.32665741e-02 -9.56971347e-01 2.38826320e-01
8.08205485e-01 -4.59502935e-01 8.29854548e-01 -1.42945337e+00
1.08764434e+00 5.04450276e-02 -1.24000020e-01 -1.20604098e+00
-3.41345847e-01 -4.27456081e-01 -1.25186011e-01 3.33009511e-01
9.68658447e-01 -6.53919756e-01 6.48046672e-01 1.33428323e+00
1.17464915e-01 -7.53549337e-01 -7.95773923e-01 -1.06108993e-01
1.34200960e-01 -6.21393621e-01 7.31862783e-01 1.48330319e+00
5.37244201e-01 3.60849470e-01 -2.31696755e-01 2.63798118e-01
7.81619787e-01 3.59693140e-01 2.88145870e-01 -1.20736516e+00
-4.69550639e-01 -8.89154896e-03 9.25270095e-02 -4.16611046e-01
9.90863740e-02 -1.11516225e+00 -2.91525692e-01 -1.61889458e+00
7.31223226e-01 -5.05855441e-01 -4.22273725e-01 6.88208580e-01
-4.68506932e-01 -2.11222500e-01 -4.89802733e-02 2.99427688e-01
-1.63852014e-02 1.89404711e-01 1.14160728e+00 1.28282949e-01
-2.02373967e-01 -8.66700038e-02 -1.39540136e+00 7.46887863e-01
5.70964575e-01 -1.13508141e+00 -5.05452514e-01 -1.13007044e-02
2.78598040e-01 5.99328697e-01 6.33024216e-01 -1.98331103e-01
3.02725047e-01 -4.28343952e-01 5.85646749e-01 -2.90453658e-02
-2.08728939e-01 -1.05204916e+00 7.46437252e-01 1.12719893e+00
-8.83861959e-01 -2.09269390e-01 -6.28334880e-02 6.16616368e-01
-1.54756969e-02 -7.91447982e-02 2.65967160e-01 -1.32776335e-01
-1.06700018e-01 -8.95653367e-02 -3.79268825e-01 -3.16600174e-01
9.80585456e-01 -6.03704415e-02 -1.26667455e-01 -5.88543892e-01
-1.11226845e+00 7.23837167e-02 8.56883004e-02 1.19318731e-01
9.00226176e-01 -1.09360480e+00 -1.00014067e+00 1.14165753e-01
1.01124547e-01 -2.71313995e-01 8.83207098e-02 9.64923561e-01
-2.09689379e-01 7.52496243e-01 3.23838517e-02 -4.80020404e-01
-1.07479429e+00 8.07105064e-01 1.96101829e-01 -1.46490216e-01
-7.10207999e-01 2.63785899e-01 6.99158430e-01 8.26637298e-02
-2.10320368e-01 -8.47096026e-01 -1.06688542e-02 -2.56331474e-01
4.62964952e-01 1.60319850e-01 -2.57062972e-01 2.27196589e-02
-3.83645296e-01 9.23891291e-02 1.57218277e-01 -2.95802593e-01
1.46816301e+00 -1.89379733e-02 -1.07019946e-01 4.81439769e-01
7.90248036e-01 -6.49270713e-02 -7.45931685e-01 -1.95207782e-02
4.06623073e-02 -2.61837810e-01 1.41369298e-01 -1.15117264e+00
-5.58565795e-01 5.34902215e-01 2.38758966e-01 -1.63445696e-02
9.26736951e-01 2.86831051e-01 5.40162362e-02 1.35078475e-01
-4.10315879e-02 -4.34122741e-01 -1.84671700e-01 -5.14059722e-01
1.12347364e+00 -1.19392145e+00 3.44994038e-01 -6.44340575e-01
-8.23583007e-01 9.17208493e-01 4.88421582e-02 2.12757096e-01
7.32559681e-01 2.91495919e-01 2.51873165e-01 -4.87773627e-01
-1.22573709e+00 2.89870113e-01 4.53951418e-01 4.72967416e-01
6.24352574e-01 5.39696813e-01 -6.04375839e-01 1.30024970e+00
-2.56875753e-01 5.14514387e-01 3.90459925e-01 4.21677768e-01
-3.92763381e-04 -9.39342141e-01 -6.32528722e-01 8.02901506e-01
-5.53256512e-01 -4.37814474e-01 -5.48504651e-01 9.23452377e-01
1.75072879e-01 9.99199390e-01 -1.12881564e-01 -3.32668662e-01
3.02942932e-01 2.76616931e-01 2.19102398e-01 -8.02426755e-01
-3.33763242e-01 1.14291146e-01 2.56603718e-01 -7.50111818e-01
-2.97269970e-01 -6.01826787e-01 -1.60413206e+00 -2.61112541e-01
-1.55468553e-01 3.73161167e-01 5.45765162e-01 1.20041025e+00
1.90212980e-01 9.15757835e-01 3.50019246e-01 1.45754099e-01
-5.18695474e-01 -8.00207138e-01 -2.84402281e-01 4.12304223e-01
2.98157364e-01 -4.78211373e-01 -3.61902654e-01 3.69967282e-01]
|
[8.48068618774414, 5.676583290100098]
|
c1fb4f31-35f9-4803-84f2-d1a0417c220d
|
weakly-supervised-object-localization-via
|
2207.10447
| null |
https://arxiv.org/abs/2207.10447v2
|
https://arxiv.org/pdf/2207.10447v2.pdf
|
Weakly Supervised Object Localization via Transformer with Implicit Spatial Calibration
|
Weakly Supervised Object Localization (WSOL), which aims to localize objects by only using image-level labels, has attracted much attention because of its low annotation cost in real applications. Recent studies leverage the advantage of self-attention in visual Transformer for long-range dependency to re-active semantic regions, aiming to avoid partial activation in traditional class activation mapping (CAM). However, the long-range modeling in Transformer neglects the inherent spatial coherence of the object, and it usually diffuses the semantic-aware regions far from the object boundary, making localization results significantly larger or far smaller. To address such an issue, we introduce a simple yet effective Spatial Calibration Module (SCM) for accurate WSOL, incorporating semantic similarities of patch tokens and their spatial relationships into a unified diffusion model. Specifically, we introduce a learnable parameter to dynamically adjust the semantic correlations and spatial context intensities for effective information propagation. In practice, SCM is designed as an external module of Transformer, and can be removed during inference to reduce the computation cost. The object-sensitive localization ability is implicitly embedded into the Transformer encoder through optimization in the training phase. It enables the generated attention maps to capture the sharper object boundaries and filter the object-irrelevant background area. Extensive experimental results demonstrate the effectiveness of the proposed method, which significantly outperforms its counterpart TS-CAM on both CUB-200 and ImageNet-1K benchmarks. The code is available at https://github.com/164140757/SCM.
|
['Xiang Wan', 'Jiong Wang', 'Ruimao Zhang', 'Haotian Bai']
|
2022-07-21
| null | null | null | null |
['weakly-supervised-object-localization', 'long-range-modeling']
|
['computer-vision', 'natural-language-processing']
|
[ 3.22010741e-02 7.25616440e-02 -3.35070401e-01 -4.56320614e-01
-6.75817788e-01 -3.74555379e-01 4.83881235e-01 -2.53708544e-03
-3.89939934e-01 4.49842572e-01 4.96943966e-02 8.74941051e-02
-1.01999104e-01 -6.39134288e-01 -9.84439433e-01 -9.89582241e-01
1.85653090e-01 1.34145498e-01 7.38734961e-01 -2.19030324e-02
4.61152568e-02 2.80677289e-01 -1.04540873e+00 1.62473813e-01
1.13363290e+00 1.28527021e+00 7.79513001e-01 -1.09656848e-01
-2.21203968e-01 7.89999664e-01 -3.34095091e-01 -1.33344218e-01
1.27420083e-01 -2.87376016e-01 -5.47426760e-01 3.95471789e-02
4.01609033e-01 -1.01286791e-01 -3.57982725e-01 1.34482265e+00
3.82298350e-01 1.72741026e-01 3.51024628e-01 -1.09502721e+00
-9.47478354e-01 6.55650437e-01 -8.26162636e-01 4.54368740e-01
-2.32622415e-01 4.06620204e-01 9.97196555e-01 -1.10157371e+00
2.47020528e-01 1.17315197e+00 4.71893519e-01 3.55389208e-01
-1.27855194e+00 -8.73507142e-01 6.89169407e-01 2.34217778e-01
-1.80381525e+00 -3.48016173e-01 1.13296545e+00 -1.75939143e-01
5.59202135e-01 1.87979247e-02 6.50384724e-01 9.57339406e-01
-1.56556353e-01 1.02674639e+00 1.00866532e+00 -1.27163157e-01
1.83709443e-01 2.68902987e-01 9.15301293e-02 7.91873515e-01
2.13911697e-01 -1.75597578e-01 -3.96014661e-01 2.26547062e-01
1.09705949e+00 2.70905972e-01 -4.83363330e-01 -4.14377898e-01
-1.19146752e+00 6.30930245e-01 1.22335577e+00 4.64640647e-01
-1.78272739e-01 3.12827915e-01 4.86769341e-02 -3.15439105e-01
6.50692701e-01 1.86135590e-01 -3.66012454e-01 3.75195235e-01
-7.52739012e-01 -1.71264067e-01 1.48103490e-01 1.06329560e+00
1.05747616e+00 -1.28633052e-01 -4.89444196e-01 9.56475079e-01
3.30065072e-01 3.04652482e-01 4.19395506e-01 -7.53181517e-01
5.14074445e-01 8.87014329e-01 1.40847461e-02 -1.01176548e+00
-1.42947733e-01 -8.85513008e-01 -7.28716612e-01 -1.52314052e-01
2.87772983e-01 2.34799162e-01 -1.02221704e+00 1.88734627e+00
4.44552004e-01 6.44357204e-01 -3.76706153e-01 1.30555260e+00
7.18247235e-01 6.29382312e-01 2.70384908e-01 6.35105222e-02
1.30127549e+00 -1.29816747e+00 -5.67500770e-01 -6.28627121e-01
3.98060679e-01 -4.24001396e-01 1.35723257e+00 -8.91222358e-02
-9.43207800e-01 -5.96847177e-01 -1.04904199e+00 -2.56797552e-01
-3.08745682e-01 2.63251513e-01 6.72764599e-01 1.36251062e-01
-8.72476399e-01 2.51461953e-01 -8.67673397e-01 -3.04837283e-02
1.00188124e+00 3.09726030e-01 -5.77020831e-02 -1.12026736e-01
-1.25768888e+00 5.15171587e-01 4.85987484e-01 4.30788219e-01
-9.31017220e-01 -8.75710487e-01 -8.55876625e-01 2.66049743e-01
6.02454484e-01 -3.87472898e-01 9.08608258e-01 -1.28368163e+00
-1.27976370e+00 6.52913928e-01 -2.21257105e-01 -2.31992781e-01
3.63764733e-01 -9.14457962e-02 -1.39343604e-01 2.98174381e-01
3.61079335e-01 9.27064776e-01 8.64773214e-01 -1.39276516e+00
-6.23940587e-01 -3.83752376e-01 2.07779765e-01 3.73413444e-01
-4.39927757e-01 -2.64677852e-01 -1.08024812e+00 -1.06322408e+00
2.08103523e-01 -7.74974644e-01 -1.68436885e-01 1.67125329e-01
-3.47821563e-01 -2.59240329e-01 7.51638055e-01 -4.14869636e-01
1.08682835e+00 -2.27908254e+00 1.11904643e-01 1.17113441e-01
1.95124313e-01 1.94966406e-01 -3.07860136e-01 -2.01833367e-01
6.63440898e-02 -3.06977611e-02 -4.16701436e-01 -3.17622960e-01
-2.53558397e-01 2.23115683e-01 -2.50352144e-01 5.54371297e-01
4.70385969e-01 1.37855208e+00 -1.02938449e+00 -6.02528691e-01
3.53395611e-01 6.89639091e-01 -5.00264466e-01 3.26426700e-02
-3.47449273e-01 4.93462056e-01 -8.21211696e-01 7.79220164e-01
8.57494175e-01 -6.12846315e-01 -8.34657624e-02 -4.71055686e-01
-1.15610234e-01 2.69103110e-01 -9.60221946e-01 1.98481965e+00
-5.94931841e-01 4.06951666e-01 7.89324418e-02 -1.12885964e+00
6.85450256e-01 -1.18099086e-01 3.34002376e-01 -9.93366897e-01
1.31657019e-01 5.79311289e-02 -8.31962153e-02 -3.37623566e-01
-1.05362236e-01 2.86164165e-01 1.56433016e-01 -1.12546422e-03
1.09800711e-01 2.65592217e-01 -9.96513143e-02 2.52521276e-01
7.80169308e-01 1.29965261e-01 -1.25383765e-01 -4.40751463e-01
6.65523410e-01 -2.63709337e-01 7.54471600e-01 5.80578923e-01
-9.82861370e-02 6.08328640e-01 2.34161377e-01 -1.31736308e-01
-4.68275368e-01 -1.13593924e+00 -2.81394809e-01 9.65498090e-01
9.58385408e-01 -1.97042897e-01 -8.42631400e-01 -9.00587320e-01
3.62303071e-02 4.28945690e-01 -7.78379738e-01 -3.92285466e-01
-7.00489342e-01 -7.46036708e-01 2.10302711e-01 7.35625327e-01
7.65819252e-01 -1.07867050e+00 -2.24432498e-01 2.43453875e-01
-2.90143073e-01 -1.13534248e+00 -7.96255171e-01 1.91454172e-01
-7.07503557e-01 -8.00255001e-01 -8.98398578e-01 -8.54998231e-01
1.02209103e+00 5.97734809e-01 8.25317800e-01 1.68314740e-01
-2.83828795e-01 5.86368963e-02 -3.00679266e-01 -1.18081987e-01
2.20055297e-01 2.25002363e-01 -3.36730450e-01 3.59799057e-01
3.43332589e-01 -4.48775262e-01 -1.03274357e+00 6.37846529e-01
-8.34206581e-01 1.21707857e-01 8.68098557e-01 1.02817273e+00
7.85119593e-01 1.93450928e-01 6.01213634e-01 -7.87146568e-01
1.12872556e-01 -5.25830209e-01 -7.58109570e-01 3.30670089e-01
-4.46665585e-01 -3.40922810e-02 4.74833816e-01 -6.29828513e-01
-1.15320218e+00 8.79597068e-02 -3.21141370e-02 -5.85347354e-01
1.05521008e-01 2.85637498e-01 -5.64219534e-01 -3.49857062e-01
2.04849586e-01 4.74559903e-01 -3.38815331e-01 -5.03324628e-01
3.40943605e-01 3.05327058e-01 3.93986255e-01 -5.16164541e-01
7.57651508e-01 7.25365877e-01 -5.10190666e-01 -4.62890625e-01
-1.30272639e+00 -5.38219392e-01 -5.00336587e-01 -1.29439428e-01
7.31063426e-01 -1.11894023e+00 -5.46678662e-01 4.13365841e-01
-9.86409307e-01 -5.94190598e-01 -3.57812285e-01 3.35607409e-01
-1.62018269e-01 -9.82839242e-03 -5.00090718e-01 -4.19241816e-01
-1.68502435e-01 -1.25769758e+00 1.31301963e+00 4.98478413e-01
2.90414065e-01 -9.18391824e-01 -4.08205718e-01 3.71700615e-01
3.87947321e-01 -1.92229539e-01 7.87791312e-01 -3.25694323e-01
-1.13717115e+00 1.20658435e-01 -8.02415133e-01 4.02118504e-01
1.72859535e-01 -5.15905440e-01 -1.12541389e+00 -2.19583005e-01
-7.24190697e-02 -1.82153657e-01 1.23652411e+00 4.78860706e-01
1.61775374e+00 -3.14543426e-01 -5.74163556e-01 8.22163880e-01
1.41348481e+00 -6.61818832e-02 4.13886487e-01 1.03768863e-01
1.04809594e+00 4.15452868e-01 6.40349030e-01 9.44723263e-02
4.66699690e-01 8.28859091e-01 5.22840261e-01 -3.39652002e-01
-4.64978188e-01 -4.60987568e-01 1.86293766e-01 5.66811740e-01
1.11974701e-01 -1.30545065e-01 -6.59945726e-01 5.43917477e-01
-1.85693920e+00 -6.51071012e-01 1.04418583e-01 2.16389465e+00
9.64645207e-01 2.70679265e-01 -2.41349012e-01 -2.68398076e-01
6.92859650e-01 2.68822253e-01 -7.55561471e-01 5.17658234e-01
-1.03605665e-01 -3.28445956e-02 5.93343675e-01 5.27182877e-01
-1.08300686e+00 1.11701572e+00 4.57742834e+00 1.30784452e+00
-1.21573699e+00 5.04843831e-01 7.84283817e-01 -1.47229806e-01
-3.22216183e-01 -2.43523251e-02 -8.93027723e-01 8.28946352e-01
2.21878603e-01 2.75829762e-01 4.09716338e-01 8.04389954e-01
1.94776744e-01 -1.18805930e-01 -8.34216058e-01 8.60792100e-01
-7.25004151e-02 -1.36550725e+00 -2.94341352e-02 -1.14942387e-01
7.86664784e-01 2.03360975e-01 1.85714930e-01 3.21689844e-01
-1.83577150e-01 -7.22122788e-01 9.97360706e-01 4.39582467e-01
7.45354235e-01 -6.71682894e-01 5.47803879e-01 2.51079947e-01
-1.47852623e+00 -2.97929913e-01 -5.43173492e-01 2.66761780e-01
-2.09592842e-02 7.87899017e-01 -3.77931476e-01 1.27029568e-01
9.11011457e-01 9.89540875e-01 -7.81817734e-01 9.62442040e-01
-4.51161802e-01 6.63499713e-01 -3.87336701e-01 1.62178367e-01
4.63496983e-01 -2.05936283e-01 4.60852236e-01 1.09832180e+00
4.92058955e-02 1.10883199e-01 2.37495393e-01 1.35421765e+00
-1.90111667e-01 -5.59849329e-02 -6.93607926e-02 2.34498367e-01
6.26396954e-01 1.36361110e+00 -9.76895571e-01 -2.43079036e-01
-5.04661083e-01 1.12510026e+00 5.32430887e-01 6.99465275e-01
-1.17230761e+00 -2.95862883e-01 4.27561045e-01 4.02702779e-01
4.87683356e-01 -5.55492863e-02 -3.33851725e-01 -1.18042862e+00
2.95155764e-01 -4.29488152e-01 1.53227344e-01 -6.97874606e-01
-1.19616425e+00 5.57915270e-01 -1.05338536e-01 -1.05538011e+00
5.54264247e-01 -2.74858564e-01 -6.07061327e-01 8.81118953e-01
-1.82790554e+00 -1.33818364e+00 -6.28672063e-01 5.31807303e-01
7.06201375e-01 1.93251118e-01 1.92306355e-01 5.37706673e-01
-9.07148480e-01 6.81329191e-01 -1.72201470e-01 3.09940502e-02
6.19240463e-01 -1.08851671e+00 5.70621230e-02 7.79421031e-01
5.45601249e-02 7.67901063e-01 1.43055663e-01 -5.94598532e-01
-1.01328588e+00 -1.49610472e+00 4.32680815e-01 -4.11550939e-01
6.29853666e-01 -7.82449484e-01 -1.27053916e+00 3.83450449e-01
-3.70511226e-02 5.57536304e-01 8.16387981e-02 -3.06485891e-01
-3.83176506e-01 -5.22199869e-01 -7.50741720e-01 4.83603776e-01
1.24559236e+00 -5.61998665e-01 -3.06182653e-01 3.09620529e-01
1.05113685e+00 -2.57102013e-01 -5.18781364e-01 4.87697840e-01
1.07635148e-01 -8.08153689e-01 1.17725468e+00 1.94898590e-01
1.41510487e-01 -6.93695247e-01 2.23037049e-01 -1.09502935e+00
-4.11122352e-01 -1.99200362e-01 -1.76516995e-01 1.53825343e+00
3.49582940e-01 -6.90099895e-01 6.36691093e-01 3.15325171e-01
-2.08662376e-01 -1.04570258e+00 -8.28984678e-01 -6.97679102e-01
-2.28251740e-01 -2.18120903e-01 5.49092174e-01 9.52859044e-01
-5.66496134e-01 3.10699016e-01 -2.17100903e-02 4.08853829e-01
6.34896994e-01 2.56479561e-01 2.34196812e-01 -9.18473959e-01
-1.44535959e-01 -5.02616227e-01 -2.43084475e-01 -1.48902369e+00
2.49280289e-01 -1.00094664e+00 2.08333313e-01 -1.41624558e+00
2.65368700e-01 -1.00542378e+00 -7.15650380e-01 6.73918307e-01
-4.20943469e-01 4.96141613e-01 -3.97911184e-02 3.33663553e-01
-8.93547535e-01 9.27551985e-01 1.36698687e+00 -3.43440086e-01
-9.93487015e-02 -1.06722988e-01 -7.41670668e-01 8.34764063e-01
6.65676832e-01 -5.54264426e-01 -7.19483256e-01 -6.66851342e-01
-3.57657336e-02 -4.99555290e-01 8.01908910e-01 -8.58345687e-01
3.12266171e-01 -1.40226632e-01 5.62115490e-01 -3.35611194e-01
2.25415394e-01 -9.88527179e-01 -1.93288460e-01 1.71465680e-01
-3.13472092e-01 -5.04043400e-01 1.18276149e-01 7.19702721e-01
-3.54055762e-01 -4.06890512e-02 9.93419945e-01 -3.06578837e-02
-9.51989174e-01 6.54436290e-01 1.39810622e-01 1.18454590e-01
1.06963050e+00 -1.04635164e-01 -2.12349638e-01 9.16847363e-02
-3.99968326e-01 4.86556560e-01 5.33883810e-01 6.61911130e-01
6.59441650e-01 -1.30971169e+00 -3.23077023e-01 3.08791071e-01
2.73785740e-01 5.04180014e-01 5.00855386e-01 1.10475004e+00
-1.83993563e-01 2.73347795e-01 1.53752327e-01 -7.55931556e-01
-6.28196537e-01 6.77366495e-01 4.45813209e-01 1.55805945e-01
-8.21092427e-01 1.16478574e+00 1.02726483e+00 -4.63301763e-02
3.70994925e-01 -4.23212975e-01 -1.38276648e-02 -1.85414806e-01
5.44411957e-01 -5.70983998e-03 -8.07229951e-02 -6.80088520e-01
-5.79252183e-01 6.24346435e-01 -1.79127216e-01 2.53585309e-01
1.18820381e+00 -4.25080299e-01 -9.53228548e-02 2.44057700e-01
1.26846564e+00 -5.11699617e-02 -1.74350953e+00 -5.85709631e-01
-1.04356438e-01 -4.86908942e-01 4.36658561e-01 -7.34628141e-01
-1.46618915e+00 9.47431982e-01 6.46910310e-01 -1.79056570e-01
1.21772110e+00 3.66568655e-01 8.01651478e-01 3.13482480e-03
3.57885927e-01 -8.97575915e-01 3.09429705e-01 2.11830541e-01
8.74933600e-01 -1.37228918e+00 -2.44555607e-01 -6.03089452e-01
-6.35107875e-01 5.22767603e-01 1.08077836e+00 -1.21050440e-01
6.10506892e-01 3.04603547e-01 -8.59256014e-02 -1.70204058e-01
-4.52864408e-01 -1.73997313e-01 4.38102871e-01 3.85454893e-01
1.88474536e-01 -1.35674268e-01 7.52483383e-02 8.51855516e-01
5.65230310e-01 -1.66577175e-01 -1.93489969e-01 6.32387161e-01
-3.76638472e-01 -6.58559740e-01 -1.04902729e-01 2.54932880e-01
-2.31691495e-01 -3.37619752e-01 -1.38442129e-01 5.64522147e-01
5.34870565e-01 5.23948014e-01 1.96210817e-01 3.20906490e-02
2.21654460e-01 -3.67640078e-01 3.22640449e-01 -4.93613362e-01
-2.74539173e-01 4.25492585e-01 -5.63066244e-01 -7.57360041e-01
-4.63373810e-01 -3.52064252e-01 -1.54603827e+00 3.09655070e-01
-6.77415729e-01 1.02485918e-01 3.88747782e-01 7.49815226e-01
4.84320641e-01 7.62835026e-01 6.04904175e-01 -8.54754508e-01
-2.87347823e-01 -8.46468687e-01 -4.29659873e-01 2.56300181e-01
3.62721115e-01 -9.84807909e-01 -1.95179880e-01 -7.41171688e-02]
|
[9.603755950927734, 0.821819543838501]
|
fdb6e2a8-4bbe-47e8-ab9e-f226dd5ecd7f
|
multi-task-pre-training-for-plug-and-play
|
2109.14739
| null |
https://arxiv.org/abs/2109.14739v2
|
https://arxiv.org/pdf/2109.14739v2.pdf
|
Multi-Task Pre-Training for Plug-and-Play Task-Oriented Dialogue System
|
Pre-trained language models have been recently shown to benefit task-oriented dialogue (TOD) systems. Despite their success, existing methods often formulate this task as a cascaded generation problem which can lead to error accumulation across different sub-tasks and greater data annotation overhead. In this study, we present PPTOD, a unified plug-and-play model for task-oriented dialogue. In addition, we introduce a new dialogue multi-task pre-training strategy that allows the model to learn the primary TOD task completion skills from heterogeneous dialog corpora. We extensively test our model on three benchmark TOD tasks, including end-to-end dialogue modelling, dialogue state tracking, and intent classification. Experimental results show that PPTOD achieves new state of the art on all evaluated tasks in both high-resource and low-resource scenarios. Furthermore, comparisons against previous SOTA methods show that the responses generated by PPTOD are more factually correct and semantically coherent as judged by human annotators.
|
['Yi Zhang', 'Yi-An Lai', 'Deng Cai', 'Arshit Gupta', 'Elman Mansimov', 'Lei Shu', 'Yixuan Su']
|
2021-09-29
| null |
https://aclanthology.org/2022.acl-long.319
|
https://aclanthology.org/2022.acl-long.319.pdf
|
acl-2022-5
|
['end-to-end-dialogue-modelling']
|
['natural-language-processing']
|
[-1.93645041e-02 5.58311641e-01 6.63457513e-02 -6.20223820e-01
-1.06473625e+00 -7.81190813e-01 1.11118889e+00 -2.18787733e-02
-5.38547218e-01 9.46459413e-01 8.46700370e-01 -1.17406659e-01
6.58843994e-01 -2.02379286e-01 1.73407242e-01 -1.00010835e-01
4.26491708e-01 1.32185400e+00 1.84387416e-01 -7.80995727e-01
6.52991235e-02 -2.65740365e-01 -7.86456227e-01 8.67433965e-01
9.28098559e-01 6.34448588e-01 2.02310875e-01 9.36577141e-01
-3.27401608e-01 1.45936537e+00 -1.11668801e+00 -8.53419185e-01
-1.51780039e-01 -6.27433181e-01 -1.71828306e+00 1.15320235e-01
1.69501573e-01 -7.15014219e-01 -7.38776550e-02 5.14116585e-01
8.57756615e-01 4.09894705e-01 7.14233637e-01 -1.03193724e+00
-3.94067168e-01 8.22519243e-01 3.44589837e-02 -1.04606628e-01
8.09730232e-01 2.74164289e-01 1.05843437e+00 -1.02849567e+00
8.57370794e-01 1.68948209e+00 5.41915059e-01 1.15083313e+00
-1.28898096e+00 -1.77923769e-01 1.50782513e-02 -1.16237797e-01
-5.15958846e-01 -7.18077600e-01 6.73967183e-01 -4.11972046e-01
1.46974146e+00 1.96458429e-01 1.34477720e-01 1.56244862e+00
6.12746999e-02 1.33700728e+00 1.19141185e+00 -5.23996949e-01
-2.54888628e-02 2.11978227e-01 2.80137658e-01 5.61542332e-01
-5.38883746e-01 -4.16898966e-01 -8.41583252e-01 -2.56782323e-01
2.40545586e-01 -6.75805986e-01 -1.86761737e-01 -2.67666746e-02
-1.17701209e+00 1.11048806e+00 -1.24615796e-01 1.56195015e-01
-3.57822508e-01 -3.74538124e-01 9.66993034e-01 5.51047683e-01
9.25803542e-01 7.93875933e-01 -5.42674243e-01 -6.36408269e-01
-3.00482005e-01 5.38253188e-01 1.50406349e+00 8.32158923e-01
1.71305001e-01 1.35673126e-02 -8.88339758e-01 1.40610695e+00
2.44556367e-01 2.20559224e-01 5.86209059e-01 -1.01490867e+00
8.88228834e-01 5.64976871e-01 2.11744264e-01 -2.54827529e-01
-5.86056232e-01 2.23025143e-01 -7.36649990e-01 -2.16991067e-01
7.05120146e-01 -6.66830540e-01 -3.43957931e-01 1.72953641e+00
4.50723380e-01 -7.04757631e-01 5.36922455e-01 7.34670579e-01
1.12198794e+00 7.62385905e-01 5.02910793e-01 -2.50514388e-01
1.44528115e+00 -1.51572216e+00 -1.22386849e+00 -3.66054773e-01
1.06220460e+00 -9.24056709e-01 1.35177457e+00 3.36508214e-01
-1.24925852e+00 -6.05996132e-01 -4.82019901e-01 -3.84682566e-01
-3.06563992e-02 2.37517372e-01 4.44116056e-01 4.61269408e-01
-9.89021838e-01 2.05132782e-01 -4.48055148e-01 -4.43172753e-01
-7.99979717e-02 1.16674509e-02 -4.30080369e-02 2.81536728e-01
-1.36967587e+00 1.45653653e+00 3.28951538e-01 -1.62585139e-01
-1.15717733e+00 -5.15950739e-01 -8.53432655e-01 -2.37336963e-01
4.25376892e-01 -6.95304871e-01 2.35100508e+00 -5.19916773e-01
-2.21312380e+00 1.01964629e+00 -1.13967143e-01 -6.90430284e-01
7.65010893e-01 -6.28232062e-01 1.91522185e-02 -3.41265574e-02
8.51266906e-02 9.67773676e-01 3.12518537e-01 -9.73349452e-01
-7.68711507e-01 -9.00576934e-02 3.82436365e-01 8.86180282e-01
-3.51558536e-01 3.44171315e-01 -4.30871993e-02 -3.98427427e-01
-5.25668502e-01 -9.51415479e-01 -1.39759198e-01 -4.89033729e-01
-6.33649766e-01 -1.01075995e+00 7.58767188e-01 -7.73197472e-01
1.10578728e+00 -1.65100133e+00 4.45339590e-01 -8.78422201e-01
2.27266103e-01 4.17541146e-01 7.76104303e-03 7.44002223e-01
4.63897943e-01 -1.46434173e-01 -4.23863567e-02 -9.38116133e-01
2.71062046e-01 1.75554574e-01 -3.39343488e-01 -2.65952319e-01
2.43405923e-01 8.88066053e-01 -9.59181786e-01 -4.73248303e-01
2.75005490e-01 -3.23046558e-02 -3.34695399e-01 9.67736721e-01
-8.28503609e-01 7.87892699e-01 -3.62943769e-01 9.49758440e-02
1.22628540e-01 -2.46254936e-01 2.86457181e-01 9.45040286e-02
1.48163764e-02 1.06631768e+00 -3.84890765e-01 2.03311229e+00
-7.57119954e-01 5.66923738e-01 7.61601999e-02 -5.22991478e-01
9.34215069e-01 8.46203685e-01 6.98509887e-02 -6.79995000e-01
8.11023191e-02 9.70375985e-02 2.39864364e-02 -5.10594726e-01
1.03184962e+00 -1.44711882e-01 -5.42860448e-01 9.36548591e-01
5.12147069e-01 -4.34648722e-01 2.37589642e-01 5.19739032e-01
7.91530669e-01 1.49978504e-01 4.54294711e-01 -1.64451405e-01
6.14073634e-01 4.29242194e-01 1.79543868e-01 6.75228953e-01
-1.60768062e-01 2.07776442e-01 5.85055888e-01 -3.58876824e-01
-9.10016596e-01 -5.49932361e-01 2.31278405e-01 1.85832810e+00
-2.59520203e-01 -3.88997406e-01 -9.54576433e-01 -1.15319133e+00
-3.92549723e-01 1.06484306e+00 -3.59678566e-01 1.86369792e-02
-6.33853912e-01 -6.42439485e-01 9.63341296e-01 3.15032035e-01
8.52769911e-01 -1.35847676e+00 -4.37994361e-01 6.17182255e-01
-6.87726676e-01 -1.46158910e+00 -5.75716376e-01 4.76950705e-02
-6.29476249e-01 -7.93723881e-01 -9.10223186e-01 -7.93941200e-01
1.27335146e-01 1.16094179e-01 1.58754492e+00 -2.14523330e-01
2.38194361e-01 4.61865515e-01 -5.11376143e-01 -3.85844648e-01
-1.26421726e+00 3.50728989e-01 -1.11258820e-01 -2.89631039e-01
4.19493616e-01 2.28976548e-01 -1.97661027e-01 4.67058152e-01
-3.81633520e-01 5.47920644e-01 1.13492578e-01 1.15337515e+00
-1.60225570e-01 -6.74268603e-01 1.00889242e+00 -1.09600258e+00
1.50876760e+00 -2.71418929e-01 -1.50317684e-01 4.27533895e-01
-4.84262049e-01 9.31668654e-02 5.80305040e-01 -3.07761401e-01
-1.89304364e+00 -4.82811145e-02 -2.75301605e-01 1.27490342e-01
-2.66682267e-01 4.76697564e-01 2.38772333e-02 5.33247113e-01
8.66593421e-01 1.86032221e-01 8.41914564e-02 -6.95663929e-01
4.83267188e-01 1.02995896e+00 3.47726107e-01 -8.88708532e-01
9.60161164e-02 -7.54446834e-02 -6.62355542e-01 -8.01981091e-01
-1.27823925e+00 -4.38888639e-01 -7.72863865e-01 -4.07513559e-01
1.00898933e+00 -1.11166990e+00 -5.90390921e-01 7.32808471e-01
-1.71956122e+00 -9.54873383e-01 -1.24011070e-01 1.19833559e-01
-5.12418926e-01 3.50085437e-01 -1.05776846e+00 -1.00662243e+00
-9.80243862e-01 -1.07385731e+00 1.04291630e+00 1.63866609e-01
-7.88933575e-01 -1.45501542e+00 4.91736829e-01 8.27505946e-01
3.32689017e-01 -3.22638929e-01 8.85529339e-01 -1.41928768e+00
1.61039054e-01 1.66851997e-01 5.74431010e-03 2.98213601e-01
9.67024714e-02 -3.99533629e-01 -1.20129049e+00 -1.70489565e-01
1.35229886e-01 -1.29389203e+00 3.19655508e-01 -5.89101948e-02
3.39553982e-01 -6.12797737e-01 3.93302888e-02 -2.97604561e-01
4.93217170e-01 1.56371519e-01 2.26609826e-01 3.53765003e-02
5.60186982e-01 1.25880730e+00 1.03094387e+00 4.13664669e-01
9.14092004e-01 1.10390127e+00 -1.31552443e-01 -4.02668826e-02
-4.00023490e-01 -3.36979061e-01 7.12857485e-01 9.63929832e-01
2.76957840e-01 -6.37337923e-01 -1.01439667e+00 5.00410795e-01
-2.11550498e+00 -6.27214491e-01 -3.37271929e-01 1.82134759e+00
1.52555490e+00 2.58602165e-02 5.51755130e-01 -5.19546211e-01
5.64450562e-01 2.44263068e-01 -3.38759065e-01 -7.71962643e-01
-2.52526831e-02 -9.32005979e-03 -4.46987420e-01 7.31861949e-01
-9.22144055e-01 1.36278903e+00 6.19679976e+00 6.68005407e-01
-7.24522173e-01 5.50734401e-01 7.43027925e-01 1.53079852e-01
1.50563354e-02 -1.54083937e-01 -1.00658262e+00 2.07763717e-01
1.07883096e+00 -3.63940984e-01 -7.97180459e-02 9.09787655e-01
1.45733625e-01 -1.29848361e-01 -1.32726431e+00 4.54948515e-01
1.24936454e-01 -9.81187940e-01 1.57899588e-01 -1.77099228e-01
6.25306726e-01 -1.27354547e-01 -2.49033436e-01 8.81710589e-01
1.01342273e+00 -8.34039629e-01 4.87548620e-01 -1.69455800e-02
5.96742034e-01 -2.60627806e-01 8.78787279e-01 5.76286972e-01
-7.34317839e-01 2.86556035e-01 -1.90667450e-01 -1.59845486e-01
4.69135255e-01 -4.85527962e-02 -1.71504521e+00 3.74222338e-01
1.28470019e-01 3.59044999e-01 -3.54632914e-01 4.63717371e-01
-4.96315092e-01 4.78806436e-01 4.98024635e-02 -4.16456014e-01
4.24457282e-01 7.64033869e-02 5.29146254e-01 1.54994392e+00
-2.67796218e-01 8.02141428e-02 5.48734784e-01 6.41807139e-01
-3.55781972e-01 2.64044464e-01 -6.37323678e-01 8.02241638e-02
5.22599757e-01 1.41073811e+00 -1.33158788e-01 -6.02488339e-01
-3.10095191e-01 1.16077363e+00 4.95975107e-01 6.20513558e-02
-4.62486982e-01 1.04233297e-02 3.31269026e-01 -3.68479043e-01
-2.17007175e-01 -2.09664464e-01 -1.98426142e-01 -1.12718344e+00
-2.94542849e-01 -1.19336581e+00 6.25173271e-01 -6.51071966e-01
-1.38810074e+00 9.44857359e-01 6.27497630e-03 -8.97856414e-01
-9.89133477e-01 -5.03605545e-01 -6.00988686e-01 8.66142452e-01
-1.15669894e+00 -1.33181667e+00 -1.73349962e-01 5.24502277e-01
1.58508718e+00 -3.78275901e-01 1.33709264e+00 -5.71991615e-02
-6.03882849e-01 4.35726374e-01 -3.31551969e-01 3.83623213e-01
1.17621267e+00 -1.62187076e+00 5.96690416e-01 4.58684117e-01
-2.50623859e-02 1.83524460e-01 7.64905334e-01 -7.82694757e-01
-9.34665859e-01 -8.67393792e-01 1.17072892e+00 -7.60134816e-01
6.34950101e-01 -5.23832917e-01 -9.64843452e-01 6.09694719e-01
1.01514852e+00 -7.75345147e-01 7.21467137e-01 4.80976522e-01
-1.75155237e-01 5.15445292e-01 -1.05108368e+00 6.00080013e-01
7.27013409e-01 -5.74244976e-01 -1.03306866e+00 7.60136485e-01
7.24236012e-01 -7.85897374e-01 -9.77518380e-01 2.35564485e-01
2.61442721e-01 -7.17095852e-01 6.46605253e-01 -9.94540215e-01
6.56036317e-01 3.94467056e-01 2.03119472e-01 -1.58644760e+00
2.22622663e-01 -1.07950270e+00 -2.63975739e-01 1.48434770e+00
6.44668996e-01 -1.59710169e-01 3.95606577e-01 9.38450873e-01
-5.07018566e-01 -4.67007428e-01 -8.32628369e-01 -5.47751069e-01
2.65727013e-01 8.81529450e-02 8.00957978e-02 9.82657194e-01
6.15660906e-01 1.50761676e+00 -6.22593880e-01 -5.24088740e-01
3.07934344e-01 -4.48939689e-02 1.15261579e+00 -1.26115823e+00
-1.80114985e-01 -3.59322906e-01 5.65530539e-01 -1.57176757e+00
4.16779816e-01 -5.68353355e-01 4.74030405e-01 -1.57646143e+00
3.60923000e-02 -4.23972815e-01 4.96929049e-01 5.91355562e-01
-5.00757277e-01 -2.30501637e-01 2.08862185e-01 4.20174420e-01
-1.07224333e+00 9.49071169e-01 1.30384278e+00 -5.64338788e-02
-5.26264250e-01 1.36745483e-01 -4.02026355e-01 5.85463047e-01
7.37473071e-01 -3.84252191e-01 -4.88103211e-01 -6.06031299e-01
-1.82082802e-01 6.17558777e-01 -3.85985076e-02 -5.78454494e-01
2.70299375e-01 1.65871307e-02 -2.59355009e-01 -4.76278722e-01
7.38111913e-01 -1.07255116e-01 -5.38976490e-01 3.42185855e-01
-9.91585851e-01 -1.70139894e-02 2.20359653e-01 2.07728311e-01
-1.80401921e-01 -4.99684274e-01 6.39404476e-01 -2.80179650e-01
-5.84213197e-01 -2.62421042e-01 -8.05728972e-01 5.20850956e-01
7.90025115e-01 2.77444005e-01 -8.27685535e-01 -9.52202678e-01
-5.54908812e-01 5.30672967e-01 6.25271630e-03 6.34692132e-01
3.03765118e-01 -9.60927367e-01 -1.13151622e+00 -4.70653564e-01
3.13665986e-01 1.25763819e-01 2.33735144e-01 6.46850288e-01
-7.36627579e-02 8.28422308e-01 -1.74573973e-01 -5.92344224e-01
-1.54909217e+00 -5.46447895e-02 4.12863761e-01 -9.87492204e-01
-5.09984910e-01 9.96958971e-01 1.41189620e-01 -9.71938908e-01
4.28732723e-01 9.32799578e-02 -4.94541019e-01 3.40915918e-01
4.12586331e-01 2.02440977e-01 2.63350620e-03 -5.15630782e-01
9.91379321e-02 -2.96096653e-01 -6.47860706e-01 -6.51169360e-01
1.02320313e+00 -3.37576330e-01 1.10362843e-01 6.69737816e-01
6.10019147e-01 -3.28381985e-01 -1.26034999e+00 -6.53216243e-01
2.73409694e-01 -2.66878828e-02 -3.41597974e-01 -1.38079631e+00
-3.09477895e-01 1.03989673e+00 5.72793223e-02 5.99412799e-01
5.08413494e-01 -1.41441869e-02 9.16200578e-01 8.66242886e-01
3.94743621e-01 -1.31426692e+00 6.75485432e-01 1.16995943e+00
1.13763106e+00 -1.33931184e+00 -3.08470905e-01 -2.40391076e-01
-1.66167843e+00 9.85870659e-01 1.22298086e+00 3.93961310e-01
-1.28781915e-01 -1.76720005e-02 4.75360215e-01 -1.75134897e-01
-1.39549446e+00 1.66154886e-03 1.52943075e-01 4.85537678e-01
9.35358405e-01 -1.50631383e-01 -2.05379665e-01 5.99335015e-01
-2.04941988e-01 -2.55414426e-01 6.18866205e-01 9.24033523e-01
-3.85159701e-01 -1.37241101e+00 5.53784310e-04 1.50238678e-01
-4.21020150e-01 -1.75997630e-01 -1.11415505e+00 5.80374658e-01
-9.29262519e-01 1.36102343e+00 -1.03727698e-01 -2.87985325e-01
5.37255228e-01 6.73016131e-01 2.62599081e-01 -1.17426658e+00
-1.36945641e+00 -3.63633446e-02 1.14464176e+00 -1.64501518e-01
-4.22659427e-01 -4.32133257e-01 -1.14551139e+00 -2.14315549e-01
-4.69134420e-01 5.59503675e-01 3.90498489e-01 1.05041742e+00
4.25841987e-01 5.58116734e-01 6.88444614e-01 -5.80433011e-01
-1.00831246e+00 -1.83227444e+00 1.05057135e-01 5.03184676e-01
-5.80354221e-02 -4.60336447e-01 8.98368135e-02 3.58111821e-02]
|
[12.713942527770996, 8.123894691467285]
|
8672bd59-1488-4398-8522-fa3f4714d7ae
|
knowledge-distillation-transfer-sets-and
|
2210.04834
| null |
https://arxiv.org/abs/2210.04834v3
|
https://arxiv.org/pdf/2210.04834v3.pdf
|
Knowledge Distillation Transfer Sets and their Impact on Downstream NLU Tasks
|
Teacher-student knowledge distillation is a popular technique for compressing today's prevailing large language models into manageable sizes that fit low-latency downstream applications. Both the teacher and the choice of transfer set used for distillation are crucial ingredients in creating a high quality student. Yet, the generic corpora used to pretrain the teacher and the corpora associated with the downstream target domain are often significantly different, which raises a natural question: should the student be distilled over the generic corpora, so as to learn from high-quality teacher predictions, or over the downstream task corpora to align with finetuning? Our study investigates this trade-off using Domain Classification (DC) and Intent Classification/Named Entity Recognition (ICNER) as downstream tasks. We distill several multilingual students from a larger multilingual LM with varying proportions of generic and task-specific datasets, and report their performance after finetuning on DC and ICNER. We observe significant improvements across tasks and test sets when only task-specific corpora is used. We also report on how the impact of adding task-specific data to the transfer set correlates with the similarity between generic and task-specific data. Our results clearly indicate that, while distillation from a generic LM benefits downstream tasks, students learn better using target domain data even if it comes at the price of noisier teacher predictions. In other words, target domain data still trumps teacher knowledge.
|
['Pan Wei', 'Gokmen Oz', 'Turan Gojayev', 'Thomas Gueudre', 'Lizhen Tan', 'Charith Peris']
|
2022-10-10
| null | null | null | null |
['intent-classification']
|
['natural-language-processing']
|
[ 1.71478629e-01 4.34317179e-02 -3.36584657e-01 -4.90439028e-01
-1.05038369e+00 -1.08491707e+00 4.22030836e-01 2.27490827e-01
-9.13309038e-01 9.07384753e-01 3.21458697e-01 -6.91170752e-01
4.46873754e-02 -5.35638750e-01 -6.46249413e-01 -4.48854268e-01
3.76840413e-01 8.74057710e-01 4.14091527e-01 -3.56287718e-01
-2.19681904e-01 3.67564231e-01 -1.40877187e+00 5.60458422e-01
1.30857623e+00 5.82076669e-01 6.25202298e-01 7.04057038e-01
-1.99396536e-01 7.62347460e-01 -7.94656515e-01 -6.33081555e-01
2.67768055e-01 -1.83129996e-01 -1.05671823e+00 -2.60793626e-01
8.33490849e-01 -2.21736178e-01 -3.17162760e-02 6.27090275e-01
5.98648190e-01 3.50273728e-01 7.01182961e-01 -9.57915187e-01
-5.73495567e-01 1.01124191e+00 -3.03785235e-01 3.49229753e-01
4.45030890e-02 2.35255703e-01 9.51073647e-01 -7.16848254e-01
6.05012238e-01 1.07315218e+00 5.67749202e-01 5.41849732e-01
-1.58419216e+00 -9.39401567e-01 7.57436752e-02 4.09759842e-02
-1.24614155e+00 -6.43684864e-01 2.30508432e-01 -4.98495758e-01
1.11047590e+00 2.32056249e-02 3.20113450e-01 1.13516176e+00
1.72853693e-01 9.38039839e-01 1.16160953e+00 -5.05019546e-01
-1.41655967e-01 5.96734047e-01 1.81265786e-01 2.16442838e-01
1.67084619e-01 2.97517851e-02 -6.23238087e-01 -3.79398353e-02
3.15752983e-01 -4.69198942e-01 -2.33175203e-01 -1.87837765e-01
-1.09584415e+00 6.80230200e-01 1.00112543e-01 3.84634167e-01
3.07261129e-03 -2.39793986e-01 4.88009125e-01 8.49301755e-01
5.10932982e-01 7.32674599e-01 -1.21926534e+00 -4.52324152e-01
-9.96693313e-01 1.58777773e-01 1.11618078e+00 1.23759913e+00
6.82659149e-01 7.62778148e-02 -1.78263620e-01 1.04808390e+00
-7.15104863e-02 5.07549644e-01 9.81669724e-01 -5.83378017e-01
8.77175987e-01 4.84829694e-01 -4.74603951e-01 -1.26850411e-01
-3.60491313e-02 -4.36493903e-01 -3.70647848e-01 -1.40015841e-01
8.20174038e-01 -6.28246844e-01 -9.14952993e-01 2.11030054e+00
3.49833339e-01 3.61324072e-01 3.31394821e-01 5.43278337e-01
7.79510677e-01 7.33708918e-01 5.23143888e-01 -6.06813803e-02
1.35806179e+00 -9.17710066e-01 -1.80890113e-01 -5.70295155e-01
1.32098615e+00 -1.07353365e+00 1.20059311e+00 5.03363550e-01
-1.06704664e+00 -7.55352318e-01 -7.84781694e-01 -5.31195462e-01
-3.02346408e-01 -9.19536203e-02 1.99643984e-01 3.35037768e-01
-1.12161267e+00 4.46567953e-01 -3.47810566e-01 -2.12722003e-01
-2.58797891e-02 4.37831253e-01 -4.12695855e-01 -2.18745723e-01
-1.21067166e+00 1.13115859e+00 6.55933797e-01 -7.04254866e-01
-9.43836451e-01 -1.15928507e+00 -6.62632644e-01 2.99973190e-01
1.67325005e-01 -4.76535052e-01 1.76354384e+00 -1.14560890e+00
-1.43994987e+00 9.14168715e-01 2.16530487e-01 -3.31274062e-01
3.90444636e-01 -2.68409818e-01 -1.37043908e-01 -3.56510192e-01
5.23373894e-02 7.98563778e-01 6.13158584e-01 -7.35084414e-01
-1.07348490e+00 -2.43016914e-01 -8.40194821e-02 6.50900126e-01
-5.47749698e-01 -1.41764909e-01 -2.24352777e-01 -6.25974953e-01
-3.57673347e-01 -9.59045649e-01 1.65080056e-02 -6.84841156e-01
-3.81587520e-02 -6.90532684e-01 7.68115699e-01 -5.55335462e-01
1.33041275e+00 -2.16386175e+00 1.64457098e-01 -3.90601531e-02
1.25021562e-01 4.46331084e-01 -6.17832661e-01 3.18088561e-01
-1.58555180e-01 2.49861442e-02 2.17207968e-01 -2.33649388e-01
-2.43382335e-01 2.79268831e-01 -3.77223969e-01 6.28099963e-02
1.27836689e-01 7.11764276e-01 -9.37846124e-01 -4.77729708e-01
-2.19225213e-01 1.91112190e-01 -8.50900829e-01 4.63416934e-01
-4.08583701e-01 5.24174750e-01 -4.46705341e-01 7.80616235e-03
2.63092995e-01 4.77647688e-03 3.33789825e-01 8.68490785e-02
-8.83262902e-02 9.54777360e-01 -1.05277824e+00 1.77136421e+00
-8.37043405e-01 6.77808404e-01 2.07738087e-01 -9.43405330e-01
8.28906715e-01 5.10673523e-01 2.05529258e-01 -8.15545321e-01
-8.42964575e-02 4.29984033e-01 5.29352486e-01 -2.83549607e-01
6.13258958e-01 -4.29581583e-01 -2.50125140e-01 7.03666925e-01
4.21483606e-01 -4.36550319e-01 1.55435234e-01 2.37794682e-01
1.06822467e+00 2.12044373e-01 2.67606854e-01 -4.63955790e-01
3.15364271e-01 3.12153012e-01 5.42512774e-01 4.25127625e-01
-7.26215616e-02 1.23699225e-01 3.27811480e-01 -1.24876454e-01
-9.17236567e-01 -1.02844846e+00 -2.07896665e-01 1.91338122e+00
-3.12625021e-01 -4.40524191e-01 -7.01806188e-01 -9.73447144e-01
2.44562164e-01 9.09559608e-01 -2.17380002e-02 -2.59965956e-01
-8.75663042e-01 -1.84791461e-01 7.64372170e-01 5.46806395e-01
1.45541370e-01 -9.47159529e-01 -2.47942492e-01 3.64763409e-01
-8.64274278e-02 -1.15203381e+00 -7.41397619e-01 6.35172307e-01
-1.11637044e+00 -7.36693084e-01 -4.25990283e-01 -9.15533900e-01
5.17854452e-01 1.64162636e-01 1.63570487e+00 -1.26644047e-02
1.16592169e-01 1.07409850e-01 -6.20825887e-02 -4.13458854e-01
-6.58835709e-01 6.83597207e-01 2.39495456e-01 -5.98026156e-01
6.70704484e-01 -6.54967546e-01 -5.20527624e-02 8.02588016e-02
-8.66732895e-01 1.86734945e-01 8.42370212e-01 8.50796819e-01
3.06584805e-01 3.31687704e-02 5.84285796e-01 -1.24157572e+00
6.65679216e-01 -6.20565891e-01 -6.28652036e-01 2.56231755e-01
-6.45780981e-01 4.88467216e-01 9.63330746e-01 -8.67678344e-01
-1.17786241e+00 -6.76744878e-02 -1.04741581e-01 -3.87673855e-01
-4.24917847e-01 4.46668148e-01 -1.54606804e-01 3.57551455e-01
8.51886690e-01 -5.23018874e-02 -3.66778851e-01 -7.72089422e-01
5.43362439e-01 8.51373315e-01 4.09815580e-01 -1.24558008e+00
6.04331255e-01 -3.07567596e-01 -6.33763492e-01 -7.26709187e-01
-1.04119241e+00 -4.61223722e-01 -8.46494496e-01 3.31664771e-01
6.59963727e-01 -1.16509485e+00 -5.11130095e-01 2.39908099e-01
-1.15889120e+00 -1.06505358e+00 -4.60113674e-01 5.04433572e-01
-3.23862016e-01 -2.38149941e-01 -8.11245918e-01 -2.48981968e-01
-1.02614939e-01 -1.57065237e+00 7.70732820e-01 1.68088675e-01
-4.16787267e-01 -1.27060115e+00 2.11159304e-01 5.82173049e-01
3.91072839e-01 -5.82763553e-01 1.29755521e+00 -1.36491311e+00
-3.62172574e-01 2.61726707e-01 -8.53618085e-02 5.04664958e-01
-3.43795307e-02 -3.06733459e-01 -1.16566098e+00 -4.22112018e-01
-1.19141407e-01 -7.05920756e-01 4.65284258e-01 -3.08684967e-02
8.01414907e-01 -4.79194880e-01 -2.61272192e-01 5.83248973e-01
1.29038513e+00 1.35698408e-01 6.19379319e-02 2.47802094e-01
8.01123798e-01 8.15673411e-01 4.85798568e-01 -1.16980664e-01
5.81819594e-01 7.38854587e-01 -3.88067603e-01 1.96254194e-01
-3.96707386e-01 -5.26987016e-01 7.47440994e-01 1.37033594e+00
4.30614918e-01 -1.95307121e-01 -1.40522242e+00 6.82491660e-01
-1.31235039e+00 -5.10003209e-01 1.38178363e-01 2.32141376e+00
1.71065366e+00 1.16724379e-01 1.63403433e-02 -3.08133096e-01
3.43913347e-01 -2.55159140e-01 -3.16885978e-01 -6.75075769e-01
3.76242399e-01 5.97376943e-01 5.79774976e-01 6.72903895e-01
-6.84514284e-01 1.30281270e+00 5.66613102e+00 1.14655852e+00
-1.47301364e+00 3.59192461e-01 5.62273502e-01 -1.24996044e-01
-3.72881562e-01 -4.68713418e-03 -1.52688503e+00 4.54202503e-01
1.65373480e+00 -3.83766979e-01 1.61259875e-01 9.51489925e-01
3.30350623e-02 1.47293610e-02 -1.54096663e+00 4.68363345e-01
-3.20421249e-01 -9.16203916e-01 3.54899354e-02 1.78449720e-01
9.49092448e-01 3.50186616e-01 -5.62444367e-02 1.00116456e+00
9.41164732e-01 -1.04750764e+00 6.01583779e-01 -9.39573273e-02
8.94094288e-01 -8.64298701e-01 5.82019508e-01 8.38762164e-01
-9.28571105e-01 6.35819286e-02 -2.24845245e-01 -1.69286340e-01
-2.60160238e-01 2.98497915e-01 -1.50836945e+00 8.35070238e-02
4.48003441e-01 4.26713347e-01 -6.63468003e-01 6.37697935e-01
-3.33123952e-01 9.02657986e-01 -3.47222924e-01 2.48188928e-01
3.02975863e-01 -6.49451762e-02 3.22036803e-01 1.39423525e+00
2.28985846e-01 9.19362232e-02 5.49708307e-01 4.58490700e-01
-3.26168448e-01 2.76291370e-01 -7.52320170e-01 -1.69986412e-01
8.05642426e-01 1.12846231e+00 -2.93494791e-01 -4.97620165e-01
-5.93572855e-01 7.87034810e-01 7.62113869e-01 2.38276228e-01
-4.37749624e-01 -2.18391448e-01 9.29613173e-01 2.65999138e-01
5.42521216e-02 -2.95198917e-01 -3.19175303e-01 -1.24758863e+00
-3.87856781e-01 -1.28330708e+00 6.17919624e-01 -4.54657465e-01
-1.34351099e+00 3.95292550e-01 1.89690351e-01 -9.63344455e-01
-4.77070332e-01 -5.23940802e-01 -5.37475646e-01 1.12858248e+00
-1.59666634e+00 -9.63755488e-01 1.32685646e-01 6.40433729e-01
8.78314376e-01 -7.77458549e-02 7.90724635e-01 4.39025611e-01
-4.47131187e-01 9.00118291e-01 1.41144097e-01 1.77276298e-01
1.37487376e+00 -1.34371161e+00 2.75675476e-01 7.10148990e-01
2.99914330e-01 6.13126338e-01 4.40861285e-01 -6.94964290e-01
-1.24021029e+00 -9.92920160e-01 1.34566557e+00 -8.44119072e-01
8.42596471e-01 -3.49142194e-01 -1.03031731e+00 1.00887275e+00
2.25080177e-01 -2.62487411e-01 8.76122117e-01 4.54386264e-01
-3.75732899e-01 -1.21366233e-01 -8.51122141e-01 5.41300297e-01
7.80412376e-01 -5.45777977e-01 -9.18557584e-01 2.01114774e-01
1.01616549e+00 -5.75339317e-01 -1.11687243e+00 2.05246359e-01
2.15221018e-01 -5.37095487e-01 7.77769029e-01 -8.62813056e-01
4.65542495e-01 2.04755679e-01 1.14065140e-01 -1.67571497e+00
-1.55884311e-01 -3.05794477e-01 2.02634007e-01 1.48732090e+00
6.78908169e-01 -4.12308484e-01 7.85948634e-01 5.83642423e-01
-2.68431723e-01 -5.58334589e-01 -6.99021280e-01 -8.56698573e-01
8.22791040e-01 -5.84142745e-01 4.89177465e-01 1.28997290e+00
-5.31071052e-02 9.56722498e-01 2.21107200e-01 -6.34728596e-02
2.48624310e-01 2.23297793e-02 8.93954873e-01 -1.29830229e+00
-4.76260066e-01 -3.43629926e-01 7.57116005e-02 -1.25423777e+00
3.95040452e-01 -1.32737589e+00 1.95298001e-01 -9.85150099e-01
-6.11509159e-02 -1.29613876e+00 -1.34304568e-01 7.72329509e-01
-3.19115698e-01 -6.08817227e-02 3.99473876e-01 1.60772115e-01
-2.68725425e-01 1.66616678e-01 1.37026775e+00 9.29863974e-02
-4.44842815e-01 4.31308411e-02 -7.21586883e-01 5.56109488e-01
5.52264154e-01 -6.39682949e-01 -7.84517050e-01 -8.61249208e-01
-8.57970957e-03 5.05326129e-02 -3.53348404e-01 -7.14783370e-01
4.01356280e-01 -3.26684713e-01 2.43225515e-01 -1.02612250e-01
7.92975426e-02 -8.51537466e-01 -3.55133951e-01 2.25938767e-01
-5.13487518e-01 1.74970269e-01 6.11376762e-01 7.19986632e-02
-1.59494713e-01 -4.45849568e-01 1.04987347e+00 -3.03412843e-02
-8.25541854e-01 1.19288750e-01 -3.47616643e-01 7.03441203e-01
8.84007633e-01 -2.18125246e-03 -3.80453140e-01 -1.72088519e-01
-5.78467429e-01 2.62086272e-01 3.42000663e-01 5.71911454e-01
1.08276509e-01 -1.14413130e+00 -8.95590305e-01 5.24708867e-01
-7.94112682e-02 3.81419450e-01 -1.68692350e-01 8.46762598e-01
-2.16141358e-01 7.14684844e-01 -1.07041687e-01 -5.78447700e-01
-1.36711907e+00 2.53699750e-01 -4.14142273e-02 -7.00615227e-01
-2.87308484e-01 1.18912339e+00 5.76460660e-01 -8.74908984e-01
2.27954507e-01 -5.63232005e-01 -2.62785051e-02 3.83842170e-01
2.82050848e-01 1.89321741e-01 2.26928025e-01 -3.93881589e-01
-6.23662909e-03 1.12077601e-01 -3.49204123e-01 -1.00863166e-01
1.25437689e+00 -5.31577738e-03 3.15205038e-01 3.07250828e-01
1.09545159e+00 4.25951958e-01 -1.07513130e+00 -4.93180126e-01
4.29991215e-01 -4.07874584e-02 -1.60523765e-02 -8.80692244e-01
-8.92550468e-01 9.61658955e-01 1.56958088e-01 -7.57540911e-02
9.53995228e-01 -1.90839302e-02 8.88255179e-01 2.95295984e-01
4.01209116e-01 -9.94367421e-01 -1.52046708e-02 1.03275788e+00
4.33878869e-01 -1.08632970e+00 -1.46216795e-01 -1.70630276e-01
-6.47545099e-01 9.16273296e-01 1.21932840e+00 4.64750193e-02
3.17295998e-01 6.40356660e-01 2.08497092e-01 2.09485203e-01
-1.26735055e+00 -1.06202886e-01 4.31800097e-01 3.84503365e-01
9.95395303e-01 1.04916029e-01 -1.32432535e-01 7.01943934e-01
-7.03447580e-01 -1.73189744e-01 3.46477836e-01 9.31678236e-01
-5.76968551e-01 -1.57961142e+00 -3.24634701e-01 4.13206369e-01
-4.67032582e-01 -4.71150219e-01 -2.70911992e-01 8.31371725e-01
4.04707342e-01 7.02898383e-01 1.52836248e-01 -2.70854503e-01
3.06613773e-01 6.30599201e-01 4.43669081e-01 -1.42631721e+00
-1.11251342e+00 -1.21403284e-01 2.50457019e-01 -3.31972450e-01
2.73051083e-01 -2.96793282e-01 -1.20478535e+00 -3.44233960e-01
-4.28232938e-01 5.72688222e-01 5.32143354e-01 1.14104831e+00
2.84193307e-01 3.52410495e-01 8.58634859e-02 -4.83402342e-01
-9.09134507e-01 -1.09189188e+00 -2.42965668e-01 2.57872045e-01
2.60248214e-01 -3.88656199e-01 -2.76095957e-01 3.26487243e-01]
|
[10.800707817077637, 8.474559783935547]
|
3ec4b88a-6e55-4ef2-b7f7-52bf087a6b0d
|
epic-fusion-audio-visual-temporal-binding-for
|
1908.08498
| null |
https://arxiv.org/abs/1908.08498v1
|
https://arxiv.org/pdf/1908.08498v1.pdf
|
EPIC-Fusion: Audio-Visual Temporal Binding for Egocentric Action Recognition
|
We focus on multi-modal fusion for egocentric action recognition, and propose a novel architecture for multi-modal temporal-binding, i.e. the combination of modalities within a range of temporal offsets. We train the architecture with three modalities -- RGB, Flow and Audio -- and combine them with mid-level fusion alongside sparse temporal sampling of fused representations. In contrast with previous works, modalities are fused before temporal aggregation, with shared modality and fusion weights over time. Our proposed architecture is trained end-to-end, outperforming individual modalities as well as late-fusion of modalities. We demonstrate the importance of audio in egocentric vision, on per-class basis, for identifying actions as well as interacting objects. Our method achieves state of the art results on both the seen and unseen test sets of the largest egocentric dataset: EPIC-Kitchens, on all metrics using the public leaderboard.
|
['Evangelos Kazakos', 'Arsha Nagrani', 'Andrew Zisserman', 'Dima Damen']
|
2019-08-22
|
epic-fusion-audio-visual-temporal-binding-for-1
|
http://openaccess.thecvf.com/content_ICCV_2019/html/Kazakos_EPIC-Fusion_Audio-Visual_Temporal_Binding_for_Egocentric_Action_Recognition_ICCV_2019_paper.html
|
http://openaccess.thecvf.com/content_ICCV_2019/papers/Kazakos_EPIC-Fusion_Audio-Visual_Temporal_Binding_for_Egocentric_Action_Recognition_ICCV_2019_paper.pdf
|
iccv-2019-10
|
['egocentric-activity-recognition']
|
['computer-vision']
|
[ 3.30083698e-01 -3.58239055e-01 8.13393742e-02 -2.71011263e-01
-1.04932535e+00 -5.10433614e-01 9.38935220e-01 -4.25583452e-01
-4.76975799e-01 4.97741818e-01 1.09350967e+00 6.67419016e-01
-2.30128467e-01 -2.97841221e-01 -5.91225326e-01 -6.78851068e-01
-2.04411492e-01 1.05040953e-01 1.40450612e-01 -1.66718394e-01
1.90739706e-02 1.64986297e-01 -2.10932779e+00 1.01787639e+00
2.47501045e-01 1.16750443e+00 -3.40540797e-01 1.11964095e+00
5.81049860e-01 1.37866437e+00 -2.39333004e-01 -2.34720990e-01
1.78479016e-01 -3.49304110e-01 -9.66755927e-01 2.08774149e-01
1.04183424e+00 -7.08437026e-01 -8.66797328e-01 3.90862912e-01
7.57344007e-01 6.19437993e-01 6.04095697e-01 -1.32119548e+00
-3.18926156e-01 3.74695003e-01 -4.78572398e-01 3.45969200e-01
8.70320201e-01 5.36589324e-01 9.73242342e-01 -8.58945847e-01
7.09089518e-01 1.51989019e+00 4.77615774e-01 4.50726449e-01
-1.26385546e+00 -4.63593692e-01 3.63401711e-01 7.50760972e-01
-9.41040099e-01 -9.65925574e-01 5.78794062e-01 -6.67221129e-01
1.54224229e+00 -5.06844148e-02 5.86325586e-01 1.86094356e+00
4.24150638e-02 1.13654423e+00 8.85835409e-01 -1.53144822e-01
8.37081447e-02 -7.17663348e-01 -1.49904579e-01 4.50923562e-01
-5.10947526e-01 2.71392792e-01 -1.46808326e+00 8.27496722e-02
7.46951997e-01 1.92866996e-01 -1.70992047e-01 -4.51860011e-01
-1.89136493e+00 1.83202401e-01 2.79738665e-01 1.86433464e-01
-8.05310547e-01 8.85408521e-01 8.56951058e-01 2.66637981e-01
4.70795751e-01 5.60893603e-02 -5.19659460e-01 -8.56021821e-01
-5.66998661e-01 2.10759237e-01 1.40173987e-01 5.77975452e-01
4.04526979e-01 3.81832831e-02 -5.98716974e-01 6.42042041e-01
9.31790695e-02 6.04600012e-01 3.67457658e-01 -1.53068364e+00
6.04375958e-01 3.82673413e-01 1.07937396e-01 -4.13386792e-01
-6.15908980e-01 -3.65843773e-02 -5.16390502e-01 3.03918332e-01
6.32370293e-01 -1.48634939e-02 -9.71711040e-01 2.13717365e+00
4.55642134e-01 9.10609365e-01 1.52623162e-01 1.04863846e+00
8.75670016e-01 6.08401969e-02 5.06651461e-01 8.85639787e-02
1.41959572e+00 -9.52753782e-01 -7.47917116e-01 1.14309065e-01
5.92757165e-01 -4.83589590e-01 8.00024927e-01 5.62563241e-01
-1.23083901e+00 -7.49084115e-01 -5.27713776e-01 -1.69693276e-01
-3.08639497e-01 1.47249371e-01 8.59874904e-01 2.90752590e-01
-1.04953671e+00 5.93145251e-01 -1.17735386e+00 -6.98212981e-01
4.14374232e-01 2.00096875e-01 -1.00381875e+00 -2.46452470e-03
-1.01857269e+00 7.62100041e-01 3.14380825e-01 -6.61123097e-02
-1.49216223e+00 -7.02987671e-01 -1.00384080e+00 -1.18593536e-01
9.41925868e-02 -1.20697093e+00 1.41380870e+00 -1.01245630e+00
-1.49204504e+00 7.26078331e-01 -2.59367287e-01 -5.77895582e-01
4.19424742e-01 -7.49391675e-01 -6.69253290e-01 7.02915788e-01
-9.66439024e-03 1.02052844e+00 1.21126759e+00 -7.85876691e-01
-9.37185884e-01 -5.39774716e-01 5.21043837e-01 4.87132251e-01
-3.74405742e-01 -4.37598079e-02 -2.16910988e-01 -4.99369025e-01
5.24385609e-02 -8.49369645e-01 1.41827896e-01 -4.33040857e-02
-1.81534723e-03 -2.94941276e-01 9.91765261e-01 -5.10927737e-01
8.17007244e-01 -2.36349821e+00 5.73352695e-01 -3.74960542e-01
2.19410270e-01 -3.07847291e-01 -3.72081757e-01 4.06920642e-01
-5.35404205e-01 -5.43827891e-01 2.36907855e-01 -9.01933968e-01
1.98018149e-01 5.07121757e-02 -3.72811317e-01 7.54980922e-01
2.53838092e-01 9.75975871e-01 -1.22291160e+00 -2.98422754e-01
8.26077700e-01 8.17192435e-01 -6.01791918e-01 1.23954199e-01
2.67408416e-02 8.82349491e-01 -1.32098407e-01 1.02897370e+00
2.49348655e-01 -3.28994021e-02 -3.38653535e-01 -5.48978567e-01
1.70225188e-01 2.16373071e-01 -1.11571264e+00 2.63291454e+00
-5.53655148e-01 5.22695005e-01 -2.04124734e-01 -5.43240607e-01
2.23019063e-01 8.47472310e-01 1.01261020e+00 -8.45852196e-01
2.39547584e-02 -1.62969440e-01 -3.59992653e-01 -7.37682700e-01
5.91142654e-01 2.14658141e-01 -3.49615782e-01 5.35750329e-01
8.57089520e-01 2.94119149e-01 3.39112222e-01 3.76165479e-01
1.48848176e+00 9.76937592e-01 6.57429844e-02 3.77608091e-01
5.01340806e-01 -2.69283384e-01 3.39170039e-01 6.49216175e-01
-5.81915677e-01 7.93354630e-01 3.16141486e-01 -1.95664808e-01
-5.93618512e-01 -1.30622375e+00 2.17017785e-01 1.76935613e+00
8.93440992e-02 -4.92827147e-01 -3.73448521e-01 -8.77424657e-01
5.77859171e-02 5.22359788e-01 -9.72126484e-01 -2.69413948e-01
-4.40107375e-01 -1.14381954e-01 7.21551955e-01 9.23499346e-01
4.94521350e-01 -1.17224395e+00 -9.37527061e-01 1.27131909e-01
-5.29387057e-01 -1.33158910e+00 -2.26245105e-01 9.98203754e-02
-7.67614245e-01 -1.26023126e+00 -6.62001491e-01 2.63012461e-02
6.23304397e-02 4.88438845e-01 1.23544240e+00 -8.15612793e-01
-2.03038394e-01 1.49559247e+00 -5.71833372e-01 -4.42869961e-03
3.33711386e-01 -3.12652379e-01 2.25194320e-01 4.41355914e-01
2.97689408e-01 -9.73559797e-01 -8.54119420e-01 2.08576918e-01
-5.42713761e-01 -2.91365236e-02 3.83500338e-01 6.77297533e-01
3.69619071e-01 -5.44652641e-01 4.65479076e-01 -1.12332508e-01
1.65654765e-03 -4.91236478e-01 1.44722667e-02 6.99215662e-03
3.51548880e-01 -1.40257329e-01 1.58339083e-01 -3.61477792e-01
-1.27348113e+00 3.69238645e-01 3.74814034e-01 -1.08145344e+00
-6.33499503e-01 1.00397676e-01 2.51463149e-02 6.31839409e-02
8.35687935e-01 1.71870701e-02 -2.92334646e-01 -4.64566052e-01
9.04672623e-01 1.85653463e-01 1.04220533e+00 -6.38314188e-01
3.30119938e-01 1.15655243e+00 1.00763671e-01 -5.25184274e-01
-5.74560046e-01 -8.09830070e-01 -8.90879810e-01 -7.68930078e-01
1.08259308e+00 -1.55022442e+00 -1.10410988e+00 5.84290862e-01
-1.04571021e+00 -2.45594040e-01 -6.37441397e-01 9.33422446e-01
-1.09113085e+00 3.08916062e-01 -4.41853583e-01 -8.44998240e-01
1.61859356e-02 -8.60177457e-01 1.75668073e+00 -1.76332071e-02
-4.99939859e-01 -7.24365413e-01 3.31296235e-01 6.21728897e-01
1.40974522e-01 6.35451853e-01 1.82511825e-02 -1.97128505e-01
-4.74798799e-01 -8.83068442e-02 -1.03578545e-01 1.25195712e-01
1.90709326e-02 -2.53417283e-01 -1.62807012e+00 -2.31710657e-01
-4.58449095e-01 -8.89804959e-01 1.47994137e+00 4.16464448e-01
8.90175462e-01 2.90927589e-01 -1.11876495e-01 6.12157702e-01
8.75836909e-01 -1.74431220e-01 6.67945683e-01 1.55245170e-01
8.36449385e-01 5.69039345e-01 7.21288919e-01 8.92363250e-01
4.86898899e-01 5.25462210e-01 8.60351205e-01 2.15007529e-01
-2.54241645e-01 8.85166451e-02 7.38814712e-01 4.71077748e-02
-5.95922709e-01 -1.67509556e-01 -6.35819077e-01 8.19284022e-01
-2.30809474e+00 -1.59605658e+00 1.13351882e-01 1.93021643e+00
2.57928163e-01 -2.21916482e-01 5.02381086e-01 2.01099068e-01
2.98790216e-01 4.76910949e-01 -5.66118062e-01 -3.00894007e-02
-2.40037084e-01 1.72164604e-01 1.77109018e-01 1.27539948e-01
-1.81480527e+00 6.99828684e-01 6.48982143e+00 4.85242456e-01
-9.85550106e-01 4.12763894e-01 7.66284466e-02 -8.42815876e-01
2.09110528e-01 -2.28126571e-01 -3.35704029e-01 2.38245383e-01
9.96180713e-01 3.11261177e-01 4.73050922e-01 5.67404449e-01
2.67662078e-01 -3.87928516e-01 -1.49764395e+00 1.32201791e+00
2.05408975e-01 -1.02583206e+00 -3.41679722e-01 -2.87287056e-01
8.16858411e-01 3.63041669e-01 1.04640938e-01 4.98963684e-01
3.71729732e-01 -8.86496246e-01 1.05114865e+00 9.80932653e-01
6.71021223e-01 -4.84770834e-01 3.47855687e-01 -6.98446855e-02
-1.54151738e+00 -4.14481461e-01 3.35365742e-01 -1.20169550e-01
4.08753097e-01 -1.89538288e-03 -1.50408074e-01 7.88321853e-01
1.05362117e+00 1.43932831e+00 -5.86605191e-01 8.27944517e-01
-1.12094410e-01 3.11378151e-01 -3.65417808e-01 8.03259909e-01
3.04883689e-01 2.91649699e-01 6.88075304e-01 1.22770810e+00
3.06555986e-01 -4.69305702e-02 2.73394976e-02 2.37450674e-01
2.84877002e-01 -4.77498829e-01 -6.32117093e-01 -3.01327351e-02
1.29884273e-01 1.20175815e+00 -1.21651523e-01 -3.55933845e-01
-5.75637639e-01 1.13764668e+00 2.27937028e-01 6.54420853e-01
-1.08200943e+00 6.41534030e-02 1.05894029e+00 -3.91057014e-01
4.68026280e-01 -2.25913227e-01 5.53534925e-02 -1.28605175e+00
-7.24210441e-02 -7.77198732e-01 1.10503459e+00 -1.13041544e+00
-1.17718899e+00 2.71498621e-01 2.01139510e-01 -1.81945181e+00
-6.94264889e-01 -3.37201387e-01 -3.65452677e-01 6.19222224e-01
-1.21858811e+00 -1.79996848e+00 -6.09573901e-01 1.22149253e+00
4.69660044e-01 -2.18886301e-01 9.89444435e-01 4.51242924e-01
-3.16020429e-01 3.38763535e-01 -2.48244748e-01 -1.36155039e-01
1.12515306e+00 -1.18844831e+00 4.71307449e-02 8.98103058e-01
2.11122528e-01 3.40347648e-01 5.23213685e-01 -2.20260397e-01
-1.55002260e+00 -9.81713235e-01 4.72405761e-01 -9.69832897e-01
7.77388275e-01 -2.44096160e-01 -2.57898957e-01 8.89364600e-01
3.79045516e-01 2.67983764e-01 6.15574300e-01 3.48702461e-01
-7.65143096e-01 -5.43513186e-02 -9.40659702e-01 4.68902349e-01
1.50074327e+00 -1.06246877e+00 -5.36544025e-01 3.56290102e-01
5.92309833e-01 -4.33559507e-01 -1.02510989e+00 4.90985006e-01
9.86869097e-01 -1.31075168e+00 1.21091044e+00 -8.06313276e-01
5.36045432e-01 -5.16090989e-01 -4.61487323e-01 -1.13368821e+00
-4.45130289e-01 -7.01535106e-01 -7.90963292e-01 9.84195590e-01
-2.44354129e-01 -3.42548281e-01 5.97166538e-01 1.31250113e-01
-2.99295664e-01 -2.58592963e-01 -1.40254462e+00 -6.69363797e-01
-7.39429295e-01 -9.49894786e-01 2.81623662e-01 7.46933997e-01
1.00187764e-01 2.28688002e-01 -6.59478068e-01 6.47193985e-03
6.50044322e-01 -6.23320341e-02 9.89920318e-01 -8.56503367e-01
-4.49467659e-01 -3.52338880e-01 -9.13482487e-01 -8.62893879e-01
2.19140202e-01 -5.79377830e-01 -2.35367328e-01 -1.31425226e+00
5.32188416e-02 3.84269387e-01 -8.26810539e-01 6.61641717e-01
9.65762287e-02 7.31111825e-01 1.72802716e-01 1.32124543e-01
-1.20139241e+00 8.38030398e-01 1.23321044e+00 -2.52073109e-01
-6.41841022e-03 -5.83299816e-01 -3.51763248e-01 7.60049284e-01
1.76853061e-01 -1.33898392e-01 -5.32093704e-01 -5.29349148e-01
-2.58092768e-02 1.84725463e-01 9.86049056e-01 -1.48500073e+00
1.59391582e-01 -1.04124472e-01 5.15129447e-01 -9.15838838e-01
1.19954002e+00 -9.23824906e-01 1.94436431e-01 4.56104055e-02
-4.40522224e-01 -1.77099288e-01 1.37562379e-01 9.62616324e-01
-2.28040278e-01 6.79884851e-01 4.19697762e-01 1.17518395e-01
-1.38549769e+00 2.69329786e-01 -1.50599390e-01 -3.13265286e-02
1.08284163e+00 -4.04498935e-01 -6.15338624e-01 -5.61691999e-01
-1.20115674e+00 3.52095813e-01 1.78515628e-01 7.14377105e-01
4.37081546e-01 -1.66876972e+00 -6.74163580e-01 5.46837412e-02
6.13587737e-01 -5.48165917e-01 9.48763967e-01 1.56055737e+00
1.51941366e-02 3.17869425e-01 -4.83287364e-01 -1.01835430e+00
-1.33810341e+00 3.33180368e-01 3.54182571e-01 -1.39287636e-01
-6.17650568e-01 9.89474535e-01 3.56053591e-01 -1.00331113e-01
4.61358100e-01 -3.07986796e-01 -3.70016605e-01 4.64773774e-01
6.58077180e-01 5.36275446e-01 -9.86652300e-02 -1.01695466e+00
-6.03851736e-01 5.39044201e-01 3.34273517e-01 -4.88660753e-01
1.20242536e+00 -1.93582743e-01 1.81511343e-01 7.41375268e-01
1.02493691e+00 -5.88019133e-01 -1.62122858e+00 -3.73721123e-01
-5.45367062e-01 -7.19039440e-01 7.45873228e-02 -7.81047523e-01
-8.75480592e-01 9.14108634e-01 9.19531643e-01 -1.63472384e-01
1.34491670e+00 7.64853954e-02 5.79783618e-01 5.23837268e-01
4.13419247e-01 -1.13184500e+00 5.76258421e-01 7.61307478e-01
9.76138115e-01 -1.24011374e+00 -1.10282283e-02 1.45885035e-01
-8.67672741e-01 9.11976695e-01 5.86814046e-01 -2.47531772e-01
4.13071901e-01 -7.76998177e-02 -1.06159747e-01 -2.18544334e-01
-1.12193179e+00 -7.98677266e-01 3.09934318e-01 7.09942222e-01
3.67332995e-01 -4.05761451e-02 4.46281523e-01 2.54387259e-01
3.64071786e-01 2.27323875e-01 -2.45362874e-02 1.10393643e+00
7.42328446e-03 -4.87587094e-01 -4.18467164e-01 1.81851029e-01
-2.16908112e-01 2.50172168e-01 -2.80315399e-01 5.84538698e-01
3.29181552e-01 1.03790748e+00 2.71507263e-01 -6.66450083e-01
4.27127272e-01 2.41764322e-01 9.20486927e-01 -2.21307635e-01
-8.28840673e-01 1.00493141e-01 5.31238616e-01 -1.48062789e+00
-1.13808584e+00 -1.03998899e+00 -8.66790414e-01 -2.88300753e-01
4.38254401e-02 -4.70748663e-01 2.31263191e-01 1.00388825e+00
5.56209266e-01 9.90994036e-01 3.48895580e-01 -1.62352943e+00
-2.57485062e-01 -1.10773659e+00 -5.24951875e-01 8.86136293e-01
4.80721027e-01 -1.17292774e+00 -3.66376072e-01 3.18267852e-01]
|
[8.299202919006348, 0.6285319328308105]
|
0da4db7e-99a2-46a6-be34-b27bab54cf16
|
fd-on-understanding-the-role-of-deep-feature
|
2305.20048
| null |
https://arxiv.org/abs/2305.20048v2
|
https://arxiv.org/pdf/2305.20048v2.pdf
|
F?D: On understanding the role of deep feature spaces on face generation evaluation
|
Perceptual metrics, like the Fr\'echet Inception Distance (FID), are widely used to assess the similarity between synthetically generated and ground truth (real) images. The key idea behind these metrics is to compute errors in a deep feature space that captures perceptually and semantically rich image features. Despite their popularity, the effect that different deep features and their design choices have on a perceptual metric has not been well studied. In this work, we perform a causal analysis linking differences in semantic attributes and distortions between face image distributions to Fr\'echet distances (FD) using several popular deep feature spaces. A key component of our analysis is the creation of synthetic counterfactual faces using deep face generators. Our experiments show that the FD is heavily influenced by its feature space's training dataset and objective function. For example, FD using features extracted from ImageNet-trained models heavily emphasize hats over regions like the eyes and mouth. Moreover, FD using features from a face gender classifier emphasize hair length more than distances in an identity (recognition) feature space. Finally, we evaluate several popular face generation models across feature spaces and find that StyleGAN2 consistently ranks higher than other face generators, except with respect to identity (recognition) features. This suggests the need for considering multiple feature spaces when evaluating generative models and using feature spaces that are tuned to nuances of the domain of interest.
|
['Guha Balakrishnan', 'Krish Kabra']
|
2023-05-31
| null | null | null | null |
['face-generation']
|
['computer-vision']
|
[ 9.56231505e-02 2.52281576e-02 6.58198893e-02 -5.76124310e-01
-3.72457206e-01 -7.51177251e-01 1.19064569e+00 -2.64096290e-01
-2.91660815e-01 5.80598950e-01 5.55063248e-01 1.16690129e-01
-2.34367803e-01 -9.76586521e-01 -6.42338932e-01 -6.19530976e-01
8.45343322e-02 3.74977775e-02 -5.77182055e-01 -2.26069212e-01
3.32862794e-01 6.04498386e-01 -1.86714935e+00 2.18289524e-01
4.65875447e-01 1.02893007e+00 -3.11832577e-01 3.28390896e-01
1.70906618e-01 3.54871005e-01 -8.21817458e-01 -9.38022494e-01
4.97411132e-01 -8.29635203e-01 -5.76817393e-01 1.03467174e-01
8.45131814e-01 -1.51241675e-01 -2.04324380e-01 1.11421406e+00
7.28005826e-01 -7.24894702e-02 1.07669675e+00 -1.56705403e+00
-1.18829560e+00 5.20407677e-01 -4.80449587e-01 1.37806818e-01
4.92546469e-01 5.50248742e-01 9.82505441e-01 -1.09904408e+00
8.74435186e-01 1.76100111e+00 8.48529994e-01 5.21962583e-01
-1.60456324e+00 -8.81960273e-01 -3.09788674e-01 6.27639145e-02
-1.67564797e+00 -8.00795436e-01 8.11558187e-01 -5.98663568e-01
3.51950198e-01 2.17555761e-01 4.39672321e-01 1.52762926e+00
3.65123838e-01 2.05629483e-01 1.39412105e+00 -3.65334719e-01
1.72481164e-01 2.91454762e-01 -6.99912190e-01 4.79338616e-01
3.53990555e-01 4.83610660e-01 -4.93848443e-01 -1.08015522e-01
7.03799903e-01 -3.04529786e-01 -2.44116873e-01 -2.93309987e-01
-9.09831047e-01 1.19656062e+00 5.41131198e-01 3.62071425e-01
-2.95860052e-01 3.11406195e-01 9.32167917e-02 2.39218041e-01
4.34488326e-01 8.55696440e-01 -5.50852157e-02 8.72284248e-02
-1.05978537e+00 5.76520562e-01 5.20594001e-01 4.80440885e-01
9.14441288e-01 2.09985688e-01 -5.77700257e-01 8.92128885e-01
3.20997834e-01 5.46617091e-01 5.58792233e-01 -1.20533252e+00
-8.68506357e-02 3.70673209e-01 -1.27346143e-01 -1.50311542e+00
-1.92710772e-01 -2.63078213e-01 -5.35513759e-01 4.41933870e-01
6.61612988e-01 -1.96706399e-01 -7.07886517e-01 2.08459830e+00
1.47653893e-01 -1.50861651e-01 -1.81278542e-01 9.51595962e-01
7.78175652e-01 2.66955435e-01 2.17565596e-01 1.19758390e-01
1.39911664e+00 -2.49320492e-01 -3.96522731e-01 -1.84811532e-01
4.28047568e-01 -9.81550097e-01 1.23210871e+00 7.22992420e-02
-9.73083794e-01 -7.52067626e-01 -9.10902619e-01 5.90866022e-02
-5.36364079e-01 1.37675241e-01 5.65421045e-01 1.17018199e+00
-1.27097154e+00 7.33883917e-01 -3.95796634e-02 -4.42435503e-01
6.94144428e-01 2.23513395e-01 -4.18101758e-01 -1.77289546e-02
-1.10836029e+00 9.00505066e-01 5.43131307e-02 -1.70781881e-01
-1.06731057e+00 -7.42233276e-01 -9.36845899e-01 1.41853103e-02
-7.82465041e-02 -6.11089289e-01 8.03339839e-01 -1.42864549e+00
-1.39240670e+00 1.06254137e+00 1.63645595e-01 -2.50842690e-01
5.71170866e-01 2.28942648e-01 -5.07805169e-01 4.81601432e-03
2.53702581e-01 1.08717406e+00 1.22884250e+00 -1.39148378e+00
-1.81301966e-01 -6.28271103e-01 6.07149117e-02 -3.61656584e-02
-3.07868421e-01 6.33021770e-03 3.03053796e-01 -9.71712053e-01
-3.16214055e-01 -1.00101709e+00 2.78837532e-01 1.69912219e-01
-4.58271056e-01 -1.49067163e-01 6.54110670e-01 -4.64762330e-01
8.62796783e-01 -2.22055817e+00 -7.38145411e-02 3.68286580e-01
2.30219275e-01 -9.61052254e-02 -3.99514735e-01 2.37446189e-01
-3.12289447e-01 4.31105614e-01 -9.51262638e-02 -1.86612561e-01
2.89702505e-01 -1.18746735e-01 -2.17445508e-01 6.93541169e-01
4.72899377e-01 9.25440490e-01 -6.46406829e-01 -3.99668515e-01
6.87497109e-02 7.23858774e-01 -7.63934314e-01 -7.84785599e-02
1.93992220e-02 3.81596297e-01 1.43761650e-01 3.73997301e-01
8.19088817e-01 2.23285526e-01 -6.93717971e-02 -2.67542243e-01
-6.82626897e-03 -3.84701043e-02 -8.51679444e-01 1.40869486e+00
-5.19036472e-01 8.43957782e-01 -2.92232275e-01 -5.37007153e-01
1.04768550e+00 1.01866797e-01 1.45459160e-01 -8.12903225e-01
2.72362471e-01 4.23820876e-02 3.41565490e-01 -1.23546779e-01
4.65871930e-01 -3.44032019e-01 1.32406369e-01 4.94500279e-01
3.04157943e-01 -2.98508406e-01 5.03731258e-02 -3.97831798e-02
7.97249973e-01 -1.86602231e-02 -6.53967634e-02 -5.43357372e-01
2.54921407e-01 -3.77419770e-01 4.07911003e-01 4.95187640e-01
-6.79599941e-02 9.18900490e-01 7.83372402e-01 -2.15461314e-01
-1.11350393e+00 -1.24074817e+00 -4.07187700e-01 1.01781058e+00
-1.73758343e-01 -2.80120492e-01 -9.66806591e-01 -7.05755889e-01
2.68502742e-01 9.07683134e-01 -1.20223284e+00 -5.35884619e-01
-1.26077518e-01 -7.04882979e-01 9.08356249e-01 2.60730296e-01
3.37876618e-01 -1.11469221e+00 -5.77955782e-01 -3.05925041e-01
9.31925699e-02 -8.33281398e-01 -4.74098593e-01 -4.55209404e-01
-3.70392412e-01 -1.05472696e+00 -7.67567337e-01 -3.67184997e-01
6.35592937e-01 -2.57046223e-02 1.18137181e+00 -1.66196451e-01
-5.07213652e-01 4.05353725e-01 -2.85256237e-01 -4.28592891e-01
-5.12805581e-01 -2.42558092e-01 1.80254072e-01 3.30012530e-01
3.24939221e-01 -5.12641430e-01 -8.03018689e-01 3.76088738e-01
-1.01533997e+00 -2.99889266e-01 5.05200028e-01 7.48901308e-01
1.59597814e-01 7.39413574e-02 6.59658551e-01 -7.62631118e-01
8.65470767e-01 -5.75021684e-01 -2.97755748e-01 3.97616997e-02
-6.49833739e-01 6.79956153e-02 4.70820159e-01 -3.60744983e-01
-1.07838476e+00 -3.56832802e-01 5.68250678e-02 -4.79938477e-01
-2.36132517e-01 3.20039913e-02 -3.45717311e-01 -1.58206690e-02
1.17620444e+00 -1.06670380e-01 1.71881497e-01 -1.15706667e-01
4.77820009e-01 3.56812000e-01 3.29221517e-01 -6.56041980e-01
8.77393484e-01 5.34526825e-01 5.21717127e-03 -8.33555639e-01
-5.09445608e-01 3.02803487e-01 -4.27903891e-01 -2.25522265e-01
8.01694393e-01 -7.18915820e-01 -7.75697589e-01 3.26459527e-01
-9.70317960e-01 -5.61963059e-02 -5.41079164e-01 4.65427548e-01
-6.16972387e-01 -3.83082107e-02 -2.30221093e-01 -6.76457644e-01
3.15159932e-02 -1.25331306e+00 1.16822243e+00 2.06571743e-01
-6.17644966e-01 -9.28867102e-01 2.98745912e-02 2.32789382e-01
5.04665077e-01 6.65472984e-01 8.13272893e-01 -3.50117683e-01
-1.49109200e-01 1.66189752e-03 -4.06569123e-01 4.68708992e-01
2.89482564e-01 3.58563036e-01 -1.41646969e+00 -2.34958589e-01
-9.35989618e-03 -3.44245195e-01 8.39274168e-01 5.00438690e-01
1.28229094e+00 -5.79403281e-01 -1.80043560e-02 6.79567337e-01
1.34424114e+00 4.14387807e-02 8.36271763e-01 -7.44185075e-02
5.31315863e-01 9.76795793e-01 1.94347814e-01 4.24805164e-01
1.42205045e-01 7.24990070e-01 3.24032992e-01 -5.80644682e-02
-3.76279444e-01 -4.95427221e-01 6.15433455e-01 -1.84436008e-01
6.40630350e-02 -6.91590831e-02 -7.09692240e-01 4.82244015e-01
-1.10989261e+00 -1.22400486e+00 1.67710781e-01 2.13013554e+00
7.44092226e-01 -6.27506077e-02 3.58678401e-01 1.30323559e-01
7.85176575e-01 2.71036893e-01 -3.88813645e-01 -6.57569587e-01
-4.82771516e-01 4.37277436e-01 3.27514410e-01 2.43322119e-01
-8.31877410e-01 7.74079859e-01 6.41379690e+00 8.19801271e-01
-1.28254557e+00 -5.09648509e-02 1.12646568e+00 -1.77434310e-01
-5.55061638e-01 -1.32683069e-01 -4.53806281e-01 6.31069005e-01
9.01611388e-01 -3.77856702e-01 4.92786705e-01 7.56473422e-01
1.70573577e-01 -4.41985438e-03 -1.34336030e+00 1.06711936e+00
3.38399738e-01 -1.31459010e+00 1.76578596e-01 3.69141966e-01
8.49749625e-01 -4.12500739e-01 8.84495497e-01 9.83523205e-02
3.91928226e-01 -1.71537721e+00 1.03783906e+00 3.89220029e-01
1.18183970e+00 -8.04083288e-01 3.86231780e-01 -3.68996620e-01
-6.68759406e-01 -1.66214228e-01 -4.24730808e-01 2.17059460e-02
-3.79791617e-01 6.75505579e-01 -1.03799069e+00 1.64852291e-02
6.81982934e-01 2.75070280e-01 -8.94166529e-01 5.96388698e-01
-1.28934428e-01 3.78962785e-01 7.07067922e-03 3.01084310e-01
9.95664075e-02 7.08941789e-03 3.27274710e-01 1.17359066e+00
5.46307445e-01 -2.87532717e-01 -5.27759731e-01 1.53966486e+00
-3.32244337e-01 8.11321512e-02 -1.05763257e+00 -1.76666260e-01
5.74943721e-01 1.28334546e+00 -5.96534669e-01 1.51318714e-01
-2.29865953e-01 8.79628181e-01 -1.24038413e-01 2.94750899e-01
-9.95263278e-01 -2.75944650e-01 1.10087550e+00 5.13831079e-01
4.69750464e-02 1.50736064e-01 -2.86177605e-01 -7.47614980e-01
-2.19653428e-01 -1.02448308e+00 2.79558636e-02 -8.80696714e-01
-1.44103050e+00 6.52894795e-01 -1.94715969e-02 -9.10693526e-01
-3.51442039e-01 -6.25387967e-01 -6.47789538e-01 9.50432718e-01
-9.80285645e-01 -1.21533847e+00 -2.79772103e-01 7.13900149e-01
1.49013981e-01 -3.27743083e-01 7.87951291e-01 8.61735493e-02
-4.08755839e-01 1.02105403e+00 -1.15751386e-01 2.82475561e-01
7.92515457e-01 -1.05517972e+00 4.65717733e-01 5.95877886e-01
4.49766725e-01 6.44533932e-01 7.58596241e-01 -3.40968192e-01
-1.02268600e+00 -1.07456052e+00 6.14247262e-01 -6.58685446e-01
2.74519533e-01 -2.94871151e-01 -3.47643375e-01 4.36919421e-01
1.96135402e-01 -2.01117601e-02 8.43298972e-01 -1.82023011e-02
-7.53322661e-01 -2.40978211e-01 -1.58416164e+00 6.51928127e-01
1.28102601e+00 -6.91891670e-01 -2.59308785e-01 3.41268666e-02
3.00919652e-01 1.54708698e-01 -8.97120237e-01 3.32920134e-01
7.93842912e-01 -1.42103064e+00 1.03490925e+00 -4.56988335e-01
7.91186750e-01 -1.98393185e-02 -3.36028486e-01 -1.73368979e+00
-5.39911687e-01 -3.64556551e-01 4.66065317e-01 1.38708079e+00
3.80464882e-01 -6.12928569e-01 6.70723081e-01 3.73165309e-01
2.61128962e-01 -5.12886763e-01 -7.96295285e-01 -6.43966258e-01
3.44940841e-01 -1.96724549e-01 1.03160393e+00 1.10551047e+00
-4.78229493e-01 3.16095978e-01 -1.70491356e-02 -2.96681017e-01
6.07849240e-01 -3.84207182e-02 7.96656609e-01 -1.37045968e+00
6.96985349e-02 -7.46820390e-01 -7.19929159e-01 -5.32598570e-02
4.80526865e-01 -1.06573117e+00 -3.42893928e-01 -9.77258563e-01
1.09474801e-01 -3.85383636e-01 3.75766046e-02 2.22334698e-01
1.72725812e-01 6.73847675e-01 3.55871588e-01 -7.48282373e-02
3.16443108e-02 6.86651886e-01 1.30996501e+00 -1.30111381e-01
2.01828688e-01 -3.36788535e-01 -1.14143276e+00 7.86424756e-01
7.68585145e-01 -3.22808057e-01 -4.65699255e-01 -4.45953816e-01
7.06589296e-02 -3.80093515e-01 8.25043261e-01 -1.00914264e+00
-3.23347032e-01 -1.99291497e-01 1.04409051e+00 2.98375815e-01
5.06437719e-01 -4.51320797e-01 3.68167430e-01 2.83770621e-01
-3.65569800e-01 2.29475960e-01 1.20307654e-01 1.50021300e-01
-1.11564972e-01 -4.46197577e-03 1.08538747e+00 -1.56524599e-01
-4.78736401e-01 1.88481763e-01 -1.58593338e-03 3.53345066e-01
9.99461055e-01 -5.10338426e-01 -2.98516124e-01 -5.48181295e-01
-5.30630887e-01 -4.72402155e-01 8.66741002e-01 7.90407896e-01
6.37069166e-01 -1.63215303e+00 -1.04649997e+00 4.75475758e-01
2.85120964e-01 -6.45125926e-01 9.38591138e-02 4.55216974e-01
-3.78097862e-01 2.88540989e-01 -5.70640385e-01 -5.80481470e-01
-9.32682991e-01 3.60935301e-01 4.23632264e-01 3.25557858e-01
6.06481321e-02 1.09007645e+00 5.65434337e-01 -3.45667690e-01
-2.52503097e-01 -3.05165034e-02 3.45514119e-02 4.34680223e-01
3.92030835e-01 4.97088492e-01 -1.60227646e-03 -1.19605911e+00
-2.82938212e-01 4.10421014e-01 2.40718707e-01 -3.54483098e-01
1.17239714e+00 9.41953585e-02 -6.88085034e-02 1.20550923e-01
1.48406994e+00 1.61732167e-01 -1.30043197e+00 2.17137009e-01
-1.43760905e-01 -9.24109697e-01 -1.13495007e-01 -7.74671733e-01
-1.34041131e+00 7.21597552e-01 9.72619474e-01 1.51475459e-01
8.90980124e-01 -1.24034630e-02 2.54978865e-01 -2.52271116e-01
1.94644660e-01 -1.08175719e+00 2.92898595e-01 3.58636640e-02
1.28205311e+00 -1.11574519e+00 -1.25848696e-01 -5.07021099e-02
-6.99623585e-01 6.52544796e-01 6.13608062e-01 -1.84854597e-01
7.07779467e-01 -5.41684292e-02 4.29522693e-02 -3.04354012e-01
-3.78544271e-01 -2.78621186e-02 3.84586483e-01 7.94988453e-01
6.82950914e-01 2.30659857e-01 -3.28566879e-01 4.36016113e-01
-9.81743157e-01 -2.99628466e-01 4.16688532e-01 1.77737907e-01
-4.41633649e-02 -1.02682924e+00 -6.05801940e-01 5.25347531e-01
-6.64340854e-01 1.07823037e-01 -6.82459235e-01 7.78426468e-01
6.94489658e-01 9.84603107e-01 3.71924430e-01 -4.99511242e-01
1.22828193e-01 2.68068407e-02 7.54995346e-01 -6.75806344e-01
-5.35984635e-01 -5.05569458e-01 -1.35404795e-01 -5.94598651e-01
-3.02986413e-01 -7.41257787e-01 -7.72853673e-01 -7.07436979e-01
-1.23714931e-01 -7.46473670e-02 7.53656864e-01 5.85225165e-01
3.85740072e-01 1.32037327e-01 9.22265232e-01 -9.07593250e-01
-5.18430233e-01 -9.67789233e-01 -6.90179825e-01 9.03075278e-01
-9.81547497e-03 -9.62522924e-01 -3.97760153e-01 1.93697978e-02]
|
[12.807619094848633, 0.9665404558181763]
|
82d52895-89d8-4386-be1d-64ab6ced40e0
|
analysis-of-numerical-integration-in-rnn
|
2305.0467
| null |
https://arxiv.org/abs/2305.04670v1
|
https://arxiv.org/pdf/2305.04670v1.pdf
|
Analysis of Numerical Integration in RNN-Based Residuals for Fault Diagnosis of Dynamic Systems
|
Data-driven modeling and machine learning are widely used to model the behavior of dynamic systems. One application is the residual evaluation of technical systems where model predictions are compared with measurement data to create residuals for fault diagnosis applications. While recurrent neural network models have been shown capable of modeling complex non-linear dynamic systems, they are limited to fixed steps discrete-time simulation. Modeling using neural ordinary differential equations, however, make it possible to evaluate the state variables at specific times, compute gradients when training the model and use standard numerical solvers to explicitly model the underlying dynamic of the time-series data. Here, the effect of solver selection on the performance of neural ordinary differential equation residuals during training and evaluation is investigated. The paper includes a case study of a heavy-duty truck's after-treatment system to highlight the potential of these techniques for improving fault diagnosis performance.
|
['Mattias Krysander', 'Daniel Jung', 'Theodor Westny', 'Arman Mohammadi']
|
2023-05-08
| null | null | null | null |
['numerical-integration']
|
['miscellaneous']
|
[ 1.21500358e-01 -8.55531916e-02 1.50711671e-01 -6.90632686e-02
-3.90622258e-01 3.02020274e-02 9.36876386e-02 -2.46328451e-02
5.48326857e-02 5.37890196e-01 -2.46127442e-01 -7.87327230e-01
-7.34633148e-01 -3.77580911e-01 -2.11731523e-01 -8.26120973e-01
-1.61792189e-01 6.54239476e-01 -1.81243539e-01 -3.76783371e-01
7.46126100e-02 1.00551593e+00 -1.29996014e+00 1.52340293e-01
5.90171933e-01 7.91340888e-01 -3.03376913e-02 8.42281818e-01
1.82806358e-01 1.17812443e+00 -7.55076706e-01 6.22716308e-01
3.51008140e-02 -4.92019385e-01 -6.02735996e-01 1.58531278e-01
-5.90585411e-01 -5.86458325e-01 -6.50367081e-01 7.10897386e-01
5.94944358e-01 4.57264096e-01 6.84441090e-01 -1.18053126e+00
4.26404877e-03 2.07075834e-01 1.54869989e-01 4.50338185e-01
2.45011702e-01 3.59356731e-01 2.16219097e-01 -8.27903986e-01
2.26862848e-01 8.23045313e-01 1.15599990e+00 2.93429852e-01
-1.32474458e+00 -1.78209752e-01 -1.77555859e-01 1.34799778e-01
-1.13844621e+00 -2.17661649e-01 9.07534719e-01 -7.63434589e-01
1.55671084e+00 2.64115065e-01 6.38520360e-01 5.48871279e-01
6.88146770e-01 3.06362867e-01 8.62521827e-01 -1.96029663e-01
4.80587959e-01 -1.45041630e-01 5.22862434e-01 3.74549568e-01
-1.54017436e-03 4.84031022e-01 6.86340034e-02 -3.44542325e-01
8.21491420e-01 2.97127485e-01 -4.52935636e-01 -1.11077234e-01
-5.14311373e-01 7.07745314e-01 2.08240524e-01 3.81237447e-01
-1.00850356e+00 1.00420497e-01 6.63827956e-01 6.72118306e-01
6.71435654e-01 7.33735025e-01 -5.47113121e-01 -2.48173803e-01
-9.78943050e-01 4.85769868e-01 8.88779819e-01 2.72345603e-01
3.75927478e-01 1.09430945e+00 1.43565804e-01 8.15379441e-01
8.31095427e-02 2.39782080e-01 6.81575119e-01 -1.10026395e+00
-8.93227160e-02 5.33287287e-01 1.56838447e-01 -6.68673337e-01
-7.35493243e-01 -4.36875880e-01 -7.58162498e-01 5.47506154e-01
1.95565417e-01 -5.83158910e-01 -1.20411527e+00 1.00283825e+00
-8.74305144e-03 3.36573720e-01 2.62602031e-01 8.54891181e-01
3.61313373e-02 8.97334278e-01 -2.89198518e-01 -6.21562123e-01
4.63402361e-01 -3.77152979e-01 -8.01293612e-01 -1.03188418e-02
1.12701011e+00 -3.27060521e-01 4.49768513e-01 3.70508969e-01
-1.37919426e+00 -3.44890624e-01 -9.18962240e-01 5.82756937e-01
-2.84536541e-01 2.73345690e-02 5.14146760e-02 7.71864504e-02
-9.26278889e-01 1.21901894e+00 -1.25806344e+00 1.98477618e-02
-2.52885759e-01 5.46478510e-01 1.06080681e-01 1.49978831e-01
-1.39929390e+00 1.46879351e+00 8.40044543e-02 7.85234392e-01
-1.03821933e+00 -1.10586858e+00 -6.79493487e-01 1.82241336e-01
1.96114793e-01 -4.21012402e-01 1.61324477e+00 -7.36934245e-01
-1.51906550e+00 3.07233576e-02 -2.66055129e-02 -7.04044819e-01
4.08642113e-01 -1.19232330e-02 -7.28749752e-01 -3.61772142e-02
-3.72038156e-01 -4.15421933e-01 6.06340706e-01 -8.10911655e-01
-1.88459143e-01 2.25013375e-01 -2.98308223e-01 4.46337871e-02
1.27371103e-01 -1.65910482e-01 2.49526650e-01 -3.34648013e-01
3.30295593e-01 -9.49330032e-01 -7.09754884e-01 -5.04798949e-01
-1.55987903e-01 8.58243108e-02 1.02170169e+00 -1.12957096e+00
1.16922212e+00 -2.00416970e+00 -6.34360239e-02 6.23316944e-01
-3.06175321e-01 3.81356835e-01 3.25544059e-01 9.40159976e-01
-7.66112149e-01 -3.74848306e-01 -4.82542485e-01 2.65844256e-01
-3.54991019e-01 2.48159200e-01 -3.16933423e-01 5.37318349e-01
5.26227534e-01 5.06907284e-01 -4.10644442e-01 1.56343788e-01
5.66853583e-01 6.05203807e-01 -2.25004345e-01 2.64909595e-01
-3.21673676e-02 2.84218580e-01 -4.31886077e-01 1.91179618e-01
-1.59209669e-02 -1.63095817e-01 4.45010047e-03 -9.14642438e-02
-4.46266197e-02 3.65238130e-01 -1.22114527e+00 8.36918294e-01
-6.98103249e-01 8.41722071e-01 1.28909245e-01 -1.46305525e+00
9.14723814e-01 7.71316290e-01 8.72083545e-01 -7.15768754e-01
2.68298715e-01 3.14377904e-01 2.73750275e-01 -7.74220645e-01
3.55865866e-01 -3.19281101e-01 3.83731991e-01 4.18284416e-01
-3.35088819e-01 -4.05896127e-01 1.05214730e-01 -1.70378342e-01
1.28089106e+00 -4.85306792e-02 -1.66298017e-01 -2.81403512e-01
3.77000183e-01 5.34509957e-01 7.53687680e-01 8.77624825e-02
2.50582159e-01 4.43592578e-01 6.15042686e-01 -4.56713647e-01
-1.22933960e+00 -4.02258694e-01 -1.45451501e-01 3.88925135e-01
-1.91815242e-01 2.12975040e-01 -6.06704712e-01 4.89407815e-02
2.20372945e-01 1.10538447e+00 -4.68354344e-01 -8.38074982e-01
-7.93759227e-01 -7.58444905e-01 2.84518421e-01 8.04551840e-01
-1.64411604e-01 -1.03322029e+00 -8.73065948e-01 8.29489827e-01
5.94238698e-01 -5.74764371e-01 1.92897335e-01 7.28867829e-01
-1.28943467e+00 -1.07592332e+00 -6.56511605e-01 -5.72108150e-01
8.57387662e-01 -4.14397389e-01 8.02216470e-01 3.41107845e-01
-6.43709481e-01 6.12469375e-01 1.63870886e-01 -2.36994177e-01
-8.51819158e-01 -2.79403150e-01 2.66660273e-01 -3.88994426e-01
-2.48735026e-01 -3.12863141e-01 -1.04647696e-01 2.90360034e-01
-8.62544060e-01 -1.26277924e-01 1.59031540e-01 1.26245880e+00
4.25187290e-01 5.78324258e-01 7.51885533e-01 -8.81958187e-01
1.11882734e+00 -5.62507987e-01 -7.15804875e-01 1.74307689e-01
-1.12690532e+00 -4.26176516e-03 6.89021051e-01 -7.18659222e-01
-1.05493832e+00 -5.31931482e-02 6.22483902e-03 -9.13645744e-01
1.14973687e-01 1.28104699e+00 4.61662173e-01 4.56270836e-02
4.39242780e-01 -1.65011026e-02 5.80505371e-01 -4.32069540e-01
-4.79643971e-01 3.91374886e-01 6.30486488e-01 -2.28827193e-01
4.71144319e-01 -2.24919364e-01 9.51339751e-02 -7.94038475e-01
1.10524101e-02 -2.33470723e-01 -2.66199768e-01 -4.59062696e-01
1.06029280e-01 -6.61841094e-01 -7.78196990e-01 6.98381782e-01
-8.46646905e-01 -8.39301109e-01 -6.07800364e-01 6.46373391e-01
-6.18989646e-01 -2.65510559e-01 -1.16559649e+00 -1.17999649e+00
-2.74727345e-01 -1.17584395e+00 4.82172221e-01 1.15704022e-01
-3.48111153e-01 -1.32790697e+00 2.08734944e-01 -4.90552574e-01
6.69923484e-01 3.85631591e-01 1.12796557e+00 -5.54498196e-01
1.08615272e-01 -7.31218636e-01 4.72497314e-01 6.52561128e-01
-4.52427864e-02 3.43442142e-01 -6.85288846e-01 -4.76323038e-01
6.66501641e-01 1.38191581e-01 2.51917571e-01 5.94456494e-01
7.40297019e-01 -5.08622751e-02 -3.73182446e-01 2.04282284e-01
1.42042267e+00 6.78894758e-01 4.87974405e-01 1.04186460e-01
5.06066084e-01 6.79786742e-01 6.35011315e-01 3.45262736e-01
-2.73386270e-01 1.97627246e-01 2.29954332e-01 -1.60367951e-01
3.60732615e-01 3.62988710e-01 4.35194016e-01 9.27057803e-01
1.20026603e-01 3.30105014e-02 -1.43439496e+00 6.76867783e-01
-1.64127731e+00 -8.48909497e-01 -5.32864273e-01 1.85373926e+00
2.97645539e-01 3.44153047e-01 -1.14345826e-01 7.56841779e-01
7.40266204e-01 -5.09989500e-01 -9.16188478e-01 -7.81473696e-01
2.02010751e-01 7.53930807e-02 6.29253507e-01 5.26818752e-01
-4.17837322e-01 1.23268351e-01 7.10169458e+00 2.33094931e-01
-1.53867972e+00 -2.17383504e-01 5.79868853e-01 -2.40867794e-01
2.47333542e-01 -4.77991067e-02 -2.67427653e-01 3.16613615e-01
1.81957674e+00 -3.94879252e-01 3.60975415e-01 7.96506345e-01
7.31197536e-01 -2.44905591e-01 -1.00863016e+00 5.59720457e-01
-3.56311023e-01 -1.12646842e+00 -6.19474113e-01 -8.76574516e-02
8.01573396e-01 -1.19349584e-01 -1.66435942e-01 4.13559228e-01
2.42974475e-01 -9.44766879e-01 2.38956213e-01 9.23277557e-01
3.98183614e-01 -8.64289641e-01 1.01227939e+00 3.57985079e-01
-7.29151428e-01 -5.82278192e-01 -2.93848664e-02 -4.33986247e-01
6.31818473e-01 7.38119841e-01 -1.03789175e+00 3.80149275e-01
4.69602913e-01 4.79554594e-01 -2.04564705e-01 1.08131564e+00
4.08122718e-01 7.81645954e-01 -2.91896135e-01 8.39759707e-02
2.98934042e-01 -9.63602364e-02 3.98976058e-01 7.07292616e-01
3.68242174e-01 9.05326679e-02 1.68857247e-01 7.97650993e-01
8.29977572e-01 -3.91722530e-01 -5.68845809e-01 -2.28474706e-01
1.47553876e-01 7.90380597e-01 -4.62324977e-01 -4.16498363e-01
-2.53623407e-02 4.85752732e-01 -1.97188973e-01 7.89760590e-01
-7.88736761e-01 -4.68734533e-01 8.02275836e-01 3.27862412e-01
-3.81648876e-02 -3.26802343e-01 -2.33306184e-01 -5.57155430e-01
-1.36111110e-01 -8.74305665e-01 1.51704252e-01 -1.13275516e+00
-9.52360809e-01 5.22273660e-01 1.76689357e-01 -1.24787533e+00
-1.32407534e+00 -6.30371332e-01 -8.35482121e-01 1.26020324e+00
-8.36330950e-01 -1.83793709e-01 1.00205548e-01 4.58072156e-01
4.85789210e-01 -1.96237639e-01 7.73880422e-01 2.67510027e-01
-1.11776888e+00 -2.69129902e-01 7.61330307e-01 -1.19406499e-01
1.73487619e-01 -9.57548380e-01 3.72079521e-01 8.52266669e-01
-6.75217927e-01 4.05306160e-01 1.24368668e+00 -9.87349510e-01
-1.51961851e+00 -7.91189551e-01 3.03054422e-01 1.92833170e-01
6.89989984e-01 3.30434710e-01 -1.63509607e+00 4.77053255e-01
-9.08018053e-02 4.05110829e-02 -5.31687438e-02 -3.53734702e-01
5.52401602e-01 -1.61118265e-02 -1.15541089e+00 3.70198429e-01
1.12961233e-01 -8.65040243e-01 -3.18702728e-01 2.03776032e-01
2.14981526e-01 -6.97431624e-01 -1.07764876e+00 7.38368332e-01
1.30812198e-01 -5.56673110e-01 6.49458766e-01 -6.21804714e-01
9.91514698e-02 -2.02012911e-01 4.26638484e-01 -1.68174827e+00
-2.94750750e-01 -5.44479311e-01 -5.23819983e-01 8.49851549e-01
4.59270269e-01 -8.92693520e-01 7.12595165e-01 1.23313177e+00
-4.45043385e-01 -1.08558583e+00 -7.88161635e-01 -6.84334040e-01
-6.14199452e-02 -6.09430850e-01 2.92853445e-01 8.47348511e-01
4.81907185e-03 -1.02997497e-01 -8.23898092e-02 4.30607706e-01
-1.08943477e-01 -2.09588021e-01 1.46076813e-01 -1.02496183e+00
-2.20231935e-01 -3.23353350e-01 -3.69401604e-01 -2.06240326e-01
2.21835583e-01 -3.41403306e-01 4.72734571e-01 -1.43785846e+00
-7.04247713e-01 -3.52197409e-01 -3.91886353e-01 2.02607483e-01
2.12373026e-02 -2.76323229e-01 -2.26094082e-01 2.13429049e-01
4.51018155e-01 2.01456472e-01 5.18703759e-01 -9.39700380e-02
-4.03141737e-01 2.07825273e-01 2.27362618e-01 5.46610653e-01
9.24437046e-01 -4.68476444e-01 -6.79061949e-01 -3.04720521e-01
-4.66872416e-02 7.38474667e-01 5.08933067e-01 -1.29413331e+00
4.95658219e-01 -1.15718447e-01 3.37764502e-01 -6.84254885e-01
2.94084460e-01 -1.05206048e+00 8.04204702e-01 9.78548765e-01
-3.56314868e-01 7.87371337e-01 4.23579633e-01 2.57180244e-01
-4.14748162e-01 -4.68612492e-01 7.25340366e-01 2.39534482e-01
-6.44846499e-01 -1.96244698e-02 -1.08084035e+00 -3.48161370e-01
9.58101332e-01 -4.36960965e-01 1.33807406e-01 -5.26775062e-01
-1.07880199e+00 2.49894202e-01 2.12250620e-01 1.98679771e-02
5.76635122e-01 -1.06513643e+00 -4.08420861e-01 3.89737338e-01
-5.67728579e-01 -8.24385211e-02 5.52214324e-01 1.00085902e+00
-7.05066681e-01 2.94073880e-01 -8.40618610e-02 -5.75221658e-01
-9.53231573e-01 4.64982539e-01 1.16578865e+00 -3.13064218e-01
-7.27924824e-01 3.50681186e-01 -4.16760653e-01 -2.84351736e-01
6.10990040e-02 -3.61552984e-01 2.26758525e-01 -9.49246585e-02
2.62093812e-01 8.15203011e-01 5.57646215e-01 -3.48208189e-01
-7.17489719e-02 1.51488140e-01 1.22453146e-01 -1.72186017e-01
1.56234884e+00 2.17847526e-01 2.61425413e-02 9.44939137e-01
1.27842021e+00 -1.07536864e+00 -1.34546733e+00 -3.96272689e-02
3.79495621e-01 3.95529807e-01 4.82196599e-01 -6.74133301e-01
-1.26720393e+00 7.25703537e-01 8.15224767e-01 6.35106027e-01
1.32465148e+00 -8.50059986e-01 6.03836238e-01 7.48715460e-01
-4.11186814e-02 -1.30258656e+00 -3.47857803e-01 7.86158502e-01
8.30310643e-01 -5.90480149e-01 -2.54819612e-03 4.52545844e-02
-5.35030663e-01 1.45162976e+00 4.63095516e-01 -6.67419910e-01
7.71689713e-01 8.14824283e-01 2.36394405e-01 -1.34766772e-01
-1.01212215e+00 4.49477792e-01 -1.41173732e-02 2.21853405e-01
4.04190242e-01 -4.25786048e-01 -5.93345263e-04 -1.39112361e-02
2.12237492e-01 2.67414123e-01 6.80926919e-01 1.31472576e+00
-3.06893010e-02 -6.22018635e-01 -6.09879136e-01 8.47459376e-01
-4.74999189e-01 2.67204851e-01 2.89546996e-02 9.57183480e-01
-6.16282880e-01 7.79910505e-01 1.76611438e-01 -2.43458793e-01
6.29527509e-01 5.17974436e-01 -5.86731452e-03 -4.72836047e-01
-9.42302883e-01 1.88152604e-02 2.65738815e-01 -4.57265168e-01
1.87391520e-01 -7.71742880e-01 -1.55035543e+00 -2.53769070e-01
-5.67799270e-01 3.67750973e-01 9.37177539e-01 1.06607533e+00
2.81496972e-01 1.30962622e+00 6.11441970e-01 -9.65803027e-01
-8.16288650e-01 -9.53312457e-01 -7.30138361e-01 6.40879478e-03
6.07701540e-01 -6.36575222e-01 -5.39975524e-01 -2.82263290e-02]
|
[6.566550254821777, 2.685302972793579]
|
0ec1b063-d7c2-4114-8027-e7ad25ad9364
|
generative-steganography-network
|
2207.13867
| null |
https://arxiv.org/abs/2207.13867v3
|
https://arxiv.org/pdf/2207.13867v3.pdf
|
Generative Steganography Network
|
Steganography usually modifies cover media to embed secret data. A new steganographic approach called generative steganography (GS) has emerged recently, in which stego images (images containing secret data) are generated from secret data directly without cover media. However, existing GS schemes are often criticized for their poor performances. In this paper, we propose an advanced generative steganography network (GSN) that can generate realistic stego images without using cover images. We firstly introduce the mutual information mechanism in GS, which helps to achieve high secret extraction accuracy. Our model contains four sub-networks, i.e., an image generator ($G$), a discriminator ($D$), a steganalyzer ($S$), and a data extractor ($E$). $D$ and $S$ act as two adversarial discriminators to ensure the visual quality and security of generated stego images. $E$ is to extract the hidden secret from generated stego images. The generator $G$ is flexibly constructed to synthesize either cover or stego images with different inputs. It facilitates covert communication by concealing the function of generating stego images in a normal generator. A module named secret block is designed to hide secret data in the feature maps during image generation, with which high hiding capacity and image fidelity are achieved. In addition, a novel hierarchical gradient decay (HGD) skill is developed to resist steganalysis detection. Experiments demonstrate the superiority of our work over existing methods.
|
['Qing Zhou', 'Zhenxing Qian', 'Ge Luo', 'Xinpeng Zhang', 'Sheng Li', 'Ping Wei']
|
2022-07-28
| null | null | null | null |
['steganalysis']
|
['computer-vision']
|
[ 9.37119305e-01 3.81679207e-01 3.12795609e-01 7.18633085e-02
-1.59341052e-01 -4.03993219e-01 4.87903386e-01 -9.22940612e-01
-5.83037660e-02 5.54151058e-01 -1.52137995e-01 -2.81792521e-01
5.11890411e-01 -1.35615361e+00 -6.68485463e-01 -1.12684786e+00
-2.50911146e-01 -3.80497456e-01 1.34043202e-01 -4.77595627e-01
2.20319375e-01 9.87178683e-02 -1.39677477e+00 2.86450684e-01
9.88265216e-01 7.67245889e-01 4.39472169e-01 7.34679818e-01
9.46002752e-02 6.34047270e-01 -7.87881255e-01 -3.14752698e-01
5.47928929e-01 -1.26395726e+00 -3.51857185e-01 4.11618471e-01
-6.06249869e-01 -5.44685781e-01 -5.97356081e-01 1.31563330e+00
4.69557941e-01 -5.31352818e-01 5.36939383e-01 -1.38555562e+00
-8.94600034e-01 7.40299821e-01 -4.77097422e-01 -3.16687018e-01
1.66872859e-01 8.05720448e-01 4.04449552e-01 -4.01053250e-01
5.80174327e-01 1.13393831e+00 2.77366519e-01 8.79265487e-01
-1.01766992e+00 -1.16491818e+00 -4.67286617e-01 -1.43994838e-01
-1.54187012e+00 -4.60902125e-01 1.02548027e+00 -8.98088068e-02
5.65303028e-01 5.25344431e-01 8.23902905e-01 8.36538255e-01
6.48852110e-01 5.50699055e-01 1.41728020e+00 -5.15579104e-01
-1.56346649e-01 4.53163475e-01 -9.76974964e-01 8.50519836e-01
5.53614497e-01 5.97286820e-01 -5.13379909e-02 5.88122830e-02
1.02535403e+00 -2.79361643e-02 -7.23666489e-01 -2.16005504e-01
-1.06896365e+00 1.01686180e+00 5.60937643e-01 4.26121265e-01
-1.28680572e-01 2.96492457e-01 -7.07971379e-02 7.39685237e-01
8.50759149e-02 7.22269043e-02 3.02510440e-01 6.10359311e-01
-5.03431678e-01 -8.20064172e-02 9.06153202e-01 1.07654035e+00
7.54828751e-01 5.62709570e-01 1.70025095e-01 2.45384142e-01
7.27131844e-01 1.04153192e+00 7.12929726e-01 -5.33912718e-01
4.88014311e-01 5.32943249e-01 -2.05837205e-01 -1.57715333e+00
2.53762245e-01 -3.13545167e-01 -1.18497729e+00 4.73122656e-01
-1.89599752e-01 -2.47292221e-01 -1.03650033e+00 1.64969492e+00
1.44557089e-01 1.46131292e-01 5.53288817e-01 6.01208448e-01
7.93391526e-01 1.00613284e+00 -2.70436317e-01 -1.61772534e-01
1.16309798e+00 -5.33321142e-01 -6.86143994e-01 -3.87782604e-01
4.66115773e-01 -8.78157020e-01 3.77098709e-01 4.59870137e-02
-1.12938011e+00 -5.37188411e-01 -1.46225381e+00 5.12906611e-01
-3.89913738e-01 -3.12804729e-01 2.44554028e-01 1.08243978e+00
-1.14247525e+00 2.81893432e-01 -2.89505869e-01 2.24595353e-01
2.89246738e-01 7.16215909e-01 -2.91415304e-01 -3.27854455e-02
-1.71693933e+00 4.75620866e-01 9.22491014e-01 7.22857565e-02
-1.07655346e+00 2.07060799e-01 -1.15380895e+00 -8.23613852e-02
7.39553422e-02 -7.15627074e-01 4.95993972e-01 -1.44997609e+00
-1.52101922e+00 9.73896265e-01 3.71042579e-01 -4.30482090e-01
4.91911322e-01 8.86426866e-01 -8.16609740e-01 2.40112841e-01
-2.35807851e-01 6.00966275e-01 1.49240947e+00 -1.62165546e+00
-5.85327625e-01 3.48199755e-02 -3.28036755e-01 1.61566645e-01
2.29747556e-02 -2.71152675e-01 -1.15076028e-01 -7.63379216e-01
3.57770741e-01 -1.06897700e+00 -1.88709036e-01 -4.45923328e-01
-6.45141065e-01 6.39875889e-01 1.19631982e+00 -7.82821000e-01
1.31400323e+00 -2.20781708e+00 -1.49760425e-01 7.46505320e-01
5.07317483e-01 6.96785569e-01 -1.53243423e-01 6.39101565e-01
-1.20735705e-01 2.86737591e-01 -5.42372465e-01 2.08400697e-01
-1.77637398e-01 1.72584534e-01 -2.11174354e-01 4.31129992e-01
1.35377832e-02 1.21883249e+00 -7.84635723e-01 -4.29094702e-01
6.77288175e-02 6.80492818e-01 -3.63986820e-01 1.69659421e-01
5.14551699e-02 4.07237053e-01 -4.91132110e-01 3.29345018e-01
9.94340837e-01 -4.43674862e-01 4.97863382e-01 5.22329062e-02
9.42212418e-02 -1.71163127e-01 -1.18817341e+00 7.75429070e-01
-8.00129846e-02 3.12486589e-01 1.47940591e-01 -7.33568609e-01
1.27547526e+00 5.08738160e-01 7.47567788e-02 -5.75717330e-01
5.57599664e-01 5.42397022e-01 1.97141007e-01 -5.98088086e-01
2.71067619e-01 -2.11271971e-01 -1.39535546e-01 7.39357650e-01
-3.83402765e-01 -2.47176006e-01 -2.73263395e-01 1.27365470e-01
9.58029389e-01 -1.45294994e-01 2.74857253e-01 -4.24731225e-02
9.07809436e-01 -3.74707133e-01 3.67666513e-01 4.11432147e-01
2.44947791e-01 4.01329190e-01 2.72115946e-01 -6.42181933e-02
-1.37313902e+00 -5.18546999e-01 6.11454785e-01 1.14683554e-01
5.95644057e-01 -1.44686559e-02 -9.53612387e-01 -5.25761724e-01
-3.57658416e-01 2.46008709e-01 -2.77884275e-01 -6.12133920e-01
-7.65837252e-01 -6.45877957e-01 8.03562701e-01 -1.35569260e-01
1.47917771e+00 -1.37551010e+00 -4.88677502e-01 3.29796046e-01
-2.41572052e-01 -6.88272536e-01 -6.56165183e-01 -3.99748802e-01
-6.79139495e-01 -1.06078923e+00 -7.19802260e-01 -1.19834137e+00
1.09150827e+00 5.34712493e-01 5.26062846e-01 5.91606557e-01
-1.07475437e-01 -1.52202725e-01 -4.72365558e-01 -3.12267274e-01
-1.22687376e+00 -1.70332506e-01 -3.23123932e-01 4.05558944e-01
1.26699984e-01 -5.50848722e-01 -8.69446993e-01 5.32324433e-01
-1.62768054e+00 4.25490469e-01 1.03566849e+00 8.41223180e-01
3.26978117e-01 4.82075155e-01 1.80421069e-01 -1.13365781e+00
5.48889935e-01 -5.31089127e-01 -5.81553698e-01 6.98512271e-02
-9.60771024e-01 9.76640582e-02 7.06959307e-01 -3.58837068e-01
-7.09045827e-01 -2.22598299e-01 -6.46932572e-02 -9.58214626e-02
2.34941423e-01 2.50609040e-01 -6.12550676e-01 -6.04199529e-01
3.52862418e-01 1.06188071e+00 6.63126469e-01 -2.45908275e-04
1.05455808e-01 8.81626189e-01 3.32744956e-01 2.51437038e-01
1.42536211e+00 6.45717919e-01 5.37921414e-02 -4.30771947e-01
2.35558897e-01 3.58405858e-01 1.30913600e-01 -6.42553046e-02
6.02455378e-01 -8.07695925e-01 -7.80637681e-01 1.04242861e+00
-9.35356796e-01 -1.16621755e-01 4.21842486e-02 2.37426996e-01
-4.38293666e-01 4.04579103e-01 -4.80461180e-01 -6.29515946e-01
-4.54901427e-01 -1.20440185e+00 5.03290713e-01 2.57027805e-01
3.91924888e-01 -9.78235722e-01 -2.53860891e-01 7.48405755e-02
7.09773779e-01 1.01790023e+00 6.18789971e-01 -1.39989987e-01
-1.05347478e+00 -2.34000340e-01 -2.07955822e-01 6.41773224e-01
4.54189897e-01 -3.30398351e-01 -5.87855577e-01 -7.61783600e-01
5.12376249e-01 1.46741554e-01 8.11915576e-01 8.07107911e-02
6.85132980e-01 -1.13875699e+00 -5.10542333e-01 1.09072423e+00
1.68754780e+00 8.70499134e-01 1.37462759e+00 3.01220119e-01
6.66491568e-01 4.24105883e-01 5.21074906e-02 1.67584628e-01
1.84502453e-01 1.04623638e-01 5.66867352e-01 -3.72872114e-01
-7.61705413e-02 -4.47281659e-01 5.84130943e-01 9.40530777e-01
-2.14312121e-01 -8.05188835e-01 -4.66234058e-01 1.83333263e-01
-1.28790092e+00 -1.14301670e+00 -1.07564524e-01 1.82595122e+00
8.27741981e-01 1.87314689e-01 -2.78070867e-01 5.91184199e-01
1.20946395e+00 5.10444880e-01 -4.45222110e-01 -2.48075262e-01
-2.98520118e-01 2.21470386e-01 9.08969879e-01 4.63002682e-01
-8.09182107e-01 8.18901718e-01 5.64025354e+00 9.77146626e-01
-1.39319026e+00 -6.80321828e-02 5.54134607e-01 6.09555602e-01
-7.45216787e-01 2.19848141e-01 -5.66141903e-01 9.87947285e-01
5.24901569e-01 -1.20065279e-01 4.87786949e-01 4.46590304e-01
-4.49906625e-02 1.66293666e-01 -2.68390447e-01 9.10735667e-01
3.40368152e-01 -1.29507673e+00 3.06204140e-01 5.35985291e-01
8.97750378e-01 -7.00925767e-01 4.64343905e-01 -2.02277824e-01
3.94749641e-01 -1.08804166e+00 3.86930346e-01 3.65064442e-01
1.38770211e+00 -8.63959432e-01 6.56363428e-01 3.58814299e-01
-1.24395812e+00 -1.89820398e-02 -2.81723797e-01 2.62414247e-01
9.50715616e-02 2.17894658e-01 -8.55260670e-01 6.84502184e-01
1.13930792e-01 3.17215532e-01 -1.44458234e-01 3.69505078e-01
-5.90889692e-01 5.42184532e-01 1.85939074e-02 4.14542109e-02
5.72005570e-01 -2.25628123e-01 8.28942358e-01 8.76664639e-01
6.78851306e-01 2.92587727e-01 -1.65455222e-01 9.17162538e-01
-9.67757851e-02 -1.59897923e-01 -1.05043888e+00 -1.54832050e-01
4.52435255e-01 9.22394216e-01 -6.92468226e-01 -3.61097127e-01
2.62323376e-02 1.18096983e+00 -6.58805490e-01 2.87569821e-01
-7.47543991e-01 -1.06363857e+00 1.62248909e-01 4.01119441e-01
5.45541346e-01 -7.97414035e-02 1.79316282e-01 -9.54060674e-01
-2.29233757e-01 -1.29656112e+00 6.20827228e-02 -5.41044652e-01
-6.98834836e-01 7.20461488e-01 -2.79660791e-01 -1.67631078e+00
-3.85019898e-01 -1.45201787e-01 -5.16009510e-01 1.09731770e+00
-1.56704676e+00 -1.32789791e+00 -4.26561087e-01 9.28348422e-01
-1.14186741e-02 -4.93011773e-01 5.46905637e-01 -8.53312165e-02
-1.38786808e-01 5.40951371e-01 1.77074417e-01 4.70660895e-01
1.38817623e-01 -6.14989460e-01 6.47445679e-01 1.08580458e+00
-4.32614744e-01 5.85661590e-01 7.41046906e-01 -1.02188218e+00
-1.31248188e+00 -1.13772190e+00 7.09536493e-01 2.54736513e-01
1.29336521e-01 -3.32923710e-01 -6.21394575e-01 6.75087392e-01
3.02398592e-01 -2.50261575e-01 4.72625345e-01 -1.45913875e+00
-2.45049834e-01 -7.14140944e-03 -1.65395439e+00 5.20227492e-01
8.72603893e-01 -2.66905427e-01 -6.35493398e-02 -1.64926127e-01
7.46816158e-01 -4.24231887e-01 -4.70401645e-01 2.32516155e-01
6.24246478e-01 -1.09061062e+00 8.54574025e-01 2.38045201e-01
5.91274261e-01 -5.36800385e-01 1.23721555e-01 -1.12069273e+00
-3.01080167e-01 -1.19652402e+00 -1.39393369e-02 9.86229837e-01
1.36606887e-01 -1.20276153e+00 6.19560301e-01 6.38731718e-02
2.34508440e-01 -2.87367493e-01 -5.09664476e-01 -9.21017885e-01
-3.77893835e-01 2.58253306e-01 1.31339538e+00 9.37421739e-01
-3.25509369e-01 -1.20710032e-02 -9.17839587e-01 9.99607816e-02
8.44654143e-01 -1.08580209e-01 7.46516109e-01 -6.65532768e-01
-3.14973235e-01 -2.29145080e-01 -5.21224737e-01 -9.37351584e-01
-3.13243270e-01 -9.24343824e-01 -4.72790226e-02 -1.16493678e+00
-1.31731123e-01 -6.80786371e-01 -2.62493989e-03 1.14917047e-01
-3.71548422e-02 7.92758942e-01 1.43191889e-01 5.90684414e-01
6.67498261e-02 2.28902057e-01 1.90017605e+00 -1.07094258e-01
-2.09327847e-01 -2.07647905e-02 -9.11851704e-01 5.12827039e-01
9.61007059e-01 -6.48493111e-01 -5.96461475e-01 -2.26234898e-01
2.01954871e-01 3.10639322e-01 5.23497283e-01 -8.45869720e-01
-3.97281758e-02 -1.03420667e-01 2.84691989e-01 9.15501360e-03
-1.06333261e-02 -9.01914835e-01 8.90271723e-01 1.26816916e+00
-2.80634891e-02 -1.75299272e-01 -5.39809704e-01 4.14224476e-01
-1.94981024e-01 -1.16735153e-01 9.46383357e-01 -5.38842201e-01
-5.57172775e-01 3.21484715e-01 -3.53442818e-01 -4.80006725e-01
1.24338996e+00 -6.84003174e-01 -2.96436757e-01 -5.91543615e-01
-3.93804312e-01 -9.38892141e-02 7.16102362e-01 1.27046272e-01
1.19878995e+00 -1.54315042e+00 -8.36930513e-01 1.02498841e+00
-1.43054426e-01 -2.43345127e-01 2.28238836e-01 4.47115034e-01
-1.01022160e+00 1.56433135e-01 -2.74779856e-01 -1.49904057e-01
-1.22294414e+00 4.80241805e-01 2.25005344e-01 -4.83461797e-01
-4.65130955e-01 7.49333978e-01 1.71691507e-01 1.40506625e-01
-3.25468898e-01 1.51635498e-01 -2.67050415e-01 -3.53708535e-01
6.77686751e-01 2.10012600e-01 -5.78570008e-01 -9.06316698e-01
1.31891683e-01 4.52194512e-01 6.50496334e-02 -1.68278277e-01
8.50323737e-01 -4.82333183e-01 -3.87498468e-01 -3.77135336e-01
1.30709624e+00 7.85930231e-02 -1.06254065e+00 -1.38947919e-01
-7.36884952e-01 -6.64024770e-01 -1.79997444e-01 -3.38781476e-01
-1.37635112e+00 4.06378597e-01 5.59431612e-01 7.10025966e-01
1.30792511e+00 -3.68357867e-01 1.38911700e+00 -1.76320061e-01
6.20070517e-01 -7.28729188e-01 5.58891892e-02 2.20316481e-02
6.04017437e-01 -1.05750763e+00 -2.72987008e-01 -4.62280363e-01
-5.67596614e-01 8.91834795e-01 3.12892616e-01 -3.54998738e-01
5.26345968e-01 2.56060362e-01 -1.38133923e-02 -2.88125724e-01
-3.14911216e-01 -1.15916431e-01 3.21566388e-02 7.92537928e-01
-2.59347856e-01 -3.82671808e-03 -6.13189101e-01 1.06798150e-01
-4.05664355e-01 -7.19259009e-02 5.41348040e-01 1.26526427e+00
-7.62344956e-01 -1.36678648e+00 -7.46002376e-01 7.53775761e-02
-5.50615728e-01 -1.14520237e-01 -1.31537691e-01 8.11379373e-01
6.20473325e-01 9.92863238e-01 -2.38070935e-01 -9.33323443e-01
-2.53158659e-01 -3.62561464e-01 1.96929961e-01 -3.47053617e-01
-4.78942871e-01 1.51155174e-01 -4.00833577e-01 -2.65868306e-01
-5.65232575e-01 3.17361727e-02 -1.02406859e+00 -7.94388473e-01
-4.34927493e-01 3.38882059e-01 6.33640170e-01 6.25222147e-01
1.35021091e-01 5.19480705e-01 1.39255345e+00 -4.42928046e-01
-1.55652910e-01 -5.18412769e-01 -7.96208024e-01 3.25960755e-01
4.32341367e-01 1.20843828e-01 -6.83334768e-01 4.72526550e-01]
|
[4.309024810791016, 8.052433967590332]
|
62168585-4205-4176-b4cf-721c4c69a02a
|
simple-unsupervised-similarity-based-aspect
|
2008.1082
| null |
https://arxiv.org/abs/2008.10820v1
|
https://arxiv.org/pdf/2008.10820v1.pdf
|
Simple Unsupervised Similarity-Based Aspect Extraction
|
In the context of sentiment analysis, there has been growing interest in performing a finer granularity analysis focusing on the specific aspects of the entities being evaluated. This is the goal of Aspect-Based Sentiment Analysis (ABSA) which basically involves two tasks: aspect extraction and polarity detection. The first task is responsible for discovering the aspects mentioned in the review text and the second task assigns a sentiment orientation (positive, negative, or neutral) to that aspect. Currently, the state-of-the-art in ABSA consists of the application of deep learning methods such as recurrent, convolutional and attention neural networks. The limitation of these techniques is that they require a lot of training data and are computationally expensive. In this paper, we propose a simple approach called SUAEx for aspect extraction. SUAEx is unsupervised and relies solely on the similarity of word embeddings. Experimental results on datasets from three different domains have shown that SUAEx achieves results that can outperform the state-of-the-art attention-based approach at a fraction of the time.
|
['Danny Suarez Vargas', 'Viviane Pereira Moreira', 'Lucas R. C. Pessutto']
|
2020-08-25
| null | null | null | null |
['aspect-extraction']
|
['natural-language-processing']
|
[ 4.96778004e-02 1.31014615e-01 -3.09462011e-01 -3.63010556e-01
-5.61959028e-01 -5.37801445e-01 9.59530652e-01 8.06622088e-01
-4.23720807e-01 4.01340276e-01 3.46490592e-01 -3.83060515e-01
7.04254676e-03 -9.19449747e-01 -2.97069669e-01 -5.35925865e-01
3.31719130e-01 4.36570704e-01 1.15940869e-02 -5.68663180e-01
5.13233840e-01 3.26431811e-01 -1.62025821e+00 3.74833167e-01
4.39530611e-01 1.24711668e+00 -3.29469800e-01 3.92673790e-01
-6.76137567e-01 8.32714915e-01 -6.68102264e-01 -6.97804987e-01
-1.80376321e-01 -3.85943085e-01 -8.58154237e-01 -5.65804504e-02
-1.06308505e-01 2.96632200e-01 2.99851000e-01 1.02272809e+00
3.58880997e-01 3.45510840e-02 6.24114692e-01 -9.67930675e-01
-5.84340990e-01 6.27951324e-01 -5.01291811e-01 1.80140778e-01
1.71627715e-01 -4.12886858e-01 1.51997256e+00 -1.02728248e+00
4.66735333e-01 8.94144475e-01 5.38582861e-01 1.82604998e-01
-7.39832461e-01 -2.62046814e-01 4.67036039e-01 1.64942086e-01
-8.86499763e-01 -4.22336161e-02 1.00941694e+00 -3.68446559e-01
1.35258055e+00 9.66621935e-02 9.13895309e-01 5.84057212e-01
1.30249813e-01 9.72359180e-01 1.17311120e+00 -5.50188005e-01
6.13525808e-01 4.53833073e-01 7.70312488e-01 2.31505305e-01
5.06693661e-01 -6.32430255e-01 -5.11328816e-01 -1.13380119e-01
-1.48217201e-01 -2.95809451e-02 -1.19500637e-01 -3.83384079e-01
-9.86562431e-01 1.07366204e+00 2.46869326e-01 6.52739167e-01
-8.03600430e-01 -2.00244278e-01 7.02332854e-01 1.98628753e-01
9.63220835e-01 7.32574224e-01 -8.40081096e-01 -3.25658441e-01
-8.41625690e-01 2.62010217e-01 1.04213059e+00 5.21476865e-01
7.38909662e-01 -1.16009019e-01 -9.54040959e-02 5.51891148e-01
3.37066650e-01 2.45700419e-01 6.74078763e-01 -1.45342767e-01
4.09011334e-01 1.29777849e+00 -7.03965640e-03 -1.03546703e+00
-3.91177535e-01 -5.08349776e-01 -4.56559747e-01 3.23731154e-01
-1.04298770e-01 -3.31000894e-01 -1.06610370e+00 1.24843395e+00
3.63417506e-01 -3.41586322e-01 2.31144726e-01 6.59453273e-01
8.97472918e-01 6.91243947e-01 7.62993246e-02 -8.39126855e-03
1.74334812e+00 -9.60205436e-01 -9.09735858e-01 -4.60384786e-01
4.71648932e-01 -8.72732222e-01 1.02837539e+00 3.13281357e-01
-9.04328823e-01 -1.64957419e-01 -1.43130577e+00 -2.08173506e-02
-1.14399254e+00 2.04346687e-01 8.86902750e-01 6.65662408e-01
-9.47607100e-01 3.11188966e-01 -7.15323627e-01 -3.33754182e-01
4.60617721e-01 4.27929908e-01 -4.40727115e-01 2.92088896e-01
-1.21691334e+00 8.24738801e-01 1.20012619e-01 4.65924665e-02
-1.44516245e-01 -5.37571430e-01 -1.10872924e+00 4.17551100e-01
1.50241032e-01 -4.19340491e-01 1.22238076e+00 -1.35387719e+00
-1.52744365e+00 8.97478104e-01 -2.32454225e-01 -4.87748533e-01
-1.57227188e-01 -5.24486184e-01 -4.43747401e-01 3.82201932e-02
2.44848028e-01 2.87969083e-01 7.48814166e-01 -9.34064984e-01
-7.87202239e-01 -6.34972572e-01 4.10315633e-01 2.46787086e-01
-5.49088776e-01 3.59615088e-01 -5.08285820e-01 -6.08271539e-01
-3.31195921e-01 -9.74727511e-01 -1.87176123e-01 -5.10260046e-01
-3.10679853e-01 -4.86019105e-01 9.66844618e-01 -3.56435269e-01
1.36870849e+00 -2.02370167e+00 7.75349513e-02 1.41712308e-01
1.06235109e-01 6.47839248e-01 8.90905857e-02 4.55762744e-01
-3.77172917e-01 1.42872423e-01 -2.92830706e-01 -4.07861054e-01
3.49669084e-02 -2.57306606e-01 -4.59559947e-01 1.88747734e-01
5.34385741e-01 8.31250012e-01 -8.01217914e-01 -6.69864491e-02
4.69009299e-03 6.92238688e-01 -3.55777472e-01 1.87724367e-01
-2.50165522e-01 -9.94702503e-02 -4.95699495e-01 6.04048312e-01
4.18128461e-01 -2.98993111e-01 1.62043020e-01 -2.16851071e-01
-3.10318142e-01 9.08104777e-01 -8.48979771e-01 1.23808122e+00
-6.65730417e-01 9.81874824e-01 -2.90786624e-01 -1.08949471e+00
1.00701237e+00 3.63566905e-01 3.09042335e-01 -5.12596250e-01
4.76558089e-01 2.24057347e-01 7.18882754e-02 -2.25610793e-01
8.98855686e-01 -1.58493444e-01 -2.71422118e-01 6.92609429e-01
2.09064469e-01 -1.81638803e-02 5.66479743e-01 8.48672315e-02
8.93296599e-01 -5.15841171e-02 7.61168838e-01 -3.95847708e-01
8.22930515e-01 4.15290184e-02 3.42134744e-01 1.88959137e-01
-4.13218662e-02 5.14232695e-01 1.03236449e+00 -5.47905743e-01
-9.03050959e-01 -4.52358961e-01 1.09865092e-01 9.77696836e-01
-1.29989877e-01 -7.42829978e-01 -7.50696719e-01 -8.91722322e-01
-2.76536673e-01 7.25452602e-01 -1.21604800e+00 4.32781037e-03
-4.08105642e-01 -8.73422623e-01 -6.78068325e-02 6.37904823e-01
4.31981951e-01 -1.45632279e+00 -7.38574266e-01 2.24872902e-01
1.09915145e-01 -1.09889197e+00 5.35228010e-03 3.85982782e-01
-6.19462132e-01 -9.21806872e-01 -7.03586698e-01 -5.30199289e-01
5.90344429e-01 6.87941760e-02 1.33398795e+00 -2.42971659e-01
-1.83341280e-02 2.24940091e-01 -7.16301382e-01 -9.13417637e-01
7.12283775e-02 6.64052546e-01 -3.23321313e-01 4.49722499e-01
1.03151500e+00 -4.97481287e-01 -6.22416735e-01 -2.50351965e-01
-9.04876292e-01 -4.32285100e-01 9.29104328e-01 4.69717950e-01
6.17113709e-01 -1.07159682e-01 5.34434855e-01 -1.31009376e+00
9.85937595e-01 -4.74488676e-01 -4.18667823e-01 5.76198548e-02
-7.28272557e-01 9.04485136e-02 5.38489044e-01 -1.15585968e-01
-9.99919593e-01 -2.32332587e-01 -1.92542151e-01 1.69503450e-01
-1.42408758e-01 1.06469774e+00 -1.27730027e-01 3.96033019e-01
2.18418062e-01 1.16205059e-01 -2.71329701e-01 -2.92354971e-01
1.82608217e-01 7.52760470e-01 -7.65680596e-02 1.09062567e-01
3.13174725e-01 5.90782762e-01 -1.40472902e-02 -9.46343303e-01
-1.32320023e+00 -7.94133425e-01 -4.34192806e-01 -1.91155598e-01
9.71527815e-01 -8.19309354e-01 -3.89078915e-01 3.59692693e-01
-1.04009652e+00 2.13331103e-01 -4.38893020e-01 3.59931558e-01
-1.96125597e-01 1.66517541e-01 -3.34096909e-01 -6.72729135e-01
-9.39721107e-01 -1.16256118e+00 1.08734751e+00 3.55102479e-01
-6.87874317e-01 -8.89288902e-01 3.88871640e-01 2.73275107e-01
6.28105402e-01 -5.31161614e-02 1.04875743e+00 -1.14785314e+00
-2.47044519e-01 -6.28466427e-01 -1.07424617e-01 3.29660147e-01
2.81133354e-01 -4.68379371e-02 -1.11144888e+00 1.41384214e-01
9.38609317e-02 -1.03796922e-01 8.49952757e-01 2.78679341e-01
6.28638804e-01 -8.88459980e-02 6.14326587e-03 1.36978313e-01
1.33152664e+00 1.19405247e-01 5.70617318e-01 7.78150022e-01
5.51672280e-01 7.92951167e-01 9.38116908e-01 3.00616443e-01
3.78319442e-01 5.11363149e-01 4.07423347e-01 -1.14336044e-01
8.18004180e-03 1.74946934e-01 4.60337698e-01 8.85755002e-01
-2.27610841e-02 -1.72375247e-01 -7.85649538e-01 1.10674417e+00
-1.87293208e+00 -7.36594081e-01 -1.25103518e-01 1.87180364e+00
6.20521128e-01 4.56550777e-01 -1.99348871e-02 5.00392437e-01
4.29177016e-01 5.38807034e-01 -2.11726904e-01 -1.12198138e+00
1.06875028e-03 4.28453356e-01 7.97507688e-02 3.25560123e-01
-1.37584925e+00 8.23234975e-01 5.25012207e+00 5.02351344e-01
-1.34259093e+00 -9.88239273e-02 6.92347229e-01 1.45312101e-01
-3.22771937e-01 -1.15900598e-01 -8.39129984e-01 1.10646054e-01
9.95060444e-01 -1.25230014e-01 -1.86305881e-01 1.17724252e+00
-1.44121170e-01 -1.83260113e-01 -6.85989738e-01 5.83140194e-01
4.33924228e-01 -1.12415779e+00 2.17401698e-01 -1.51208565e-01
8.18767548e-01 1.63572550e-01 1.11619331e-01 3.66231620e-01
-1.57205224e-01 -7.11246610e-01 4.53220218e-01 2.43017152e-01
3.11698884e-01 -1.05611312e+00 1.32848191e+00 -2.39949182e-01
-1.17656660e+00 1.94026172e-01 -1.42695650e-01 -1.12421706e-01
1.08883321e-01 9.39782679e-01 -5.50953269e-01 4.99288112e-01
8.39092672e-01 8.88924897e-01 -4.57842231e-01 6.93729937e-01
-6.48880363e-01 6.35858476e-01 1.73705560e-03 -5.81566274e-01
4.84507769e-01 -2.20226526e-01 5.26085794e-01 1.35427713e+00
1.41984239e-01 -2.74827421e-01 -2.02479243e-01 4.72795576e-01
-2.13539913e-01 4.41456199e-01 -5.59813678e-01 -4.16042298e-01
-1.68281153e-01 1.71143794e+00 -9.86552477e-01 -4.91669714e-01
-6.55004442e-01 6.50596380e-01 3.32219929e-01 6.14024512e-02
-3.80293041e-01 -8.09263468e-01 9.01603937e-01 -8.09711814e-02
1.03932536e+00 1.20825984e-01 -4.59218562e-01 -1.06641436e+00
2.03175038e-01 -8.83748412e-01 2.38364175e-01 -5.40859103e-01
-9.84195352e-01 1.18647146e+00 -4.20363814e-01 -1.17802668e+00
-4.55390036e-01 -7.91360855e-01 -7.87961960e-01 8.17445874e-01
-1.86548543e+00 -1.19731462e+00 -1.73612326e-01 2.50069350e-01
6.22350395e-01 -2.16107622e-01 9.56945777e-01 1.97892711e-01
-2.85029560e-01 1.81499913e-01 -2.76496500e-01 1.07289627e-01
4.64535236e-01 -1.44047296e+00 5.65333128e-01 7.45037913e-01
6.62959516e-02 5.88937938e-01 8.37306082e-01 -3.87867749e-01
-1.02516532e+00 -9.26812470e-01 1.68852091e+00 -4.00656164e-01
9.81475830e-01 -2.33277544e-01 -7.04891324e-01 5.22118092e-01
6.53026998e-01 -1.85519174e-01 1.07307935e+00 5.27928591e-01
-3.15374136e-01 -1.88490450e-01 -6.50578678e-01 6.14253640e-01
1.54235423e-01 -6.01027250e-01 -8.38187277e-01 -2.15017214e-01
5.42153001e-01 5.72280288e-02 -6.40712798e-01 5.12915730e-01
4.73094046e-01 -9.56027269e-01 5.89460552e-01 -5.50536096e-01
6.95195377e-01 -4.45670903e-01 1.90865900e-02 -1.40788698e+00
-1.87415835e-02 -1.90626875e-01 -1.26925603e-01 1.43475616e+00
8.39090586e-01 -5.13848603e-01 6.58848703e-01 3.80646557e-01
1.22955874e-01 -1.20369208e+00 -4.63977754e-01 -1.52067676e-01
-2.49203593e-01 -4.65473711e-01 6.19152784e-01 7.82265782e-01
1.83879003e-01 9.99971509e-01 1.06132381e-01 -2.93484926e-02
3.31836194e-02 7.33327270e-01 5.56232333e-01 -1.18675923e+00
-5.75822368e-02 -6.86282039e-01 -5.57341218e-01 -3.91989917e-01
1.68305606e-01 -4.23228145e-01 -1.91532239e-01 -1.78340042e+00
1.21031977e-01 1.61262780e-01 -5.86707592e-01 3.34787637e-01
-4.52901572e-01 1.81938589e-01 6.56094104e-02 -1.62498951e-01
-7.92015612e-01 6.89287663e-01 6.13931239e-01 -3.29823792e-01
-2.07876101e-01 2.55942971e-01 -1.01952338e+00 9.70068991e-01
1.00394511e+00 -4.61848468e-01 -3.53678167e-01 -1.35783479e-01
9.10087168e-01 -6.34610236e-01 -2.33675107e-01 -7.25711763e-01
1.25650436e-01 4.60399538e-01 -3.10475789e-02 -8.14570487e-01
3.78340304e-01 -7.37177968e-01 -5.87583780e-01 2.36822084e-01
-3.76288891e-01 3.85795355e-01 2.71889985e-01 3.08788836e-01
-8.04757237e-01 -2.55218297e-01 3.74531746e-01 9.59952772e-02
-6.25732720e-01 5.43333357e-03 -5.49832106e-01 4.48775403e-02
8.35861504e-01 2.89857030e-01 -6.83833063e-02 -4.09894675e-01
-4.43465859e-01 -2.41109908e-01 8.66652131e-02 5.25126755e-01
4.98011500e-01 -1.02831650e+00 -4.79849577e-01 -2.16902699e-02
4.85426843e-01 -1.98647842e-01 -1.03100101e-02 8.07405889e-01
-2.53015429e-01 7.08730102e-01 4.46091592e-02 -1.69270694e-01
-1.44159055e+00 6.16559446e-01 6.06123395e-02 -8.94585907e-01
-2.77125180e-01 5.60638487e-01 1.26433477e-01 -4.99371797e-01
-1.62293360e-01 -2.83241093e-01 -1.12984586e+00 6.69927776e-01
6.64904833e-01 5.14787845e-02 5.40044904e-01 -7.45545089e-01
-5.22663355e-01 6.91534996e-01 -3.76918942e-01 -1.68147773e-01
1.57621276e+00 1.57691553e-01 -3.94893616e-01 5.90266705e-01
1.13675988e+00 1.83801621e-01 -4.69627082e-01 3.61790620e-02
2.09645957e-01 5.98226814e-03 3.07641238e-01 -8.56634915e-01
-1.08303988e+00 1.02348363e+00 2.60053694e-01 6.49995804e-01
1.15315652e+00 -2.16696155e-03 6.20403171e-01 4.98146564e-01
-8.83379355e-02 -1.21495068e+00 -2.84175068e-01 7.71716237e-01
7.25268543e-01 -1.22444177e+00 1.35247141e-01 -2.26705357e-01
-8.47712219e-01 9.27743495e-01 3.62064302e-01 -2.79406339e-01
9.49120820e-01 8.12686607e-02 3.85768920e-01 -5.86002231e-01
-8.14467728e-01 -5.02127886e-01 5.22749603e-01 1.91530526e-01
8.48929167e-01 -5.93312606e-02 -7.50460923e-01 7.88250983e-01
-1.73839882e-01 -2.31945962e-01 4.56055313e-01 1.08173823e+00
-2.22171187e-01 -1.09738612e+00 7.93325529e-03 4.12261516e-01
-1.19993615e+00 -2.86381483e-01 -7.27951467e-01 7.79160619e-01
-2.76395589e-01 9.59203482e-01 -1.50253445e-01 -7.71744624e-02
4.65749949e-01 3.32606919e-02 -1.05659500e-01 -6.58350527e-01
-1.02157795e+00 4.80333194e-02 3.77049297e-01 -4.84735489e-01
-8.71004522e-01 -6.38677537e-01 -1.11145711e+00 1.29262477e-01
-2.75861531e-01 4.99020696e-01 1.09258175e+00 1.19527638e+00
6.29853725e-01 8.07368875e-01 6.14487469e-01 -6.35825515e-01
-1.47118028e-02 -1.04706800e+00 -4.75192815e-01 3.88942331e-01
3.59989405e-01 -4.32144731e-01 -2.20503584e-01 -3.07175338e-01]
|
[11.345892906188965, 6.750380516052246]
|
d5d151da-4ad7-4aae-8239-11e88069b41e
|
uniform-pac-guarantees-for-model-based-rl
|
2305.0835
| null |
https://arxiv.org/abs/2305.08350v1
|
https://arxiv.org/pdf/2305.08350v1.pdf
|
Uniform-PAC Guarantees for Model-Based RL with Bounded Eluder Dimension
|
Recently, there has been remarkable progress in reinforcement learning (RL) with general function approximation. However, all these works only provide regret or sample complexity guarantees. It is still an open question if one can achieve stronger performance guarantees, i.e., the uniform probably approximate correctness (Uniform-PAC) guarantee that can imply both a sub-linear regret bound and a polynomial sample complexity for any target learning accuracy. We study this problem by proposing algorithms for both nonlinear bandits and model-based episodic RL using the general function class with a bounded eluder dimension. The key idea of the proposed algorithms is to assign each action to different levels according to its width with respect to the confidence set. The achieved uniform-PAC sample complexity is tight in the sense that it matches the state-of-the-art regret bounds or sample complexity guarantees when reduced to the linear case. To the best of our knowledge, this is the first work for uniform-PAC guarantees on bandit and RL that goes beyond linear cases.
|
['Quanquan Gu', 'Jiafan He', 'Yue Wu']
|
2023-05-15
| null | null | null | null |
['open-question']
|
['natural-language-processing']
|
[-3.93303344e-03 4.19050992e-01 -6.93803132e-01 -2.36827046e-01
-1.15893781e+00 -6.80652678e-01 6.82248026e-02 3.53359401e-01
-4.82645929e-01 1.43044829e+00 -1.88633502e-01 -4.84163612e-01
-6.86337471e-01 -8.99143279e-01 -1.09624457e+00 -9.89015460e-01
-1.46612868e-01 7.70127535e-01 2.02469438e-01 -3.48604210e-02
2.41220400e-01 4.07048494e-01 -1.51855707e+00 -1.51277659e-02
1.01967490e+00 1.51455092e+00 -1.28440738e-01 5.78184366e-01
-4.59087593e-03 5.75635135e-01 -4.92505252e-01 -4.93389219e-01
5.08062124e-01 -6.52364612e-01 -7.85106599e-01 -1.94183186e-01
1.27712935e-01 -5.58515072e-01 -9.79140215e-03 1.05067122e+00
4.13832128e-01 1.52203292e-01 3.38707954e-01 -1.18865335e+00
-4.76235777e-01 1.06870222e+00 -5.35688818e-01 -4.51764651e-02
8.48434791e-02 -4.02767062e-01 1.18343318e+00 1.65479243e-01
2.82388806e-01 1.15297687e+00 5.40479362e-01 6.45261109e-01
-1.31579578e+00 -6.29419506e-01 3.66493762e-01 1.46275580e-01
-8.29424322e-01 -3.21626328e-02 3.22905511e-01 -1.49530560e-01
4.81610924e-01 4.43844825e-01 7.71982372e-01 6.17529035e-01
-2.94818103e-01 1.17973328e+00 1.34584200e+00 -4.80734944e-01
7.75541246e-01 3.33816290e-01 7.11362958e-02 5.65907598e-01
5.68981111e-01 2.20931411e-01 -2.38210008e-01 -2.40642622e-01
7.90306270e-01 -1.70072764e-01 -4.45206583e-01 -6.73880994e-01
-5.81560850e-01 1.02131188e+00 2.79237330e-01 2.86515772e-01
-3.14880282e-01 5.22524714e-01 2.27898985e-01 7.84878016e-01
6.09370112e-01 3.54319036e-01 -7.24544764e-01 -2.75330484e-01
-1.01969516e+00 5.43031335e-01 1.30025673e+00 8.74235570e-01
5.39774477e-01 5.33633120e-02 -1.70635179e-01 5.20708084e-01
-1.45594338e-02 5.07659912e-01 2.30674803e-01 -9.20350552e-01
4.75070268e-01 4.77917604e-02 8.37784410e-01 -3.96611691e-01
-2.40881547e-01 -1.03867269e+00 -3.51870745e-01 3.51607412e-01
9.31973636e-01 -2.37093210e-01 -1.51457369e-01 2.02344584e+00
6.13258481e-01 3.00969630e-02 1.97094604e-01 8.38782370e-01
-8.78670663e-02 5.81757247e-01 -5.18221974e-01 -8.13162267e-01
8.78163338e-01 -9.31484580e-01 -5.92174768e-01 5.16132042e-02
6.21204138e-01 -3.14923376e-01 9.24492717e-01 6.90756559e-01
-1.14248502e+00 7.32578784e-02 -1.08552778e+00 3.66367638e-01
8.67829621e-02 -8.19264203e-02 8.42164874e-01 1.21294892e+00
-6.29949510e-01 8.46558750e-01 -8.74245942e-01 -1.94553673e-01
3.66941720e-01 2.51259029e-01 -2.98205633e-02 -1.02459185e-01
-7.58267939e-01 5.81855714e-01 4.15222555e-01 -1.62287176e-01
-8.29260886e-01 -8.89964402e-01 -3.90196264e-01 1.14425205e-01
9.08387899e-01 -5.38687766e-01 1.68649292e+00 -1.26768279e+00
-1.80849922e+00 3.62342685e-01 1.08619750e-01 -1.10850739e+00
1.01391053e+00 -3.91561687e-01 2.22270802e-01 -3.80773507e-02
-3.41186762e-01 -6.39176071e-02 6.65374756e-01 -1.15219247e+00
-9.49532211e-01 -6.86650872e-01 6.83514178e-01 1.96479276e-01
-2.76782244e-01 -5.54890811e-01 2.07040027e-01 -3.32401901e-01
-7.40636364e-02 -8.10914040e-01 -1.61009401e-01 9.81627107e-02
5.19212186e-02 -3.97848308e-01 2.04761729e-01 -4.24962521e-01
1.10694289e+00 -1.89505720e+00 1.29136279e-01 6.44102395e-02
-4.77027804e-01 -3.73165519e-03 1.15904726e-01 5.23552895e-01
3.45468789e-01 -6.02925569e-02 -1.51683297e-03 -1.42225191e-01
4.94206488e-01 4.07431960e-01 -7.15907991e-01 8.15660238e-01
-5.20940006e-01 5.74394464e-01 -9.15468633e-01 -2.03426206e-03
-2.80435421e-02 -1.04828142e-01 -8.31687391e-01 1.30990759e-01
-6.33753598e-01 8.99779722e-02 -6.21036053e-01 2.65137434e-01
6.68479443e-01 -8.52315575e-02 3.47304165e-01 1.92980111e-01
-8.38310272e-02 1.78420708e-01 -1.35697067e+00 1.35451043e+00
-4.90694374e-01 2.16839164e-02 3.52158934e-01 -1.44939530e+00
6.79622114e-01 1.72051221e-01 3.70202899e-01 -3.26358378e-01
-1.18382350e-02 6.19411647e-01 -2.48686731e-01 -1.75173834e-01
2.72685707e-01 -5.57043135e-01 -2.91681644e-02 3.83105993e-01
-1.00295365e-01 1.22627430e-01 1.46825835e-01 -4.42175657e-01
8.53157222e-01 3.97042096e-01 2.68358350e-01 -4.87072617e-01
3.90777826e-01 -3.21734011e-01 4.33200240e-01 1.24572885e+00
7.25601017e-02 -3.19380965e-03 8.22175860e-01 -2.49917015e-01
-9.51397181e-01 -7.95201302e-01 -2.11245179e-01 1.26722777e+00
1.07331216e-01 4.45742421e-02 -6.92812026e-01 -6.01564169e-01
4.28203434e-01 9.25121605e-01 -8.74183714e-01 5.41511998e-02
-1.27873480e-01 -5.74209332e-01 3.53534162e-01 2.57277578e-01
4.79361802e-01 -5.78132272e-01 -7.31292248e-01 3.30130547e-01
2.49039158e-01 -8.17093134e-01 -1.83565930e-01 3.01855475e-01
-1.08011210e+00 -9.51509833e-01 -7.55923629e-01 -3.96376848e-01
2.64354020e-01 -1.67809695e-01 6.55054510e-01 -3.28570276e-01
3.06581259e-01 5.20936012e-01 -4.42129463e-01 -4.50727701e-01
-1.16652377e-01 9.26363692e-02 -1.76043928e-01 3.75940092e-02
-1.89331397e-01 -4.68482047e-01 -7.31906891e-01 1.54861420e-01
-8.94188285e-01 -6.38683066e-02 5.04119098e-01 9.82836723e-01
7.84114301e-01 2.31864363e-01 6.35044932e-01 -8.72347057e-01
4.86646116e-01 -3.99032354e-01 -1.11427259e+00 4.67881083e-01
-8.25049460e-01 5.78195035e-01 9.34456468e-01 -3.96724552e-01
-8.90504122e-01 -1.77329153e-01 -3.31651382e-02 -6.25685975e-02
1.87697336e-01 4.20564264e-01 1.59019887e-01 -7.55477883e-03
4.77399200e-01 5.49587786e-01 -1.54022053e-01 -5.48528731e-01
2.93550760e-01 5.30603290e-01 2.05795050e-01 -1.01689506e+00
4.79374617e-01 5.23359954e-01 3.55689108e-01 -5.25141478e-01
-1.48552716e+00 -2.27011457e-01 1.16574831e-01 -3.75565104e-02
2.56595820e-01 -4.15423453e-01 -1.13621962e+00 5.85231818e-02
-5.29700041e-01 -6.36724889e-01 -6.62338853e-01 5.94062626e-01
-1.26987231e+00 3.36809605e-01 -3.70677561e-01 -1.79803872e+00
-3.66617471e-01 -6.50510430e-01 6.88185096e-01 1.34811684e-01
5.08203864e-01 -6.33271754e-01 1.09099537e-01 2.29075894e-01
3.12368631e-01 3.64253312e-01 5.80704093e-01 -5.54660857e-01
-4.78675544e-01 3.14510725e-02 4.46199477e-02 3.30300212e-01
-1.93436906e-01 -5.66131890e-01 -5.36386073e-01 -6.76805019e-01
1.94139600e-01 -6.15297854e-01 9.17367518e-01 5.99719644e-01
1.14143598e+00 -8.47875655e-01 -4.32355739e-02 6.13726974e-01
1.61007154e+00 2.25107431e-01 4.30656105e-01 5.52943885e-01
-1.81609362e-01 2.62650430e-01 1.09873855e+00 1.06211042e+00
-2.49603810e-03 5.95273435e-01 5.37299275e-01 5.96996069e-01
4.85007763e-01 -3.49913269e-01 3.22655767e-01 -5.67852668e-02
-2.45611966e-02 -1.04072101e-01 -1.75118282e-01 5.27432144e-01
-2.37400675e+00 -1.09858608e+00 2.93699175e-01 2.76940560e+00
1.06606662e+00 2.05883667e-01 4.87932235e-01 2.95787275e-01
6.44351959e-01 -7.36973882e-02 -9.83903825e-01 -5.57280898e-01
1.43229559e-01 2.09132284e-01 8.70883644e-01 6.28067911e-01
-8.03214431e-01 6.25508904e-01 5.44563627e+00 1.14738870e+00
-8.53455842e-01 2.59692073e-01 6.29791200e-01 -5.25590897e-01
-2.47930229e-01 9.74080786e-02 -8.52245331e-01 4.34672564e-01
8.41844141e-01 -3.66947085e-01 1.06333160e+00 1.32773221e+00
-6.69582412e-02 -1.84190333e-01 -9.42619681e-01 9.03649509e-01
-2.16207087e-01 -1.11189580e+00 -3.54321271e-01 2.06775293e-01
8.59408081e-01 -3.39871347e-01 1.05390884e-01 5.44385433e-01
3.02945465e-01 -6.89212263e-01 7.48021126e-01 5.40077448e-01
5.37154853e-01 -1.37992263e+00 7.15459466e-01 7.44324684e-01
-7.45463371e-01 -5.42496562e-01 -6.17596805e-01 -1.76331311e-01
-7.16321170e-02 5.10453463e-01 -4.42493170e-01 7.56190956e-01
4.80568409e-01 2.54826456e-01 3.55892628e-01 1.40920794e+00
-9.36328322e-02 7.46993840e-01 -7.21181214e-01 -5.31021237e-01
5.02600849e-01 -3.67656261e-01 3.26207101e-01 7.68380940e-01
5.62277675e-01 -4.13165949e-02 1.71969265e-01 4.93940830e-01
1.12113766e-02 3.26046556e-01 -9.39852223e-02 -1.31475732e-01
4.46030825e-01 9.16579247e-01 -5.28545022e-01 -1.13761194e-01
-1.55184329e-01 8.93382967e-01 5.66731453e-01 9.01806578e-02
-8.78068984e-01 -2.20798939e-01 5.48333287e-01 -3.00325789e-02
8.86043549e-01 -2.87664775e-02 -8.42477381e-02 -9.53935206e-01
3.02789211e-01 -4.86363918e-01 7.94466674e-01 -7.90588334e-02
-1.24222755e+00 1.29437804e-01 -9.90680233e-02 -8.28755379e-01
-2.12287635e-01 -5.20648062e-01 1.00853220e-01 3.96306247e-01
-1.59906578e+00 -7.68306077e-01 2.60736883e-01 4.23797280e-01
3.07305932e-01 1.06358286e-02 7.62966931e-01 -5.50548881e-02
-1.51032418e-01 8.85323763e-01 9.68396068e-01 -6.17899120e-01
3.86097848e-01 -1.55380905e+00 -7.35805273e-01 4.23579425e-01
-1.33202016e-01 8.79954919e-02 1.08375633e+00 -1.80984482e-01
-1.59695041e+00 -7.91335881e-01 2.64260769e-01 7.84323812e-02
8.01193833e-01 -1.71905249e-01 -4.07961220e-01 6.71731114e-01
-1.82057917e-01 -3.79760601e-02 5.27084172e-01 3.80704850e-01
-2.61276007e-01 -6.94808006e-01 -1.50991118e+00 3.28260958e-01
1.01513207e+00 5.08821048e-02 -3.62494946e-01 4.29569483e-01
8.04259479e-01 -6.03756845e-01 -1.06468773e+00 2.79827297e-01
8.16108167e-01 -1.03025186e+00 6.55645013e-01 -6.25076830e-01
2.06183009e-02 -1.04209043e-01 -3.18265796e-01 -1.16547585e+00
-1.55387667e-03 -9.33975458e-01 -5.47216952e-01 9.72859979e-01
2.71687746e-01 -8.53014469e-01 9.40447330e-01 3.92690688e-01
1.37265995e-01 -1.18713939e+00 -1.38615775e+00 -1.27262604e+00
3.73301774e-01 -5.38112879e-01 7.09859490e-01 5.19065797e-01
1.45384446e-01 -2.67831068e-02 -7.61905611e-01 9.71069559e-02
7.05033839e-01 8.66054177e-01 6.65405452e-01 -1.19776988e+00
-1.16334021e+00 -5.81519365e-01 -1.01545326e-01 -1.46366882e+00
1.22480974e-01 -5.93696654e-01 -1.41455650e-01 -1.36216795e+00
7.72117004e-02 -7.08777905e-01 -3.19359124e-01 2.50523388e-01
2.72500247e-01 -2.93966889e-01 9.16393101e-02 -1.21976592e-01
-9.19774354e-01 7.89956808e-01 1.31991160e+00 1.32573217e-01
-2.89070994e-01 6.18812621e-01 -7.58806765e-01 3.42444390e-01
1.12320900e+00 -4.94946957e-01 -5.55087090e-01 -7.52735734e-02
4.51578468e-01 5.27806401e-01 1.90495397e-03 -9.62971926e-01
-2.67798062e-02 -4.34705764e-01 -2.40695290e-02 -4.57264543e-01
1.71811908e-01 -9.21093822e-01 8.43755156e-02 8.60747814e-01
-6.88127160e-01 -5.16781807e-01 1.14730440e-01 9.44497228e-01
3.34736258e-01 -5.65928042e-01 8.91593158e-01 3.74549143e-02
1.40380248e-01 2.74546653e-01 -8.44327956e-02 1.66886643e-01
1.16651261e+00 9.44644138e-02 -2.27734923e-01 -7.29179382e-01
-5.73981762e-01 2.30267316e-01 1.14513941e-01 -1.43729493e-01
2.01824337e-01 -1.18341088e+00 -6.40690982e-01 -3.34937662e-01
-2.20696762e-01 -3.84811938e-01 2.65329838e-01 8.91390324e-01
-2.07339525e-01 6.49580240e-01 7.37356991e-02 -2.38293439e-01
-1.07431948e+00 8.72980058e-01 5.37782073e-01 -5.36493778e-01
-4.76626009e-01 6.01340115e-01 -1.01237439e-01 -2.34242231e-01
5.74200511e-01 -5.11938214e-01 2.96076536e-01 -2.18504086e-01
6.52062237e-01 5.23060381e-01 -2.21877411e-01 2.04169333e-01
-3.59461829e-03 1.93971112e-01 4.06591669e-02 -3.13531339e-01
1.48436666e+00 3.68254893e-02 7.57379904e-02 5.60063004e-01
7.97330499e-01 -9.76970345e-02 -1.51312673e+00 -4.37869012e-01
-8.59924108e-02 -5.25541067e-01 6.70939088e-02 -9.51746821e-01
-9.26812887e-01 4.94520217e-01 6.66228652e-01 6.80587232e-01
1.14911687e+00 5.63225783e-02 6.00608349e-01 7.68217266e-01
8.83769155e-01 -1.23106003e+00 -3.79648209e-01 3.14769059e-01
9.17961359e-01 -6.92075610e-01 5.39267249e-03 2.81004813e-02
-2.79267073e-01 1.22015250e+00 3.36473733e-01 -4.19766635e-01
3.91931653e-01 2.36661628e-01 -7.26852596e-01 4.64643300e-01
-8.41258168e-01 -4.63797331e-01 -7.25415861e-03 1.81400493e-01
1.65278301e-01 3.12164396e-01 -1.09893250e+00 9.13699746e-01
-4.06204611e-01 2.25268513e-01 9.85658988e-02 7.11966574e-01
-9.00727749e-01 -1.12163317e+00 -3.19938123e-01 4.54314291e-01
-7.91295946e-01 2.66375333e-01 -4.00248244e-02 8.41433346e-01
-3.53195578e-01 9.29861486e-01 -2.47675806e-01 1.98643416e-01
1.54917449e-01 -1.17330775e-01 1.23065531e+00 -5.27881645e-02
-2.29642823e-01 1.14843778e-01 1.57646105e-01 -5.71488678e-01
-3.38821650e-01 -5.85135579e-01 -1.15968573e+00 -4.53542441e-01
-5.18630207e-01 7.30688274e-01 5.14060497e-01 9.14369702e-01
5.76946400e-02 7.48524442e-02 5.78186274e-01 -3.97906661e-01
-1.38514316e+00 -8.85175049e-01 -9.93728340e-01 1.59037462e-03
3.90424281e-01 -5.24009168e-01 -4.63665277e-01 -8.01675081e-01]
|
[4.512645244598389, 3.309234619140625]
|
05bea9f3-0e41-41a8-89fa-9f7273a81c48
|
a-robotic-visual-grasping-design-rethinking
|
2209.07459
| null |
https://arxiv.org/abs/2209.07459v2
|
https://arxiv.org/pdf/2209.07459v2.pdf
|
A Robotic Visual Grasping Design: Rethinking Convolution Neural Network with High-Resolutions
|
High-resolution representations are important for vision-based robotic grasping problems. Existing works generally encode the input images into low-resolution representations via sub-networks and then recover high-resolution representations. This will lose spatial information, and errors introduced by the decoder will be more serious when multiple types of objects are considered or objects are far away from the camera. To address these issues, we revisit the design paradigm of CNN for robotic perception tasks. We demonstrate that using parallel branches as opposed to serial stacked convolutional layers will be a more powerful design for robotic visual grasping tasks. In particular, guidelines of neural network design are provided for robotic perception tasks, e.g., high-resolution representation and lightweight design, which respond to the challenges in different manipulation scenarios. We then develop a novel grasping visual architecture referred to as HRG-Net, a parallel-branch structure that always maintains a high-resolution representation and repeatedly exchanges information across resolutions. Extensive experiments validate that these two designs can effectively enhance the accuracy of visual-based grasping and accelerate network training. We show a series of comparative experiments in real physical environments at Youtube: https://youtu.be/Jhlsp-xzHFY.
|
['Zhen Kan', 'Mingyu Cai', 'Ziyang Chen', 'Shaochen Wang', 'Zhangli Zhou']
|
2022-09-15
| null | null | null | null |
['robotic-grasping']
|
['robots']
|
[ 1.72965571e-01 -1.60204321e-01 -2.43852600e-01 -2.50017881e-01
-2.59099990e-01 -3.31345618e-01 1.35868236e-01 -3.16586256e-01
-7.27238879e-02 4.77946520e-01 1.17674820e-01 1.03612252e-01
-2.04826772e-01 -9.20301497e-01 -1.25631988e+00 -6.32773101e-01
-6.98249638e-02 -1.09951839e-01 3.20958048e-01 -3.22054207e-01
2.93713480e-01 6.64024591e-01 -1.66095269e+00 7.87015796e-01
4.60803211e-01 1.22610080e+00 1.15418410e+00 5.48372269e-01
-7.16760233e-02 7.84975886e-01 -5.72188914e-01 6.22614324e-02
4.72530484e-01 2.18447879e-01 -8.11610341e-01 -2.86425620e-01
2.02820957e-01 -9.87882197e-01 -7.59742022e-01 1.09894586e+00
4.07546818e-01 -1.02311537e-01 5.00940681e-01 -8.40091050e-01
-1.32118082e+00 1.05623245e+00 -7.86408722e-01 -7.02871308e-02
1.81103393e-01 2.38474309e-01 4.31031853e-01 -9.01568651e-01
5.03994644e-01 1.86108124e+00 5.16527116e-01 8.13426077e-01
-9.76796210e-01 -6.93198919e-01 3.18224192e-01 2.88022339e-01
-8.59243095e-01 -3.20258774e-02 8.14805746e-01 -2.93420613e-01
7.08343863e-01 -1.41676217e-01 4.74593341e-01 1.55529761e+00
2.78379261e-01 8.89281452e-01 7.52709270e-01 -7.25336224e-02
-3.41974124e-02 -5.34562349e-01 8.73835385e-02 5.66179335e-01
6.26814663e-01 2.22594738e-01 -5.17210305e-01 2.22303540e-01
1.67895031e+00 4.97699887e-01 -5.41310132e-01 -5.20274758e-01
-1.58612955e+00 4.19155598e-01 1.14757776e+00 1.38855144e-01
-5.92297971e-01 6.34989440e-01 3.88288140e-01 1.55461177e-01
-2.21185058e-01 3.96010280e-01 -2.80729383e-01 2.80759901e-01
-7.76255131e-02 1.95018575e-01 4.27998483e-01 1.66207278e+00
4.86248404e-01 7.17896819e-02 -1.67704552e-01 1.12577510e+00
3.03792566e-01 3.07546288e-01 2.38281086e-01 -1.28531849e+00
4.92822140e-01 4.95646298e-01 1.47944182e-01 -8.74211252e-01
-2.93867081e-01 -2.79818680e-02 -1.10521412e+00 5.49980700e-01
3.31793010e-01 1.32120252e-01 -1.12961090e+00 1.45193422e+00
-6.00266866e-02 -4.40335304e-01 2.16921538e-01 1.42900538e+00
1.16054678e+00 7.01532185e-01 1.01388186e-01 2.73962170e-01
1.31966913e+00 -1.10260534e+00 -8.03116500e-01 -1.07293993e-01
-2.72279561e-01 -5.73673308e-01 1.13406742e+00 3.83264869e-01
-1.23124754e+00 -9.07649398e-01 -1.31514585e+00 -4.05707300e-01
-3.40600103e-01 4.27075207e-01 7.61169195e-01 -3.27846497e-01
-6.70395255e-01 9.06181216e-01 -8.19881856e-01 -1.45969450e-01
7.21368253e-01 2.39891171e-01 -3.00838441e-01 -4.56732154e-01
-9.31766272e-01 9.23581123e-01 6.52747273e-01 5.07896960e-01
-1.19098008e+00 -4.78433013e-01 -7.85143971e-01 2.53222529e-02
1.84880599e-01 -3.80009711e-01 1.41090274e+00 -6.01795435e-01
-1.42252243e+00 5.16941905e-01 2.80841202e-01 5.72109502e-03
3.59159231e-01 -4.75130379e-01 2.28584871e-01 3.60884368e-01
-1.29127085e-01 9.00877237e-01 8.78447711e-01 -1.74769032e+00
-4.42731261e-01 -4.57414538e-01 4.81393546e-01 -4.82576080e-02
-1.13170438e-01 -9.32513103e-02 -2.31616333e-01 -6.68649137e-01
4.76666242e-01 -5.33756614e-01 -2.68983319e-02 7.14303017e-01
-1.24148145e-01 -2.43327260e-01 1.20892763e+00 -5.39504290e-01
4.11336541e-01 -1.93341064e+00 4.84859318e-01 -3.43830258e-01
1.98610291e-01 3.43952388e-01 -3.95217717e-01 3.40857387e-01
9.26393345e-02 -8.69259909e-02 2.43997071e-02 2.83814728e-01
-1.31054357e-01 3.85724545e-01 -4.87810642e-01 1.91581458e-01
3.22335899e-01 1.21490800e+00 -9.04374838e-01 -1.73792049e-01
3.66199970e-01 6.78765595e-01 -2.94271588e-01 5.60859323e-01
-4.03358310e-01 5.40476382e-01 -6.50483310e-01 9.59192336e-01
9.60568428e-01 -3.54363501e-01 1.19317122e-01 -7.43010461e-01
-2.05576420e-01 -1.87231526e-01 -7.76679039e-01 2.00377464e+00
-4.15557355e-01 5.50905526e-01 5.29016972e-01 -9.65608418e-01
1.30156863e+00 -4.95346263e-02 2.48944327e-01 -8.60330462e-01
2.78418541e-01 1.11778870e-01 -2.22096607e-01 -8.06225657e-01
3.84731352e-01 3.79254311e-01 1.23872846e-01 1.06438939e-02
9.87685248e-02 -2.03579180e-02 -5.45751117e-02 -1.88045710e-01
7.67071843e-01 5.77528596e-01 5.01602478e-02 -2.04931080e-01
-5.36194555e-02 -3.30796801e-02 4.33429092e-01 7.17139959e-01
-8.29947367e-02 8.01316202e-01 3.41940045e-01 -6.35731995e-01
-1.43744469e+00 -1.11416256e+00 -2.26119310e-01 1.06638467e+00
8.12710106e-01 1.23705119e-02 -4.84181732e-01 -5.75577170e-02
1.94862053e-01 1.81671027e-02 -6.01816058e-01 -1.98242828e-01
-9.64885950e-01 -3.04972559e-01 1.91933081e-01 1.01980603e+00
8.79924893e-01 -1.50650740e+00 -1.25292242e+00 1.74021050e-01
-7.34273195e-02 -1.14371800e+00 1.80810705e-01 2.72489518e-01
-9.25234675e-01 -1.20384157e+00 -1.15876222e+00 -1.23035073e+00
7.54092574e-01 7.70675600e-01 7.41264820e-01 6.95034713e-02
-6.02987468e-01 2.85101205e-01 -8.06400955e-01 -4.32846040e-01
-1.64291859e-01 -1.13090046e-01 -1.31710559e-01 -7.32695162e-01
1.30023018e-01 -4.87654120e-01 -9.45670664e-01 2.38346055e-01
-1.04320145e+00 1.96102068e-01 1.00441170e+00 9.33496773e-01
5.67591548e-01 -4.94289815e-01 5.65235198e-01 -8.65823328e-02
7.32008815e-01 -3.98398817e-01 -5.64498365e-01 4.58598256e-01
7.70245492e-02 4.48063016e-02 7.18389034e-01 -8.47187817e-01
-9.01283503e-01 1.55004814e-01 1.36456117e-01 -1.00273502e+00
-5.31944074e-02 9.39496905e-02 9.53187123e-02 -1.61675066e-01
6.55036449e-01 9.28222090e-02 1.60793200e-01 -6.70351326e-01
3.58262628e-01 8.23378444e-01 6.45548940e-01 -8.22352409e-01
3.59945714e-01 3.05536211e-01 -1.66199490e-01 -7.27969050e-01
-5.08988798e-01 1.21528711e-02 -7.75161922e-01 -4.04199570e-01
8.37212980e-01 -9.79536891e-01 -1.11602843e+00 5.60000718e-01
-1.56425190e+00 -4.85529244e-01 -1.11802280e-01 4.55958486e-01
-6.93501711e-01 3.11668485e-01 -9.77684855e-01 -5.84782660e-01
-5.25718272e-01 -1.32419145e+00 1.22164977e+00 5.38752317e-01
3.37785602e-01 -1.15373857e-01 -5.61438024e-01 3.44962180e-02
5.57708263e-01 2.93231159e-01 9.71930385e-01 2.55151391e-01
-9.53097582e-01 2.25885794e-01 -8.77916396e-01 3.73879701e-01
1.88444346e-01 -1.60718635e-01 -7.22488582e-01 -4.72049594e-01
-3.55882823e-01 -7.94376194e-01 1.04047036e+00 3.56693387e-01
1.80746472e+00 -2.74915069e-01 -3.04216057e-01 7.18404651e-01
1.53614283e+00 3.06462198e-01 7.66087532e-01 3.20457548e-01
7.08657324e-01 5.52580833e-01 5.89448035e-01 3.89693499e-01
4.90491353e-02 5.35529912e-01 1.05706859e+00 -7.74633745e-03
-2.95869559e-01 -3.09119612e-01 1.24127708e-01 6.27782762e-01
-3.78864020e-01 -2.86336597e-02 -7.28568435e-01 3.99637163e-01
-1.99409521e+00 -7.92659402e-01 1.03888929e-01 1.73809290e+00
6.22485876e-01 -2.68154621e-01 -2.04478666e-01 -2.71393627e-01
9.16865349e-01 1.19307481e-01 -8.81770849e-01 -2.78564215e-01
-6.41057864e-02 8.04287717e-02 5.71776271e-01 -4.25735004e-02
-8.08362186e-01 8.86779249e-01 6.33795023e+00 3.52075458e-01
-1.22097135e+00 -7.44359568e-02 5.78951649e-02 -5.69252251e-03
7.03117670e-03 -3.99028003e-01 -4.70127016e-01 3.48548442e-01
5.31820692e-02 2.52554595e-01 5.38590670e-01 1.27556324e+00
-2.87368566e-01 -3.31049748e-02 -1.24429381e+00 1.23428714e+00
-9.53704864e-02 -1.33863235e+00 3.43200058e-01 -2.95503289e-01
4.35906321e-01 8.77090842e-02 -5.96208349e-02 1.58263594e-01
3.49846601e-01 -1.12704420e+00 9.14908588e-01 5.38574219e-01
1.11003006e+00 -4.39054877e-01 4.50156033e-01 1.47967577e-01
-1.28662336e+00 -5.76249182e-01 -1.22686768e+00 4.04750258e-02
4.30971347e-02 8.27873722e-02 -1.22536704e-01 4.86555278e-01
1.35770226e+00 7.46266723e-01 8.40929225e-02 9.19955254e-01
-1.63454279e-01 -2.90877908e-01 4.64393497e-02 -2.84157902e-01
1.61907554e-01 2.30137363e-01 3.53959560e-01 1.06244373e+00
2.79881090e-01 3.67802352e-01 2.90655755e-02 1.42819452e+00
-1.46078140e-01 -4.60943282e-01 -7.35625267e-01 3.95255201e-02
5.07326305e-01 1.09925699e+00 -3.71328026e-01 -8.68585184e-02
-2.59338796e-01 8.67033601e-01 7.02264607e-01 4.28325146e-01
-6.72331035e-01 -6.28428638e-01 7.62145281e-01 -3.09393983e-02
5.40204704e-01 -7.23600149e-01 -8.49402100e-02 -1.01439714e+00
3.42448264e-01 -6.41121209e-01 -1.44394726e-01 -1.23368859e+00
-1.20459485e+00 6.19218111e-01 8.71972442e-02 -1.33471620e+00
2.24529147e-01 -1.25581539e+00 -2.58640140e-01 6.45434856e-01
-1.74019349e+00 -1.05530965e+00 -9.72194314e-01 4.95127529e-01
9.69133079e-01 -3.25417109e-02 6.18037224e-01 1.06326282e-01
-2.23919034e-01 1.91203967e-01 9.83235240e-02 3.21764410e-01
3.84160042e-01 -6.51642203e-01 8.85585137e-03 3.40212852e-01
-7.45106161e-01 7.24852145e-01 2.90220410e-01 -5.59783459e-01
-1.93155921e+00 -1.03652823e+00 -2.76973903e-01 1.23572135e-02
2.72568583e-01 -4.27961826e-01 -1.04187965e+00 6.69522703e-01
1.14985444e-01 8.34319799e-04 -1.78247243e-01 -3.46989423e-01
-4.45469350e-01 -7.43170083e-02 -1.16314924e+00 5.75166523e-01
1.43346596e+00 -2.52544165e-01 -7.03186274e-01 2.68987119e-01
8.71665299e-01 -6.07095242e-01 -1.11516333e+00 9.39082742e-01
9.89743650e-01 -7.27328658e-01 1.25424278e+00 -5.39978743e-01
1.17913330e+00 -2.02739909e-01 -4.10702080e-01 -1.13858926e+00
-7.10646391e-01 -6.18698597e-02 -2.71351188e-01 6.55168593e-01
-5.89958653e-02 -3.38056028e-01 4.44756180e-01 1.43148616e-01
-2.03025192e-01 -8.60915720e-01 -5.25838077e-01 -8.09768617e-01
2.39831656e-01 1.89880177e-01 5.57765961e-01 5.26078105e-01
1.27925232e-01 -5.48799075e-02 -3.76163006e-01 2.53090024e-01
7.62841761e-01 3.92856985e-01 5.40884495e-01 -1.00107801e+00
2.07865745e-01 -4.67392296e-01 -3.60087782e-01 -1.55620933e+00
-4.12184298e-02 -6.14004612e-01 4.77772832e-01 -2.00674224e+00
3.73079628e-01 -4.60816503e-01 -3.61485332e-02 6.41354382e-01
2.86354184e-01 5.62149426e-03 3.51181030e-01 5.50535619e-01
-3.39253277e-01 6.92191184e-01 1.96643519e+00 -3.43911141e-01
1.68668702e-01 -5.29693067e-01 -5.31856954e-01 4.54720527e-01
9.71626699e-01 5.07913679e-02 -2.26035073e-01 -1.10598195e+00
-2.65937686e-01 2.45625839e-01 6.98367000e-01 -1.06575298e+00
1.62135825e-01 -2.93706357e-01 8.62140596e-01 -5.71330130e-01
4.74311680e-01 -9.36245501e-01 -8.78686309e-02 6.55403376e-01
-6.09251618e-01 1.87659170e-02 2.17297345e-01 4.97658134e-01
-1.09496288e-01 -2.02193588e-01 9.38380003e-01 -6.43177032e-01
-1.07658482e+00 3.33175898e-01 -1.00064166e-01 -5.83292663e-01
1.09559608e+00 -3.97136986e-01 -8.53852451e-01 -4.69084121e-02
-5.76279521e-01 3.44936281e-01 3.88030201e-01 9.16153133e-01
1.20321405e+00 -1.46485579e+00 -6.13214314e-01 1.65323779e-01
1.53252438e-01 5.53220689e-01 4.53498363e-01 1.94164261e-01
-8.26413810e-01 3.19054186e-01 -1.03411376e+00 -7.91348696e-01
-9.90548968e-01 7.54533410e-01 2.03643709e-01 3.98103058e-01
-1.13093650e+00 7.86654651e-01 3.56416494e-01 -2.94188112e-01
6.96597636e-01 -5.50053179e-01 -2.24844381e-01 -5.35804570e-01
6.07299626e-01 2.81308353e-01 -4.68107373e-01 -2.41543129e-01
-1.13923654e-01 8.35666478e-01 -2.30945662e-01 5.26886821e-01
1.55754721e+00 4.95433286e-02 -2.01983944e-01 1.62576646e-01
1.13471162e+00 -8.58068168e-01 -1.77607906e+00 -2.60808051e-01
-4.99236405e-01 -6.57773256e-01 -2.79044420e-01 -7.61868596e-01
-1.31849706e+00 1.05647910e+00 6.88329041e-01 -2.44794525e-02
9.64592695e-01 2.33532503e-01 6.65602803e-01 8.03323150e-01
8.43121886e-01 -1.12403727e+00 4.85101968e-01 6.26067996e-01
1.73932767e+00 -1.27625346e+00 -9.61243659e-02 -5.15771210e-01
-4.47025239e-01 1.54130399e+00 1.23036170e+00 -5.39288342e-01
3.52359712e-01 4.87789184e-01 -1.63429767e-01 -2.34243870e-01
-5.73245585e-01 -1.01943603e-02 -2.38121077e-01 8.95158589e-01
1.45688772e-01 9.62455943e-02 -6.61505386e-02 4.42887127e-01
1.02185614e-01 1.05990596e-01 4.31535900e-01 1.28040445e+00
-7.17821658e-01 -5.25986493e-01 -2.74847776e-01 2.68215686e-01
-1.23946734e-01 2.59244740e-01 -1.51075885e-01 5.84698796e-01
1.96663756e-02 5.97017467e-01 9.93684828e-02 -5.56906223e-01
5.92076302e-01 -5.85373640e-01 9.98520076e-01 -2.98254907e-01
-1.52181461e-01 -2.91442335e-01 -3.40227664e-01 -9.23232257e-01
-6.12507045e-01 1.01058334e-02 -1.28231108e+00 -2.06772164e-01
-1.27663061e-01 -3.28714758e-01 9.87734973e-01 3.32171708e-01
4.20698136e-01 9.55612361e-01 3.43942970e-01 -1.53870487e+00
-8.39387178e-01 -1.02256250e+00 -3.61832052e-01 1.38409361e-01
4.22842264e-01 -9.67101276e-01 7.14447871e-02 -4.94124033e-02]
|
[5.77485466003418, -0.8965104222297668]
|
6c45bb75-cbce-480e-a859-a6ca69d6f159
|
convolutional-sequence-to-sequence-model-for
|
1805.00655
| null |
http://arxiv.org/abs/1805.00655v1
|
http://arxiv.org/pdf/1805.00655v1.pdf
|
Convolutional Sequence to Sequence Model for Human Dynamics
|
Human motion modeling is a classic problem in computer vision and graphics.
Challenges in modeling human motion include high dimensional prediction as well
as extremely complicated dynamics.We present a novel approach to human motion
modeling based on convolutional neural networks (CNN). The hierarchical
structure of CNN makes it capable of capturing both spatial and temporal
correlations effectively. In our proposed approach,a convolutional long-term
encoder is used to encode the whole given motion sequence into a long-term
hidden variable, which is used with a decoder to predict the remainder of the
sequence. The decoder itself also has an encoder-decoder structure, in which
the short-term encoder encodes a shorter sequence to a short-term hidden
variable, and the spatial decoder maps the long and short-term hidden variable
to motion predictions. By using such a model, we are able to capture both
invariant and dynamic information of human motion, which results in more
accurate predictions. Experiments show that our algorithm outperforms the
state-of-the-art methods on the Human3.6M and CMU Motion Capture datasets. Our
code is available at the project website.
|
['Zhen Zhang', 'Chen Li', 'Wee Sun Lee', 'Gim Hee Lee']
|
2018-05-02
|
convolutional-sequence-to-sequence-model-for-1
|
http://openaccess.thecvf.com/content_cvpr_2018/html/Li_Convolutional_Sequence_to_CVPR_2018_paper.html
|
http://openaccess.thecvf.com/content_cvpr_2018/papers/Li_Convolutional_Sequence_to_CVPR_2018_paper.pdf
|
cvpr-2018-6
|
['human-pose-forecasting', 'human-dynamics']
|
['computer-vision', 'computer-vision']
|
[-9.80461612e-02 -1.75383970e-01 -4.36007053e-01 -8.06360170e-02
-1.12389596e-02 -5.93992062e-02 5.75900733e-01 -5.77107072e-01
-3.16654563e-01 4.79610592e-01 5.35085380e-01 -5.33194803e-02
6.16162956e-01 -7.48541534e-01 -7.51424611e-01 -6.41481757e-01
-2.53862560e-01 2.54418194e-01 6.43042147e-01 -6.46906197e-02
-3.12376232e-03 3.45071882e-01 -1.42048621e+00 4.15096015e-01
2.63913833e-02 9.22047257e-01 5.17277718e-01 1.24128962e+00
5.25232479e-02 1.41799533e+00 -8.15194920e-02 6.40227273e-03
3.66243497e-02 -5.61343789e-01 -7.99278259e-01 -9.04576033e-02
2.24313792e-02 -6.20117068e-01 -7.98826993e-01 5.81870735e-01
4.49609101e-01 1.68829262e-01 4.62717921e-01 -1.14368105e+00
-3.71061951e-01 1.60699934e-01 -4.79728222e-01 1.22423694e-01
3.94163430e-01 3.92319769e-01 8.03217232e-01 -7.43168592e-01
9.56715941e-01 1.36925244e+00 7.07005441e-01 9.18314695e-01
-7.96237171e-01 -4.91856933e-01 -6.48865849e-02 4.27239448e-01
-1.11466241e+00 -3.71979564e-01 7.63723731e-01 -8.41668010e-01
1.21748638e+00 1.10196739e-01 1.06219888e+00 1.01369333e+00
8.08315337e-01 9.64891195e-01 3.60592417e-02 -1.90342337e-01
7.01823533e-02 -5.70904911e-01 2.52635386e-02 8.78461361e-01
-2.51532823e-01 3.49261999e-01 -4.93362010e-01 -3.47818397e-02
1.21057689e+00 9.61464420e-02 -1.51746571e-01 -3.95489603e-01
-1.42864895e+00 6.52329445e-01 4.26126093e-01 1.54223427e-01
-6.22373700e-01 8.56076717e-01 4.30757821e-01 -2.80388743e-01
2.38446578e-01 -2.02023387e-01 -3.08766425e-01 -4.59512353e-01
-1.12244308e+00 6.99908912e-01 6.91396475e-01 8.77564728e-01
7.81345844e-01 1.15232110e-01 -2.31417328e-01 4.55552399e-01
4.53754842e-01 2.79077381e-01 7.14759529e-01 -1.35477436e+00
3.60367030e-01 2.70397484e-01 3.13140601e-01 -1.25378668e+00
-5.27750373e-01 -4.79761139e-02 -1.17765999e+00 2.86971211e-01
1.55406877e-01 -2.40148529e-01 -9.32389855e-01 1.73197925e+00
1.32467762e-01 7.07690716e-01 -1.82385355e-01 1.09588277e+00
5.83638787e-01 1.01475346e+00 1.26854971e-01 -8.14303830e-02
1.16361296e+00 -1.44895911e+00 -9.10387099e-01 -2.75080085e-01
8.17048907e-01 -5.22644103e-01 4.93243456e-01 -1.75972223e-01
-1.25293815e+00 -9.57222700e-01 -1.04054999e+00 -3.59749198e-01
2.41952892e-02 7.36396983e-02 3.74560177e-01 -6.94909617e-02
-1.22666430e+00 6.34523690e-01 -1.33596516e+00 -2.11682543e-01
1.50793478e-01 2.77850211e-01 -3.27337921e-01 7.27214739e-02
-1.21049464e+00 8.34298551e-01 2.18537956e-01 1.56938657e-01
-9.08592820e-01 -2.50947982e-01 -1.18986738e+00 6.77915215e-02
-1.46081358e-01 -1.29473269e+00 1.42924559e+00 -7.78744340e-01
-1.54789031e+00 4.95542496e-01 -6.03044808e-01 -6.02939308e-01
5.86463809e-01 -2.78796256e-01 -9.99329835e-02 1.21366955e-01
8.31491500e-02 1.00882399e+00 7.03003228e-01 -8.48010719e-01
-7.27091610e-01 -8.66845548e-02 -2.08908185e-01 2.30137095e-01
1.83160007e-01 -7.71803856e-02 -8.96416843e-01 -8.18641245e-01
-5.11992723e-02 -1.26970994e+00 -6.10914469e-01 2.28419185e-01
-1.31241977e-01 2.49819085e-01 1.09984684e+00 -7.61289716e-01
1.71475136e+00 -1.91361725e+00 4.61201072e-01 -2.88259447e-01
2.68408954e-01 4.10523444e-01 4.39424627e-02 3.45819771e-01
-1.26116619e-01 -8.59592110e-02 -1.60263553e-01 -6.55853748e-01
-1.93826884e-01 3.13380629e-01 -2.30790570e-01 4.32065874e-01
-2.63812020e-02 1.25958645e+00 -9.03316438e-01 -4.05772060e-01
5.51083565e-01 5.66122413e-01 -6.20144069e-01 4.42643255e-01
-3.25744987e-01 6.54842317e-01 -3.90206426e-01 3.14097732e-01
3.75557631e-01 -4.55597907e-01 -7.88422301e-02 1.49204850e-01
-2.86777794e-01 1.33531705e-01 -9.77618873e-01 1.78940570e+00
-1.90600470e-01 9.60315406e-01 -2.47248694e-01 -6.21343255e-01
6.46709263e-01 4.68884975e-01 6.42582059e-01 -4.98023301e-01
3.12823243e-02 -1.53920939e-02 -2.07643718e-01 -7.38914490e-01
7.54163146e-01 -4.47438844e-02 6.72206357e-02 3.65098506e-01
-2.53969878e-01 2.95767426e-01 -9.17184576e-02 -2.76131779e-02
1.13698399e+00 5.39329588e-01 4.06879365e-01 8.98764208e-02
7.03626275e-01 3.13477814e-02 7.66390562e-01 3.88707250e-01
-4.38520133e-01 8.00269544e-01 3.76497328e-01 -9.93091226e-01
-1.44818056e+00 -5.88224113e-01 5.14850497e-01 7.33344913e-01
1.66882634e-01 -5.71754038e-01 -5.94671190e-01 -4.32851613e-02
-1.39438465e-01 2.73182482e-01 -8.09230626e-01 -1.00273810e-01
-1.12120426e+00 -1.95656031e-01 4.02497739e-01 9.68928993e-01
5.46168327e-01 -1.25881851e+00 -1.09096599e+00 5.18827319e-01
-3.51769388e-01 -1.03059590e+00 -6.75565958e-01 -3.04357022e-01
-8.68721724e-01 -7.51108050e-01 -1.04211295e+00 -8.62988949e-01
1.17307700e-01 3.04129660e-01 9.93099689e-01 2.42201269e-01
-2.23044038e-01 -6.75301701e-02 -1.43149078e-01 -1.38327451e-02
-3.49941283e-01 -1.78130984e-01 -2.01327205e-01 -4.31419499e-02
2.66708463e-01 -3.49680156e-01 -8.79555643e-01 2.10104063e-01
-7.22750008e-01 6.41825318e-01 2.99564987e-01 8.36334646e-01
4.85203207e-01 -3.32959056e-01 -1.10155322e-01 -5.36394656e-01
9.76102054e-03 -5.77242017e-01 -4.36500072e-01 -9.97471586e-02
2.34146821e-04 4.60878983e-02 4.63310510e-01 -4.94519591e-01
-8.89805019e-01 5.57586074e-01 -2.28389397e-01 -7.90411949e-01
-3.92381754e-03 4.77137536e-01 4.45914082e-02 2.13892639e-01
2.63107061e-01 2.92383015e-01 5.70519865e-02 -3.75200927e-01
2.61497021e-01 2.44775176e-01 7.92914450e-01 8.55648071e-02
4.17042553e-01 7.25167036e-01 2.73170561e-01 -9.30243075e-01
-2.90586472e-01 -5.42881191e-01 -1.06139576e+00 -5.00683248e-01
1.46566796e+00 -1.14603627e+00 -6.93047941e-01 8.51736009e-01
-1.61915588e+00 -5.57412565e-01 1.42551726e-02 6.04222238e-01
-1.02118719e+00 3.67520273e-01 -1.14419901e+00 -7.09653556e-01
-6.27180412e-02 -1.33685231e+00 1.22069347e+00 2.75416411e-02
-6.39270961e-01 -1.29002571e+00 4.39736247e-01 3.38527374e-02
2.09362790e-01 5.55858791e-01 7.06239343e-01 1.17016636e-01
-8.40075970e-01 -1.79379806e-01 1.71847388e-01 -3.49864960e-02
-2.00782493e-01 1.22563355e-01 -6.38832629e-01 -9.38082635e-02
-1.95484329e-02 -3.12268436e-02 1.01933908e+00 9.26645756e-01
1.00214016e+00 -1.08394086e-01 -5.92252076e-01 8.94800365e-01
1.18680286e+00 2.20341817e-01 1.09201443e+00 4.78596836e-01
1.14250255e+00 4.67054278e-01 5.80965757e-01 5.87276161e-01
7.53243446e-01 8.80787492e-01 3.82052511e-01 -1.13824703e-01
-1.83174044e-01 -3.65594357e-01 3.60511929e-01 9.48955178e-01
-2.69439638e-01 -3.33932489e-01 -1.07161891e+00 5.17585278e-01
-2.34760475e+00 -1.43602157e+00 -4.79799360e-01 1.79762614e+00
4.44105238e-01 -9.99167040e-02 2.22128466e-01 -1.82339456e-02
6.59015238e-01 5.55993199e-01 -3.18073481e-01 -2.91357428e-01
9.06201899e-02 -2.30430707e-01 3.92579556e-01 7.12854147e-01
-1.19267130e+00 1.10517550e+00 6.92563820e+00 3.21333736e-01
-1.20069826e+00 -1.57699764e-01 4.70718205e-01 -4.93160598e-02
-1.75020751e-02 -3.98678482e-02 -6.99119449e-01 5.15875816e-01
1.13568354e+00 -1.72425091e-01 2.82681752e-02 8.30363750e-01
6.19904280e-01 -7.08953962e-02 -1.08309722e+00 1.04908812e+00
-6.77191094e-02 -1.66868341e+00 1.84313059e-01 2.67022729e-01
7.43903935e-01 -1.41737476e-01 -1.50618359e-01 1.32614365e-02
2.83725746e-02 -1.19154775e+00 9.56886411e-01 9.47534859e-01
6.27334595e-01 -7.03168154e-01 6.37741208e-01 7.88441062e-01
-1.77911496e+00 4.82170358e-02 -4.17618573e-01 -4.97219294e-01
6.35302722e-01 4.28923070e-02 -4.69325989e-01 2.89689779e-01
5.52711546e-01 1.30205655e+00 -1.07960507e-01 8.37429762e-01
3.22332978e-02 3.12576711e-01 2.90892571e-01 1.83344454e-01
3.27156097e-01 -7.64460638e-02 4.27567095e-01 1.35765159e+00
3.99700344e-01 3.69002730e-01 2.44847447e-01 6.71887755e-01
2.65506715e-01 -3.51043731e-01 -7.74171054e-01 4.67480235e-02
8.20268616e-02 7.63031960e-01 -3.93396735e-01 -6.77127302e-01
-4.82846200e-01 1.27267599e+00 2.81463921e-01 4.40526038e-01
-9.05126631e-01 9.78822336e-02 9.67561066e-01 7.97353387e-02
4.93695647e-01 -6.62576973e-01 -3.15746307e-01 -1.21636987e+00
-7.93514699e-02 -3.44306737e-01 2.36702845e-01 -9.44900095e-01
-6.70105875e-01 5.28974652e-01 -1.69913858e-01 -1.56058335e+00
-9.64864969e-01 -4.82631743e-01 -6.26724958e-01 1.14884067e+00
-1.25086224e+00 -1.21393836e+00 -4.49822098e-01 5.68324685e-01
7.65572190e-01 -6.86646765e-03 7.86513329e-01 2.85708457e-01
-3.57259721e-01 5.64051531e-02 -1.07690588e-01 2.64234155e-01
3.94381642e-01 -8.45038235e-01 9.90091205e-01 7.69230247e-01
-2.45694295e-01 2.95542866e-01 7.33620644e-01 -9.94632006e-01
-1.26865184e+00 -1.29546976e+00 1.29650378e+00 -5.27462423e-01
4.40199703e-01 -9.53296125e-02 -9.65020657e-01 1.08175492e+00
3.04421242e-02 7.65895918e-02 4.82039988e-01 -6.14891648e-01
1.40912920e-01 5.62236369e-01 -4.39664871e-01 7.79231489e-01
9.08343434e-01 -5.01319349e-01 -4.55768228e-01 -1.76074758e-01
7.91762829e-01 -7.37646759e-01 -4.65859026e-01 3.46359938e-01
1.03751862e+00 -1.10960209e+00 1.03890133e+00 -5.24719298e-01
9.87181783e-01 -3.30874801e-01 -1.96969118e-02 -8.94353390e-01
-8.59713733e-01 -4.69472021e-01 -7.48263299e-01 3.88572633e-01
9.60141197e-02 5.85092194e-02 1.09263206e+00 6.17543519e-01
-5.75617515e-02 -8.90945435e-01 -7.66338527e-01 -4.53634739e-01
-2.42402647e-02 -5.56052268e-01 3.27495158e-01 7.29350984e-01
-2.02650502e-01 1.05332844e-01 -1.14467466e+00 9.11846012e-02
3.17968756e-01 -7.29012489e-02 9.86003697e-01 -7.37799048e-01
-3.40922326e-01 -3.24621588e-01 -7.74140120e-01 -1.69554818e+00
2.49157071e-01 -3.88909876e-01 1.76819772e-01 -1.57545459e+00
3.50083500e-01 2.40432203e-01 1.32504568e-01 4.41720970e-02
-2.72239834e-01 2.06443816e-01 3.55528235e-01 4.89767343e-01
-3.95768881e-01 6.62104666e-01 1.43010509e+00 2.32911985e-02
-2.08110556e-01 6.84157014e-02 2.24593371e-01 9.17801201e-01
4.85151201e-01 -3.13230038e-01 -2.74403602e-01 -6.12561464e-01
-2.72116810e-01 7.43652463e-01 6.32051647e-01 -1.20891345e+00
5.88713348e-01 -2.55084485e-01 7.52493978e-01 -1.00708294e+00
6.50136292e-01 -6.18383706e-01 6.03095591e-01 9.27069783e-01
-3.84413332e-01 4.77713883e-01 1.84184294e-02 5.70871830e-01
-3.34395677e-01 2.13158429e-01 7.83548832e-01 -3.27491492e-01
-1.09688580e+00 7.21272767e-01 -5.98777652e-01 -4.66082543e-01
1.01953816e+00 -3.01701486e-01 3.17185223e-02 -8.32306862e-01
-8.52812469e-01 2.54386038e-01 6.32850230e-01 6.17864966e-01
8.34395230e-01 -1.68034160e+00 -5.68591297e-01 2.07728460e-01
-1.75578311e-01 -1.25965372e-01 4.61164981e-01 5.26030183e-01
-9.44941163e-01 6.97820187e-01 -4.97558802e-01 -8.33296597e-01
-1.29899573e+00 5.38477540e-01 4.06722605e-01 -2.98387527e-01
-8.26126814e-01 6.67162001e-01 4.71794695e-01 2.18525808e-02
1.64368704e-01 -2.86525488e-01 -3.81223649e-01 -3.53549123e-01
8.49219799e-01 4.92630363e-01 -6.37438893e-01 -1.33409429e+00
-3.26814353e-01 7.69106209e-01 4.08379793e-01 -3.53285909e-01
1.14312625e+00 -3.97806734e-01 -5.84076792e-02 7.79228628e-01
1.49105155e+00 -5.19755840e-01 -1.57445776e+00 -5.38584702e-02
-2.67220214e-02 -4.54235137e-01 -2.08156556e-01 -8.59602690e-02
-9.11766350e-01 1.15654397e+00 3.09285313e-01 -2.65680283e-01
9.78379011e-01 -2.28966057e-01 1.29454434e+00 -5.67700267e-02
4.54500496e-01 -8.76558602e-01 2.18811899e-01 1.07898533e+00
7.79695988e-01 -1.01455855e+00 -2.29006633e-01 -2.71603495e-01
-8.21467638e-01 1.30335331e+00 6.21773958e-01 -2.73705781e-01
7.62367845e-01 2.12810025e-01 5.44105023e-02 -7.98083395e-02
-1.23056042e+00 -9.90327150e-02 3.64355713e-01 4.61313248e-01
6.88922286e-01 1.32093072e-01 -1.05141103e-01 4.41733390e-01
-1.95864707e-01 3.81150275e-01 4.76816326e-01 1.00165522e+00
-5.07988095e-01 -9.91688967e-01 -2.72640169e-01 -7.34377280e-02
-2.54754305e-01 1.16872303e-01 -6.13728277e-02 6.80752158e-01
1.00400820e-01 4.92199004e-01 2.95937091e-01 -7.27410078e-01
8.09634849e-02 -2.73049343e-03 1.41157106e-01 -4.16380137e-01
-3.55160654e-01 2.71579027e-01 -9.30991024e-02 -1.08018732e+00
-4.42910701e-01 -5.95077693e-01 -1.49981546e+00 -5.12054443e-01
3.68360966e-01 -3.37764740e-01 4.93677706e-01 7.78979301e-01
3.62432718e-01 6.45629048e-01 2.93671310e-01 -1.38991404e+00
-1.23437494e-01 -8.26926470e-01 -3.93678427e-01 5.28183520e-01
8.26090395e-01 -5.36038220e-01 -4.25350107e-02 6.35575652e-01]
|
[7.32023811340332, -0.16830916702747345]
|
1e51abf8-ed76-4df8-a1d5-45dfb96586b6
|
towards-a-music-language-mapping
| null | null |
https://aclanthology.org/L18-1482
|
https://aclanthology.org/L18-1482.pdf
|
Towards a music-language mapping
| null |
['Francesca Bonin', 'Michele Berlingerio']
|
2018-05-01
|
towards-a-music-language-mapping-1
|
https://aclanthology.org/L18-1482
|
https://aclanthology.org/L18-1482.pdf
|
lrec-2018-5
|
['lexical-analysis']
|
['natural-language-processing']
|
[-8.63703638e-02 1.71006292e-01 -6.22772932e-01 -4.08054382e-01
-8.41685571e-03 -9.08429027e-01 6.55310392e-01 -6.53472245e-01
-2.85945535e-01 1.06888819e+00 -4.63127941e-02 -1.01159286e+00
-3.91567826e-01 -9.63214397e-01 -4.95059669e-01 -6.31337762e-01
-9.79754329e-01 7.25764990e-01 3.30370307e-01 -6.93831444e-01
7.03166842e-01 7.88774848e-01 -1.68942046e+00 7.18545914e-01
7.04417467e-01 8.52217197e-01 2.49141872e-01 1.14950800e+00
-1.95044339e-01 1.55633950e+00 -7.48382092e-01 -5.46825826e-01
3.13719302e-01 -1.23176083e-01 -7.22945035e-01 -1.01074085e-01
9.28529128e-02 -8.59008506e-02 -2.09758401e-01 9.22211111e-01
5.37373662e-01 4.49454933e-02 1.08379531e+00 -1.42548037e+00
-5.91619551e-01 6.10313773e-01 -4.01565880e-02 1.21627934e-01
1.03678203e+00 -5.39447069e-01 1.19919395e+00 -1.13026452e+00
7.20913768e-01 1.26888943e+00 8.66221786e-01 5.44149756e-01
-1.22286928e+00 -1.94712028e-01 -3.26822817e-01 -9.51717794e-02
-1.46558487e+00 -3.25250506e-01 4.25783843e-02 -2.08119690e-01
1.66093647e+00 1.26596653e+00 1.20609856e+00 1.01401424e+00
1.26658809e+00 8.34431887e-01 1.04267764e+00 -5.13792276e-01
3.35295945e-01 3.66983831e-01 1.54683650e-01 6.33519173e-01
8.40953708e-01 5.26628852e-01 -7.06372619e-01 -9.13127720e-01
9.33553874e-01 -2.94925272e-01 1.71355158e-01 -5.05680561e-01
-9.05919552e-01 6.91228509e-01 1.78732842e-01 3.83959889e-01
-1.39880210e-01 9.89067405e-02 1.26390755e-01 5.30987144e-01
-2.58292928e-02 6.47037446e-01 -9.11868811e-01 -1.33165747e-01
-8.71728659e-01 5.10332465e-01 1.25398111e+00 1.52653182e+00
1.24482810e-01 2.94908643e-01 -9.34252143e-02 3.17179203e-01
8.92314315e-01 1.01808000e+00 4.28362608e-01 -1.36146402e+00
-6.87414408e-02 1.72361732e-01 5.01781464e-01 -8.52631688e-01
-6.33224547e-01 -9.64177120e-03 -8.93263519e-01 4.49267089e-01
3.49161088e-01 4.57367361e-01 -8.02827001e-01 5.07305264e-01
4.33481112e-02 -2.34125629e-01 4.53833073e-01 5.55570945e-02
4.99930978e-01 3.76208365e-01 -1.34477139e-01 -5.73289394e-01
1.06082785e+00 -1.36716676e+00 -1.35299087e+00 2.33215362e-01
9.05734658e-01 -1.07320261e+00 4.35900748e-01 5.33875942e-01
-1.55548143e+00 -1.37560293e-01 -1.08699942e+00 1.78573877e-01
-7.27255583e-01 -3.14239264e-01 8.57801437e-01 1.43120694e+00
-1.60129595e+00 9.73287821e-01 -4.91727620e-01 6.59165755e-02
1.20568443e-02 8.24621081e-01 -2.64718989e-03 4.62812334e-01
-1.33193445e+00 1.08501506e+00 2.22979754e-01 -1.21242590e-01
-1.65216476e-01 -2.13068098e-01 -8.23704481e-01 -5.63443303e-01
-4.78693932e-01 -5.29636025e-01 1.44139910e+00 -2.59346128e-01
-1.65295815e+00 9.71794367e-01 -1.42069459e-01 -1.97814897e-01
6.14786744e-01 -1.28011424e-02 -8.31891418e-01 2.42498964e-01
-1.89849049e-01 5.76383233e-01 9.28263724e-01 -1.35132408e+00
-7.59897232e-01 -1.67359829e-01 -1.23336017e-01 2.66287565e-01
-1.25510961e-01 1.89734384e-01 2.11616129e-01 -1.12999000e-01
3.27147305e-01 -7.20919967e-01 -2.53068686e-01 -5.32041907e-01
-1.46512717e-01 -7.10518599e-01 7.70373225e-01 -4.51523662e-01
1.83705616e+00 -1.67618537e+00 -2.06720144e-01 3.98590982e-01
3.57815564e-01 -1.24705513e-03 2.40583986e-01 1.08380008e+00
-2.76906848e-01 7.32199550e-01 4.11965609e-01 -1.10722095e-01
2.24991128e-01 4.85861301e-01 -4.16602850e-01 3.05609167e-01
-1.29282743e-01 1.15307164e+00 -1.16605783e+00 -5.23096442e-01
4.11106765e-01 9.42391157e-02 -4.58732933e-01 4.79237735e-01
2.99364805e-01 1.47170946e-01 -3.56553018e-01 1.39399457e+00
1.15709066e+00 -1.31984919e-01 1.45911396e-01 5.30878425e-01
-3.79135728e-01 3.55090618e-01 -6.96863770e-01 1.05554795e+00
7.08333924e-02 5.00173986e-01 1.02364108e-01 -7.94621468e-01
3.33247900e-01 8.61540735e-01 4.37155962e-01 -9.67555881e-01
-2.26398129e-02 6.92409754e-01 1.12803578e-01 -6.07703328e-01
7.58228302e-01 3.94563079e-02 -3.64872098e-01 6.40070081e-01
-2.37588286e-01 -6.59476995e-01 9.64643434e-02 3.08100313e-01
6.22585893e-01 -5.18246442e-02 5.62923312e-01 -1.05613089e+00
7.86340594e-01 -1.82965681e-01 -1.81299388e-01 1.03415680e+00
-3.09923887e-01 3.19085121e-01 2.99841821e-01 -6.65102363e-01
-6.45341039e-01 -1.12307119e+00 -4.89381433e-01 1.30636716e+00
3.24267983e-01 -4.39044595e-01 -9.54439282e-01 -2.49762803e-01
1.77620783e-01 6.89606130e-01 -5.90509653e-01 3.84124845e-01
-5.03739953e-01 -8.32535863e-01 7.39044368e-01 3.45434904e-01
-5.07752821e-02 -1.33414865e+00 -6.58416986e-01 1.25490099e-01
-2.20292807e-01 -6.63697243e-01 -6.23428151e-02 4.48765576e-01
-1.35989368e+00 -5.18594682e-01 -6.66252747e-02 -8.20914626e-01
5.87345481e-01 2.46782884e-01 1.27047324e+00 5.39230824e-01
-2.31483161e-01 4.26904231e-01 -1.21292919e-01 -4.95818377e-01
-4.59671497e-01 -8.00336525e-02 5.28869390e-01 -5.87835789e-01
5.19427478e-01 -2.50617653e-01 -7.29350567e-01 5.37953973e-01
-6.88540697e-01 1.62748516e-01 1.79803044e-01 1.04410267e+00
1.35816500e-01 -9.34035778e-02 1.22507080e-01 -6.38007045e-01
8.72274399e-01 -1.69219792e-01 -3.78732830e-01 5.77745810e-02
-6.77108407e-01 -3.74140263e-01 3.21430594e-01 -3.25342178e-01
-1.01981449e+00 -4.87835288e-01 -9.82677937e-02 2.45538145e-01
1.11353043e-02 -1.46784872e-01 6.47139177e-02 -5.24923325e-01
8.02199244e-01 9.25758183e-02 1.99174434e-02 -6.80815242e-03
3.01039815e-01 7.09525108e-01 -6.82967342e-03 -6.68678164e-01
8.44880998e-01 4.91470337e-01 7.98524171e-02 -9.57177758e-01
-1.52186140e-01 -2.60129690e-01 -9.51962709e-01 -6.54426932e-01
6.56643391e-01 -6.78531289e-01 -9.10833478e-01 3.91110867e-01
-9.38691139e-01 -3.38627815e-01 -3.91645581e-01 4.25431967e-01
-1.01278400e+00 2.75717527e-02 -3.90154392e-01 -1.27895141e+00
-5.10977268e-01 -1.02017939e+00 9.43384409e-01 5.30070923e-02
-5.10597289e-01 -1.26927447e+00 5.87685481e-02 2.71537274e-01
1.81734428e-01 -1.73075795e-01 6.90226793e-01 -2.38256708e-01
-4.24233019e-01 -1.53791070e-01 2.34436691e-02 -1.39755070e-01
1.70832314e-02 4.95917559e-01 -9.81751978e-01 -5.31145096e-01
6.65065646e-02 -1.92070693e-01 -1.08835101e-01 6.52520418e-01
5.91872573e-01 -2.29931593e-01 -8.56000841e-01 5.40386558e-01
1.38545322e+00 3.85070026e-01 5.32770038e-01 7.28214979e-01
1.41836226e-01 5.53460240e-01 9.17806149e-01 4.63203549e-01
1.30579369e-02 3.28798652e-01 2.40537539e-01 1.49327129e-01
1.11720070e-01 -1.54819340e-01 3.77893507e-01 1.16112018e+00
-8.18235934e-01 -2.69281328e-01 -5.07867396e-01 4.42987174e-01
-1.72482407e+00 -1.40330648e+00 -4.32368398e-01 6.90478683e-01
6.25676990e-01 1.56016424e-01 -1.48347050e-01 3.35214496e-01
4.99015123e-01 -2.03574806e-01 -1.19133167e-01 -1.06291151e+00
-1.43546045e-01 3.15233678e-01 7.37729073e-01 1.00061214e+00
-7.20721722e-01 1.03317809e+00 1.29781246e+01 1.02230716e+00
2.21112028e-01 1.03134915e-01 5.16071796e-01 3.48020852e-01
-4.36954498e-01 -4.56139445e-02 -1.04416132e+00 2.72933897e-02
1.38140702e+00 -4.30666685e-01 6.85999811e-01 5.44219851e-01
3.44648361e-01 -4.23268199e-01 -1.26188684e+00 5.26221812e-01
9.73738134e-02 -1.40886843e+00 -2.83300440e-04 6.85225725e-01
7.73699820e-01 -5.08050561e-01 6.22419357e-01 3.24184299e-01
6.09259963e-01 -1.14389277e+00 8.60300779e-01 2.53660440e-01
1.03040910e+00 -6.05088234e-01 5.67372203e-01 1.68872893e-01
-1.14389896e+00 -2.20873043e-01 -8.77727985e-01 -1.00755692e+00
3.93533185e-02 -1.81779593e-01 -4.29956943e-01 3.48861217e-01
9.58353162e-01 2.99398601e-01 -3.93658698e-01 9.95779395e-01
-4.78694476e-02 1.04875881e-02 -2.71853864e-01 -4.48467314e-01
4.83122796e-01 -3.54241252e-01 4.66730654e-01 1.00164843e+00
2.48499006e-01 3.51035744e-01 -9.84472036e-02 4.01770771e-01
5.45058846e-01 3.29446048e-02 -1.19659424e+00 -1.78908288e-01
2.83276141e-01 9.16795909e-01 -4.83487815e-01 -4.22520459e-01
-2.00212970e-01 8.62069130e-01 -3.55488248e-02 5.01107454e-01
-6.11489356e-01 -4.35615242e-01 9.72222984e-01 -1.27327025e-01
-1.14700586e-01 -3.48497719e-01 -6.23769283e-01 -7.30352640e-01
-5.89872956e-01 -4.54965204e-01 5.93606755e-02 -5.53365827e-01
-1.39813089e+00 5.79277515e-01 -2.27688253e-02 -1.40553558e+00
-6.99901402e-01 -1.27676582e+00 -4.76714373e-01 4.92853165e-01
-1.11898029e+00 -1.10984349e+00 2.50124663e-01 4.52870727e-01
1.64141744e-01 -5.34416080e-01 1.39563632e+00 3.57715860e-02
1.00637585e-01 9.24474537e-01 6.69434488e-01 -7.37814724e-01
5.56605101e-01 -1.27867436e+00 5.68737745e-01 -1.37897313e-01
-4.31265175e-01 9.05828118e-01 6.28349900e-01 -5.39804697e-01
-1.41196322e+00 -3.66917729e-01 1.08350635e+00 -9.83769417e-01
6.55218959e-01 -3.86345625e-01 4.23767231e-02 7.88592756e-01
7.15902448e-01 -6.18741751e-01 8.21781039e-01 -1.83753878e-01
1.80774391e-01 5.75296998e-01 -1.39248300e+00 6.12354755e-01
1.66275799e+00 -4.63594139e-01 -6.25784039e-01 7.60327101e-01
8.13696027e-01 -6.94087505e-01 -1.30082703e+00 3.34633321e-01
8.65424156e-01 -8.75409484e-01 1.61978090e+00 -1.32660246e+00
-4.43697497e-02 2.81152606e-01 -2.61993498e-01 -9.32519078e-01
-5.96193194e-01 -1.23518765e+00 -5.33532679e-01 -5.83747849e-02
5.96577883e-01 -1.13057327e+00 3.42365682e-01 8.74560475e-01
-2.82833427e-01 -6.42737269e-01 -1.06996536e+00 -1.32016802e+00
-3.35779637e-02 -1.45572275e-01 4.90409225e-01 7.63798356e-01
6.83744550e-01 1.09839931e-01 -6.36873543e-02 -1.00294888e-01
5.33176839e-01 9.55312885e-03 4.41501856e-01 -1.34294486e+00
3.86843324e-01 -5.75816095e-01 -3.07655483e-01 -9.45992947e-01
-8.85957032e-02 -8.12076271e-01 -6.53862000e-01 -1.28511906e+00
-8.32044985e-03 -1.91056758e-01 -1.11109078e-01 -1.65725678e-01
3.67937148e-01 2.13746816e-01 1.20859891e-02 1.03788137e-01
-3.71160030e-01 6.18435517e-02 1.29639816e+00 6.91750320e-05
-1.62315920e-01 4.85058486e-01 -4.81304944e-01 7.84440815e-01
8.58408585e-02 -2.99253196e-01 -6.78878546e-01 6.11881316e-02
6.69384480e-01 4.61409837e-02 3.24159935e-02 -7.42885649e-01
5.37211418e-01 -3.75702560e-01 4.78586555e-01 -1.32223868e+00
1.30741090e-01 -9.61415648e-01 6.95283338e-02 9.40189242e-01
2.74610907e-01 1.20424610e-02 8.66204947e-02 5.39500564e-02
-1.44871444e-01 -5.70943117e-01 9.21121240e-01 -4.07591403e-01
-4.92852688e-01 -5.20386267e-03 -1.03226590e+00 8.97834301e-02
9.92593169e-01 -7.84614205e-01 -3.59281451e-01 -4.20183957e-01
-8.29068601e-01 -1.95836127e-02 6.50830388e-01 3.03609259e-02
7.30431557e-01 -1.51530886e+00 -2.30721906e-01 7.18729138e-01
-3.22939813e-01 -3.74400020e-01 -1.70157343e-01 6.58265352e-01
-1.32361674e+00 1.02442718e+00 -5.41665435e-01 -4.55340147e-01
-1.14228773e+00 4.78126436e-01 4.28307921e-01 -2.41845414e-01
-2.15481281e-01 1.11954463e+00 2.71224789e-02 -8.23025763e-01
1.85185194e-01 -9.21545625e-02 -7.54407048e-01 3.55081353e-03
6.88606799e-01 1.05194807e+00 -2.91290224e-01 -6.04341030e-01
-4.56784427e-01 6.65885091e-01 2.32151806e-01 -2.87484169e-01
9.17833567e-01 -2.19243199e-01 -9.89108324e-01 4.28274393e-01
8.38715494e-01 -1.36269778e-01 -3.69319022e-02 4.16855574e-01
1.25943512e-01 -8.02164078e-01 -4.32406247e-01 -3.55811834e-01
-1.85641110e-01 5.54822803e-01 5.34874737e-01 8.99602413e-01
8.57008278e-01 -3.02566767e-01 8.18335712e-01 9.66778398e-01
5.72402716e-01 -1.68019545e+00 -2.13140488e-01 6.89524531e-01
9.12339568e-01 -9.28350806e-01 5.44190466e-01 -7.27165341e-01
-4.14997995e-01 1.32979155e+00 4.68304873e-01 -1.55325383e-01
1.27306652e+00 5.74917436e-01 1.14069022e-02 -3.70670199e-01
-9.44949508e-01 1.12705544e-01 3.75366658e-01 1.11147714e+00
5.06513238e-01 5.07374525e-01 -9.66500878e-01 3.21953118e-01
-7.28706717e-01 -2.34555230e-01 4.90474731e-01 1.41972518e+00
-6.43810987e-01 -1.20391917e+00 -7.23931909e-01 4.61561680e-01
-5.49773455e-01 -1.16372630e-01 -5.28106689e-01 8.46754074e-01
-5.76629937e-02 1.49448860e+00 -1.97535474e-03 -5.03491640e-01
4.11356747e-01 1.41089618e-01 7.21762300e-01 -1.23501487e-01
-9.04846430e-01 4.15413082e-01 3.84890139e-01 -1.23056793e+00
-8.58632207e-01 -1.05834293e+00 -1.40667629e+00 -1.19437599e+00
-5.12782812e-01 1.89310342e-01 3.83317530e-01 3.90289724e-01
-2.06836104e-01 2.85260603e-02 9.80917513e-01 -1.07949340e+00
-3.90341938e-01 -9.39418614e-01 -1.00026262e+00 -7.84516707e-02
2.89751232e-01 -8.00943017e-01 -7.83523321e-01 2.83909619e-01]
|
[-7.519640922546387, 3.5545969009399414]
|
f27cf850-dd21-4a3f-bd65-7ed5e28e5598
|
vanillanet-the-power-of-minimalism-in-deep
|
2305.12972
| null |
https://arxiv.org/abs/2305.12972v2
|
https://arxiv.org/pdf/2305.12972v2.pdf
|
VanillaNet: the Power of Minimalism in Deep Learning
|
At the heart of foundation models is the philosophy of "more is different", exemplified by the astonishing success in computer vision and natural language processing. However, the challenges of optimization and inherent complexity of transformer models call for a paradigm shift towards simplicity. In this study, we introduce VanillaNet, a neural network architecture that embraces elegance in design. By avoiding high depth, shortcuts, and intricate operations like self-attention, VanillaNet is refreshingly concise yet remarkably powerful. Each layer is carefully crafted to be compact and straightforward, with nonlinear activation functions pruned after training to restore the original architecture. VanillaNet overcomes the challenges of inherent complexity, making it ideal for resource-constrained environments. Its easy-to-understand and highly simplified architecture opens new possibilities for efficient deployment. Extensive experimentation demonstrates that VanillaNet delivers performance on par with renowned deep neural networks and vision transformers, showcasing the power of minimalism in deep learning. This visionary journey of VanillaNet has significant potential to redefine the landscape and challenge the status quo of foundation model, setting a new path for elegant and effective model design. Pre-trained models and codes are available at https://github.com/huawei-noah/VanillaNet and https://gitee.com/mindspore/models/tree/master/research/cv/vanillanet.
|
['DaCheng Tao', 'Jianyuan Guo', 'Yunhe Wang', 'Hanting Chen']
|
2023-05-22
| null | null | null | null |
['philosophy']
|
['miscellaneous']
|
[-1.37283355e-01 2.41108820e-01 -1.61467746e-01 -2.46103197e-01
-8.62853676e-02 -4.31353450e-01 7.43699372e-01 -3.53005707e-01
-4.56430227e-01 3.31321895e-01 2.66517907e-01 -6.72144175e-01
-1.74210686e-02 -5.28376579e-01 -5.82937360e-01 -5.01137614e-01
2.28194427e-02 3.82087231e-02 -1.17682494e-01 -5.29902399e-01
8.48480612e-02 4.59628403e-01 -1.44258761e+00 7.86376521e-02
4.88581002e-01 8.18148792e-01 4.26779717e-01 6.77921832e-01
-2.96849851e-02 1.20774615e+00 -3.52598608e-01 -5.66987097e-01
3.37925792e-01 -1.53483674e-01 -7.46587873e-01 -3.38099062e-01
6.63583517e-01 -4.09870893e-01 -6.58155859e-01 7.83712268e-01
5.16263425e-01 -6.57489374e-02 1.89182058e-01 -1.28536367e+00
-1.10359609e+00 5.85131884e-01 -3.28473598e-01 4.40833569e-01
-2.68515408e-01 5.83989859e-01 1.38117135e+00 -1.06649423e+00
2.43175060e-01 1.12236404e+00 9.60566401e-01 7.89356351e-01
-1.24976015e+00 -6.53025746e-01 3.82310390e-01 2.46572182e-01
-1.42818773e+00 -8.32216978e-01 5.48913240e-01 -4.09429431e-01
1.21029246e+00 2.97424883e-01 8.80564332e-01 1.13737154e+00
2.29194060e-01 1.05944943e+00 6.49373591e-01 -3.90524507e-01
5.53856082e-02 1.07080698e-01 -6.25778884e-02 1.00715196e+00
1.94496021e-01 2.38412321e-01 -2.44629413e-01 3.35939229e-01
8.77381623e-01 1.81491345e-01 -4.16335672e-01 -3.49180579e-01
-9.43952680e-01 7.98193991e-01 7.25244820e-01 5.05385339e-01
-2.12749124e-01 5.78314543e-01 4.13609892e-01 2.67032444e-01
9.46080908e-02 8.83326113e-01 -5.00573277e-01 -1.96060821e-01
-1.06338823e+00 2.70901889e-01 6.58525050e-01 7.89006770e-01
6.05691850e-01 6.70181870e-01 1.56780973e-01 9.02742922e-01
3.10651600e-01 2.45328158e-01 5.47913373e-01 -1.27798676e+00
-9.24679041e-02 4.97208297e-01 -3.47772390e-01 -1.11710966e+00
-4.18732703e-01 -9.68797386e-01 -9.08522189e-01 4.80591953e-01
2.01158330e-01 -5.24633713e-02 -1.08196795e+00 1.72622812e+00
-1.55294493e-01 -1.03368640e-01 -1.00456730e-01 7.92764068e-01
8.41274321e-01 6.39288962e-01 1.62221164e-01 4.76288408e-01
1.25558484e+00 -1.34574270e+00 -2.54277319e-01 -7.23216772e-01
4.10513788e-01 -5.77978015e-01 1.36758089e+00 3.99182379e-01
-1.26544511e+00 -4.26188856e-01 -1.26413143e+00 -3.58234137e-01
-2.74366587e-01 -4.23974246e-02 9.89991486e-01 5.58960676e-01
-1.38477600e+00 6.79921031e-01 -9.27242637e-01 -4.25920695e-01
8.53738070e-01 4.54714864e-01 -3.32053661e-01 -9.16446224e-02
-9.11971986e-01 1.19702411e+00 3.37003201e-01 4.02293742e-01
-1.02541876e+00 -8.64219844e-01 -7.61613011e-01 1.45348817e-01
3.44345003e-01 -9.36270535e-01 1.59387100e+00 -1.08572686e+00
-1.37857497e+00 7.24051416e-01 3.22118774e-02 -7.06516504e-01
3.18885595e-01 -1.86839476e-01 -3.51029217e-01 -1.10068992e-01
-1.36260197e-01 9.42858160e-01 6.42719150e-01 -1.02050495e+00
-5.12655258e-01 -4.62754294e-02 3.29480886e-01 9.38718244e-02
-4.80402619e-01 -1.06771849e-01 -5.38844705e-01 -4.88641143e-01
-2.10656151e-01 -6.86183333e-01 -4.58117276e-01 2.15108097e-01
-3.26566160e-01 2.39686910e-02 7.61876941e-01 -5.04596651e-01
1.18791354e+00 -2.21205974e+00 -1.96141675e-01 -2.47101501e-01
8.34564626e-01 5.89978576e-01 -3.79202187e-01 4.73015308e-01
-1.66776568e-01 3.38366866e-01 -2.02910051e-01 -3.41091335e-01
1.40708670e-01 3.52449328e-01 -2.23510876e-01 3.17964464e-01
2.77678072e-01 1.32167184e+00 -9.04346883e-01 -2.14779660e-01
3.74185443e-01 7.37209857e-01 -7.37599134e-01 -4.69282381e-02
-5.60776517e-02 5.01139015e-02 -1.00634888e-01 7.48436511e-01
4.49955761e-01 -5.03062129e-01 7.60941803e-02 -3.18849325e-01
-3.65059584e-01 4.43943739e-01 -8.41782093e-01 1.57363749e+00
-4.90595073e-01 1.11013842e+00 4.95795727e-01 -1.05617845e+00
7.66762316e-01 3.66865918e-02 1.30365297e-01 -1.05419040e+00
4.33299392e-01 1.36784062e-01 1.96111068e-01 -4.72191095e-01
5.70285201e-01 -1.29898682e-01 1.84329644e-01 1.58050612e-01
1.73288807e-01 -2.68435925e-01 1.46079823e-01 2.51619309e-01
1.00938523e+00 1.66173652e-01 1.77158728e-01 -4.20069724e-01
8.77810866e-02 1.24978997e-01 5.67250311e-01 7.25188613e-01
-3.29759538e-01 5.28909564e-01 2.54561931e-01 -8.19762945e-01
-1.29818881e+00 -1.24343336e+00 -5.65424077e-02 1.22244215e+00
-3.22204351e-01 -6.49664402e-01 -5.91859519e-01 -1.18949853e-01
-3.41815166e-02 7.31471181e-01 -6.46857142e-01 -2.61911690e-01
-5.95871866e-01 -6.00569308e-01 7.49658406e-01 4.69452351e-01
6.99757159e-01 -9.07989621e-01 -8.23094189e-01 2.28544503e-01
6.93513155e-02 -9.34454083e-01 -1.29019111e-01 3.13423395e-01
-8.69781077e-01 -7.41998255e-01 -5.78628004e-01 -8.53409588e-01
4.40864593e-01 2.99945295e-01 1.43084288e+00 3.34430039e-01
-5.48661113e-01 1.13448232e-01 5.64627126e-02 -6.20781422e-01
-2.76756853e-01 2.92695791e-01 -5.17948437e-03 -4.52226669e-01
3.58946741e-01 -8.40173841e-01 -7.68862307e-01 -6.25781044e-02
-7.36058593e-01 3.75142634e-01 7.70185769e-01 7.98615575e-01
2.16052815e-01 -1.14965327e-01 3.65124464e-01 -6.61566556e-01
6.22808993e-01 -3.23885560e-01 -5.48950374e-01 -5.55439619e-03
-8.23537588e-01 -3.82721387e-02 6.67738378e-01 -2.23837912e-01
-6.48439050e-01 -3.79794776e-01 -2.54407346e-01 -4.38813686e-01
7.78899044e-02 4.95718509e-01 1.94056794e-01 -5.09451441e-02
9.50364530e-01 9.11971107e-02 2.60921955e-01 -5.26453257e-01
6.20676994e-01 4.58287358e-01 8.01871479e-01 -4.17185605e-01
8.38476896e-01 4.39997554e-01 -3.86178672e-01 -9.56284940e-01
-7.65167832e-01 -9.89531353e-03 -2.46585652e-01 8.14075172e-02
6.54586971e-01 -9.67941821e-01 -6.75895393e-01 3.10589224e-01
-1.00703335e+00 -6.51973307e-01 -5.90307534e-01 1.19894341e-01
-1.73080832e-01 1.44963771e-01 -6.47499502e-01 -5.28472662e-01
-5.19872606e-01 -9.78095531e-01 3.29183519e-01 4.99417573e-01
-3.67947012e-01 -1.06935239e+00 -1.86867472e-02 4.32325840e-01
9.74700928e-01 6.05419055e-02 8.86047781e-01 -4.17692751e-01
-7.52223611e-01 -1.17417604e-01 -3.29636902e-01 7.74495363e-01
-6.50266558e-02 1.99642882e-01 -1.15013409e+00 -2.41124138e-01
-1.11493908e-01 -3.46943438e-01 9.23943281e-01 2.70382762e-01
1.06887293e+00 -5.57125866e-01 -3.00715715e-02 9.74467099e-01
1.55740428e+00 9.95709375e-02 5.95180035e-01 6.40478551e-01
8.04208219e-01 3.38406414e-01 -2.41658494e-01 3.39215696e-01
6.79431081e-01 2.01578319e-01 7.08986044e-01 -4.26883817e-01
-2.34507963e-01 -2.22200572e-01 2.26712182e-01 7.86360860e-01
-8.68179649e-02 -1.00867003e-01 -1.12691534e+00 7.25047052e-01
-1.70941496e+00 -1.13515067e+00 9.92743596e-02 1.78746903e+00
7.11544335e-01 3.91591012e-01 -9.91181508e-02 -1.01395480e-01
2.14504614e-01 4.61456001e-01 -4.77313548e-01 -8.55202198e-01
-1.46845788e-01 3.16945463e-01 4.39466298e-01 4.63247538e-01
-9.10108268e-01 1.20473778e+00 6.56386757e+00 7.35543847e-01
-1.40738070e+00 1.32632971e-01 6.41966224e-01 -5.02279758e-01
-3.81640851e-01 1.14444084e-01 -6.74178898e-01 2.05080926e-01
1.01478875e+00 -2.83886671e-01 7.15092957e-01 8.83546472e-01
2.95482308e-01 1.71108171e-01 -9.87079561e-01 8.38146508e-01
-4.35613431e-02 -1.96513653e+00 -7.06597045e-02 1.31221060e-02
4.54290688e-01 9.69127655e-01 4.11628813e-01 4.82328385e-01
6.91822231e-01 -1.45810759e+00 9.95510519e-01 1.71085760e-01
6.76729560e-01 -4.48374361e-01 3.74330312e-01 1.56052023e-01
-9.68333960e-01 -4.45502967e-01 -5.08027852e-01 -4.67375427e-01
-7.52617940e-02 3.98928702e-01 -7.54971087e-01 6.05386868e-03
8.77536118e-01 5.96709907e-01 -6.24687135e-01 1.00487399e+00
-1.65664807e-01 5.07828534e-01 -3.44176263e-01 6.08288264e-03
6.70137942e-01 1.97088923e-02 3.24817955e-01 1.40626860e+00
-5.81550151e-02 -1.45341516e-01 -9.23063755e-02 1.05876338e+00
-1.19273745e-01 -1.71142537e-02 -8.00614476e-01 -2.28670999e-01
5.71685016e-01 1.34328008e+00 -4.43303138e-01 -1.68722495e-01
-5.81065416e-01 5.59966028e-01 3.81124020e-01 4.80135500e-01
-8.44449937e-01 -2.78907448e-01 7.98460662e-01 1.40565380e-01
4.13749427e-01 -3.78381491e-01 -4.99654561e-01 -1.05158734e+00
-9.37627703e-02 -1.10225689e+00 9.23321024e-02 -7.61352539e-01
-9.33312237e-01 8.03524733e-01 -3.52236241e-01 -7.31104434e-01
-1.98721308e-02 -7.89003551e-01 -9.26113069e-01 7.41547227e-01
-1.56633723e+00 -1.28350294e+00 -4.21359003e-01 4.32112992e-01
6.11598253e-01 -1.67928681e-01 7.68680394e-01 3.06793571e-01
-7.58337438e-01 8.34193945e-01 1.24586977e-01 2.67467529e-01
2.14068875e-01 -1.01144040e+00 9.32205975e-01 7.98366308e-01
2.15197489e-01 7.98305750e-01 7.52900064e-01 -3.11563201e-02
-1.40079165e+00 -8.21592629e-01 7.51993656e-01 -5.80655813e-01
8.80023360e-01 -4.94838506e-01 -8.00833166e-01 8.88964593e-01
4.51874226e-01 -1.60257965e-01 5.23809016e-01 1.39308840e-01
-6.90930486e-01 -2.16374323e-01 -8.05073202e-01 1.08112955e+00
9.63223338e-01 -6.13355160e-01 -5.66035986e-01 1.60359800e-01
8.54651809e-01 -8.32669586e-02 -5.42573750e-01 2.92736083e-01
8.12097549e-01 -1.13694859e+00 1.04694557e+00 -3.72173131e-01
4.47906286e-01 -1.23130842e-04 -2.43665069e-01 -1.02542067e+00
-6.66870117e-01 -1.01880920e+00 -3.00473012e-02 1.01015329e+00
6.89344525e-01 -7.17272758e-01 9.57751811e-01 6.85813248e-01
-4.30751085e-01 -1.20481312e+00 -7.41623819e-01 -7.05228806e-01
3.01917642e-01 -6.06227934e-01 4.53482032e-01 1.01113033e+00
-1.50407076e-01 5.42343080e-01 -1.79815829e-01 -1.80483893e-01
4.91982669e-01 -3.99389207e-01 6.71558082e-01 -1.07212722e+00
-5.98539710e-01 -1.06174386e+00 -3.21823537e-01 -1.03027546e+00
-2.67919421e-01 -1.04280508e+00 -4.05944288e-01 -1.77483273e+00
-4.35663834e-02 -4.60712135e-01 -2.86484778e-01 7.81498432e-01
3.93204719e-01 5.36528349e-01 6.09600663e-01 3.32631707e-01
-3.42241675e-01 4.38980877e-01 1.19118309e+00 -2.80063808e-01
-2.16661274e-01 -3.20457667e-01 -1.46509361e+00 8.22444260e-01
1.00705564e+00 -8.10033977e-02 -5.41417122e-01 -1.04108357e+00
2.79226512e-01 -5.99583149e-01 5.81912220e-01 -1.12653756e+00
3.42697591e-01 -5.10700271e-02 2.81576365e-01 -1.23232991e-01
4.87453908e-01 -6.69910014e-01 6.88356087e-02 5.09726465e-01
-2.79635727e-01 3.61397743e-01 5.31640351e-01 2.51021888e-02
6.06594719e-02 -1.29581988e-01 8.88946235e-01 -5.04852116e-01
-1.02545977e+00 2.89529443e-01 -3.24937850e-01 1.81874856e-01
7.41374552e-01 -5.38561761e-01 -6.01431429e-01 -2.50194848e-01
-5.29625893e-01 4.41900134e-01 5.69273591e-01 5.58391213e-01
5.79345882e-01 -9.71569479e-01 -6.39605463e-01 3.50881577e-01
-7.49841481e-02 1.25573397e-01 3.16325516e-01 7.11991906e-01
-8.40279877e-01 3.84599268e-01 -4.26143467e-01 -4.07440364e-01
-8.14694643e-01 4.21938419e-01 6.38292491e-01 -1.40123563e-02
-8.89215291e-01 1.24407351e+00 3.34115267e-01 -3.07134330e-01
2.72581398e-01 -3.26637149e-01 4.38570119e-02 -1.79070622e-01
5.44644356e-01 2.29054257e-01 -1.14941434e-03 -2.81671613e-01
-3.99151981e-01 2.32434422e-01 -4.12761688e-01 2.39075366e-02
1.57927442e+00 3.08585111e-02 -1.87952787e-01 2.13439018e-01
1.09254742e+00 -2.65671760e-01 -1.47975862e+00 -1.38259098e-01
-1.29726842e-01 -7.64960200e-02 4.74128842e-01 -1.02009225e+00
-1.16554248e+00 1.12997365e+00 4.14896280e-01 4.22477052e-02
1.02513409e+00 -1.04456566e-01 6.97929025e-01 5.66604435e-01
1.74743291e-02 -1.02312052e+00 3.31347920e-02 7.44231641e-01
8.13520491e-01 -9.19481039e-01 -1.70105044e-02 2.51562893e-01
-4.44163442e-01 8.15018475e-01 8.06523442e-01 -2.13617861e-01
6.72205925e-01 4.69945043e-01 2.66155839e-01 -4.14531738e-01
-1.04503644e+00 -6.00282624e-02 -2.16826648e-01 6.81726336e-01
4.77471948e-01 -1.28682787e-02 1.24699965e-01 3.80688727e-01
-6.43326044e-01 -1.75288804e-02 5.49332201e-01 8.23554814e-01
-5.94837487e-01 -8.26517999e-01 -1.03991657e-01 3.57584983e-01
-5.25133729e-01 -6.09720767e-01 -1.36856586e-01 1.11309266e+00
1.27745584e-01 7.87830293e-01 1.47522837e-01 -5.72343171e-01
3.68951499e-01 4.06409092e-02 4.08565670e-01 -4.34658408e-01
-6.85980618e-01 -1.09117083e-01 1.05919637e-01 -6.29044533e-01
1.48523867e-01 -2.34401226e-01 -1.12735868e+00 -9.51664627e-01
-3.12120188e-02 9.21010897e-02 8.06915104e-01 7.60812521e-01
6.55991077e-01 5.57067156e-01 3.10975283e-01 -9.42427635e-01
-7.32443929e-01 -5.69047213e-01 -2.29980424e-01 -1.92031637e-01
6.31862283e-01 -2.61963516e-01 -2.43535176e-01 -1.01781443e-01]
|
[8.90909194946289, 2.6294686794281006]
|
d06b27a3-68b8-45e6-8750-59e0b11169ef
|
region-proposal-networks-with-contextual
|
1812.1033
| null |
http://arxiv.org/abs/1812.10330v1
|
http://arxiv.org/pdf/1812.10330v1.pdf
|
Region Proposal Networks with Contextual Selective Attention for Real-Time Organ Detection
|
State-of-the-art methods for object detection use region proposal networks
(RPN) to hypothesize object location. These networks simultaneously predicts
object bounding boxes and \emph{objectness} scores at each location in the
image. Unlike natural images for which RPN algorithms were originally designed,
most medical images are acquired following standard protocols, thus organs in
the image are typically at a similar location and possess similar geometrical
characteristics (e.g. scale, aspect-ratio, etc.). Therefore, medical image
acquisition protocols hold critical localization and geometric information that
can be incorporated for faster and more accurate detection. This paper presents
a novel attention mechanism for the detection of organs by incorporating
imaging protocol information. Our novel selective attention approach (i)
effectively shrinks the search space inside the feature map, (ii) appends
useful localization information to the hypothesized proposal for the detection
architecture to learn where to look for each organ, and (iii) modifies the
pyramid of regression references in the RPN by incorporating organ- and
modality-specific information, which results in additional time reduction. We
evaluated the proposed framework on a dataset of 768 chest X-ray images
obtained from a diverse set of sources. Our results demonstrate superior
performance for the detection of the lung field compared to the
state-of-the-art, both in terms of detection accuracy, demonstrating an
improvement of $>7\%$ in Dice score, and reduced processing time by $27.53\%$
due to fewer hypotheses.
|
['Antonio R. Porras', 'Awais Mansoor', 'Marius George Linguraru']
|
2018-12-26
| null | null | null | null |
['organ-detection']
|
['medical']
|
[ 2.02917919e-01 5.76512180e-02 -1.94829658e-01 -2.53601819e-01
-7.06602573e-01 -4.01496649e-01 3.03329557e-01 6.38981700e-01
-5.37039042e-01 3.46180707e-01 -5.86269833e-02 -2.03593954e-01
-2.82661468e-01 -6.96368158e-01 -5.73977649e-01 -7.25784004e-01
-5.99620454e-02 3.89422417e-01 5.48817873e-01 3.06071669e-01
2.89945096e-01 9.04896915e-01 -1.14257717e+00 1.27860174e-01
5.26042938e-01 1.17093122e+00 7.19879627e-01 5.74196994e-01
3.55507433e-02 6.32280350e-01 -3.69473249e-01 -1.11240618e-01
3.02013308e-01 -2.56066084e-01 -5.23266137e-01 4.50180955e-02
4.95636076e-01 -3.99380654e-01 -5.89645982e-01 9.54464555e-01
6.28146112e-01 1.36557087e-01 8.52714062e-01 -6.26446009e-01
-5.26437044e-01 3.17925513e-01 -7.52949893e-01 6.09406829e-01
-9.04627815e-02 1.80722713e-01 9.58942771e-01 -9.99777675e-01
4.30157185e-01 8.72896314e-01 6.03835702e-01 3.89320314e-01
-8.85713577e-01 -5.38235068e-01 -1.29481733e-01 3.94379310e-02
-1.62636900e+00 1.64794661e-02 6.65741384e-01 -3.09866041e-01
7.92222142e-01 3.76355290e-01 4.52206701e-01 3.95952195e-01
4.19687688e-01 6.10155642e-01 7.53629506e-01 -4.71542299e-01
-3.92830558e-02 4.04474765e-01 -2.29362175e-02 1.13783348e+00
4.37146217e-01 -9.83290374e-02 -1.64698720e-01 -6.67116418e-02
1.23113787e+00 5.64324260e-01 -3.28970045e-01 -4.88152057e-01
-1.42380655e+00 7.35569358e-01 1.12780046e+00 3.88013631e-01
-7.77277052e-01 9.73195583e-02 2.23042727e-01 -6.46849751e-01
1.07670650e-01 5.56568503e-01 -2.26919889e-01 4.83356297e-01
-7.42262959e-01 -4.71668318e-02 3.86435956e-01 7.55415976e-01
3.40852290e-01 -2.11665913e-01 -5.91790080e-01 6.41903102e-01
4.04494166e-01 6.09121621e-01 4.15807068e-01 -6.73180401e-01
4.81300801e-01 8.03080261e-01 -5.85492626e-02 -1.02784562e+00
-8.20315123e-01 -9.06986058e-01 -9.02130246e-01 -7.51004964e-02
4.03295457e-01 5.91440313e-02 -1.13152826e+00 1.39782810e+00
4.49384034e-01 -8.33030343e-02 -3.20320427e-01 1.09939420e+00
1.03148389e+00 4.93603885e-01 2.20114499e-01 -1.22706875e-01
1.82290101e+00 -8.96463275e-01 -3.65269899e-01 -9.47293416e-02
3.69890392e-01 -7.17064559e-01 8.39936316e-01 4.05827537e-02
-1.19808412e+00 -6.77200556e-01 -1.01042509e+00 1.66363046e-01
-1.12869784e-01 4.68157828e-01 4.79327381e-01 4.98450011e-01
-7.61173487e-01 4.21983778e-01 -8.22106779e-01 -4.50021207e-01
7.86077082e-01 6.54354393e-01 -1.77793965e-01 -9.32840779e-02
-6.04472578e-01 8.57944727e-01 3.74296874e-01 1.32754087e-01
-8.17497790e-01 -9.29087877e-01 -6.97599173e-01 3.70300442e-01
5.40105999e-01 -7.87981808e-01 9.82068002e-01 -3.90083522e-01
-1.07279551e+00 6.28982723e-01 -2.94622276e-02 -3.60754818e-01
4.00400013e-01 5.32489792e-02 -1.04971491e-01 6.53267026e-01
1.17990837e-01 9.45314944e-01 6.93685770e-01 -1.12199366e+00
-7.01186836e-01 -3.45841318e-01 -5.33024110e-02 3.42734873e-01
-2.37066761e-01 9.56087559e-02 -6.78541958e-01 -6.70726061e-01
5.31475008e-01 -7.91930974e-01 -4.07571554e-01 4.07169193e-01
-4.58442390e-01 -1.72327176e-01 7.05873787e-01 -5.87861598e-01
1.08625853e+00 -2.00490737e+00 -2.73583680e-01 3.85019332e-01
4.69846159e-01 1.71768770e-01 1.73674017e-01 -2.00936899e-01
-2.29545496e-02 9.93263628e-03 -1.64598852e-01 -2.58984447e-01
-3.85757357e-01 8.21796805e-02 9.25575867e-02 8.18775117e-01
7.50706345e-02 9.50833142e-01 -7.13720798e-01 -1.03481555e+00
5.81469595e-01 5.76361775e-01 -5.05347550e-01 1.41668886e-01
2.01307386e-01 4.33327526e-01 -6.02892816e-01 8.04257572e-01
5.54324925e-01 -5.68228126e-01 7.26352632e-03 -5.06333172e-01
-1.09319799e-01 1.79346185e-02 -1.11689913e+00 1.39482212e+00
-5.21091759e-01 4.09556180e-01 -4.58239168e-02 -6.16474092e-01
7.28118479e-01 4.28102165e-01 8.16496670e-01 -4.45821583e-01
3.05975050e-01 1.01653546e-01 3.29594940e-01 -4.87194836e-01
1.60728827e-01 -1.03846721e-01 1.96628317e-01 3.38552266e-01
-1.47601375e-02 -1.10688761e-01 1.02416970e-01 -4.88227792e-02
9.28566813e-01 -4.08765823e-01 5.56658030e-01 -2.81267583e-01
7.62585819e-01 -1.23742692e-01 4.13510442e-01 1.04409766e+00
-3.52510303e-01 7.93385923e-01 1.36232704e-01 -5.19430935e-01
-9.14937913e-01 -1.19668961e+00 -4.96345937e-01 8.13086331e-01
3.01080823e-01 1.30351737e-01 -5.77506363e-01 -9.11110401e-01
-1.15296915e-01 4.69634861e-01 -7.37900794e-01 2.25200117e-01
-8.67845476e-01 -5.87549627e-01 2.39148170e-01 8.51917982e-01
4.71494853e-01 -1.28053117e+00 -1.16318703e+00 1.56059237e-02
-9.24572721e-02 -9.98815477e-01 -5.37467837e-01 2.15879306e-01
-1.08859158e+00 -1.14001608e+00 -8.32671821e-01 -8.44183624e-01
1.23061085e+00 3.57072830e-01 9.55796480e-01 3.11528653e-01
-8.60744059e-01 2.74306029e-01 -3.71228009e-02 -5.37424147e-01
-1.28103897e-01 3.36767249e-02 -8.29875022e-02 -2.58268535e-01
-6.15028478e-02 5.87852299e-02 -1.09332192e+00 4.57015902e-01
-8.22669864e-01 -1.59871593e-01 9.81395364e-01 8.96796346e-01
8.89624774e-01 1.21352419e-01 2.46453017e-01 -6.34512186e-01
9.28982720e-02 -2.38578543e-01 -6.84284806e-01 3.35534751e-01
-4.18427736e-01 -1.76317096e-01 4.82569158e-01 -3.17173064e-01
-9.11226988e-01 1.69617891e-01 2.42447004e-01 -5.29254436e-01
-2.00572401e-01 7.34915137e-02 2.32860014e-01 -3.60785037e-01
5.71248770e-01 2.38058686e-01 -7.39966854e-02 -1.40250355e-01
1.26550972e-01 2.94420004e-01 5.26589990e-01 -1.88240752e-01
7.06878126e-01 7.08988070e-01 4.18583095e-01 -8.41720700e-01
-7.11855233e-01 -7.37876117e-01 -8.36493850e-01 -1.83774158e-01
1.00431705e+00 -5.81626058e-01 -8.08952510e-01 -1.73049718e-01
-1.06757796e+00 3.23300242e-01 -3.98894161e-01 9.79210258e-01
-2.47784525e-01 3.24272722e-01 -6.08565867e-01 -7.60630906e-01
-7.09563196e-01 -1.25988674e+00 9.27533269e-01 5.03947973e-01
-3.73760611e-02 -8.31465304e-01 -3.94814909e-01 2.40223870e-01
5.22355378e-01 -3.22684906e-02 9.70063865e-01 -6.83805585e-01
-8.32722545e-01 -4.61854279e-01 -7.61667013e-01 5.41497804e-02
1.45044476e-01 -1.84576377e-01 -7.37749994e-01 -2.99685866e-01
1.81706712e-01 2.60507405e-01 7.23964095e-01 9.74043071e-01
1.49098015e+00 -7.92962238e-02 -6.39121771e-01 5.56829095e-01
1.45008790e+00 3.91684741e-01 3.08828861e-01 1.01364642e-01
6.82875574e-01 5.22780538e-01 5.39562523e-01 3.89642715e-01
4.96675447e-03 5.06640136e-01 7.53085673e-01 -5.90805352e-01
-2.42220238e-01 2.51633506e-02 -2.72280455e-01 4.29580897e-01
-1.52050033e-01 -1.88435301e-01 -9.58461940e-01 6.48584545e-01
-1.39350390e+00 -6.39471889e-01 -9.09253024e-03 2.10819983e+00
3.34499121e-01 9.02161598e-02 5.63519076e-02 -3.19148690e-01
8.15122128e-01 -7.65415728e-02 -6.12530351e-01 1.76382422e-01
2.82663167e-01 2.97540665e-01 7.96719790e-01 1.26756519e-01
-1.30478573e+00 4.26266462e-01 5.92208242e+00 6.23776317e-01
-1.19999397e+00 1.06183335e-01 7.12135613e-01 -1.93339929e-01
2.44410530e-01 -2.65168488e-01 -8.41731131e-01 1.51638806e-01
3.75964135e-01 2.20226511e-01 -1.15341082e-01 8.57095480e-01
1.27628610e-01 -1.27246737e-01 -1.07572854e+00 9.25910175e-01
2.72139817e-01 -1.37269914e+00 -1.06984876e-01 1.16856985e-01
4.45750058e-01 1.49732485e-01 1.61757216e-01 2.39505302e-02
-4.49167967e-01 -9.54526186e-01 4.60909605e-01 3.90559196e-01
6.69404685e-01 -6.00095272e-01 1.02627623e+00 2.40358397e-01
-1.26584065e+00 -1.80602148e-01 -5.12839317e-01 5.56309938e-01
8.23497176e-02 2.41787180e-01 -1.43133283e+00 4.46515173e-01
7.52783239e-01 2.02031121e-01 -7.38415062e-01 1.38785052e+00
-1.55780971e-01 3.85226279e-01 -4.59355026e-01 -2.35529870e-01
2.45000392e-01 1.13926388e-01 5.25194347e-01 1.18839478e+00
3.57724160e-01 2.68665045e-01 2.18357056e-01 9.17965531e-01
-6.73365369e-02 3.68735880e-01 -2.82094777e-01 4.42228317e-01
3.10167700e-01 1.50182664e+00 -1.32228303e+00 -3.81441444e-01
-2.93641657e-01 5.61956704e-01 -9.43771005e-02 4.91644256e-02
-1.06519949e+00 -3.34964722e-01 2.36656442e-02 5.44040680e-01
6.60795271e-01 9.33877155e-02 -2.38550097e-01 -6.39796257e-01
-2.28678696e-02 -2.88033098e-01 5.94776869e-01 -6.74122989e-01
-1.08646119e+00 7.23953485e-01 3.05473190e-02 -1.24765205e+00
1.13335185e-01 -7.03478515e-01 -5.45314550e-01 8.04547071e-01
-1.32289553e+00 -1.16209137e+00 -4.73635077e-01 3.56514215e-01
6.51079357e-01 4.14636135e-02 6.06771648e-01 2.58616269e-01
-5.83887994e-01 5.64401090e-01 6.55830875e-02 3.15050036e-01
4.87026483e-01 -1.20828617e+00 -8.04973245e-02 6.72584772e-01
1.53625563e-01 7.62199581e-01 3.13313365e-01 -4.85806555e-01
-1.22087324e+00 -1.09908545e+00 5.10048330e-01 -3.92682254e-01
3.18319738e-01 -1.37058601e-01 -7.50979364e-01 3.15770596e-01
-5.33877574e-02 5.19674599e-01 5.08099496e-01 -2.70198166e-01
2.46624202e-02 -1.78556308e-01 -1.47879148e+00 3.91947776e-01
5.18745482e-01 -2.11001426e-01 -4.62518692e-01 5.33528745e-01
5.44077396e-01 -5.30936241e-01 -9.45285618e-01 5.57526112e-01
6.05484009e-01 -8.73658121e-01 1.19902194e+00 -2.36292303e-01
1.52932420e-01 -4.43869233e-01 -1.18656717e-01 -5.56219697e-01
-6.22676611e-01 -5.22445440e-02 -3.78162041e-02 7.03878999e-01
4.88558829e-01 -4.69882578e-01 9.82408226e-01 4.25744027e-01
-2.42945135e-01 -1.23448896e+00 -9.74285722e-01 -2.97801077e-01
-1.71850339e-01 -3.10698748e-01 2.50286639e-01 5.09999335e-01
-4.61258650e-01 7.37605542e-02 1.47543341e-01 3.93799901e-01
6.69937730e-01 1.01397201e-01 2.96613216e-01 -9.80272114e-01
-3.08102638e-01 -5.60809493e-01 -4.60538030e-01 -8.89063537e-01
-4.36957359e-01 -8.47749114e-01 6.22714162e-02 -1.67631030e+00
5.94135106e-01 -5.32283604e-01 -5.85674405e-01 4.23989922e-01
-1.53873414e-01 4.84469682e-01 2.93696284e-01 3.82634252e-01
-5.19508600e-01 1.07612155e-01 1.36168325e+00 2.21780278e-02
-1.49507627e-01 2.46872753e-01 -5.47453582e-01 9.29959357e-01
6.07445419e-01 -5.87491572e-01 -1.89836785e-01 -1.23111419e-01
-3.13899875e-01 2.20999554e-01 4.87632006e-01 -1.24112058e+00
3.47879410e-01 1.30973890e-01 9.84246075e-01 -1.05884159e+00
3.67071867e-01 -9.57682133e-01 -3.33389997e-01 8.77992153e-01
-2.19469592e-01 4.57922891e-02 2.26747379e-01 6.02681935e-01
4.89806607e-02 -3.91422451e-01 1.08303022e+00 -4.21898007e-01
-3.51478726e-01 3.51939052e-01 -2.61126328e-02 -3.18392664e-01
1.30026448e+00 -4.06283975e-01 -1.10923953e-01 2.31400169e-02
-6.47522688e-01 6.36136010e-02 1.76581159e-01 8.60949140e-03
7.38908947e-01 -8.97225976e-01 -5.73785663e-01 2.91583300e-01
7.53616020e-02 2.71939278e-01 4.16430891e-01 1.22988617e+00
-8.43172669e-01 7.52033472e-01 2.43384987e-02 -1.04732549e+00
-1.31236172e+00 5.90764642e-01 5.02668440e-01 -4.22583461e-01
-7.04299092e-01 1.09407353e+00 6.68622792e-01 -1.86456993e-01
2.21472204e-01 -6.22322023e-01 -2.44054049e-01 -1.67332783e-01
3.83619219e-01 3.37465316e-01 1.57521084e-01 -6.48122132e-01
-6.32183790e-01 8.02795291e-01 -3.63015175e-01 2.52399266e-01
1.16834712e+00 4.22970802e-02 1.78256258e-02 -1.11458093e-01
1.00639486e+00 -9.78928208e-02 -1.11847150e+00 -3.50567847e-01
-3.25751811e-01 -5.38943052e-01 2.56314188e-01 -8.90272737e-01
-1.09565771e+00 8.46297026e-01 1.00653076e+00 -6.94331527e-02
1.06564891e+00 3.89984041e-01 4.59913641e-01 1.93367302e-01
9.66518074e-02 -7.11775124e-01 3.03030819e-01 8.90266076e-02
7.40081668e-01 -1.42365897e+00 4.86577809e-01 -4.40460712e-01
-6.30962968e-01 1.15298092e+00 7.92500556e-01 -9.96429920e-02
6.36065125e-01 5.68987243e-03 -9.61414054e-02 -5.37965775e-01
-2.92622685e-01 5.22271544e-02 8.44545066e-01 3.61212462e-01
4.48810488e-01 -3.14531028e-02 -5.85403517e-02 3.91419590e-01
2.36044675e-01 -4.81343508e-01 1.35604724e-01 8.91484737e-01
-6.17079675e-01 -4.38495755e-01 -7.73416162e-01 7.36486793e-01
-6.54184937e-01 -2.30445918e-02 1.03411198e-01 9.74692345e-01
3.40351433e-01 5.18907487e-01 2.46026650e-01 2.31264234e-01
3.64524215e-01 -4.05479401e-01 5.43639004e-01 -7.28523433e-01
-7.00254321e-01 2.54034430e-01 -4.91016150e-01 -4.66349125e-01
-1.03072286e-01 -5.86055219e-01 -1.55175209e+00 2.85584301e-01
-6.57789409e-01 -5.15530780e-02 8.06264639e-01 7.01702416e-01
1.54769912e-01 7.97209263e-01 4.99799520e-01 -9.08364713e-01
-5.87688565e-01 -9.11526203e-01 -3.88600141e-01 2.00135425e-01
1.98477045e-01 -7.34623194e-01 -9.28594097e-02 -2.46656537e-01]
|
[15.153302192687988, -2.308587074279785]
|
e38d96ee-3134-4f42-b98e-f5789bcf2c27
|
vidosat-high-dimensional-sparsifying
|
1710.00947
| null |
http://arxiv.org/abs/1710.00947v1
|
http://arxiv.org/pdf/1710.00947v1.pdf
|
VIDOSAT: High-dimensional Sparsifying Transform Learning for Online Video Denoising
|
Techniques exploiting the sparsity of images in a transform domain have been
effective for various applications in image and video processing. Transform
learning methods involve cheap computations and have been demonstrated to
perform well in applications such as image denoising and medical image
reconstruction. Recently, we proposed methods for online learning of
sparsifying transforms from streaming signals, which enjoy good convergence
guarantees, and involve lower computational costs than online synthesis
dictionary learning. In this work, we apply online transform learning to video
denoising. We present a novel framework for online video denoising based on
high-dimensional sparsifying transform learning for spatio-temporal patches.
The patches are constructed either from corresponding 2D patches in successive
frames or using an online block matching technique. The proposed online video
denoising requires little memory, and offers efficient processing. Numerical
experiments compare the performance to the proposed video denoising scheme but
fixing the transform to be 3D DCT, as well as prior schemes such as dictionary
learning-based schemes, and the state-of-the-art VBM3D and VBM4D on several
video data sets, demonstrating the promising performance of the proposed
methods.
|
['Saiprasad Ravishankar', 'Bihan Wen', 'Yoram Bresler']
|
2017-10-03
| null | null | null | null |
['video-denoising']
|
['computer-vision']
|
[ 4.04528320e-01 -3.83636087e-01 -8.14648867e-02 -2.23677337e-01
-1.05607188e+00 -6.33601705e-03 1.79964349e-01 5.05579561e-02
-3.77093703e-01 3.23366195e-01 1.40966535e-01 -7.11703226e-02
-2.03822583e-01 -5.33604622e-01 -8.40109468e-01 -1.09092033e+00
-2.78304666e-01 1.34871051e-01 3.43796164e-01 -1.59575865e-01
1.70367081e-02 1.02842353e-01 -1.48729813e+00 3.80825728e-01
5.93622923e-01 1.27537370e+00 2.47592896e-01 4.83636200e-01
2.62443498e-02 6.45594835e-01 -1.56977460e-01 -2.38614619e-01
7.26079345e-01 -6.28056586e-01 -2.63037622e-01 6.10438466e-01
7.23917305e-01 -5.69385111e-01 -8.54145467e-01 1.11235416e+00
6.68183982e-01 2.86652118e-01 2.16924697e-01 -7.26046085e-01
-5.06849110e-01 2.75469661e-01 -7.40099192e-01 4.07338500e-01
3.17467868e-01 -2.13318735e-01 7.26792395e-01 -1.06672299e+00
6.84456170e-01 1.16927946e+00 9.31531191e-01 3.07889551e-01
-1.27014720e+00 -5.50199628e-01 -4.49390747e-02 6.73102498e-01
-1.23536229e+00 -7.58747637e-01 9.54363167e-01 -1.92995906e-01
8.59215498e-01 9.50211566e-03 8.58319223e-01 6.44232333e-01
5.23746967e-01 9.31764781e-01 1.15569198e+00 -5.30372739e-01
2.74839580e-01 -4.39345717e-01 -3.18015039e-01 8.21603417e-01
-3.74427177e-02 1.62818000e-01 -9.42586601e-01 -2.85405517e-01
1.08761358e+00 4.16198105e-01 -3.93410355e-01 -2.62427449e-01
-1.18097544e+00 8.71810853e-01 2.95983795e-02 2.83353895e-01
-7.50925243e-01 9.47963893e-02 6.59293354e-01 8.56765926e-01
9.86348808e-01 -3.30332696e-01 -3.49992603e-01 2.91387718e-02
-1.35040057e+00 -1.33898348e-01 6.39026225e-01 8.65120232e-01
7.02810466e-01 3.99343878e-01 1.65873021e-01 8.50680113e-01
1.72970697e-01 6.74334884e-01 4.47382480e-01 -1.17771089e+00
4.29595590e-01 -2.83087432e-01 -1.70024171e-01 -1.32161558e+00
-3.42969559e-02 -2.28155181e-02 -1.28436673e+00 1.50610618e-02
1.90484777e-01 2.35285789e-01 -8.55954409e-01 1.29213953e+00
6.31346881e-01 9.26759601e-01 -6.59346730e-02 8.66201103e-01
6.23310626e-01 9.50242341e-01 -2.61271179e-01 -9.52142656e-01
8.51027489e-01 -6.87968314e-01 -1.09729946e+00 2.27523595e-01
4.15380090e-01 -9.95366037e-01 5.88921785e-01 9.72459972e-01
-1.32349086e+00 -8.03470969e-01 -7.31933713e-01 -1.12166973e-02
4.83223170e-01 -1.56037584e-01 5.21056294e-01 5.23339033e-01
-1.30536973e+00 8.37850034e-01 -1.40081167e+00 -2.46622220e-01
5.19601166e-01 3.28797728e-01 -2.74543583e-01 -4.85272497e-01
-8.76010299e-01 5.14537513e-01 6.52699471e-02 1.83226883e-01
-1.39462304e+00 -7.18032002e-01 -1.14707768e+00 -1.53782606e-01
3.06150854e-01 -5.72771013e-01 8.40217233e-01 -1.02430129e+00
-1.50441039e+00 8.31795633e-01 -4.44380939e-01 -6.06015384e-01
3.23707193e-01 -3.02635133e-01 -3.48818690e-01 6.99665010e-01
1.77894384e-02 2.68151850e-01 1.71628141e+00 -8.20086539e-01
-4.57548171e-01 -3.23211312e-01 -2.70215601e-01 6.59524798e-02
-5.96394181e-01 -1.82095885e-01 -2.51590669e-01 -1.13534093e+00
6.82118773e-01 -5.51355958e-01 -4.59766090e-01 4.29036170e-01
2.41068229e-01 2.16704011e-01 1.16276228e+00 -1.24729085e+00
1.10336542e+00 -2.35138917e+00 4.23027962e-01 2.45909020e-01
-2.37155892e-02 3.27833593e-02 -2.57523000e-01 4.06249106e-01
-5.80788888e-02 -4.30957377e-01 -3.47100347e-01 -5.48048794e-01
-3.76788944e-01 6.15892589e-01 -4.02095646e-01 1.01819766e+00
-2.33527541e-01 2.62622267e-01 -7.06106365e-01 -5.33571482e-01
3.70863050e-01 6.92094505e-01 -7.10822761e-01 1.34520873e-01
1.00690275e-01 7.37272203e-01 -3.64699930e-01 7.00082421e-01
6.60857439e-01 -2.20100135e-02 1.29162252e-01 -5.92600226e-01
-8.53557438e-02 -8.47775638e-02 -1.41854608e+00 2.07815862e+00
-6.70753896e-01 5.36693335e-01 6.95434153e-01 -1.69107163e+00
9.29578483e-01 8.54424894e-01 9.36206341e-01 -6.79619372e-01
-1.03707565e-02 4.21034247e-01 -4.89667654e-01 -7.16470540e-01
6.08360209e-02 -5.81555605e-01 5.44222116e-01 3.71424109e-01
3.73681158e-01 -4.07860786e-01 3.87257665e-01 -6.32836521e-02
1.03951514e+00 4.00869772e-02 4.32220668e-01 -3.13574970e-01
7.96700239e-01 -3.06620210e-01 6.32818460e-01 6.68902338e-01
6.48644343e-02 5.00994623e-01 -2.78933700e-02 -5.65175474e-01
-1.17341757e+00 -7.72772133e-01 -6.75280020e-02 8.10006320e-01
8.06676149e-02 -3.03879678e-01 -5.08773506e-01 -1.66046068e-01
-7.69927576e-02 1.03621259e-01 -2.02045441e-01 -7.88826495e-02
-9.90025878e-01 -4.00401592e-01 -4.07240428e-02 4.29026723e-01
5.16693890e-01 -6.74592495e-01 -2.52677470e-01 5.04135311e-01
-2.23800793e-01 -1.12130344e+00 -6.46437466e-01 5.73996753e-02
-1.62896061e+00 -9.67918277e-01 -8.81494164e-01 -1.30294383e+00
6.22484803e-01 7.25982130e-01 9.17904317e-01 1.79163530e-01
-3.65187287e-01 9.04024065e-01 -3.90185595e-01 1.21753804e-01
-3.23061377e-01 -4.09262717e-01 2.27241948e-01 4.48425382e-01
-2.82291025e-01 -1.00645006e+00 -6.51774764e-01 4.59431738e-01
-1.28931510e+00 -3.22425157e-01 3.47312301e-01 1.25435817e+00
1.26683581e+00 6.09791398e-01 3.34571600e-01 -8.09278429e-01
1.50848359e-01 -2.45017007e-01 -7.26406395e-01 1.11130718e-02
-4.10421848e-01 -2.73237973e-01 8.36359620e-01 -4.07895237e-01
-1.13436067e+00 1.45369247e-01 -4.39796418e-01 -8.87583911e-01
2.77984619e-01 6.77208483e-01 2.11415589e-02 -4.60907400e-01
4.51925904e-01 7.03645766e-01 3.54904115e-01 -6.25841439e-01
1.42422169e-01 2.58081198e-01 4.70891774e-01 -5.95316410e-01
9.75632489e-01 1.09874499e+00 7.13150948e-02 -1.18758881e+00
-8.34000230e-01 -9.16304111e-01 -5.48173368e-01 -1.07934430e-01
7.09991336e-01 -1.16233408e+00 -3.14602464e-01 5.54113209e-01
-1.04102898e+00 -3.35875422e-01 -3.79319221e-01 6.97391033e-01
-8.95584226e-01 8.76674592e-01 -1.05622745e+00 -4.66118038e-01
-3.82291794e-01 -9.26308572e-01 1.10135150e+00 -2.61911601e-01
2.74945349e-01 -1.43807733e+00 -5.87058924e-02 3.19215596e-01
3.79999757e-01 1.34283632e-01 6.83404267e-01 -4.51377444e-02
-6.27487719e-01 -3.74276079e-02 3.15885514e-01 7.67905772e-01
3.90958756e-01 -5.92981398e-01 -5.78791618e-01 -9.58032906e-01
9.06528413e-01 -2.70799488e-01 1.03350186e+00 9.25227106e-01
1.21225250e+00 -3.42714399e-01 -9.39557105e-02 1.07884419e+00
1.72481537e+00 1.67853683e-01 6.76724076e-01 1.20681241e-01
5.34258068e-01 3.12190205e-01 7.72486329e-01 8.26380670e-01
-2.34060541e-01 2.66207457e-01 2.98046827e-01 -6.73525967e-04
-1.55435383e-01 -3.25838625e-02 5.11800230e-01 1.44314718e+00
-2.25097731e-01 4.27160040e-02 -5.10375977e-01 6.35363638e-01
-1.81944311e+00 -1.13143563e+00 1.12659587e-02 2.15242958e+00
8.86638463e-01 -3.01273391e-02 -1.03587963e-01 4.89624918e-01
5.82699001e-01 5.28900981e-01 -3.63830566e-01 6.59081787e-02
-1.00140221e-01 7.99763978e-01 4.21442389e-01 5.62937319e-01
-1.21999431e+00 5.08166790e-01 6.42281055e+00 1.03507507e+00
-1.05906487e+00 5.16014457e-01 5.68634093e-01 1.22481488e-01
-1.63488194e-01 -7.86800385e-02 -5.29969633e-01 5.46690784e-02
7.13118494e-01 -6.52773008e-02 5.99428654e-01 5.99428356e-01
6.28139436e-01 8.09284206e-03 -1.09139311e+00 1.47758305e+00
2.94229060e-01 -1.45866394e+00 6.43932670e-02 -3.02895844e-01
1.01949394e+00 -2.98787355e-01 1.26631930e-01 -4.74285185e-02
-4.00555998e-01 -5.90457618e-01 5.61183929e-01 3.25345278e-01
6.30720258e-01 -6.61057830e-01 6.33915901e-01 3.93150330e-01
-1.31305242e+00 -4.66727912e-02 -6.56843662e-01 3.28493416e-02
4.80154455e-01 9.91508484e-01 -2.08655804e-01 4.48900670e-01
9.84231174e-01 1.50304699e+00 1.38944194e-01 9.96400058e-01
-1.40803412e-01 9.81634796e-01 -3.77290756e-01 6.05173707e-01
9.22098830e-02 -6.04837477e-01 6.80901408e-01 9.82907236e-01
7.02664316e-01 4.34954196e-01 4.74554926e-01 1.40680656e-01
4.99210358e-02 1.98624253e-01 -5.10377407e-01 1.90739974e-01
-1.07226558e-02 9.76798654e-01 -5.86743236e-01 -3.86051774e-01
-7.69446731e-01 1.08502305e+00 -5.04473746e-01 3.45134526e-01
-7.15874970e-01 8.89665335e-02 3.84751081e-01 3.47990692e-01
8.83859932e-01 -4.99768436e-01 1.32056132e-01 -1.21755648e+00
1.50946975e-01 -1.28575909e+00 5.92175066e-01 -3.41709375e-01
-1.36523128e+00 3.08544546e-01 -9.60593894e-02 -1.63508356e+00
7.47542530e-02 -4.86624837e-01 -4.18615729e-01 2.73355126e-01
-1.66795802e+00 -9.71487820e-01 -2.55448580e-01 1.13315392e+00
8.54900181e-01 -2.55724102e-01 4.93962854e-01 7.21272111e-01
2.23737322e-02 2.94336975e-01 6.17907047e-01 -1.55704677e-01
6.94519401e-01 -7.48343587e-01 -7.32236952e-02 7.86215603e-01
3.99189144e-01 2.28802353e-01 6.25209451e-01 -6.15157366e-01
-1.87709117e+00 -1.00292397e+00 4.65980053e-01 3.95095199e-01
6.04158163e-01 -1.30351158e-02 -1.04038560e+00 5.67049921e-01
2.59533823e-01 5.05487263e-01 6.29707158e-01 -3.52802753e-01
-5.90595603e-02 -5.66447616e-01 -1.21821868e+00 2.14258730e-01
1.08376467e+00 -5.40070415e-01 -4.39980179e-01 7.07049549e-01
4.29051995e-01 -6.17208600e-01 -1.06717968e+00 8.53781775e-02
2.70490944e-01 -9.13449705e-01 1.26820683e+00 -2.52074003e-01
4.11439598e-01 -1.60278812e-01 -3.26824188e-01 -1.18141425e+00
-3.43897372e-01 -1.04224288e+00 -3.00938010e-01 8.64789903e-01
-1.57119438e-01 -5.18514752e-01 7.50153959e-01 -1.45989224e-01
-3.78516138e-01 -5.32126486e-01 -1.36180508e+00 -8.32614839e-01
-3.66818666e-01 -4.34487373e-01 -1.27862126e-01 8.29318941e-01
-3.28391343e-01 -1.84851214e-01 -7.75491118e-01 2.82137811e-01
1.19714999e+00 1.18260384e-01 4.72182006e-01 -7.62334108e-01
-8.03363681e-01 1.53356656e-01 -5.35830677e-01 -1.53855276e+00
1.45040214e-01 -8.76004577e-01 1.31404474e-02 -1.31849933e+00
4.18545865e-02 -1.27023369e-01 -3.73433471e-01 2.41304830e-01
1.72285840e-01 3.71740580e-01 -1.42160282e-01 3.15760106e-01
-5.05678415e-01 7.42413938e-01 1.41837561e+00 -4.87422884e-01
1.64928474e-02 -4.07356955e-02 -8.50643218e-02 7.93182075e-01
2.76449680e-01 -6.18548036e-01 -4.69821066e-01 -7.04058707e-01
-2.47492269e-02 6.98135972e-01 2.67824739e-01 -9.08211231e-01
4.50883001e-01 3.84544246e-02 1.55185804e-01 -5.76574683e-01
5.22861421e-01 -1.12785351e+00 1.44012645e-01 7.50684381e-01
-7.57368058e-02 -6.70997947e-02 6.80139735e-02 9.76873219e-01
-7.86064029e-01 -1.30799145e-01 9.71789777e-01 -2.37720132e-01
-8.34704876e-01 6.41100526e-01 -2.87331551e-01 -6.01451136e-02
7.47435987e-01 -3.91165793e-01 5.79369903e-01 -7.45215654e-01
-1.20151365e+00 -1.57846898e-01 1.21732585e-01 -1.13673247e-01
1.12574673e+00 -1.34611368e+00 -8.94962907e-01 5.36868095e-01
-5.14349878e-01 -1.20345578e-01 4.85253572e-01 1.19011176e+00
-7.95556486e-01 -9.26747322e-02 -7.86540136e-02 -1.02161551e+00
-1.31575763e+00 5.50812244e-01 1.14990994e-02 -2.24164411e-01
-1.03772068e+00 8.35430205e-01 1.30397648e-01 7.52965808e-02
2.39172980e-01 -2.91908681e-01 2.97989935e-01 -5.78251220e-02
5.80961585e-01 5.56916177e-01 1.97808191e-01 -3.30576599e-01
-7.14184567e-02 8.37176740e-01 7.27426037e-02 4.28834483e-02
1.93000817e+00 -2.34894186e-01 -4.28604066e-01 2.02091724e-01
1.35243201e+00 -1.25559211e-01 -1.32808340e+00 -5.99580646e-01
-1.00267716e-01 -8.97185564e-01 3.40008467e-01 1.64199516e-01
-1.59094322e+00 6.35281801e-01 8.32660794e-01 6.76259547e-02
1.64863861e+00 -2.84570098e-01 1.48377514e+00 3.84824842e-01
8.12325835e-01 -1.16618538e+00 4.51620042e-01 3.01978648e-01
7.45352507e-01 -1.29773486e+00 3.64843428e-01 -6.60575449e-01
3.13996971e-02 1.25264335e+00 -9.26300511e-02 -6.32659376e-01
1.17357516e+00 2.69700974e-01 -6.69687912e-02 8.40947926e-02
-6.61015570e-01 1.72856539e-01 1.22909896e-01 5.27296007e-01
2.42919400e-01 -4.33738887e-01 -5.02187431e-01 2.80332807e-02
4.01161492e-01 1.23968802e-01 2.53510356e-01 1.00426733e+00
-4.33699280e-01 -1.10985005e+00 -7.33737826e-01 3.43755215e-01
-5.31600535e-01 -1.26513854e-01 5.34786701e-01 5.94381273e-01
5.11891730e-02 8.03658783e-01 -2.96624955e-02 -1.52675509e-02
2.53396153e-01 -2.60641575e-01 1.04606938e+00 -4.89278316e-01
-3.16228181e-01 7.40146577e-01 -2.25394219e-01 -6.66273296e-01
-9.88136649e-01 -9.94281828e-01 -9.11825418e-01 -1.04999326e-01
-2.30764776e-01 1.51448166e-02 2.88482547e-01 7.97643363e-01
-3.29201072e-02 9.77515727e-02 9.74393249e-01 -1.08890414e+00
-7.88474917e-01 -5.23823678e-01 -8.58928561e-01 5.66236913e-01
4.14263368e-01 -5.19009590e-01 -4.48958725e-01 7.03862786e-01]
|
[11.567971229553223, -2.1513333320617676]
|
d7bdecd6-b487-4373-9cdf-ec7f115f4278
|
a-diffusion-probabilistic-prior-for-low-dose
|
2305.15887
| null |
https://arxiv.org/abs/2305.15887v1
|
https://arxiv.org/pdf/2305.15887v1.pdf
|
A Diffusion Probabilistic Prior for Low-Dose CT Image Denoising
|
Low-dose computed tomography (CT) image denoising is crucial in medical image computing. Recent years have been remarkable improvement in deep learning-based methods for this task. However, training deep denoising neural networks requires low-dose and normal-dose CT image pairs, which are difficult to obtain in the clinic settings. To address this challenge, we propose a novel fully unsupervised method for low-dose CT image denoising, which is based on denoising diffusion probabilistic model -- a powerful generative model. First, we train an unconditional denoising diffusion probabilistic model capable of generating high-quality normal-dose CT images from random noise. Subsequently, the probabilistic priors of the pre-trained diffusion model are incorporated into a Maximum A Posteriori (MAP) estimation framework for iteratively solving the image denoising problem. Our method ensures the diffusion model produces high-quality normal-dose CT images while keeping the image content consistent with the input low-dose CT images. We evaluate our method on a widely used low-dose CT image denoising benchmark, and it outperforms several supervised low-dose CT image denoising methods in terms of both quantitative and visual performance.
|
['Xiaokun Liang', 'Shan Tan', 'Songhui Diao', 'Yaoqin Xie', 'Xuan Liu']
|
2023-05-25
| null | null | null | null |
['computed-tomography-ct']
|
['methodology']
|
[ 4.48755383e-01 9.18709785e-02 4.19951975e-01 -5.43376923e-01
-1.43685710e+00 -1.18977241e-01 4.08313245e-01 1.84695572e-01
-7.09498644e-01 3.68674487e-01 4.81062382e-01 1.67272910e-02
-1.88058570e-01 -1.11482835e+00 -5.61053276e-01 -1.34245598e+00
6.22795336e-02 6.60391867e-01 2.60262638e-01 -1.37603413e-02
-1.63190126e-01 2.87467301e-01 -7.58156240e-01 5.54710269e-01
7.89029956e-01 7.50276864e-01 5.68077147e-01 8.98845017e-01
2.02675253e-01 8.92941177e-01 -3.29124838e-01 -1.07319109e-01
2.12850168e-01 -7.86850035e-01 -6.29196763e-01 -4.56088297e-02
4.27266844e-02 -7.18226671e-01 -6.98826253e-01 1.30245876e+00
8.37833583e-01 1.06682397e-01 1.05011594e+00 -5.82238495e-01
-5.53553760e-01 3.18509638e-01 -4.47484583e-01 4.84934181e-01
-1.36027792e-02 3.56580585e-01 2.80584484e-01 -4.68220949e-01
5.36005378e-01 1.12767100e+00 7.56927192e-01 5.77192724e-01
-1.34463656e+00 -1.50686234e-01 -4.92769092e-01 -1.85964838e-01
-8.45429301e-01 -1.23081625e-01 7.30278373e-01 -4.22734767e-01
5.77214479e-01 -2.23886478e-03 5.84335148e-01 1.02289510e+00
9.71010327e-01 4.37914759e-01 1.32328045e+00 -1.73921257e-01
2.54807353e-01 -4.67884600e-01 -2.67169565e-01 7.24416852e-01
7.10843503e-02 2.82401025e-01 -1.79122597e-01 -1.66420937e-01
1.09216881e+00 3.75304431e-01 -5.32117426e-01 -3.93652171e-02
-1.07396591e+00 8.55831563e-01 7.97057986e-01 5.70850909e-01
-1.09581161e+00 5.23440182e-01 2.28051171e-01 -2.37241760e-02
8.43234062e-01 -2.39964068e-01 1.96475551e-01 7.96011165e-02
-1.04932594e+00 -8.95229876e-02 5.56835592e-01 3.54753375e-01
2.71149814e-01 -1.84444041e-04 -3.79379272e-01 7.69892275e-01
6.12320065e-01 4.54899341e-01 5.15644372e-01 -9.64222193e-01
5.92149310e-02 -2.84498371e-03 -3.07632387e-01 -7.04890430e-01
-2.89236158e-01 -3.78412426e-01 -1.78354895e+00 4.12072003e-01
2.67593265e-01 9.93159860e-02 -1.47212732e+00 1.51628518e+00
3.79929960e-01 2.83043563e-01 -4.09105308e-02 9.35336888e-01
9.57369268e-01 8.08114290e-01 4.75098729e-01 -6.14989281e-01
1.22544098e+00 -6.14376664e-01 -1.01486301e+00 -1.67395279e-01
3.73091072e-01 -8.62753034e-01 7.96547055e-01 5.70676208e-01
-1.46500945e+00 -2.49659240e-01 -1.07711732e+00 -1.23541638e-01
2.74187505e-01 -2.97785997e-01 3.97622824e-01 7.45807171e-01
-1.14944386e+00 9.45412815e-01 -1.43363512e+00 9.13145989e-02
8.56336117e-01 2.86866575e-01 -3.70019615e-01 -7.56157160e-01
-9.32564497e-01 7.36140490e-01 9.55084711e-02 2.88554490e-01
-1.61455023e+00 -9.21640575e-01 -7.85652876e-01 -1.90103486e-01
-1.03809170e-01 -1.17468154e+00 1.30498588e+00 -6.08780921e-01
-1.42001104e+00 8.01832676e-01 -1.06534280e-01 -3.06659549e-01
7.81593621e-01 1.25693828e-01 -1.32422592e-03 5.00722885e-01
1.40309632e-01 6.15391672e-01 1.01612389e+00 -1.52052534e+00
-1.11171968e-01 -3.78336161e-01 -5.98105371e-01 6.31197467e-02
1.96273640e-01 -1.69494629e-01 -3.26448321e-01 -9.39464390e-01
6.06873453e-01 -5.47542036e-01 -7.74102569e-01 2.13120133e-01
-1.96553245e-01 1.50927380e-01 5.00106275e-01 -8.86658669e-01
7.93385744e-01 -2.08352613e+00 8.95268396e-02 3.11742872e-01
6.42449856e-01 -1.77574471e-01 8.79995301e-02 -6.15726523e-02
-1.57718331e-01 1.28859788e-01 -9.01644468e-01 -6.31559134e-01
-4.14957583e-01 6.35503232e-01 1.63856044e-01 6.75453782e-01
-1.17320821e-01 8.42803121e-01 -9.74226713e-01 -6.07866168e-01
2.47596160e-01 9.82445657e-01 -3.91794205e-01 5.12363255e-01
6.58181980e-02 9.43292320e-01 -3.48269612e-01 2.44255751e-01
9.83425260e-01 -8.09092000e-02 -3.42694484e-02 -4.10146117e-01
2.67207384e-01 -1.90244526e-01 -6.70339584e-01 2.05835891e+00
-5.39340496e-01 1.60350755e-01 2.17057228e-01 -7.65855789e-01
3.64140511e-01 5.21478653e-01 9.58454072e-01 -7.43781388e-01
4.25511122e-01 2.69720048e-01 -1.21448070e-01 -7.16485620e-01
-1.18559614e-01 -1.10830605e+00 2.20312253e-01 6.21429563e-01
6.24769405e-02 -7.91668475e-01 -8.75936747e-02 3.37378263e-01
1.32451952e+00 -2.42801040e-01 -3.08605611e-01 -1.94000021e-01
2.23183677e-01 -1.85023814e-01 4.24752742e-01 8.58576655e-01
-3.15471083e-01 1.30502141e+00 3.73289824e-01 -2.43643582e-01
-1.13146305e+00 -1.36194694e+00 -1.56733006e-01 4.31039423e-01
-6.47752732e-02 -7.63552357e-03 -1.07114768e+00 -5.88452101e-01
-5.05956948e-01 6.25680268e-01 -7.76295900e-01 -2.40786895e-01
-6.48419023e-01 -1.52380300e+00 3.98244828e-01 4.64064181e-01
5.39829552e-01 -9.26882148e-01 -1.63398981e-01 4.20546263e-01
-4.44589734e-01 -9.24509108e-01 -4.97886479e-01 5.35670102e-01
-1.33153832e+00 -8.28812182e-01 -1.18545353e+00 -9.67876971e-01
9.21709836e-01 7.13548809e-03 1.30969512e+00 2.79945999e-01
-5.03448963e-01 3.63188595e-01 8.97628143e-02 -1.14765882e-01
-9.81665492e-01 -5.97921193e-01 -2.83514112e-01 -1.90474793e-01
-1.72646016e-01 -7.58033097e-01 -1.12675130e+00 5.14190160e-02
-1.60530233e+00 -4.87659387e-02 7.86396980e-01 1.09504664e+00
1.25362968e+00 6.20077908e-01 2.33662799e-01 -9.88215566e-01
7.14743733e-01 -4.87754226e-01 -2.24675149e-01 -8.38218629e-02
-5.56631744e-01 1.32470146e-01 2.64553756e-01 -2.35360146e-01
-1.40841305e+00 9.95699987e-02 -9.97862220e-01 -1.21740103e-01
-1.23075671e-01 4.29252118e-01 8.25772807e-02 -1.68582067e-01
8.16118479e-01 5.52286386e-01 1.57157853e-01 -3.05521727e-01
1.76511198e-01 3.24677616e-01 6.86790049e-01 -5.16368330e-01
5.81073105e-01 9.38259542e-01 2.97175467e-01 -6.08778358e-01
-8.66929829e-01 -2.62706965e-01 -6.15626872e-01 -2.16783792e-01
1.45914996e+00 -8.03378165e-01 -5.14436901e-01 7.52341568e-01
-1.10888696e+00 -5.10522246e-01 -3.77894461e-01 6.34941876e-01
-7.04920471e-01 7.61430681e-01 -1.28370833e+00 -2.75665015e-01
-5.13849020e-01 -1.71000779e+00 1.05522943e+00 2.60832831e-02
2.77674884e-01 -1.17799461e+00 3.30005914e-01 1.84610888e-01
6.21719718e-01 5.04303157e-01 1.18009734e+00 7.62888417e-02
-6.00593746e-01 -1.78485498e-01 -1.89816877e-01 9.58101153e-01
1.51646569e-01 -5.62689543e-01 -8.61016512e-01 -2.70597577e-01
8.53786767e-01 -2.50113428e-01 1.11974359e+00 1.24478030e+00
1.45091367e+00 2.20072657e-01 -4.35728505e-02 8.73676240e-01
1.72626841e+00 -3.35185118e-02 9.43487048e-01 -6.60709664e-02
6.21616185e-01 2.29881912e-01 1.29137048e-02 2.56432325e-01
1.12970106e-01 2.34031156e-02 5.93623936e-01 -4.44024593e-01
-6.08732462e-01 -1.51581746e-02 -1.43595144e-01 7.75492549e-01
-1.68113112e-01 -3.25056285e-01 -8.96287084e-01 4.73436236e-01
-1.43354952e+00 -7.21304238e-01 -5.33874214e-01 1.90172112e+00
9.42234755e-01 2.24792451e-01 -3.63676548e-01 5.69000728e-02
3.73870194e-01 -8.11179131e-02 -2.74325311e-01 2.64044285e-01
-4.81730998e-02 5.64599514e-01 3.82787585e-01 5.72593331e-01
-1.03498125e+00 3.10757339e-01 6.36923409e+00 1.01083767e+00
-8.89534116e-01 7.76224136e-01 1.12275696e+00 1.28586471e-01
-3.86771709e-01 -3.30523312e-01 -1.66896991e-02 3.86271089e-01
7.27987468e-01 1.74611256e-01 -3.07817329e-02 4.03070688e-01
5.80669343e-01 -5.31464398e-01 -9.48219895e-01 1.12027931e+00
-2.80352328e-02 -1.33896339e+00 2.15137508e-02 1.01845726e-01
8.98669899e-01 1.45859420e-01 2.07727477e-01 5.30734956e-02
4.15048301e-01 -1.27840793e+00 2.43379220e-01 7.13201284e-01
5.39005220e-01 -8.79095972e-01 1.02367270e+00 4.66898888e-01
-5.83701551e-01 4.12244797e-01 -4.81625736e-01 6.68789625e-01
6.22080445e-01 1.11832583e+00 -3.20141107e-01 6.35475338e-01
8.60499918e-01 5.90855479e-01 -1.31298795e-01 1.05038929e+00
-3.95824939e-01 6.68048680e-01 -4.17644352e-01 7.77188540e-01
3.80236328e-01 -2.92484343e-01 2.68341005e-01 1.15909374e+00
3.69823903e-01 5.37550390e-01 -5.34491567e-03 8.11452985e-01
-2.88495868e-01 -2.22845301e-01 -3.28045428e-01 4.49543118e-01
-4.33038682e-01 1.18798220e+00 -9.83667970e-01 -4.33636278e-01
5.03276801e-03 1.26959121e+00 -1.76830769e-01 3.21733922e-01
-6.46265090e-01 3.63614708e-01 1.69766713e-02 2.36243144e-01
-9.92601812e-02 -2.73485363e-01 -4.98308420e-01 -8.96422923e-01
-2.53963172e-01 -6.82626605e-01 5.54572701e-01 -8.95138383e-01
-1.65068758e+00 9.39328849e-01 -3.25586125e-02 -9.78909612e-01
1.21856583e-02 -3.08433503e-01 -7.79320955e-01 9.46857691e-01
-1.64742231e+00 -1.05114186e+00 -6.12152755e-01 7.16518819e-01
4.64198351e-01 4.13535774e-01 6.14158034e-01 4.91432697e-01
-2.08788857e-01 3.01888045e-02 3.55864108e-01 4.28520521e-04
6.34833038e-01 -1.42481673e+00 -5.53986542e-02 8.34393740e-01
-4.16634679e-01 1.94195390e-01 6.59311533e-01 -8.31398547e-01
-1.14337051e+00 -1.07693076e+00 4.67662871e-01 -2.60187387e-01
3.02810878e-01 7.14057386e-02 -9.63430345e-01 4.17175770e-01
5.29716730e-01 3.88832182e-01 5.63997149e-01 -7.90247619e-01
2.86261082e-01 6.61369339e-02 -1.72724903e+00 3.10359687e-01
6.08251870e-01 -2.16186360e-01 -4.89111274e-01 7.29379475e-01
4.53990430e-01 -6.43117845e-01 -1.27469218e+00 3.59555304e-01
-2.11951453e-02 -9.08075690e-01 1.30632281e+00 -8.07635710e-02
7.93888092e-01 -2.03883462e-02 8.15941021e-02 -1.51672494e+00
-3.49145114e-01 -4.04634356e-01 3.18889081e-01 7.18246043e-01
2.30812877e-02 -3.58522773e-01 8.72865319e-01 3.70217085e-01
-3.97309303e-01 -6.37635529e-01 -1.01943207e+00 -4.74684328e-01
4.58943784e-01 -6.63856864e-01 -1.41855285e-01 7.17552006e-01
-7.41226971e-01 -4.91743907e-02 -2.03942642e-01 1.82408690e-01
1.31073809e+00 -4.97849613e-01 5.97896837e-02 -7.16638684e-01
-4.05510336e-01 -3.89865279e-01 -1.29727349e-01 -1.07809925e+00
-1.84192151e-01 -9.58565176e-01 4.98276979e-01 -2.05411339e+00
4.51417714e-01 -2.48364061e-01 -2.18994290e-01 1.73775718e-01
-1.57957003e-01 6.06131196e-01 -3.38628352e-01 2.79268861e-01
-1.31466269e-01 7.18750417e-01 1.85523927e+00 -3.63266289e-01
1.55217364e-01 1.29868045e-01 -2.84085304e-01 8.51476490e-01
3.82777750e-01 -1.29628503e+00 -2.98616171e-01 -5.66281497e-01
4.46227305e-02 3.01643819e-01 6.05650902e-01 -8.93363059e-01
4.26989138e-01 1.50873557e-01 6.93341613e-01 -4.97810543e-01
2.02745512e-01 -9.11916673e-01 7.71152228e-02 8.87174070e-01
-1.02418132e-01 -1.55455142e-01 5.08027943e-03 7.99932003e-01
-4.57635134e-01 -4.28599536e-01 1.48169029e+00 -5.10778248e-01
-1.03982158e-01 5.47361970e-01 -6.85197115e-01 -5.14148250e-02
7.07491040e-01 4.47810180e-02 1.29084691e-01 -4.85018134e-01
-1.24143541e+00 -1.69061512e-01 1.22031933e-02 -2.61218727e-01
1.05202019e+00 -1.27282524e+00 -1.05671728e+00 9.57795829e-02
-3.24274987e-01 4.43599999e-01 8.28659832e-01 1.30423999e+00
-9.45333123e-01 -3.19174141e-01 1.68967144e-05 -1.00370920e+00
-9.01569188e-01 2.92335272e-01 7.66585886e-01 -7.76956201e-01
-8.73475611e-01 9.70806301e-01 5.21688879e-01 -2.41196766e-01
-1.04169138e-01 -6.39262795e-01 2.61180222e-01 -5.79632878e-01
3.60504568e-01 9.63032804e-03 4.25480187e-01 -4.32309061e-01
-4.13739793e-02 5.66488862e-01 -1.48115844e-01 -1.66741714e-01
1.84334242e+00 -1.19888760e-01 -3.18591684e-01 -9.94898006e-02
1.22292018e+00 -4.86020416e-01 -1.35646009e+00 -1.58116117e-01
-3.85933787e-01 -4.34947819e-01 8.55775952e-01 -7.07422912e-01
-1.34823537e+00 9.83864963e-01 1.18842089e+00 2.91783866e-02
1.37907553e+00 -7.75456652e-02 1.09413064e+00 5.02516069e-02
3.24533135e-02 -7.84808934e-01 4.57864940e-01 1.66146860e-01
9.63264525e-01 -1.47361934e+00 1.75799713e-01 -4.75848585e-01
-4.17158395e-01 1.04403210e+00 9.46416482e-02 -4.37426984e-01
1.03531563e+00 5.51876843e-01 2.68603683e-01 -5.57156980e-01
-2.94934869e-01 2.76013482e-02 1.77004248e-01 6.69915736e-01
4.33824301e-01 -3.05061251e-01 -1.96832672e-01 4.66038465e-01
2.75681287e-01 3.73775400e-02 4.73070741e-01 1.08483982e+00
-3.54282051e-01 -1.01820731e+00 -4.97333467e-01 2.72413403e-01
-7.39946306e-01 -2.15742975e-01 2.94917524e-01 4.58131582e-01
5.57466457e-03 8.16510379e-01 -2.99286544e-01 1.92395404e-01
2.86671251e-01 -4.16358352e-01 7.49402761e-01 -6.38193846e-01
-8.78338754e-01 6.71967089e-01 -4.73916441e-01 -4.51883286e-01
-4.63297307e-01 -3.41815919e-01 -1.23980784e+00 -2.94849902e-01
9.10645872e-02 -1.60312414e-01 8.76874387e-01 9.99820173e-01
-2.55173326e-01 9.47351277e-01 3.94625098e-01 -1.00297379e+00
-5.13072312e-01 -9.87777591e-01 -5.99650443e-01 6.24407291e-01
3.13214362e-01 -1.72093555e-01 -4.15400922e-01 3.29775065e-01]
|
[13.461468696594238, -2.519710063934326]
|
83d4c8d3-9649-45ee-ae83-aec6ae99d45d
|
surrogate-based-black-box-optimization-method
|
2110.03522
| null |
https://arxiv.org/abs/2110.03522v1
|
https://arxiv.org/pdf/2110.03522v1.pdf
|
Surrogate-Based Black-Box Optimization Method for Costly Molecular Properties
|
AI-assisted molecular optimization is a very active research field as it is expected to provide the next-generation drugs and molecular materials. An important difficulty is that the properties to be optimized rely on costly evaluations. Machine learning methods are investigated with success to predict these properties, but show generalization issues on less known areas of the chemical space. We propose here a surrogate-based black box optimization method, to tackle jointly the optimization and machine learning problems. It consists in optimizing the expected improvement of the surrogate of a molecular property using an evolutionary algorithm. The surrogate is defined as a Gaussian Process Regression (GPR) model, learned on a relevant area of the search space with respect to the property to be optimized. We show that our approach can successfully optimize a costly property of interest much faster than a purely metaheuristic approach.
|
['Benoit Da Mota', 'Beatrice Duval', 'Thomas Cauchy', 'Jules Leguy']
|
2021-10-01
| null | null | null | null |
['gpr', 'gpr']
|
['computer-vision', 'miscellaneous']
|
[ 6.65004075e-01 9.44915712e-02 -2.55216628e-01 -1.18027434e-01
-7.55959511e-01 -3.58145088e-01 4.32052940e-01 6.95448399e-01
-5.37476301e-01 1.24984026e+00 -3.27514827e-01 -1.49807855e-01
-5.47122478e-01 -9.21110570e-01 -8.00879538e-01 -1.20183039e+00
-8.78566578e-02 8.68388236e-01 1.26787489e-02 -3.10932606e-01
6.59721792e-01 6.88170195e-01 -1.44034576e+00 -1.00219294e-01
1.38590622e+00 9.76348519e-01 2.06174254e-01 4.52460438e-01
8.68691057e-02 7.77684599e-02 -4.15818959e-01 -3.48500878e-01
1.02815935e-02 -5.45173705e-01 -7.37519503e-01 -4.56134915e-01
-1.23332322e-01 6.53322279e-01 4.53785330e-01 1.10620224e+00
7.05274403e-01 3.47583801e-01 9.66440558e-01 -8.17634881e-01
-2.05759242e-01 1.25453815e-01 -7.84159377e-02 -1.14206575e-01
4.08588260e-01 1.92930043e-01 9.32266414e-01 -4.40947115e-01
7.21944213e-01 9.47849274e-01 3.17277074e-01 5.96667886e-01
-1.50969505e+00 -2.58458942e-01 -1.68974489e-01 4.25708055e-01
-1.30755329e+00 4.00393680e-02 6.90127790e-01 -4.18765008e-01
1.04658294e+00 3.61013651e-01 7.72567630e-01 7.61556566e-01
5.95396936e-01 6.09482825e-01 1.04573810e+00 -3.34631979e-01
9.37545300e-01 4.47908700e-01 -2.83217549e-01 5.52447855e-01
3.02897483e-01 3.73396575e-01 -5.83223581e-01 -5.01237750e-01
-8.70079547e-02 -2.49955326e-01 -3.09517980e-01 -6.89651191e-01
-7.88391292e-01 1.08272159e+00 3.69873524e-01 3.22689593e-01
-1.00557971e+00 7.61349872e-02 -1.75891623e-01 -5.65675013e-02
5.26646435e-01 1.39149094e+00 -8.30522478e-01 -4.05625939e-01
-8.37786436e-01 4.89362150e-01 1.04489076e+00 1.37501046e-01
7.70370185e-01 -1.37791365e-01 -1.36269964e-02 4.31650549e-01
3.81973028e-01 3.23106170e-01 2.01240405e-01 -4.29307163e-01
2.57523566e-01 7.48378515e-01 1.09946407e-01 -8.15216601e-01
-5.45101941e-01 -6.36057258e-01 -3.56950074e-01 4.34066266e-01
3.24816465e-01 -3.06546479e-01 -6.21733606e-01 1.37952840e+00
6.50640965e-01 -1.54077560e-01 2.32202504e-02 5.62401116e-01
2.39461794e-01 8.69876623e-01 3.09788287e-01 -6.45441353e-01
9.81946886e-01 -8.72846007e-01 -4.46431458e-01 2.10197829e-02
5.65508485e-01 -6.54733181e-01 5.26367962e-01 6.97000384e-01
-1.17704225e+00 -1.22915290e-01 -1.23703861e+00 6.48229957e-01
-6.80407882e-01 -2.38088250e-01 7.52491474e-01 9.41360772e-01
-6.74793422e-01 1.46579182e+00 -6.48014963e-01 -2.27082670e-02
5.81551135e-01 9.36880112e-01 -2.83779532e-01 1.69649273e-01
-9.33517754e-01 1.18838358e+00 8.57623696e-01 1.51951402e-01
-7.25015342e-01 -9.50903475e-01 -5.85022628e-01 9.57239270e-02
7.14767516e-01 -8.89897645e-01 6.55510485e-01 -9.97641742e-01
-2.12575912e+00 5.54939449e-01 -1.54963642e-01 -5.36542296e-01
4.78514403e-01 -1.22525126e-01 -2.56430715e-01 -3.51942033e-02
-4.92085099e-01 4.55758393e-01 1.11456740e+00 -1.02127099e+00
-3.61232460e-01 -4.06827062e-01 -4.74165380e-01 1.69783428e-01
-1.72598854e-01 -1.00419469e-01 9.12887678e-02 -3.60913754e-01
-1.98724642e-01 -9.69155610e-01 -7.58238018e-01 -2.56787628e-01
-5.20286143e-01 -2.06230864e-01 3.76375407e-01 -4.50981855e-01
1.22275651e+00 -1.37321079e+00 8.26642931e-01 6.83013618e-01
-1.41216576e-01 4.94867802e-01 -1.30379423e-01 6.37178838e-01
-1.97470516e-01 3.45565528e-01 -5.92031956e-01 7.94220567e-02
-2.93592691e-01 -8.88321027e-02 1.23127839e-02 5.72331727e-01
5.01945913e-01 8.10230851e-01 -9.55348849e-01 -1.67093903e-01
-9.82778370e-02 5.27081847e-01 -7.73403168e-01 1.54587120e-01
-7.34472632e-01 6.26898110e-01 -9.37715113e-01 7.76467562e-01
5.27358890e-01 -6.48982003e-02 4.64203544e-02 -1.62930474e-01
-2.99447984e-01 -1.29203826e-01 -9.96477723e-01 1.38397408e+00
-2.09918573e-01 1.34001732e-01 -3.04312289e-01 -1.14252770e+00
1.05816460e+00 1.49446145e-01 8.65279496e-01 -4.37289834e-01
3.03797722e-01 5.41912735e-01 3.06964934e-01 -2.75020182e-01
2.08597749e-01 -5.50898254e-01 3.94790053e-01 1.87531799e-01
-1.62754990e-02 -5.51454186e-01 1.94340602e-01 -6.77972794e-01
1.06540012e+00 5.38915694e-01 6.30304217e-01 -4.10247058e-01
9.11620915e-01 3.12211812e-01 4.76727545e-01 5.21057248e-01
1.61602646e-01 1.97895080e-01 3.34235638e-01 -5.19725025e-01
-1.11482728e+00 -5.34488976e-01 -2.50376374e-01 5.67921162e-01
3.84210534e-02 -3.16101074e-01 -7.39130378e-01 -6.95147276e-01
-9.48275104e-02 8.63458157e-01 -6.22489154e-01 -5.21531582e-01
-6.05779171e-01 -1.37345183e+00 -3.55734080e-02 -5.43222539e-02
-2.31173784e-01 -1.12966919e+00 -5.55466175e-01 7.14042962e-01
6.45836353e-01 -5.09169638e-01 7.52504840e-02 5.43415248e-01
-1.04234457e+00 -1.05955720e+00 -6.98576450e-01 -3.06882113e-01
5.34157038e-01 -5.41782498e-01 9.98757660e-01 5.79913333e-02
-6.12517357e-01 -4.68642674e-02 -2.43504420e-01 -8.36508334e-01
-6.00064695e-01 7.10515231e-02 3.53541374e-02 2.93107331e-01
5.07398397e-02 -6.55232131e-01 -6.32889152e-01 1.99243322e-01
-6.95740521e-01 -3.27258587e-01 6.27063334e-01 9.86532569e-01
1.01855290e+00 2.29244813e-01 5.06435812e-01 -6.60046577e-01
6.57303452e-01 -4.19020116e-01 -8.88526618e-01 4.18088526e-01
-8.28804255e-01 8.38672221e-01 6.97915554e-01 -4.80083257e-01
-6.94369376e-01 2.60147363e-01 -2.89427549e-01 7.57214576e-02
-4.04612767e-03 5.73294818e-01 -4.59848076e-01 -6.25005484e-01
4.99498338e-01 1.28898606e-01 -9.50077623e-02 -5.44007480e-01
1.09625228e-01 1.97321564e-01 -3.73591557e-02 -9.61173117e-01
6.39092505e-01 3.12724471e-01 6.99425817e-01 -8.17680418e-01
-4.25809324e-01 -4.45571244e-01 -4.57789570e-01 -6.31806022e-03
8.00086260e-01 -1.31157607e-01 -1.13393223e+00 -7.20089525e-02
-1.08611405e+00 -1.25763461e-01 -4.36525881e-01 6.93896413e-01
-9.14071560e-01 1.34278446e-01 3.42313230e-01 -9.77210045e-01
-5.27899623e-01 -1.32675636e+00 9.91366923e-01 5.31475425e-01
-3.35933834e-01 -1.04800904e+00 5.40553391e-01 2.06669211e-01
3.10125947e-01 5.37404358e-01 1.30962813e+00 -9.44322467e-01
-5.92420161e-01 -4.40710068e-01 3.39656770e-01 1.93598837e-01
-1.15980394e-03 2.00891271e-01 -6.82613909e-01 -1.52543232e-01
2.00325996e-01 1.74582228e-02 5.78686774e-01 6.38520241e-01
1.13833785e+00 -1.68233290e-01 -6.00850105e-01 7.27725089e-01
1.55417168e+00 8.74117434e-01 8.01134169e-01 5.43795764e-01
2.20034942e-01 6.62985682e-01 8.79940093e-01 4.17164296e-01
-3.78423154e-01 1.01590168e+00 6.06154740e-01 7.42777586e-02
5.37118375e-01 -3.76853123e-02 1.17045745e-01 1.14928838e-02
-5.28653622e-01 -3.04007739e-01 -8.62993658e-01 1.31642669e-01
-1.80524004e+00 -9.26865160e-01 -1.22219464e-02 2.46526504e+00
9.69394445e-01 3.83420363e-02 1.35494277e-01 3.69138531e-02
5.53493083e-01 -3.42131853e-01 -7.42497921e-01 -7.47168124e-01
-2.23217368e-01 8.00431430e-01 5.62003136e-01 4.25066233e-01
-8.51592839e-01 7.51289308e-01 6.16606092e+00 1.19887912e+00
-1.09938109e+00 -3.70498151e-01 7.71379411e-01 7.41855125e-04
-2.75844336e-01 1.54674232e-01 -8.78854871e-01 4.28157508e-01
9.98747289e-01 -3.03320765e-01 5.23279548e-01 7.21745014e-01
3.96056712e-01 -3.08230191e-01 -1.20059919e+00 7.74736941e-01
-7.44580477e-02 -1.48792505e+00 8.48851427e-02 4.30252731e-01
9.76192832e-01 -4.69163448e-01 1.35517746e-01 -1.24611393e-01
-2.68339306e-01 -1.37105775e+00 4.04830337e-01 8.09762001e-01
2.22819880e-01 -1.06263316e+00 6.37095928e-01 4.01239395e-01
-6.26604676e-01 -6.67148381e-02 -2.14580491e-01 2.84568995e-01
2.30087280e-01 4.25217986e-01 -8.86964202e-01 5.63650310e-01
2.59167910e-01 3.82227659e-01 -4.17405576e-01 1.75471473e+00
-1.50414214e-01 3.47839206e-01 -5.12300193e-01 -7.45550156e-01
3.34451228e-01 -7.38890409e-01 1.03688943e+00 6.81645989e-01
5.14497340e-01 -3.01839024e-01 -6.10785447e-02 1.11090374e+00
2.05839828e-01 6.66049719e-01 -3.65888774e-01 -4.91425604e-01
1.04779704e-02 9.89696860e-01 -7.36834288e-01 -6.94485381e-02
3.00591826e-01 8.34721267e-01 -8.64854176e-03 1.81695104e-01
-7.84865141e-01 -2.89757162e-01 6.34079814e-01 -1.03930302e-01
3.86672586e-01 1.03637144e-01 -2.26609021e-01 -7.04499602e-01
-2.14716434e-01 -8.75286341e-01 1.85008928e-01 -3.01242650e-01
-9.66539979e-01 6.61554456e-01 -1.54146373e-01 -9.75719690e-01
-2.24434316e-01 -9.19089079e-01 -6.02181196e-01 9.63746786e-01
-1.57650518e+00 -5.22670090e-01 4.34368432e-01 2.26691663e-01
3.74910921e-01 -3.92845571e-01 8.46569419e-01 -4.74334545e-02
-5.94601572e-01 2.14183599e-01 5.35508335e-01 -1.00069356e+00
4.58712518e-01 -1.24592471e+00 -2.83919811e-01 3.58646929e-01
-9.65946838e-02 3.68064761e-01 1.27707052e+00 -7.19404519e-01
-1.70939195e+00 -6.41954064e-01 6.62655234e-01 -2.58204669e-01
7.35604525e-01 9.84510854e-02 -6.35134041e-01 -3.43926609e-01
-3.95510793e-02 -3.01676184e-01 7.28801787e-01 1.68039631e-02
3.49240839e-01 3.61148529e-02 -1.25356638e+00 5.48894763e-01
6.75787270e-01 -2.00124015e-03 -2.06527337e-01 4.94828075e-01
4.21168983e-01 -7.35581219e-02 -1.02027702e+00 5.66220284e-01
4.29620206e-01 -5.80239475e-01 1.19591236e+00 -1.02887988e+00
2.25624442e-01 -3.27719271e-01 9.79197323e-02 -1.57650781e+00
1.56979561e-01 -1.18596554e+00 -3.78605604e-01 6.42467439e-01
8.45307708e-01 -6.39391065e-01 1.13556671e+00 6.09476089e-01
1.70106247e-01 -1.46491730e+00 -1.07787716e+00 -1.02887142e+00
1.31024227e-01 -4.86116931e-02 6.84697270e-01 4.67067689e-01
-7.50372484e-02 4.36462641e-01 -2.15072140e-01 4.83515672e-03
5.18167198e-01 3.04468870e-01 4.02634233e-01 -1.34858644e+00
-5.59993684e-01 -8.54745686e-01 -4.27186847e-01 -4.15797234e-01
3.93121503e-02 -5.74965656e-01 4.17874334e-03 -1.15543103e+00
-8.65684226e-02 -3.85803610e-01 -1.64832547e-01 -1.80282518e-02
-2.95454681e-01 -2.39837885e-01 4.05075327e-02 -1.61948398e-01
-2.53122807e-01 8.87475312e-01 1.21579468e+00 -3.57733637e-01
-6.97564840e-01 5.05764008e-01 -5.55025280e-01 6.08585715e-01
9.03661072e-01 -8.85633290e-01 -5.59801199e-02 4.35139418e-01
6.32542849e-01 6.79915994e-02 -9.84383598e-02 -9.63308871e-01
5.42190596e-02 -5.37113786e-01 9.40354243e-02 -3.16089571e-01
5.60197413e-01 -9.27542925e-01 4.67625350e-01 8.09883058e-01
-1.44512221e-01 -2.77953625e-01 1.12908460e-01 7.03547060e-01
-2.41379678e-01 -7.61873245e-01 7.27974713e-01 1.49490476e-01
-3.04572850e-01 5.17573893e-01 -1.45463571e-01 -3.70940328e-01
1.24116063e+00 -4.69088316e-01 1.49986297e-01 -4.23551500e-02
-7.53406763e-01 8.14232044e-03 3.78688216e-01 6.30382076e-02
5.41765809e-01 -8.91123414e-01 -7.30404019e-01 -2.65768886e-01
-4.27156724e-02 -3.84620249e-01 -1.19745679e-01 8.68805468e-01
-7.25311160e-01 5.86027622e-01 -1.96564980e-02 -4.22342598e-01
-1.29518962e+00 8.49589169e-01 5.91083109e-01 -6.44042730e-01
-3.53358127e-02 7.69585729e-01 -2.35167682e-01 -1.34961084e-01
-1.30660579e-01 -5.24597242e-03 -2.65679568e-01 8.80156532e-02
3.95319939e-01 6.01636350e-01 3.33255947e-01 -3.91561508e-01
-5.21866381e-01 8.26109350e-01 3.43539596e-01 1.87289789e-01
1.85351169e+00 4.81786430e-01 -2.25811377e-01 -5.83771355e-02
1.35075521e+00 -1.02990553e-01 -1.09849739e+00 2.83637404e-01
4.35444474e-01 -2.16042131e-01 1.49502292e-01 -1.03920245e+00
-5.69213986e-01 6.27236843e-01 5.42729557e-01 -1.93126127e-02
1.10756373e+00 -1.78129062e-01 5.03092766e-01 6.78036392e-01
2.34244511e-01 -1.31055999e+00 -1.14487045e-01 1.66369379e-01
9.66976225e-01 -1.21940136e+00 3.50861311e-01 -3.80874723e-01
-4.09114569e-01 1.49556386e+00 7.38461465e-02 2.91284248e-02
4.98235136e-01 -2.04990104e-01 -6.06265008e-01 -3.65921170e-01
-7.05829620e-01 -2.67568588e-01 6.92358553e-01 5.97329497e-01
2.08980724e-01 -4.51976620e-02 -7.35022902e-01 2.43074566e-01
1.60197511e-01 -2.04136029e-01 -7.56037980e-02 9.27033961e-01
-4.68872458e-01 -1.69889200e+00 -4.70961124e-01 1.93752140e-01
-3.52205247e-01 -2.47965649e-01 -5.37960708e-01 5.32436669e-01
1.93280727e-01 8.62317979e-01 -5.75529218e-01 5.42030111e-02
3.05611223e-01 -3.17033976e-02 7.51053154e-01 -6.04005098e-01
-6.66819930e-01 2.28865482e-02 1.24527819e-01 -6.59356773e-01
-2.96008617e-01 -5.94767511e-01 -9.27855432e-01 8.24201256e-02
-6.45211816e-01 7.54256189e-01 1.22948265e+00 1.10988951e+00
2.37818405e-01 3.34603339e-01 8.10267270e-01 -9.95421708e-01
-4.95297074e-01 -2.53318787e-01 -3.45098108e-01 4.13980661e-03
4.20241617e-02 -6.85711980e-01 -9.17916000e-02 -2.35405058e-01]
|
[5.1240458488464355, 5.308169364929199]
|
398468a0-1a0d-4518-90c9-4db70c2e75ec
|
binaural-signal-representations-for-joint
|
2209.059
| null |
https://arxiv.org/abs/2209.05900v1
|
https://arxiv.org/pdf/2209.05900v1.pdf
|
Binaural Signal Representations for Joint Sound Event Detection and Acoustic Scene Classification
|
Sound event detection (SED) and Acoustic scene classification (ASC) are two widely researched audio tasks that constitute an important part of research on acoustic scene analysis. Considering shared information between sound events and acoustic scenes, performing both tasks jointly is a natural part of a complex machine listening system. In this paper, we investigate the usefulness of several spatial audio features in training a joint deep neural network (DNN) model performing SED and ASC. Experiments are performed for two different datasets containing binaural recordings and synchronous sound event and acoustic scene labels to analyse the differences between performing SED and ASC separately or jointly. The presented results show that the use of specific binaural features, mainly the Generalized Cross Correlation with Phase Transform (GCC-phat) and sines and cosines of phase differences, result in a better performing model in both separate and joint tasks as compared with baseline methods based on logmel energies only.
|
['Annamaria Mesaros', 'Daniel Aleksander Krause']
|
2022-09-13
| null | null | null | null |
['sound-event-detection', 'scene-classification']
|
['audio', 'computer-vision']
|
[ 3.58130276e-01 -6.06598556e-01 8.57613146e-01 -5.24904490e-01
-8.78456831e-01 -5.42754352e-01 8.77051830e-01 5.10164976e-01
-7.52853453e-01 3.73732358e-01 4.05554354e-01 -4.34246734e-02
-2.98740298e-01 -3.67865235e-01 -5.03269315e-01 -7.72400796e-01
-3.65487814e-01 -4.20121802e-03 4.49789792e-01 -1.40805215e-01
1.81603849e-01 4.64173079e-01 -2.13581657e+00 1.84971273e-01
2.50174135e-01 1.22992444e+00 3.47205251e-01 1.14480555e+00
7.14145228e-02 6.88622713e-01 -9.16081011e-01 4.49973764e-03
1.65534437e-01 -5.37266552e-01 -5.16251922e-01 -3.84388983e-01
4.76089358e-01 -4.74677458e-02 2.53171790e-02 1.00205767e+00
1.11439025e+00 5.65186322e-01 6.36774957e-01 -9.84982014e-01
2.72496402e-01 7.28088796e-01 -1.04923248e-01 6.54475212e-01
4.54457462e-01 8.14322755e-03 1.13019884e+00 -7.66796529e-01
-1.36571117e-02 1.00197184e+00 7.79934764e-01 -2.37533525e-01
-1.10455060e+00 -6.29652083e-01 -3.89126718e-01 6.61491454e-01
-1.28178728e+00 -7.23675847e-01 1.14951158e+00 -3.96490037e-01
1.00648844e+00 4.37187999e-01 4.06794488e-01 7.55875409e-01
3.69144715e-02 5.54692090e-01 1.26903450e+00 -7.57628202e-01
4.23904806e-01 1.95899695e-01 5.05994521e-02 -1.46342620e-01
-3.61803889e-01 3.21379513e-01 -8.05053711e-01 2.90676602e-03
2.18558758e-01 -6.40914559e-01 -3.81520867e-01 -2.20398102e-02
-8.72614622e-01 6.72259986e-01 2.23064065e-01 6.96442604e-01
-3.89200836e-01 2.60751247e-01 5.26642263e-01 2.12874964e-01
3.84021521e-01 5.06612659e-01 -3.14940065e-01 -5.34007192e-01
-1.06926286e+00 3.26222271e-01 7.92950928e-01 9.57868844e-02
6.95333123e-01 5.38639903e-01 4.23977803e-03 1.23481727e+00
1.84312940e-01 2.86018461e-01 5.42685568e-01 -5.78962922e-01
-1.03139199e-01 -1.42188773e-01 -1.19078971e-01 -9.53556836e-01
-7.79205322e-01 -6.29971087e-01 -4.05356318e-01 1.02450520e-01
5.50951779e-01 -2.47099787e-01 -6.96547925e-01 1.80399048e+00
2.79470474e-01 4.80654031e-01 -1.58148110e-01 9.51710880e-01
7.52918005e-01 7.09893227e-01 1.26295209e-01 -9.29042399e-02
1.53100216e+00 -4.47752893e-01 -8.30439925e-01 -1.39224634e-01
3.50560427e-01 -1.19037271e+00 1.05033278e+00 5.36994100e-01
-1.04500926e+00 -1.03353310e+00 -1.13073492e+00 5.32189235e-02
-6.84163928e-01 1.37640193e-01 2.40676135e-01 7.75869727e-01
-1.05423307e+00 5.47986567e-01 -6.56156480e-01 -1.88522920e-01
-7.00551569e-02 2.76622683e-01 -2.43761186e-02 3.81310433e-01
-1.30783653e+00 7.28903294e-01 1.82922572e-01 5.53630888e-02
-8.67123961e-01 -7.75321066e-01 -7.91297555e-01 3.58711481e-01
1.12091370e-01 4.22931872e-02 1.53846216e+00 -7.31027901e-01
-1.49206710e+00 5.59085727e-01 1.24327257e-01 -7.75881171e-01
2.03120887e-01 -5.73432326e-01 -7.37520218e-01 2.06441537e-01
-1.94450751e-01 4.41858172e-01 8.75734091e-01 -1.01870072e+00
-6.75881803e-01 -4.33089472e-02 -3.42715979e-01 3.18652123e-01
-2.06064314e-01 5.09724796e-01 7.51939788e-02 -6.42438173e-01
-5.67405559e-02 -5.42950809e-01 6.38442934e-02 -4.03623164e-01
-4.00693327e-01 -2.40419433e-01 7.60239422e-01 -8.66249442e-01
1.02132726e+00 -2.43549204e+00 -9.07242373e-02 8.78232941e-02
-2.99833924e-01 2.42891654e-01 1.56475250e-02 3.77469450e-01
-3.61289859e-01 -4.64578420e-01 -2.27800772e-01 -4.45366830e-01
3.27549845e-01 -1.75592646e-01 -3.61258924e-01 3.17348778e-01
1.75143257e-01 1.97420463e-01 -5.62103331e-01 -1.23412438e-01
5.01327038e-01 5.73951364e-01 -3.69948685e-01 2.24985570e-01
-5.17420545e-02 4.45342183e-01 1.95429757e-01 1.61861122e-01
6.00487053e-01 5.99752307e-01 -1.57537073e-01 -2.59822071e-01
-3.44557613e-01 7.61109829e-01 -1.54701412e+00 1.53864193e+00
-6.07643068e-01 1.07781768e+00 1.57367647e-01 -8.56716633e-01
9.68100071e-01 6.37929559e-01 3.26876879e-01 -9.01747406e-01
3.12188119e-01 3.90948772e-01 3.76988649e-01 -3.07708323e-01
4.78310645e-01 -3.69169563e-01 4.61122915e-02 4.41950679e-01
4.66337353e-01 -3.56579065e-01 1.50884986e-01 -2.95247346e-01
8.78188014e-01 -1.30033344e-01 1.86142474e-01 -3.21186453e-01
6.97959125e-01 -5.73494911e-01 1.71828911e-01 7.27833629e-01
-2.94867158e-01 6.83083594e-01 3.23731184e-01 -1.38335362e-01
-7.68848598e-01 -1.11397576e+00 -3.46887171e-01 1.34634602e+00
-2.05272898e-01 -3.52088302e-01 -6.62137628e-01 -1.29828930e-01
-4.17884707e-01 9.44572270e-01 -2.37221345e-01 -2.13420853e-01
-6.42309248e-01 -7.61971414e-01 8.14364910e-01 4.19786602e-01
3.48989934e-01 -1.28650248e+00 -9.40123677e-01 3.68485153e-01
-5.97984456e-02 -1.25694931e+00 -1.50091611e-02 8.50104749e-01
-2.34764621e-01 -7.09371269e-01 -5.38940847e-01 -5.45794904e-01
-4.35972363e-01 -4.74804752e-02 9.24638569e-01 -5.19244552e-01
-6.42681658e-01 5.60021281e-01 -4.66089815e-01 -9.32022333e-01
-4.22570825e-01 -2.72498190e-01 1.37469873e-01 2.87123203e-01
3.37513983e-01 -1.00640249e+00 -4.42988843e-01 5.03426120e-02
-1.08677554e+00 -3.08409423e-01 3.56767833e-01 2.89873004e-01
2.18583450e-01 3.38859648e-01 5.86441159e-01 -3.05960536e-01
7.31801510e-01 -2.72150695e-01 -4.83546764e-01 -2.21574634e-01
1.49935991e-01 -3.49191010e-01 5.46125233e-01 -3.67694587e-01
-1.06338060e+00 -2.07160544e-02 -5.18781245e-01 -2.38296360e-01
-9.15890574e-01 4.19694483e-01 -1.85711220e-01 9.97498855e-02
7.10584104e-01 3.20877761e-01 -4.47066188e-01 -8.25630546e-01
-8.48018900e-02 5.76714218e-01 6.13188148e-01 -2.75707960e-01
5.09289801e-01 3.08742136e-01 1.09695598e-01 -1.28395700e+00
-6.16148055e-01 -7.19183564e-01 -6.38016760e-01 -3.37084085e-01
1.04435420e+00 -7.36717820e-01 -4.17984098e-01 8.07388544e-01
-1.05388236e+00 -1.33759171e-01 -4.90939528e-01 1.00299144e+00
-4.09365177e-01 2.77859956e-01 -2.79277205e-01 -1.22760737e+00
9.20059755e-02 -1.21218586e+00 1.11418140e+00 1.58736467e-01
-2.08627090e-01 -7.97749817e-01 4.79549021e-01 6.31362274e-02
5.87737679e-01 1.76459059e-01 9.11499023e-01 -1.12130225e+00
-2.02602684e-01 -2.50878394e-01 1.11476175e-01 7.12350786e-01
1.77274212e-01 -9.50338766e-02 -1.76086581e+00 1.81058913e-01
3.68669599e-01 -2.08625272e-01 8.66408527e-01 6.21596158e-01
9.78244662e-01 2.04271331e-01 3.29996735e-01 3.76183003e-01
1.31764853e+00 5.47363281e-01 5.50953627e-01 9.44232866e-02
3.39190125e-01 7.89924800e-01 2.96816438e-01 7.19444394e-01
1.92898344e-02 9.02163804e-01 4.23814148e-01 -2.62406692e-02
-3.92566919e-01 1.09329797e-01 4.76504087e-01 9.95133936e-01
9.26847234e-02 -2.27959514e-01 -9.54097033e-01 6.92366958e-01
-1.12978244e+00 -7.55510867e-01 -3.92530680e-01 2.27929378e+00
7.27621794e-01 1.45418793e-01 2.00621411e-01 9.80799615e-01
6.31484210e-01 3.23318928e-01 1.23630814e-01 -7.20307887e-01
-2.52274156e-01 1.04515147e+00 6.70821741e-02 3.23825777e-01
-1.44511235e+00 5.28092504e-01 5.64248323e+00 1.01449752e+00
-1.32564235e+00 9.39087570e-02 1.84868529e-01 -1.07942678e-01
1.49988517e-01 -1.42134473e-01 -4.57921505e-01 2.74226427e-01
1.46128607e+00 3.99920493e-01 2.06730992e-01 5.32304168e-01
2.07273141e-01 -4.94383007e-01 -1.18097055e+00 8.49419117e-01
1.05599158e-01 -8.53832722e-01 -2.57989973e-01 -3.18706334e-01
2.59434342e-01 1.70509487e-01 2.07032338e-01 2.12307408e-01
-6.86430782e-02 -8.62021863e-01 9.61357236e-01 3.34308892e-01
3.18346798e-01 -7.16508687e-01 7.37692416e-01 6.04664795e-02
-1.28971899e+00 7.74514377e-02 -7.97694642e-03 -1.16638958e-01
3.77339751e-01 5.40989459e-01 -1.00456691e+00 7.07867742e-01
1.02553225e+00 3.70817453e-01 -5.40842235e-01 1.48627710e+00
-2.70818118e-02 1.07410991e+00 -7.50741243e-01 -4.99273697e-03
1.49811149e-01 1.36701018e-01 9.52661753e-01 1.62087166e+00
2.92006582e-01 -4.27616566e-01 -2.52939582e-01 6.34883463e-01
3.92043978e-01 3.65522027e-01 -2.95792371e-01 -5.50237261e-02
2.69708186e-01 1.22099364e+00 -8.60920072e-01 4.12130766e-02
-2.80675709e-01 4.79727060e-01 -3.31307381e-01 2.77665347e-01
-8.14229012e-01 -6.43552244e-01 6.56631231e-01 -1.44861653e-01
4.87692356e-01 -3.13759685e-01 -1.60224020e-01 -3.87692094e-01
-4.55657132e-02 -6.33852839e-01 3.28525931e-01 -9.43579495e-01
-9.55717504e-01 6.77606821e-01 1.00791037e-01 -1.24756348e+00
-2.72175878e-01 -5.45401037e-01 -9.89122391e-01 8.12016666e-01
-1.63808584e+00 -8.21380436e-01 -1.01529077e-01 6.55353248e-01
4.48099911e-01 -6.29175976e-02 8.36248040e-01 6.27449989e-01
-2.66294092e-01 4.01620328e-01 -7.23738000e-02 -4.94898818e-02
6.57715142e-01 -1.40474296e+00 -7.27097213e-04 7.60617435e-01
7.60865569e-01 1.26544282e-01 1.10887146e+00 -3.29976380e-02
-7.32953489e-01 -7.55908430e-01 8.41302335e-01 8.95767007e-03
6.99679077e-01 -5.63736439e-01 -9.28819001e-01 1.05789386e-01
4.90776598e-01 -1.81762740e-01 9.11815882e-01 -6.47888184e-02
-1.64861962e-01 -3.12053800e-01 -7.13783443e-01 1.87950015e-01
3.28995198e-01 -8.36165607e-01 -6.06756389e-01 1.76537275e-01
6.02367699e-01 -4.60875109e-02 -5.63131690e-01 5.19165516e-01
3.00915450e-01 -1.35919106e+00 9.23524976e-01 -9.94976759e-02
1.21523894e-01 -4.50660110e-01 -5.13697743e-01 -1.36661839e+00
1.82521939e-01 -5.58371365e-01 2.72542268e-01 1.28081119e+00
1.43094838e-01 -4.34183896e-01 1.19854659e-01 -2.50170112e-01
-4.77561831e-01 -1.52303815e-01 -1.14771497e+00 -6.54308438e-01
-2.46632025e-01 -1.10597551e+00 3.94810647e-01 7.15362906e-01
-5.00900626e-01 3.53982389e-01 -2.40054205e-01 4.08604026e-01
2.49905363e-01 -7.19842911e-02 6.49497092e-01 -1.40013742e+00
-4.92707223e-01 -5.37022591e-01 -6.93545640e-01 -5.63882649e-01
-1.04114503e-01 -5.65197885e-01 3.94666791e-01 -1.12114024e+00
-4.54281867e-01 -1.48360610e-01 -7.30586171e-01 1.59213930e-01
9.16368663e-02 4.38449472e-01 1.13165535e-01 -2.68071145e-01
-2.63947815e-01 6.28191113e-01 5.78414321e-01 2.65016705e-01
-1.73184291e-01 3.32061529e-01 -2.71286629e-02 9.80601132e-01
6.22071624e-01 -4.86955523e-01 -2.65102148e-01 -2.21755311e-01
1.94469720e-01 5.24475165e-02 6.62685812e-01 -1.62380564e+00
4.31171715e-01 3.72083902e-01 1.15506895e-01 -7.83837795e-01
9.20767009e-01 -7.41952002e-01 5.13879955e-02 2.04811305e-01
-3.64546567e-01 -2.38484368e-01 6.29553437e-01 4.46739942e-01
-6.96023881e-01 -3.36864233e-01 8.20763350e-01 1.00330323e-01
-7.55213499e-01 -4.64588612e-01 -7.24571824e-01 -2.23476633e-01
5.93495548e-01 -2.42272615e-01 1.18971236e-01 -6.12677515e-01
-9.39862370e-01 -4.87200677e-01 -5.09140193e-01 3.82053643e-01
3.78281146e-01 -9.19227779e-01 -6.85045779e-01 2.61332929e-01
-2.39086971e-01 -1.51888117e-01 7.07946301e-01 1.00229061e+00
-4.78523910e-01 5.37534475e-01 -2.81141669e-01 -7.89535522e-01
-1.43536496e+00 6.20555086e-03 5.57575941e-01 3.41131426e-02
-1.79348469e-01 1.32599044e+00 3.48864168e-01 -3.98875743e-01
5.86959660e-01 -5.22790849e-01 -5.53846836e-01 3.86310279e-01
3.47523183e-01 4.99610454e-01 4.88646120e-01 -7.45629132e-01
-3.38895321e-01 3.19950938e-01 2.36210302e-01 -4.95003998e-01
1.45674372e+00 -9.20834765e-02 -6.52379617e-02 9.33253407e-01
1.31275511e+00 2.49199226e-01 -9.75484788e-01 -2.63607144e-01
1.76014796e-01 -1.81229532e-01 4.07933146e-01 -9.37644422e-01
-7.22721457e-01 1.36121202e+00 9.76118386e-01 6.38387382e-01
1.43990338e+00 -7.16939345e-02 5.76441884e-01 6.16137013e-02
-6.92681894e-02 -1.11630177e+00 1.62390321e-01 5.22726655e-01
7.36167848e-01 -8.31403375e-01 -1.89455673e-01 -1.49368018e-01
-5.82830191e-01 1.10442412e+00 2.93497771e-01 -1.26637429e-01
8.97831500e-01 2.30671868e-01 2.03233808e-01 -1.07810602e-01
-4.67073709e-01 -5.33391416e-01 5.28710723e-01 4.28495824e-01
4.73607868e-01 -5.13390303e-02 3.76360640e-02 7.00521529e-01
-7.81715930e-01 -6.45637512e-01 3.24000388e-01 8.62003624e-01
-5.07642269e-01 -8.48034143e-01 -4.87532437e-01 1.57741949e-01
-8.40174079e-01 -1.41508013e-01 -4.13191885e-01 6.61346793e-01
3.53798717e-01 1.02441418e+00 3.43365014e-01 -4.74763125e-01
3.45685780e-01 4.99098241e-01 2.89262295e-01 -5.36614597e-01
-9.74342942e-01 5.65335870e-01 6.60074130e-02 -1.84572190e-01
-6.11905515e-01 -7.78540254e-01 -1.06251895e+00 3.89084578e-01
-4.86606866e-01 6.52070865e-02 1.26377606e+00 1.08263183e+00
-4.48350385e-02 1.01236999e+00 6.87442362e-01 -9.92926657e-01
-3.31577569e-01 -1.21284199e+00 -8.10401797e-01 2.29283154e-01
5.66464782e-01 -5.58031678e-01 -6.56826973e-01 1.15228601e-01]
|
[15.19804573059082, 5.4057135581970215]
|
aa7d701f-8882-4e24-b9aa-73ccc0aa210d
|
weakly-supervised-facial-action-unit
| null | null |
http://openaccess.thecvf.com/content_cvpr_2018/html/Peng_Weakly_Supervised_Facial_CVPR_2018_paper.html
|
http://openaccess.thecvf.com/content_cvpr_2018/papers/Peng_Weakly_Supervised_Facial_CVPR_2018_paper.pdf
|
Weakly Supervised Facial Action Unit Recognition Through Adversarial Training
|
Current works on facial action unit (AU) recognition typically require fully AU-annotated facial images for supervised AU classifier training. AU annotation is a time-consuming, expensive, and error-prone process. While AUs are hard to annotate, facial expression is relatively easy to label. Furthermore, there exist strong probabilistic dependencies between expressions and AUs as well as dependencies among AUs. Such dependencies are referred to as domain knowledge. In this paper, we propose a novel AU recognition method that learns AU classifiers from domain knowledge and expression-annotated facial images through adversarial training. Specifically, we first generate pseudo AU labels according to the probabilistic dependencies between expressions and AUs as well as correlations among AUs summarized from domain knowledge. Then we propose a weakly supervised AU recognition method via an adversarial process, in which we simultaneously train two models: a recognition model R, which learns AU classifiers, and a discrimination model D, which estimates the probability that AU labels generated from domain knowledge rather than the recognized AU labels from R. The training procedure for R maximizes the probability of D making a mistake. By leveraging the adversarial mechanism, the distribution of recognized AUs is closed to AU prior distribution from domain knowledge. Furthermore, the proposed weakly supervised AU recognition can be extended to semi-supervised learning scenarios with partially AU-annotated images. Experimental results on three benchmark databases demonstrate that the proposed method successfully leverages the summarized domain knowledge to weakly supervised AU classifier learning through an adversarial process, and thus achieves state-of-the-art performance.
|
['Shangfei Wang', 'Guozhu Peng']
|
2018-06-01
| null | null | null |
cvpr-2018-6
|
['facial-action-unit-detection']
|
['computer-vision']
|
[ 5.79185486e-01 4.70504016e-01 -2.34588474e-01 -6.37373030e-01
-9.87332582e-01 -5.62777758e-01 4.33149338e-01 -5.11743128e-01
-5.88995442e-02 7.36364484e-01 -1.07358478e-01 3.15413684e-01
5.40830195e-01 -9.46123064e-01 -9.18498874e-01 -1.08479810e+00
3.23423505e-01 3.13562363e-01 -2.97663450e-01 -5.95485754e-02
-5.74386775e-01 3.13522637e-01 -1.52670228e+00 3.07535052e-01
6.58445179e-01 1.61568439e+00 -5.02952099e-01 4.25287217e-01
-2.31316756e-03 1.30446374e+00 -6.02388918e-01 -6.74721301e-01
2.60260701e-01 -7.55990326e-01 -7.11195648e-01 5.60635507e-01
2.81111538e-01 -6.12593532e-01 -4.00369465e-01 1.26482356e+00
1.71128094e-01 -1.86447967e-02 9.22525346e-01 -1.53873718e+00
-9.69752669e-01 2.86965698e-01 -5.33253729e-01 -5.52523732e-01
4.06167835e-01 8.22889954e-02 1.01829302e+00 -8.29373121e-01
4.10378784e-01 1.16804159e+00 3.60195935e-01 1.13820505e+00
-9.96715903e-01 -1.11833918e+00 2.00055405e-01 -6.63902098e-03
-1.60048771e+00 -3.90354782e-01 9.10122216e-01 -3.87896419e-01
4.27796125e-01 -2.36146338e-02 3.66309881e-01 1.40448022e+00
-3.80588561e-01 1.12745106e+00 1.33025527e+00 -3.36628109e-01
2.42339939e-01 1.76261291e-01 -2.14382429e-02 9.15644586e-01
-4.37270582e-01 1.90685615e-01 -3.90272766e-01 -3.24043036e-01
6.30413234e-01 -1.10274861e-02 -1.34446755e-01 5.79513684e-02
-4.01553005e-01 6.15824640e-01 2.62023419e-01 -3.27446982e-02
-2.73302406e-01 1.02048092e-01 3.66225004e-01 2.52345979e-01
5.32829165e-01 -2.41804227e-01 -2.11160406e-01 -8.95655826e-02
-3.08641344e-01 -6.33682758e-02 6.71311080e-01 1.16578972e+00
1.16434968e+00 1.66018680e-01 -2.20621005e-01 9.52763200e-01
2.94758886e-01 8.70902419e-01 2.41542876e-01 -7.54625738e-01
1.56418219e-01 7.18419790e-01 1.00981601e-01 -8.13119948e-01
5.04537560e-02 1.57389835e-01 -7.45132148e-01 2.62920469e-01
5.18528998e-01 -4.33678091e-01 -8.08470488e-01 2.15102696e+00
2.54559487e-01 5.74706256e-01 4.75941300e-01 8.69275510e-01
8.75591516e-01 6.42494202e-01 2.84624487e-01 -3.62781376e-01
9.82844293e-01 -8.31962764e-01 -8.65131497e-01 -6.49455353e-04
8.18240464e-01 -4.66929764e-01 8.53745937e-01 2.69565731e-01
-7.83912122e-01 -3.36637050e-01 -7.43276358e-01 1.67870700e-01
6.14526346e-02 6.34162605e-01 5.85953057e-01 6.53367639e-01
-6.35039210e-01 2.40466654e-01 -7.42898107e-01 -8.07992667e-02
8.78951490e-01 4.55033392e-01 -6.33987963e-01 -1.67311773e-01
-1.12341869e+00 5.87251902e-01 2.72796810e-01 2.50956297e-01
-1.29468966e+00 -2.43038371e-01 -1.26045132e+00 -1.37123451e-01
5.54554880e-01 -1.03173681e-01 1.36466050e+00 -1.84499788e+00
-1.87191784e+00 1.21324587e+00 -2.57726461e-01 -2.08265483e-01
4.42087531e-01 -2.38353595e-01 -4.37630773e-01 2.25079894e-01
-1.19919352e-01 4.25290972e-01 1.24050367e+00 -1.38984537e+00
-4.96775895e-01 -5.14412582e-01 9.81495753e-02 8.11929554e-02
-5.24159074e-01 7.44530633e-02 -2.87148684e-01 -5.13454139e-01
-3.37645084e-01 -1.02680039e+00 1.84455529e-01 4.09269631e-01
-2.04349488e-01 -5.70207119e-01 1.01799214e+00 -4.09551144e-01
8.73313129e-01 -2.28222418e+00 -4.82315235e-02 3.08826596e-01
4.87977825e-02 2.70635784e-01 -2.39727184e-01 -1.65249854e-01
-1.01496778e-01 -2.81667084e-01 -3.37596595e-01 -3.40168983e-01
8.56396481e-02 7.12709129e-01 -4.59209859e-01 5.31799436e-01
8.08056414e-01 9.39570189e-01 -1.12221217e+00 -4.52127576e-01
-5.14752865e-02 2.93330789e-01 -2.40660518e-01 8.73328447e-01
-3.13605726e-01 5.05515933e-01 -6.82118058e-01 1.22295344e+00
7.07746983e-01 -6.69871941e-02 1.99021012e-01 -1.78012952e-01
6.12358987e-01 -4.84899402e-01 -7.95070946e-01 1.32996488e+00
-5.45822918e-01 1.59293309e-01 1.84141770e-01 -1.33346629e+00
1.41287398e+00 4.14979249e-01 3.82309556e-01 -2.51425862e-01
6.04785740e-01 1.52447879e-01 -3.56385857e-01 -4.40332919e-01
-2.46721089e-01 -5.09526670e-01 -2.26633489e-01 6.62554145e-01
4.02884960e-01 -1.31396368e-01 -2.55713791e-01 -8.49829018e-02
1.01167428e+00 4.90671366e-01 3.57059270e-01 3.10946018e-01
8.09993565e-01 -2.60854274e-01 8.15843523e-01 2.82161087e-01
-5.01633167e-01 2.81959146e-01 6.43469512e-01 -3.26985538e-01
-4.98470426e-01 -1.11907399e+00 -1.08074024e-01 1.10498250e+00
7.64821321e-02 -1.72896519e-01 -9.39962327e-01 -1.40927815e+00
-1.58234283e-01 2.94841439e-01 -1.08538485e+00 -5.47871768e-01
-1.36111140e-01 -4.17082429e-01 9.34687078e-01 7.40294933e-01
6.05343759e-01 -1.13927674e+00 4.56890091e-02 -1.68180898e-01
9.60228667e-02 -1.39120448e+00 -3.77926707e-01 -5.63491136e-03
-3.19979608e-01 -1.23371124e+00 -7.47254670e-01 -8.33249450e-01
1.10567367e+00 -1.79425925e-01 7.39331961e-01 -1.01591535e-01
-1.62520148e-02 5.67281544e-01 -5.58097661e-01 -4.83725190e-01
-6.41390383e-01 -7.54503787e-01 4.00485396e-01 8.81626606e-01
7.96236694e-01 -4.50531095e-01 -1.97332293e-01 5.20012259e-01
-9.61856127e-01 -3.24414611e-01 7.39872694e-01 1.24891496e+00
8.89727473e-01 -1.00749306e-01 5.77499747e-01 -1.08517969e+00
2.02875808e-01 -5.45171797e-01 -4.27610010e-01 3.52084011e-01
-1.52505532e-01 -1.27197146e-01 8.46015930e-01 -7.71351874e-01
-1.44473290e+00 6.10660851e-01 -2.04002008e-01 -1.03203285e+00
-4.97643679e-01 1.47864878e-01 -6.88316584e-01 -3.19239736e-01
6.49813116e-01 2.62516797e-01 2.87787527e-01 7.59816393e-02
4.36895847e-01 9.38351095e-01 6.33541226e-01 -1.01249802e+00
8.02258015e-01 7.22297072e-01 7.89752044e-03 -7.52940953e-01
-1.35013962e+00 -1.36488602e-01 -6.43879652e-01 -4.28383350e-01
8.77320051e-01 -1.14394855e+00 -5.84563553e-01 8.01635861e-01
-1.00843751e+00 -4.52058345e-01 -2.67933607e-01 1.97152823e-01
-7.21031725e-01 3.92559975e-01 -5.81226587e-01 -9.42211509e-01
-1.50850609e-01 -9.67592359e-01 1.25916326e+00 2.96311468e-01
-2.08055675e-01 -8.84729207e-01 -2.13105842e-01 6.02317810e-01
-3.20989996e-01 6.25122249e-01 4.00160849e-01 -5.76021433e-01
-2.32838035e-01 -5.57289124e-01 -3.71647805e-01 1.05467010e+00
5.26144147e-01 -7.51492055e-03 -1.26754975e+00 1.09682202e-01
-2.00959846e-01 -1.35894358e+00 3.92396271e-01 -2.29698107e-01
1.32466495e+00 -6.01745129e-01 -1.03088364e-01 5.00698984e-01
9.14883196e-01 7.48084262e-02 5.07319093e-01 -3.50482464e-01
7.57118404e-01 7.51962066e-01 1.04680014e+00 7.11366177e-01
6.24588840e-02 6.48855031e-01 5.01174569e-01 -3.12033892e-02
1.91296011e-01 -4.94517356e-01 7.26562619e-01 3.47422510e-01
-1.32271707e-01 1.01466373e-01 -5.28705478e-01 1.82677478e-01
-1.87081420e+00 -8.10780883e-01 3.02563608e-01 1.95325589e+00
1.11703777e+00 -2.04715773e-01 3.95043753e-02 -9.15727615e-02
5.44955671e-01 3.58831882e-02 -7.97425866e-01 -4.11971688e-01
-2.16697514e-01 4.07781571e-01 7.55737275e-02 2.33307630e-01
-1.18286288e+00 1.18660462e+00 5.42219162e+00 9.37162876e-01
-1.01181459e+00 2.76437163e-01 7.83205628e-01 1.72864139e-01
-9.25349593e-02 -2.40854159e-01 -6.42209411e-01 3.35169941e-01
6.03219569e-01 -8.44462439e-02 1.85855538e-01 1.34467018e+00
-8.47180486e-02 2.05914527e-01 -1.29439223e+00 1.17552292e+00
5.36227465e-01 -8.00078511e-01 2.15722799e-01 -1.02015980e-01
9.07251000e-01 -5.51000476e-01 -1.21753467e-02 5.53390265e-01
5.99144518e-01 -1.12162006e+00 5.00720620e-01 3.90343845e-01
1.13949394e+00 -8.33572507e-01 8.20754409e-01 3.62940490e-01
-1.15360284e+00 -3.20232213e-02 -3.07531476e-01 -1.84767306e-01
-1.80424660e-01 1.36334822e-01 -4.73538846e-01 3.20947886e-01
6.03535116e-01 1.09028840e+00 -9.76085197e-03 -1.68523967e-01
-9.61134195e-01 6.97402358e-01 -2.26285547e-01 -4.88691442e-02
2.05405757e-01 -3.88944805e-01 2.69605294e-02 9.91393268e-01
6.08575679e-02 5.64898789e-01 3.79516155e-01 8.99697602e-01
-4.74882692e-01 2.25749731e-01 -8.21477056e-01 -1.99808821e-01
1.94994643e-01 1.30011439e+00 -2.57073399e-02 -2.56867200e-01
-7.33716547e-01 1.34693527e+00 6.95246279e-01 3.64811271e-01
-8.70353341e-01 1.63030282e-01 8.02588224e-01 -3.88651013e-01
-2.67239995e-02 2.81533569e-01 1.60506770e-01 -1.12720692e+00
6.36677295e-02 -8.72127533e-01 5.03202856e-01 -8.00945640e-01
-1.59627390e+00 7.38167405e-01 -2.61370242e-01 -1.53572917e+00
-2.91425049e-01 -8.72186005e-01 -5.68454504e-01 7.23392248e-01
-1.50866497e+00 -1.57136977e+00 -5.01838982e-01 9.62039351e-01
1.55547664e-01 -2.97345757e-01 1.19558847e+00 4.25544232e-02
-6.43655717e-01 1.08218479e+00 -2.64947027e-01 7.08851039e-01
7.59682953e-01 -1.00233209e+00 -6.17617249e-01 6.58537805e-01
1.64206862e-01 1.30626187e-01 2.03682125e-01 -4.09920871e-01
-1.17109096e+00 -1.42350578e+00 3.38676900e-01 -3.44307184e-01
9.83802676e-01 -3.61739427e-01 -8.59644353e-01 8.61460388e-01
-2.51095086e-01 7.61405647e-01 1.20270336e+00 -1.37579352e-01
-7.36785412e-01 -1.98635116e-01 -1.19616044e+00 3.29398245e-01
8.90338421e-01 -8.18217516e-01 -3.74576330e-01 4.57351327e-01
1.81051180e-01 -4.40726846e-01 -9.87602055e-01 5.06320536e-01
5.16235769e-01 -6.61032319e-01 4.23923820e-01 -7.63902009e-01
5.30397236e-01 -2.24933818e-01 -2.88771600e-01 -1.10979760e+00
2.71189094e-01 -6.42448664e-01 -3.10442299e-01 1.46224284e+00
2.75586247e-01 -4.56700146e-01 1.09035039e+00 8.22112858e-01
1.64861858e-01 -8.47820520e-01 -8.92688215e-01 -8.06231439e-01
-4.64166747e-03 -4.98520643e-01 4.25162196e-01 1.03256059e+00
-2.20073573e-02 2.31239989e-01 -5.74165523e-01 3.11959326e-01
6.12211823e-01 1.98950544e-01 9.93334234e-01 -8.20739269e-01
-3.55503917e-01 -1.37482464e-01 -6.05536401e-01 -1.01792812e+00
1.18780899e+00 -7.47219980e-01 4.05077070e-01 -7.75412202e-01
1.78483516e-01 -4.75033909e-01 -1.70293614e-01 9.54952240e-01
-3.44976395e-01 4.71932471e-01 -1.62444368e-01 2.27844536e-01
-5.32843411e-01 9.97408509e-01 1.30226445e+00 -4.32673723e-01
8.68421048e-02 1.99328929e-01 -3.79278839e-01 1.02787685e+00
5.36706209e-01 -2.12768957e-01 -4.47689682e-01 -6.44026026e-02
-2.15620622e-01 4.29449193e-02 2.40815103e-01 -5.81055880e-01
-2.18412653e-01 -1.93717763e-01 1.86537355e-01 9.45982113e-02
4.76462692e-01 -1.00229394e+00 -4.07972634e-01 -7.98714627e-03
-1.68567240e-01 -7.57500708e-01 7.37133846e-02 7.75673628e-01
-5.73159993e-01 -4.36631627e-02 9.02945876e-01 -4.07222435e-02
-8.88234138e-01 8.06135416e-01 -1.53394908e-01 -5.40118814e-02
1.56840312e+00 -6.92833140e-02 -2.98190359e-02 -7.08580971e-01
-1.05970788e+00 1.91389337e-01 2.59244412e-01 4.21051532e-01
6.31515682e-01 -1.59316242e+00 -6.83197141e-01 1.77098513e-01
6.65407658e-01 1.43048123e-01 1.52939364e-01 5.13396502e-01
8.99169035e-03 -1.75133556e-01 -2.38740444e-01 -5.67964077e-01
-1.52185380e+00 4.38706785e-01 4.24934477e-01 -3.77341993e-02
-5.06090932e-02 1.11515355e+00 5.01040518e-01 -5.39463937e-01
3.02340180e-01 1.94713816e-01 -1.53580487e-01 -9.65111628e-02
7.27835238e-01 -1.03080846e-01 -3.09890836e-01 -1.24412847e+00
-1.63598627e-01 6.28814697e-01 3.65680940e-02 2.48168766e-01
8.88996661e-01 7.62871429e-02 -1.18937559e-01 2.64528722e-01
1.25005579e+00 -1.79677218e-01 -1.63031101e+00 -6.47582531e-01
-3.85598600e-01 -3.71436149e-01 -2.84596890e-01 -5.11164129e-01
-1.18157399e+00 9.07193542e-01 3.65604937e-01 -3.53232503e-01
1.44182706e+00 2.90811241e-01 7.19034910e-01 3.70720327e-01
4.62781370e-01 -9.88709509e-01 4.60577786e-01 2.47788280e-01
7.87227929e-01 -1.49386692e+00 -3.25783640e-01 -7.19209373e-01
-1.07067323e+00 1.02896070e+00 1.20971859e+00 -2.05908462e-01
7.31516540e-01 3.32400292e-01 4.91704315e-01 8.72688517e-02
-5.00079274e-01 -4.24152106e-01 2.04167753e-01 8.32082987e-01
1.61917746e-01 2.15887800e-01 2.37973165e-02 1.14032948e+00
1.61503702e-01 2.17445150e-01 2.03953490e-01 7.65855312e-01
-7.48064369e-02 -1.17040300e+00 -2.51882851e-01 2.50015050e-01
-4.51125652e-01 2.19754979e-01 -8.28872561e-01 5.48414946e-01
3.10694456e-01 8.34571958e-01 -1.53995969e-03 -6.34735405e-01
1.93072036e-01 1.12713501e-01 5.68813264e-01 -7.60698378e-01
-1.41556486e-02 -2.37120137e-01 5.85302524e-02 -6.96957052e-01
-6.36596441e-01 -5.90849638e-01 -1.42703152e+00 1.32640898e-01
-2.69064873e-01 1.14992633e-01 5.71589470e-02 1.27757609e+00
6.41516894e-02 5.17460443e-02 1.06698883e+00 -5.28772354e-01
-6.72170639e-01 -1.03028870e+00 -7.36312032e-01 9.14820850e-01
4.60004248e-02 -8.67456257e-01 -3.76643181e-01 4.02385235e-01]
|
[13.634804725646973, 1.5684739351272583]
|
3a06a97a-f88c-42ca-853b-3a2a9088b8d2
|
deep-representation-of-facial-geometric-and
|
1511.03015
| null |
http://arxiv.org/abs/1511.03015v1
|
http://arxiv.org/pdf/1511.03015v1.pdf
|
Deep Representation of Facial Geometric and Photometric Attributes for Automatic 3D Facial Expression Recognition
|
In this paper, we present a novel approach to automatic 3D Facial Expression
Recognition (FER) based on deep representation of facial 3D geometric and 2D
photometric attributes. A 3D face is firstly represented by its geometric and
photometric attributes, including the geometry map, normal maps, normalized
curvature map and texture map. These maps are then fed into a pre-trained deep
convolutional neural network to generate the deep representation. Then the
facial expression prediction is simplyachieved by training linear SVMs over the
deep representation for different maps and fusing these SVM scores. The
visualizations show that the deep representation provides a complete and highly
discriminative coding scheme for 3D faces. Comprehensive experiments on the
BU-3DFE database demonstrate that the proposed deep representation can
outperform the widely used hand-crafted descriptors (i.e., LBP, SIFT, HOG,
Gabor) and the state-of-art approaches under the same experimental protocols.
|
['Zongben Xu', 'Liming Chen', 'Huibin Li', 'Jian Sun', 'Dong Wang']
|
2015-11-10
| null | null | null | null |
['3d-facial-expression-recognition']
|
['computer-vision']
|
[-5.67238405e-02 -1.62820801e-01 -1.99988373e-02 -9.11122382e-01
-3.81590098e-01 -2.44074374e-01 6.48454249e-01 -2.46777579e-01
-2.72021797e-02 3.28555167e-01 -9.43089128e-02 4.72384356e-02
1.98116362e-01 -8.04922223e-01 -5.39253056e-01 -8.50258589e-01
-3.05459321e-01 2.86162168e-01 -2.91097373e-01 -3.67676169e-01
3.83013070e-01 1.24274719e+00 -2.07319856e+00 2.67053276e-01
1.67081431e-01 1.85044920e+00 -5.26883841e-01 3.92111957e-01
-3.88487965e-01 6.97866797e-01 -5.40344000e-01 -4.39138561e-01
1.96668699e-01 -1.84434876e-01 -5.38980246e-01 2.65106916e-01
6.16900027e-01 -5.62529922e-01 -2.68105805e-01 9.65746343e-01
4.60412681e-01 -4.49347161e-02 1.07399464e+00 -1.39178252e+00
-7.23403156e-01 -5.27317584e-01 -8.17441940e-01 -1.30695775e-01
7.48068869e-01 -3.07890654e-01 5.44437766e-01 -1.34357059e+00
6.60061598e-01 1.54553735e+00 6.77453935e-01 3.17197829e-01
-1.06878233e+00 -7.76617110e-01 -2.64947534e-01 1.90453559e-01
-1.52937686e+00 -5.20709217e-01 1.12459803e+00 -5.78989446e-01
1.01795554e+00 1.33089647e-01 7.30446637e-01 1.02262866e+00
2.47221619e-01 6.20156050e-01 1.35627604e+00 -3.90465587e-01
1.49487972e-01 5.59800155e-02 -2.10295260e-01 1.31451786e+00
-3.21810365e-01 -3.44503969e-02 -4.42310691e-01 -2.97588080e-01
8.30844939e-01 2.76397951e-02 1.94717482e-01 -4.09337014e-01
-3.76071692e-01 8.87781560e-01 3.97644252e-01 1.02712370e-01
-6.27956450e-01 2.43731905e-02 4.63162035e-01 4.24766570e-01
8.32822382e-01 -1.69211701e-01 -2.38578916e-01 -1.87547803e-02
-7.75945246e-01 3.44263494e-01 6.68753982e-01 6.86767876e-01
1.25529289e+00 2.17239276e-01 -9.08849612e-02 1.01427519e+00
4.72222090e-01 6.73862517e-01 3.28313529e-01 -9.13951814e-01
-1.31761536e-01 8.13303590e-01 -9.26319286e-02 -1.62437761e+00
-4.69646186e-01 2.62777358e-01 -8.51010621e-01 7.91030228e-01
1.10313848e-01 7.48546869e-02 -1.02568531e+00 1.46678638e+00
4.03044820e-01 3.38632683e-03 -1.51824817e-01 1.02360058e+00
1.19235563e+00 5.23697674e-01 9.01311710e-02 2.21804008e-01
1.13452566e+00 -5.47704458e-01 -6.73191249e-01 3.07595044e-01
5.26432574e-01 -7.57455111e-01 5.18981040e-01 2.90880024e-01
-9.56056237e-01 -7.89074242e-01 -1.00770211e+00 -9.44592357e-02
-6.22609675e-01 5.20642519e-01 6.84012413e-01 5.89173436e-01
-1.29041791e+00 5.89476228e-01 -6.30697072e-01 -3.43533725e-01
8.18214178e-01 6.02573574e-01 -8.84900868e-01 1.51360616e-01
-1.04242778e+00 9.07312393e-01 -4.34899926e-02 1.19543537e-01
-8.09303820e-01 -2.12163046e-01 -1.17978799e+00 2.31618602e-02
-4.34629411e-01 -8.53705555e-02 8.42084765e-01 -1.22654021e+00
-1.92684937e+00 1.59858489e+00 -2.01121002e-01 1.90617874e-01
1.28760338e-01 -1.77249741e-02 -3.72827351e-01 3.06524307e-01
-9.61462855e-02 7.71459222e-01 1.06004024e+00 -1.00731516e+00
-1.48744047e-01 -8.06168735e-01 -2.87301205e-02 5.34128100e-02
-2.49099895e-01 3.23094726e-01 -2.13726982e-01 -4.06337023e-01
2.21799433e-01 -7.78001964e-01 1.20217375e-01 5.23160458e-01
-2.55090117e-01 -4.08999562e-01 1.09765100e+00 -6.89389825e-01
5.60798109e-01 -2.24970961e+00 1.61652729e-01 4.89446461e-01
2.22108345e-02 2.24368080e-01 -1.77295402e-01 7.05099627e-02
-4.08120573e-01 -1.72910869e-01 1.27920538e-01 -3.87913138e-01
2.15168595e-01 -1.00100860e-02 -3.08580399e-02 8.45382154e-01
7.62016237e-01 6.83391213e-01 -6.16287410e-01 -4.84234393e-01
4.11308378e-01 9.15321589e-01 -3.76234889e-01 4.46404666e-01
7.13867173e-02 1.60354257e-01 -4.80329186e-01 1.02834380e+00
1.15508854e+00 8.47672597e-02 -3.49380411e-02 -2.14573234e-01
4.94237617e-02 -1.27878472e-01 -8.25442255e-01 1.56568539e+00
-4.31076944e-01 7.22835839e-01 1.08663291e-01 -1.20007110e+00
1.80979514e+00 2.83237129e-01 5.65223277e-01 -6.20299816e-01
6.05292201e-01 2.55798578e-01 -6.49460375e-01 -6.66476488e-01
1.15869127e-01 1.44852372e-02 -8.31499100e-02 4.01517570e-01
5.08508027e-01 1.30332671e-02 -4.33593750e-01 -4.30813164e-01
6.43877447e-01 3.70298922e-01 9.86364633e-02 -2.75114357e-01
9.42942381e-01 -4.07343894e-01 1.12851672e-01 4.99836765e-02
-2.77571738e-01 3.59233081e-01 8.25357676e-01 -8.72786045e-01
-9.65709269e-01 -8.30976963e-01 -3.47118616e-01 1.11563694e+00
-7.01884106e-02 -2.37075880e-01 -8.61961186e-01 -7.12195277e-01
3.79829139e-01 -2.03712266e-02 -1.05839729e+00 -1.47642612e-01
-3.28683943e-01 -3.82006913e-01 5.01044452e-01 4.09970164e-01
5.76605678e-01 -9.15928006e-01 -5.58611631e-01 -1.93906277e-01
3.35297823e-01 -8.98859501e-01 1.06637172e-01 1.53265268e-01
-5.16782045e-01 -7.26510465e-01 -7.99709439e-01 -1.05801308e+00
8.11231732e-01 -2.71144286e-02 6.39391005e-01 1.36276662e-01
-4.44343418e-01 4.08305228e-02 -2.09617570e-01 -2.88102001e-01
-2.30837509e-01 -4.23488349e-01 1.18374690e-01 5.60721099e-01
9.16626453e-01 -5.81366956e-01 -4.94343638e-01 1.77325204e-01
-4.01648194e-01 -1.32883757e-01 5.01758397e-01 8.90502930e-01
6.95026577e-01 -3.72936457e-01 1.82490572e-01 -4.13111806e-01
4.77852643e-01 -3.13196808e-01 -5.61155915e-01 7.05975816e-02
-5.04859313e-02 1.07127257e-01 3.34478170e-01 -2.78130710e-01
-9.17401791e-01 4.21820432e-01 -4.59601402e-01 -6.87138140e-01
-5.91750801e-01 2.58342862e-01 -3.01390201e-01 -6.92228556e-01
6.03413701e-01 1.78530544e-01 4.65549201e-01 -4.88879442e-01
1.71375751e-01 9.80722070e-01 2.57669538e-01 -5.11730611e-01
4.42708015e-01 5.40532947e-01 4.42355841e-01 -8.56352091e-01
-6.05600417e-01 -3.30832377e-02 -9.86521363e-01 -4.56233948e-01
9.48620260e-01 -8.66849065e-01 -9.00541902e-01 7.76063621e-01
-1.45270097e+00 1.07432462e-01 2.74641186e-01 1.30161449e-01
-8.43061566e-01 8.56711715e-02 -6.90897226e-01 -9.41489756e-01
-2.11266011e-01 -1.26339853e+00 1.70184267e+00 2.15885043e-01
-1.31156728e-01 -7.96801448e-01 -1.31921442e-02 8.14627763e-03
3.41776073e-01 9.02907431e-01 8.60869050e-01 -4.04204428e-01
-1.15603104e-01 -4.81428862e-01 -5.04935563e-01 3.99625748e-01
4.97696213e-02 3.02246273e-01 -1.47002113e+00 3.65673602e-02
-1.33854464e-01 -7.53229320e-01 5.97670197e-01 1.44404799e-01
1.65598452e+00 -1.63726538e-01 -9.54378098e-02 1.03649449e+00
1.24051023e+00 2.33715296e-01 5.15560806e-01 1.78884774e-01
4.02065963e-01 7.87094772e-01 5.30991316e-01 7.52128780e-01
1.75948933e-01 8.55387628e-01 5.62055707e-01 -5.18004060e-01
2.02808946e-01 -1.55521736e-01 2.43004367e-01 2.60442078e-01
-4.94724005e-01 4.14616644e-01 -6.87987626e-01 -1.61376312e-01
-1.60974860e+00 -9.66158748e-01 3.34838480e-01 1.76076770e+00
4.74141806e-01 -2.79257029e-01 -3.17283794e-02 2.10475072e-01
5.86607277e-01 1.76201463e-01 -2.87155837e-01 -1.08048034e+00
-2.45037198e-01 7.08845615e-01 1.19438162e-02 2.38563538e-01
-1.34654963e+00 1.16085362e+00 6.49674892e+00 5.61135411e-01
-1.64512300e+00 -2.04277277e-01 9.42576051e-01 2.56170779e-01
4.54058982e-02 -6.89323664e-01 -7.19938576e-01 1.31951988e-01
8.16770375e-01 9.63973403e-02 2.21543357e-01 1.18767607e+00
-1.18430883e-01 1.90267578e-01 -9.38344240e-01 1.38300085e+00
4.81490433e-01 -1.24866223e+00 1.42637372e-01 2.23678932e-01
4.64680076e-01 -3.51376295e-01 2.36216187e-01 1.64565429e-01
4.24849913e-02 -1.38998413e+00 6.17936075e-01 6.39222860e-01
1.21302378e+00 -9.29919779e-01 7.79825687e-01 -3.08268428e-01
-8.93564165e-01 -3.01285610e-02 -6.63006365e-01 -1.26449682e-04
-2.73198754e-01 9.83417183e-02 -6.61951959e-01 2.45854050e-01
8.85032833e-01 8.86239946e-01 -3.67871940e-01 4.19154465e-01
-1.98039040e-01 7.05000758e-02 -1.52831838e-01 -1.35216355e-01
3.19229037e-01 -3.17581445e-01 3.55169512e-02 1.44293535e+00
5.74559927e-01 1.04033060e-01 -1.80411130e-01 9.21334028e-01
-1.22197252e-02 3.40347052e-01 -8.77677381e-01 -5.97582199e-02
7.63123408e-02 1.56256557e+00 -4.54237759e-01 -3.08940917e-01
-4.58405107e-01 1.11217022e+00 4.53867793e-01 4.01368499e-01
-7.46377349e-01 -6.17390931e-01 1.09319282e+00 2.27660127e-02
4.53992248e-01 -1.53310880e-01 4.49888371e-02 -8.05435002e-01
-2.44724959e-01 -5.82851946e-01 1.46408260e-01 -1.11136448e+00
-1.07857478e+00 9.72372413e-01 -2.85639673e-01 -9.77868676e-01
-4.40683872e-01 -1.18882990e+00 -5.24240732e-01 1.08917618e+00
-1.58317983e+00 -1.22706521e+00 -6.03311837e-01 9.42870915e-01
1.11074257e-03 -5.92309356e-01 1.39505374e+00 2.77658373e-01
-3.29171658e-01 7.43695617e-01 -1.16017029e-01 3.98125976e-01
5.02076805e-01 -9.22937810e-01 2.15135440e-01 -5.93354069e-02
-1.73471719e-02 1.70504034e-01 3.60220611e-01 -7.84025639e-02
-1.53198719e+00 -9.76882935e-01 9.33010995e-01 -1.55967608e-01
4.85367030e-01 -7.40462542e-01 -8.31145465e-01 3.17573160e-01
-1.13280587e-01 4.23298568e-01 8.07972252e-01 -7.43880421e-02
-6.24235988e-01 -3.03591222e-01 -1.43013501e+00 1.18979365e-01
8.95867586e-01 -8.05922508e-01 -2.98785478e-01 2.94313967e-01
1.36774942e-01 -4.39997673e-01 -1.00825596e+00 3.05786252e-01
8.29289496e-01 -1.27256477e+00 9.25169051e-01 -9.39763904e-01
5.78210056e-01 7.61699155e-02 -3.36638093e-01 -1.08815384e+00
-3.56026769e-01 -1.99779436e-01 6.48257369e-03 7.79294789e-01
-5.58971614e-02 -3.15542877e-01 9.33169544e-01 1.89603820e-01
2.31607687e-02 -1.09273028e+00 -1.06991053e+00 -4.12162662e-01
-2.78882198e-02 -5.78123890e-02 1.02808762e+00 9.41091120e-01
-1.14854895e-01 -1.29140213e-01 -1.22770198e-01 -5.39488904e-02
4.22193557e-01 3.72423202e-01 8.37618887e-01 -1.33403528e+00
3.11068892e-01 -5.23514330e-01 -1.38776183e+00 -7.73217440e-01
8.16330433e-01 -1.02115107e+00 -2.72079617e-01 -9.42048252e-01
7.70902708e-02 -3.35179150e-01 -4.14770454e-01 7.94377923e-01
4.51112896e-01 6.59741759e-01 -1.58106923e-01 -1.74303070e-01
-3.25529337e-01 8.00270319e-01 1.23020995e+00 -1.09274946e-01
9.25648808e-02 -3.81867737e-01 -4.10720468e-01 7.38393366e-01
5.53938925e-01 2.56593851e-03 5.55798970e-02 -2.22914398e-01
-2.68254399e-01 1.02719881e-01 5.93408227e-01 -7.50245333e-01
-2.18854487e-01 -1.42659172e-01 1.05892563e+00 -4.56575006e-01
7.28377402e-01 -6.91040039e-01 -6.75506815e-02 1.45014867e-01
-1.97624564e-01 -4.85913381e-02 3.14142823e-01 2.87075285e-02
-4.74048018e-01 2.57311732e-01 9.50917482e-01 2.09318325e-01
-6.91419125e-01 6.40338540e-01 -2.34626085e-01 -4.92884904e-01
9.86655533e-01 -5.50788283e-01 -2.95992475e-02 -4.40493047e-01
-8.28318655e-01 -4.13697362e-01 4.20964539e-01 4.76509899e-01
8.82749677e-01 -1.88568664e+00 -6.23437166e-01 9.57407176e-01
1.76216125e-01 -4.26386863e-01 1.50325134e-01 6.57533169e-01
-6.04415476e-01 2.03039303e-01 -1.07168150e+00 -8.25376630e-01
-1.38207412e+00 2.89529830e-01 5.89923441e-01 2.92205840e-01
-2.50223219e-01 9.00029421e-01 1.88677743e-01 -4.51427609e-01
2.32843101e-01 3.72613370e-02 -3.97232145e-01 8.59485753e-03
6.43077016e-01 1.45475939e-02 1.62944540e-01 -1.32935810e+00
-6.17054105e-01 1.10060227e+00 2.54777014e-01 2.59157568e-01
1.63340914e+00 3.18536103e-01 -3.43156040e-01 -1.64720956e-02
1.85333812e+00 -4.29940850e-01 -1.17374623e+00 -5.81623055e-04
-1.73348069e-01 -7.03424633e-01 9.42491740e-02 -4.26605195e-01
-1.37170911e+00 1.21318686e+00 8.80282104e-01 -8.01257566e-02
1.21974397e+00 2.66715121e-02 4.10468221e-01 3.39917928e-01
3.12359691e-01 -9.50951338e-01 2.18150690e-01 5.79141259e-01
1.12391233e+00 -1.06020534e+00 -2.21072584e-01 -2.38629267e-01
-4.25966322e-01 1.61243474e+00 5.37186086e-01 -4.54190701e-01
1.07548809e+00 2.13406369e-01 1.97369143e-01 -4.73963499e-01
-4.26600397e-01 -6.17437251e-02 4.62571144e-01 6.86547995e-01
6.27895236e-01 -1.01978645e-01 2.34435871e-01 4.17853296e-01
-3.70567232e-01 9.83474404e-02 -4.06145602e-02 6.37222946e-01
-4.28539872e-01 -7.50743032e-01 -2.80736417e-01 2.25547314e-01
-2.90135264e-01 5.10405362e-01 -5.62283218e-01 7.16685414e-01
2.81093866e-01 4.54714149e-01 4.68308419e-01 -6.25127792e-01
2.54819036e-01 1.86571062e-01 7.98417211e-01 -2.03161940e-01
-1.30238399e-01 -1.09746791e-02 -2.62157898e-02 -9.11220491e-01
-4.84666437e-01 -5.09040356e-01 -1.15998578e+00 -3.97826493e-01
2.20918790e-01 -2.03564897e-01 8.94707263e-01 6.70074463e-01
4.41022426e-01 -1.46478429e-01 9.64732349e-01 -1.33894861e+00
-4.59404349e-01 -7.40910232e-01 -7.39197254e-01 8.56211185e-01
3.76949430e-01 -1.25663280e+00 -3.62246543e-01 -1.02123395e-02]
|
[13.508399963378906, 1.2746680974960327]
|
8ff9e82e-6356-431b-b154-c21b1d648339
|
low-resource-unsupervised-nmt-diagnosing-the
| null | null |
https://aclanthology.org/2020.eamt-1.10
|
https://aclanthology.org/2020.eamt-1.10.pdf
|
Low-Resource Unsupervised NMT: Diagnosing the Problem and Providing a Linguistically Motivated Solution
|
Unsupervised Machine Translation has been advancing our ability to translate without parallel data, but state-of-the-art methods assume an abundance of monolingual data. This paper investigates the scenario where monolingual data is limited as well, finding that current unsupervised methods suffer in performance under this stricter setting. We find that the performance loss originates from the poor quality of the pretrained monolingual embeddings, and we offer a potential solution: dependency-based word embeddings. These embeddings result in a complementary word representation which offers a boost in performance of around 1.5 BLEU points compared to standard word2vec when monolingual data is limited to 1 million sentences per language. We also find that the inclusion of sub-word information is crucial to improving the quality of the embeddings.
|
['Gertjan van Noord', 'Antonio Toral', 'Lukas Edman']
| null | null | null | null |
eamt-2020-11
|
['unsupervised-machine-translation']
|
['natural-language-processing']
|
[-1.65948018e-01 -1.25550985e-01 -6.43946767e-01 -8.71962085e-02
-1.23456347e+00 -8.89623821e-01 1.04043102e+00 2.48793468e-01
-7.90342689e-01 8.48119140e-01 8.59769762e-01 -8.90524387e-01
3.18861306e-01 -3.35415751e-01 -5.81902385e-01 -4.31463242e-01
2.58166075e-01 6.45893216e-01 -2.44686276e-01 -7.35830307e-01
4.76745181e-02 2.07285240e-01 -7.98115134e-01 -2.37135217e-01
1.03358376e+00 1.90056339e-01 5.15217260e-02 5.22118866e-01
-4.01283026e-01 1.08179692e-02 -3.35178316e-01 -6.75892115e-01
5.55452049e-01 -3.19582164e-01 -7.87158310e-01 -1.33753255e-01
4.26676393e-01 -2.36218035e-01 -5.98888993e-01 9.39417481e-01
6.89385772e-01 -1.25682101e-01 4.89593595e-01 -7.15658247e-01
-1.04709613e+00 6.69983804e-01 -2.97241122e-01 4.41105813e-01
2.42569789e-01 1.12703435e-01 1.42610288e+00 -1.25012875e+00
8.88701618e-01 9.97121215e-01 6.17141426e-01 3.35579693e-01
-1.28181374e+00 -2.64024943e-01 -1.53658092e-01 1.98387071e-01
-1.38661885e+00 -7.63845921e-01 3.52082938e-01 -3.59512061e-01
1.50960541e+00 -1.46713018e-01 4.75760281e-01 1.25417364e+00
6.30146325e-01 3.33065361e-01 1.16888356e+00 -8.11007619e-01
-2.93934137e-01 3.31938595e-01 -3.15022729e-02 3.70811254e-01
4.24698949e-01 7.56874979e-02 -4.73753005e-01 -1.27924249e-01
6.04876637e-01 -2.21180186e-01 -5.27917445e-02 -2.20380679e-01
-1.50480151e+00 1.08360934e+00 1.36488065e-01 6.66122258e-01
-1.72079340e-01 -8.69303271e-02 5.14129877e-01 7.75804639e-01
7.37450361e-01 7.03241646e-01 -5.25289893e-01 -4.72198933e-01
-1.01278174e+00 4.65725316e-03 7.22725451e-01 9.37367618e-01
7.47682154e-01 8.34262520e-02 1.14410572e-01 9.17316437e-01
-8.72824639e-02 6.17045641e-01 7.33381689e-01 -4.49962676e-01
8.16242397e-01 2.45183483e-01 2.75099147e-02 -6.38524115e-01
-5.71607389e-02 -6.66659832e-01 -2.74728775e-01 -4.07172263e-01
4.03590381e-01 -2.63030261e-01 -8.66255641e-01 1.69795597e+00
1.36156296e-02 -4.55414891e-01 2.11455286e-01 7.42816925e-01
1.44145519e-01 7.07147896e-01 -2.10045919e-01 -1.92354515e-01
1.08108485e+00 -1.12088251e+00 -8.66619885e-01 -5.06801665e-01
7.98979700e-01 -1.38018095e+00 1.19104409e+00 -7.29582459e-02
-8.99830163e-01 -3.26126575e-01 -1.22095990e+00 -3.72844428e-01
-5.10492504e-01 -2.19235525e-01 6.21994734e-01 6.62750840e-01
-1.20359337e+00 3.98074478e-01 -6.68902516e-01 -8.54623735e-01
-3.17781083e-02 3.76996636e-01 -7.40957439e-01 -5.63081443e-01
-1.32462335e+00 1.54568815e+00 -4.53856662e-02 -1.60518110e-01
-4.45449889e-01 -5.81494987e-01 -8.33401501e-01 -1.67659029e-01
-5.88626601e-02 -4.24709529e-01 9.47575033e-01 -8.00525665e-01
-1.21212864e+00 7.20660150e-01 -3.03844035e-01 -2.42507249e-01
4.29509401e-01 -2.86347628e-01 -5.01952946e-01 -8.34271163e-02
3.42993349e-01 4.44429845e-01 4.68236208e-01 -9.49823618e-01
-5.25052905e-01 -2.63525397e-01 -9.60106328e-02 4.65666652e-01
-8.52395713e-01 2.95364171e-01 -4.94600445e-01 -5.93382537e-01
-8.66528451e-02 -9.72592473e-01 -3.41173887e-01 -5.46978533e-01
-7.46922893e-03 -2.14357615e-01 3.55522305e-01 -9.73588884e-01
1.25267279e+00 -1.83605683e+00 3.35947961e-01 -1.39398679e-01
-1.13430291e-01 1.84750646e-01 -5.83793402e-01 1.01839387e+00
1.82186335e-01 3.91950577e-01 -4.30750884e-02 -3.07741880e-01
-3.94284725e-03 4.52228874e-01 -7.68151730e-02 5.18320918e-01
3.46009195e-01 1.04554796e+00 -9.90061224e-01 -3.70335191e-01
-7.82719441e-03 5.23959637e-01 -2.85199910e-01 -6.91480190e-02
2.34762579e-01 2.67551899e-01 -7.99795017e-02 5.62470198e-01
4.89715189e-01 2.15220302e-01 6.21227026e-01 1.78378895e-01
-3.32644999e-01 8.57973754e-01 -5.21947920e-01 2.00136375e+00
-7.60015190e-01 9.32949424e-01 -2.29745194e-01 -7.04118788e-01
7.43516207e-01 3.91508967e-01 3.94650757e-01 -8.46738338e-01
1.56866103e-01 6.21449590e-01 5.23921669e-01 -3.21736723e-01
7.08487749e-01 -4.41404492e-01 -1.81133598e-01 6.56611562e-01
2.75021166e-01 -1.02794133e-01 3.39624345e-01 1.76242188e-01
1.06718302e+00 9.87989083e-02 2.23166510e-01 -6.80762947e-01
-3.10138497e-03 4.34521794e-01 5.99352419e-01 4.27524596e-01
-1.54165685e-01 4.81840849e-01 1.67712405e-01 -3.30908000e-01
-1.75162077e+00 -1.20260203e+00 -2.37937540e-01 1.12669456e+00
-2.95824677e-01 -5.36246896e-01 -4.80768412e-01 -7.43554473e-01
-3.44809294e-02 7.03731120e-01 -3.79248649e-01 2.90564690e-02
-7.80092478e-01 -1.06798756e+00 6.30462468e-01 4.94354308e-01
-2.07872704e-01 -5.09423435e-01 6.55601472e-02 5.14802992e-01
-2.88265169e-01 -1.19465625e+00 -6.99027002e-01 3.58685732e-01
-1.03629637e+00 -5.93752444e-01 -6.13375485e-01 -9.60128069e-01
6.16515219e-01 2.74726123e-01 1.30205417e+00 -2.31160238e-01
8.35536495e-02 3.01695634e-02 -5.95443428e-01 -1.01385020e-01
-4.64739889e-01 5.26229739e-01 5.57034671e-01 -4.97215778e-01
8.67216051e-01 -6.27159834e-01 -3.08987409e-01 -7.97211453e-02
-8.51725042e-01 -3.91141474e-01 8.09553981e-01 1.09285057e+00
2.10627869e-01 -4.64057863e-01 6.19253993e-01 -6.43619239e-01
9.61706519e-01 -5.94243705e-01 -1.37501016e-01 1.64967328e-01
-1.09289515e+00 1.96319357e-01 7.02454209e-01 -5.11746883e-01
-5.79735518e-01 -5.12005150e-01 -1.96787953e-01 -1.39963822e-02
2.36613363e-01 7.76266813e-01 2.24819899e-01 9.22438949e-02
6.06567800e-01 2.69109756e-01 5.39478846e-02 -6.10968709e-01
6.37496650e-01 9.20885623e-01 1.48305431e-01 -5.97906590e-01
9.18501318e-01 4.96298559e-02 -3.95101100e-01 -7.53551483e-01
-4.20446098e-01 -5.98971248e-01 -8.14464211e-01 2.58000135e-01
6.33136213e-01 -1.14530241e+00 3.32588047e-01 -2.40457028e-01
-1.19767964e+00 -1.87885612e-01 -1.91038504e-01 8.63781631e-01
-3.88823509e-01 3.69116545e-01 -9.04216528e-01 -3.52203578e-01
-3.52124423e-01 -1.36074162e+00 6.44564211e-01 -3.16800445e-01
-4.46635485e-01 -1.29734445e+00 5.37794471e-01 3.25855821e-01
6.98666692e-01 -1.81757644e-01 1.19605577e+00 -8.04952860e-01
-2.33073547e-01 -3.20293158e-01 -1.02991037e-01 4.88435030e-01
4.71379042e-01 -2.71088332e-01 -5.53102851e-01 -5.55743098e-01
-1.93678111e-01 -2.35534623e-01 5.99572539e-01 -1.17370307e-01
-9.33450386e-02 -2.27877885e-01 4.20085825e-02 3.63652647e-01
1.60009003e+00 -1.53328434e-01 4.37236458e-01 4.55937833e-01
7.66318321e-01 4.91064787e-01 3.37869406e-01 -1.02231167e-01
5.09835005e-01 6.15137637e-01 -1.16645128e-01 -1.92650795e-01
-2.98145264e-01 -3.85203153e-01 6.86463952e-01 1.97619295e+00
1.19395377e-02 -1.72657222e-02 -1.19408190e+00 1.10865736e+00
-1.57491517e+00 -6.51393890e-01 -9.26736817e-02 2.24332619e+00
1.18164027e+00 1.14693694e-01 -1.09699056e-01 -2.14166954e-01
3.97238791e-01 2.36924648e-01 -1.90795243e-01 -9.76813018e-01
-2.70208269e-01 5.85839868e-01 9.84823287e-01 8.38225126e-01
-6.30125284e-01 1.40075207e+00 7.41818285e+00 6.93444967e-01
-1.16662192e+00 5.83234608e-01 7.94943422e-02 -1.20701559e-01
-6.38144374e-01 2.15039656e-01 -6.75996542e-01 2.98680127e-01
1.34057391e+00 -4.44142699e-01 5.46286404e-01 3.24791491e-01
2.18390211e-01 2.37925991e-01 -1.11576319e+00 6.03280962e-01
3.38057846e-01 -1.14214492e+00 2.08605766e-01 4.41668838e-01
1.04609215e+00 6.55385375e-01 1.14725187e-01 3.71351033e-01
4.07838494e-01 -1.09770894e+00 3.95593971e-01 4.92559886e-03
1.10753977e+00 -9.18972313e-01 9.53199804e-01 3.11291963e-01
-8.38898718e-01 2.10020497e-01 -6.04175866e-01 -2.30577230e-01
1.91314608e-01 4.49774265e-01 -8.65637898e-01 7.45629966e-01
2.03674361e-01 7.00572431e-01 -5.29773653e-01 6.89106584e-01
-2.38703862e-01 6.82052791e-01 -2.71564156e-01 4.99323085e-02
6.83330536e-01 -4.22703296e-01 4.12262142e-01 1.46224916e+00
4.51720983e-01 -5.89895844e-01 1.50184795e-01 1.20339438e-01
-3.04634511e-01 5.82591176e-01 -8.97100210e-01 -4.87201512e-01
3.94993633e-01 9.12407339e-01 -3.00265312e-01 -1.56771183e-01
-8.32708120e-01 1.22826290e+00 6.85559750e-01 4.10972506e-01
-3.86598200e-01 -3.60071033e-01 9.97593641e-01 3.95266563e-02
2.49894202e-01 -8.60841632e-01 -3.20082456e-01 -1.43447781e+00
1.41899332e-01 -1.18502951e+00 -1.40524611e-01 -1.60453379e-01
-1.46368682e+00 6.82953656e-01 -4.22103971e-01 -1.04926813e+00
-3.62427264e-01 -6.53304040e-01 -3.05947095e-01 1.21026969e+00
-1.54268014e+00 -1.27002203e+00 6.16565347e-01 1.10829264e-01
8.11689258e-01 -1.98394507e-01 1.09941483e+00 5.18214524e-01
-4.03759956e-01 9.19431508e-01 7.77906716e-01 1.79119751e-01
1.18866611e+00 -1.15948272e+00 8.15567136e-01 9.75371122e-01
5.27884603e-01 1.11201000e+00 8.13420594e-01 -6.07979834e-01
-1.83016002e+00 -8.53609741e-01 1.87743545e+00 -9.69449222e-01
1.15833938e+00 -5.76033294e-01 -5.11475801e-01 8.01298141e-01
7.35839427e-01 -1.99938297e-01 9.73994076e-01 5.83365858e-01
-6.79169714e-01 1.23184994e-01 -7.37531722e-01 9.05546844e-01
9.39419627e-01 -9.93591487e-01 -8.90019476e-01 3.89143258e-01
8.63555491e-01 6.56623989e-02 -1.06893349e+00 1.51329860e-01
6.98245227e-01 -3.46916914e-01 6.47990584e-01 -9.73591685e-01
5.71522176e-01 7.21867606e-02 -4.73896831e-01 -1.78037822e+00
-3.50641608e-01 -5.55081964e-01 2.95112520e-01 1.08904827e+00
8.89681935e-01 -6.88887835e-01 3.83707941e-01 5.29086329e-02
-1.17010102e-01 -9.01410520e-01 -1.02037632e+00 -1.12810564e+00
9.78114963e-01 -3.20970058e-01 4.19934332e-01 1.18121457e+00
3.50562692e-01 7.79424369e-01 -4.13203686e-01 -2.14979947e-01
4.41236824e-01 -2.13337660e-01 5.73018909e-01 -8.28296661e-01
-9.81555581e-02 -3.49752456e-01 -2.95970351e-01 -1.14199305e+00
1.87061876e-01 -1.19103050e+00 1.67815306e-03 -1.77958667e+00
3.45923454e-01 -3.78446311e-01 -4.00570095e-01 9.48169604e-02
-3.16054106e-01 5.69093883e-01 1.16361655e-01 2.14374006e-01
-1.12493671e-01 4.59701329e-01 1.06878519e+00 -1.01320259e-01
-2.74999775e-02 -7.19864786e-01 -7.82403588e-01 2.56407559e-01
8.21110606e-01 -5.57274401e-01 -1.70332491e-01 -1.22646618e+00
1.68403119e-01 -3.75533670e-01 -2.95936882e-01 -6.14816785e-01
5.77544905e-02 -1.40314072e-01 1.71906844e-01 -1.28888980e-01
2.23747790e-01 -6.67199731e-01 -2.26015598e-01 2.20460013e-01
-2.12538093e-01 7.37315416e-01 1.69322342e-01 3.49579632e-01
-2.94109851e-01 -4.02176417e-02 3.52423310e-01 -6.33193627e-02
-3.84329975e-01 1.49749786e-01 -5.05453348e-01 2.52901733e-01
4.25297499e-01 -2.11569667e-02 -2.27420434e-01 -2.38732249e-01
-3.96476448e-01 -2.66049784e-02 7.43669391e-01 7.76557982e-01
1.62158594e-01 -1.60000134e+00 -9.64540541e-01 1.31873190e-01
2.85354316e-01 -8.05326998e-01 -3.48656386e-01 9.85977173e-01
-4.90346313e-01 6.77404881e-01 -2.38692984e-01 -2.27257803e-01
-8.56535077e-01 3.04894269e-01 -2.50168771e-01 -4.08108175e-01
-3.86217833e-01 5.67455113e-01 -4.01210248e-01 -6.28428578e-01
-1.86181441e-01 2.57876068e-02 2.31663913e-01 2.72272974e-01
1.49669915e-01 1.97775140e-01 2.77447343e-01 -9.75378811e-01
-4.03785884e-01 5.76296926e-01 -4.23662901e-01 -6.15461767e-01
1.36335349e+00 -2.31821150e-01 -9.84683335e-02 5.77347398e-01
1.46657681e+00 5.92813790e-01 -6.53646827e-01 -3.95879984e-01
1.53753385e-01 -4.64529425e-01 6.03640713e-02 -6.45496249e-01
-5.13288081e-01 9.82275903e-01 3.71457338e-01 -2.43235558e-01
6.29133582e-01 -1.28412217e-01 1.14072132e+00 3.73895258e-01
6.48687422e-01 -1.21691144e+00 -3.49397928e-01 9.58209455e-01
5.28490722e-01 -1.29196644e+00 2.66227666e-02 -1.42188137e-02
-4.65132773e-01 1.01384175e+00 -3.53847019e-04 -1.85121611e-01
3.46001595e-01 2.30323106e-01 5.88528395e-01 2.32127771e-01
-6.87492132e-01 -3.96902114e-01 1.53909594e-01 4.59954649e-01
8.02516460e-01 2.46121764e-01 -9.02179718e-01 2.60374427e-01
-3.59292418e-01 -4.78094429e-01 4.27762538e-01 9.02217150e-01
-2.88770527e-01 -1.86755073e+00 -2.22811863e-01 3.62656415e-01
-7.03950703e-01 -7.54803717e-01 -4.34610963e-01 8.77322078e-01
1.03029879e-02 1.15336776e+00 1.28766298e-01 -4.95732069e-01
2.32037798e-01 5.52081943e-01 5.92845321e-01 -8.09549034e-01
-5.96972883e-01 1.69056907e-01 2.82878518e-01 -3.93508226e-01
-7.69501030e-02 -5.14802516e-01 -7.42089808e-01 -6.29427433e-01
-2.38934696e-01 5.02978206e-01 8.96290064e-01 9.21530664e-01
2.99726844e-01 1.19325057e-01 6.46531045e-01 -5.46954572e-01
-8.29150617e-01 -1.21474218e+00 -1.39912248e-01 3.30278814e-01
3.21239501e-01 -3.33578110e-01 -3.69969368e-01 -2.92896647e-02]
|
[11.338127136230469, 10.219405174255371]
|
b98f2980-3f89-4130-a6d6-07fb484a3dda
|
med7-a-transferable-clinical-natural-language
|
2003.01271
| null |
https://arxiv.org/abs/2003.01271v2
|
https://arxiv.org/pdf/2003.01271v2.pdf
|
Med7: a transferable clinical natural language processing model for electronic health records
|
The field of clinical natural language processing has been advanced significantly since the introduction of deep learning models. The self-supervised representation learning and the transfer learning paradigm became the methods of choice in many natural language processing application, in particular in the settings with the dearth of high quality manually annotated data. Electronic health record systems are ubiquitous and the majority of patients' data are now being collected electronically and in particular in the form of free text. Identification of medical concepts and information extraction is a challenging task, yet important ingredient for parsing unstructured data into structured and tabulated format for downstream analytical tasks. In this work we introduced a named-entity recognition model for clinical natural language processing. The model is trained to recognise seven categories: drug names, route, frequency, dosage, strength, form, duration. The model was first self-supervisedly pre-trained by predicting the next word, using a collection of 2 million free-text patients' records from MIMIC-III corpora and then fine-tuned on the named-entity recognition task. The model achieved a lenient (strict) micro-averaged F1 score of 0.957 (0.893) across all seven categories. Additionally, we evaluated the transferability of the developed model using the data from the Intensive Care Unit in the US to secondary care mental health records (CRIS) in the UK. A direct application of the trained NER model to CRIS data resulted in reduced performance of F1=0.762, however after fine-tuning on a small sample from CRIS, the model achieved a reasonable performance of F1=0.944. This demonstrated that despite a close similarity between the data sets and the NER tasks, it is essential to fine-tune on the target domain data in order to achieve more accurate results.
|
['Alejo Nevado-Holgado', 'Nemanja Vaci', 'Qiang Liu', 'Andrey Kormilitzin']
|
2020-03-03
| null | null | null | null |
['medical-named-entity-recognition']
|
['natural-language-processing']
|
[ 3.96183729e-01 3.76532108e-01 2.59130783e-02 -4.85212743e-01
-1.15122378e+00 -6.34450197e-01 3.83706301e-01 9.52066422e-01
-1.09876132e+00 8.50208580e-01 4.80948091e-01 -5.97372830e-01
-3.67401540e-01 -6.65588319e-01 -3.20529073e-01 -2.98494011e-01
-2.00678393e-01 6.20186150e-01 -2.62106091e-01 6.78354353e-02
2.58045457e-02 5.15789449e-01 -1.05107844e+00 7.29864717e-01
8.14987242e-01 6.87540710e-01 1.56832993e-01 7.40356624e-01
-3.29327703e-01 1.00978553e+00 -7.04981863e-01 -4.00332630e-01
7.82496706e-02 -3.76726031e-01 -9.03249621e-01 -2.52825975e-01
-2.09841594e-01 -1.22162290e-01 3.83770070e-03 6.88595235e-01
7.97072232e-01 -1.31392786e-02 6.69652820e-01 -3.84362340e-01
-3.68549973e-01 6.25193179e-01 -6.54473230e-02 2.24801391e-01
4.46342438e-01 3.13257761e-02 8.21251571e-01 -5.68973660e-01
8.11876714e-01 7.43146479e-01 8.23800206e-01 6.80887878e-01
-1.04910624e+00 -7.06879497e-01 -4.67099488e-01 -1.65277094e-01
-1.28861773e+00 -4.99411345e-01 -1.08105298e-02 -6.17062747e-01
1.34594238e+00 6.39077350e-02 1.22161619e-01 9.65361774e-01
4.65715975e-01 3.20599794e-01 1.17737532e+00 -5.91811955e-01
3.04039270e-01 5.27068913e-01 1.86127260e-01 4.46328521e-01
2.54595935e-01 3.45457136e-03 -5.51615506e-02 -4.23066497e-01
5.73741794e-01 8.90846271e-03 -4.53531481e-02 2.46366173e-01
-1.19884729e+00 9.52426255e-01 3.27468604e-01 7.61234403e-01
-5.84858239e-01 -5.83685040e-01 6.98625624e-01 2.39132717e-01
4.36363429e-01 6.87996984e-01 -1.01382935e+00 -3.09430540e-01
-9.14588213e-01 -1.07189305e-01 1.10195613e+00 7.36321390e-01
4.36835364e-02 -3.85104597e-01 -2.91640610e-01 1.00853837e+00
1.36415124e-01 9.48260725e-02 9.29247439e-01 -1.82779372e-01
6.53633833e-01 7.87223876e-01 -4.93741371e-02 -7.00208247e-01
-8.83724809e-01 -3.21035534e-01 -9.35496211e-01 -2.87284732e-01
5.29773653e-01 -5.09773612e-01 -1.11624038e+00 1.55681705e+00
1.00827157e-01 -3.25193524e-01 5.74870288e-01 3.89473885e-01
9.04656112e-01 5.61964869e-01 7.98023045e-01 -3.16394985e-01
1.65624249e+00 -3.97144556e-01 -8.16665828e-01 -9.74633619e-02
1.03738308e+00 -7.35734820e-01 4.13602173e-01 2.85419464e-01
-8.23738933e-01 -2.98918158e-01 -6.78180635e-01 -1.03100548e-02
-7.72143841e-01 8.94373376e-03 4.90520537e-01 6.77776814e-01
-8.86166871e-01 7.14747965e-01 -8.34067941e-01 -5.64694464e-01
6.54153943e-01 6.80907786e-01 -7.59944975e-01 -1.81410074e-01
-1.37221789e+00 1.08530569e+00 6.09238923e-01 -1.38338298e-01
-3.24112386e-01 -8.73176932e-01 -9.14385378e-01 2.05443144e-01
-9.15330946e-02 -6.52373135e-01 1.30752003e+00 -7.09473252e-01
-1.10213411e+00 1.09085810e+00 1.45386323e-01 -5.30455410e-01
4.93067145e-01 3.95514295e-02 -8.40368092e-01 8.05651993e-02
2.44165644e-01 3.34077835e-01 2.43998945e-01 -6.03522420e-01
-8.36074233e-01 -4.55383718e-01 -4.61494714e-01 -1.22675888e-01
-7.46729225e-02 3.53308022e-01 1.53905051e-02 -5.52477717e-01
-3.70148271e-01 -6.02518022e-01 -4.69116777e-01 -5.78475773e-01
-2.20644757e-01 -2.58248061e-01 -1.42148390e-01 -8.64728868e-01
1.42743492e+00 -2.07959056e+00 -3.40318054e-01 1.44748241e-01
2.29509473e-01 4.89390105e-01 -4.42501940e-02 6.37319326e-01
-6.44547343e-01 2.55124360e-01 -5.24878979e-01 5.09365462e-02
-2.14237049e-01 4.62968759e-02 1.57795534e-01 4.24168169e-01
6.21324122e-01 8.22124362e-01 -1.07231450e+00 -4.61500078e-01
2.83507667e-02 6.83440030e-01 -4.71934408e-01 5.05516827e-01
2.53917456e-01 4.05522048e-01 -5.32228708e-01 3.23592514e-01
3.35731626e-01 -2.58811444e-01 2.62530714e-01 1.42616662e-03
-1.56060115e-01 5.09799302e-01 -9.21926677e-01 1.52842462e+00
-4.46900785e-01 1.90729350e-01 -7.70128518e-02 -9.44199145e-01
7.97373354e-01 6.63728356e-01 9.31172907e-01 -7.46016085e-01
3.33215386e-01 2.49539167e-01 2.89834946e-01 -7.50192165e-01
1.23297572e-01 -6.39157593e-01 -2.94729978e-01 3.21088403e-01
4.79456410e-02 3.31923068e-01 1.23279989e-01 -2.49908436e-02
1.54652596e+00 -2.89146572e-01 9.20180321e-01 -2.49114454e-01
5.79075634e-01 2.54022002e-01 5.40527403e-01 4.63712782e-01
-1.65845811e-01 5.45251667e-01 3.17797542e-01 -3.39333296e-01
-8.45060050e-01 -7.52197683e-01 -7.27023482e-01 7.14894414e-01
-8.26787770e-01 -4.76793259e-01 -8.00282300e-01 -7.19487906e-01
-1.79577902e-01 8.07566285e-01 -5.84433675e-01 -1.21517077e-01
-4.55262154e-01 -8.70172083e-01 7.55319595e-01 3.91927809e-01
1.77989081e-01 -1.56098330e+00 -7.52553940e-01 7.06977308e-01
9.03877988e-02 -1.13073957e+00 -3.23811114e-01 6.15276933e-01
-7.24053502e-01 -1.21128631e+00 -7.79291391e-01 -8.54239166e-01
4.45824981e-01 -6.31977797e-01 1.11403775e+00 -2.49967992e-01
-5.05699992e-01 3.60543311e-01 -4.06915575e-01 -6.93722129e-01
-8.28429818e-01 2.28430212e-01 -3.67184170e-02 -1.37258515e-01
9.21449721e-01 -7.75219202e-02 -5.02704799e-01 -7.63083175e-02
-1.16281080e+00 -3.29388082e-01 8.06342244e-01 9.76672351e-01
3.91741782e-01 -8.35236069e-03 8.26005161e-01 -1.42308211e+00
7.89827824e-01 -7.05926836e-01 -2.04654858e-01 1.57494232e-01
-6.18973970e-01 9.47640985e-02 7.70155489e-01 -2.93615550e-01
-8.57836664e-01 2.94998944e-01 -5.64507425e-01 2.61510283e-01
-6.92712605e-01 8.27293158e-01 -9.06330813e-03 5.48135281e-01
8.33933294e-01 -3.72274555e-02 -1.00058094e-01 -6.11505926e-01
-3.45777199e-02 1.31288946e+00 3.30531836e-01 -1.54567108e-01
2.24161610e-01 -1.16216354e-01 -2.76086777e-01 -9.27477360e-01
-7.21859396e-01 -7.96370506e-01 -7.05445886e-01 5.51621735e-01
1.12140429e+00 -8.98086607e-01 -6.71633840e-01 2.31324613e-01
-1.00955009e+00 -3.02367002e-01 -3.44300985e-01 6.09222353e-01
-3.63043338e-01 1.37948290e-01 -7.51839042e-01 -7.09410608e-01
-7.68706441e-01 -8.81408155e-01 8.36219668e-01 5.36818691e-02
-5.84055483e-01 -1.13823676e+00 4.64328915e-01 2.12215275e-01
4.02127057e-01 3.46651226e-01 1.17681205e+00 -1.47263491e+00
3.44414681e-01 -4.54916298e-01 -1.85473487e-01 1.80392221e-01
6.36353076e-01 -5.31626284e-01 -9.22693849e-01 -1.13772310e-01
2.16086447e-01 -1.89586103e-01 6.39001846e-01 2.83149600e-01
8.27628970e-01 -2.08057418e-01 -3.08704674e-01 2.78935671e-01
1.39172912e+00 6.49329185e-01 6.98512435e-01 3.80392402e-01
4.76701796e-01 6.95582271e-01 2.54221410e-01 4.08120185e-01
3.31722856e-01 2.90960252e-01 -3.14232737e-01 -2.34181061e-01
2.35119149e-01 -9.67915058e-02 1.30767986e-01 6.58292353e-01
2.25649104e-01 -1.13780620e-02 -1.32832170e+00 5.70557296e-01
-1.48196292e+00 -7.67303646e-01 7.79679883e-03 2.19721746e+00
1.20964503e+00 1.12711765e-01 7.66538605e-02 -8.04660469e-03
6.66134953e-01 -6.22651517e-01 -4.61396456e-01 -8.13708007e-01
3.02770853e-01 7.80855477e-01 6.53674304e-01 1.55644149e-01
-1.13676941e+00 7.17191219e-01 6.05203342e+00 4.38427746e-01
-8.93355548e-01 -1.23609401e-01 8.55381131e-01 2.85305381e-01
2.80871034e-01 -6.07139289e-01 -7.66147017e-01 4.84614015e-01
1.84347546e+00 -8.80104154e-02 1.64324418e-01 5.21722019e-01
4.64780748e-01 4.99157496e-02 -1.30265152e+00 1.07847977e+00
-6.17939234e-02 -1.09954321e+00 -8.00669268e-02 1.37659147e-01
4.08495009e-01 1.75003201e-01 -4.05724198e-01 5.45298755e-01
2.22928941e-01 -1.42872643e+00 4.23167683e-02 4.34277534e-01
1.02449334e+00 -7.50929415e-01 1.08813512e+00 4.67175275e-01
-8.49201500e-01 -2.24865243e-01 -2.38248438e-01 1.26775950e-01
7.27169812e-02 5.55391610e-01 -1.35158873e+00 5.72135091e-01
6.60057425e-01 6.75455153e-01 -3.50648344e-01 9.99539316e-01
1.18564837e-01 5.90099156e-01 -1.95622087e-01 8.13105106e-02
2.66816407e-01 -4.26066369e-02 -4.48609367e-02 1.66431451e+00
1.72620654e-01 5.03820598e-01 -5.43565117e-02 4.59649205e-01
-1.88681170e-01 5.37609398e-01 -5.32199085e-01 -5.67457318e-01
8.93836617e-02 1.32053089e+00 -6.99018180e-01 -3.11103612e-01
-4.61854249e-01 7.65329480e-01 3.37356359e-01 1.00346699e-01
-4.77857441e-01 -6.54219210e-01 4.70607698e-01 2.28715509e-01
3.23461503e-01 2.19383135e-01 -2.18835995e-01 -9.19003129e-01
-3.86278808e-01 -1.14421988e+00 8.28557134e-01 -2.83344030e-01
-1.69459283e+00 9.92336869e-01 -2.16648296e-01 -1.02521181e+00
-5.14099956e-01 -9.24582362e-01 -1.83936775e-01 1.17046452e+00
-1.36660004e+00 -7.78963566e-01 2.25432798e-01 5.83452880e-01
4.12553698e-01 -3.15945685e-01 1.46922171e+00 6.59378767e-01
-4.87316906e-01 7.18693912e-01 2.10561186e-01 6.68129027e-01
9.70037341e-01 -1.25473571e+00 2.28757039e-01 2.37950817e-01
-1.13655835e-01 8.58964026e-01 3.19651604e-01 -6.17411375e-01
-9.33156967e-01 -1.21861994e+00 1.45822072e+00 -6.05323732e-01
6.66845679e-01 -1.99814394e-01 -1.09468448e+00 4.23115253e-01
1.14898466e-01 -2.78979689e-01 1.37518322e+00 -1.11100994e-01
-1.39749959e-01 1.33828789e-01 -1.62574959e+00 1.57205164e-01
5.38980603e-01 -5.06860554e-01 -9.51125264e-01 4.10747081e-01
4.77803588e-01 -2.76808679e-01 -1.38577867e+00 1.83997169e-01
3.65528047e-01 -4.68965709e-01 6.28168821e-01 -1.18535161e+00
5.12934923e-01 7.67111108e-02 -2.49532796e-03 -1.28043532e+00
-3.38841438e-01 -5.34175992e-01 3.62488657e-01 1.16794801e+00
8.84013414e-01 -6.42460585e-01 5.28802812e-01 1.02843666e+00
-3.85654718e-02 -6.52632654e-01 -8.17447782e-01 -2.46710926e-01
2.73842633e-01 -3.57663453e-01 3.20627570e-01 1.10870194e+00
4.52560097e-01 6.45203233e-01 1.62910149e-01 -1.03721149e-01
1.09812655e-01 -2.78510571e-01 1.88935906e-01 -1.31857026e+00
-2.10733041e-01 -2.17009306e-01 -5.52783132e-01 -2.42647558e-01
-3.15829776e-02 -1.16023123e+00 4.06653024e-02 -1.81116641e+00
3.21865648e-01 -4.45271403e-01 -5.74478149e-01 7.26225257e-01
-1.48914382e-01 -9.70006958e-02 -9.78500322e-02 2.15607509e-02
-2.30528340e-01 -1.32558599e-01 7.57421732e-01 -9.93694440e-02
-6.36557460e-01 1.63632885e-01 -9.52132821e-01 3.93635720e-01
8.55490565e-01 -8.45288336e-01 -7.77120739e-02 -1.62642628e-01
1.39928833e-01 1.05007552e-01 -2.75946110e-01 -5.69661140e-01
2.33650371e-01 1.85315311e-01 6.35357738e-01 -2.66801298e-01
-2.68627048e-01 -1.06426013e+00 1.28760457e-01 6.18775547e-01
-5.69605470e-01 1.93958327e-01 5.56026578e-01 4.53060597e-01
-8.21440741e-02 -2.52503872e-01 7.05037296e-01 -2.99570471e-01
-4.81348515e-01 9.56135988e-02 -6.44961655e-01 1.67936906e-01
8.44761610e-01 -7.29479492e-02 2.89145093e-02 1.15711704e-01
-1.02461934e+00 3.00309490e-02 -5.04509471e-02 3.64382625e-01
3.59887391e-01 -8.72485936e-01 -8.44396174e-01 2.77468592e-01
3.03834170e-01 -2.89983805e-02 1.62098199e-01 5.91662705e-01
-6.28051519e-01 8.44351649e-01 -1.83618516e-01 -3.35421324e-01
-1.01226914e+00 7.60284245e-01 2.05139101e-01 -8.20496261e-01
-7.17587948e-01 2.56921679e-01 2.29480416e-02 -5.82476377e-01
1.17882565e-01 -5.17184019e-01 -6.38804436e-01 2.17832878e-01
8.14942777e-01 9.06544477e-02 5.00151932e-01 -7.02094376e-01
-6.07926130e-01 3.21301073e-01 -3.58141422e-01 3.73051129e-02
1.77417552e+00 3.14912081e-01 6.28219247e-02 2.34688744e-01
1.44507205e+00 -8.97637159e-02 -5.11674464e-01 5.11582755e-02
5.21575391e-01 2.06993893e-01 1.60009917e-02 -1.18246388e+00
-6.93953276e-01 5.87728679e-01 7.43362486e-01 9.53756794e-02
1.06955051e+00 -7.00739399e-02 5.44444442e-01 4.69658464e-01
1.15294270e-01 -8.27728629e-01 -6.50130928e-01 5.92649996e-01
4.26620573e-01 -1.18590474e+00 -1.64495856e-01 -1.65287852e-01
-7.13237762e-01 1.03667140e+00 2.52078827e-02 1.10442661e-01
8.12813163e-01 3.59040797e-01 2.37751409e-01 -3.29718620e-01
-4.74985898e-01 3.87826152e-02 3.06425095e-01 6.12256944e-01
1.03626847e+00 7.82755241e-02 -4.23234135e-01 1.10122859e+00
-1.46444991e-01 3.46885622e-01 5.01600206e-01 1.01306534e+00
4.96291481e-02 -1.24269879e+00 -1.99765339e-01 8.15190673e-01
-1.09515727e+00 -3.33340615e-01 -3.65118831e-01 8.05676699e-01
2.65269540e-02 1.03981209e+00 4.32152897e-02 2.79848874e-02
7.82478452e-01 4.30304080e-01 9.13237259e-02 -1.09708285e+00
-1.21786749e+00 -7.56169558e-02 1.98977143e-01 -3.52586180e-01
-4.24908936e-01 -5.42710841e-01 -1.44345021e+00 7.19069466e-02
-1.13571145e-01 4.11957264e-01 6.60569370e-01 9.55512345e-01
5.28889477e-01 6.63953960e-01 3.45378309e-01 -2.16808319e-01
-5.61085463e-01 -1.07208979e+00 -5.66273868e-01 5.51170766e-01
3.46490949e-01 -1.64736554e-01 -2.15234891e-01 2.86385119e-01]
|
[8.428345680236816, 8.734039306640625]
|
3fb2759a-a398-4adf-a6bb-4e8a16b092cf
|
unsupervised-language-agnostic-wer
|
2303.05046
| null |
https://arxiv.org/abs/2303.05046v1
|
https://arxiv.org/pdf/2303.05046v1.pdf
|
Unsupervised Language agnostic WER Standardization
|
Word error rate (WER) is a standard metric for the evaluation of Automated Speech Recognition (ASR) systems. However, WER fails to provide a fair evaluation of human perceived quality in presence of spelling variations, abbreviations, or compound words arising out of agglutination. Multiple spelling variations might be acceptable based on locale/geography, alternative abbreviations, borrowed words, and transliteration of code-mixed words from a foreign language to the target language script. Similarly, in case of agglutination, often times the agglutinated, as well as the split forms, are acceptable. Previous work handled this problem by using manually identified normalization pairs and applying them to both the transcription and the hypothesis before computing WER. In this paper, we propose an automatic WER normalization system consisting of two modules: spelling normalization and segmentation normalization. The proposed system is unsupervised and language agnostic, and therefore scalable. Experiments with ASR on 35K utterances across four languages yielded an average WER reduction of 13.28%. Human judgements of these automatically identified normalization pairs show that our WER-normalized evaluation is highly consistent with the perceived quality of ASR output.
|
['Rupeshkumar Mehta', 'Manish Gupta', 'Ankur Gupta', 'Rahul Ambavat', 'Satarupa Guha']
|
2023-03-09
| null | null | null | null |
['transliteration']
|
['natural-language-processing']
|
[ 3.19087356e-01 -2.67877817e-01 4.03009415e-01 -5.18682301e-01
-8.15384865e-01 -8.41221333e-01 4.60376799e-01 4.38455909e-01
-7.33232796e-01 7.20519125e-01 2.68568635e-01 -6.74545944e-01
9.56905335e-02 -3.86963814e-01 -3.34590942e-01 -4.68814760e-01
5.14694750e-01 4.92481977e-01 2.48546094e-01 -4.37191129e-01
2.98588276e-01 6.26207411e-01 -1.31745636e+00 2.63813764e-01
1.14989424e+00 5.17334700e-01 1.92228958e-01 9.53326762e-01
-4.53870028e-01 4.93562698e-01 -1.43905163e+00 -5.12512982e-01
1.61811665e-01 -5.96378684e-01 -8.86059165e-01 2.78475463e-01
2.07687870e-01 6.03943318e-02 4.55636866e-02 1.22120547e+00
3.19466501e-01 2.73464650e-01 5.36629140e-01 -7.51667798e-01
-6.33521736e-01 8.69566262e-01 -2.96985563e-02 3.57367873e-01
5.97630858e-01 1.21115714e-01 8.33370566e-01 -9.44157243e-01
3.29282850e-01 1.20778549e+00 2.30998933e-01 6.48712575e-01
-8.47892523e-01 -4.72443253e-01 -1.11730294e-02 -2.56989211e-01
-1.60551155e+00 -6.80950940e-01 3.42310667e-01 -3.36649597e-01
1.18918097e+00 6.01457715e-01 2.14793727e-01 8.20760608e-01
-2.02082142e-01 3.86728317e-01 1.17520988e+00 -9.21362460e-01
3.54460061e-01 3.60463947e-01 3.21770728e-01 4.26274478e-01
2.78551102e-01 -2.78615415e-01 -3.99162173e-01 3.33827212e-02
2.51894921e-01 -5.77080846e-01 -4.11941320e-01 3.93241256e-01
-1.25186920e+00 4.79277223e-01 -2.35294849e-01 6.51910365e-01
-1.63648218e-01 -5.20533442e-01 4.26633000e-01 3.55224997e-01
1.93958402e-01 6.60243750e-01 -4.80392128e-01 -4.49433357e-01
-1.08120227e+00 8.44188686e-03 8.41375113e-01 1.11591113e+00
4.03114438e-01 4.49299186e-01 -2.36284956e-01 1.22971761e+00
3.01944792e-01 8.67232382e-01 9.57672596e-01 -4.91857469e-01
5.62556744e-01 6.33288026e-01 1.39945954e-01 -6.37762189e-01
-4.36098911e-02 -2.98289865e-01 -3.73731256e-01 2.89973114e-02
6.73121810e-01 -7.70940930e-02 -1.25966644e+00 1.53590965e+00
-5.50791398e-02 -6.54444098e-01 5.13230920e-01 8.22418869e-01
8.10364485e-01 8.47561359e-01 1.25119969e-01 -4.49666977e-01
1.32955623e+00 -9.78385270e-01 -1.15175021e+00 -3.38793814e-01
6.52751684e-01 -1.29210591e+00 1.44306993e+00 4.01752710e-01
-1.14536285e+00 -4.54280645e-01 -1.11814237e+00 1.96163490e-01
-6.66975975e-01 1.62735388e-01 -1.68108508e-01 1.13266945e+00
-1.00488234e+00 2.76140541e-01 -4.32766557e-01 -5.87572873e-01
-4.02280837e-01 1.67529687e-01 -2.54880905e-01 7.67294392e-02
-1.19971168e+00 1.00256550e+00 4.68419552e-01 1.68065026e-01
-3.87718916e-01 -5.24280109e-02 -8.21879566e-01 -1.10769056e-01
1.63517654e-01 1.00950427e-01 1.38519001e+00 -1.29363835e+00
-1.70667410e+00 9.86931026e-01 -3.29216450e-01 -2.43591934e-01
3.14642400e-01 -2.34264612e-01 -1.10854816e+00 -3.12037826e-01
-1.72957718e-01 1.19079821e-01 5.87049663e-01 -9.20972764e-01
-5.28601348e-01 -2.80992538e-01 -3.74711424e-01 3.14230084e-01
-8.63624215e-02 6.65042281e-01 -1.86319917e-01 -7.46153593e-01
1.16449863e-01 -7.18368411e-01 1.11133657e-01 -7.99623430e-01
-1.97036043e-01 -2.12670282e-01 3.95216167e-01 -1.07561707e+00
1.69934678e+00 -2.32884336e+00 -2.48118073e-01 4.75869358e-01
-2.86793649e-01 8.24596286e-01 -1.31643787e-01 2.55667746e-01
-1.40483737e-01 3.08901727e-01 -3.86980981e-01 2.64171120e-02
8.49203244e-02 4.55474138e-01 -1.68392316e-01 2.04073429e-01
3.17733049e-01 5.45901179e-01 -8.73839438e-01 -3.73841375e-01
1.97421327e-01 3.19774091e-01 -2.99286805e-02 4.32123393e-01
1.04654230e-01 8.33140165e-02 1.19266964e-01 6.64834142e-01
4.53512520e-01 4.36869889e-01 6.53029904e-02 1.15176082e-01
-3.42052639e-01 8.02338302e-01 -1.41989684e+00 1.19826114e+00
-5.76995075e-01 4.82220918e-01 -8.06411505e-02 -6.44693017e-01
1.22881329e+00 4.94075954e-01 -3.81114781e-01 -6.41751707e-01
2.79756397e-01 7.74632514e-01 4.40020740e-01 -2.71087021e-01
8.03066909e-01 1.22750893e-01 -9.31841433e-02 1.38936758e-01
1.17318951e-01 -3.56228769e-01 3.70907903e-01 -1.16536736e-01
1.01743388e+00 -1.64444387e-01 7.22863913e-01 -3.02070946e-01
8.92239571e-01 -2.23499492e-01 4.79814887e-01 6.48176849e-01
-4.11980987e-01 8.02268028e-01 2.61890858e-01 3.01880557e-02
-1.06852424e+00 -1.19220746e+00 -1.13983944e-01 1.03470135e+00
-1.49698332e-01 -3.90458494e-01 -1.17189300e+00 -4.01527435e-01
-5.38507104e-01 1.14302564e+00 -2.15633512e-02 8.13396946e-02
-7.51393378e-01 -4.03318912e-01 1.01478696e+00 2.78208643e-01
1.73090085e-01 -1.28035128e+00 -2.52936006e-01 4.01860982e-01
-1.79885879e-01 -1.25956106e+00 -5.27564466e-01 2.51748800e-01
-3.81410778e-01 -7.55600810e-01 -6.59065008e-01 -7.32202709e-01
5.50358951e-01 -3.69933099e-02 8.94128621e-01 2.83028305e-01
1.12032823e-01 1.94736496e-01 -7.10459709e-01 -3.91473979e-01
-1.26342642e+00 -1.12016210e-02 2.39047229e-01 4.62387428e-02
6.16439283e-01 6.29651546e-02 -2.50222199e-02 4.45269048e-01
-1.08660495e+00 -4.49839979e-01 4.05670941e-01 5.06330550e-01
3.03831071e-01 -3.14276576e-01 3.86093974e-01 -6.89459741e-01
8.63630474e-01 -1.29615828e-01 -4.68186796e-01 5.57253718e-01
-4.39017594e-01 2.37888217e-01 9.22373295e-01 -3.89295906e-01
-1.03270042e+00 -9.53780636e-02 -5.54007173e-01 1.01406701e-01
-7.24518239e-01 2.43567020e-01 -6.15055859e-01 2.60467231e-01
8.70376527e-01 1.48949906e-01 -1.97577775e-01 -4.29311961e-01
9.52481031e-02 1.38043678e+00 6.05583668e-01 -5.10722280e-01
7.52066374e-01 -4.37368453e-01 -7.15426922e-01 -1.26668620e+00
-4.58374411e-01 -6.98175073e-01 -5.38988292e-01 -1.64744124e-01
8.69762838e-01 -6.02338970e-01 -5.42038903e-02 7.01919496e-01
-1.46824670e+00 -1.47648722e-01 -2.89048553e-01 5.53571582e-01
1.74022019e-02 6.31336808e-01 -4.71882671e-01 -9.32678640e-01
-2.76080787e-01 -1.44375443e+00 8.37191463e-01 2.52846956e-01
-6.86432898e-01 -7.37784386e-01 -1.34936571e-01 3.97542417e-01
4.89448816e-01 -2.53285021e-01 7.91400135e-01 -1.23721123e+00
1.53123319e-01 -1.41837165e-01 3.08591034e-02 7.81095862e-01
4.22660261e-01 3.69032621e-01 -9.33744013e-01 -7.74060115e-02
-1.15655065e-01 -2.39996314e-02 2.97623873e-01 -1.14645824e-01
6.24334157e-01 -4.97312725e-01 3.94290656e-01 8.94165263e-02
1.00993061e+00 5.73300481e-01 7.40115702e-01 2.65947700e-01
5.70698500e-01 6.33317530e-01 5.30639470e-01 1.99247271e-01
-9.91257653e-02 5.95210016e-01 -9.78207439e-02 5.36265522e-02
-1.68926060e-01 9.31656808e-02 8.17358792e-01 1.53439212e+00
1.30290017e-01 -6.84526563e-01 -1.27488542e+00 5.32457650e-01
-1.17289138e+00 -6.05628192e-01 -4.29143161e-01 2.42608380e+00
1.00780213e+00 2.56943882e-01 -3.95850502e-02 5.01985013e-01
1.04481971e+00 8.06085616e-02 1.18827559e-01 -1.30221021e+00
-3.85910809e-01 5.07149756e-01 5.67985654e-01 9.62995052e-01
-7.28578508e-01 1.19223976e+00 6.00192404e+00 9.03549790e-01
-1.01778185e+00 -2.36145183e-02 3.20619285e-01 3.24583143e-01
-3.76407236e-01 -1.45320937e-01 -8.77911270e-01 4.08920437e-01
1.28206086e+00 -2.04209030e-01 6.28023207e-01 5.52618682e-01
1.23455994e-01 -4.72114328e-03 -9.17446196e-01 9.57941294e-01
3.70090544e-01 -5.57752788e-01 2.92889684e-01 -4.08084661e-01
4.89008874e-01 -1.26253292e-01 -2.70002216e-01 2.18443751e-01
2.21376270e-01 -1.04023957e+00 1.02373421e+00 1.21988289e-01
6.87363148e-01 -7.35576332e-01 1.02064323e+00 1.26274079e-01
-1.02077878e+00 3.35562348e-01 -2.00107560e-01 -1.34727329e-01
-4.90367189e-02 3.82276237e-01 -1.00942457e+00 3.30978006e-01
3.89843285e-01 -1.75878972e-01 -8.32248390e-01 7.65778720e-01
-4.02743012e-01 9.58108902e-01 -3.39925438e-01 -4.14142489e-01
2.95249134e-01 -1.10132851e-01 8.39935362e-01 1.81711376e+00
3.26148123e-01 3.97988856e-02 -5.51306084e-02 4.87403214e-01
1.85431868e-01 8.00803602e-01 -3.60765040e-01 -2.19254196e-01
6.67072296e-01 1.01846433e+00 -8.73182893e-01 -3.48723441e-01
-3.00273299e-01 1.07193649e+00 1.07293442e-01 4.42572117e-01
-6.58198595e-01 -7.47984946e-01 6.87513053e-01 -5.56511581e-02
-2.05985866e-02 -3.48633111e-01 -3.01647037e-01 -9.59657311e-01
2.49256551e-01 -1.34714353e+00 1.49240375e-01 -5.78335166e-01
-1.01688349e+00 1.10401320e+00 -2.88194060e-01 -1.09382296e+00
-3.52037512e-02 -7.83206344e-01 -3.69320959e-01 1.11783183e+00
-1.20120466e+00 -4.96392757e-01 -1.32366106e-01 2.06545651e-01
8.60251307e-01 -3.92495394e-01 9.74407196e-01 3.92773539e-01
-6.31587923e-01 1.01193142e+00 -7.83621967e-02 3.30893070e-01
9.10491049e-01 -1.40797460e+00 4.50456291e-01 1.49074471e+00
3.50888133e-01 6.62861526e-01 8.78273308e-01 -5.08798361e-01
-8.10876191e-01 -9.37380314e-01 1.36951363e+00 -3.09666932e-01
6.83472216e-01 -3.64575803e-01 -1.26744401e+00 3.28187585e-01
4.55461532e-01 -3.27007860e-01 7.37413883e-01 -1.44658759e-01
-3.10869902e-01 -1.22983955e-01 -9.23481345e-01 7.64329612e-01
7.24161923e-01 -4.72905576e-01 -9.30969238e-01 1.78905427e-01
8.32120478e-01 -4.39340353e-01 -6.64410532e-01 1.57909736e-01
2.41936132e-01 -7.09064603e-01 3.20327252e-01 -3.97692204e-01
-2.56685428e-02 -5.68055391e-01 -4.00075257e-01 -1.49735892e+00
-7.85586610e-02 -8.21786940e-01 4.25843686e-01 1.68556857e+00
9.26460803e-01 -4.70533669e-01 4.47532237e-02 7.56020606e-01
-4.62020278e-01 -1.31989783e-02 -8.56594920e-01 -9.54369187e-01
-8.64407942e-02 -5.26221693e-01 7.54250765e-01 8.16107929e-01
5.45938909e-02 4.26242650e-01 9.66812894e-02 3.04369897e-01
1.63212359e-01 -6.98794723e-01 5.51828206e-01 -9.06524658e-01
-1.50882125e-01 -6.42874360e-01 -5.77662408e-01 -7.08543837e-01
1.80135950e-01 -7.19918370e-01 5.75326204e-01 -1.19657183e+00
-5.15505672e-01 -1.91521108e-01 -3.16079199e-01 4.21121478e-01
-2.31638968e-01 -1.46877021e-02 2.00873554e-01 -1.96078748e-01
-4.63797927e-01 2.40260765e-01 6.33276999e-01 3.25753391e-02
-4.77728486e-01 -7.16084018e-02 -2.74175048e-01 8.54406238e-01
1.06988275e+00 -5.03951371e-01 -8.93170759e-03 -5.41157126e-01
8.28851238e-02 -1.55688524e-01 -1.24447435e-01 -1.18150353e+00
9.94886383e-02 -2.49129295e-01 2.26239171e-02 -2.66196370e-01
-2.30823174e-01 -8.03553402e-01 -6.93359375e-02 2.47358903e-01
-3.71553540e-01 4.56581682e-01 1.95677832e-01 -8.72774329e-03
-3.12719494e-01 -6.70124650e-01 9.30242419e-01 -3.14539089e-03
-7.16953635e-01 -3.22908372e-01 -8.17121685e-01 3.05614054e-01
7.25657284e-01 -6.01819515e-01 -1.48942009e-01 -1.44431472e-01
-3.07256877e-01 -1.89282849e-01 4.66000557e-01 6.24768138e-01
4.75733578e-01 -1.00661254e+00 -8.74155521e-01 4.51094091e-01
2.98357904e-01 -3.68806094e-01 -2.89539903e-01 4.32949036e-01
-9.29260731e-01 2.34432593e-01 -9.96505935e-03 -2.40520939e-01
-1.44647062e+00 2.20488831e-01 3.70572329e-01 -6.37405887e-02
-1.10304132e-01 6.56560063e-01 -2.16076195e-01 -6.09630942e-01
4.50798959e-01 -6.65920675e-01 -3.01128298e-01 -1.76370919e-01
5.60604334e-01 4.40058470e-01 7.08798051e-01 -1.04630911e+00
-5.79547644e-01 3.23671520e-01 1.38629973e-02 -4.64920431e-01
6.01030469e-01 -2.04108730e-01 -7.54542649e-02 7.23554254e-01
9.77645636e-01 6.19848490e-01 -3.89999330e-01 -1.02950685e-01
3.59814376e-01 -2.65054584e-01 -3.38567346e-01 -9.21852052e-01
-4.67871755e-01 7.46289849e-01 3.77216995e-01 3.23568523e-01
9.92861271e-01 -2.64340788e-01 7.58865297e-01 4.87419605e-01
2.65572350e-02 -1.57483900e+00 -3.65223348e-01 1.08240664e+00
8.60197425e-01 -1.02726126e+00 -5.67918122e-01 -2.22330391e-01
-7.24244833e-01 1.19996595e+00 6.08851671e-01 2.06733704e-01
3.73470634e-01 2.48742074e-01 4.64818895e-01 2.72667736e-01
-3.40659946e-01 -2.12986618e-01 3.71264279e-01 5.63983023e-01
8.86855781e-01 3.30966324e-01 -9.11976576e-01 4.39835966e-01
-7.32390225e-01 -7.69586504e-01 7.45516300e-01 6.84544683e-01
-6.92588627e-01 -1.15358746e+00 -7.28775978e-01 1.21286742e-01
-6.44969523e-01 -3.61709207e-01 -6.74218118e-01 6.29877150e-01
3.37550491e-02 1.40554428e+00 1.35764897e-01 -1.87010750e-01
7.15642452e-01 4.55301464e-01 1.95959479e-01 -8.03342044e-01
-8.68014574e-01 2.10306849e-02 2.48965323e-01 -1.95343181e-01
-2.50959456e-01 -5.95963895e-01 -1.31456900e+00 -1.39355317e-01
-4.21614885e-01 4.64711994e-01 7.89919436e-01 1.17067766e+00
-2.38500103e-01 4.28007156e-01 5.32367229e-01 8.15926399e-03
-5.07428706e-01 -1.13046110e+00 -6.03311598e-01 5.67371249e-01
2.38721654e-01 -1.30282581e-01 -6.86660528e-01 1.79222107e-01]
|
[14.230024337768555, 7.000101089477539]
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.