Datasets:
J0nasW
/
paperswithcode

Dataset card Data Studio Files
xet
Community
1
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
5884c871-aed2-4bf1-8b64-b0d4d65845fc
aaa-fair-evaluation-for-abuse-detection
null
null
https://dl.acm.org/doi/abs/10.1145/3447535.3462484
https://dl.acm.org/doi/pdf/10.1145/3447535.3462484
AAA: Fair Evaluation for Abuse Detection Systems Wanted
User-generated web content is rife with abusive language that can harm others and discourage participation. Thus, a primary research aim is to develop abuse detection systems that can be used to alert and support human moderators of online communities. Such systems are notoriously hard to develop and evaluate. Even when they appear to achieve satisfactory performance on current evaluation metrics, they may fail in practice on new data. This is partly because datasets commonly used in this field suffer from selection bias, and consequently, existing supervised models overrely on cue words such as group identifiers (e.g., gay and black) which are not inherently abusive. Although there are attempts to mitigate this bias, current evaluation metrics do not adequately quantify their progress. In this work, we introduce Adversarial Attacks against Abuse (AAA), a new evaluation strategy and associated metric that better captures a model’s performance on certain classes of hard-to-classify microposts, and for example penalises systems which are biased on low-level lexical features. It does so by adversarially modifying the model developer’s training and test data to generate plausible test samples dynamically. We make AAA available as an easy-to-use tool, and show its effectiveness in error analysis by comparing the AAA performance of several state-of-the-art models on multiple datasets. This work will inform the development of detection systems and contribute to the fight against abusive language online.
['Roberto Navigli', 'Rocco Tripodi', 'Björn Ross', 'Michele Bevilacqua', 'Agostina Calabrese']
2021-06-21
null
null
null
acm-web-science-2021-6
['abuse-detection']
['natural-language-processing']
[ 1.82174072e-02 -2.48401016e-01 -4.05108631e-01 -3.83736521e-01 -5.17749906e-01 -7.63234675e-01 8.63499939e-01 4.33111459e-01 -6.53046250e-01 7.41904557e-01 7.80898556e-02 -5.92505336e-01 1.61088817e-02 -8.29565823e-01 -3.78034413e-01 -5.63508645e-02 -2.60858238e-01 3.37791711e-01 2.06837222e-01 -4.20115530e-01 3.80784303e-01 4.90644008e-01 -1.18503726e+00 4.41949099e-01 1.06466758e+00 2.48161182e-01 -7.75210261e-01 5.58395624e-01 -1.22367337e-01 7.83436060e-01 -9.88586187e-01 -1.01426780e+00 7.10412562e-02 -4.98919249e-01 -8.10404539e-01 -2.89927065e-01 6.91185534e-01 -6.48757100e-01 -7.67137036e-02 1.27104485e+00 4.67381835e-01 -2.87540495e-01 7.19721377e-01 -1.45238590e+00 -6.87911272e-01 8.94392848e-01 -4.62902337e-01 5.72635829e-01 6.41310155e-01 4.26312476e-01 8.77607882e-01 -6.73212111e-01 7.74013519e-01 1.34399474e+00 7.83272088e-01 7.27021456e-01 -1.45302057e+00 -1.04703319e+00 -1.80687681e-01 1.00826062e-02 -1.17654395e+00 -5.41036189e-01 6.40635788e-01 -6.42690182e-01 7.85738885e-01 5.49028814e-01 5.41027248e-01 1.85577691e+00 -5.44481585e-03 5.82218826e-01 1.33293366e+00 -3.60359132e-01 2.53980875e-01 4.87907827e-01 3.98259282e-01 5.27732670e-01 7.00942934e-01 7.93179125e-03 -4.82620984e-01 -8.78368318e-01 1.12891391e-01 -3.90420973e-01 -9.99269914e-03 -1.61692455e-01 -4.40656573e-01 1.37838686e+00 2.70046473e-01 6.97248459e-01 -2.96645135e-01 -2.05775708e-01 4.60399270e-01 2.92979658e-01 5.95594406e-01 7.99286962e-01 1.40568882e-01 -4.76095915e-01 -1.09264195e+00 6.76913798e-01 1.20458090e+00 2.88769871e-01 2.45482177e-01 4.02064063e-02 -2.24767649e-03 1.20849907e+00 2.72106588e-01 3.67461205e-01 4.51009363e-01 -8.43508422e-01 2.18235537e-01 6.68146193e-01 -1.00949481e-01 -1.18429101e+00 -2.96342283e-01 -3.78354639e-01 -4.46279556e-01 4.88855124e-01 6.82243645e-01 -2.22346529e-01 -5.45378745e-01 1.59533727e+00 1.98309466e-01 -1.69125021e-01 -4.65184718e-01 8.08667243e-01 2.55822301e-01 2.08534926e-01 5.16101778e-01 -7.31589720e-02 8.88020992e-01 -3.88725966e-01 -7.08537042e-01 -4.93415982e-01 1.01387835e+00 -6.83725119e-01 1.07394516e+00 5.15668690e-01 -1.09868944e+00 8.30979422e-02 -1.22724378e+00 3.48200083e-01 -7.93156505e-01 -8.49264324e-01 7.38929987e-01 1.57020044e+00 -7.62172520e-01 7.66443670e-01 -5.06875157e-01 -4.45125103e-01 6.33621514e-01 1.03791445e-01 -3.53166342e-01 -6.78810775e-02 -1.50142539e+00 1.38080633e+00 2.66982429e-03 -3.81314695e-01 -6.61233604e-01 -5.89907110e-01 -6.90520942e-01 -2.39684433e-01 1.96134612e-01 -2.24405363e-01 1.14564037e+00 -1.32036877e+00 -1.03222358e+00 1.04308689e+00 4.15251940e-01 -6.42335057e-01 7.75558710e-01 -2.44706407e-01 -7.01893866e-01 -5.92798851e-02 2.69884408e-01 1.08655602e-01 9.24282968e-01 -1.25835347e+00 -2.68578351e-01 -4.86682832e-01 2.53906518e-01 -2.77645111e-01 -8.85311365e-01 6.89405143e-01 3.02203715e-01 -7.07859218e-01 -5.59564173e-01 -7.72054136e-01 -9.36719254e-02 7.73430094e-02 -3.18450600e-01 -3.81538970e-03 8.86460423e-01 -8.69585276e-01 1.77742982e+00 -1.78988647e+00 -1.52373195e-01 5.21381736e-01 1.64549112e-01 9.44859564e-01 1.35856792e-01 4.24182594e-01 5.04438020e-03 1.03371441e+00 -3.67449939e-01 -3.15163642e-01 -1.34851366e-01 1.95564732e-01 -1.05016202e-01 6.98889613e-01 2.80949593e-01 4.71864462e-01 -9.77500498e-01 -5.15988231e-01 4.83990759e-02 3.55866432e-01 -7.87740409e-01 8.75392929e-02 -2.74498239e-02 -3.54551040e-02 -1.92497164e-01 6.78854644e-01 5.33199787e-01 1.15594223e-01 1.02798060e-01 4.43596303e-01 -3.81438360e-02 4.62621391e-01 -1.02758729e+00 9.49571609e-01 -2.77184069e-01 5.73350549e-01 3.39720577e-01 -5.22268891e-01 7.39763677e-01 -8.64406973e-02 2.05321848e-01 -4.93833035e-01 1.94263220e-01 5.60882092e-01 5.19867301e-01 -5.72933018e-01 3.27258795e-01 -6.72452897e-02 -1.72106177e-02 7.60436594e-01 -1.72573060e-01 5.05300574e-02 2.91724116e-01 5.36560893e-01 1.38869774e+00 -3.20918351e-01 3.46213102e-01 -2.79857554e-02 4.48073715e-01 -6.81983232e-02 1.88823298e-01 9.74089324e-01 -6.04783654e-01 1.74602091e-01 6.24102056e-01 -2.28150830e-01 -1.12954199e+00 -8.49217772e-01 -4.31349874e-01 1.19538009e+00 -3.66475403e-01 -5.58124781e-01 -9.33559597e-01 -1.17543840e+00 2.86928922e-01 1.16628921e+00 -5.86950302e-01 -4.09446180e-01 -4.15041625e-01 -8.73153329e-01 1.28530252e+00 1.12206303e-01 3.00560743e-01 -8.81877542e-01 -4.05279636e-01 4.08982396e-01 -9.94440466e-02 -7.11515427e-01 -1.03664204e-01 -2.26014286e-01 -3.95932198e-01 -1.36561954e+00 -4.00820136e-01 6.39717057e-02 2.53141701e-01 3.35367471e-02 1.02789974e+00 6.49946868e-01 -2.58362710e-01 2.19771191e-01 -6.16707325e-01 -4.67157483e-01 -9.98185039e-01 1.48116201e-01 2.63273358e-01 4.73128632e-03 8.78595829e-01 -5.66061676e-01 -3.08972329e-01 2.28302419e-01 -1.18349254e+00 -5.68698049e-01 2.44583681e-01 6.46057248e-01 -4.45462018e-01 -4.57966626e-01 7.49131560e-01 -1.44183934e+00 1.25059223e+00 -9.85261500e-01 -6.52623102e-02 -1.72083944e-01 -9.05730903e-01 -3.21457148e-01 5.24607539e-01 -6.95872366e-01 -6.44257367e-01 -5.18155396e-01 -4.54471469e-01 -1.86908022e-01 -1.35801762e-01 5.53339839e-01 9.92965028e-02 -2.73578703e-01 1.48764181e+00 -3.12965930e-01 1.74488857e-01 -5.33892095e-01 7.71152824e-02 1.22894919e+00 -6.08806163e-02 -4.35694516e-01 9.43404019e-01 2.18158752e-01 -4.26205665e-01 -9.78764772e-01 -5.69344223e-01 -4.12519127e-01 -3.61553311e-01 -3.26316476e-01 3.77967060e-01 -3.63270700e-01 -3.65691632e-01 6.52958274e-01 -1.09599936e+00 -2.70481080e-01 3.77750933e-01 3.87450308e-02 8.40332638e-03 7.17617214e-01 -6.79370642e-01 -1.16516030e+00 -3.35330188e-01 -7.75323212e-01 4.07193959e-01 -2.36264393e-02 -9.27432358e-01 -9.55373466e-01 2.34461173e-01 5.61939895e-01 7.17589378e-01 5.41230381e-01 8.42468619e-01 -1.28996551e+00 1.94498718e-01 -4.54089224e-01 2.19231211e-02 7.72081792e-01 -1.98476031e-01 5.68942964e-01 -9.30017769e-01 -3.65613014e-01 -1.82191983e-01 -4.97248560e-01 4.64366555e-01 -3.37730885e-01 9.48342443e-01 -7.13205755e-01 -2.46334568e-01 1.29387856e-01 1.08540893e+00 -7.70104602e-02 6.06668472e-01 6.08021379e-01 4.41937119e-01 8.55981946e-01 2.47926876e-01 4.61374879e-01 3.52345556e-02 8.38506222e-01 4.19108570e-01 1.67625636e-01 2.91866530e-02 -2.99717367e-01 6.30537391e-01 1.57108098e-01 1.81522548e-01 -1.04314178e-01 -1.19651949e+00 4.75085467e-01 -1.46672273e+00 -1.13401294e+00 -5.55957019e-01 2.16626406e+00 1.12521148e+00 6.53366327e-01 6.11841619e-01 3.76084834e-01 5.83206892e-01 2.38676175e-01 -1.93777665e-01 -9.89117563e-01 -8.55358467e-02 4.23664212e-01 4.15982068e-01 7.91728675e-01 -9.65250909e-01 8.30145836e-01 6.32483292e+00 6.05375528e-01 -9.51649308e-01 3.33872795e-01 4.89388674e-01 -2.84194857e-01 -2.62275815e-01 -4.83221151e-02 -6.68894291e-01 8.90732527e-01 1.23227453e+00 -1.67160273e-01 5.57823658e-01 1.07718372e+00 1.61984086e-01 -1.13608763e-02 -1.05400670e+00 5.61639249e-01 4.61120397e-01 -1.04791367e+00 9.91164986e-03 3.41469735e-01 4.64974821e-01 -7.91322291e-02 1.87769495e-02 5.63589513e-01 4.79396671e-01 -1.29841459e+00 5.17172217e-01 1.33479699e-01 3.87958556e-01 -7.54143178e-01 6.93352044e-01 5.36692262e-01 -2.39523631e-02 -2.72132009e-01 -5.25760613e-02 -2.91576356e-01 2.90398262e-02 4.48284030e-01 -8.90627742e-01 -1.30827442e-01 6.55608356e-01 3.51704329e-01 -1.07424212e+00 9.78507161e-01 -2.44790584e-01 1.09282064e+00 -4.11257297e-01 -3.44064951e-01 3.32592785e-01 1.33520409e-01 7.88544178e-01 1.56382239e+00 -2.28771731e-01 -1.44374251e-01 2.53221914e-02 8.80049407e-01 -8.69431794e-02 2.77457088e-01 -1.09039998e+00 -2.75144964e-01 6.12931311e-01 1.14244032e+00 -1.54058591e-01 -1.69522583e-01 -2.87907690e-01 7.49927759e-01 5.15380621e-01 6.80889487e-02 -6.59869790e-01 -2.70291537e-01 8.52609396e-01 7.11103082e-01 -4.85592037e-01 -9.56359953e-02 -4.43645537e-01 -1.04496074e+00 -1.41044781e-01 -1.66044760e+00 5.80063283e-01 -3.38008732e-01 -1.67447710e+00 3.13341528e-01 2.04439461e-02 -6.95775449e-01 -6.80209398e-01 -6.04655504e-01 -4.92068768e-01 7.76820362e-01 -1.03380799e+00 -1.17970228e+00 -2.36848518e-02 4.53085601e-01 1.98659688e-01 -3.53584141e-02 9.56983089e-01 4.99835908e-01 -6.39162242e-01 9.31316018e-01 -1.70475498e-01 5.11773288e-01 9.62951660e-01 -1.15808082e+00 3.13305169e-01 9.26449060e-01 1.81721032e-01 1.00471199e+00 9.54295933e-01 -9.09317732e-01 -8.09567273e-01 -7.36818910e-01 8.49509537e-01 -9.43110406e-01 1.00871086e+00 -4.80170846e-01 -1.26509070e+00 6.26143754e-01 -6.35981262e-02 -2.89161026e-01 1.04107475e+00 4.10092026e-01 -9.35681403e-01 3.16811591e-01 -1.46319973e+00 7.35205948e-01 1.10015059e+00 -4.54655230e-01 -5.63272059e-01 4.42469031e-01 -4.63329479e-02 3.01636895e-03 -7.33990610e-01 1.39054671e-01 5.51562905e-01 -1.28993404e+00 7.35335231e-01 -1.18362868e+00 6.26471221e-01 3.05937320e-01 9.09538344e-02 -1.23565960e+00 -2.89156348e-01 -6.65883839e-01 -2.77047396e-01 1.43317652e+00 4.95634735e-01 -7.65962899e-01 5.84659517e-01 9.77137029e-01 4.32975441e-01 -4.92620915e-01 -9.90677595e-01 -7.38348067e-01 4.14028704e-01 -4.97769773e-01 2.61054456e-01 1.52491331e+00 3.12419862e-01 3.73438120e-01 -4.26177889e-01 -4.88260463e-02 7.07510650e-01 -6.71863079e-01 9.30015147e-01 -1.19709253e+00 -2.05635563e-01 -8.46852183e-01 -5.29487312e-01 -8.91444087e-03 2.42374361e-01 -8.78789842e-01 -5.76441526e-01 -9.67835426e-01 1.40029848e-01 -6.27569079e-01 1.14235982e-01 6.25874937e-01 -2.99559563e-01 4.97476041e-01 2.48812482e-01 1.29188433e-01 -1.82169080e-01 1.05738277e-02 4.69058841e-01 -1.11344703e-01 -7.45222569e-02 -4.96059507e-02 -7.88262904e-01 8.14310908e-01 8.52560461e-01 -7.75435746e-01 2.62234434e-02 2.19959170e-02 3.57389331e-01 -5.18037260e-01 6.09695137e-01 -9.26799417e-01 -1.75489530e-01 -1.99152499e-01 3.07561129e-01 1.72914416e-01 2.12126642e-01 -5.55846214e-01 -7.03347847e-02 6.26460552e-01 -4.07314986e-01 -1.26636893e-01 5.99943008e-03 3.00454766e-01 2.85140246e-01 -7.58693933e-01 9.55818713e-01 -7.01967627e-02 -1.44428372e-01 2.63024885e-02 -6.24340236e-01 2.59450793e-01 1.10806572e+00 -8.92882571e-02 -6.17525518e-01 -6.14061713e-01 -4.13029194e-01 2.00507790e-02 8.02310884e-01 6.96105599e-01 3.24289888e-01 -1.02338278e+00 -1.05287743e+00 1.02927469e-01 2.07653657e-01 -9.14611936e-01 -2.02549577e-01 4.82853562e-01 -5.33002436e-01 -1.22027278e-01 -4.34631377e-01 -1.98879108e-01 -1.49646568e+00 4.65660989e-01 5.03177524e-01 -3.01400483e-01 2.20938716e-02 6.16787553e-01 -7.33525932e-01 -2.83187807e-01 1.24872178e-01 4.89779532e-01 -2.60979056e-01 1.23388149e-01 7.78029263e-01 6.12494469e-01 -7.86252320e-02 -7.57296622e-01 -4.69687581e-01 -3.94345492e-01 -5.25706112e-01 -2.40790561e-01 1.12845325e+00 4.34457868e-01 -3.13556910e-01 3.40141833e-01 1.11246347e+00 5.36619544e-01 -5.46553075e-01 5.81515655e-02 1.98132023e-01 -1.01321375e+00 -1.19686127e-01 -1.06764531e+00 -4.88920301e-01 7.22889781e-01 5.39252043e-01 8.34592879e-01 3.39740753e-01 -3.53999734e-01 5.20439804e-01 -2.99375486e-02 3.07901889e-01 -1.05778170e+00 3.04315269e-01 2.94062674e-01 8.72133672e-01 -1.22010708e+00 2.29207814e-01 -4.35684264e-01 -4.21218008e-01 8.14456582e-01 9.02101815e-01 -9.79750454e-02 3.36510658e-01 1.30003110e-01 1.21790364e-01 -1.77616850e-01 -3.74876231e-01 8.26325864e-02 -7.30989734e-03 7.38490701e-01 5.71881413e-01 1.20969042e-01 -1.02847159e+00 5.31700432e-01 -1.82743281e-01 -1.18833199e-01 8.06882024e-01 8.78742278e-01 -3.18958580e-01 -1.46976984e+00 -5.03360927e-01 8.25459003e-01 -9.63562310e-01 -2.54388154e-02 -1.22513294e+00 8.88922691e-01 5.21173403e-02 1.22011638e+00 -3.73609275e-01 -5.26413620e-01 3.17487061e-01 3.41120720e-01 1.47597149e-01 -7.72204399e-01 -1.27597022e+00 -5.82332850e-01 7.43928909e-01 -4.48977858e-01 -3.57280336e-02 -7.20008194e-01 -5.49688578e-01 -9.65927124e-01 -2.74014145e-01 -6.75594285e-02 6.63903415e-01 7.57453501e-01 2.40032479e-01 -1.05274506e-01 6.24722958e-01 -3.58067811e-01 -1.17338729e+00 -1.30560601e+00 -2.09318638e-01 1.00946593e+00 5.39322384e-02 -7.26361215e-01 -6.44097805e-01 -4.88273770e-01]
[8.6823148727417, 10.465112686157227]
d242db40-fb74-429c-9486-6adfa982d862
unsupervised-parallel-corpus-mining-on-web
2009.08595
null
https://arxiv.org/abs/2009.08595v1
https://arxiv.org/pdf/2009.08595v1.pdf
Unsupervised Parallel Corpus Mining on Web Data
With a large amount of parallel data, neural machine translation systems are able to deliver human-level performance for sentence-level translation. However, it is costly to label a large amount of parallel data by humans. In contrast, there is a large-scale of parallel corpus created by humans on the Internet. The major difficulty to utilize them is how to filter them out from the noise website environments. Current parallel data mining methods all require labeled parallel data as the training source. In this paper, we present a pipeline to mine the parallel corpus from the Internet in an unsupervised manner. On the widely used WMT'14 English-French and WMT'16 English-German benchmarks, the machine translator trained with the data extracted by our pipeline achieves very close performance to the supervised results. On the WMT'16 English-Romanian and Romanian-English benchmarks, our system produces new state-of-the-art results, 39.81 and 38.95 BLEU scores, even compared with supervised approaches.
['Zihang Dai', 'Yiming Yang', 'Guokun Lai']
2020-09-18
null
null
null
null
['parallel-corpus-mining']
['natural-language-processing']
[ 1.91008478e-01 -4.09647733e-01 -4.50547218e-01 -2.87828594e-01 -1.48210955e+00 -6.93189085e-01 6.05217636e-01 -2.58105490e-02 -8.32404852e-01 8.99773479e-01 4.36514197e-03 -6.72967255e-01 3.88922274e-01 -6.19147241e-01 -7.89837956e-01 -2.41542131e-01 2.49258906e-01 1.07015526e+00 -9.49028358e-02 -5.77993453e-01 1.32656842e-01 -4.44816351e-02 -9.24345076e-01 5.96036911e-01 1.13810527e+00 5.16232669e-01 4.85806555e-01 4.87337172e-01 -3.60932320e-01 1.34038448e-01 -5.96706331e-01 -6.79930210e-01 6.23652160e-01 -8.29391122e-01 -1.05926037e+00 -2.45208338e-01 3.67544711e-01 5.94308600e-02 -1.30444437e-01 1.08742261e+00 6.07625306e-01 -2.89724112e-01 2.74164915e-01 -7.72467017e-01 -5.50207019e-01 9.17920887e-01 -6.35110795e-01 2.83632845e-01 2.98566878e-01 -2.76522133e-02 1.08087361e+00 -1.18366516e+00 1.04747200e+00 1.04554641e+00 5.11122525e-01 4.31169778e-01 -1.22801006e+00 -7.28294730e-01 -2.53262848e-01 1.30619228e-01 -1.26601827e+00 -4.75198686e-01 3.87153327e-01 -8.43989030e-02 1.44848573e+00 2.95636803e-01 4.14366603e-01 1.42187572e+00 4.73561823e-01 8.75052810e-01 1.38110423e+00 -8.15069675e-01 -1.12140469e-01 8.56679454e-02 5.41048124e-02 5.70643365e-01 -9.69893932e-02 8.77736807e-02 -7.76397705e-01 -1.52245358e-01 2.20880270e-01 -3.41907650e-01 -4.86550294e-02 2.52458513e-01 -1.66916788e+00 7.56935537e-01 2.96823885e-02 3.34466815e-01 -4.40546662e-01 -4.18209344e-01 7.30476320e-01 9.35934186e-01 7.73183584e-01 6.00554883e-01 -8.32335711e-01 -4.77004349e-01 -9.84540164e-01 5.59472069e-02 1.11209214e+00 1.30232751e+00 7.79016018e-01 -4.20858592e-01 6.15523979e-02 1.11714005e+00 -9.20664966e-02 8.73343527e-01 6.92161500e-01 -5.34560204e-01 1.14637864e+00 4.98467743e-01 -1.44641548e-01 -4.92161661e-01 -1.84272662e-01 -5.99352360e-01 -9.75759387e-01 -2.27586895e-01 4.56736356e-01 -3.21843714e-01 -7.59221911e-01 1.54249585e+00 2.48031802e-02 -7.07022429e-01 3.77393484e-01 7.99294770e-01 4.03978586e-01 8.84724319e-01 -3.55345279e-01 -5.15445411e-01 1.17134714e+00 -1.46394658e+00 -7.05138266e-01 -4.72416282e-01 7.59940088e-01 -1.45603013e+00 1.38241529e+00 3.88468295e-01 -1.05828929e+00 -4.88468498e-01 -9.05356407e-01 -3.51855680e-02 -2.85551757e-01 1.68452397e-01 3.75243992e-01 2.66628087e-01 -8.97553980e-01 8.19051087e-01 -7.37369955e-01 -7.34157503e-01 2.21382454e-01 3.51483554e-01 -4.52242881e-01 -3.69670719e-01 -1.25224781e+00 1.12539065e+00 2.85524696e-01 -2.40238547e-01 -7.17402458e-01 -3.60791534e-01 -4.18435186e-01 -2.02324376e-01 1.76600426e-01 -6.80111945e-01 1.50531673e+00 -1.24129426e+00 -1.56020975e+00 9.51602817e-01 -4.17286843e-01 -3.89370978e-01 6.73405826e-01 -2.61485010e-01 -3.77313823e-01 -1.64679199e-01 4.64704752e-01 7.06194818e-01 4.47138995e-01 -8.65527391e-01 -9.24742460e-01 -2.99570709e-01 -3.19576085e-01 2.63650715e-01 -4.03345853e-01 5.17617643e-01 -6.08143687e-01 -4.88629162e-01 1.20209463e-01 -1.06812990e+00 -3.43218654e-01 -5.46418130e-01 -3.64152104e-01 -5.16375065e-01 5.91341197e-01 -9.32217598e-01 1.10404193e+00 -1.75440109e+00 1.38755426e-01 -6.05754256e-02 -1.56103596e-01 2.78185099e-01 -6.36998057e-01 7.39698768e-01 1.77562967e-01 1.76915228e-01 -1.77346244e-01 -4.66582358e-01 -1.98933944e-01 3.66975248e-01 -9.74546298e-02 3.41090381e-01 2.15621844e-01 9.48946595e-01 -1.07199728e+00 -4.98049796e-01 -3.64825189e-01 -6.45339936e-02 -3.86121422e-01 2.20627800e-01 -2.57287741e-01 5.00321925e-01 -2.43281081e-01 5.50090373e-01 3.61220747e-01 -2.40596250e-01 2.98061311e-01 3.80739570e-01 -1.76377639e-01 8.56106043e-01 -4.60221857e-01 2.33795452e+00 -5.99171400e-01 7.90179670e-01 -1.34933412e-01 -5.81127763e-01 1.06830645e+00 4.58902687e-01 2.92278796e-01 -1.01709366e+00 1.45273115e-02 7.75936306e-01 4.12055373e-01 -3.20940286e-01 6.10405505e-01 -1.42507687e-01 -2.44394168e-01 8.53948593e-01 1.54580697e-01 4.57654335e-02 6.57797515e-01 4.87311929e-02 1.25413764e+00 1.32205024e-01 9.05877426e-02 -4.47017342e-01 2.79822916e-01 5.88453591e-01 9.33227241e-01 4.97428834e-01 6.71977103e-02 6.32383466e-01 2.42855906e-01 -7.31454372e-01 -1.55325675e+00 -7.42195487e-01 1.44135877e-01 9.95492756e-01 -3.80761147e-01 -8.09767663e-01 -8.19594324e-01 -9.96260405e-01 -5.44258177e-01 4.81851012e-01 -1.20913088e-01 2.26132646e-01 -9.32197034e-01 -8.94313574e-01 6.14455819e-01 1.80485025e-01 5.53010941e-01 -9.97794390e-01 3.77361812e-02 3.41565847e-01 -9.26961958e-01 -1.17437661e+00 -7.48741508e-01 3.00550401e-01 -8.91936958e-01 -5.64919472e-01 -5.72363317e-01 -1.04062366e+00 5.17342150e-01 1.77481920e-01 1.63089895e+00 -2.13723674e-01 2.88440377e-01 -5.91183782e-01 -4.38637674e-01 -4.34845269e-01 -9.93560910e-01 8.37249875e-01 2.39115119e-01 -4.97444034e-01 7.99294472e-01 -5.49446344e-01 -1.72309473e-01 5.11969507e-01 -4.96711582e-01 3.88368636e-01 8.71738970e-01 1.13677430e+00 7.05247164e-01 -3.63955438e-01 6.62757516e-01 -8.02439153e-01 8.11817944e-01 -3.56509387e-01 -5.40205538e-01 2.53616542e-01 -9.04226363e-01 -3.99642065e-02 1.04203570e+00 -4.83838409e-01 -7.24576414e-01 -5.68095371e-02 -3.28287810e-01 -1.34260997e-01 -1.26080081e-01 6.37786984e-01 -9.25799981e-02 3.66635144e-01 9.42515254e-01 3.06185424e-01 -4.66393158e-02 -6.64489150e-01 1.39148951e-01 1.12143421e+00 2.03482360e-01 -6.77035153e-01 7.54717827e-01 -7.14851469e-02 -3.26032162e-01 -5.69353998e-01 -6.75915897e-01 -4.02603745e-01 -8.27002585e-01 2.45528877e-01 5.35161316e-01 -9.63085413e-01 -1.45260423e-01 3.29832494e-01 -1.47358775e+00 -3.58181328e-01 -9.39839184e-02 5.74708343e-01 -4.73102063e-01 3.72090042e-01 -1.08143544e+00 -1.23990007e-01 -9.32595909e-01 -1.35436416e+00 8.78493249e-01 -1.64260969e-01 -5.98597944e-01 -6.99207246e-01 2.98746616e-01 4.58114147e-01 2.54744917e-01 -2.81697601e-01 1.06544530e+00 -8.85491073e-01 -3.42637271e-01 -1.57514334e-01 -4.15467992e-02 4.88634944e-01 1.68284133e-01 -3.71642113e-01 -5.97071648e-01 -4.65407431e-01 -1.99217722e-02 -4.43650454e-01 4.60750252e-01 -1.74148411e-01 4.72298771e-01 -2.46790707e-01 -1.17647596e-01 3.66720825e-01 1.28442490e+00 1.17737040e-01 3.16075295e-01 5.20791590e-01 4.11945850e-01 5.89628637e-01 6.12301826e-01 -7.10082129e-02 2.09099859e-01 5.88735282e-01 -1.85601681e-01 -1.71549231e-01 -1.20090015e-01 -5.69975793e-01 6.47911429e-01 1.80736208e+00 3.82368360e-03 -2.28817567e-01 -1.16981244e+00 5.23407698e-01 -1.77190101e+00 -5.58681130e-01 -4.19068247e-01 1.94550824e+00 1.22668540e+00 3.00528675e-01 -4.62522788e-04 -6.45080954e-02 6.47909284e-01 -3.66586745e-01 -2.52119094e-01 -6.49369001e-01 -3.14181775e-01 2.54633814e-01 6.04452431e-01 4.65133607e-01 -7.92342722e-01 1.30783105e+00 6.60895443e+00 9.97096956e-01 -1.09799242e+00 4.52487618e-01 5.50816476e-01 -8.14738497e-02 -2.22716883e-01 9.73803625e-02 -8.22601557e-01 4.09411520e-01 1.62346387e+00 -1.94250271e-01 6.97385013e-01 4.82528538e-01 2.94058502e-01 1.67517617e-01 -1.28717589e+00 9.30608630e-01 -4.35911417e-02 -1.28925514e+00 8.48643407e-02 3.39732856e-01 1.04797840e+00 8.57502699e-01 -2.41701499e-01 4.49632823e-01 3.93263876e-01 -8.36886942e-01 6.33416176e-01 9.46571305e-02 9.03993130e-01 -8.13345909e-01 8.11405659e-01 9.36672688e-01 -6.06940091e-01 3.21853071e-01 -5.66107213e-01 -2.43523970e-01 1.11906558e-01 6.98932827e-01 -9.31925952e-01 8.25225830e-01 6.43362820e-01 7.60485291e-01 -3.57256919e-01 4.96296376e-01 -5.05491734e-01 8.64093661e-01 -2.89304912e-01 -8.83971751e-02 4.36217159e-01 -5.00884414e-01 4.80968952e-01 1.32514644e+00 3.51563543e-01 -4.02200788e-01 2.55931288e-01 6.00460291e-01 -5.44106841e-01 6.16584003e-01 -8.04868817e-01 -2.19761685e-01 2.29197338e-01 1.12870908e+00 -4.68773633e-01 -5.08432686e-01 -6.30602062e-01 1.30755532e+00 6.10751748e-01 3.82491857e-01 -5.46878695e-01 -2.75168598e-01 4.91558969e-01 -1.52920514e-01 -1.49070293e-01 -4.33391988e-01 -4.07775283e-01 -1.38046408e+00 3.24639887e-01 -1.67986715e+00 1.65455833e-01 -3.36399794e-01 -1.61912417e+00 1.17606246e+00 -4.81901526e-01 -1.24120498e+00 -5.14512539e-01 -6.61751926e-01 -3.94285917e-01 1.33604598e+00 -1.31249368e+00 -1.05176115e+00 3.27821225e-01 3.81286323e-01 1.01939130e+00 -6.63272440e-01 1.22166824e+00 5.59807420e-01 -5.37642002e-01 6.71256065e-01 4.87259269e-01 4.44094181e-01 1.01061654e+00 -1.05487621e+00 1.17747879e+00 1.00052750e+00 5.40359259e-01 6.44571841e-01 4.92380649e-01 -6.13626003e-01 -1.65800071e+00 -9.46889937e-01 1.80187023e+00 -6.11920178e-01 6.47393227e-01 -6.57854259e-01 -6.97245181e-01 6.59813344e-01 8.06387126e-01 -3.43616962e-01 7.42424905e-01 4.35352772e-01 -3.33322257e-01 -3.67633663e-02 -7.38234282e-01 5.80078363e-01 1.15395617e+00 -4.81627762e-01 -7.25421190e-01 6.58905745e-01 6.80289209e-01 -3.77857089e-01 -9.12141144e-01 2.44901761e-01 3.61208498e-01 -5.55384636e-01 4.91166413e-01 -8.33857000e-01 8.88379097e-01 -4.99820448e-02 -2.15812311e-01 -1.73175049e+00 -1.13851212e-01 -9.10018325e-01 4.06744659e-01 1.00335610e+00 1.10045946e+00 -5.96336901e-01 7.03992963e-01 2.81703547e-02 -2.89380968e-01 -8.62860501e-01 -1.03540933e+00 -9.42131579e-01 6.81051433e-01 -2.09532142e-01 4.66343552e-01 9.80916083e-01 1.96758986e-01 1.17422438e+00 -3.84899765e-01 -3.05351526e-01 5.11244893e-01 3.18510175e-01 9.22015309e-01 -1.04249918e+00 -2.92803288e-01 -4.26996410e-01 1.80998355e-01 -1.25244915e+00 3.57793421e-01 -1.38518167e+00 1.00238226e-01 -1.55226755e+00 4.61489230e-01 -5.26710927e-01 -2.28249449e-02 4.01787549e-01 -9.74299535e-02 2.85990298e-01 3.65406759e-02 5.98529100e-01 -4.77126628e-01 3.84908050e-01 1.34125936e+00 -2.95060754e-01 -7.00185299e-02 -1.57584861e-01 -3.90964836e-01 4.71981823e-01 1.13342333e+00 -8.81120265e-01 -5.36250882e-02 -1.10477006e+00 2.47489020e-01 4.56756767e-04 -3.94425243e-01 -6.54275715e-01 2.29248121e-01 -8.67728442e-02 2.54385382e-01 -6.71761096e-01 -1.19342431e-02 -5.30981600e-01 -4.91411611e-02 4.55611438e-01 -1.50543302e-01 7.92120457e-01 1.09975182e-01 1.39512122e-01 -4.36766565e-01 -2.60504242e-03 6.72714233e-01 -3.46602499e-01 -1.76663160e-01 2.11673260e-01 -5.74739337e-01 2.75081933e-01 3.99704039e-01 3.54617208e-01 -3.65099639e-01 -1.06435515e-01 -2.22397685e-01 1.86777204e-01 4.83917683e-01 7.92680025e-01 2.90536135e-01 -1.35955429e+00 -1.22030020e+00 3.39624137e-01 1.96214616e-01 -1.97285667e-01 -2.98257589e-01 9.00279224e-01 -4.83769566e-01 5.60985327e-01 -3.32182646e-01 -6.67636514e-01 -1.08791566e+00 5.40565789e-01 5.22497930e-02 -6.81514084e-01 -6.00588739e-01 4.16513532e-01 -4.31797296e-01 -8.52470100e-01 -1.17163798e-02 -1.93490341e-01 4.55797434e-01 -1.88095942e-01 3.78193945e-01 2.19961941e-01 5.79749823e-01 -5.26682079e-01 -5.25852777e-02 8.69268030e-02 -3.59599113e-01 -4.18905199e-01 1.24600208e+00 -6.79305345e-02 -5.87792754e-01 3.76816005e-01 1.36707342e+00 7.03669488e-02 -5.98316073e-01 -5.55860758e-01 4.67616171e-01 -3.38182390e-01 -4.42601979e-01 -1.04994738e+00 -5.75716615e-01 1.02585757e+00 3.34829837e-01 -2.90606588e-01 9.02499139e-01 -2.26497501e-01 1.22945487e+00 7.90642440e-01 7.58714199e-01 -1.27189004e+00 -2.53981322e-01 1.08201826e+00 7.51018405e-01 -1.49913251e+00 -4.57239479e-01 -2.01859131e-01 -3.37321132e-01 1.04415107e+00 4.69178885e-01 1.79524943e-01 1.67423621e-01 3.81592810e-01 6.48422599e-01 2.10799575e-01 -9.42090333e-01 1.30702630e-01 1.72843963e-01 1.46815211e-01 7.79425025e-01 1.90481126e-01 -8.89584959e-01 4.79110271e-01 -5.23242712e-01 -2.14661688e-01 3.72170389e-01 8.61196995e-01 -2.40050793e-01 -1.93683434e+00 -1.98783949e-01 3.53824884e-01 -7.65023530e-01 -5.02748013e-01 -6.66169524e-01 6.44901812e-01 -1.22369170e-01 1.26988709e+00 -1.52025402e-01 -5.82503378e-01 3.94360036e-01 4.00809646e-01 3.51658374e-01 -7.42451072e-01 -8.94370079e-01 2.18404219e-01 4.68531013e-01 -4.04375762e-01 -1.55100882e-01 -6.12111270e-01 -1.04516792e+00 -5.89446783e-01 -1.86321691e-01 6.29170954e-01 8.07615101e-01 9.41237807e-01 3.70990038e-01 1.33363962e-01 7.05410123e-01 -3.89689147e-01 -8.26810062e-01 -1.45260537e+00 -7.75002465e-02 4.50168937e-01 -1.05681345e-01 1.78963453e-01 5.09773120e-02 -4.79334109e-02]
[11.568166732788086, 10.311986923217773]
c1c372b1-78a2-49a5-a649-e5028bbe78f0
generalizing-a-person-retrieval-model-hetero
null
null
http://openaccess.thecvf.com/content_ECCV_2018/html/Zhun_Zhong_Generalizing_A_Person_ECCV_2018_paper.html
http://openaccess.thecvf.com/content_ECCV_2018/papers/Zhun_Zhong_Generalizing_A_Person_ECCV_2018_paper.pdf
Generalizing A Person Retrieval Model Hetero- and Homogeneously
Person re-identification (re-ID) poses unique challenges for unsupervised domain adaptation (UDA) in that classes in the source and target sets (domains) are entirely different and that image variations are largely caused by cameras. Given a labeled source training set and an unlabeled target training set, we aim to improve the generalization ability of re-ID models on the target testing set. To this end, we introduce a Hetero-Homogeneous Learning (HHL) method. Our method enforces two properties simultaneously: 1) camera invariance, learned via positive pairs formed by unlabeled target images and their camera style transferred counterparts; 2) domain connectedness, by regarding source / target images as negative matching pairs to the target / source images. The first property is implemented by homogeneous learning because training pairs are collected from the same domain. The second property is achieved by heterogeneous learning because we sample training pairs from both the source and target domains. On Market-1501, DukeMTMC-reID and CUHK03, we show that the two properties contribute indispensably and that very competitive re-ID UDA accuracy is achieved. Code is available at: https://github.com/zhunzhong07/HHL
['Zhun Zhong', 'Shaozi Li', 'Yi Yang', 'Liang Zheng']
2018-09-01
null
null
null
eccv-2018-9
['person-retrieval']
['computer-vision']
[ 7.41225109e-02 -1.52607098e-01 -3.81923527e-01 -4.71694559e-01 -6.79549932e-01 -7.94698417e-01 8.42424393e-01 -4.18585658e-01 -4.59279865e-01 8.79296899e-01 1.17837436e-01 2.91585326e-01 8.89317393e-02 -5.58622181e-01 -8.54116499e-01 -5.49201131e-01 3.18783194e-01 8.28667402e-01 9.37243402e-02 -1.47783220e-01 -2.67895460e-01 4.09082681e-01 -1.37787354e+00 1.46476462e-01 9.68053579e-01 6.36335850e-01 2.44041961e-02 4.94616479e-01 1.12977624e-01 6.47191882e-01 -4.03730750e-01 -7.88633287e-01 8.21253061e-01 -6.80518270e-01 -8.92350554e-01 3.76159966e-01 7.97998905e-01 -3.17039877e-01 -5.10333776e-01 1.32025492e+00 4.89783108e-01 1.22886114e-01 9.29611504e-01 -1.62892759e+00 -9.97086525e-01 1.41223922e-01 -6.07220829e-01 -4.93288934e-02 4.33579504e-01 -1.61888618e-02 6.38859451e-01 -8.63827348e-01 8.09400797e-01 1.21421719e+00 5.94596207e-01 1.12354016e+00 -1.39272475e+00 -9.84886765e-01 2.88209189e-02 3.65632415e-01 -1.57851851e+00 -6.13035142e-01 8.25562716e-01 -5.97062707e-01 2.33887210e-01 3.09888236e-02 2.87521124e-01 1.49783134e+00 -4.65377569e-01 7.36263037e-01 1.24607623e+00 -5.69859684e-01 1.01212613e-01 8.16697598e-01 3.30994576e-01 2.81955659e-01 2.04598561e-01 3.07650536e-01 -3.97321880e-01 -3.19882929e-02 7.97944188e-01 9.13307220e-02 -3.90013039e-01 -7.25825667e-01 -1.12737024e+00 7.03211963e-01 2.57492840e-01 2.35292628e-01 7.35504031e-02 -5.56579590e-01 3.57480735e-01 5.62876284e-01 1.80992231e-01 1.86104968e-01 -3.28044564e-01 2.77360260e-01 -4.23921108e-01 1.23543710e-01 6.18803024e-01 1.45926702e+00 9.22408104e-01 -2.00857252e-01 2.45012324e-02 1.14713597e+00 -7.64700547e-02 9.75671649e-01 8.61051738e-01 -7.23721743e-01 5.73802888e-01 5.25942326e-01 2.80423075e-01 -7.03437567e-01 -1.55001298e-01 -1.38842672e-01 -9.93469954e-01 7.57915452e-02 8.83682907e-01 -8.12719390e-02 -8.37390959e-01 1.96687341e+00 2.01691389e-01 2.95753419e-01 3.64214420e-01 7.21360505e-01 7.41391897e-01 3.39358926e-01 1.06260926e-02 2.52576508e-02 1.20269680e+00 -1.02350485e+00 -3.60172749e-01 -3.35587978e-01 4.93256122e-01 -5.59550405e-01 1.02079058e+00 2.91204154e-02 -7.76082933e-01 -1.04614055e+00 -1.08094513e+00 4.16109711e-02 -5.40597379e-01 4.46304083e-01 -3.17506306e-02 7.94113100e-01 -9.35709298e-01 2.23166913e-01 -1.42136857e-01 -7.82210171e-01 3.74439329e-01 4.37047422e-01 -8.61155510e-01 -5.08864641e-01 -1.07100677e+00 7.82241166e-01 3.99843484e-01 -4.19934243e-01 -7.62083590e-01 -6.25106275e-01 -7.56343961e-01 -2.62381971e-01 3.12110744e-02 -5.16769707e-01 1.13229525e+00 -1.52595651e+00 -1.29609716e+00 1.54002011e+00 -2.42358342e-01 -3.77715886e-01 8.07866693e-01 4.68517886e-03 -8.10891449e-01 2.85990015e-02 4.46436644e-01 6.87983036e-01 9.10959125e-01 -1.50583518e+00 -7.26485670e-01 -7.49500215e-01 -2.12691769e-01 2.30353609e-01 -5.63594878e-01 -1.56193227e-01 -6.68122292e-01 -6.02677584e-01 -2.28583187e-01 -1.13473749e+00 1.65252611e-01 -2.16859713e-01 -3.83111805e-01 -4.48011495e-02 7.60259688e-01 -7.56958127e-01 6.43487215e-01 -2.19658756e+00 1.63588852e-01 3.45119238e-01 4.63063978e-02 4.06758547e-01 -3.61459464e-01 7.59949684e-02 -5.34352958e-01 -9.13844556e-02 -2.35334694e-01 -3.91768366e-01 -1.12102717e-01 7.13748783e-02 -2.26033077e-01 5.39203346e-01 1.25270467e-02 7.64019549e-01 -8.32022488e-01 -4.62384284e-01 2.08008572e-01 1.49623677e-01 -1.59965292e-01 5.67153633e-01 3.40046197e-01 9.11875486e-01 -1.85633048e-01 5.04411936e-01 9.03659403e-01 -1.03537910e-01 2.93406337e-01 -2.30064169e-01 2.59110808e-01 -3.19552630e-01 -1.30831051e+00 1.33750272e+00 -1.74861446e-01 3.61674130e-01 -1.72336727e-01 -1.06976342e+00 9.88110781e-01 2.88429469e-01 3.62423003e-01 -1.03976464e+00 -3.26240323e-02 2.72770137e-01 -2.58135796e-01 -3.00201416e-01 2.62678862e-01 -3.89520451e-02 -1.08351409e-01 4.01613921e-01 4.94719356e-01 2.96756804e-01 1.80169016e-01 9.38115120e-02 4.62291032e-01 4.48524468e-02 3.91097814e-01 -3.02796721e-01 7.40044355e-01 4.63402644e-02 8.04535687e-01 9.09726918e-01 -5.91326058e-01 8.82160366e-01 1.27049536e-01 -3.43840033e-01 -1.50211430e+00 -1.39956665e+00 -2.33906046e-01 1.05525374e+00 4.27260339e-01 9.73642915e-02 -7.64609516e-01 -9.96134400e-01 8.66609346e-03 3.96917373e-01 -8.16244006e-01 -2.15126187e-01 -5.50606549e-01 -3.51986855e-01 5.80814719e-01 6.69650614e-01 8.95346403e-01 -5.31811535e-01 2.72806942e-01 -1.94645211e-01 -2.85166681e-01 -1.37719023e+00 -8.54965448e-01 -2.43483782e-01 -6.12152219e-01 -1.26194978e+00 -1.38811100e+00 -1.11466134e+00 9.19089615e-01 5.82135439e-01 1.05494964e+00 -2.50424951e-01 -2.55260225e-02 8.77558172e-01 -3.83916616e-01 -1.61311015e-01 -4.96578872e-01 -1.58427328e-01 5.64910889e-01 4.42646205e-01 7.34986484e-01 -4.60723519e-01 -2.76925474e-01 7.61911452e-01 -5.47446787e-01 -1.25248367e-02 3.32802117e-01 9.28476572e-01 5.23829818e-01 -8.86230767e-02 4.80691105e-01 -1.10662925e+00 1.09960824e-01 -4.46445614e-01 -5.82900405e-01 5.90073705e-01 -4.24589574e-01 -2.86211014e-01 7.18676090e-01 -7.14053214e-01 -1.26807678e+00 3.62077236e-01 1.99403971e-01 -6.06280982e-01 -6.42083168e-01 -2.77781934e-01 -6.57825470e-01 -2.07915287e-02 9.07549143e-01 3.87354672e-01 4.38055806e-02 -4.41257179e-01 2.91937023e-01 7.71198452e-01 9.85872149e-01 -6.58402145e-01 1.24905694e+00 6.36554062e-01 -4.17037964e-01 -8.44007850e-01 -4.30919886e-01 -7.74502397e-01 -1.18757188e+00 -1.96315140e-01 7.78661072e-01 -1.29309607e+00 -2.65435666e-01 7.60844707e-01 -8.76029670e-01 -3.23971540e-01 -2.87822694e-01 5.68903923e-01 -4.08621490e-01 4.96948749e-01 -3.01020831e-01 -4.53969955e-01 -6.73496872e-02 -9.46772873e-01 7.55121410e-01 4.76399779e-01 -6.17409647e-02 -1.08208585e+00 5.88136576e-02 3.91643614e-01 7.22754896e-02 2.67503113e-02 5.60998440e-01 -1.12378943e+00 -1.98318422e-01 -3.77499104e-01 -4.27376509e-01 5.82098722e-01 4.05967772e-01 -4.77763504e-01 -1.16571641e+00 -6.62034392e-01 -1.22215837e-01 -2.87928671e-01 5.54808319e-01 1.88257754e-01 7.79890060e-01 -1.83194488e-01 -4.42118555e-01 6.88320994e-01 1.29108727e+00 2.53032535e-01 5.76645672e-01 4.61290181e-01 9.71910298e-01 7.21541584e-01 4.11530793e-01 2.83739537e-01 4.47314948e-01 9.71208632e-01 -1.61611393e-01 -3.01521569e-01 -3.74188632e-01 -5.75076640e-01 4.45039243e-01 3.78397763e-01 -1.67503700e-01 6.07330166e-02 -9.37161565e-01 6.40673935e-01 -1.81894851e+00 -1.07989883e+00 -5.46729863e-02 2.64889193e+00 6.87071502e-01 -1.75746188e-01 6.23553813e-01 -2.52549082e-01 1.32974696e+00 -2.78599113e-01 -7.72459447e-01 3.43783736e-01 -4.51227427e-01 -1.48058280e-01 7.36903906e-01 4.39048678e-01 -1.27912092e+00 7.96358824e-01 5.32550097e+00 7.77956545e-01 -8.11733961e-01 2.90165424e-01 6.90738201e-01 1.77510381e-01 1.36525435e-02 -3.11319172e-01 -1.09430230e+00 6.32561564e-01 7.53954709e-01 -3.82475853e-01 3.50953847e-01 1.01137900e+00 -2.36440554e-01 3.49491447e-01 -1.31319380e+00 1.42067170e+00 3.21952671e-01 -1.07422423e+00 1.41293973e-01 7.73347691e-02 9.71122682e-01 -2.19416812e-01 1.26434654e-01 3.92100543e-01 4.94911641e-01 -7.16125309e-01 6.37732327e-01 2.02169761e-01 1.16390455e+00 -6.60659850e-01 7.14558661e-01 1.66878745e-01 -1.08887589e+00 -1.77706942e-01 -6.98221803e-01 3.11948925e-01 -1.98475108e-01 1.23970278e-01 -7.07267880e-01 6.34294152e-01 9.55522597e-01 1.08146548e+00 -8.00199270e-01 7.57399142e-01 -1.95042148e-01 3.80459309e-01 6.54084235e-02 6.46037221e-01 -3.17802817e-01 -4.27426308e-01 4.51676279e-01 1.05134928e+00 1.11866757e-01 -8.12369734e-02 1.55358180e-01 7.79200137e-01 -2.42805332e-01 3.03250481e-03 -8.37829292e-01 3.44801635e-01 5.14378369e-01 8.94211948e-01 -3.10862839e-01 -4.76521999e-01 -6.70053542e-01 1.42513609e+00 3.33094090e-01 7.12287128e-01 -8.20150018e-01 -9.36542451e-02 7.15044618e-01 2.95683946e-02 8.97123516e-02 1.27261549e-01 -1.24617927e-01 -1.62324023e+00 9.91228223e-02 -8.97818625e-01 7.20452368e-01 -5.12244105e-01 -1.86444545e+00 6.46699727e-01 1.04615048e-01 -1.70955074e+00 -2.54000813e-01 -7.28367805e-01 -4.29401487e-01 1.02508593e+00 -1.51480961e+00 -1.49663949e+00 -5.40013850e-01 1.28402376e+00 4.19411898e-01 -6.88768744e-01 7.58219719e-01 5.53592026e-01 -6.78174913e-01 1.16676676e+00 5.53835988e-01 7.50006735e-01 1.31354940e+00 -1.19329774e+00 3.08992356e-01 9.47273314e-01 7.02146767e-03 5.38937688e-01 2.54655987e-01 -4.73069668e-01 -1.05389106e+00 -1.21486485e+00 6.78592324e-01 -6.74736321e-01 3.05329531e-01 -4.66107845e-01 -9.60940480e-01 1.04934537e+00 1.34858638e-02 -2.79849079e-02 8.13441217e-01 4.28435355e-02 -8.14545631e-01 -2.67054141e-01 -1.28633165e+00 4.62742746e-01 1.01060545e+00 -7.60767519e-01 -5.73129237e-01 5.15604615e-01 2.69098520e-01 -2.72240639e-01 -8.64023566e-01 2.29871482e-01 4.72135276e-01 -8.92480195e-01 1.23151851e+00 -7.22061396e-01 1.29030198e-01 -3.27121675e-01 -2.79680014e-01 -1.26011014e+00 -4.12903607e-01 -1.13063984e-01 3.05341840e-01 1.70053542e+00 2.70731628e-01 -9.71858919e-01 6.84311211e-01 9.31464434e-01 2.96705484e-01 1.80527017e-01 -7.65612006e-01 -1.35459447e+00 4.08297449e-01 1.72497500e-02 6.65913105e-01 1.31500387e+00 -2.06711367e-01 3.83240968e-01 -6.56593502e-01 2.76458919e-01 9.31377351e-01 6.40919954e-02 1.13069642e+00 -1.31615293e+00 -3.15522134e-01 -1.96510777e-01 -3.19194049e-01 -1.07399678e+00 3.97675037e-01 -8.91200483e-01 -2.71086603e-01 -7.77025223e-01 6.14975750e-01 -5.98262668e-01 -1.77846879e-01 4.42751020e-01 -8.46010447e-02 5.00328600e-01 3.25259596e-01 7.99914777e-01 -3.98890764e-01 3.97557944e-01 9.85932648e-01 -3.23084772e-01 -2.24032626e-01 1.94519579e-01 -6.15301311e-01 5.77323854e-01 7.86097348e-01 -2.44148940e-01 -3.30707222e-01 -4.01987433e-01 -6.07838571e-01 -1.40704766e-01 6.18506491e-01 -1.13377750e+00 2.12666735e-01 -2.92555541e-02 7.40204215e-01 -3.98741104e-02 2.49609321e-01 -8.37670922e-01 2.28632182e-01 2.79893756e-01 -4.01119739e-01 -1.53112710e-01 9.61595923e-02 5.57546377e-01 -1.60232514e-01 -2.44059622e-01 1.23352087e+00 -1.61773562e-01 -1.22580016e+00 5.00045180e-01 2.04626784e-01 1.62054390e-01 1.08385611e+00 -4.98086274e-01 -3.99133652e-01 -3.75607252e-01 -7.58014262e-01 6.01354130e-02 9.03543174e-01 6.17989004e-01 2.99608946e-01 -1.59520745e+00 -9.21039879e-01 3.69265109e-01 6.06078506e-01 -3.45220625e-01 4.56904560e-01 5.80484569e-01 -3.24144065e-02 3.12040329e-01 -6.30218506e-01 -6.34382010e-01 -1.36224353e+00 8.13291967e-01 5.14804542e-01 -6.89442456e-02 -4.28230137e-01 7.60077655e-01 7.00201690e-01 -8.43452871e-01 1.66639388e-01 6.56963229e-01 -2.20357552e-01 -1.29145131e-01 7.18245149e-01 4.27713633e-01 -2.27213219e-01 -1.21703780e+00 -3.46840203e-01 7.43262589e-01 -2.98218936e-01 -2.62955576e-02 9.09386218e-01 -3.51572841e-01 2.11940035e-01 8.13263804e-02 1.50389588e+00 -1.28809199e-01 -1.25618458e+00 -7.04284191e-01 -8.68526846e-02 -6.09820962e-01 -6.44938886e-01 -7.45144367e-01 -9.61028039e-01 6.44722700e-01 1.12548769e+00 -1.57018393e-01 1.11345303e+00 2.13158429e-01 5.71933746e-01 1.32211030e-01 3.80743444e-01 -1.16729128e+00 1.38650000e-01 3.37654322e-01 6.94805920e-01 -1.68580461e+00 -3.46060067e-01 -3.42682719e-01 -9.07038510e-01 8.23406577e-01 8.97159755e-01 -4.36290056e-02 3.53239357e-01 -2.79181808e-01 1.79166511e-01 3.84820402e-01 5.45768887e-02 -4.29357082e-01 2.75685906e-01 1.23380184e+00 8.92851576e-02 2.49364346e-01 1.49654612e-01 7.02657700e-01 -5.88000901e-02 -1.35479331e-01 3.59950632e-01 5.85926354e-01 1.21567681e-01 -1.41399109e+00 -6.85079873e-01 9.81435850e-02 -5.84384687e-02 1.89536661e-01 -5.75359344e-01 1.00260985e+00 1.86689064e-01 8.44719648e-01 2.23189760e-02 -5.30001521e-01 5.06843150e-01 6.67410493e-02 5.06533504e-01 -5.25122702e-01 -1.41542479e-01 -3.00447822e-01 -1.77770987e-01 -2.07389459e-01 -4.72993463e-01 -7.45019197e-01 -7.26509988e-01 -3.40638310e-01 -3.79906595e-03 2.08089501e-01 1.78755373e-01 7.63124168e-01 4.06026542e-01 -7.56859556e-02 8.27821314e-01 -7.66665816e-01 -4.97874796e-01 -8.26687217e-01 -8.55135977e-01 1.06429124e+00 2.84465522e-01 -7.67763317e-01 -3.73111665e-01 5.72021842e-01]
[14.755678176879883, 1.0423527956008911]
e738d847-5f8a-4d2c-97d5-911a34a193f3
a-zero-shot-framework-for-sketch-based-image
1807.11724
null
http://arxiv.org/abs/1807.11724v1
http://arxiv.org/pdf/1807.11724v1.pdf
A Zero-Shot Framework for Sketch-based Image Retrieval
Sketch-based image retrieval (SBIR) is the task of retrieving images from a natural image database that correspond to a given hand-drawn sketch. Ideally, an SBIR model should learn to associate components in the sketch (say, feet, tail, etc.) with the corresponding components in the image having similar shape characteristics. However, current evaluation methods simply focus only on coarse-grained evaluation where the focus is on retrieving images which belong to the same class as the sketch but not necessarily having the same shape characteristics as in the sketch. As a result, existing methods simply learn to associate sketches with classes seen during training and hence fail to generalize to unseen classes. In this paper, we propose a new benchmark for zero-shot SBIR where the model is evaluated in novel classes that are not seen during training. We show through extensive experiments that existing models for SBIR that are trained in a discriminative setting learn only class specific mappings and fail to generalize to the proposed zero-shot setting. To circumvent this, we propose a generative approach for the SBIR task by proposing deep conditional generative models that take the sketch as an input and fill the missing information stochastically. Experiments on this new benchmark created from the "Sketchy" dataset, which is a large-scale database of sketch-photo pairs demonstrate that the performance of these generative models is significantly better than several state-of-the-art approaches in the proposed zero-shot framework of the coarse-grained SBIR task.
['Ashish Mishra', 'Anurag Mittal', 'Shiva Krishna Reddy', 'Sasi Kiran Yelamarthi']
2018-07-31
null
null
null
null
['sketch-based-image-retrieval']
['computer-vision']
[ 3.33815277e-01 -3.95851701e-01 -1.69950575e-01 -4.16062146e-01 -1.06842732e+00 -6.22755289e-01 1.00100732e+00 -2.84485728e-01 -4.70780768e-02 4.75321889e-01 -1.73436925e-02 1.89337566e-01 -3.13928187e-01 -9.83790696e-01 -9.38744843e-01 -7.41527557e-01 4.16169614e-01 9.46383357e-01 2.15077683e-01 -1.80610180e-01 2.59686053e-01 5.51213443e-01 -1.88744283e+00 4.62308556e-01 2.68988192e-01 8.94542277e-01 3.10380250e-01 5.86744189e-01 -2.54563600e-01 3.77828479e-01 -6.58733428e-01 -5.03144622e-01 3.64097744e-01 -3.85162562e-01 -5.10765851e-01 2.13738799e-01 1.10528708e+00 -6.02933288e-01 -5.52045882e-01 8.98252249e-01 4.10950184e-01 3.29376131e-01 9.88697410e-01 -1.35911691e+00 -1.13691437e+00 3.06634083e-02 -3.52898240e-01 -1.44858614e-01 3.50124478e-01 -3.33245695e-01 1.23011637e+00 -1.50380540e+00 1.00778377e+00 1.29425395e+00 2.87521511e-01 7.80745327e-01 -1.32347059e+00 -6.73557937e-01 1.04147434e-01 9.14048180e-02 -1.76262724e+00 -3.56820017e-01 9.85926330e-01 -1.61808744e-01 5.09256542e-01 2.63246596e-01 2.89156288e-01 1.24228525e+00 -3.28425795e-01 1.08477151e+00 8.02157044e-01 -3.88440877e-01 3.27390730e-01 2.56246746e-01 -2.84043346e-02 5.43040872e-01 5.38518094e-02 -1.51317976e-02 -6.50505900e-01 -2.96661884e-01 9.37044859e-01 6.41676009e-01 -7.74751278e-03 -9.08919096e-01 -1.19225109e+00 9.14799154e-01 4.43972468e-01 4.33764011e-01 -2.44774923e-01 2.66680747e-01 1.36474967e-01 3.26351225e-01 3.27488035e-01 6.79875165e-02 -3.88276167e-02 3.28356415e-01 -1.13330841e+00 4.03695017e-01 7.31512666e-01 1.39891553e+00 1.04115856e+00 -2.84800559e-01 -4.69029456e-01 1.17496884e+00 2.06880160e-02 7.87052095e-01 2.60887891e-01 -6.84035540e-01 3.07089150e-01 4.74952132e-01 1.87661871e-01 -8.83066654e-01 5.54426789e-01 -2.05276489e-01 -7.99503028e-01 1.75576121e-01 2.84573603e-02 6.64326012e-01 -1.21966588e+00 1.75882256e+00 1.90167800e-02 2.63650954e-01 5.00859991e-02 8.87750685e-01 1.14839339e+00 7.04718173e-01 -1.27156004e-01 2.51542270e-01 1.20504057e+00 -9.98071015e-01 -3.43254238e-01 -3.34639102e-02 -1.79732129e-01 -8.68156672e-01 1.39816391e+00 2.47650295e-01 -9.25991416e-01 -7.51668870e-01 -1.13523602e+00 -8.88018310e-02 -6.29142225e-01 3.80767941e-01 2.12233767e-01 5.11011720e-01 -1.00809240e+00 5.20618439e-01 -2.67329663e-01 -6.12502515e-01 3.55140775e-01 -3.45495790e-02 -5.19306898e-01 -5.92222095e-01 -8.19852233e-01 4.95780498e-01 6.69369474e-02 -2.69256353e-01 -1.35995531e+00 -5.92616081e-01 -6.90612137e-01 3.22375745e-01 3.77929986e-01 -7.35274136e-01 1.00097454e+00 -9.72227812e-01 -1.15430784e+00 1.01600122e+00 -3.26764196e-01 3.54216322e-02 4.42762554e-01 -1.95978489e-03 -2.30401099e-01 3.04768443e-01 1.83879361e-01 8.72102857e-01 1.35038602e+00 -1.73791015e+00 -4.53185737e-01 -3.40425998e-01 1.98414147e-01 -3.13478336e-02 -3.26482564e-01 -1.44347042e-01 -7.23938644e-01 -8.40388298e-01 1.64380297e-02 -8.78097832e-01 2.18771771e-01 2.96718627e-01 -8.45964998e-02 -3.59179705e-01 1.23214293e+00 -2.59283721e-01 9.18611825e-01 -2.18491507e+00 1.30079731e-01 2.04656079e-01 -7.45743886e-02 3.01811993e-01 -4.93806094e-01 7.69734681e-01 -5.99701852e-02 6.77723736e-02 -1.90478727e-01 -5.61204612e-01 1.72196269e-01 5.46676576e-01 -7.86956310e-01 7.51619712e-02 2.24087462e-01 1.04301667e+00 -1.06831872e+00 -4.63766128e-01 2.75615364e-01 5.62264621e-01 -2.94100761e-01 3.87098670e-01 -1.42528281e-01 1.41005591e-01 -2.99618512e-01 8.09110403e-01 7.45422244e-01 -3.00628185e-01 2.52769217e-02 -1.62817642e-01 4.74922538e-01 -2.42372304e-01 -1.16244507e+00 1.99772072e+00 -5.75185776e-01 3.76306802e-01 -2.43522525e-01 -9.87748802e-01 1.12593126e+00 3.30098599e-01 2.19671085e-01 -5.98042309e-01 -4.19476956e-01 1.60290137e-01 -5.03589690e-01 -1.32348537e-01 4.84295785e-01 -4.10867155e-01 -8.57332870e-02 5.99223673e-01 4.26291853e-01 -2.44622037e-01 1.08118579e-01 4.14596766e-01 8.60329151e-01 5.95375970e-02 1.00760236e-01 -1.12766251e-01 4.37973440e-01 -3.82458776e-01 1.51098490e-01 1.09554815e+00 2.45651096e-01 1.05078995e+00 1.75415352e-02 -5.80260694e-01 -1.39960635e+00 -1.66249943e+00 -7.47655407e-02 1.09688938e+00 3.42711091e-01 -4.37336296e-01 -2.72118479e-01 -6.68159246e-01 5.55246435e-02 5.18332183e-01 -7.57715583e-01 -7.91528299e-02 -1.34654060e-01 -1.42802179e-01 3.00095320e-01 5.14396906e-01 3.30407649e-01 -1.24022341e+00 -3.77084911e-01 2.83384733e-02 -3.46941990e-03 -8.75606298e-01 -6.12517953e-01 -3.45822155e-01 -6.89325988e-01 -9.18066561e-01 -1.30333889e+00 -1.05356932e+00 9.84291732e-01 6.61299050e-01 1.36632538e+00 2.77583718e-01 -5.84764421e-01 8.38867486e-01 -3.72500837e-01 -2.26572156e-01 -1.68231949e-01 -2.36338526e-01 -2.48516262e-01 3.66644561e-01 2.27713376e-01 -5.91970980e-01 -7.57497370e-01 5.13530552e-01 -1.32344615e+00 -1.29173279e-01 7.30783105e-01 1.27410340e+00 7.59048879e-01 -7.35532045e-02 6.84138417e-01 -7.97696590e-01 4.69360590e-01 -4.19555157e-01 -2.66482562e-01 7.71061003e-01 -4.18592453e-01 1.56643435e-01 5.26064456e-01 -6.79144681e-01 -1.01099825e+00 1.04935877e-01 8.81835446e-02 -1.02949870e+00 -2.77323276e-01 1.70534611e-01 -1.81807593e-01 4.46818173e-02 3.25206995e-01 6.15895748e-01 -1.10890798e-01 -5.94706595e-01 5.19695163e-01 5.18456340e-01 4.88201916e-01 -9.19375360e-01 9.26196754e-01 6.76720321e-01 3.97314318e-02 -9.01686728e-01 -6.52427018e-01 -7.86041439e-01 -5.92566848e-01 -1.82998031e-02 5.29219866e-01 -8.64219189e-01 -2.38724485e-01 1.51875854e-01 -1.05220079e+00 9.88419876e-02 -4.15924877e-01 4.17444967e-02 -7.54911780e-01 4.57156569e-01 -2.47478992e-01 -9.49658096e-01 -5.11441588e-01 -9.38479602e-01 1.73403549e+00 1.15881488e-01 1.29202917e-01 -7.44852304e-01 1.43888071e-01 -1.62517037e-02 3.67454946e-01 -8.95023718e-02 1.14145565e+00 -5.72134018e-01 -9.00974929e-01 -4.03960705e-01 -3.93716246e-01 4.00229275e-01 2.52597541e-01 2.36643590e-02 -1.02844715e+00 -4.32569236e-01 -4.09607828e-01 -5.21569610e-01 1.00456810e+00 1.12402644e-02 1.16577566e+00 -1.63761467e-01 -3.22842121e-01 3.20196599e-01 1.73516881e+00 2.15498164e-01 8.23364139e-01 -2.94272274e-01 4.02751029e-01 4.60427999e-01 7.48586059e-01 3.80704403e-01 2.43753567e-02 8.71691048e-01 2.86389410e-01 1.49432393e-02 -3.09862792e-01 -7.59664476e-01 -1.22998981e-02 4.61497873e-01 5.21918945e-02 -2.76706874e-01 -5.08684397e-01 9.57968414e-01 -1.93833792e+00 -1.22943509e+00 4.11580831e-01 2.55368495e+00 6.11388564e-01 -2.88602114e-01 -4.36796667e-03 -9.24226865e-02 7.25729585e-01 1.63498059e-01 -3.42367113e-01 -1.12508461e-01 -6.10600151e-02 6.17509544e-01 -1.30514458e-01 3.12634766e-01 -1.02249026e+00 1.04237700e+00 5.80731153e+00 1.06469357e+00 -8.91686559e-01 1.14901081e-01 3.76859128e-01 -1.10482145e-02 -5.18133640e-01 1.27902046e-01 -7.34337986e-01 2.70525545e-01 3.15068454e-01 -1.26009777e-01 6.97443664e-01 9.84891355e-01 -5.64915657e-01 1.00753322e-01 -1.49361455e+00 1.33682513e+00 5.27925789e-01 -1.26082790e+00 6.05527401e-01 -1.24501206e-01 9.65469778e-01 -4.10592347e-01 3.35566849e-01 4.77595508e-01 -2.03101113e-02 -1.03389168e+00 6.07521892e-01 8.76361191e-01 1.15548337e+00 -6.46963179e-01 3.93793851e-01 3.57721835e-01 -1.26792538e+00 1.39625594e-01 -8.44913304e-01 3.65701854e-01 -2.08900839e-01 1.12990752e-01 -8.58671486e-01 6.09395504e-01 6.18915558e-01 6.48253918e-01 -6.13758087e-01 9.49812710e-01 -6.42362311e-02 1.16140477e-01 -5.57667948e-02 -3.96947423e-03 1.58591360e-01 -2.23221302e-01 4.76967782e-01 9.81797278e-01 5.74616492e-01 -3.53889093e-02 7.06222951e-02 1.14749575e+00 -2.38410741e-01 -4.04976262e-03 -1.12042344e+00 -4.12486307e-02 4.50680673e-01 1.22257316e+00 -5.10990143e-01 -7.98951805e-01 -5.26215494e-01 1.28037786e+00 2.94483691e-01 5.87168396e-01 -5.44836164e-01 -4.06005710e-01 6.66442037e-01 6.48314506e-03 7.47369766e-01 3.61290276e-02 4.52645607e-02 -1.20449221e+00 1.72870666e-01 -5.66368043e-01 3.20725590e-01 -1.01615143e+00 -1.75862288e+00 5.48117280e-01 1.22307934e-01 -1.40616131e+00 -3.99050415e-01 -4.93722469e-01 -6.93894446e-01 9.82461154e-01 -1.37403536e+00 -1.51139045e+00 -4.96568173e-01 8.77215028e-01 9.02810216e-01 -3.56374681e-01 1.09249592e+00 3.82623017e-01 1.59833759e-01 6.48207486e-01 2.70062178e-01 1.45361468e-01 9.49760199e-01 -1.07891583e+00 4.07583684e-01 5.40575922e-01 7.82479703e-01 8.72581005e-01 3.64250362e-01 -6.12110078e-01 -1.47497177e+00 -1.19374907e+00 8.68085265e-01 -6.07863665e-01 3.69095773e-01 -6.16284907e-01 -8.27281713e-01 5.78541756e-01 -4.41257916e-02 3.59684467e-01 3.38443100e-01 -3.39948311e-02 -7.48095989e-01 -2.95189142e-01 -1.11614990e+00 4.13384050e-01 1.17857015e+00 -1.03103364e+00 -7.30095267e-01 2.67063797e-01 2.76806772e-01 4.64568622e-02 -7.47689128e-01 2.65089691e-01 9.88156438e-01 -7.15771139e-01 1.50759685e+00 -8.39849591e-01 5.81539750e-01 -2.46329680e-01 -4.81667221e-01 -1.14185333e+00 -2.74071068e-01 -2.87767313e-02 4.40738425e-02 1.21903360e+00 1.92356985e-02 -1.54345766e-01 8.01578760e-01 3.00472617e-01 2.59956211e-01 -5.12912750e-01 -8.34340036e-01 -1.14543545e+00 -8.26818198e-02 -6.59342948e-03 6.86879933e-01 7.17998981e-01 -5.56896389e-01 4.21135932e-01 -6.51922286e-01 -7.65919015e-02 8.19761097e-01 6.96417093e-01 1.04459202e+00 -1.34124434e+00 -2.81226158e-01 -3.26600641e-01 -4.90749836e-01 -9.41130161e-01 1.77888364e-01 -8.73418033e-01 1.61246866e-01 -1.58041966e+00 7.07558513e-01 -4.67147768e-01 -4.25514966e-01 3.39112431e-01 -2.58022159e-01 4.94601309e-01 4.09045368e-01 4.05425787e-01 -6.54081047e-01 6.65154278e-01 1.13616800e+00 -5.84195375e-01 2.85895705e-01 2.89260764e-02 -3.05434257e-01 2.26691738e-01 3.08719754e-01 -3.91415030e-01 -7.82716990e-01 -2.99769282e-01 2.23180279e-02 1.62651628e-01 8.54168236e-01 -6.99895084e-01 1.87467530e-01 -7.36319926e-03 3.97883296e-01 -8.10858011e-01 6.88800752e-01 -1.05649519e+00 4.46522593e-01 1.00741588e-01 -3.60507280e-01 -1.71155870e-01 -2.12040424e-01 8.31308722e-01 -3.81086677e-01 -3.62625271e-01 6.13477290e-01 -2.45466426e-01 -8.23742390e-01 6.28756404e-01 1.43089190e-01 -1.70521647e-01 8.84121299e-01 -1.66471288e-01 -2.33361751e-01 -6.40030801e-01 -7.41469502e-01 -2.74802595e-01 6.84501827e-01 6.70830905e-01 1.08293343e+00 -1.78898203e+00 -6.79043174e-01 3.68759871e-01 7.67448902e-01 -2.17103586e-01 3.88403773e-01 2.26722315e-01 -4.60939333e-02 4.81697291e-01 -3.97830486e-01 -6.15947783e-01 -1.18411553e+00 1.06926787e+00 3.77243571e-02 -1.37264267e-01 -6.94458485e-01 7.51373351e-01 7.55230606e-01 -5.22449017e-01 2.77611822e-01 -3.86266708e-02 1.83876514e-01 -1.25986621e-01 5.41088223e-01 4.40900065e-02 1.51150212e-01 -4.92706001e-01 -2.82670379e-01 7.21106231e-01 -4.13847752e-02 -1.89964518e-01 1.24198294e+00 1.55126616e-01 8.08718503e-02 5.53529859e-01 1.32537806e+00 -1.08038925e-01 -1.09457195e+00 -6.41577303e-01 -1.61809295e-01 -9.16455328e-01 -3.00871938e-01 -7.82765448e-01 -1.05392861e+00 1.05169797e+00 5.67196906e-01 -1.80630460e-01 9.42373216e-01 3.02407503e-01 6.55859530e-01 5.97604930e-01 6.56394243e-01 -9.29311275e-01 4.89237309e-01 2.06270665e-01 1.28375804e+00 -1.25659931e+00 -2.57352412e-01 -1.18182644e-01 -5.19659817e-01 1.18928099e+00 3.52777392e-01 -4.52793866e-01 5.83899975e-01 -1.46779746e-01 -2.97578126e-01 -2.41858125e-01 -7.89177716e-01 -2.09744647e-01 5.77376902e-01 5.53698897e-01 1.26062185e-01 1.23465873e-01 -6.83367625e-02 3.94286722e-01 3.40136230e-01 2.95806117e-02 6.51182681e-02 9.72428381e-01 -2.09708437e-01 -1.40918863e+00 -2.67003357e-01 4.31530178e-01 1.38851494e-01 -1.15405180e-01 -6.83064401e-01 7.02404559e-01 -1.90383479e-01 5.97458184e-01 1.22236870e-01 -2.33617157e-01 2.92483151e-01 2.37229764e-01 8.07148099e-01 -7.58428752e-01 -2.89019868e-02 -3.41485180e-02 -2.59778589e-01 -4.01262462e-01 -3.84074330e-01 -4.90628779e-01 -6.47379458e-01 -1.89410280e-02 5.32162885e-05 2.67336927e-02 6.77261949e-01 6.80379450e-01 3.39201659e-01 1.12075679e-01 6.26062572e-01 -9.45018053e-01 -6.76653028e-01 -8.74945462e-01 -8.59331667e-01 8.06275368e-01 2.34772310e-01 -9.27315831e-01 -1.16295166e-01 6.83531463e-02]
[11.627346992492676, 0.6364580392837524]
e11c2bbf-886a-408a-9032-ac2885511e99
understanding-parameter-sharing-in
2306.0938
null
https://arxiv.org/abs/2306.09380v1
https://arxiv.org/pdf/2306.09380v1.pdf
Understanding Parameter Sharing in Transformers
Parameter sharing has proven to be a parameter-efficient approach. Previous work on Transformers has focused on sharing parameters in different layers, which can improve the performance of models with limited parameters by increasing model depth. In this paper, we study why this approach works from two perspectives. First, increasing model depth makes the model more complex, and we hypothesize that the reason is related to model complexity (referring to FLOPs). Secondly, since each shared parameter will participate in the network computation several times in forward propagation, its corresponding gradient will have a different range of values from the original model, which will affect the model convergence. Based on this, we hypothesize that training convergence may also be one of the reasons. Through further analysis, we show that the success of this approach can be largely attributed to better convergence, with only a small part due to the increased model complexity. Inspired by this, we tune the training hyperparameters related to model convergence in a targeted manner. Experiments on 8 machine translation tasks show that our model achieves competitive performance with only half the model complexity of parameter sharing models.
['Jingbo Zhu', 'Tong Xiao', 'Xiaohui Wang', 'Zhexi Zhang', 'Mingxuan Wang', 'Ye Lin']
2023-06-15
null
null
null
null
['machine-translation']
['natural-language-processing']
[-2.17146724e-02 3.01781535e-01 -4.59749788e-01 -2.33163863e-01 -3.01256448e-01 -4.26829398e-01 3.10004920e-01 -3.44257080e-03 -6.07059598e-01 6.15807593e-01 -3.42497528e-02 -4.34600562e-01 -7.81972036e-02 -6.18594706e-01 -9.17922199e-01 -5.46271741e-01 1.11774586e-01 7.08880603e-01 4.39220101e-01 -3.48879904e-01 3.28297853e-01 5.43464720e-01 -1.22468674e+00 1.84954718e-01 7.10332930e-01 5.84227681e-01 3.65994632e-01 4.13712054e-01 -2.31137782e-01 5.86676002e-01 -6.94976330e-01 -6.31955147e-01 1.34857789e-01 -4.60361600e-01 -9.66193438e-01 -2.98470974e-01 3.88565585e-02 -2.69079596e-01 1.65965810e-01 9.34609413e-01 3.33681881e-01 -1.14178829e-01 3.94228637e-01 -1.15325165e+00 8.87549520e-02 1.27739406e+00 -2.22843096e-01 2.55923681e-02 -2.69633383e-01 -1.42270520e-01 1.03461158e+00 -7.43574321e-01 5.80932558e-01 1.18670905e+00 6.36545300e-01 5.55278540e-01 -1.31064749e+00 -5.94706059e-01 6.02560997e-01 1.92666233e-01 -1.38860774e+00 -3.00645441e-01 7.30052114e-01 -6.40580580e-02 9.64602113e-01 2.46944696e-01 8.95188272e-01 8.56352925e-01 1.63555056e-01 8.39011550e-01 9.30550694e-01 -5.39232731e-01 2.03247637e-01 5.71651220e-01 3.14125493e-02 5.09640872e-01 3.39304358e-01 -1.73477143e-01 -4.11308408e-01 -1.47155896e-01 7.22890496e-01 -3.65187109e-01 -4.54319686e-01 -4.84856635e-01 -8.80977571e-01 8.96975756e-01 5.76912522e-01 7.51970589e-01 -2.03314841e-01 4.45996463e-01 2.59996861e-01 5.81872344e-01 3.94690633e-01 9.24742222e-01 -9.13953662e-01 -4.06915635e-01 -8.73338997e-01 1.83614090e-01 1.09922206e+00 5.57577550e-01 9.65918243e-01 -2.63698250e-01 2.09822267e-01 8.48211169e-01 4.67236519e-01 2.94593930e-01 5.93072832e-01 -7.11873353e-01 5.09698212e-01 8.03742647e-01 2.08376963e-02 -7.91094959e-01 -7.25549519e-01 -7.92303681e-01 -5.61055243e-01 -8.23096856e-02 7.66301036e-01 -1.10027179e-01 -6.45386994e-01 2.04114866e+00 1.94852322e-01 -1.26171291e-01 -1.78719133e-01 9.17780042e-01 2.14744598e-01 5.91814160e-01 -2.52958722e-02 -1.82565212e-01 1.25173306e+00 -1.16174531e+00 -5.67244232e-01 -4.08156008e-01 1.18042636e+00 -8.94675851e-01 1.21662784e+00 3.41890961e-01 -1.13236666e+00 -3.01882982e-01 -9.91419375e-01 2.01279059e-01 -2.57457793e-01 2.71252871e-01 6.07098639e-01 7.79995978e-01 -8.55355799e-01 9.31317985e-01 -1.03075683e+00 -3.73352557e-01 1.31479621e-01 6.01472616e-01 1.16836004e-01 2.54212648e-01 -1.31888604e+00 1.17147005e+00 4.50337142e-01 -2.69918935e-04 -2.41402686e-01 -9.92897272e-01 -2.93830633e-01 3.92463863e-01 3.52576703e-01 -8.57349157e-01 1.23204780e+00 -1.19373250e+00 -1.82468665e+00 3.70812833e-01 -3.83152694e-01 -6.80218935e-01 8.18767369e-01 -1.99161470e-01 1.42201602e-01 -2.52703484e-02 -5.73166370e-01 6.28047228e-01 8.05755675e-01 -1.27683342e+00 -3.76536667e-01 -2.55731702e-01 2.93548584e-01 2.32081547e-01 -6.80659831e-01 1.38457436e-02 -5.98639011e-01 -2.24062830e-01 2.19521731e-01 -1.29378390e+00 -2.70528585e-01 -2.71208078e-01 -1.13706790e-01 -1.04921304e-01 2.97231376e-01 -3.44632536e-01 1.66830003e+00 -1.85215104e+00 4.18182999e-01 3.92184287e-01 9.18091089e-02 3.79207075e-01 -4.40875143e-02 5.48161685e-01 3.17861848e-02 5.56013525e-01 -1.14652291e-02 -4.81626809e-01 -1.38225257e-01 1.85362265e-01 -2.57750332e-01 1.94187596e-01 1.20847635e-01 8.51396561e-01 -4.95817393e-01 -2.72496194e-01 -8.11133906e-02 5.71672499e-01 -6.65183783e-01 -2.10375860e-01 -2.79947281e-01 1.86920911e-01 -4.64131445e-01 2.44770125e-02 7.64230072e-01 -4.04100835e-01 4.42398787e-01 -1.44477725e-01 -8.17087218e-02 5.08637130e-01 -1.31340003e+00 1.28930342e+00 -6.89106762e-01 5.19433081e-01 3.02018188e-02 -7.03551412e-01 6.58332646e-01 2.99550027e-01 1.38603151e-01 -5.97713888e-01 1.76514983e-01 6.66990757e-01 3.88020694e-01 3.23283151e-02 5.55188596e-01 1.09256871e-01 5.70912838e-01 6.55619025e-01 -3.40606064e-01 -3.97542603e-02 1.37694553e-01 3.12879793e-02 7.19868898e-01 6.39155060e-02 -1.22416772e-01 -2.12316036e-01 5.33000886e-01 6.18435722e-03 5.46065986e-01 7.92213261e-01 -2.22478062e-02 2.95808375e-01 8.02297354e-01 -2.39486098e-01 -1.21089447e+00 -3.30687910e-01 2.01342534e-02 1.09806883e+00 5.86843714e-02 -7.65555263e-01 -1.04637671e+00 -5.05888879e-01 -2.88470536e-01 5.24555326e-01 -6.29042447e-01 -3.17635894e-01 -1.10697603e+00 -9.21175599e-01 5.27667403e-01 4.78347629e-01 4.69817966e-01 -7.46754885e-01 -9.71055269e-01 2.84012139e-01 -1.89143136e-01 -9.25603867e-01 -1.62699997e-01 3.06090802e-01 -1.36039913e+00 -8.32792997e-01 -7.56932199e-01 -4.98054266e-01 6.63771808e-01 2.60760397e-01 1.15171897e+00 6.44890726e-01 2.99487352e-01 -1.70371979e-01 -3.89537752e-01 -4.40904528e-01 -8.45753253e-01 9.35337782e-01 -2.41217494e-01 -2.58883029e-01 1.07411958e-01 -4.37273443e-01 -6.16995394e-01 6.15910709e-01 -8.56509626e-01 3.73888850e-01 5.20323992e-01 7.28996634e-01 6.60263896e-02 -1.63069076e-03 3.18714321e-01 -1.02220714e+00 6.95648193e-01 -2.10273668e-01 -6.94972217e-01 3.15922320e-01 -1.03966951e+00 3.39515656e-01 7.59730577e-01 -6.49187386e-01 -8.43039393e-01 -2.31893867e-01 1.13536827e-01 -3.91672313e-01 2.63124168e-01 6.41279876e-01 1.64298043e-01 -9.51252058e-02 4.87679958e-01 2.78472193e-02 1.77474409e-01 -5.89741528e-01 1.36377558e-01 2.61295229e-01 -5.67645252e-01 -4.64473307e-01 7.32739866e-01 3.29800993e-01 6.50963262e-02 -3.77043545e-01 -5.56313753e-01 -1.25652418e-01 -3.43252093e-01 1.26581252e-01 4.37508374e-01 -6.99951172e-01 -5.69194734e-01 5.59822142e-01 -1.27661467e+00 -6.30314410e-01 7.73511231e-02 7.35822380e-01 -3.49493623e-01 1.00512635e-02 -7.16309309e-01 -5.97642958e-01 -2.28298560e-01 -1.34888291e+00 5.80673873e-01 8.08295682e-02 -2.64151067e-01 -1.16578329e+00 -1.35919198e-01 2.02430993e-01 7.66553938e-01 -5.58437943e-01 1.18473625e+00 -7.28001773e-01 -6.27461910e-01 2.27007754e-02 -6.58364743e-02 2.62988776e-01 -2.09352523e-01 1.39609084e-01 -7.81709373e-01 -5.54293871e-01 3.07683740e-02 1.56131322e-02 8.20060492e-01 2.24967211e-01 9.75213230e-01 -1.43502176e-01 -4.83130664e-01 4.73198175e-01 1.42798781e+00 -1.62080020e-01 5.19203067e-01 6.33972883e-01 5.82099557e-01 5.58044016e-01 5.46287298e-01 3.48155677e-01 3.46309543e-01 1.08167350e+00 3.52894932e-01 -3.14146906e-01 -8.85685757e-02 -1.49628103e-01 3.87200832e-01 7.79116988e-01 -1.06418312e-01 -2.09184676e-01 -9.21246350e-01 2.89453924e-01 -2.02448916e+00 -5.36044836e-01 -1.43620133e-01 2.30315399e+00 7.94308543e-01 2.58489072e-01 1.90851942e-01 1.65773377e-01 5.77089489e-01 -1.07056230e-01 -3.84219319e-01 -6.73540592e-01 4.53110524e-02 -7.97574446e-02 8.13356340e-01 5.49916446e-01 -3.95602524e-01 1.07005537e+00 6.25547934e+00 7.76272476e-01 -1.54737508e+00 6.73136711e-02 5.46818614e-01 -2.66253650e-01 -3.64021808e-01 2.32042730e-01 -9.64684248e-01 5.95541179e-01 1.19829595e+00 -1.17372468e-01 5.17180145e-01 5.92526495e-01 2.11426198e-01 1.93890817e-02 -1.13526332e+00 4.81179297e-01 -1.61548838e-01 -1.19774234e+00 2.10363790e-01 3.26759309e-01 6.04871452e-01 1.98989451e-01 -1.43126532e-01 2.06625029e-01 -1.08027779e-01 -8.54127109e-01 9.02010322e-01 1.37665227e-01 -7.06149673e-04 -8.70275974e-01 9.10862923e-01 4.98965293e-01 -8.71594846e-01 -1.88463613e-01 -4.58405405e-01 -1.70021936e-01 1.15199782e-01 6.09723926e-01 -1.09274137e+00 4.51843411e-01 6.90986872e-01 1.32155582e-01 -6.60468757e-01 1.04468787e+00 -3.29518557e-01 9.68923688e-01 -5.55789590e-01 -1.86569571e-01 2.39696175e-01 -1.72881484e-01 5.18355966e-01 9.35996473e-01 1.97749048e-01 -5.63800275e-01 -2.11960465e-01 1.03068137e+00 -3.42860557e-02 2.98488557e-01 -1.50290132e-01 2.60185692e-02 6.91408515e-01 1.04940307e+00 -7.76454210e-01 -1.92986235e-01 -2.59685367e-01 7.72358179e-01 4.75896537e-01 2.53459454e-01 -1.07034862e+00 -8.19894895e-02 3.63409072e-01 2.79488564e-01 5.98687172e-01 -1.16060123e-01 -3.74621809e-01 -1.05042112e+00 1.23747967e-01 -8.94305348e-01 2.98418431e-03 -4.02116209e-01 -6.47884190e-01 5.99568546e-01 8.77973512e-02 -9.45387185e-01 -2.33474240e-01 -3.45318556e-01 -3.46482247e-01 9.47329462e-01 -1.85079551e+00 -9.35828447e-01 7.02681541e-02 2.19014302e-01 2.53694624e-01 1.54197514e-01 7.98862875e-01 4.08355355e-01 -7.34291911e-01 9.95698571e-01 -7.18815764e-03 -2.62774795e-01 7.87957907e-01 -9.99683976e-01 4.87608045e-01 6.41647279e-01 -2.73229051e-02 9.66385663e-01 7.64619470e-01 -3.31457466e-01 -1.28771329e+00 -6.01238012e-01 1.24412465e+00 -2.76325047e-01 6.67023063e-01 -3.90150070e-01 -1.13161314e+00 5.10721684e-01 -1.14038669e-01 -4.15177256e-01 5.24695992e-01 4.34644401e-01 -2.95073479e-01 2.26822067e-02 -8.98383617e-01 7.66476512e-01 8.49060953e-01 -1.50123775e-01 -1.84593096e-01 2.05910906e-01 7.13471293e-01 -4.68213201e-01 -9.27681625e-01 2.95746416e-01 4.75082636e-01 -9.41649199e-01 7.08117902e-01 -4.77224469e-01 3.91678363e-01 -9.71497223e-02 1.60937890e-01 -1.49805081e+00 -1.83544740e-01 -6.17078781e-01 -2.43665487e-01 1.04166949e+00 1.03061998e+00 -1.10530961e+00 8.29110324e-01 7.81183481e-01 2.37317339e-01 -1.19384527e+00 -8.62211049e-01 -8.77239287e-01 5.01616895e-01 -2.67846495e-01 9.96187747e-01 7.50737667e-01 -3.46335769e-01 3.50613117e-01 -2.75634110e-01 -6.94158673e-02 -3.61129232e-02 3.54145616e-02 7.61617780e-01 -1.23058748e+00 -5.57663202e-01 -5.47267199e-01 1.40846602e-03 -1.31456614e+00 2.46061422e-02 -6.85042977e-01 -2.54248768e-01 -1.23865151e+00 1.23965934e-01 -8.93396199e-01 -2.50586778e-01 4.11534846e-01 -2.70787388e-01 1.50806010e-01 7.07524955e-01 5.83338141e-01 -2.30948061e-01 2.38733545e-01 1.28488410e+00 3.03972244e-01 -4.47783798e-01 1.32519662e-01 -5.84349573e-01 6.61925137e-01 1.06927490e+00 -5.68507016e-01 -5.25171220e-01 -7.86176682e-01 9.27952051e-01 -3.25961560e-01 1.08290888e-01 -7.57945836e-01 1.52361661e-01 4.13817726e-02 -1.02831628e-02 -1.96701556e-01 3.52519721e-01 -1.05325246e+00 4.29076888e-02 8.14845085e-01 -2.80633330e-01 1.39835954e-01 3.04551125e-01 1.82209507e-01 -1.61177551e-04 -5.98058045e-01 7.16152489e-01 -5.99286705e-03 -2.15935737e-01 -2.22330183e-01 -4.11651164e-01 -3.74001592e-01 5.63993752e-01 -3.96138549e-01 -1.55735478e-01 -3.19174141e-01 -4.68125552e-01 1.63675219e-01 6.58002079e-01 4.89232987e-01 -1.51340544e-01 -9.12245810e-01 -5.11828244e-01 1.20763749e-01 -1.51701972e-01 -2.27345616e-01 -1.74182467e-02 1.23288906e+00 -3.86809468e-01 6.06853426e-01 2.43396431e-01 -5.87612629e-01 -1.37852037e+00 1.36752680e-01 5.65240562e-01 -6.17599428e-01 -3.66022766e-01 7.79027820e-01 -6.73819855e-02 -4.65240657e-01 2.85803318e-01 -5.11875212e-01 4.75107972e-03 1.24788076e-01 3.49683315e-01 4.43665534e-01 2.54332840e-01 -2.39579603e-01 -4.84735012e-01 7.36338258e-01 -4.90831077e-01 -6.98485076e-02 1.16872847e+00 -1.94242075e-01 -8.43049139e-02 3.32879484e-01 1.13185215e+00 6.06317520e-02 -1.12614250e+00 -2.28627324e-01 -1.64790168e-01 -2.30695456e-01 1.37884125e-01 -9.44529712e-01 -1.16033566e+00 7.50927925e-01 4.98469353e-01 2.27008507e-01 8.33946705e-01 -1.58844456e-01 6.81369960e-01 4.50124681e-01 5.25685489e-01 -1.25115097e+00 -1.97968632e-01 6.11875892e-01 4.93537158e-01 -1.12351120e+00 1.05964206e-01 -7.06203640e-01 -4.15022105e-01 1.23862684e+00 4.00853932e-01 1.94936663e-01 3.62357140e-01 5.27574830e-02 1.69558331e-01 2.99000721e-02 -9.06347573e-01 1.39063001e-01 8.50357395e-03 -9.46556926e-02 6.73820138e-01 -5.58933616e-02 -6.51859701e-01 3.55148524e-01 -4.60145593e-01 1.02463335e-01 4.99447227e-01 7.00926483e-01 -3.90392005e-01 -1.61115825e+00 -2.33673275e-01 7.72826746e-02 -4.68226492e-01 -1.27975926e-01 -3.86534870e-01 9.06453073e-01 -1.25076860e-01 9.70800459e-01 6.10946193e-02 -1.18636124e-01 2.74489999e-01 1.53069749e-01 6.83027565e-01 -3.48502785e-01 -1.02295530e+00 -1.06494516e-01 5.84595725e-02 -5.93425393e-01 -2.00513318e-01 -2.71178663e-01 -1.38990498e+00 -4.66739833e-01 -7.18447983e-01 3.54579061e-01 1.03826857e+00 8.41777861e-01 5.75413346e-01 5.64995885e-01 5.51640570e-01 -4.64042068e-01 -8.74484301e-01 -9.96777654e-01 -5.84457926e-02 2.07264423e-01 -2.16722954e-02 -4.82761413e-01 -7.38824964e-01 -4.24823225e-01]
[8.681144714355469, 3.637951374053955]
d7efe3c0-7409-468e-a0e0-13aa3f1251d2
segment-fusion-hierarchical-context-fusion
null
null
http://openaccess.thecvf.com//content/CVPR2022/html/Thyagharajan_Segment-Fusion_Hierarchical_Context_Fusion_for_Robust_3D_Semantic_Segmentation_CVPR_2022_paper.html
http://openaccess.thecvf.com//content/CVPR2022/papers/Thyagharajan_Segment-Fusion_Hierarchical_Context_Fusion_for_Robust_3D_Semantic_Segmentation_CVPR_2022_paper.pdf
Segment-Fusion: Hierarchical Context Fusion for Robust 3D Semantic Segmentation
3D semantic segmentation is a fundamental building block for several scene understanding applications such as autonomous driving, robotics and AR/VR. Several state-of-the-art semantic segmentation models suffer from the part-misclassification problem, wherein parts of the same object are labelled incorrectly. Previous methods have utilized hierarchical, iterative methods to fuse semantic and instance information, but they lack learnability in context fusion, and are computationally complex and heuristic driven. This paper presents Segment-Fusion, a novel attention-based method for hierarchical fusion of semantic and instance information to address the part misclassifications. The presented method includes a graph segmentation algorithm for grouping points into segments that pools point-wise features into segment-wise features, a learnable attention-based network to fuse these segments based on their semantic and instance features, and followed by a simple yet effective connected component labelling algorithm to convert segment features to instance labels. Segment-Fusion can be flexibly employed with any network architecture for semantic/instance segmentation. It improves the qualitative and quantitative performance of several semantic segmentation backbones by upto 5% on the ScanNet and S3DIS datasets.
['Sreenivas Subramoney', 'Om Ji Omer', 'Prashant Laddha', 'Benjamin Ummenhofer', 'Anirud Thyagharajan']
2022-01-01
null
null
null
cvpr-2022-1
['robust-3d-semantic-segmentation']
['computer-vision']
[ 6.20460033e-01 7.39569187e-01 -2.19534039e-01 -7.83896685e-01 -6.78806722e-01 -3.95774841e-01 3.94383669e-01 4.63080764e-01 -1.19015567e-01 3.26987892e-01 -3.15249950e-01 -2.79161662e-01 -3.14906329e-01 -8.35979283e-01 -7.74201930e-01 -4.98255402e-01 1.79254651e-01 8.76967072e-01 8.67856443e-01 -2.08674282e-01 4.10116345e-01 7.52074182e-01 -2.05299020e+00 1.41417548e-01 1.09373271e+00 1.31874907e+00 3.18921179e-01 3.35855663e-01 -7.28052199e-01 3.64353359e-01 -4.61353600e-01 -1.11397214e-01 3.75862807e-01 -2.78849244e-01 -1.25370419e+00 4.23590004e-01 5.49674094e-01 1.21135525e-01 1.21134005e-01 1.26840091e+00 2.03742832e-01 2.55349815e-01 5.72177291e-01 -1.70305789e+00 -2.49892443e-01 5.85956931e-01 -4.98806775e-01 -3.03898342e-02 2.48357922e-01 -1.55651957e-01 8.10489595e-01 -6.08561158e-01 4.57400709e-01 1.46233964e+00 7.00535059e-01 3.40513468e-01 -9.17056739e-01 -4.65403140e-01 4.64499086e-01 4.26569402e-01 -1.25480771e+00 2.13441215e-02 7.94075072e-01 -4.10432607e-01 1.05542493e+00 2.81417072e-01 9.33846533e-01 4.33858752e-01 -3.17673236e-02 9.18381333e-01 9.31964695e-01 -2.06370130e-01 2.77643442e-01 3.74829061e-02 5.16372621e-01 8.34122360e-01 2.53991604e-01 -2.29123071e-01 -7.26358443e-02 3.15131545e-01 7.02634990e-01 2.58092046e-01 2.05233142e-01 -7.25305796e-01 -9.30876195e-01 8.87837768e-01 8.85786414e-01 2.22799987e-01 -3.45001817e-01 1.98201671e-01 2.39781082e-01 -7.24099278e-02 5.90881169e-01 3.25277865e-01 -6.13362730e-01 3.84057134e-01 -1.11631691e+00 2.13977978e-01 7.13802516e-01 1.32668066e+00 1.30220759e+00 -9.99904647e-02 -1.77186713e-01 7.33352661e-01 4.58831042e-01 3.00720066e-01 3.34888488e-01 -1.16678023e+00 1.74128890e-01 1.29812920e+00 -3.51441681e-01 -8.85703862e-01 -7.68566430e-01 -3.40303093e-01 -4.83854115e-01 2.74662524e-01 5.28291147e-03 2.89208651e-01 -1.89790571e+00 1.32758355e+00 7.24029183e-01 3.09914887e-01 -1.05235010e-01 1.03267241e+00 1.17377412e+00 1.59167409e-01 3.42146546e-01 3.63826990e-01 1.43879735e+00 -1.21529412e+00 -5.05160272e-01 -4.11868483e-01 6.88304126e-01 -4.69386995e-01 6.01538479e-01 3.04031689e-02 -1.11391425e+00 -6.99735522e-01 -1.18142879e+00 -3.36011529e-01 -9.23968852e-01 -2.77557224e-01 6.30419195e-01 6.38406157e-01 -1.17836487e+00 6.19082987e-01 -8.25177491e-01 -5.34971952e-01 8.48600149e-01 6.03302658e-01 -3.88570458e-01 -2.37913638e-01 -8.61982107e-01 9.11351681e-01 9.31511223e-01 1.15089774e-01 -8.83589089e-01 -5.02191484e-01 -1.29066885e+00 9.03610960e-02 6.74780905e-01 -7.97088623e-01 1.13974345e+00 -1.01936388e+00 -1.14540505e+00 1.03261411e+00 -2.67776269e-02 -6.30083740e-01 3.33551705e-01 -7.85491914e-02 -1.76518500e-01 3.52744043e-01 4.39548552e-01 1.19244146e+00 9.17532206e-01 -1.65644574e+00 -9.99053121e-01 -6.43943787e-01 1.01772390e-01 4.67430055e-01 6.04087234e-01 -5.21887362e-01 -6.38727665e-01 -3.40453625e-01 9.43292797e-01 -7.34266639e-01 -5.80760360e-01 -2.43611664e-01 -6.16061687e-01 -4.02488381e-01 1.30555308e+00 -5.91783583e-01 6.98544919e-01 -1.84317148e+00 1.63683519e-01 4.66290295e-01 3.35170388e-01 2.62612104e-01 -3.16040330e-02 -5.57433106e-02 1.36114374e-01 1.86214373e-01 -8.58646035e-01 -3.65534782e-01 -4.42071725e-03 4.56403553e-01 2.32133105e-01 2.06402615e-01 3.48500013e-01 9.54990149e-01 -1.03744602e+00 -7.00954914e-01 7.12629080e-01 3.46358240e-01 -5.65294504e-01 1.07648239e-01 -4.27015245e-01 3.60620260e-01 -5.17264724e-01 8.76515210e-01 7.23472118e-01 -1.66915897e-02 -2.64071822e-01 -1.78937003e-01 3.57121862e-02 1.39504358e-01 -1.17359936e+00 1.99454665e+00 -7.99707621e-02 3.47815245e-01 2.65686959e-02 -1.47996044e+00 1.03775406e+00 -6.45089746e-02 6.66295409e-01 -6.81204200e-01 3.56034756e-01 3.11157763e-01 -3.78903747e-01 -4.46964622e-01 7.51861930e-01 -1.68483574e-02 -3.41075242e-01 -1.59280635e-02 3.00028920e-01 -7.04945922e-01 1.42714188e-01 8.11532587e-02 9.26083028e-01 3.97396088e-01 -4.46555875e-02 -2.51537949e-01 6.70471072e-01 5.72123170e-01 5.84244847e-01 7.35924780e-01 -2.69687384e-01 8.43169987e-01 2.07161203e-01 -2.73815572e-01 -8.69321167e-01 -9.34424639e-01 -3.42643335e-02 9.83039498e-01 8.08140755e-01 8.68265796e-03 -1.24639475e+00 -8.93344462e-01 2.21394479e-01 8.69641304e-01 -4.48612273e-01 -4.12147045e-01 -3.58537525e-01 -2.90505946e-01 2.74391264e-01 6.74407423e-01 8.14240158e-01 -1.21945274e+00 -6.80862427e-01 2.73905337e-01 -9.20605883e-02 -1.13372159e+00 -1.48004442e-02 5.04339457e-01 -8.78596663e-01 -1.27129960e+00 -4.38889503e-01 -1.00327384e+00 8.27734888e-01 5.47807634e-01 1.00985217e+00 3.32391351e-01 -4.07991499e-01 3.43988836e-01 -4.37812805e-01 -5.12374818e-01 -3.31686974e-01 3.05823982e-01 -4.42998439e-01 -9.02536958e-02 3.54117513e-01 -1.41466245e-01 -5.29704869e-01 4.60372925e-01 -8.86680424e-01 9.88762602e-02 6.04067802e-01 4.47298110e-01 1.03553510e+00 1.19708098e-01 5.03274322e-01 -1.16950059e+00 8.19433108e-02 -4.96888906e-01 -4.83376563e-01 1.47076517e-01 -5.96621990e-01 1.19969845e-02 7.61397136e-03 1.85396418e-01 -1.00240993e+00 3.71569872e-01 -2.02422857e-01 -5.45235217e-01 -6.56055152e-01 2.68308789e-01 -3.86389762e-01 -2.18824640e-01 3.08398396e-01 -1.04235880e-01 7.94148911e-03 -4.61146802e-01 7.86743224e-01 6.50400400e-01 6.88212693e-01 -2.79244155e-01 4.38820720e-01 5.59642315e-01 9.47381184e-02 -7.81389058e-01 -1.03827548e+00 -1.05829704e+00 -9.82814670e-01 -2.01456845e-01 1.34023881e+00 -8.03502440e-01 -3.09561312e-01 6.25530958e-01 -9.11818564e-01 -3.82658482e-01 -4.83685911e-01 9.69888791e-02 -8.80379617e-01 3.20881665e-01 -1.72731221e-01 -5.22956669e-01 -1.14990875e-01 -1.34765804e+00 1.60527134e+00 4.83952016e-01 8.05638060e-02 -8.72195244e-01 -4.80830699e-01 7.68792212e-01 8.43388028e-03 3.95998925e-01 9.12775457e-01 -8.55532765e-01 -6.64083362e-01 -1.92217231e-01 -6.24964893e-01 3.19244444e-01 -5.43110929e-02 -2.64210522e-01 -8.91007364e-01 -4.29778807e-02 -2.47437522e-01 -1.10100012e-03 1.04468977e+00 4.56361592e-01 1.25033307e+00 -6.23955764e-02 -7.07273066e-01 5.49236059e-01 1.42361629e+00 3.51505041e-01 5.11630893e-01 1.28517881e-01 1.24906659e+00 7.19689369e-01 7.41544008e-01 -1.00121267e-01 5.81977546e-01 3.86921287e-01 8.53812218e-01 -4.23246682e-01 -3.31656545e-01 -1.73555672e-01 -2.71091461e-01 3.61792356e-01 1.13486350e-01 -2.11511955e-01 -9.54733253e-01 7.77012110e-01 -2.01346803e+00 -5.70767760e-01 -3.72975945e-01 1.89051414e+00 2.07293063e-01 3.01406175e-01 1.15768984e-01 4.06915456e-01 1.03411674e+00 -1.84349164e-01 -7.10806429e-01 -4.15013194e-01 4.55020219e-02 2.33015090e-01 9.63051498e-01 4.29963857e-01 -1.21245182e+00 1.38898134e+00 5.55666542e+00 7.72930324e-01 -7.15984106e-01 1.52673051e-01 5.20825386e-01 5.05068064e-01 -2.34113961e-01 1.10777862e-01 -6.98081493e-01 1.14096701e-01 6.01884484e-01 3.46466899e-01 3.64218839e-02 9.35903788e-01 -8.08992162e-02 -4.72074687e-01 -8.84540319e-01 7.26755261e-01 2.13719189e-01 -1.07721567e+00 1.45051658e-01 -1.84133545e-01 8.21046114e-01 7.67209232e-02 -3.49191576e-01 3.27861547e-01 4.02474523e-01 -9.51380074e-01 9.36356664e-01 5.07410109e-01 4.57109421e-01 -8.79500329e-01 9.81602132e-01 2.30107009e-01 -1.35353422e+00 -7.69356489e-02 -2.61526287e-01 2.39466906e-01 2.62945026e-01 3.42189610e-01 -8.56511533e-01 7.82018065e-01 9.20066953e-01 6.92384183e-01 -6.18839025e-01 1.32076395e+00 -2.47123569e-01 2.13213220e-01 -4.28712279e-01 2.49963433e-01 6.45543933e-01 -2.85155386e-01 4.80797410e-01 9.63039100e-01 1.71323255e-01 1.08690113e-01 5.41506588e-01 8.87406409e-01 9.62381139e-02 -2.07209066e-01 -4.38551664e-01 1.40160546e-01 3.45401108e-01 1.19268370e+00 -1.52322114e+00 -6.68757200e-01 -2.03065977e-01 1.02534604e+00 8.83321241e-02 2.77954280e-01 -7.09725559e-01 -5.53502083e-01 4.72909331e-01 2.03309759e-01 5.47628939e-01 -1.24253266e-01 -7.13292658e-01 -4.78337377e-01 -1.64948314e-01 -2.11638272e-01 4.40955788e-01 -8.14869523e-01 -9.18685615e-01 4.94732380e-01 3.74730200e-01 -8.65432560e-01 1.52284484e-02 -3.76264781e-01 -3.59923750e-01 5.61173558e-01 -1.57859695e+00 -1.34250224e+00 -6.96518779e-01 6.02636516e-01 8.11401248e-01 1.98664561e-01 3.68245840e-01 1.55137837e-01 -4.56224024e-01 1.45541564e-01 -4.78413939e-01 -5.01747392e-02 -3.95489000e-02 -1.38920069e+00 4.79062408e-01 6.62102222e-01 -3.98256853e-02 3.08892373e-02 5.36128342e-01 -8.88988554e-01 -7.02425957e-01 -1.47786701e+00 7.03952432e-01 -3.55568141e-01 2.18568012e-01 -2.15740383e-01 -1.00079811e+00 5.91689527e-01 -1.40905127e-01 9.55651700e-02 1.75099522e-01 -3.00801575e-01 -7.43459165e-02 1.82763904e-01 -1.61194658e+00 2.73156971e-01 1.45057750e+00 -3.10075819e-01 -8.31602275e-01 3.03260863e-01 1.18604410e+00 -5.64936101e-01 -7.19961166e-01 7.38066196e-01 1.60491154e-01 -1.00167489e+00 1.01607835e+00 -3.92993957e-01 -6.38524294e-02 -6.17911398e-01 -7.48260692e-02 -1.13053393e+00 -2.88048893e-01 -1.49518445e-01 1.57680497e-01 1.05837750e+00 3.17950249e-01 -5.01139402e-01 9.22921062e-01 4.79670376e-01 -8.89442444e-01 -5.92208505e-01 -9.57578003e-01 -6.37503207e-01 -2.15655476e-01 -5.63740849e-01 9.11681056e-01 7.01438487e-01 -4.59233880e-01 1.55592471e-01 4.34642643e-01 3.19201767e-01 6.75608814e-01 1.70628905e-01 4.90440875e-01 -1.53389990e+00 4.50538844e-01 -7.98442125e-01 -1.01284397e+00 -7.38227010e-01 1.69562027e-01 -1.16570842e+00 4.06894118e-01 -2.25295138e+00 -2.22786859e-01 -6.84450388e-01 -2.48758271e-01 6.67525768e-01 -6.32308275e-02 4.26348329e-01 1.96041897e-01 -2.23287139e-02 -8.38579178e-01 4.00701404e-01 1.19394541e+00 -3.64483654e-01 -4.29600865e-01 1.02833673e-01 -4.84464675e-01 7.62232542e-01 8.00317705e-01 -5.26595891e-01 -5.54302692e-01 -1.24082513e-01 -1.95597380e-01 -7.31606632e-02 6.07630074e-01 -1.26107359e+00 2.06079498e-01 1.03349455e-01 9.41040739e-02 -1.08966792e+00 2.56080449e-01 -1.16432118e+00 1.66767895e-01 3.36806446e-01 -9.32362825e-02 -2.60325223e-01 2.32422039e-01 6.79732800e-01 -2.54535973e-01 -2.41952509e-01 7.61517704e-01 -4.72599715e-01 -1.34295034e+00 3.96284968e-01 -1.62966996e-01 4.38273549e-02 1.33011532e+00 -7.81953752e-01 -1.61315314e-02 2.28379041e-01 -1.01570630e+00 5.29577315e-01 4.01787102e-01 6.92435324e-01 6.17492020e-01 -1.16018999e+00 -3.62433583e-01 4.43572551e-01 2.03795895e-01 8.13786209e-01 4.46096778e-01 7.55031943e-01 -7.32917964e-01 4.51609284e-01 -2.91986108e-01 -1.10193634e+00 -1.01608109e+00 5.37714660e-01 4.41305965e-01 1.96456403e-01 -5.64218938e-01 9.21392500e-01 9.30331126e-02 -6.91153884e-01 1.33998036e-01 -6.84056163e-01 -3.04405004e-01 1.72991559e-01 -1.86864123e-01 5.56336761e-01 3.36373180e-01 -1.07851636e+00 -4.24065500e-01 9.04473424e-01 1.39411241e-01 2.08666578e-01 9.60501730e-01 -3.32541913e-01 -6.95344433e-03 3.15274000e-01 1.15205359e+00 -9.70150113e-01 -1.35652900e+00 -1.16515100e-01 1.43503010e-01 -8.49419311e-02 2.13892877e-01 -7.06336915e-01 -1.26180601e+00 8.11977983e-01 5.55333316e-01 4.27207857e-01 1.04153764e+00 4.12866950e-01 8.72142851e-01 7.80700594e-02 5.20818532e-01 -1.34422445e+00 -3.36667299e-01 4.05201107e-01 5.87622166e-01 -1.32072997e+00 -3.94280069e-02 -8.87346447e-01 -4.71407264e-01 7.11804032e-01 7.11478233e-01 -1.34366021e-01 7.07694054e-01 -2.49227554e-01 -1.38731292e-02 -6.58238649e-01 1.64729413e-02 -8.15002322e-01 4.86935854e-01 8.44780922e-01 -2.33145699e-01 2.25151330e-01 -1.17628030e-01 3.45329702e-01 -1.38426423e-01 -3.36867005e-01 1.11272931e-01 1.04192829e+00 -9.32331920e-01 -6.23411238e-01 -2.43821308e-01 6.41015232e-01 -1.63172632e-01 3.16157371e-01 -4.03638601e-01 9.36954081e-01 6.46162748e-01 1.00982583e+00 4.04831797e-01 -3.89188647e-01 4.56385881e-01 9.49788243e-02 2.84692615e-01 -8.13386619e-01 -5.77429891e-01 -6.07025772e-02 -3.31035107e-02 -8.43051612e-01 -7.67337501e-01 -6.47720456e-01 -1.88948941e+00 1.33215383e-01 -5.03411770e-01 -4.69428375e-02 9.78754342e-01 1.26208866e+00 3.79848033e-01 7.42746294e-01 4.21883225e-01 -1.21415913e+00 1.36110276e-01 -7.16766655e-01 -4.48085457e-01 5.29069483e-01 2.22913489e-01 -1.07959628e+00 -3.62042010e-01 -7.65263736e-02]
[8.16026496887207, -2.906137466430664]
df84bcb5-1674-4acd-b488-7e16912b187f
unified-smoke-and-fire-detection-in-an
2202.07954
null
https://arxiv.org/abs/2202.07954v1
https://arxiv.org/pdf/2202.07954v1.pdf
Unified smoke and fire detection in an evolutionary framework with self-supervised progressive data augment
Few researches have studied simultaneous detection of smoke and flame accompanying fires due to their different physical natures that lead to uncertain fluid patterns. In this study, we collect a large image data set to re-label them as a multi-label image classification problem so as to identify smoke and flame simultaneously. In order to solve the generalization ability of the detection model on account of the movable fluid objects with uncertain shapes like fire and smoke, and their not compactible natures as well as the complex backgrounds with high variations, we propose a data augment method by random image stitch to deploy resizing, deforming, position variation, and background altering so as to enlarge the view of the learner. Moreover, we propose a self-learning data augment method by using the class activation map to extract the highly trustable region as new data source of positive examples to further enhance the data augment. By the mutual reinforcement between the data augment and the detection model that are performed iteratively, both modules make progress in an evolutionary manner. Experiments show that the proposed method can effectively improve the generalization performance of the model for concurrent smoke and fire detection.
['helin sun', 'zhongyan lu', 'Hongyong Wang', 'Su Yang', 'Hang Zhang']
2022-02-16
null
null
null
null
['multi-label-image-classification', 'self-learning', 'fire-detection']
['computer-vision', 'natural-language-processing', 'time-series']
[ 5.52353680e-01 -5.36243677e-01 1.52802214e-01 -3.89593579e-02 1.76665280e-02 -7.06286311e-01 4.04839993e-01 -2.42836729e-01 -3.99855286e-01 4.13723111e-01 -2.64368266e-01 -5.93396090e-03 -4.32350524e-02 -8.71303082e-01 -4.93634939e-01 -1.12707496e+00 4.36784655e-01 3.23372453e-01 6.45707786e-01 -5.39173074e-02 3.00130069e-01 4.18999046e-01 -1.74549913e+00 4.18671966e-01 1.21358621e+00 7.25471616e-01 4.80159163e-01 5.45706809e-01 -4.76438701e-01 7.59209871e-01 -3.77157241e-01 5.46630807e-02 4.45843637e-01 -4.08579141e-01 -3.41753989e-01 4.24641162e-01 2.79158831e-01 -4.95198727e-01 -8.94833729e-02 1.33298957e+00 4.62322682e-01 1.82402924e-01 8.04550588e-01 -1.47754514e+00 -6.76413298e-01 4.49215800e-01 -1.01396656e+00 3.10428232e-01 -5.26499823e-02 3.02335918e-01 2.55229473e-01 -8.60482037e-01 2.62998909e-01 1.29744625e+00 6.89303458e-01 8.64846051e-01 -5.66974044e-01 -1.07006884e+00 5.97523749e-01 1.73700988e-01 -1.50867569e+00 -6.47786781e-02 1.09510756e+00 -3.28005046e-01 1.54984355e-01 4.92101222e-01 9.24596488e-01 9.62240398e-01 -1.69135496e-01 7.61485815e-01 1.50973308e+00 -3.09038073e-01 1.24779865e-01 6.40084326e-01 -5.47409765e-02 1.01156926e+00 3.09658527e-01 4.01027411e-01 1.05741166e-01 4.49607894e-03 8.27259243e-01 4.74290222e-01 -2.11716741e-01 8.93216357e-02 -8.91663551e-01 6.43850923e-01 5.39359331e-01 2.61207461e-01 -4.80057225e-02 -8.45213607e-02 9.89778191e-02 1.00107379e-01 5.02820075e-01 1.73391849e-01 -3.75231683e-01 5.49294591e-01 -6.03796959e-01 1.83897346e-01 5.03186762e-01 7.55864620e-01 8.39725077e-01 1.61630824e-01 -1.90118223e-01 6.95765376e-01 4.98241246e-01 8.51871669e-01 6.52745843e-01 -4.11785334e-01 3.84652585e-01 9.59890544e-01 -3.36141512e-02 -1.07089746e+00 -4.56429273e-01 -4.01926666e-01 -1.06448567e+00 6.03478074e-01 2.40952045e-01 -3.60350549e-01 -1.05931318e+00 1.52532935e+00 8.30947161e-01 8.05178583e-01 1.19395867e-01 8.47799063e-01 1.00277209e+00 7.66026020e-01 2.04205066e-01 -5.18701971e-01 1.17613864e+00 -1.15557742e+00 -6.96366370e-01 -3.28443527e-01 2.11204082e-01 -5.47950149e-01 8.07564318e-01 2.46669263e-01 -5.55441916e-01 -7.39103496e-01 -9.24179733e-01 5.36299944e-01 -4.76483941e-01 -4.69992263e-03 4.00564104e-01 6.50338888e-01 -4.79868919e-01 4.76697832e-01 -3.39036226e-01 -7.46368915e-02 5.90417802e-01 1.71690017e-01 1.54390186e-01 -1.32584617e-01 -1.15890622e+00 6.80068076e-01 7.31690943e-01 3.61450195e-01 -8.48181248e-01 -6.57963157e-01 -3.02672207e-01 -4.51917201e-01 6.11775935e-01 -4.65169549e-01 7.13236749e-01 -1.17267859e+00 -1.13604712e+00 5.49269915e-01 3.41536462e-01 3.40931028e-01 8.91706586e-01 8.49866122e-02 -6.44201458e-01 -2.40016114e-02 -2.98481762e-01 5.17270684e-01 1.19193876e+00 -1.75841284e+00 -1.07991779e+00 -4.14566755e-01 -1.66403651e-01 6.10235333e-01 -5.20672023e-01 -8.35344568e-02 3.16310078e-02 -6.08873188e-01 1.91627607e-01 -1.02868211e+00 -4.13030684e-01 2.14400575e-01 -7.90144131e-02 -1.22164540e-01 1.42308795e+00 -4.14387822e-01 1.28408945e+00 -2.12380672e+00 9.65665504e-02 2.97455907e-01 1.84659988e-01 4.38440144e-01 -1.66808486e-01 -2.42838740e-01 3.84630598e-02 4.25721914e-01 -5.14347732e-01 1.94514260e-01 -5.95567286e-01 3.96968305e-01 -1.11951148e-02 2.76556581e-01 3.04537594e-01 6.35258079e-01 -1.09277391e+00 -9.27275896e-01 3.15187663e-01 2.79388577e-01 -4.33607772e-02 2.76019692e-01 -4.28561121e-01 5.95254600e-01 -7.49088168e-01 6.18596077e-01 1.17063856e+00 2.63895094e-02 -2.69505829e-01 -3.73009086e-01 -1.92442536e-01 -9.08133328e-01 -1.65244198e+00 1.11853695e+00 -1.41945943e-01 -1.96370244e-01 2.87256807e-01 -4.58781600e-01 1.08989096e+00 2.30772480e-01 4.54038262e-01 -2.86625803e-01 3.29369217e-01 8.75612125e-02 -2.08286881e-01 -1.09083271e+00 8.14125836e-02 -4.91566509e-01 4.17307645e-01 5.58899641e-01 -4.84020621e-01 -3.16468686e-01 -1.63706064e-01 -5.92284426e-02 8.19633245e-01 -5.58835641e-02 -8.71460363e-02 -2.15343341e-01 6.49951518e-01 -1.04263961e-01 7.19964147e-01 4.38939720e-01 -2.78177530e-01 5.37731290e-01 -3.65625054e-01 -4.76146907e-01 -8.68497133e-01 -8.55556011e-01 -1.23222381e-01 9.34223175e-01 7.53924966e-01 2.76181012e-01 -6.05796039e-01 -1.01051271e+00 9.82333645e-02 3.26739728e-01 -6.96497858e-01 -4.57693040e-01 -6.21416748e-01 -1.06187057e+00 3.77573729e-01 6.23331368e-01 9.78363097e-01 -1.29526341e+00 -5.59937298e-01 8.57283846e-02 -8.42067897e-02 -7.30023742e-01 -4.29367781e-01 2.12140083e-01 -7.28805661e-01 -1.20961261e+00 -7.27910221e-01 -1.06720531e+00 7.16950059e-01 6.28517628e-01 5.29075503e-01 7.32742429e-01 -3.70761812e-01 1.55694470e-01 -5.62016070e-01 -5.68094075e-01 -7.03281999e-01 -3.49175662e-01 -7.65993819e-02 2.10945442e-01 -1.59019772e-02 -5.56684613e-01 -4.88202989e-01 3.53768319e-01 -1.42534459e+00 2.50822723e-01 4.67361927e-01 5.97673118e-01 1.69653639e-01 7.34004080e-01 6.68951809e-01 -8.59342933e-01 6.38871014e-01 -5.03798068e-01 -4.85609621e-01 6.47575259e-01 -8.09462845e-01 -1.22004412e-01 5.03578544e-01 -9.17677879e-01 -1.51138198e+00 2.96072900e-01 2.38448367e-01 -7.70665348e-01 -3.57613206e-01 -3.77992205e-02 -9.85730961e-02 -3.23634535e-01 7.06794858e-01 3.15492868e-01 1.23396419e-01 -1.77329034e-01 4.69082654e-01 7.27645874e-01 4.11013693e-01 -3.57610703e-01 1.44592464e+00 4.67670977e-01 6.99036345e-02 -4.74611849e-01 -7.12256074e-01 -5.27143061e-01 -7.91867077e-01 -7.90121138e-01 1.01465416e+00 -7.09102571e-01 -7.09000111e-01 9.41677153e-01 -1.14762461e+00 -5.44984974e-02 -1.35482073e-01 1.96543738e-01 1.42612055e-01 6.20346844e-01 -5.66142976e-01 -1.26686275e+00 -4.70154464e-01 -9.06523645e-01 7.25452542e-01 6.58701122e-01 5.82442820e-01 -9.50733900e-01 -2.12709401e-02 1.46709040e-01 3.53598297e-01 5.32887697e-01 8.50756347e-01 -3.29671383e-01 -5.62257349e-01 2.62342393e-01 -2.68062472e-01 3.48555118e-01 4.33754355e-01 2.85939068e-01 -1.07154071e+00 -1.41995609e-01 2.83843994e-01 -2.66536891e-01 1.13373685e+00 -1.48633868e-02 9.52616215e-01 -4.33637440e-01 -4.84134734e-01 4.88032430e-01 1.52407253e+00 6.79419160e-01 2.70162433e-01 3.07938963e-01 1.04703999e+00 6.60003603e-01 6.69339180e-01 5.59604347e-01 -8.85189325e-02 2.78807022e-02 8.81846905e-01 -3.09257030e-01 -1.57797158e-01 -3.08970772e-02 1.20810300e-01 5.74826896e-01 -2.39245355e-01 -3.31749171e-01 -5.59049308e-01 1.01981647e-01 -1.90321720e+00 -1.22488081e+00 -3.28183830e-01 1.72351122e+00 7.83745170e-01 -1.11943819e-01 3.44384857e-03 1.28545687e-01 1.28823841e+00 8.50830227e-02 -7.40653574e-01 2.67536223e-01 -2.50886530e-01 -2.76626676e-01 3.45312029e-01 3.37881804e-01 -1.32706416e+00 7.01663971e-01 5.89890385e+00 9.29386079e-01 -1.14679253e+00 -5.34884632e-02 5.75999141e-01 1.78169236e-01 -6.30770743e-01 -2.87381560e-01 -7.06592441e-01 5.96886754e-01 -1.79492291e-02 1.96832523e-01 8.79640162e-01 5.02505779e-01 -2.08476231e-01 3.70442569e-02 -5.04270673e-01 7.96765327e-01 1.18298911e-01 -9.45141315e-01 2.97489502e-02 -2.99206436e-01 1.00372553e+00 -1.98954061e-01 1.11230783e-01 9.28786099e-02 3.58743191e-01 -6.51678920e-01 7.38043666e-01 5.97010493e-01 5.54712117e-01 -4.40781474e-01 4.79956388e-01 6.93348289e-01 -1.55449426e+00 -5.77100992e-01 -3.87713850e-01 1.29416719e-01 -1.03069760e-01 6.45199597e-01 -7.83746541e-01 6.19789898e-01 4.29228485e-01 5.54872334e-01 -8.00169587e-01 9.84637618e-01 -7.69188330e-02 2.46946499e-01 -5.44463992e-01 -1.18441552e-01 1.01436667e-01 -4.29980367e-01 4.72077429e-01 1.07368898e+00 3.12528670e-01 4.01721895e-01 6.77172303e-01 8.71521771e-01 2.95411080e-01 5.98778874e-02 -6.19732797e-01 2.44990811e-01 5.30591488e-01 1.46894705e+00 -1.03559613e+00 -4.00481611e-01 -1.42154366e-01 8.08665752e-01 1.48396820e-01 2.80645639e-01 -1.06569827e+00 1.88880950e-01 -1.17155045e-01 -3.42093743e-02 -6.12573475e-02 1.51999801e-01 -1.96165949e-01 -9.24791574e-01 -4.92841415e-02 -6.88090146e-01 5.91868758e-01 -7.78079629e-01 -1.51505613e+00 7.35088289e-01 -9.21089426e-02 -1.27820385e+00 2.85164624e-01 -3.28873485e-01 -8.53412807e-01 5.65309763e-01 -1.64143193e+00 -1.36207294e+00 -9.15815711e-01 8.39201033e-01 5.04040897e-01 -1.25948936e-01 6.24660254e-01 3.13253313e-01 -6.55855596e-01 1.26299933e-01 4.75531779e-02 -1.69497862e-01 7.14501083e-01 -9.93056178e-01 -2.58850008e-01 6.91807508e-01 -8.37581083e-02 -1.25205806e-02 4.01674092e-01 -1.01528013e+00 -9.53477383e-01 -1.54382253e+00 -8.61833319e-02 -2.63482004e-01 2.22848639e-01 -1.47949666e-01 -1.06592059e+00 2.32063875e-01 -1.21144615e-02 1.67128012e-01 1.78195447e-01 -3.55848521e-01 -2.63384908e-01 -2.82767773e-01 -1.48144710e+00 4.31644887e-01 1.18243217e+00 -9.77970958e-02 -4.92328316e-01 6.08320713e-01 7.29505181e-01 -3.53971303e-01 -3.49239379e-01 7.88208544e-01 3.94558430e-01 -7.60918021e-01 8.53995383e-01 -3.05305779e-01 3.57346863e-01 -6.17738545e-01 1.46424055e-01 -1.13767099e+00 -7.20485687e-01 -1.43826604e-01 1.36128962e-01 1.56152356e+00 1.84904993e-01 -4.67342943e-01 5.66657424e-01 3.97993535e-01 -2.70333211e-03 -5.90124547e-01 -7.88260102e-01 -5.51359177e-01 -1.86839789e-01 9.33600813e-02 6.97339714e-01 1.20883334e+00 -2.24551201e-01 4.27744836e-02 -3.53863627e-01 4.68251616e-01 4.23715681e-01 2.52831370e-01 2.20758244e-01 -1.53142643e+00 3.52203008e-03 -3.24489534e-01 1.86026350e-01 -5.61047018e-01 2.81660631e-02 -1.00800014e+00 3.69126737e-01 -1.34744561e+00 6.00955307e-01 -1.01367211e+00 -6.22254372e-01 5.84476590e-01 -6.48398042e-01 1.29940093e-01 3.12863111e-01 3.04936856e-01 -4.96606112e-01 5.76594114e-01 1.80912161e+00 -5.53580880e-01 -3.34842175e-01 2.28438467e-01 -7.12682784e-01 8.72649670e-01 8.04357827e-01 -4.91643041e-01 -6.35131180e-01 -4.89035994e-01 -9.46235359e-02 -4.47388506e-03 4.60800558e-01 -1.12133992e+00 2.52014041e-01 -6.53542757e-01 5.83033919e-01 -4.72211599e-01 1.25109985e-01 -1.39085972e+00 1.94623172e-01 7.35086918e-01 -1.91873357e-01 1.47774473e-01 3.24675053e-01 9.24247324e-01 3.35096806e-01 -6.13225996e-01 8.87902975e-01 -5.95894217e-01 -7.63132751e-01 5.37100792e-01 -2.80316740e-01 -1.11841977e-01 1.50413144e+00 -1.69351846e-01 -4.97958660e-01 -3.42231914e-02 -3.85707110e-01 3.28275293e-01 2.23876923e-01 5.15332520e-01 7.02831626e-01 -1.29217768e+00 -7.12181330e-01 3.89018625e-01 -2.38437191e-01 2.49690756e-01 6.03182316e-01 3.11238110e-01 -2.15989515e-01 -6.09790981e-01 -3.46651047e-01 -5.74552834e-01 -1.53581560e+00 1.03198719e+00 4.24729764e-01 -1.16181239e-01 -4.87960488e-01 9.01343524e-01 4.01801050e-01 -4.26919073e-01 1.18840054e-01 -3.00918698e-01 -4.51277345e-01 7.21523613e-02 5.06321192e-01 4.68423575e-01 -3.29020709e-01 -1.12516768e-01 -2.91598618e-01 8.26500952e-01 1.89480949e-02 1.99922159e-01 1.02493250e+00 -1.19921900e-01 -2.42037013e-01 3.61742914e-01 7.26336718e-01 -7.77107552e-02 -1.37970603e+00 -1.26278430e-01 -5.87496996e-01 -5.91921747e-01 -8.19798652e-03 -1.05250084e+00 -1.43901503e+00 6.50774777e-01 9.76763606e-01 2.81447351e-01 1.31425869e+00 -2.56973147e-01 5.00367999e-01 2.51133531e-01 1.60373356e-02 -1.05931723e+00 3.81149620e-01 2.09645256e-01 8.10766280e-01 -1.33120036e+00 6.49690032e-02 -6.57411933e-01 -5.73178649e-01 1.21317172e+00 1.07707286e+00 2.49338616e-02 7.05599129e-01 7.27182746e-01 3.55351031e-01 -4.01767045e-02 -4.50189352e-01 -1.74073651e-01 7.26951007e-03 8.07030857e-01 -4.21588987e-01 -1.80899665e-01 -1.14799164e-01 3.71046096e-01 4.55877930e-01 -1.25739332e-02 2.89537460e-01 8.83772314e-01 -9.49687719e-01 -8.08741808e-01 -7.29280055e-01 4.73404497e-01 1.46028355e-01 2.81512067e-02 -4.12618458e-01 6.50658071e-01 7.86381483e-01 1.19473612e+00 -5.14199957e-02 -8.38765383e-01 1.01626582e-01 -7.53231719e-02 4.76640582e-01 -2.84931839e-01 -7.36229062e-01 8.62474367e-02 -3.39028686e-01 1.91560194e-01 -6.56818986e-01 -4.43092972e-01 -1.40172577e+00 -2.88095269e-02 -9.61477518e-01 4.01412658e-02 4.12451148e-01 1.12518287e+00 -2.96213686e-01 8.45458448e-01 1.10044730e+00 -5.78731477e-01 -7.65140057e-01 -8.48733664e-01 -6.58249557e-01 6.67367637e-01 2.48823166e-01 -5.89845479e-01 -6.08134687e-01 6.38686866e-02]
[10.854293823242188, -1.2221972942352295]
a2958e12-be4e-4b3f-a96e-546a6ca06c49
cs4ml-a-general-framework-for-active-learning
2306.00945
null
https://arxiv.org/abs/2306.00945v1
https://arxiv.org/pdf/2306.00945v1.pdf
CS4ML: A general framework for active learning with arbitrary data based on Christoffel functions
We introduce a general framework for active learning in regression problems. Our framework extends the standard setup by allowing for general types of data, rather than merely pointwise samples of the target function. This generalization covers many cases of practical interest, such as data acquired in transform domains (e.g., Fourier data), vector-valued data (e.g., gradient-augmented data), data acquired along continuous curves, and, multimodal data (i.e., combinations of different types of measurements). Our framework considers random sampling according to a finite number of sampling measures and arbitrary nonlinear approximation spaces (model classes). We introduce the concept of generalized Christoffel functions and show how these can be used to optimize the sampling measures. We prove that this leads to near-optimal sample complexity in various important cases. This paper focuses on applications in scientific computing, where active learning is often desirable, since it is usually expensive to generate data. We demonstrate the efficacy of our framework for gradient-augmented learning with polynomials, Magnetic Resonance Imaging (MRI) using generative models and adaptive sampling for solving PDEs using Physics-Informed Neural Networks (PINNs).
['Nick Dexter', 'Juan M. Cardenas', 'Ben Adcock']
2023-06-01
null
null
null
null
['active-learning', 'active-learning']
['methodology', 'natural-language-processing']
[ 5.56382060e-01 2.40555078e-01 -2.25796953e-01 -3.28124583e-01 -1.23512685e+00 -3.53257746e-01 4.71492738e-01 3.06233913e-01 -7.47225761e-01 1.05094039e+00 -1.28577977e-01 -1.29877508e-01 -5.33046007e-01 -7.99421787e-01 -9.53412592e-01 -1.25961804e+00 -3.77405673e-01 7.64262199e-01 -7.48871565e-02 -1.00684583e-01 3.10082078e-01 8.88644338e-01 -1.14131641e+00 -2.85723835e-01 1.25235772e+00 8.48764837e-01 1.24093957e-01 7.28768349e-01 2.17538755e-02 4.97692496e-01 -5.57923257e-01 8.61413553e-02 2.17220888e-01 -2.50976831e-01 -5.63158512e-01 2.09908545e-01 4.00048375e-01 1.11174509e-02 1.65057465e-01 1.01548564e+00 6.08365595e-01 5.64618886e-01 9.24214423e-01 -1.15087807e+00 -4.70689505e-01 3.85415167e-01 -3.39831740e-01 1.00642800e-01 -1.69622339e-02 -1.64215222e-01 6.25498533e-01 -1.11419702e+00 4.60028559e-01 8.71033549e-01 7.07964242e-01 5.23298800e-01 -1.89977586e+00 -2.75432944e-01 -1.69442922e-01 2.33829498e-01 -1.01525509e+00 -5.19286990e-01 1.02584016e+00 -6.86704457e-01 2.09436178e-01 4.74258095e-01 8.37095678e-01 1.18682575e+00 3.23618233e-01 8.95564139e-01 1.26591849e+00 -7.13014901e-01 7.31759429e-01 2.14830756e-01 3.79904181e-01 4.91730988e-01 2.01186746e-01 3.92576270e-02 -3.18608016e-01 -7.16381252e-01 9.46253836e-01 3.06923062e-01 -5.52505076e-01 -8.65770519e-01 -1.13651490e+00 1.20677125e+00 1.18017219e-01 2.47647643e-01 -6.68720841e-01 -2.09079191e-01 1.51660100e-01 3.50625008e-01 8.90845120e-01 6.76817656e-01 -3.36447746e-01 2.13328302e-01 -1.16556668e+00 3.23702365e-01 8.64234984e-01 7.43482530e-01 7.39059865e-01 1.93019301e-01 -8.56184736e-02 1.05475163e+00 2.07150519e-01 3.95712376e-01 1.93105757e-01 -9.97319400e-01 1.03869401e-02 2.68084724e-02 2.33021408e-01 -5.08586705e-01 -5.25884211e-01 -5.37861943e-01 -1.00947094e+00 5.44030368e-01 6.28511846e-01 -8.14964846e-02 -6.47785544e-01 1.55491781e+00 4.17357683e-01 8.01303089e-02 -2.79273868e-01 8.12227011e-01 5.60335517e-01 4.26524550e-01 -2.34964907e-01 -1.05061162e+00 6.34921193e-01 -6.55827999e-01 -6.47831500e-01 2.55364984e-01 4.24859822e-01 -5.78812957e-01 1.12068081e+00 7.97956288e-01 -1.37213266e+00 -3.66680413e-01 -8.13220203e-01 9.86460671e-02 -2.07428783e-01 -1.41402438e-01 4.30144429e-01 4.60178733e-01 -1.02912486e+00 9.22438145e-01 -1.08322537e+00 5.92740066e-02 4.75997329e-01 3.70278507e-01 -3.57655920e-02 8.00306350e-02 -9.09324586e-01 6.10643744e-01 1.63757473e-01 3.54377627e-02 -7.27847397e-01 -1.21200562e+00 -6.92007184e-01 -2.43762508e-01 4.57835793e-01 -4.70449865e-01 9.54926670e-01 -9.44809556e-01 -1.32965934e+00 5.90233326e-01 4.17211354e-02 -4.33236420e-01 8.93390477e-01 -1.32455751e-01 -4.37431410e-02 1.71611786e-01 -1.75309241e-01 2.97094673e-01 1.15567648e+00 -9.03033435e-01 1.25496894e-01 -5.17901540e-01 2.09711090e-01 6.04602620e-02 -3.24001402e-01 -3.03933531e-01 3.44033718e-01 -6.42493308e-01 2.06517026e-01 -1.00878942e+00 -6.57160580e-01 2.84729153e-01 -3.61807942e-01 -1.73906013e-01 6.73913956e-01 -4.82803077e-01 6.60156906e-01 -2.05745602e+00 5.20114124e-01 3.18410218e-01 4.60468233e-01 -1.70146778e-01 1.74392670e-01 3.34667683e-01 -3.28600764e-01 -1.82896122e-01 -8.88931274e-01 -2.43576750e-01 -2.12789923e-01 3.32081079e-01 -1.29198387e-01 8.70857775e-01 2.32184872e-01 8.38489950e-01 -9.55841184e-01 -5.78225672e-01 1.31314993e-01 4.75115299e-01 -5.24250150e-01 1.19855888e-01 -1.01485781e-01 1.12596667e+00 -4.63761419e-01 3.11376154e-01 4.92746651e-01 -3.37880164e-01 -3.15447420e-01 -3.72125283e-02 -1.15285084e-01 -1.05651796e-01 -1.20763946e+00 1.63011491e+00 -7.61543751e-01 5.57774723e-01 1.61874130e-01 -1.91497338e+00 9.54085410e-01 3.84882987e-01 1.01624846e+00 -1.58718631e-01 -2.49621093e-01 4.68936920e-01 -3.59043032e-02 -6.14209890e-01 1.02364220e-01 -2.43000597e-01 2.62005389e-01 3.76887202e-01 4.35842909e-02 -2.60066152e-01 2.30894014e-01 -1.04851782e-01 8.59603703e-01 -5.43412156e-02 6.40341103e-01 -6.78693175e-01 5.59191227e-01 -1.96962401e-01 2.84311086e-01 8.67491305e-01 8.55800509e-02 5.75942695e-01 4.79533136e-01 -1.96612746e-01 -1.17562807e+00 -1.07889390e+00 -7.70265639e-01 9.14300263e-01 -3.35501194e-01 1.13432400e-01 -4.77768719e-01 -4.89242017e-01 -2.27551728e-01 6.04947209e-01 -7.70861030e-01 -5.65115623e-02 -8.95536005e-01 -1.10622716e+00 1.76699106e-02 3.15465957e-01 1.94093823e-01 -1.00497019e+00 -5.57888210e-01 2.68319309e-01 1.34607166e-01 -6.76827729e-01 -2.54952401e-01 4.73853409e-01 -1.43412387e+00 -1.06038034e+00 -1.23720789e+00 -4.54088807e-01 8.03637803e-01 -2.69983232e-01 1.19292271e+00 -3.69017661e-01 -4.04850632e-01 8.71368825e-01 -1.11479521e-01 -2.90570438e-01 -4.17057484e-01 -3.37787382e-02 -1.00619875e-01 1.44854277e-01 -2.48008341e-01 -7.55388021e-01 -4.55222130e-01 1.16040006e-01 -1.02042592e+00 -2.03442559e-01 4.43050265e-01 1.31668770e+00 9.85114872e-01 -3.24737698e-01 5.78090549e-01 -1.25138402e+00 6.91436529e-01 -5.89949250e-01 -7.37189949e-01 3.18090290e-01 -5.03798604e-01 1.85055837e-01 7.38207698e-01 -9.71911073e-01 -6.87459469e-01 2.64763422e-02 -3.61991711e-02 -4.16448981e-01 2.19943132e-02 5.26574135e-01 1.05684035e-01 -5.04282892e-01 1.14630222e+00 2.60520309e-01 1.22049652e-01 -2.40912080e-01 1.11019842e-01 4.19541568e-01 2.87290514e-01 -8.32002699e-01 5.17541945e-01 4.76553112e-01 5.01195431e-01 -1.36925054e+00 -6.87702358e-01 -3.57204795e-01 -7.52664149e-01 -2.60875553e-01 4.09962177e-01 -1.99133128e-01 -6.01352394e-01 1.73725843e-01 -9.25007522e-01 -4.29939866e-01 -9.31034863e-01 7.57878661e-01 -1.21244836e+00 2.96959847e-01 -5.60733438e-01 -1.14294982e+00 -4.62843746e-01 -1.38946414e+00 1.05198359e+00 -1.50454119e-01 -2.88185142e-02 -1.50465214e+00 3.10484648e-01 -3.70093733e-02 5.45319438e-01 5.93351603e-01 9.64557469e-01 -7.73174167e-01 -3.75278383e-01 -1.76178310e-02 5.29151440e-01 3.48576486e-01 1.58810255e-03 -6.37854859e-02 -7.73642540e-01 -3.41017663e-01 6.17915273e-01 -4.44577932e-01 7.77941823e-01 9.39679980e-01 1.63620496e+00 -4.53817338e-01 -1.72413900e-01 6.16183579e-01 1.08624077e+00 2.81424761e-01 5.48983932e-01 -2.38363191e-01 4.50326771e-01 6.45712733e-01 4.05390620e-01 3.43925059e-01 -4.47565556e-01 8.05577338e-01 1.26961634e-01 -2.16095760e-01 4.25915301e-01 4.59144264e-01 4.33970839e-02 8.30603778e-01 -3.24575841e-01 2.01977849e-01 -7.62142539e-01 1.95294440e-01 -1.72776496e+00 -9.25431728e-01 -2.28611484e-01 2.68367720e+00 1.03699183e+00 8.88906121e-02 2.43768573e-01 6.07403100e-01 7.26041019e-01 -1.00051217e-01 -1.01867115e+00 5.36206551e-03 3.85762118e-02 7.18087614e-01 4.12124336e-01 6.28143132e-01 -1.12527835e+00 -2.17672288e-02 6.19570160e+00 9.83265400e-01 -1.24927235e+00 5.21811187e-01 6.61398530e-01 2.22183522e-02 -3.26456964e-01 -2.05034494e-01 -3.55917782e-01 3.79310817e-01 7.75548637e-01 -1.72876239e-01 3.84966075e-01 8.33529353e-01 1.35984957e-01 1.31429017e-01 -1.24627411e+00 1.04202974e+00 -1.84122957e-02 -1.29524219e+00 -5.64683855e-01 1.42348751e-01 7.37336576e-01 -1.38848752e-01 3.58830653e-02 1.33318916e-01 -8.66222158e-02 -9.06296194e-01 5.54592192e-01 7.99727321e-01 8.04315925e-01 -4.98557329e-01 1.60867766e-01 5.67303121e-01 -5.96565783e-01 7.40123168e-02 -4.43327129e-01 3.99552435e-01 1.78822026e-01 1.10273874e+00 -6.57079041e-01 3.15807015e-01 1.81178063e-01 8.45136046e-01 -4.31547351e-02 1.41621697e+00 2.52330840e-01 7.18834817e-01 -5.76476812e-01 -1.92661479e-01 -2.00627163e-01 -6.37055159e-01 9.16612744e-01 8.36133659e-01 3.28272700e-01 -1.90673899e-02 2.26196066e-01 1.04128039e+00 2.28018805e-01 3.46397460e-01 -5.81915915e-01 1.60869986e-01 2.55539238e-01 1.33708000e+00 -8.55008185e-01 -2.68997490e-01 -2.49042317e-01 6.41887188e-01 3.63315880e-01 6.97379470e-01 -5.53927720e-01 -1.98379666e-01 2.01972853e-02 4.62043166e-01 -7.96860233e-02 -3.57739538e-01 -5.88880442e-02 -1.18622506e+00 5.29301865e-03 -5.89211762e-01 3.30341756e-01 -3.91584873e-01 -1.36033487e+00 2.38401443e-01 4.59977180e-01 -1.22822797e+00 -3.79958302e-01 -6.23089492e-01 -4.97917324e-01 8.00448954e-01 -1.06635654e+00 -5.65922678e-01 -1.39786124e-01 6.98644698e-01 5.56256413e-01 -1.40389949e-01 7.79014528e-01 3.29172254e-01 -5.17500378e-02 2.97757924e-01 4.12233770e-01 -5.82393818e-02 3.15745622e-01 -1.53042614e+00 -7.79582933e-02 3.37968946e-01 1.94401816e-01 7.44689703e-01 7.77636945e-01 -3.45672011e-01 -1.37970638e+00 -7.75732338e-01 2.34494328e-01 -2.67348494e-02 5.78498006e-01 -5.06046891e-01 -1.10053122e+00 5.56827605e-01 -3.69827539e-01 4.21780556e-01 6.63529098e-01 8.86678025e-02 1.86504632e-01 -1.23843215e-01 -1.36496603e+00 3.56879354e-01 8.86309266e-01 -3.19187909e-01 -2.28518322e-01 8.64604592e-01 3.72725397e-01 -5.06736100e-01 -1.15195429e+00 5.40605247e-01 5.05904257e-01 -8.23036492e-01 1.23917902e+00 -7.33270288e-01 2.13231474e-01 2.58404106e-01 -7.58800879e-02 -1.52095222e+00 -1.35513678e-01 -6.95850909e-01 -4.84214783e-01 4.81293797e-01 2.46457323e-01 -1.00162828e+00 6.19158864e-01 2.79510617e-01 -3.63174587e-01 -1.23980308e+00 -1.08645201e+00 -7.29312658e-01 3.67600411e-01 -3.84215951e-01 1.54505923e-01 1.03648067e+00 -3.23410988e-01 5.92219867e-02 -1.28825411e-01 -2.25965634e-01 1.11913073e+00 2.16049273e-02 3.74633551e-01 -1.57002473e+00 -7.08763838e-01 -3.89531493e-01 -4.71720636e-01 -1.01986861e+00 2.53636539e-01 -8.45603466e-01 -3.68090272e-02 -1.19361317e+00 7.74053531e-03 -9.73898590e-01 -4.50190380e-02 1.11552672e-02 -8.85868818e-02 3.96016315e-02 -1.25496626e-01 4.25135165e-01 -7.56085441e-02 4.68836546e-01 1.53978658e+00 -1.63843736e-01 -2.75457233e-01 5.64360857e-01 -2.25851938e-01 6.76172853e-01 5.95141768e-01 -4.07278389e-01 -5.80819547e-01 1.97773606e-01 3.27531755e-01 5.15755534e-01 5.86201906e-01 -8.72059345e-01 2.41673782e-01 -3.66997868e-01 3.80126029e-01 -4.35151041e-01 5.19477248e-01 -7.17745960e-01 2.43577927e-01 3.96987081e-01 -7.94044018e-01 -2.95700073e-01 -3.72821331e-01 4.34407622e-01 5.91068179e-04 -6.79856658e-01 1.05893826e+00 -4.78462800e-02 -1.10397686e-03 4.84597951e-01 3.25374007e-02 3.39663744e-01 8.37454617e-01 -1.69445932e-01 1.00640170e-02 -4.07678634e-01 -1.35692418e+00 -2.67301738e-01 7.24001303e-02 -1.63194373e-01 6.27610445e-01 -1.35464334e+00 -7.89792299e-01 3.02590907e-01 4.27345559e-03 2.26987913e-01 1.01413399e-01 1.48773170e+00 -2.78846949e-01 3.94689031e-02 3.87565084e-02 -9.61418450e-01 -8.82880449e-01 4.46232945e-01 4.77685273e-01 -2.84720331e-01 -6.34161115e-01 5.09915411e-01 2.06210211e-01 -5.61084569e-01 1.85581237e-01 -4.28223759e-01 -3.35249871e-01 1.98408693e-01 3.40690732e-01 6.60275042e-01 3.64694774e-01 -2.28726894e-01 1.60548016e-01 5.32392502e-01 4.81855124e-02 -2.02561870e-01 1.43035030e+00 3.72819364e-01 -3.37864235e-02 1.09868729e+00 1.36110497e+00 -1.10952239e-02 -1.12973440e+00 -4.50397730e-01 -1.49693877e-01 -1.10282280e-01 -4.81680445e-02 -2.78432995e-01 -1.09193659e+00 8.94880056e-01 5.58046758e-01 5.97376466e-01 1.03104949e+00 6.48888722e-02 1.30445138e-01 4.54377890e-01 7.06060708e-01 -9.16546106e-01 9.42079350e-02 1.59388572e-01 1.11108708e+00 -1.08544898e+00 1.80809200e-01 -3.12884718e-01 -9.76333618e-02 1.37380564e+00 -5.65308779e-02 -4.32522714e-01 8.77216876e-01 2.48402268e-01 -3.26430857e-01 -1.54460045e-02 -3.98944438e-01 1.25142336e-01 5.61616361e-01 5.18937826e-01 5.75930595e-01 1.13228969e-01 -6.12788141e-01 -5.80148213e-03 -6.41562343e-02 -3.82258333e-02 5.40603161e-01 9.71058667e-01 -2.42480874e-01 -9.72288013e-01 -6.08797193e-01 8.40571880e-01 -3.71147811e-01 1.16873253e-02 1.58507615e-01 7.32185662e-01 -7.99484551e-02 5.23023486e-01 1.26683265e-01 4.96529013e-01 9.93084684e-02 9.69739258e-02 9.72139001e-01 -5.38569570e-01 -7.51087964e-02 2.55832046e-01 -2.83786207e-01 -2.76893348e-01 -8.32057595e-01 -9.31926787e-01 -9.20602918e-01 2.20988579e-02 -4.08359766e-01 3.37742537e-01 6.12366498e-01 9.39243793e-01 -1.94196403e-01 3.75816017e-01 6.66765988e-01 -8.54098320e-01 -8.09413135e-01 -1.10484481e+00 -8.29105020e-01 2.34570295e-01 6.64218247e-01 -9.13496614e-01 -5.79471886e-01 -6.58578873e-02]
[6.855490684509277, 4.083220481872559]
d42dea73-1962-4db3-b425-72ec2a22d8bd
sharpness-aware-minimization-an-implicit
2302.11836
null
https://arxiv.org/abs/2302.11836v3
https://arxiv.org/pdf/2302.11836v3.pdf
On Statistical Properties of Sharpness-Aware Minimization: Provable Guarantees
Sharpness-Aware Minimization (SAM) is a recent optimization framework aiming to improve the deep neural network generalization, through obtaining flatter (i.e. less sharp) solutions. As SAM has been numerically successful, recent papers have studied the theoretical aspects of the framework and have shown SAM solutions are indeed flat. However, there has been limited theoretical exploration regarding statistical properties of SAM. In this work, we directly study the statistical performance of SAM, and present a new theoretical explanation of why SAM generalizes well. To this end, we study two statistical problems, neural networks with a hidden layer and kernel regression, and prove under certain conditions, SAM has smaller prediction error over Gradient Descent (GD). Our results concern both convex and non-convex settings, and show that SAM is particularly well-suited for non-convex problems. Additionally, we prove that in our setup, SAM solutions are less sharp as well, showing our results are in agreement with the previous work. Our theoretical findings are validated using numerical experiments on numerous scenarios, including deep neural networks.
['Rahul Mazumder', 'Kayhan Behdin']
2023-02-23
null
null
null
null
['stochastic-optimization']
['methodology']
[-3.00482810e-02 3.03324431e-01 -1.42875031e-01 -3.18948656e-01 -5.62347889e-01 -2.76972413e-01 7.37749040e-02 -2.12274701e-03 -6.39320314e-01 9.94280577e-01 -1.06730826e-01 -2.94255853e-01 -4.84269828e-01 -4.87552345e-01 -1.01948869e+00 -1.10103917e+00 -1.55996799e-01 1.21751474e-02 -2.84788102e-01 -8.32743421e-02 4.11317162e-02 3.96192610e-01 -1.21928096e+00 -2.72183120e-01 9.96958196e-01 1.10818696e+00 -4.80364487e-02 5.31147480e-01 3.95848423e-01 4.50644255e-01 -2.43478477e-01 -6.01994276e-01 4.54655737e-01 -1.71512708e-01 -5.69742560e-01 -1.00556962e-01 7.16838837e-01 3.39306258e-02 -4.61849533e-02 1.43428230e+00 3.72664213e-01 2.66764849e-01 4.88834053e-01 -1.40192997e+00 -5.76188028e-01 7.58791625e-01 -6.39148116e-01 1.65351316e-01 -4.87348884e-01 -3.11452091e-01 1.15943241e+00 -1.05599988e+00 1.22816972e-01 1.01158547e+00 1.17308736e+00 7.29799867e-01 -1.13199770e+00 -5.68301439e-01 3.42369825e-01 6.84327334e-02 -1.21042097e+00 -3.85660052e-01 6.11249387e-01 -2.41043836e-01 4.78416473e-01 3.69819611e-01 3.19491327e-01 8.81980658e-01 1.21855274e-01 1.02474809e+00 1.27813947e+00 -5.45888901e-01 1.51345372e-01 5.13486207e-01 6.45425975e-01 9.44044530e-01 5.42001486e-01 -1.98754147e-02 -3.43812168e-01 8.28375816e-02 6.80915356e-01 -2.01089889e-01 -6.63399637e-01 -5.33707798e-01 -8.46884310e-01 1.04392147e+00 4.97003615e-01 3.87684911e-01 -3.51744384e-01 1.33898064e-01 3.64096493e-01 3.55423093e-01 7.63484895e-01 3.59704673e-01 -5.00269711e-01 3.28614786e-02 -1.00348735e+00 2.77570873e-01 1.04467773e+00 8.26382339e-01 5.51265001e-01 2.33506218e-01 1.45494901e-02 8.33441377e-01 9.87381041e-02 3.97906512e-01 4.01728600e-01 -7.71614134e-01 5.49760222e-01 6.31468818e-02 -7.32591227e-02 -1.22023869e+00 -6.77790821e-01 -9.74083781e-01 -1.41330469e+00 1.92579567e-01 7.08994329e-01 -3.32284093e-01 -4.42622095e-01 2.18089247e+00 -1.47784412e-01 1.46181226e-01 7.39658847e-02 9.98836696e-01 3.47510248e-01 4.07294333e-01 -2.25069374e-01 -4.05123293e-01 7.60350466e-01 -8.39731753e-01 -4.80912447e-01 -1.09547339e-01 5.70225418e-01 -1.68398976e-01 1.11385727e+00 6.07289195e-01 -1.24965799e+00 -2.62000173e-01 -9.94522870e-01 6.97579607e-02 -2.39818335e-01 3.63166630e-01 6.71733320e-01 9.95477021e-01 -1.38985825e+00 9.66103733e-01 -9.94372010e-01 -3.38881046e-01 5.26499212e-01 3.59081149e-01 -3.15747678e-01 2.03682929e-01 -9.69965339e-01 8.21738780e-01 3.27613980e-01 4.34719890e-01 -3.39861512e-01 -8.78726304e-01 -7.97454715e-01 1.33785635e-01 3.03067207e-01 -6.81619585e-01 1.14166248e+00 -1.10776758e+00 -1.51316667e+00 7.96239197e-01 -2.18982160e-01 -9.49837565e-01 8.94471645e-01 -3.87914479e-01 2.13200599e-01 -1.24198020e-01 -2.83249289e-01 3.45453650e-01 8.08930933e-01 -1.10823810e+00 -4.94189739e-01 -6.24630272e-01 2.12697372e-01 -1.39250875e-01 -6.77810729e-01 -2.65015393e-01 -2.88805198e-02 -6.10150039e-01 -6.86262995e-02 -8.73946369e-01 -3.86515647e-01 -1.28979072e-01 -5.83748281e-01 -9.52781588e-02 2.53271669e-01 -5.62257886e-01 1.19658196e+00 -2.01369309e+00 2.02202171e-01 4.22900140e-01 4.89791036e-01 2.44209751e-01 2.45507360e-02 1.22314179e-02 -2.75419861e-01 3.37943017e-01 -4.76466209e-01 -5.65472543e-01 3.41643870e-01 2.61607409e-01 -3.76872510e-01 9.22210574e-01 1.03915147e-01 9.17626083e-01 -4.87398535e-01 -1.98859602e-01 1.04777388e-01 4.67099249e-01 -7.18379200e-01 -2.35185772e-01 1.60304472e-01 1.60395190e-01 -3.12172890e-01 4.04356629e-01 9.37682390e-01 -1.71279252e-01 -3.78697366e-02 -2.78398216e-01 -1.55882403e-01 -1.53594598e-01 -1.09140217e+00 1.05066073e+00 -6.78417146e-01 9.41949844e-01 4.70904201e-01 -1.78101385e+00 5.48535526e-01 -1.59827143e-01 3.05734873e-01 -3.45163196e-01 2.79212564e-01 1.50903031e-01 -2.20544696e-01 -4.08329487e-01 3.28802794e-01 -4.43877161e-01 3.16105902e-01 1.64638594e-01 -5.50007485e-02 4.28680927e-01 1.54322401e-01 -1.19777724e-01 6.76580846e-01 -2.66815335e-01 3.76676351e-01 -7.99408078e-01 4.37830746e-01 -2.89550871e-01 4.19227362e-01 1.14400268e+00 -2.20843613e-01 5.80447257e-01 7.79571831e-01 8.61916095e-02 -9.18100476e-01 -1.09053862e+00 -4.26803380e-01 1.03518152e+00 1.39387146e-01 -4.90240902e-02 -1.10822403e+00 -3.92287344e-01 1.13175865e-02 6.70733690e-01 -9.39677656e-01 -2.69155920e-01 -6.88242495e-01 -1.20253694e+00 6.33331716e-01 6.03901684e-01 4.88796175e-01 -5.38338721e-01 -5.10798514e-01 -9.62350667e-02 1.98308736e-01 -1.19794488e+00 -3.37770909e-01 3.97931427e-01 -1.06003964e+00 -1.03163683e+00 -1.16172230e+00 -7.44968176e-01 6.87925935e-01 2.44215667e-01 1.07427704e+00 -3.52695137e-02 -2.33356096e-02 4.95652139e-01 -1.29617110e-01 -4.60238546e-01 -2.68119395e-01 3.85859340e-01 3.87299836e-01 1.38049275e-01 8.78025293e-02 -6.02058232e-01 -4.23443288e-01 1.69897228e-01 -9.41103935e-01 6.31930605e-02 8.07705164e-01 9.39673245e-01 3.76935452e-01 1.18875362e-01 6.08126998e-01 -7.97375679e-01 8.88489544e-01 -5.80995381e-01 -7.59558082e-01 2.70468205e-01 -1.01024342e+00 3.45592171e-01 9.59041476e-01 -3.80513072e-01 -9.15015757e-01 -2.01476857e-01 -1.84253573e-01 -4.03440118e-01 6.22821487e-02 6.41105473e-01 5.99309802e-02 -2.70893246e-01 5.51801622e-01 2.01434642e-01 8.85485262e-02 -5.45085073e-01 2.15889260e-01 2.94347525e-01 4.27278250e-01 -7.29116678e-01 9.42847192e-01 5.36318421e-01 2.74626911e-01 -1.00194585e+00 -1.15748298e+00 -2.76333421e-01 -6.33831739e-01 -1.04221359e-01 5.28323293e-01 -4.66316968e-01 -9.46754158e-01 4.31197762e-01 -8.24115455e-01 -4.06334966e-01 -9.65895280e-02 5.98906696e-01 -6.28061116e-01 3.53464186e-01 -5.54181814e-01 -1.07505012e+00 -3.80857795e-01 -9.58600998e-01 6.53398395e-01 3.61824453e-01 2.61031836e-01 -1.52534652e+00 -6.14163950e-02 -3.27575170e-02 3.28261137e-01 2.25434884e-01 7.37154365e-01 -5.74819744e-01 -1.51276350e-01 -8.21631867e-03 -4.73448128e-01 7.87592411e-01 -1.72695398e-01 -9.15038064e-02 -9.61243808e-01 -3.85890931e-01 2.65608579e-01 -1.02122031e-01 1.36930907e+00 8.41755927e-01 1.43019903e+00 -5.55420220e-01 -8.90882462e-02 9.20576632e-01 1.52006912e+00 -4.08367395e-01 3.66239727e-01 4.01410192e-01 5.75056076e-01 5.66268086e-01 3.25337797e-01 4.32413012e-01 1.80556774e-01 4.94600385e-01 6.19387031e-01 -2.30440989e-01 3.37413907e-01 2.02000737e-01 4.95067030e-01 6.72372580e-01 -3.19696069e-01 -9.08747911e-02 -5.25118947e-01 3.96930128e-01 -2.06100678e+00 -9.46852982e-01 -3.26115996e-01 2.31654668e+00 1.03275204e+00 1.18039407e-01 3.70916516e-01 2.31049761e-01 6.89065218e-01 -3.53814550e-02 -6.31847382e-01 -3.71627629e-01 -3.64443004e-01 4.59692359e-01 9.53757942e-01 4.28185403e-01 -1.21867955e+00 6.53993547e-01 6.37814283e+00 7.89052606e-01 -1.23839903e+00 1.02904521e-01 7.86276996e-01 -1.93426713e-01 2.99526583e-02 -5.24085581e-01 -6.95047796e-01 2.44547710e-01 6.62850380e-01 -3.01392257e-01 3.51825207e-01 1.05663025e+00 2.25685984e-01 8.80488902e-02 -1.11024284e+00 1.05672896e+00 -3.38935256e-02 -1.34577179e+00 -4.03074503e-01 2.04997867e-01 7.31335282e-01 1.07761234e-01 5.42825997e-01 3.01802129e-01 -1.19155515e-02 -1.15662873e+00 7.89161384e-01 5.18288910e-01 4.19300765e-01 -9.18267488e-01 8.84977818e-01 4.21576440e-01 -8.06840062e-01 -6.86439574e-02 -6.16390824e-01 -1.08319819e-01 -1.23702809e-01 9.84982789e-01 -4.74920094e-01 6.32698298e-01 6.60366416e-01 8.32980633e-01 -5.62734962e-01 1.11662042e+00 1.51665375e-01 7.91524410e-01 -4.03093040e-01 -1.73411027e-01 3.82785410e-01 -4.13686126e-01 6.45881355e-01 1.48281038e+00 2.24918023e-01 -1.57598689e-01 -1.74399897e-01 1.09962940e+00 -1.83019355e-01 2.21534699e-01 -3.93529683e-01 9.69054550e-02 -9.14035290e-02 1.18329012e+00 -5.74015617e-01 -2.92381831e-03 -3.92000258e-01 8.32220674e-01 5.76116562e-01 4.97946590e-01 -9.58238900e-01 -2.50056267e-01 8.43793273e-01 -1.81752458e-01 1.71795011e-01 -2.91854888e-01 -7.18021810e-01 -1.33997691e+00 3.85332346e-01 -5.72393775e-01 2.34778881e-01 -1.54388517e-01 -1.41111910e+00 3.38120103e-01 6.47498220e-02 -8.13774586e-01 6.19623140e-02 -1.23630381e+00 -5.21969736e-01 6.06740534e-01 -1.48771536e+00 -5.74143291e-01 -1.90573409e-01 4.07016009e-01 4.10837591e-01 3.47308218e-02 3.99504244e-01 4.46282119e-01 -1.07375503e+00 1.09332550e+00 5.88489175e-01 1.26829281e-01 3.42981339e-01 -1.48731101e+00 1.90414954e-03 8.71276557e-01 2.35148862e-01 8.02732825e-01 9.80217516e-01 -1.14585824e-01 -1.37209141e+00 -9.74727571e-01 3.50132972e-01 -2.03256235e-01 7.85664856e-01 -1.96146145e-01 -9.55245912e-01 4.97355998e-01 -6.38417602e-02 -2.32905075e-01 4.94329631e-01 2.42810085e-01 -1.98951006e-01 -9.05769616e-02 -1.00455928e+00 6.46730542e-01 1.00100410e+00 -1.14937991e-01 -3.30478787e-01 2.14684561e-01 2.92247385e-01 -3.25233042e-01 -6.46780372e-01 5.32611549e-01 6.15944445e-01 -1.33965695e+00 9.73162174e-01 -7.36579597e-01 4.69607323e-01 2.59805173e-01 -3.22728842e-01 -1.41313148e+00 -2.25453302e-02 -5.21496475e-01 -1.84043422e-01 9.65986371e-01 4.65784281e-01 -9.72642362e-01 8.62606406e-01 5.05348325e-01 -2.47331098e-01 -1.18962073e+00 -9.58502114e-01 -1.24234247e+00 6.93263948e-01 -6.08350217e-01 -4.03350070e-02 9.10366714e-01 -4.77954000e-02 -9.32905227e-02 -4.33449268e-01 2.16472641e-01 6.57531142e-01 -2.75544897e-02 5.36048353e-01 -1.18960404e+00 -2.66914845e-01 -1.09308779e+00 -3.42986405e-01 -1.25866604e+00 5.42921245e-01 -9.13787007e-01 1.05823889e-01 -1.06398273e+00 2.96151757e-01 -4.51437145e-01 -4.49726015e-01 2.79064298e-01 -2.40965635e-01 2.92753071e-01 2.30569735e-01 -6.35429099e-03 -3.90461713e-01 5.69281399e-01 9.19338822e-01 -2.78855152e-02 -1.66227087e-01 5.78521371e-01 -9.38939393e-01 9.46848631e-01 1.03034139e+00 -1.99723035e-01 -2.22524390e-01 -2.32228234e-01 4.31859255e-01 -5.42327106e-01 5.80412149e-01 -9.71258521e-01 1.04673266e-01 1.98531486e-02 1.91594467e-01 -2.77269632e-01 1.75492197e-01 -7.74325669e-01 -5.63623965e-01 4.76212025e-01 -6.34120524e-01 -1.56203926e-01 1.87212244e-01 5.25831044e-01 4.86523695e-02 -5.54083049e-01 1.10374749e+00 2.05810532e-01 -3.60823870e-01 4.83678401e-01 -2.94414442e-02 3.92841190e-01 6.98098719e-01 -1.19165190e-01 5.31146340e-02 -5.78906536e-01 -7.43432403e-01 1.06600173e-01 1.10124223e-01 -1.57886706e-02 4.49670702e-01 -1.11917520e+00 -7.01864779e-01 -6.25777245e-02 -2.77947068e-01 -1.20245121e-01 -4.19391356e-02 1.60056019e+00 -3.07406068e-01 6.26428366e-01 1.37844294e-01 -5.69872737e-01 -1.13356388e+00 3.22934031e-01 5.97953558e-01 -1.79454952e-01 -6.96497083e-01 1.06013227e+00 4.14833218e-01 -2.87023574e-01 6.14484251e-01 -9.37339187e-01 -7.61014642e-03 4.71853502e-02 3.81533980e-01 6.66124642e-01 1.11308314e-01 -3.51615340e-01 -2.96280146e-01 5.70882440e-01 -4.82936464e-02 -3.91482562e-02 1.43638015e+00 -1.65205255e-01 1.76649615e-02 4.81989980e-01 1.59850311e+00 5.63204214e-02 -1.41100204e+00 -3.88114154e-01 1.72790855e-01 -8.86107460e-02 1.29433200e-01 -2.59114742e-01 -1.39099741e+00 9.51370656e-01 4.09005523e-01 6.17823660e-01 1.27968085e+00 -4.07678783e-02 6.50423884e-01 6.61941886e-01 8.41435194e-02 -1.07627213e+00 -2.46183887e-01 6.07286036e-01 8.94274652e-01 -1.29521477e+00 -1.76147029e-01 -3.98780942e-01 -3.57194126e-01 1.38020313e+00 4.27267730e-01 -2.91030407e-01 7.37986088e-01 1.85860381e-01 -2.90365130e-01 9.46781263e-02 -4.83527601e-01 -2.11312950e-01 3.67759734e-01 4.22167212e-01 2.84127206e-01 -2.88996045e-02 -2.03247428e-01 8.94025803e-01 -6.65573895e-01 -1.25044525e-01 4.92829949e-01 2.79953003e-01 -3.61868531e-01 -5.54051340e-01 -2.12085977e-01 3.63543779e-01 -7.00293899e-01 -3.80131751e-01 -3.36276352e-01 9.57324982e-01 -2.86313444e-01 8.03391695e-01 -1.06206059e-01 -2.26896316e-01 1.14026174e-01 -1.53021157e-01 8.04404557e-01 -2.27482751e-01 -2.10593179e-01 -3.97839934e-01 -1.61874369e-01 -5.64555049e-01 -4.79362309e-01 -5.43564081e-01 -1.06352139e+00 -4.06316996e-01 -5.21060705e-01 2.88212560e-02 7.23925233e-01 1.23207247e+00 2.44693868e-02 4.06961590e-01 6.40578032e-01 -7.84704030e-01 -1.00334179e+00 -7.55340278e-01 -6.47960544e-01 -4.50632907e-03 7.64110148e-01 -6.53349102e-01 -8.46294045e-01 -1.84249774e-01]
[7.791507244110107, 3.837566375732422]
88f95762-c772-4ab0-af9e-820042f6f5cf
select-answer-and-explain-interpretable-multi
1911.00484
null
https://arxiv.org/abs/1911.00484v4
https://arxiv.org/pdf/1911.00484v4.pdf
Select, Answer and Explain: Interpretable Multi-hop Reading Comprehension over Multiple Documents
Interpretable multi-hop reading comprehension (RC) over multiple documents is a challenging problem because it demands reasoning over multiple information sources and explaining the answer prediction by providing supporting evidences. In this paper, we propose an effective and interpretable Select, Answer and Explain (SAE) system to solve the multi-document RC problem. Our system first filters out answer-unrelated documents and thus reduce the amount of distraction information. This is achieved by a document classifier trained with a novel pairwise learning-to-rank loss. The selected answer-related documents are then input to a model to jointly predict the answer and supporting sentences. The model is optimized with a multi-task learning objective on both token level for answer prediction and sentence level for supporting sentences prediction, together with an attention-based interaction between these two tasks. Evaluated on HotpotQA, a challenging multi-hop RC data set, the proposed SAE system achieves top competitive performance in distractor setting compared to other existing systems on the leaderboard.
['Bo-Wen Zhou', 'Guangtao Wang', 'Xiaodong He', 'Ming Tu', 'Jing Huang', 'Kevin Huang']
2019-11-01
null
null
null
null
['multi-hop-reading-comprehension']
['natural-language-processing']
[ 5.39180875e-01 3.09116155e-01 -6.99867159e-02 -8.21577668e-01 -1.52740848e+00 -4.30410177e-01 2.45156020e-01 7.27178812e-01 -4.31604058e-01 8.89391720e-01 6.90711856e-01 -2.93451458e-01 -4.55023110e-01 -2.24899158e-01 -7.59117723e-01 -1.17745779e-01 4.38674480e-01 9.88735318e-01 4.28813368e-01 -3.06425244e-01 7.48713493e-01 -2.14528114e-01 -1.55352306e+00 1.11740875e+00 1.62496078e+00 7.90608525e-01 7.02221036e-01 1.12999344e+00 -3.43127996e-01 1.56156421e+00 -7.92207479e-01 -5.57242334e-01 -3.09386581e-01 -5.65452635e-01 -1.46077001e+00 -2.04206645e-01 6.58589065e-01 -2.91875064e-01 3.65639091e-01 6.12997472e-01 5.85533082e-01 3.25604767e-01 6.19258642e-01 -9.85695541e-01 -8.36383700e-01 8.64270985e-01 -4.74024504e-01 5.57310462e-01 6.66136861e-01 -2.48024985e-01 1.46322680e+00 -1.11946082e+00 3.03054333e-01 1.64861691e+00 -3.46510322e-03 5.20572066e-01 -1.02253914e+00 -2.50656545e-01 5.10495901e-01 9.02337313e-01 -7.13744044e-01 -3.29949468e-01 5.99446058e-01 -1.11070544e-01 1.05345535e+00 6.21130466e-01 5.60115352e-02 9.72146034e-01 2.94371843e-01 1.28990793e+00 9.39251482e-01 -5.82955778e-01 -3.12947780e-02 2.00269192e-01 9.30846691e-01 7.19046712e-01 -2.25136027e-01 -7.52859652e-01 -1.04525149e+00 9.64200944e-02 -4.59673792e-01 -1.65631011e-01 -4.68592882e-01 2.32118502e-01 -9.60274756e-01 8.32611024e-01 3.12248617e-01 -1.86954185e-01 -3.51313829e-01 -4.20016870e-02 3.45599294e-01 5.12667775e-01 5.49009204e-01 5.83166122e-01 -6.87966287e-01 -1.18693545e-01 -7.02582717e-01 4.18539464e-01 7.67342269e-01 9.05099154e-01 4.18088138e-01 -5.37457883e-01 -1.20030439e+00 1.08739519e+00 3.23390633e-01 5.44818103e-01 3.06759000e-01 -1.12470162e+00 1.25177407e+00 6.18802726e-01 3.73152047e-02 -9.53218102e-01 -5.51232874e-01 -7.06507385e-01 -5.48638642e-01 -4.29622859e-01 8.40061828e-02 -1.59179866e-02 -4.43539888e-01 1.58162177e+00 2.65962362e-01 -2.11484820e-01 8.93147886e-02 9.60502982e-01 1.10580194e+00 7.06661999e-01 4.70724376e-03 -1.48035556e-01 1.51037049e+00 -1.68790424e+00 -1.07243657e+00 -4.43923324e-01 6.30690634e-01 -8.19005311e-01 1.43002546e+00 5.02065897e-01 -1.22516656e+00 -6.31335795e-01 -8.47972870e-01 -8.43409717e-01 1.28789604e-01 3.86389971e-01 5.16371354e-02 -9.94634777e-02 -6.93649530e-01 -3.23356614e-02 -1.68187201e-01 7.80636817e-03 3.31103504e-01 2.43632525e-01 1.55285653e-02 -5.35956621e-01 -1.22523832e+00 1.10817301e+00 1.36809900e-01 4.37965617e-02 -6.40381217e-01 -5.77519417e-01 -5.01434028e-01 5.18012881e-01 5.79834580e-01 -1.02596474e+00 1.51914346e+00 -6.90721571e-01 -1.40561390e+00 5.57055652e-01 -7.65204966e-01 -5.24473667e-01 3.32642972e-01 -8.32680881e-01 -1.68448791e-01 3.68310660e-01 4.84178931e-01 5.85523248e-01 8.32172632e-01 -1.25636303e+00 -7.76467621e-01 -4.31297034e-01 1.53520808e-01 5.94494164e-01 -2.74378508e-01 3.86542864e-02 -4.16831553e-01 -2.16548249e-01 7.19104558e-02 -4.48546916e-01 1.98686972e-01 -5.34564257e-01 -7.87258148e-01 -8.27162325e-01 5.64151883e-01 -1.00325096e+00 1.25424302e+00 -1.54112589e+00 5.69154918e-01 -1.99432492e-01 4.41872746e-01 1.33114338e-01 -5.49973249e-01 4.14270133e-01 1.95532545e-01 -1.31406561e-01 2.70884503e-02 -8.10313702e-01 -9.35091972e-02 1.23779499e-03 -5.13004720e-01 -1.81991115e-01 4.22849000e-01 7.36695230e-01 -9.10991251e-01 -6.43107235e-01 -2.88170040e-01 -7.63629153e-02 -5.30878186e-01 5.30211151e-01 -7.29819655e-01 5.03389776e-01 -6.49037123e-01 3.94327581e-01 6.36492848e-01 -4.90086704e-01 -1.62054628e-01 2.57863309e-02 1.45291537e-01 5.64089715e-01 -6.53650403e-01 1.66736495e+00 -6.32626414e-01 5.89057088e-01 -8.58478546e-02 -9.03811514e-01 9.77657199e-01 -2.02651266e-02 -2.15701059e-01 -1.01450658e+00 -1.19318746e-01 1.70160845e-01 2.16194376e-01 -1.08666337e+00 6.40805602e-01 2.30638072e-01 7.18036070e-02 6.86290264e-01 -9.98618901e-02 1.60448760e-01 2.62647927e-01 8.14899027e-01 1.26414716e+00 -1.50489360e-01 -4.67313826e-02 -1.81246087e-01 1.01795328e+00 1.77992531e-03 2.64288872e-01 9.92922068e-01 7.02455863e-02 5.63802183e-01 6.46611512e-01 -1.44229040e-01 -4.12221938e-01 -8.38311255e-01 2.71068484e-01 1.52219963e+00 3.29198331e-01 -4.48315263e-01 -6.21257663e-01 -9.83494282e-01 7.20338300e-02 1.36557293e+00 -4.75687534e-01 -3.19837928e-01 -5.67082107e-01 -8.09292048e-02 8.98597166e-02 5.17787635e-01 4.60511833e-01 -9.73920107e-01 -4.23750520e-01 2.14639097e-01 -8.95916104e-01 -1.08639109e+00 -5.63875258e-01 3.86277884e-01 -4.35908705e-01 -1.01391232e+00 -3.90180230e-01 -7.16029286e-01 6.05787396e-01 5.32359242e-01 1.54142880e+00 4.39642131e-01 7.63492882e-02 3.41593832e-01 -6.34351492e-01 -8.14050853e-01 -2.33239219e-01 3.54642510e-01 -2.87842631e-01 4.49450500e-02 5.29548228e-01 1.01254486e-01 -6.31710768e-01 1.88092291e-01 -5.05664587e-01 3.83533150e-01 6.32102132e-01 1.04577553e+00 6.10383093e-01 -2.93245614e-01 1.12261224e+00 -9.54929531e-01 1.34820557e+00 -6.23414874e-01 -1.42286599e-01 8.54766846e-01 -4.80195343e-01 3.90251875e-01 7.83529222e-01 -1.26175582e-02 -1.45647728e+00 -3.64663631e-01 -8.26353133e-02 -6.62584743e-03 2.63477921e-01 5.51379085e-01 -4.86020222e-02 5.53671718e-01 4.66518313e-01 1.71570778e-01 -2.05145627e-01 -5.43868661e-01 2.79751182e-01 9.05068576e-01 2.85846889e-01 -6.39186561e-01 3.86177778e-01 -8.13870654e-02 -1.78951353e-01 -3.29979509e-01 -1.94491720e+00 -6.30649924e-01 -5.14347911e-01 -3.38308811e-01 9.53896999e-01 -8.56774628e-01 -9.05124843e-01 -4.02273163e-02 -1.86863804e+00 9.04140174e-02 1.18217237e-01 2.63024628e-01 -3.82680535e-01 4.63767461e-02 -4.56268460e-01 -9.70920265e-01 -6.64713025e-01 -1.11256504e+00 1.20853961e+00 2.33620301e-01 -4.58494514e-01 -7.67974496e-01 -5.67923374e-02 1.45348787e+00 2.39509895e-01 -3.35441232e-01 1.43933439e+00 -1.01326406e+00 -8.50737810e-01 -3.77212930e-03 -3.66772592e-01 3.62864971e-01 -2.85822660e-01 -4.20352638e-01 -8.44893038e-01 2.05027200e-02 8.47734362e-02 -9.77699339e-01 1.23385656e+00 1.81869492e-01 1.70811534e+00 -3.97101551e-01 -1.14756107e-01 -1.16253093e-01 9.05235648e-01 -2.38120183e-01 2.03390077e-01 2.06773385e-01 7.36758888e-01 1.06518972e+00 9.38204646e-01 2.97792912e-01 1.00376344e+00 5.61908603e-01 4.43951160e-01 1.22994915e-01 -1.35212198e-01 -1.79376364e-01 1.92119882e-01 1.10200644e+00 3.48154604e-01 -8.53661001e-01 -6.81067824e-01 3.79949152e-01 -2.20670962e+00 -9.75411892e-01 -8.00787151e-01 1.82614028e+00 9.03632641e-01 1.45985499e-01 -2.59264916e-01 2.44690347e-02 4.30078089e-01 -6.97843358e-02 -6.86178148e-01 -6.92991793e-01 -2.50660479e-01 1.95456102e-01 -2.56172955e-01 8.69159043e-01 -6.67891681e-01 7.25808978e-01 5.28180599e+00 8.17256927e-01 -4.45066482e-01 2.83039212e-01 7.08519995e-01 -2.52717167e-01 -6.40160859e-01 -1.06999040e-01 -9.65168118e-01 3.13610345e-01 7.66129375e-01 -1.50748387e-01 2.01759920e-01 4.09281582e-01 4.80374217e-01 -5.21577299e-01 -1.23629487e+00 4.15110856e-01 6.21324301e-01 -1.21011114e+00 3.19429934e-01 -6.61408603e-01 5.00034392e-01 -2.93908775e-01 6.73978776e-02 4.39684451e-01 1.37483655e-02 -1.17039073e+00 7.34596610e-01 9.51104939e-01 1.00662231e-01 -7.59867132e-01 9.36400294e-01 8.31590176e-01 -6.64847374e-01 -3.89770806e-01 -4.10103291e-01 -7.83504918e-02 2.42147058e-01 5.85089505e-01 -6.22085869e-01 5.47607243e-01 6.12260342e-01 5.00971496e-01 -7.90222704e-01 8.07685256e-01 -6.46021426e-01 6.99950159e-01 2.14407846e-01 -5.07715881e-01 1.30280867e-01 1.42819434e-01 4.34464455e-01 9.21334147e-01 7.58777410e-02 3.06852669e-01 7.92379752e-02 8.02861035e-01 -3.48536611e-01 2.93174565e-01 -2.83503141e-02 3.94631535e-01 5.79206586e-01 1.20015228e+00 -9.19824988e-02 -3.62761557e-01 -3.20239574e-01 1.14701271e+00 1.09574711e+00 2.14137375e-01 -6.77625358e-01 -3.68672192e-01 -4.90758754e-02 -1.49068385e-01 8.91971365e-02 2.62125790e-01 -4.57898498e-01 -1.09295547e+00 4.23660189e-01 -1.10288322e+00 6.00389540e-01 -1.10943401e+00 -1.38669968e+00 5.76362967e-01 -1.79273576e-01 -9.02365804e-01 -1.60858423e-01 -4.02649194e-01 -7.81980574e-01 1.13241434e+00 -1.88920617e+00 -9.60768044e-01 -4.43942040e-01 4.13633645e-01 1.20366216e+00 -1.05455674e-01 6.81587934e-01 4.66373004e-02 -6.77087784e-01 5.22504866e-01 3.50161269e-02 -5.33884168e-01 8.25166166e-01 -1.54129767e+00 -2.19433501e-01 6.06238425e-01 5.58431372e-02 5.86109042e-01 8.05475116e-01 -5.28803051e-01 -1.19214237e+00 -9.99763608e-01 1.43137717e+00 -8.10956955e-01 2.52138078e-01 -3.24738741e-01 -1.13563979e+00 4.90838051e-01 6.79663420e-01 -5.85411191e-01 9.24462557e-01 4.51758981e-01 -3.10535312e-01 -2.28411809e-01 -8.26201499e-01 3.62974107e-01 7.65842021e-01 -3.56659859e-01 -8.88196766e-01 9.33939517e-01 1.11559594e+00 -3.16917837e-01 -4.64425862e-01 4.88463268e-02 1.76486716e-01 -8.96541357e-01 7.67643809e-01 -1.00580311e+00 1.06659150e+00 -1.46642163e-01 -4.07248642e-03 -1.49923646e+00 -2.90113181e-01 -1.84179932e-01 -2.91540414e-01 1.25293612e+00 8.57845604e-01 -7.01341555e-02 2.94759870e-01 4.94393289e-01 -4.70919460e-01 -1.07744026e+00 -8.42064321e-01 -3.45078290e-01 -1.62877142e-02 -1.35541886e-01 4.10009027e-01 3.29356372e-01 7.01673329e-03 1.26712489e+00 -4.79964048e-01 9.14080665e-02 6.17748678e-01 3.23481292e-01 6.06346309e-01 -1.11695850e+00 -3.62141132e-01 -9.60919261e-02 5.32779932e-01 -1.53382897e+00 5.74698210e-01 -1.05007648e+00 2.77188599e-01 -2.00511861e+00 6.17741644e-01 1.40429474e-02 -1.27829209e-01 1.85439624e-02 -9.39545572e-01 -5.65248966e-01 2.75942475e-01 1.67557150e-01 -1.40059686e+00 9.31444466e-01 1.54755926e+00 -3.47531646e-01 1.48084193e-01 1.48686945e-01 -8.59792829e-01 5.38969576e-01 6.60949647e-01 -5.56570947e-01 -8.27052951e-01 -1.11591816e+00 3.62386107e-01 4.42156702e-01 1.22656964e-01 -7.20923007e-01 4.79433298e-01 4.98217978e-02 3.96228552e-01 -1.09723151e+00 4.74040210e-01 -4.89207536e-01 -8.29700708e-01 2.85979092e-01 -1.31896639e+00 2.79039860e-01 -1.00781821e-01 8.42610240e-01 -3.20736349e-01 -4.99132276e-01 4.25998539e-01 1.47162750e-02 -3.87001336e-01 -9.45078805e-02 -2.34320372e-01 5.26034653e-01 7.75304437e-01 2.75591165e-01 -9.40380633e-01 -7.62867689e-01 -4.55232620e-01 1.11813807e+00 -5.59907973e-01 5.99452674e-01 1.02847886e+00 -9.83669698e-01 -1.19687784e+00 -1.54848889e-01 3.54760170e-01 2.57112216e-02 5.31934857e-01 9.34312284e-01 -1.39191985e-01 7.36070633e-01 1.70582131e-01 -5.72886765e-01 -1.64134669e+00 1.94948018e-01 1.54469416e-01 -7.24651396e-01 -2.44749367e-01 1.30368149e+00 -1.76685065e-01 -5.82101822e-01 4.08118546e-01 -1.72862113e-01 -7.60266721e-01 1.74697071e-01 7.53861248e-01 4.97480243e-01 2.02024296e-01 -1.36091918e-01 -2.86081105e-01 3.38698179e-01 -5.60856462e-01 5.04574068e-02 1.19648838e+00 -5.50861120e-01 -2.91412383e-01 4.85917568e-01 1.05017674e+00 -2.55475957e-02 -8.39726150e-01 -3.17752123e-01 2.31768236e-01 -3.88300270e-01 -1.52173951e-01 -1.25942767e+00 -4.63140905e-01 1.19756496e+00 1.07675292e-01 1.00901917e-01 9.16762948e-01 2.64781803e-01 9.78608429e-01 7.07050025e-01 -2.56387949e-01 -1.20618200e+00 7.87671864e-01 8.47346485e-01 1.40614045e+00 -1.56476724e+00 -1.10135339e-01 -5.43307483e-01 -9.07561123e-01 1.14376533e+00 1.28479695e+00 3.78349036e-01 6.97253123e-02 -2.76779324e-01 1.48185551e-01 -3.54718626e-01 -1.57087541e+00 -2.49728505e-02 6.48323178e-01 2.12561950e-01 6.48771942e-01 -1.04446568e-01 -5.32628834e-01 8.57395887e-01 -1.34448379e-01 -3.50109875e-01 4.15360868e-01 5.52864969e-01 -8.27598393e-01 -8.64437521e-01 -2.86584377e-01 8.61061156e-01 -2.91034877e-01 -2.47016162e-01 -5.83999395e-01 1.47630528e-01 -1.49615988e-01 1.60747063e+00 6.87420135e-03 -2.79607803e-01 4.14120227e-01 6.98954239e-02 2.37584010e-01 -7.40877092e-01 -8.11169803e-01 -5.09840310e-01 4.86480802e-01 -4.49170321e-01 -3.01038593e-01 -4.99182314e-01 -1.37307942e+00 -6.87304586e-02 -4.62365001e-01 3.25565755e-01 3.12493086e-01 1.40976870e+00 6.08338296e-01 1.04631293e+00 6.39207482e-01 5.95342293e-02 -9.37567532e-01 -1.11367905e+00 -1.14489861e-01 3.77001584e-01 4.73011523e-01 -3.70525658e-01 -2.25045934e-01 -2.38820553e-01]
[11.314056396484375, 8.009955406188965]
e9445d9c-61c0-4735-8bbd-3e2a235589c5
boosting-image-based-mutual-gaze-detection
2010.07811
null
https://arxiv.org/abs/2010.07811v2
https://arxiv.org/pdf/2010.07811v2.pdf
Boosting Image-based Mutual Gaze Detection using Pseudo 3D Gaze
Mutual gaze detection, i.e., predicting whether or not two people are looking at each other, plays an important role in understanding human interactions. In this work, we focus on the task of image-based mutual gaze detection, and propose a simple and effective approach to boost the performance by using an auxiliary 3D gaze estimation task during the training phase. We achieve the performance boost without additional labeling cost by training the 3D gaze estimation branch using pseudo 3D gaze labels deduced from mutual gaze labels. By sharing the head image encoder between the 3D gaze estimation and the mutual gaze detection branches, we achieve better head features than learned by training the mutual gaze detection branch alone. Experimental results on three image datasets show that the proposed approach improves the detection performance significantly without additional annotations. This work also introduces a new image dataset that consists of 33.1K pairs of humans annotated with mutual gaze labels in 29.2K images.
['Bradley Green', 'Yukun Zhu', 'Xuhui Jia', 'Raviteja Vemulapalli', 'Ching-Hui Chen', 'Bardia Doosti']
2020-10-15
null
null
null
null
['mutual-gaze']
['computer-vision']
[ 1.16848432e-01 3.34468096e-01 -1.09107532e-01 -6.62185550e-01 -2.91242301e-01 -1.40687615e-01 3.88120621e-01 -7.43930861e-02 -5.83897412e-01 2.32503965e-01 -1.09782971e-01 -1.03644408e-01 1.51609749e-01 -1.16215460e-01 -7.95959651e-01 -7.50934541e-01 1.23346999e-01 1.02650061e-01 2.88979888e-01 6.06376193e-02 5.06052494e-01 -1.48228094e-01 -1.92539227e+00 -7.04604089e-02 6.50147796e-01 1.02060056e+00 4.17486310e-01 7.25692272e-01 2.05563217e-01 9.88706410e-01 -5.47191381e-01 -3.53970051e-01 7.75132794e-03 -5.95908284e-01 -9.25644219e-01 1.85373694e-01 6.88199461e-01 -4.79318827e-01 1.78176433e-01 8.60025823e-01 3.13603550e-01 2.57520545e-02 5.26170611e-01 -1.75148141e+00 -5.17249346e-01 2.51460105e-01 -1.33430791e+00 4.28852826e-01 7.56465018e-01 8.32995623e-02 9.68440592e-01 -7.67349839e-01 2.12752566e-01 1.15000248e+00 4.52070028e-01 8.95010948e-01 -1.02877867e+00 -1.07294738e+00 1.49500951e-01 4.91774648e-01 -1.62701583e+00 -5.97599328e-01 7.11606205e-01 -5.23256779e-01 5.22006154e-01 2.96130806e-01 2.89067864e-01 7.46067226e-01 -1.70024469e-01 1.03999627e+00 1.16094935e+00 -8.20979178e-01 -4.61346447e-01 3.03445607e-01 4.59748805e-01 9.70769465e-01 6.10628724e-02 1.79350838e-01 -9.36924279e-01 1.70497850e-01 4.02815789e-01 -2.42131762e-02 -5.31519651e-01 -3.30556571e-01 -1.04443419e+00 6.35117769e-01 8.42876375e-01 1.62133515e-01 4.70903590e-02 1.64360493e-01 -1.37742450e-02 -6.49776123e-03 5.43449104e-01 5.24187796e-02 -9.71437022e-02 -2.06850339e-02 -7.64210820e-01 4.54257205e-02 4.18412000e-01 1.07855892e+00 1.10240352e+00 -7.61711061e-01 -1.14585795e-01 5.54826677e-01 8.92217696e-01 7.88233101e-01 1.94664717e-01 -7.43380010e-01 3.43114316e-01 7.80333161e-01 1.07830755e-01 -1.03674579e+00 -5.63290358e-01 1.48978755e-01 -3.61246079e-01 2.33829945e-01 5.79974890e-01 -1.75647616e-01 -7.41151750e-01 1.80872798e+00 6.82557464e-01 1.08855799e-01 -2.45398626e-01 1.23700309e+00 9.75105584e-01 2.57605791e-01 1.87353209e-01 -1.14011437e-01 1.77464294e+00 -1.15132654e+00 -9.12345886e-01 -2.41845191e-01 9.84165430e-01 -8.15890372e-01 1.01479340e+00 1.04222685e-01 -9.30351675e-01 -6.27640128e-01 -8.77803802e-01 -3.08269709e-01 -5.41839190e-02 2.69611090e-01 3.28787684e-01 6.38472915e-01 -1.10568833e+00 -1.33262694e-01 -5.98806202e-01 -4.76772964e-01 3.20277244e-01 5.98844409e-01 -4.54581290e-01 1.40664086e-01 -9.55838203e-01 7.69773602e-01 9.63110179e-02 1.39441028e-01 -5.44352710e-01 -2.89886236e-01 -1.00535119e+00 -3.59186716e-02 5.00794947e-01 -4.18111503e-01 1.39486921e+00 -1.12346709e+00 -1.32914650e+00 1.24284458e+00 -7.80224025e-01 -2.07189113e-01 2.43098989e-01 -3.78716499e-01 -7.14033097e-02 1.76570281e-01 2.18891546e-01 1.09532511e+00 1.07433379e+00 -1.14327705e+00 -9.66462374e-01 -6.54003441e-01 2.44828880e-01 4.66175526e-01 -2.77502984e-01 3.24723154e-01 -7.15661049e-01 5.10200188e-02 -1.11252338e-01 -1.16992891e+00 3.86068434e-01 1.12788096e-01 -5.17292023e-01 -7.59140849e-01 1.05831170e+00 -4.61237371e-01 1.11322474e+00 -2.11295462e+00 7.05624521e-02 -6.51615625e-03 6.73785567e-01 1.46994486e-01 2.10447684e-01 -2.16020361e-01 -2.60961000e-02 -1.11029908e-01 -1.06524657e-02 -9.72400725e-01 -2.07203060e-01 -1.41326696e-01 1.76623791e-01 4.52596426e-01 1.99780017e-01 9.59672987e-01 -8.12691689e-01 -9.50445950e-01 1.35197476e-01 5.23198724e-01 -3.60728294e-01 5.09102106e-01 2.26991579e-01 6.71153367e-01 -1.34588704e-01 3.19429547e-01 7.07496941e-01 -6.66915059e-01 5.85062318e-02 -1.95347634e-03 -9.97197349e-03 9.67223868e-02 -5.46139777e-01 1.48944545e+00 -2.43121833e-01 9.76138175e-01 -2.24177375e-01 -5.91364324e-01 8.32045376e-01 1.25100508e-01 4.58604693e-02 -8.20520461e-01 5.27491987e-01 -2.88933337e-01 8.61335322e-02 -7.20929384e-01 3.63887310e-01 1.75867379e-01 -1.18564419e-01 9.79288280e-01 1.76590845e-01 5.14389634e-01 5.85582219e-02 1.10814787e-01 5.54138482e-01 1.31625250e-01 3.07369888e-01 -1.62840232e-01 9.44759130e-01 -3.61776024e-01 2.59573430e-01 2.14888394e-01 -6.39964402e-01 4.56246883e-01 5.31383812e-01 -2.57036090e-01 -5.67873597e-01 -4.54886496e-01 1.24777980e-01 1.60379708e+00 6.49913728e-01 -4.91416812e-01 -1.18976426e+00 -1.09761262e+00 -2.43241355e-01 4.04571295e-01 -1.22273672e+00 -8.29811916e-02 -5.77668846e-01 -4.50384021e-01 4.03687567e-01 3.94539028e-01 6.24976218e-01 -8.38301480e-01 -9.21389878e-01 -8.29677999e-01 -4.50207740e-01 -1.05962408e+00 -8.91863704e-01 -1.94183186e-01 -3.20151389e-01 -1.44116879e+00 -8.49945009e-01 -8.96149635e-01 9.69262123e-01 9.24820244e-01 8.51895094e-01 5.04353166e-01 -6.89294115e-02 3.39499593e-01 -3.79782706e-01 -5.06258786e-01 -4.85157296e-02 2.69687563e-01 6.47974312e-02 3.10221523e-01 8.49099040e-01 5.68787083e-02 -5.66262484e-01 6.94130242e-01 -2.02149570e-01 2.24198684e-01 4.55670655e-01 6.58895254e-01 2.13129371e-01 -5.30330360e-01 2.44530573e-01 -8.81328642e-01 2.75778681e-01 -1.69621319e-01 -5.13444424e-01 3.68979037e-01 -5.25944054e-01 1.20520420e-01 -2.63036042e-01 -2.10469171e-01 -1.32050061e+00 1.75849140e-01 2.39595190e-01 -5.06202579e-01 -4.32857066e-01 -8.55401456e-02 3.31544993e-03 -3.01403850e-01 6.11449540e-01 -4.51380536e-02 2.80133992e-01 -4.12083477e-01 2.15088293e-01 1.07357824e+00 2.39825919e-01 1.36337906e-01 5.78194141e-01 3.84834766e-01 -1.07328296e-01 -7.69775212e-01 -1.32709444e+00 -7.96200991e-01 -9.57060456e-01 -5.23274064e-01 1.23476875e+00 -9.27155137e-01 -1.53226209e+00 5.50329745e-01 -1.09951162e+00 -1.93423882e-01 4.53106076e-01 4.31265563e-01 -3.69266421e-01 3.21286410e-01 -2.21558675e-01 -9.48093653e-01 -5.21231234e-01 -1.16547918e+00 1.46194100e+00 5.36217153e-01 -3.33130419e-01 -8.62921715e-01 1.73265189e-02 6.16579235e-01 -1.05566718e-01 -2.15900660e-01 2.82889009e-01 -4.57582355e-01 -6.20143294e-01 6.36525499e-03 -6.13149166e-01 1.12133510e-01 1.72191665e-01 -2.13119179e-01 -1.23675942e+00 -9.30445641e-02 -6.90556914e-02 -3.67913604e-01 6.37465596e-01 1.10635146e-01 7.83063710e-01 1.45217683e-02 -7.50718474e-01 4.53414947e-01 8.88491511e-01 -2.95875128e-02 3.62208396e-01 5.22831781e-03 1.02496481e+00 1.09104824e+00 1.00897169e+00 1.45552710e-01 8.23629677e-01 7.05387115e-01 5.30721784e-01 -7.49039138e-03 -1.91652671e-01 -1.96291819e-01 2.57183015e-01 5.43204546e-01 -9.52078402e-02 -2.57209569e-01 -1.17097306e+00 3.37939352e-01 -1.79308593e+00 -8.00173163e-01 -5.02905607e-01 2.08120131e+00 7.32614696e-01 -2.97644399e-02 3.55544716e-01 1.70696422e-01 1.10178065e+00 -1.33320764e-01 -3.30409825e-01 -1.64500568e-02 4.00915504e-01 -2.65071720e-01 3.35350156e-01 5.89636266e-01 -1.24751961e+00 8.31574082e-01 6.18043947e+00 3.49564821e-01 -1.10379267e+00 3.04666847e-01 4.32284087e-01 -1.77285925e-01 3.18955928e-01 -1.78783089e-01 -1.31392705e+00 6.04180813e-01 7.65132010e-01 1.63812369e-01 4.25603949e-02 5.55630147e-01 2.76701041e-02 -5.56243241e-01 -1.23444235e+00 1.37839210e+00 6.80069804e-01 -6.73364162e-01 -5.09997129e-01 2.65155613e-01 3.88556451e-01 -2.54523873e-01 2.19232962e-01 -5.04435562e-02 -1.07110724e-01 -8.54895413e-01 5.93194902e-01 6.55909002e-01 7.60854542e-01 -6.64080203e-01 7.39349127e-01 6.01534128e-01 -9.97260094e-01 -5.46709113e-02 1.92235664e-01 -2.21495643e-01 1.97195441e-01 -4.06307951e-02 -1.12955058e+00 1.52046263e-01 1.06135106e+00 8.32000434e-01 -8.24048579e-01 8.85880888e-01 -6.41428292e-01 3.62891555e-01 -1.89947873e-01 -1.58103332e-01 -1.02588899e-01 1.31242007e-01 3.28181714e-01 7.27175772e-01 -1.45642459e-01 8.02214351e-03 -2.00829878e-01 7.83169687e-01 -1.45776257e-01 -1.54390275e-01 -4.09547776e-01 4.98270482e-01 2.20811769e-01 1.23076189e+00 -5.39840400e-01 -1.36566952e-01 -3.84198546e-01 1.06029320e+00 5.10770023e-01 1.13185011e-01 -1.06136966e+00 -5.70540547e-01 5.77392340e-01 1.36969581e-01 2.31930718e-01 1.24670133e-01 4.59066499e-03 -8.39427948e-01 -2.58894376e-02 -2.46280119e-01 2.62081474e-01 -1.10253012e+00 -8.90563786e-01 4.78440225e-01 1.51209370e-03 -1.06929994e+00 -4.14649159e-01 -4.09715623e-01 -3.87930691e-01 9.71795082e-01 -1.65934658e+00 -1.35856211e+00 -9.06966269e-01 5.92659354e-01 2.57731795e-01 3.41288954e-01 6.06304169e-01 1.70491710e-01 -7.80422151e-01 9.49995756e-01 -6.05499744e-01 1.25232488e-01 1.08714485e+00 -1.12110567e+00 2.18628675e-01 5.77570319e-01 1.32390652e-02 4.81990159e-01 5.68232656e-01 -2.60155201e-01 -7.47747838e-01 -6.02044761e-01 1.28718472e+00 -8.53930354e-01 2.50992507e-01 -6.21987581e-01 -7.82358289e-01 8.65949333e-01 6.14878416e-01 -1.76289335e-01 9.43828881e-01 3.72944415e-01 -2.49657840e-01 2.93359347e-02 -9.65511978e-01 3.75527978e-01 1.19717038e+00 -5.61583221e-01 -7.94930458e-01 1.13062866e-01 7.82224834e-01 -4.55332130e-01 -4.40796703e-01 1.94855064e-01 6.06711864e-01 -9.49299753e-01 7.27223039e-01 -1.15486003e-01 1.58231348e-01 -4.05324817e-01 2.96838641e-01 -7.79390872e-01 -2.98364982e-02 -4.55848694e-01 -2.70554513e-01 1.21971536e+00 7.62424245e-02 -4.76418436e-01 7.83981383e-01 6.17311537e-01 3.31028968e-01 -4.66555983e-01 -5.93814552e-01 -1.98139593e-01 -4.75087315e-01 -2.56130798e-03 5.81725597e-01 7.54750431e-01 3.18063021e-01 8.53498757e-01 -4.63370800e-01 1.58921406e-01 7.73019254e-01 2.78035589e-02 1.20963871e+00 -1.40688491e+00 8.39551687e-02 -1.51575327e-01 -5.88690639e-01 -1.52995431e+00 2.80862510e-01 -3.80684376e-01 3.30119222e-01 -8.87047768e-01 4.14942384e-01 -3.99025083e-01 -2.95361072e-01 5.65024734e-01 -5.30312896e-01 6.64924800e-01 2.76795238e-01 5.33954024e-01 -1.05281615e+00 3.35950524e-01 1.34972680e+00 1.28335804e-01 -4.90795337e-02 3.30633149e-02 -7.82436132e-01 8.00785780e-01 6.21249855e-01 -4.15879279e-01 -4.16544378e-01 -6.06717527e-01 6.47076368e-02 -1.38899192e-01 5.12630582e-01 -8.46162200e-01 6.55054390e-01 2.65465140e-01 2.28097394e-01 -7.69714355e-01 4.51086015e-01 -6.73709154e-01 -5.43254375e-01 2.25149110e-01 -3.52023154e-01 -1.75349176e-01 -2.04748791e-02 6.42599285e-01 -1.04475975e-01 -3.43864918e-01 7.26170003e-01 2.77683616e-01 -7.77428508e-01 5.83869405e-02 2.84647290e-02 -2.10277930e-01 1.47961545e+00 -1.66572675e-01 -2.42859498e-01 -3.34615499e-01 -8.86609256e-01 5.32581627e-01 5.91573298e-01 4.87490803e-01 4.54748034e-01 -1.09324276e+00 -3.10385406e-01 4.23183203e-01 4.92803007e-01 2.31973417e-02 2.09132254e-01 1.25565934e+00 -1.05971336e-01 5.87373316e-01 -1.98943034e-01 -1.22830212e+00 -2.00660992e+00 7.01074481e-01 3.30063492e-01 -6.65201806e-03 -7.26313144e-02 1.27599061e+00 8.54833066e-01 -3.02452445e-01 2.19076484e-01 -1.44833792e-02 -6.47665977e-01 1.25716686e-01 9.31768537e-01 2.84364909e-01 -1.61706284e-01 -1.46177149e+00 -5.32723665e-01 8.06980073e-01 -3.23316574e-01 6.92555234e-02 8.15921128e-01 -8.71165991e-01 -1.67024825e-02 5.06690383e-01 1.40010929e+00 -3.33779842e-01 -1.18489778e+00 -3.74686509e-01 -8.38705003e-02 -6.77942574e-01 8.98181126e-02 -5.26136577e-01 -1.07248318e+00 1.24115562e+00 8.65730762e-01 1.87691346e-01 1.22611666e+00 3.94275993e-01 5.77315629e-01 1.56637713e-01 3.33854228e-01 -5.81725895e-01 2.42240071e-01 2.31835067e-01 5.68117261e-01 -1.77926469e+00 -1.45289749e-01 -6.14501476e-01 -7.11859941e-01 6.64494932e-01 1.00256813e+00 6.11969270e-02 7.39735484e-01 -4.05727953e-01 1.53482810e-01 -4.54421580e-01 -6.16147220e-01 -6.03238523e-01 5.54103732e-01 5.35218894e-01 4.75270063e-01 -3.84017499e-03 -9.21013877e-02 2.65365541e-01 -1.01902798e-01 -7.58173093e-02 1.14530891e-01 8.72434437e-01 -4.37493175e-01 -9.40667033e-01 -4.19002682e-01 1.99906975e-01 -3.87514114e-01 3.04605607e-02 -5.00363886e-01 8.00116360e-01 2.50100404e-01 1.13126051e+00 4.93247658e-01 -6.18048787e-01 3.93696986e-02 7.57746100e-02 6.56680226e-01 -5.55338740e-01 -2.94591129e-01 -1.30002871e-01 -2.01027378e-01 -5.85665107e-01 -1.06444085e+00 -8.00628483e-01 -1.05261803e+00 -4.01994020e-01 -9.19165909e-01 1.69487774e-01 5.56943655e-01 1.07341778e+00 5.40332317e-01 2.79388070e-01 6.22014165e-01 -1.00313902e+00 -1.27740026e-01 -1.14665079e+00 -3.83451670e-01 4.32914257e-01 7.27922738e-01 -1.15418065e+00 -3.07052791e-01 3.77513409e-01]
[14.127790451049805, 0.0425601527094841]
dda13201-e8d4-4763-adce-04663093def1
secure-deep-learning-based-distributed
2307.01559
null
https://arxiv.org/abs/2307.01559v1
https://arxiv.org/pdf/2307.01559v1.pdf
Secure Deep Learning-based Distributed Intelligence on Pocket-sized Drones
Palm-sized nano-drones are an appealing class of edge nodes, but their limited computational resources prevent running large deep-learning models onboard. Adopting an edge-fog computational paradigm, we can offload part of the computation to the fog; however, this poses security concerns if the fog node, or the communication link, can not be trusted. To tackle this concern, we propose a novel distributed edge-fog execution scheme that validates fog computation by redundantly executing a random subnetwork aboard our nano-drone. Compared to a State-of-the-Art visual pose estimation network that entirely runs onboard, a larger network executed in a distributed way improves the $R^2$ score by +0.19; in case of attack, our approach detects it within 2s with 95% probability.
['Daniele Palossi', 'Alessandro Giusti', 'Elia Cereda']
2023-07-04
null
null
null
null
['pose-estimation']
['computer-vision']
[-4.91839051e-01 6.15763724e-01 1.46392852e-01 1.01231873e-01 -5.78592271e-02 -7.32134879e-01 8.96987766e-02 -3.56499910e-01 -6.14142179e-01 8.01995218e-01 -6.43870413e-01 -3.17327410e-01 3.11244335e-02 -1.18011332e+00 -1.08533192e+00 -5.84908009e-01 -4.18627322e-01 4.46253181e-01 6.51789188e-01 -1.61251277e-01 -4.79633003e-01 4.28290039e-01 -1.41176176e+00 -2.80096263e-01 4.00372088e-01 1.46501601e+00 -4.02911961e-01 7.51823127e-01 7.07427025e-01 5.26423395e-01 -9.11283970e-01 -6.61372125e-01 7.36848235e-01 7.50301108e-02 -8.07909444e-02 -4.62481171e-01 5.24410367e-01 -6.99918747e-01 -4.03293878e-01 1.02821064e+00 4.42457110e-01 -7.84134492e-02 -5.23242503e-02 -1.85531056e+00 1.75536141e-01 2.43741468e-01 -1.25048205e-01 2.54422203e-02 3.09231947e-03 1.59036934e-01 6.67223275e-01 -1.95790648e-01 6.94612801e-01 3.81013155e-01 8.31066668e-01 4.29431289e-01 -3.88166189e-01 -9.48926449e-01 1.18832633e-01 -1.92076907e-01 -1.55608809e+00 -1.98877156e-01 4.99220490e-01 1.84640229e-01 8.67566168e-01 3.36726308e-01 1.06608200e+00 9.15784180e-01 4.59680021e-01 2.00969145e-01 8.00092995e-01 1.55200452e-01 6.75223053e-01 7.88707510e-02 -3.66597593e-01 9.75068152e-01 8.82298708e-01 5.38123369e-01 -8.87761235e-01 -2.95099080e-01 3.20325911e-01 -3.48830558e-02 -2.50552326e-01 -8.54841322e-02 -6.97458446e-01 5.07463098e-01 7.99959302e-01 -2.13164359e-01 -4.99942064e-01 7.46935904e-01 3.02807480e-01 3.27719927e-01 5.64247131e-01 6.45845979e-02 -3.74948919e-01 -2.86545843e-01 -1.19747949e+00 4.05582376e-02 1.13364685e+00 1.24095595e+00 6.62902534e-01 1.83191016e-01 4.02775258e-01 -7.42337465e-01 4.30021465e-01 7.12044418e-01 -2.56828904e-01 -5.10866821e-01 1.23654500e-01 5.13168216e-01 1.33338138e-01 -9.28411782e-01 -5.88397384e-01 -5.70482492e-01 -6.05517983e-01 6.32681310e-01 2.26179779e-01 -8.45333934e-01 -5.31476855e-01 1.23023355e+00 7.62926996e-01 5.78999221e-01 -5.99476993e-02 1.45108700e+00 5.14548540e-01 1.72008216e-01 -2.66836703e-01 1.67420253e-01 1.46079421e+00 -7.96417058e-01 -3.65411073e-01 -3.14943373e-01 3.42240453e-01 -3.58660519e-01 3.60885471e-01 6.03610098e-01 -8.41585994e-01 -5.62870502e-02 -1.45520890e+00 3.51871759e-01 -6.66682839e-01 -2.26319171e-02 8.77844214e-01 1.13442147e+00 -1.44475865e+00 4.79301393e-01 -1.00509584e+00 -4.42518443e-02 5.88041663e-01 7.69983828e-01 -1.97244599e-01 1.29350185e-01 -1.01736867e+00 6.52929723e-01 1.41092539e-01 1.30451098e-01 -1.30748808e+00 -5.64414263e-01 -3.81906927e-01 6.16173632e-02 4.43592578e-01 -9.42049503e-01 6.40576899e-01 -7.17151046e-01 -1.33807600e+00 6.22340918e-01 2.61211812e-01 -7.68083096e-01 9.85376239e-01 -3.82864624e-01 -4.68335837e-01 3.25850040e-01 -2.73323059e-01 4.14603382e-01 1.30436301e+00 -9.54657972e-01 -4.66882855e-01 -5.21378636e-01 6.38883889e-01 -2.51538366e-01 -3.67926151e-01 -2.10109621e-01 -3.44880342e-01 -1.17803037e-01 -3.11458021e-01 -1.38844395e+00 -3.21773946e-01 3.15165877e-01 -4.00844067e-01 3.73994142e-01 1.25944495e+00 -1.35356754e-01 7.16000021e-01 -2.13817000e+00 -4.17237550e-01 4.55850005e-01 8.10065925e-01 4.16663021e-01 3.78079712e-01 1.00710414e-01 5.86968303e-01 2.27629114e-02 3.92506778e-01 -6.22717619e-01 2.17284128e-01 2.93279123e-02 -3.15783590e-01 9.26817238e-01 -2.80179739e-01 8.44585598e-01 -9.85610068e-01 -1.52630076e-01 -6.20186776e-02 6.43382549e-01 -6.03334904e-01 2.24339217e-01 -2.24353537e-01 2.29049176e-02 -5.38007736e-01 9.55660582e-01 8.29629242e-01 -2.38059253e-01 4.59636837e-01 -1.02980159e-01 9.33563709e-02 2.05244496e-01 -8.96737158e-01 1.24963939e+00 -3.59866440e-01 7.87868679e-01 4.97940421e-01 -2.15789527e-01 6.82881594e-01 4.07841682e-01 1.63879663e-01 -3.44100326e-01 6.63387954e-01 1.02436483e-01 -3.39847028e-01 1.05057150e-01 6.75583005e-01 4.43197936e-02 -9.69808772e-02 7.08024800e-01 -1.99875966e-01 1.31979316e-01 -3.38150799e-01 2.06267625e-01 1.70198369e+00 -4.86760922e-02 -2.96345711e-01 -1.22269280e-01 -2.38589972e-01 1.96866587e-01 3.87620121e-01 6.75539434e-01 -4.81037736e-01 1.62297800e-01 4.05831397e-01 -7.14077652e-01 -5.35056233e-01 -1.05955267e+00 5.46018302e-01 8.58671308e-01 4.42865998e-01 -8.09278488e-01 -9.52986836e-01 -1.06628013e+00 1.42942354e-01 2.38653585e-01 -4.53011364e-01 -2.87692308e-01 -1.99130282e-01 -6.48966312e-01 1.04654741e+00 4.58961755e-01 7.06931710e-01 -3.96553189e-01 -1.45922685e+00 -2.96663679e-02 3.56403977e-01 -1.50996184e+00 -2.29526475e-01 2.29531433e-02 -5.54900289e-01 -1.33703530e+00 6.93735993e-03 -8.02567154e-02 9.05506134e-01 5.47799945e-01 1.15917039e+00 5.27531624e-01 -2.03222912e-02 3.73218119e-01 -4.80545133e-01 -5.34790218e-01 1.08399130e-01 1.32853389e-02 3.07525098e-01 -7.00097382e-02 2.37069488e-01 -6.62642241e-01 -7.99835026e-01 2.11149871e-01 -5.97785115e-01 -3.82184178e-01 2.29918346e-01 2.34871507e-01 3.87818605e-01 3.38075221e-01 7.21235648e-02 -7.86440909e-01 2.60640919e-01 -3.92958492e-01 -1.31768560e+00 -1.40261307e-01 -4.97878343e-01 -3.90871108e-01 8.78191829e-01 -2.19867259e-01 -9.45067406e-02 1.70990631e-01 4.44956005e-01 -9.28007424e-01 2.73025423e-01 1.27333209e-01 -1.73194110e-01 -9.93168712e-01 3.08359295e-01 -2.40647241e-01 -1.65803358e-01 1.23479493e-01 2.42116228e-01 2.99993277e-01 1.95977598e-01 -7.70957395e-03 1.29000950e+00 1.01640463e+00 4.54129845e-01 -6.60286307e-01 -4.22443926e-01 4.14345823e-02 -9.21787228e-04 -6.69296086e-01 6.99037075e-01 -1.24605739e+00 -1.42720699e+00 1.41898125e-01 -1.07417333e+00 -4.47395861e-01 5.79523817e-02 4.23275143e-01 -1.39365112e-02 -1.79600552e-01 -2.54113406e-01 -8.16867292e-01 -6.02028906e-01 -8.64178240e-01 1.02850020e+00 2.87957549e-01 -5.88762760e-02 -8.00295353e-01 -2.44487792e-01 3.28872710e-01 6.70131981e-01 6.63293839e-01 -5.01825698e-02 -5.02484858e-01 -1.15667021e+00 -8.24305594e-01 -1.61342248e-01 -1.07564881e-01 -3.27618450e-01 3.11424788e-02 -1.15799463e+00 -6.04835689e-01 -2.08819911e-01 -3.92622426e-02 3.80343020e-01 -7.44136889e-03 7.46682048e-01 -3.41263682e-01 -4.73839402e-01 9.95704353e-01 1.44023728e+00 -2.39979610e-01 4.98516709e-01 2.45967835e-01 6.12317502e-01 -1.71319291e-01 6.52300298e-01 6.50077105e-01 4.47564483e-01 3.78748804e-01 1.39968324e+00 -1.19544417e-01 2.17354804e-01 -2.28750065e-01 6.33883178e-01 6.39911368e-02 -4.14453149e-01 -6.79718375e-01 -5.20454347e-01 1.12784907e-01 -1.52945495e+00 -4.38176751e-01 4.47373325e-03 2.22012281e+00 -6.34435341e-02 6.18063509e-01 2.17948049e-01 -1.88029051e-01 4.77607638e-01 4.08509016e-01 -4.26531434e-01 -3.06047648e-01 1.50945827e-01 2.96924084e-01 1.34680068e+00 1.74721673e-01 -1.02086616e+00 1.00784791e+00 6.27497721e+00 4.65604961e-01 -1.41954327e+00 4.64978993e-01 2.59673387e-01 -7.07825959e-01 1.28666414e-02 1.18618473e-01 -8.60587895e-01 7.48565555e-01 1.28425574e+00 -1.69800278e-02 5.83119512e-01 1.23686326e+00 9.60842669e-02 -2.73691863e-01 -7.49629319e-01 7.76308715e-01 1.35956973e-01 -1.49625075e+00 -4.66950566e-01 5.87651849e-01 5.16422749e-01 5.84256291e-01 -2.00875968e-01 -5.21827266e-02 5.39536238e-01 -8.49204123e-01 9.20151472e-01 1.98193908e-01 8.46074402e-01 -9.83727038e-01 9.55353141e-01 3.90027642e-01 -1.25597215e+00 2.66098708e-01 -4.58110869e-01 -4.31484550e-01 1.71366856e-01 8.49878490e-01 -8.94297123e-01 5.30276418e-01 9.47937012e-01 -1.28389284e-01 -5.13163209e-01 7.43480504e-01 -5.99350095e-01 5.82598090e-01 -8.07670593e-01 -3.25805306e-01 2.45936513e-01 -6.64477721e-02 7.70172238e-01 5.53463757e-01 4.03286934e-01 3.66743281e-02 1.39802963e-01 8.05820584e-01 -4.42417592e-01 -7.23808587e-01 -8.36171627e-01 7.79576674e-02 6.07537091e-01 1.55012751e+00 -8.46948266e-01 -2.54899740e-01 -1.88187137e-01 1.09449375e+00 2.23830581e-01 7.25675672e-02 -1.35871029e+00 -3.54676753e-01 1.02933300e+00 3.63321126e-01 3.39946538e-01 -3.93229157e-01 -1.20761142e-04 -9.69355524e-01 2.63646603e-01 -4.58851129e-01 8.39309618e-02 -7.24519253e-01 -7.32113600e-01 9.90940213e-01 -6.01831138e-01 -1.42489648e+00 1.49385119e-02 -5.66653550e-01 -7.12652266e-01 1.45362288e-01 -1.34754348e+00 -1.49978197e+00 -7.24568605e-01 6.77857995e-01 -5.06818235e-01 4.34337072e-02 7.71150053e-01 2.86503851e-01 -4.25797135e-01 9.96072292e-01 -2.79720843e-01 2.66284794e-01 2.88034588e-01 -7.96847105e-01 5.63926816e-01 1.31342411e+00 9.31373239e-03 2.01978430e-01 6.52323842e-01 -8.30663562e-01 -2.03510141e+00 -1.29429495e+00 4.74733979e-01 -3.82169813e-01 6.97342038e-01 -7.94962943e-01 1.63388997e-02 7.45161474e-01 -1.81050561e-02 7.87368953e-01 4.85218853e-01 -2.27485389e-01 -2.35811010e-01 -3.34937066e-01 -1.47190666e+00 5.56241930e-01 1.17949212e+00 -4.92240816e-01 2.11933732e-01 4.61537004e-01 7.49183238e-01 -6.31656706e-01 -7.38240540e-01 1.14669994e-01 6.47842586e-01 -1.07668591e+00 6.22264683e-01 -4.18739319e-01 -3.46831381e-01 -6.29466474e-01 -3.58058289e-02 -1.12420368e+00 3.07282627e-01 -1.21519971e+00 -4.36019748e-01 6.45335555e-01 1.28201336e-01 -1.04439378e+00 1.40047193e+00 7.13162303e-01 -1.67774364e-01 -3.73153359e-01 -1.46941841e+00 -9.92511630e-01 -7.10910082e-01 -5.09587049e-01 8.34758222e-01 5.57426274e-01 -9.69391987e-02 -1.20772034e-01 -3.60806137e-01 8.64024043e-01 7.04727829e-01 -1.47192618e-02 1.18624103e+00 -1.10685730e+00 -3.38887721e-01 1.58557639e-01 -9.68706012e-01 -6.21554315e-01 4.68370132e-02 -3.99326503e-01 -9.16784629e-02 -8.65609825e-01 -6.97978854e-01 -4.40616071e-01 -8.07604566e-02 6.18842483e-01 3.73739809e-01 8.14996719e-01 4.44597691e-01 -3.37669820e-01 -1.12040806e+00 1.85555637e-01 8.10964406e-01 -5.83939217e-02 1.27259731e-01 1.46189809e-01 -3.81592035e-01 7.99952030e-01 8.36333334e-01 -7.22664356e-01 -3.71108502e-01 -3.87033641e-01 7.57571340e-01 -4.81369197e-02 8.15865219e-01 -1.55385995e+00 6.37733757e-01 2.42881596e-01 3.45877320e-01 -2.53642321e-01 5.71648121e-01 -1.41911340e+00 5.00034213e-01 5.42614162e-01 7.98194885e-01 3.21984857e-01 2.92055637e-01 5.94256580e-01 1.69839457e-01 3.75595182e-01 2.26417035e-01 -3.04554217e-02 -3.24405998e-01 5.42923093e-01 -5.32262325e-01 -1.11589588e-01 1.33082306e+00 -1.98686421e-01 -8.82976890e-01 -4.79965121e-01 -4.25085396e-01 1.10534325e-01 9.38152015e-01 1.32539049e-01 5.11553288e-01 -7.74056911e-01 -1.39174759e-01 4.02484089e-01 -2.24089801e-01 3.70019339e-02 9.21746120e-02 9.34899807e-01 -8.46514761e-01 -4.15722234e-03 -3.23481001e-02 -1.33134052e-01 -1.25890839e+00 3.67354333e-01 3.90290856e-01 -9.56258774e-02 -6.01001859e-01 1.08907211e+00 -4.54641491e-01 1.71240732e-01 7.45731965e-02 -1.16259374e-01 4.94850248e-01 -5.51507808e-02 5.22342980e-01 4.29807603e-01 4.97230440e-01 -4.41366106e-01 -8.05526197e-01 1.74215123e-01 3.32663298e-01 -2.48695491e-03 1.11182094e+00 -1.66745074e-02 -1.64103627e-01 -3.81874412e-01 7.45987058e-01 3.77542645e-01 -1.43605888e+00 4.74114180e-01 -5.59431791e-01 -5.24640203e-01 4.84446853e-01 -4.84551847e-01 -1.66077673e+00 3.13738316e-01 5.49866378e-01 3.48830670e-01 1.23515844e+00 -1.87299564e-01 1.03343689e+00 5.30768275e-01 1.14348841e+00 -9.78885472e-01 -2.88584560e-01 3.18934709e-01 7.78027810e-03 -9.58670497e-01 2.79368162e-01 -5.98135769e-01 -7.95139670e-02 8.74879837e-01 7.45477796e-01 -6.11152172e-01 9.61917579e-01 6.90367758e-01 5.53662553e-02 -6.90387249e-01 -7.61955619e-01 -1.26425862e-01 -2.58794248e-01 7.46780992e-01 -4.63936120e-01 3.20794165e-01 3.12151760e-01 8.13746333e-01 -6.01550579e-01 1.09777249e-01 6.10034108e-01 1.15217781e+00 -3.37460607e-01 -4.68986064e-01 -2.70733804e-01 3.02836657e-01 -2.81830937e-01 -8.23631063e-02 -5.14380395e-01 8.62820685e-01 3.39215130e-01 1.01162732e+00 2.71748900e-01 -9.97015119e-01 -6.03684932e-02 -4.04411316e-01 2.12298095e-01 -2.39647046e-01 -9.65505123e-01 -2.30489418e-01 3.24595988e-01 -1.09076893e+00 -1.21707097e-01 5.42851873e-02 -1.18735075e+00 -9.13327992e-01 -4.22270417e-01 1.14508390e-01 7.46729076e-01 8.85001600e-01 7.22180843e-01 4.58410114e-01 5.89247763e-01 -9.74125803e-01 -5.79063147e-02 -2.86995351e-01 -7.29764640e-01 -4.91400659e-01 1.82734936e-01 -5.97456276e-01 -7.30277121e-01 -4.46968585e-01]
[8.1185302734375, 2.36484694480896]
473deff2-2a11-41f1-82a0-67d0cbb1b082
efficient-multi-task-scene-analysis-with-rgb
2306.05242
null
https://arxiv.org/abs/2306.05242v1
https://arxiv.org/pdf/2306.05242v1.pdf
Efficient Multi-Task Scene Analysis with RGB-D Transformers
Scene analysis is essential for enabling autonomous systems, such as mobile robots, to operate in real-world environments. However, obtaining a comprehensive understanding of the scene requires solving multiple tasks, such as panoptic segmentation, instance orientation estimation, and scene classification. Solving these tasks given limited computing and battery capabilities on mobile platforms is challenging. To address this challenge, we introduce an efficient multi-task scene analysis approach, called EMSAFormer, that uses an RGB-D Transformer-based encoder to simultaneously perform the aforementioned tasks. Our approach builds upon the previously published EMSANet. However, we show that the dual CNN-based encoder of EMSANet can be replaced with a single Transformer-based encoder. To achieve this, we investigate how information from both RGB and depth data can be effectively incorporated in a single encoder. To accelerate inference on robotic hardware, we provide a custom NVIDIA TensorRT extension enabling highly optimization for our EMSAFormer approach. Through extensive experiments on the commonly used indoor datasets NYUv2, SUNRGB-D, and ScanNet, we show that our approach achieves state-of-the-art performance while still enabling inference with up to 39.1 FPS on an NVIDIA Jetson AGX Orin 32 GB.
['Horst-Michael Gross', 'Leonard Rabes', 'Robin Schmidt', 'Daniel Seichter', 'Söhnke Benedikt Fischedick']
2023-06-08
null
null
null
null
['panoptic-segmentation', 'scene-classification']
['computer-vision', 'computer-vision']
[ 1.89618960e-01 -3.95843863e-01 1.07276358e-01 -5.06192923e-01 -3.17891866e-01 -7.37914741e-01 2.76590288e-01 -1.20150827e-01 -8.19672823e-01 4.19094294e-01 -6.16698802e-01 -7.63174713e-01 7.32009485e-02 -9.20527875e-01 -1.24421847e+00 -4.77497667e-01 9.14400518e-02 2.75170535e-01 5.66168189e-01 -1.98072597e-01 3.32111210e-01 4.99857992e-01 -1.91436291e+00 6.37972951e-02 7.79259145e-01 1.43662024e+00 5.91324270e-01 1.01805174e+00 1.64112255e-01 8.81872237e-01 -3.56626004e-01 -1.55727223e-01 4.93941247e-01 3.63880366e-01 -8.01708043e-01 -6.31156703e-03 6.12932205e-01 -8.38169158e-01 -2.94673264e-01 7.88890541e-01 3.17730725e-01 1.21121310e-01 2.02525053e-02 -1.29605424e+00 1.29362509e-01 -1.08985707e-01 -3.66747409e-01 9.76887792e-02 6.96884394e-02 2.45789304e-01 5.11052549e-01 -6.24125540e-01 4.26777810e-01 1.20054698e+00 5.93202949e-01 5.28448932e-02 -8.07797432e-01 -3.87735128e-01 1.89729836e-02 3.64635915e-01 -1.31153309e+00 -3.62874061e-01 2.94193298e-01 -1.86860517e-01 1.36600089e+00 1.01547286e-01 6.96328402e-01 6.74328983e-01 3.52739573e-01 6.92021608e-01 8.70581448e-01 1.00540690e-01 3.98713589e-01 -2.03073218e-01 -1.31922841e-01 9.88147438e-01 3.20862949e-01 -2.42281511e-01 -4.53634858e-01 2.70077765e-01 9.31804836e-01 -6.61075637e-02 6.50633425e-02 -4.22814488e-01 -1.29364383e+00 6.85816348e-01 7.84728229e-01 -3.55062753e-01 -3.22541416e-01 6.13681912e-01 4.89652067e-01 5.49370088e-02 2.73306131e-01 1.76002055e-01 -5.45285583e-01 -5.10999203e-01 -5.20732224e-01 2.41770148e-01 9.12549078e-01 1.16090691e+00 1.11053681e+00 -4.58276793e-02 3.13530236e-01 4.63529646e-01 4.37058568e-01 7.98201740e-01 3.14859599e-01 -1.28305995e+00 5.27606428e-01 4.22317982e-01 2.37981919e-02 -7.12867975e-01 -5.52089810e-01 -9.38766822e-02 -6.16712093e-01 3.52601349e-01 2.39784539e-01 -2.83786684e-01 -9.19184864e-01 1.08914685e+00 6.63900912e-01 1.59741580e-01 1.12932011e-01 1.11994517e+00 6.93174183e-01 5.99842668e-01 -2.72788823e-01 5.14492512e-01 1.49207771e+00 -1.24908113e+00 -2.51170605e-01 -5.04999220e-01 7.18041241e-01 -6.75774097e-01 9.37859535e-01 5.13562322e-01 -7.18932331e-01 -4.80734438e-01 -1.44151688e+00 -7.07788289e-01 -3.79894197e-01 4.39268202e-01 1.07895792e+00 5.60213506e-01 -1.04265058e+00 5.72510123e-01 -1.44004667e+00 -3.01724941e-01 3.70412797e-01 6.45700276e-01 -2.86394864e-01 -2.09726661e-01 -6.65933192e-01 8.69766116e-01 4.57160383e-01 2.06905380e-01 -8.25976849e-01 -5.18243611e-01 -9.90866721e-01 -4.78722081e-02 6.35769784e-01 -7.59963930e-01 1.46374846e+00 -4.11995053e-01 -1.91833293e+00 4.17721987e-01 -1.16466612e-01 -7.01390684e-01 2.81245679e-01 -4.40253973e-01 1.85557559e-01 2.72528589e-01 -3.79070304e-02 8.45054448e-01 6.91703975e-01 -8.33756685e-01 -7.88591027e-01 -4.21744585e-01 5.63257635e-01 4.14146364e-01 -2.02374414e-01 -3.02765429e-01 -6.53361440e-01 1.39553603e-02 1.33931130e-01 -1.25092912e+00 -3.41801435e-01 4.62937415e-01 -4.60428804e-01 4.69677858e-02 1.20017660e+00 -4.06451374e-01 3.79355878e-01 -2.18469000e+00 -5.67972194e-04 6.24759719e-02 1.78519279e-01 3.03457439e-01 2.12425724e-01 -7.42986053e-02 5.90226531e-01 -3.15060526e-01 -1.03033595e-01 -6.56022251e-01 -2.22286545e-02 6.53977871e-01 -1.29349187e-01 5.87067366e-01 6.54671416e-02 5.78995347e-01 -6.97660089e-01 -4.50733095e-01 7.38748074e-01 6.84325337e-01 -8.15468967e-01 4.77880724e-02 -2.30039731e-01 3.86193812e-01 -3.77332389e-01 7.75842905e-01 8.01886141e-01 -3.63511741e-01 1.12754032e-01 -1.94682419e-01 -5.18447638e-01 5.12320280e-01 -1.09343493e+00 2.10002708e+00 -7.67681539e-01 9.27208364e-01 3.84346098e-01 -1.00717092e+00 6.93817973e-01 -2.03910053e-01 3.52921277e-01 -8.66434097e-01 3.41451943e-01 4.34267700e-01 -1.86472341e-01 -2.80168712e-01 1.06854236e+00 3.27351093e-01 -7.77897760e-02 8.74179453e-02 4.22409326e-02 -4.62083906e-01 2.89900303e-01 6.53818175e-02 1.14098179e+00 3.74719441e-01 3.87567319e-02 -2.41635203e-01 2.00675607e-01 4.42399561e-01 4.21202183e-01 6.20633066e-01 -8.22690055e-02 1.91322669e-01 3.12103927e-01 -4.77091253e-01 -1.15789688e+00 -9.34750617e-01 -2.08633810e-01 8.84973943e-01 4.63380188e-01 -4.84162927e-01 -5.87177336e-01 -2.54622281e-01 1.49303377e-01 2.52285689e-01 -1.22845225e-01 2.34237224e-01 -6.25269353e-01 -6.75784171e-01 5.52902341e-01 6.61341131e-01 1.07370067e+00 -4.59745198e-01 -1.60809612e+00 1.99439406e-01 1.08065203e-01 -1.71074963e+00 1.07853234e-01 4.11524087e-01 -7.57183015e-01 -1.02879679e+00 -1.50921434e-01 -3.79834145e-01 2.78998733e-01 6.99722409e-01 8.16397130e-01 -1.18169136e-01 -4.03059751e-01 2.67856926e-01 -2.33369917e-01 -3.80342424e-01 1.50503725e-01 2.75357634e-01 3.07756253e-02 -4.27825570e-01 -5.56646548e-02 -5.07206738e-01 -7.38477468e-01 1.52435064e-01 -7.77944922e-01 4.43466276e-01 5.38230836e-01 4.41585392e-01 6.70208812e-01 -9.25704017e-02 -1.48500293e-01 -4.17651802e-01 -8.21552724e-02 -4.82092083e-01 -1.14629233e+00 -8.64294022e-02 -3.64038348e-01 1.27153829e-01 7.23325908e-01 -1.09049298e-01 -9.20782506e-01 4.87735391e-01 -3.99187416e-01 -4.99347240e-01 1.51569575e-01 4.13714767e-01 1.32777631e-01 -4.80021685e-01 2.61532575e-01 6.53301552e-02 1.50333837e-01 -2.09787622e-01 2.85795569e-01 8.66953671e-01 8.70424390e-01 -4.79532987e-01 3.04128796e-01 8.25895786e-01 2.96517849e-01 -1.06202304e+00 -5.58905482e-01 -4.40918982e-01 -5.22675455e-01 -2.78249800e-01 8.84757340e-01 -1.23104310e+00 -1.33297491e+00 6.47312701e-01 -1.18271935e+00 -6.95031047e-01 2.18717530e-01 5.82478464e-01 -6.10612273e-01 1.88279405e-01 -4.66051936e-01 -4.79774892e-01 -3.54509175e-01 -1.73941660e+00 1.54789007e+00 3.34491581e-01 1.52254239e-01 -6.33568287e-01 -1.52154163e-01 3.24160665e-01 5.25508821e-01 3.46431620e-02 4.51437414e-01 -1.75874811e-02 -1.12730122e+00 -1.57539435e-02 -5.04314303e-01 1.44513860e-01 -9.03639346e-02 5.64441308e-02 -9.34194684e-01 -2.47482747e-01 -1.85157329e-01 -5.54303050e-01 6.79308593e-01 1.58688486e-01 1.46319795e+00 5.63066527e-02 -1.91148415e-01 1.19537449e+00 1.56680751e+00 7.79146776e-02 6.66570604e-01 6.18309915e-01 1.06174231e+00 1.24053955e-01 7.99222410e-01 5.37977219e-01 9.43780541e-01 9.36938524e-01 9.14559066e-01 1.34147406e-01 4.79608998e-02 1.73328638e-01 2.16782942e-01 6.46605611e-01 -2.20276453e-02 -2.45093733e-01 -8.57099771e-01 1.99828655e-01 -1.89154255e+00 -4.15540636e-01 -2.71228880e-01 2.02410412e+00 3.63011569e-01 1.49229381e-04 -2.96528071e-01 4.72729467e-02 2.09373832e-01 7.16787949e-02 -8.51642668e-01 -5.29415727e-01 1.42672837e-01 2.43180618e-01 1.12383902e+00 3.46067339e-01 -1.28145540e+00 9.78593230e-01 5.54329634e+00 5.10307610e-01 -1.54194057e+00 3.81332636e-02 2.19993457e-01 -3.37848276e-01 9.74528715e-02 1.07089095e-02 -8.52699816e-01 3.85824680e-01 1.16313446e+00 1.26258269e-01 6.37414634e-01 1.31051302e+00 5.20776622e-02 -6.46246970e-01 -8.70197594e-01 1.15409851e+00 -7.73932338e-02 -1.25065947e+00 -4.16752726e-01 1.97356060e-01 2.52463520e-01 6.00728333e-01 -1.14131138e-01 2.56597877e-01 3.23249102e-01 -7.33313978e-01 8.45926404e-01 1.23732381e-01 8.05076659e-01 -7.15254843e-01 5.63437521e-01 3.06885272e-01 -1.35589504e+00 6.24028668e-02 -5.21041751e-01 -3.37627351e-01 2.54777938e-01 6.73388124e-01 -9.60678399e-01 6.67860389e-01 1.01684153e+00 7.65296340e-01 -5.34971118e-01 8.46926093e-01 -7.80955404e-02 1.02909692e-01 -8.41562569e-01 -7.22423121e-02 3.50371361e-01 -4.45803953e-03 2.07807094e-01 8.29494178e-01 6.37258589e-01 -6.44318713e-03 2.39358664e-01 4.95781273e-01 2.17119623e-02 -3.97948891e-01 -5.10022044e-01 2.29611814e-01 4.52070266e-01 1.43768859e+00 -9.17321682e-01 -4.87150729e-01 -3.66328627e-01 1.26469076e+00 4.75343645e-01 -1.28233507e-01 -1.10196280e+00 -5.08913457e-01 9.89439428e-01 -2.96553284e-01 5.03128529e-01 -9.37423587e-01 -4.08493817e-01 -1.27317274e+00 2.63605356e-01 -4.74209011e-01 -1.43253237e-01 -9.64277148e-01 -4.52073425e-01 5.65434754e-01 -1.54599234e-01 -1.12826860e+00 -3.77271265e-01 -1.08737350e+00 -2.01939344e-01 5.52190363e-01 -1.69926274e+00 -8.77960205e-01 -8.92696500e-01 6.57404721e-01 4.98037249e-01 3.16618562e-01 6.27199411e-01 4.51290518e-01 -7.13111937e-01 3.39303017e-02 7.66058341e-02 -1.98693648e-01 3.55758190e-01 -1.29852641e+00 6.18277669e-01 9.31670249e-01 -1.67651251e-01 3.39847863e-01 5.09645462e-01 -5.93344629e-01 -2.31562495e+00 -1.18498969e+00 1.39233604e-01 -1.09202169e-01 5.05140841e-01 -4.02403235e-01 -5.49519598e-01 8.19175541e-01 7.20591471e-02 3.43631297e-01 2.44470015e-01 -1.38161570e-01 -8.18658397e-02 -1.93058461e-01 -9.75101650e-01 4.21698809e-01 1.06380248e+00 -3.68837476e-01 9.24124476e-03 2.90025890e-01 8.07259381e-01 -1.13778222e+00 -8.54009330e-01 5.10930955e-01 7.47700751e-01 -1.02408481e+00 1.03761804e+00 1.21081121e-01 5.41377962e-01 -4.08672631e-01 -4.44326758e-01 -1.15413737e+00 2.68233120e-01 -2.94661254e-01 -2.71958351e-01 5.86532474e-01 -5.96733019e-02 -7.36495435e-01 7.79505789e-01 5.02844036e-01 -6.10007107e-01 -7.58347392e-01 -1.05968273e+00 -6.97830677e-01 -4.92633015e-01 -7.87642837e-01 6.60833597e-01 4.12954122e-01 -3.02863151e-01 3.35565299e-01 -2.38332972e-01 4.71159548e-01 6.40942216e-01 1.67146146e-01 1.17051327e+00 -8.78722310e-01 -1.46049619e-01 -2.81343423e-02 -5.46553135e-01 -1.42854869e+00 -1.82380923e-03 -5.31213820e-01 3.78296077e-01 -1.55833316e+00 -2.51370758e-01 -6.08716369e-01 4.30815667e-01 5.57667434e-01 1.48891956e-01 4.32394803e-01 2.69040585e-01 7.82128647e-02 -8.64579678e-01 5.37086546e-01 1.03682435e+00 4.25335579e-02 2.50145793e-02 -3.01497459e-01 -2.29623750e-01 7.97648430e-01 6.80596411e-01 -8.71396586e-02 -4.40732479e-01 -1.08692265e+00 3.61888081e-01 5.16484454e-02 7.43953109e-01 -1.45411384e+00 5.45939982e-01 1.03545450e-01 2.57296264e-01 -8.13122153e-01 8.12334180e-01 -6.68335438e-01 -1.14477791e-01 5.29923439e-01 4.66476619e-01 2.23414212e-01 4.97715324e-01 1.85358241e-01 -9.42267030e-02 1.38879463e-01 4.67661798e-01 -5.72304567e-03 -1.31959665e+00 1.98895082e-01 -5.09425521e-01 -4.26341414e-01 9.43798363e-01 -4.04845439e-02 -7.78022468e-01 -7.34876618e-02 -1.58608928e-01 4.47846204e-01 5.54177403e-01 3.06928158e-01 5.92012465e-01 -1.03905451e+00 -1.97947711e-01 2.70074189e-01 -8.88463706e-02 7.66693830e-01 3.31768185e-01 9.21844244e-01 -1.28092456e+00 6.78251088e-01 -3.41133803e-01 -1.06573009e+00 -1.09347260e+00 1.49213716e-01 2.66582787e-01 5.36624230e-02 -6.15825474e-01 7.35663772e-01 -1.04967989e-01 -5.02480924e-01 -3.64540555e-02 -7.32888520e-01 2.08698586e-01 -2.98267692e-01 3.29768389e-01 4.96563792e-01 4.74608362e-01 -3.48284543e-01 -5.39724827e-01 3.50422233e-01 1.79022983e-01 -6.08405881e-02 1.16821456e+00 -2.49862343e-01 -1.15455642e-01 2.13141531e-01 1.33733988e+00 -4.58722264e-01 -1.61872411e+00 1.85395211e-01 -4.12269950e-01 -4.94373113e-01 3.52874398e-01 -3.52120966e-01 -1.00694263e+00 9.38677371e-01 5.79951108e-01 -9.54831168e-02 1.06884098e+00 -2.41689160e-01 1.08917356e+00 8.82651627e-01 9.19015944e-01 -1.00100350e+00 -3.92644294e-02 8.30987334e-01 3.86130512e-01 -1.42419171e+00 1.24713182e-01 -6.52433395e-01 -3.43392074e-01 1.11953115e+00 8.02775681e-01 -1.96584910e-01 3.42404932e-01 4.94890958e-01 -2.09610060e-01 -2.22825542e-01 -5.72260916e-01 -3.09977114e-01 -8.29427838e-02 1.92180559e-01 1.04111014e-02 8.56228918e-02 2.21698657e-01 3.14038768e-02 -5.98326325e-01 7.50675276e-02 5.37332833e-01 1.44061053e+00 -3.70291173e-01 -7.94719517e-01 -2.48944670e-01 4.26886857e-01 -2.64517814e-01 -3.04940511e-02 2.10839584e-01 7.37891972e-01 1.90169904e-02 8.30091596e-01 5.96728861e-01 -6.60341501e-01 1.88446164e-01 -4.76570725e-01 7.07129240e-01 -4.44754213e-01 -2.69773811e-01 -1.88146114e-01 2.14958549e-01 -1.16328263e+00 -3.87376994e-01 -5.43514490e-01 -1.26873124e+00 -5.82303166e-01 4.85399440e-02 -4.38235968e-01 1.53826165e+00 1.07181311e+00 5.59054017e-01 7.43333340e-01 3.84247929e-01 -1.34618664e+00 -2.83313364e-01 -7.16576397e-01 -2.72530377e-01 -2.78774410e-01 3.68490577e-01 -7.87778020e-01 3.37909795e-02 -2.51620770e-01]
[8.539623260498047, -2.282526969909668]
1c9a6bba-c51a-4ed2-8e2a-66e1b5d67a2d
float-factorized-learning-of-object
2203.16168
null
https://arxiv.org/abs/2203.16168v1
https://arxiv.org/pdf/2203.16168v1.pdf
FLOAT: Factorized Learning of Object Attributes for Improved Multi-object Multi-part Scene Parsing
Multi-object multi-part scene parsing is a challenging task which requires detecting multiple object classes in a scene and segmenting the semantic parts within each object. In this paper, we propose FLOAT, a factorized label space framework for scalable multi-object multi-part parsing. Our framework involves independent dense prediction of object category and part attributes which increases scalability and reduces task complexity compared to the monolithic label space counterpart. In addition, we propose an inference-time 'zoom' refinement technique which significantly improves segmentation quality, especially for smaller objects/parts. Compared to state of the art, FLOAT obtains an absolute improvement of 2.0% for mean IOU (mIOU) and 4.8% for segmentation quality IOU (sqIOU) on the Pascal-Part-58 dataset. For the larger Pascal-Part-108 dataset, the improvements are 2.1% for mIOU and 3.9% for sqIOU. We incorporate previously excluded part attributes and other minor parts of the Pascal-Part dataset to create the most comprehensive and challenging version which we dub Pascal-Part-201. FLOAT obtains improvements of 8.6% for mIOU and 7.5% for sqIOU on the new dataset, demonstrating its parsing effectiveness across a challenging diversity of objects and parts. The code and datasets are available at floatseg.github.io.
['Ravikiran Sarvadevabhatla', 'Pradeep Shenoy', 'Pranav Gupta', 'Rishubh Singh']
2022-03-30
null
http://openaccess.thecvf.com//content/CVPR2022/html/Singh_FLOAT_Factorized_Learning_of_Object_Attributes_for_Improved_Multi-Object_Multi-Part_CVPR_2022_paper.html
http://openaccess.thecvf.com//content/CVPR2022/papers/Singh_FLOAT_Factorized_Learning_of_Object_Attributes_for_Improved_Multi-Object_Multi-Part_CVPR_2022_paper.pdf
cvpr-2022-1
['scene-parsing', '2d-semantic-segmentation']
['computer-vision', 'computer-vision']
[ 3.99549991e-01 1.81863621e-01 -1.34233847e-01 -5.31415880e-01 -1.24265170e+00 -7.87534535e-01 1.26111448e-01 1.86225906e-01 -3.76570314e-01 3.84513408e-01 -2.26831302e-01 6.20344393e-02 1.14992909e-01 -5.14687538e-01 -1.07990873e+00 -3.70188445e-01 3.14690232e-01 7.16494799e-01 1.01412237e+00 9.01282728e-02 6.65093064e-02 2.93548554e-01 -1.65151882e+00 4.54674631e-01 8.60863566e-01 1.09009016e+00 3.97761345e-01 9.51227963e-01 -3.61465514e-01 5.68414927e-01 -5.95562875e-01 -5.06532192e-01 4.53103334e-01 -1.73412293e-01 -1.15983629e+00 2.24100158e-01 1.16892338e+00 -1.25721648e-01 1.93278238e-01 9.25216556e-01 2.34455779e-01 -7.68697076e-03 2.59224802e-01 -1.27837765e+00 -1.30713850e-01 4.85414326e-01 -7.40924180e-01 -6.27513677e-02 9.90270823e-02 -7.98463821e-02 1.12898731e+00 -5.69259882e-01 6.56965017e-01 1.39687455e+00 7.52372980e-01 5.53494871e-01 -1.14418936e+00 -4.32590276e-01 2.69804388e-01 7.93341249e-02 -1.10286784e+00 -2.53533214e-01 2.41560116e-01 -4.05099422e-01 1.05829322e+00 3.92447442e-01 3.59723091e-01 3.68924290e-01 1.59468263e-01 1.25952423e+00 1.03967345e+00 -2.25375831e-01 1.12764038e-01 -1.70720086e-01 5.36909938e-01 7.53529310e-01 3.20159644e-01 -5.21519601e-01 -1.10710464e-01 1.21046439e-01 3.58363569e-01 3.72170582e-02 1.85275778e-01 -4.14671123e-01 -1.10809493e+00 6.75755918e-01 5.70413411e-01 2.61135958e-03 -5.05540408e-02 2.24734485e-01 4.61035758e-01 -1.33386940e-01 4.16442215e-01 4.14594024e-01 -1.04954958e+00 -2.00994998e-01 -8.34526837e-01 4.09295559e-01 7.61293530e-01 1.35362518e+00 1.05286205e+00 -4.05079335e-01 -1.17622592e-01 1.19610858e+00 4.58914451e-02 5.02508819e-01 1.86303660e-01 -1.48092067e+00 6.20670199e-01 1.02041185e+00 -3.44137251e-02 -2.91293710e-01 -7.09311366e-01 -3.66586059e-01 -3.43868315e-01 1.54482916e-01 6.13384962e-01 7.46100917e-02 -1.34347701e+00 1.47975588e+00 7.48390496e-01 -2.08727330e-01 -1.06880940e-01 5.77260554e-01 9.28241372e-01 5.18966734e-01 3.21550846e-01 2.65247971e-01 1.73364520e+00 -1.56773400e+00 -3.32050771e-01 -7.08834350e-01 7.17550695e-01 -8.97608876e-01 1.17450225e+00 3.33654255e-01 -9.21685576e-01 -7.54365206e-01 -7.15705574e-01 -4.32766408e-01 -3.30463022e-01 1.78587690e-01 7.87522793e-01 4.36372399e-01 -8.48151922e-01 5.21356463e-01 -8.79697800e-01 -2.97286302e-01 7.99700975e-01 4.29938316e-01 -3.47352237e-01 -4.53436852e-01 -3.55869979e-01 3.70788425e-01 7.81461060e-01 -3.77247512e-01 -5.46699226e-01 -8.56056094e-01 -1.06674147e+00 6.24079630e-02 8.49132955e-01 -4.32274014e-01 1.61835027e+00 -8.13159108e-01 -1.16676581e+00 8.43988061e-01 -2.32218832e-01 -3.77742916e-01 5.58340967e-01 -5.40229380e-01 -7.64094889e-02 2.40569860e-01 6.49882555e-01 1.40095139e+00 3.77587378e-01 -1.15200555e+00 -1.18437994e+00 -3.29932302e-01 2.67159998e-01 9.66951624e-02 1.99408740e-01 1.38126332e-02 -1.02554059e+00 -4.36017036e-01 3.27549517e-01 -1.11977816e+00 -2.97140896e-01 8.65761414e-02 -5.49796283e-01 -3.57559323e-01 9.61719453e-01 -5.95713079e-01 8.20115864e-01 -2.18736911e+00 -1.34078398e-01 -4.33509260e-01 1.30710021e-01 2.78399885e-01 -3.13847512e-01 -7.42694810e-02 1.46990031e-01 -1.55689418e-02 -6.14388049e-01 -6.45700276e-01 -8.52051526e-02 4.57031786e-01 3.24487090e-01 1.48566976e-01 4.41295594e-01 9.64499116e-01 -6.47232294e-01 -7.64362931e-01 3.40111405e-01 1.72086820e-01 -9.22494829e-01 9.09612179e-02 -4.58156317e-01 2.84131348e-01 -2.97851026e-01 9.81249094e-01 8.89682710e-01 -3.21046084e-01 -1.61012277e-01 -1.18127204e-01 -9.11001340e-02 2.91616857e-01 -1.30317235e+00 1.90658915e+00 -3.51193994e-01 3.41689110e-01 1.28342181e-01 -5.02428830e-01 6.28835022e-01 -1.81970879e-01 5.68083346e-01 -7.18540192e-01 4.19123247e-02 1.77857265e-01 9.81193334e-02 -1.30068853e-01 6.86064541e-01 6.51013553e-02 -4.72629160e-01 1.35984374e-02 1.92732170e-01 -3.67018342e-01 7.94360280e-01 2.11754814e-01 1.35862660e+00 3.85574877e-01 3.36745530e-01 -4.50322717e-01 3.89790595e-01 4.63065237e-01 8.89232576e-01 6.76940084e-01 -4.46138233e-01 8.53585422e-01 4.95709270e-01 -3.00901175e-01 -9.06958163e-01 -1.02247274e+00 -4.23281938e-01 1.26139152e+00 3.97386998e-01 -4.76696551e-01 -1.12990952e+00 -1.01343250e+00 6.76921234e-02 6.83929861e-01 -3.94025832e-01 3.95083040e-01 -7.82826483e-01 -6.68377280e-01 2.26510808e-01 8.60868752e-01 7.99360156e-01 -1.11967373e+00 -7.11554527e-01 2.47724742e-01 -2.33619317e-01 -1.61372030e+00 -4.80875850e-01 2.58301407e-01 -8.56781781e-01 -1.31293976e+00 -5.19555926e-01 -7.25051641e-01 5.23455679e-01 1.90805167e-01 1.38035214e+00 -9.62717906e-02 -5.14206469e-01 1.54597327e-01 -5.22973418e-01 -3.96399230e-01 -4.64241505e-01 1.81964219e-01 -5.19735515e-01 -3.74202013e-01 9.16067883e-02 -1.95581503e-02 -5.51321745e-01 4.98676687e-01 -7.87964165e-01 3.97305101e-01 8.17093194e-01 5.00012636e-01 1.08291650e+00 -1.23726107e-01 1.46346554e-01 -1.29495692e+00 -4.90386426e-01 -2.41812691e-01 -8.47906053e-01 1.58623144e-01 -3.28328192e-01 5.12539819e-02 5.94176590e-01 -1.84890553e-01 -1.01319468e+00 4.70606089e-01 -3.50940973e-01 -1.41121119e-01 -6.14773512e-01 -2.43693069e-01 -4.68914241e-01 9.98036712e-02 3.39817256e-01 -1.37146413e-01 -3.81771356e-01 -8.27830553e-01 3.43323827e-01 5.19108951e-01 7.05047727e-01 -5.66319227e-01 4.46361214e-01 3.38315368e-01 -9.35413614e-02 -4.82499987e-01 -1.17201328e+00 -9.48675871e-01 -7.73801684e-01 1.81373596e-01 1.14806426e+00 -1.07832789e+00 -4.86942559e-01 5.64630985e-01 -1.08275962e+00 -6.21371090e-01 -4.25452262e-01 1.33391351e-01 -4.96235043e-01 4.25314516e-01 -7.49457359e-01 -3.69370371e-01 -3.93809766e-01 -1.32299292e+00 1.66764140e+00 2.15113729e-01 -8.39538723e-02 -5.39396822e-01 -1.69654623e-01 9.35840666e-01 -1.16769843e-01 2.48097762e-01 6.44482791e-01 -6.51508152e-01 -7.17356265e-01 -1.60527751e-01 -5.68523109e-01 4.62450475e-01 3.75827099e-03 -8.15788582e-02 -8.71733963e-01 -2.27755055e-01 -3.57039720e-01 -3.08626682e-01 8.02536607e-01 2.39706382e-01 1.26984286e+00 -8.85394216e-02 -3.99189144e-01 5.33476949e-01 1.61732936e+00 2.35848516e-01 3.96058798e-01 2.73872942e-01 1.01281369e+00 4.50733662e-01 1.10961354e+00 1.83916301e-01 6.60605431e-01 7.00446248e-01 5.36967933e-01 8.74061510e-03 -5.83326817e-01 -3.75339016e-02 1.12467013e-01 7.06401110e-01 3.00543189e-01 -3.39323550e-01 -1.07645595e+00 7.23764837e-01 -1.70049667e+00 -5.01203299e-01 -6.06567621e-01 1.80976248e+00 5.88564515e-01 3.80328387e-01 2.38671720e-01 -1.92589108e-02 7.02652097e-01 -6.60267100e-02 -6.23028696e-01 -3.78048688e-01 5.49151041e-02 3.14999908e-01 6.61796570e-01 4.60284919e-01 -1.45431781e+00 1.28240836e+00 5.67251348e+00 9.28918660e-01 -5.52826405e-01 2.50374079e-01 7.20621765e-01 -2.77172006e-03 1.18720338e-01 9.74205062e-02 -1.28894126e+00 3.54032934e-01 8.62428486e-01 4.85445052e-01 -1.10058384e-02 1.25709593e+00 -2.31569588e-01 -6.65195167e-01 -1.06201494e+00 7.97296226e-01 1.02725243e-02 -9.79059219e-01 -1.96015283e-01 -3.78578790e-02 7.81556249e-01 3.80437434e-01 -5.34640551e-01 5.28567255e-01 3.01183999e-01 -4.99732226e-01 9.86987889e-01 -8.94527510e-02 6.77115619e-01 -6.61517859e-01 7.28183627e-01 4.37715352e-01 -1.46620178e+00 1.23738870e-03 -3.58508199e-01 7.69989863e-02 7.71637484e-02 6.10985339e-01 -7.28335083e-01 5.46844900e-01 1.07985342e+00 6.04728043e-01 -1.10637391e+00 1.10384107e+00 -1.63789347e-01 6.03432059e-01 -5.71735084e-01 1.95269272e-01 2.43588522e-01 3.68455201e-02 3.17885667e-01 1.37999010e+00 -7.42339417e-02 -2.08606981e-02 5.24489045e-01 6.21878803e-01 -2.02647775e-01 7.15866461e-02 4.56189029e-02 2.30224133e-01 2.49364778e-01 1.56839919e+00 -1.14666045e+00 -6.25980020e-01 -5.33290923e-01 1.13348043e+00 4.21582431e-01 -1.80957332e-01 -1.03822005e+00 -4.54766333e-01 7.65648007e-01 -3.15883420e-02 7.99211621e-01 -8.80507007e-03 -3.80179316e-01 -9.56220806e-01 2.21310019e-01 -7.71865547e-01 4.93115485e-01 -5.23077965e-01 -9.51503694e-01 5.81438899e-01 2.03732789e-01 -9.68641162e-01 -4.17675376e-02 -9.00481939e-01 -2.38812402e-01 4.01818544e-01 -1.35942245e+00 -1.32675600e+00 -3.34286660e-01 2.24032938e-01 1.06308627e+00 3.87271106e-01 6.27746463e-01 4.13977981e-01 -7.49799371e-01 6.01557851e-01 -1.43173575e-01 1.77561983e-01 5.03971934e-01 -1.60826397e+00 8.01159441e-01 9.76620078e-01 2.04171404e-01 3.63817550e-02 4.11561728e-01 -6.09550357e-01 -1.09082639e+00 -1.43891287e+00 7.44942665e-01 -7.51012206e-01 3.60714406e-01 -5.93780518e-01 -8.23663592e-01 7.86624968e-01 -1.61497042e-01 3.96377146e-01 3.65735561e-01 -9.45786387e-03 -4.70021516e-01 1.12279668e-03 -1.41553056e+00 3.59009176e-01 1.27056110e+00 -4.10787202e-02 -3.91525298e-01 3.89840335e-01 1.07435441e+00 -6.79674089e-01 -1.13919699e+00 7.74319530e-01 3.97644192e-01 -1.12820816e+00 8.20917904e-01 -8.60083550e-02 3.01564664e-01 -4.08929259e-01 -4.70179975e-01 -8.00959408e-01 -3.08038205e-01 -1.55162603e-01 5.68895340e-02 1.38972282e+00 4.19764876e-01 -5.26846468e-01 9.29856300e-01 5.30793965e-01 -5.92178881e-01 -7.07251310e-01 -8.92441213e-01 -7.63347566e-01 7.35845417e-02 -7.03092635e-01 3.80000591e-01 3.93742889e-01 -7.10036039e-01 3.68231922e-01 1.95240468e-01 1.56107917e-01 6.01755142e-01 4.95492965e-01 9.30891037e-01 -1.12063837e+00 -5.31137526e-01 -2.50812143e-01 -4.93999094e-01 -1.04696119e+00 -2.71579754e-02 -9.30402398e-01 3.73376846e-01 -1.78503180e+00 4.99977857e-01 -5.49950063e-01 -1.24581136e-01 1.01389945e+00 -5.05904377e-01 5.01195908e-01 5.56820869e-01 1.11108452e-01 -1.16502428e+00 1.96340650e-01 1.27274513e+00 -1.14886269e-01 -6.81054778e-03 1.13369711e-01 -6.41641200e-01 8.87583792e-01 8.83284748e-01 -7.07188547e-01 -5.79607710e-02 -5.28904974e-01 -4.33295578e-01 -1.97247490e-01 2.62688875e-01 -1.40540648e+00 -3.03718656e-01 6.92735240e-02 3.59230727e-01 -9.39457834e-01 5.38776875e-01 -6.38743639e-01 7.16783181e-02 5.61342895e-01 8.85133743e-02 -1.08711846e-01 6.17194295e-01 4.14167136e-01 -1.20997898e-01 -1.88582361e-01 9.48803961e-01 -2.04320610e-01 -1.17669117e+00 2.71210730e-01 6.02077432e-02 4.00273591e-01 1.17486012e+00 -2.79610813e-01 -3.88842463e-01 3.87414366e-01 -5.27866840e-01 3.98495942e-01 5.19280910e-01 5.38131416e-01 2.06833765e-01 -9.24364626e-01 -6.05566621e-01 9.00419205e-02 3.35571021e-01 5.83835244e-01 3.12628955e-01 5.77767253e-01 -7.00935304e-01 3.89062732e-01 -8.56077969e-02 -9.75960851e-01 -1.57920671e+00 3.62668067e-01 1.28409639e-01 -3.58769059e-01 -5.75906873e-01 1.05859482e+00 4.50117528e-01 -8.36773753e-01 -9.85020027e-02 -7.88940966e-01 1.54496342e-01 -1.39693171e-01 2.81050056e-01 6.67569458e-01 6.14896640e-02 -7.82092690e-01 -5.59397757e-01 7.97496200e-01 -2.26973191e-01 2.72463143e-01 1.28462601e+00 9.32554826e-02 -1.31287530e-01 2.36429811e-01 1.38748384e+00 -3.19071412e-02 -1.63382900e+00 -4.38734367e-02 2.23527357e-01 -2.99497664e-01 -2.65069395e-01 -1.14398682e+00 -1.00611174e+00 6.79709375e-01 6.94531739e-01 -5.65651357e-02 1.05850863e+00 5.24384499e-01 9.45089400e-01 2.93741524e-01 5.29792845e-01 -9.32336628e-01 -1.62806153e-01 5.83377779e-01 6.52577996e-01 -1.40922606e+00 1.24804348e-01 -1.03496599e+00 -6.02567136e-01 8.24284077e-01 9.52670038e-01 -1.55737682e-03 3.05741549e-01 3.97443205e-01 1.69235006e-01 -1.61397457e-01 -3.71666431e-01 -3.93320829e-01 4.24886197e-01 1.71996474e-01 3.32182705e-01 2.93792337e-01 -9.66873914e-02 3.30855846e-01 -2.41937935e-01 -4.03265893e-01 1.52298421e-01 1.10467458e+00 -5.97791731e-01 -1.30103052e+00 -2.96123862e-01 5.00355601e-01 -6.32037699e-01 1.06502168e-01 -2.21354693e-01 9.87183034e-01 5.04748702e-01 9.78577316e-01 2.33674467e-01 -2.87842378e-02 5.72967112e-01 7.72776082e-02 4.53395307e-01 -1.02455330e+00 -6.77187860e-01 3.48514408e-01 1.79567128e-01 -9.77059364e-01 -4.16100681e-01 -7.62950420e-01 -1.68099225e+00 1.13393135e-01 -3.64128143e-01 -1.29781827e-01 7.76594996e-01 7.60005891e-01 4.67178762e-01 7.87414372e-01 1.19464569e-01 -8.73073161e-01 -1.23949423e-01 -9.08018172e-01 -3.29568714e-01 5.23713946e-01 1.01233991e-02 -5.64198792e-01 -3.77374440e-02 -1.17656924e-01]
[9.34360122680664, 0.5362877249717712]
b173ad43-22a7-4772-a49a-8ec5f0ecb6b5
disproving-xai-myths-with-formal-methods
2306.01744
null
https://arxiv.org/abs/2306.01744v1
https://arxiv.org/pdf/2306.01744v1.pdf
Disproving XAI Myths with Formal Methods -- Initial Results
The advances in Machine Learning (ML) in recent years have been both impressive and far-reaching. However, the deployment of ML models is still impaired by a lack of trust in how the best-performing ML models make predictions. The issue of lack of trust is even more acute in the uses of ML models in high-risk or safety-critical domains. eXplainable artificial intelligence (XAI) is at the core of ongoing efforts for delivering trustworthy AI. Unfortunately, XAI is riddled with critical misconceptions, that foster distrust instead of building trust. This paper details some of the most visible misconceptions in XAI, and shows how formal methods have been used, both to disprove those misconceptions, but also to devise practically effective alternatives.
['Joao Marques-Silva']
2023-05-13
null
null
null
null
['misconceptions']
['miscellaneous']
[-1.61511809e-01 8.42638552e-01 -2.65145689e-01 -5.13325155e-01 -2.61364043e-01 -2.65863359e-01 6.88918352e-01 3.54287177e-01 1.08246788e-01 8.48008215e-01 7.87711218e-02 -8.35154712e-01 -3.08940321e-01 -5.41827202e-01 -8.50800216e-01 -3.05508912e-01 -1.67713722e-03 6.18525803e-01 -2.39721596e-01 -1.60565972e-01 2.86029011e-01 2.74827749e-01 -1.12272966e+00 3.60904157e-01 9.41624284e-01 6.75553143e-01 -6.21464789e-01 2.82978296e-01 2.36091807e-01 1.73087502e+00 -4.95084465e-01 -7.89991617e-01 -5.72893806e-02 -4.05397207e-01 -9.37335610e-01 -4.91084009e-01 -1.05834670e-01 -4.08175230e-01 5.44092841e-02 9.69355643e-01 -2.16106296e-01 -6.38513565e-01 5.07684410e-01 -2.11555624e+00 -5.96124351e-01 9.68895257e-01 6.59281015e-02 -2.92864621e-01 2.79778600e-01 5.15274048e-01 9.89087999e-01 -2.98832387e-01 4.01211709e-01 1.28286612e+00 9.72203732e-01 7.50907123e-01 -1.07074165e+00 -9.75917518e-01 -8.41035172e-02 2.40262970e-01 -1.11824942e+00 -5.12045264e-01 4.73186195e-01 -5.99881291e-01 1.06092989e+00 4.66562420e-01 8.12296689e-01 1.07114148e+00 8.61900091e-01 6.32088840e-01 1.29225791e+00 -5.56143105e-01 4.24100995e-01 7.79632330e-01 2.44637430e-01 6.89367592e-01 1.00695884e+00 5.74435651e-01 -6.47805870e-01 -5.18972218e-01 3.60664517e-01 6.42067865e-02 -8.45048055e-02 -1.43367141e-01 -1.04132020e+00 9.53072131e-01 2.93965667e-01 4.93853599e-01 -3.93618017e-01 3.22069198e-01 1.97148144e-01 2.67288297e-01 3.61853987e-01 8.42898965e-01 -8.76910627e-01 -3.13169807e-01 -6.58420205e-01 2.05198884e-01 1.03242624e+00 6.02716446e-01 5.57538629e-01 4.73574065e-02 7.02201426e-01 -3.30299535e-03 8.37884605e-01 2.23281160e-01 6.20811656e-02 -1.27576375e+00 -1.54586658e-01 6.54916227e-01 2.81563848e-01 -1.25123084e+00 -2.35967115e-01 -3.23573947e-01 -6.46888018e-01 8.48115325e-01 1.31495863e-01 -2.92737246e-01 -5.10768473e-01 1.47935247e+00 7.44388178e-02 -1.04774408e-01 3.13986152e-01 6.57855272e-01 4.11437064e-01 4.43215132e-01 3.36870074e-01 1.94533505e-02 5.82715094e-01 -4.45482045e-01 -6.55870140e-01 -6.78732336e-01 8.36616457e-01 -1.92685187e-01 6.28551304e-01 8.07886600e-01 -9.93792474e-01 1.30630843e-02 -1.41766167e+00 1.96008220e-01 -1.06957830e-01 -6.37507200e-01 1.24525833e+00 7.56919980e-01 -8.04708660e-01 7.74986327e-01 -8.34325671e-01 -7.14899227e-02 6.30380571e-01 3.22659373e-01 -5.93251228e-01 -1.16480239e-01 -1.10072482e+00 1.55867422e+00 2.10690856e-01 2.58505732e-01 -1.01029503e+00 -8.16621721e-01 -8.75552118e-01 -1.05327949e-01 3.99712920e-01 -6.14561141e-01 1.02909243e+00 -1.39658868e+00 -8.48223329e-01 4.78490442e-01 2.21818402e-01 -8.02403092e-01 5.85176170e-01 -2.29777068e-01 -5.86950064e-01 -2.42030278e-01 -5.19194081e-03 4.23109829e-01 4.34349388e-01 -1.52054870e+00 -5.07032871e-01 -3.62328529e-01 -1.32516101e-01 -4.33132559e-01 5.92070743e-02 3.20317000e-01 7.35309362e-01 3.20092589e-02 -2.03510642e-01 -8.05582404e-01 -3.18068206e-01 1.52881026e-01 -2.24365979e-01 -1.81963444e-01 5.44349790e-01 -5.92764616e-01 1.12543356e+00 -1.85988259e+00 -4.41636533e-01 2.10513711e-01 6.63983047e-01 1.88311413e-01 4.89336312e-01 6.41557097e-01 -2.70973332e-02 6.70856893e-01 -3.68534699e-02 2.07538038e-01 2.03395069e-01 4.90582228e-01 -3.31241757e-01 2.71567851e-01 3.80404145e-01 8.05033326e-01 -9.92437243e-01 -4.50321972e-01 6.38107210e-02 3.91534328e-01 -2.73620754e-01 8.73639658e-02 -2.17832103e-01 2.02783599e-01 -4.36225563e-01 6.82636857e-01 2.20984429e-01 -3.36799234e-01 5.18274546e-01 3.51214170e-01 -8.48876759e-02 4.69587743e-01 -5.66731930e-01 1.07173836e+00 -1.51733667e-01 6.77620888e-01 6.50480688e-02 -6.76323414e-01 7.86384046e-01 5.92164099e-01 3.21277648e-01 -4.02731866e-01 1.74132511e-01 3.87300640e-01 3.83765370e-01 -6.44538164e-01 -1.02128536e-01 -6.51627541e-01 -5.20840324e-02 8.54241490e-01 -2.74217576e-01 -4.58466887e-01 -7.00459003e-01 2.35790133e-01 1.04856396e+00 2.42242023e-01 4.01578277e-01 -2.63402522e-01 -9.64270979e-02 5.13869762e-01 1.06081522e+00 4.12582487e-01 -4.72829700e-01 2.19143361e-01 5.87562442e-01 -1.00903332e+00 -9.32389975e-01 -5.83408117e-01 -2.76437420e-02 2.07774058e-01 -2.63511419e-01 -5.48137546e-01 -6.38068378e-01 -8.93650949e-01 1.56580210e-01 1.58833921e+00 -7.03762352e-01 -5.81192791e-01 1.01626202e-01 -2.83634871e-01 6.20323598e-01 5.46625078e-01 1.35862067e-01 -1.05645335e+00 -1.27376914e+00 1.34985700e-01 1.24309905e-01 -7.34342694e-01 4.60510910e-01 1.00497179e-01 -7.10152566e-01 -1.11230695e+00 3.37155253e-01 6.09225221e-02 5.56492507e-01 -1.08254299e-01 1.24632013e+00 7.69349694e-01 -1.32086977e-01 2.59650022e-01 -2.35928863e-01 -1.13151038e+00 -1.20328605e+00 -5.32952249e-01 1.15670465e-01 -6.27156794e-01 6.99444711e-01 -3.71498197e-01 -2.61198550e-01 2.66126096e-01 -6.97142363e-01 2.02400133e-01 5.89505732e-01 5.99751830e-01 -3.05372238e-01 1.75202757e-01 8.10568571e-01 -1.18332171e+00 4.78341699e-01 -6.74173832e-01 -3.42750371e-01 5.26386499e-01 -1.50474334e+00 -9.65893939e-02 2.82682866e-01 -3.24130841e-02 -9.94573116e-01 -4.68321532e-01 4.70679700e-02 5.37185483e-02 -2.64982790e-01 5.95187426e-01 -1.22991666e-01 -5.28471023e-02 9.14293170e-01 -6.00296438e-01 4.76321459e-01 -5.59539627e-03 1.04208954e-01 6.49962842e-01 7.07792118e-02 -6.04303956e-01 5.80649734e-01 1.42467633e-01 -2.45386764e-01 -4.26619887e-01 -7.80812204e-01 5.51964700e-01 -3.74667972e-01 -3.73418719e-01 4.60553557e-01 -5.80109060e-01 -8.49919319e-01 2.07417775e-02 -1.02303052e+00 -6.05730534e-01 -2.52867937e-01 3.81484181e-01 -3.79714340e-01 1.62003527e-03 -3.53472888e-01 -1.15181315e+00 -4.04258132e-01 -1.04149187e+00 2.18349248e-01 8.92745913e-04 -9.20953095e-01 -1.01312721e+00 4.69461456e-02 6.57360613e-01 4.77399290e-01 5.02867162e-01 1.30425942e+00 -9.56669331e-01 -3.88470948e-01 -6.39853656e-01 2.18111351e-02 5.26429236e-01 -4.40569362e-03 4.26882893e-01 -1.07957304e+00 1.21578336e-01 2.26611838e-01 -6.73889875e-01 -1.02397941e-01 6.17919527e-02 4.58839923e-01 -8.73264968e-01 -4.17886645e-01 -1.31536052e-01 1.18921518e+00 5.20318329e-01 5.87679744e-01 4.90238130e-01 2.13360190e-01 9.92270827e-01 9.89163518e-01 3.05930346e-01 5.39488494e-01 1.21591896e-01 4.53749478e-01 8.11417848e-02 3.93095613e-01 -4.46134031e-01 2.67673016e-01 2.17195496e-01 -1.06407478e-01 3.11317861e-01 -1.56855357e+00 3.81090939e-01 -2.11622787e+00 -9.71734464e-01 -2.86500067e-01 2.10947657e+00 8.81709158e-01 6.32675529e-01 -2.07208320e-01 3.27283502e-01 -2.81956475e-02 -3.72546613e-01 -4.42982644e-01 -8.03877950e-01 2.86036521e-01 -1.47985935e-01 5.79667352e-02 5.66292346e-01 -4.92690384e-01 7.77568579e-01 7.54671764e+00 1.61041066e-01 -7.78629422e-01 1.04653314e-02 8.88091683e-01 1.59463614e-01 -6.77123010e-01 6.20864093e-01 -3.01975608e-01 1.47081405e-01 1.27995968e+00 -2.83294499e-01 2.38610670e-01 1.07281566e+00 1.29251420e-01 -3.28560531e-01 -1.31311440e+00 1.18888833e-01 -7.88238943e-02 -1.48883152e+00 -3.54583949e-01 -9.92303193e-02 3.76752973e-01 -8.40723291e-02 -1.72528714e-01 7.52290189e-02 8.17298591e-01 -1.54500782e+00 9.85221803e-01 4.86372113e-01 2.96658844e-01 -9.43063855e-01 9.89386618e-01 5.91504931e-01 -6.92893099e-03 -1.70895249e-01 -4.10234690e-01 -6.78696036e-01 -3.16898674e-01 6.29294097e-01 -1.01782799e+00 2.87048727e-01 7.31078565e-01 5.74948549e-01 -4.19545084e-01 5.61560929e-01 -4.65778381e-01 7.54891992e-01 -9.95879173e-02 -9.05163884e-02 1.54016748e-01 -8.46944973e-02 3.99928354e-02 9.15263534e-01 -2.23737851e-01 1.70564190e-01 -4.42082316e-01 1.11564958e+00 5.95845997e-01 -6.29009783e-01 -1.03421760e+00 -3.96099210e-01 3.85059536e-01 1.05425274e+00 -1.64521173e-01 -2.56978780e-01 -2.75520086e-01 3.17012370e-01 7.69854560e-02 4.43863124e-02 -6.62265599e-01 2.09175721e-01 7.20522344e-01 1.12508528e-01 -5.49409747e-01 -5.08770905e-02 -6.51595771e-01 -8.66261482e-01 -2.96807885e-01 -1.45176482e+00 1.95920765e-01 -9.67424154e-01 -1.06234908e+00 5.53150117e-01 3.00786663e-02 -6.63722575e-01 -5.31790495e-01 -3.81769836e-01 -5.68950415e-01 5.92706621e-01 -1.26013029e+00 -1.55335498e+00 8.72766227e-02 1.11638322e-01 -1.62559897e-02 -2.38551246e-03 1.11981106e+00 -2.77105808e-01 -4.06911373e-01 3.75963867e-01 -5.04119217e-01 -1.95963144e-01 5.56562066e-01 -8.63564432e-01 2.90981084e-01 5.17124057e-01 1.42853204e-02 8.58514726e-01 1.11524856e+00 -7.92719662e-01 -1.49861479e+00 -6.44128919e-01 1.15612423e+00 -9.54029083e-01 8.38081777e-01 -1.98099494e-01 -9.52153683e-01 1.30797565e+00 3.15801084e-01 -3.00314158e-01 9.18697417e-01 2.96645939e-01 -5.80767751e-01 -8.10099915e-02 -1.50284493e+00 4.62258786e-01 5.42168081e-01 -4.40919191e-01 -6.60250902e-01 3.23966473e-01 7.61293828e-01 3.23237181e-01 -7.66009390e-01 3.35883528e-01 7.83282220e-01 -1.28355515e+00 5.71448505e-01 -1.01262784e+00 5.38315356e-01 4.73866649e-02 6.82707876e-02 -1.03232050e+00 -1.98007926e-01 -7.29951024e-01 2.88448900e-01 1.22806358e+00 7.49824584e-01 -9.59425271e-01 7.51785696e-01 1.81887877e+00 5.14588924e-03 -7.33372509e-01 -5.51372826e-01 -4.68612105e-01 2.91525662e-01 -8.92294109e-01 5.71583688e-01 1.27594566e+00 7.29046524e-01 2.09126830e-01 -3.55160385e-01 1.69773996e-01 8.38442862e-01 -1.03789940e-01 7.38019526e-01 -1.51440191e+00 -2.65011102e-01 -2.63944656e-01 -4.44700271e-01 2.61523217e-01 1.10374903e-02 -4.44300801e-01 9.76708308e-02 -1.40766513e+00 2.83415437e-01 -6.80175960e-01 -1.38618350e-01 8.69413853e-01 2.97170877e-02 -3.62477452e-01 3.36434007e-01 2.42970943e-01 -1.32149100e-01 6.26289705e-03 7.03692019e-01 -7.25280941e-02 2.34180704e-01 1.53967058e-02 -1.15536523e+00 1.25053513e+00 9.24867392e-01 -6.60630703e-01 -4.57167119e-01 -3.56069237e-01 9.12352026e-01 1.23056248e-01 6.64002955e-01 -9.19678271e-01 2.79582977e-01 -6.02091432e-01 2.32629269e-01 -1.23814382e-01 1.02037638e-01 -1.22143233e+00 7.98171699e-01 8.23488176e-01 -5.03226519e-01 -7.20822066e-02 2.60560960e-01 1.29771322e-01 2.46294867e-02 -4.76024151e-01 6.94150984e-01 -1.09475024e-01 -2.50466645e-01 -1.36990413e-01 -2.67696857e-01 -3.27710718e-01 1.15208936e+00 1.07003406e-01 -3.56131107e-01 -4.47294593e-01 -3.12482655e-01 1.88578829e-01 8.14042151e-01 1.53599188e-01 8.62335205e-01 -8.36656153e-01 -7.19179869e-01 8.85950774e-02 6.58618510e-02 -1.55012697e-01 -2.48094618e-01 7.35950470e-01 -5.08525193e-01 4.62717026e-01 -3.32089752e-01 -4.77002971e-02 -9.95366037e-01 7.80182004e-01 3.73734921e-01 1.05410427e-01 -7.53195107e-01 7.05673814e-01 -7.57527649e-02 -1.30418092e-01 3.25634122e-01 -2.30173007e-01 1.82154536e-01 -5.51081955e-01 6.70161784e-01 1.89380735e-01 -1.98046327e-01 -9.09138396e-02 -5.39693236e-01 -1.44980550e-02 -1.41894177e-01 -1.66914180e-01 1.58107364e+00 3.18219244e-01 -4.43017513e-01 6.84052944e-01 4.44028080e-01 -2.98418999e-01 -1.01002717e+00 3.91807407e-01 3.64598095e-01 -5.29136240e-01 1.52040869e-01 -1.54441249e+00 -5.48423648e-01 1.08415627e+00 9.16891731e-03 2.40839884e-01 6.21903896e-01 -1.99324653e-01 4.85718608e-01 2.83612132e-01 8.21782768e-01 -8.45889211e-01 -3.54804277e-01 -2.30742291e-01 1.08446443e+00 -1.32028496e+00 3.20452869e-01 -8.65301490e-02 -1.08141327e+00 9.33010459e-01 5.75113714e-01 1.69007137e-01 7.44911492e-01 4.85699624e-01 3.28547984e-01 -4.47499007e-01 -1.28632617e+00 6.66731596e-01 -1.44860059e-01 6.57079518e-01 4.60070223e-01 3.40237677e-01 -2.26956517e-01 8.24854195e-01 2.42853854e-02 4.45916533e-01 6.12286747e-01 9.65159416e-01 -3.85971159e-01 -8.75029981e-01 -4.25774872e-01 1.57302007e-01 -4.87943858e-01 7.60666206e-02 -9.32483912e-01 1.10738409e+00 1.81394801e-01 1.38927567e+00 -4.67317224e-01 -6.66039705e-01 -1.85508355e-01 2.11364537e-01 1.69953629e-01 -3.11363727e-01 -8.20868909e-01 -2.85245031e-01 5.65950811e-01 -7.27050602e-01 -1.16723567e-01 -5.88286161e-01 -1.40303302e+00 -7.00546801e-01 -2.34421015e-01 4.37567085e-01 9.40872669e-01 9.92269695e-01 2.84350425e-01 7.96949305e-03 3.40433002e-01 -6.82258382e-02 -7.37610519e-01 -6.79665148e-01 -2.48341590e-01 1.51029482e-01 3.12809378e-01 -3.50135684e-01 -4.18103725e-01 -3.45842272e-01]
[8.859902381896973, 6.065372467041016]
4a79ea37-6b0f-465d-942a-1b09c3848525
multi-objective-resource-optimization-of
2202.02139
null
https://arxiv.org/abs/2202.02139v1
https://arxiv.org/pdf/2202.02139v1.pdf
Multi Objective Resource Optimization of Wireless Network Based on Cross Domain Virtual Network Embedding
The rapid development of virtual network architecture makes it possible for wireless network to be widely used. With the popularity of artificial intelligence (AI) industry in daily life, efficient resource allocation of wireless network has become a problem. Especially when network users request wireless network resources from different management domains, they still face many practical problems. From the perspective of virtual network embedding (VNE), this paper designs and implements a multi-objective optimization VNE algorithm for wireless network resource allocation. Resource allocation in virtual network is essentially a problem of allocating underlying resources for virtual network requests (VNRs). According to the proposed objective formula, we consider the optimization mapping cost, network delay and VNR acceptance rate. VNE is completed by node mapping and link mapping. In the experiment and simulation stage, it is compared with other VNE algorithms, the cross domain VNE algorithm proposed in this paper is optimal in the above three indicators. This shows the effectiveness of the algorithm in wireless network resource allocation.
['Peiying Zhang', 'Qifeng Sun', 'Youxiang Duan', 'Tao Dong', 'Chao Wang']
2022-02-03
null
null
null
null
['network-embedding']
['methodology']
[ 8.14237744e-02 -1.41468763e-01 -6.36491418e-01 4.91074994e-02 5.21586180e-01 -3.35367799e-01 -2.29988813e-01 -3.06866288e-01 -5.40036023e-01 1.19886196e+00 -3.92744541e-01 -5.75713694e-01 -8.22873354e-01 -1.07039416e+00 3.21044207e-01 -4.16765749e-01 -4.06240582e-01 5.07352829e-01 4.59825562e-04 -1.35698140e-01 2.88174540e-01 7.52168298e-01 -9.48772132e-01 -7.32320666e-01 5.98264098e-01 1.01244998e+00 6.55533552e-01 3.07779044e-01 -6.35911047e-01 2.92267203e-01 -8.33703220e-01 -1.96176112e-01 3.70915592e-01 -2.07907811e-01 -1.01180458e+00 3.96216027e-02 -9.62630689e-01 -2.48906195e-01 -3.78010750e-01 8.58066261e-01 7.92440116e-01 3.57680172e-01 4.51452374e-01 -1.90199745e+00 -1.95588231e-01 5.88169396e-01 -7.40424752e-01 3.97949994e-01 8.55992958e-02 -5.74141920e-01 8.38074028e-01 -6.19002223e-01 8.25991988e-01 9.42891955e-01 1.91823646e-01 5.69636881e-01 -9.52257931e-01 -8.81221890e-01 1.88343525e-01 3.91080201e-01 -1.59612143e+00 -1.97607830e-01 8.31467450e-01 6.32401230e-03 6.29371464e-01 5.69733560e-01 6.60454392e-01 3.59885901e-01 8.29285681e-02 2.40375608e-01 2.93039739e-01 -5.31295002e-01 3.23703021e-01 5.72520792e-01 -5.85978925e-01 1.35073453e-01 3.29735458e-01 -1.55066267e-01 1.85839325e-01 -1.37854978e-01 1.13075650e+00 -8.00914913e-02 -4.46243644e-01 -5.03484905e-01 -9.41154778e-01 6.09425902e-01 3.58489066e-01 4.67962116e-01 -6.71560764e-01 2.08810151e-01 4.63215172e-01 6.20262742e-01 2.68213660e-01 1.60602942e-01 -4.82591331e-01 -1.14733003e-01 -8.36245000e-01 -3.26466560e-01 8.38394463e-01 1.09627008e+00 3.21676850e-01 1.83251649e-01 3.65858525e-01 1.15339553e+00 5.32588542e-01 5.67774057e-01 3.23216431e-02 -1.04366899e+00 2.87640423e-01 3.35644484e-01 -7.26004243e-02 -1.14267576e+00 -5.56098461e-01 -6.21518552e-01 -1.16997254e+00 1.51470721e-01 -2.62250006e-01 -5.85485876e-01 -2.31690750e-01 1.79184890e+00 3.99812639e-01 3.23100209e-01 1.74512282e-01 7.51911461e-01 5.95108569e-01 9.96165156e-01 2.11169139e-01 -8.81672204e-01 8.41634333e-01 -7.25685239e-01 -1.14926696e+00 2.90061921e-01 3.20282072e-01 -1.01112175e+00 2.17270121e-01 1.21329655e-03 -9.59126055e-01 -1.62857249e-01 -9.81789768e-01 8.41368079e-01 -2.20369354e-01 -1.57380417e-01 6.99611962e-01 9.15127099e-01 -1.25581503e+00 1.68684632e-01 -2.88863108e-02 -8.96807551e-01 2.63521850e-01 8.15064371e-01 -3.30897182e-01 1.40475780e-01 -1.34014297e+00 7.17168987e-01 7.29144156e-01 7.67386705e-02 -4.39476967e-01 -5.53067744e-01 -4.81501818e-01 2.50991106e-01 6.63142979e-01 -4.94195342e-01 7.48527884e-01 -9.54265177e-01 -1.43689442e+00 2.24977121e-01 1.05868258e-01 1.50358438e-01 4.34215307e-01 6.12392128e-01 -1.09695482e+00 5.09337336e-02 -9.55240056e-02 2.64505744e-01 3.90238464e-01 -1.30142260e+00 -8.14976692e-01 2.95710206e-01 3.54273260e-01 2.17012301e-01 -6.17484391e-01 3.54062140e-01 -6.15034640e-01 -5.21177053e-01 1.68839529e-01 -5.69629550e-01 -5.49748182e-01 2.43174508e-01 -2.96156615e-01 3.05453967e-02 1.11449873e+00 -2.65991360e-01 1.70875621e+00 -1.99817443e+00 4.52084899e-01 1.07803273e+00 4.95272815e-01 3.20666045e-01 -1.57142773e-01 4.35159355e-01 -1.47637844e-01 5.62803864e-01 2.79287640e-02 2.42370948e-01 -1.30641311e-01 2.09172845e-01 3.38755399e-01 1.35173261e-01 -3.79559249e-01 2.32659951e-01 -7.94054747e-01 -8.18864286e-01 3.33649129e-01 2.63329983e-01 -4.78769571e-01 1.73890963e-01 2.74809092e-01 2.30072364e-01 -7.78909028e-01 5.89369655e-01 1.05098665e+00 -2.74127871e-01 8.10543656e-01 -2.19204098e-01 -1.27544746e-01 -5.98643422e-01 -1.51576269e+00 1.32835257e+00 -7.72162378e-01 7.46420562e-01 4.35959637e-01 -1.25350225e+00 7.22202301e-01 7.44175613e-01 1.07434869e+00 -7.49234855e-01 4.19043839e-01 4.95912164e-01 7.31238946e-02 -6.99130774e-01 3.07203293e-01 -3.34010385e-02 1.33542940e-01 4.99589056e-01 -2.45896801e-02 4.62391078e-01 1.24280438e-01 1.67074785e-01 9.46985126e-01 -2.79382020e-01 4.60965753e-01 -2.49425411e-01 8.10540736e-01 -3.72597456e-01 6.21840179e-01 3.25367302e-01 -5.20852923e-01 -3.38050991e-01 4.12653387e-01 -1.31661698e-01 -1.16683888e+00 -1.02834022e+00 -2.06873640e-01 7.30117202e-01 7.58098662e-01 8.29299092e-02 -4.70026821e-01 -3.90466124e-01 -1.63660139e-01 1.96027979e-01 -1.23989142e-01 1.11063281e-02 -2.90173531e-01 -5.39332509e-01 2.77525932e-01 -1.49105459e-01 6.17386818e-01 -1.08105314e+00 -3.35922778e-01 6.78347051e-01 -2.47607321e-01 -1.22393000e+00 -2.75275886e-01 -9.01341066e-02 -7.08967328e-01 -8.18562865e-01 -8.80238295e-01 -9.93693113e-01 5.92295110e-01 5.34855604e-01 8.11438859e-01 5.90977609e-01 -5.14310598e-01 4.41489369e-01 -5.66531241e-01 -1.70670167e-01 -1.73338547e-01 4.03354585e-01 3.14171135e-01 -1.78398117e-01 -1.39628381e-01 -8.78431976e-01 -4.47311819e-01 6.43129528e-01 -8.38776767e-01 -8.51910263e-02 6.27125919e-01 4.32260454e-01 3.55735362e-01 8.99040282e-01 1.04810619e+00 -6.51137471e-01 7.78313816e-01 -9.95571196e-01 -6.46016479e-01 5.20210981e-01 -4.75203842e-01 -1.74409077e-01 6.13647640e-01 -3.08494300e-01 -8.02257776e-01 -7.53320813e-01 -9.92268790e-04 -8.68157819e-02 4.59130645e-01 6.93487287e-01 -6.69668555e-01 -3.41062069e-01 -2.00164523e-02 -1.36944398e-01 6.98678866e-02 -2.97513485e-01 -1.10225804e-01 9.67915952e-01 -3.22106570e-01 -2.62409121e-01 1.00700068e+00 1.11322202e-01 4.75356340e-01 -1.04165614e+00 3.67297903e-02 -2.03539684e-01 -1.38461530e-01 -6.51874542e-01 6.34439409e-01 -5.00958622e-01 -1.08802497e+00 -1.12863034e-01 -1.04312706e+00 3.54956016e-02 2.11382806e-01 6.54992640e-01 -1.08386368e-01 3.51217389e-01 1.01297304e-01 -1.01979947e+00 -2.47798398e-01 -1.10468268e+00 -1.84843674e-01 4.14042473e-01 1.12435445e-02 -1.30119157e+00 -4.10470814e-01 -2.37107247e-01 9.35345054e-01 3.63723069e-01 9.75912511e-01 7.00995177e-02 -6.51404917e-01 6.28800690e-02 -6.58792794e-01 8.98103341e-02 2.43949309e-01 2.07970411e-01 -1.56200752e-01 -3.92467350e-01 -3.94600958e-01 2.89524496e-01 -1.59636810e-02 5.62338531e-01 1.37564659e+00 -1.53996602e-01 -6.35768473e-01 6.50033951e-01 2.17352414e+00 7.19896317e-01 6.40939415e-01 5.20924449e-01 3.56686622e-01 6.95672452e-01 8.60777259e-01 8.96256864e-01 9.95768607e-02 8.24679673e-01 9.96245563e-01 -4.03634012e-01 2.96174139e-01 2.53532290e-01 -9.31215659e-02 7.20657587e-01 -1.81691766e-01 -1.00274706e+00 -5.09915709e-01 2.68470854e-01 -1.76595223e+00 -1.14487195e+00 2.41316408e-01 2.07409906e+00 -1.20534398e-01 1.15712225e-01 2.40274761e-02 3.80077183e-01 1.11095715e+00 4.10588905e-02 -4.00177687e-01 -6.36257708e-01 1.03068992e-01 8.85809734e-02 8.80339146e-01 4.24950987e-01 -6.03589833e-01 6.26648664e-01 5.91928768e+00 1.16172540e+00 -1.07338333e+00 1.57994479e-01 2.89474398e-01 5.08592613e-02 -4.91178960e-01 -2.81218961e-02 -6.24499507e-02 6.50496423e-01 6.11098230e-01 -4.94978160e-01 9.78166580e-01 5.41782498e-01 6.51903629e-01 3.13085765e-02 -3.91501546e-01 1.19238663e+00 -5.62290132e-01 -1.42671394e+00 1.57077774e-01 4.16992724e-01 3.99316907e-01 -3.86589557e-01 -1.39070004e-01 2.32410780e-03 -2.17480600e-01 -1.03976214e+00 3.45642775e-01 1.17556795e-01 1.14562941e+00 -1.11962259e+00 1.00644171e+00 1.23184239e-02 -1.55307937e+00 -3.22247326e-01 -4.28682059e-01 9.88249108e-02 6.13084018e-01 3.80436420e-01 -4.74527061e-01 9.37311709e-01 3.52061898e-01 3.18376184e-01 4.34144378e-01 1.36676979e+00 2.63226360e-01 1.50252014e-01 -1.90219656e-01 -2.46582389e-01 5.56383468e-02 -4.34517711e-01 6.40948951e-01 5.55305839e-01 5.49649179e-01 2.86851346e-01 1.32630527e-01 5.72950184e-01 -3.27257752e-01 6.13045454e-01 -3.18719387e-01 -2.28635706e-02 1.23162806e+00 1.33195519e+00 -8.21409822e-01 2.06681527e-02 -3.39892507e-01 7.33494341e-01 -1.01931758e-01 6.60115361e-01 -1.13921225e+00 -1.09508407e+00 1.04547298e+00 -5.56406565e-02 -7.79403225e-02 -1.43517166e-01 6.41228780e-02 -5.71317434e-01 -9.46042016e-02 -3.75327915e-01 8.80082604e-03 -4.73391086e-01 -6.60442472e-01 7.62781322e-01 -9.11600515e-02 -1.48728013e+00 3.68537724e-01 -4.55798924e-01 -5.64686656e-01 9.23933327e-01 -1.67132139e+00 -4.29535061e-01 -3.05493921e-01 5.45668960e-01 3.36476862e-01 -5.95146716e-01 9.17782426e-01 1.08507299e+00 -8.21502388e-01 7.47860372e-01 3.23067069e-01 -6.00843318e-02 1.31640330e-01 -4.85704124e-01 -3.77561152e-01 7.88311958e-01 -4.28121358e-01 3.28380674e-01 7.09873497e-01 -2.50188142e-01 -1.11029553e+00 -7.08802342e-01 6.50440931e-01 6.58436835e-01 3.77064466e-01 -6.10482693e-02 -2.23802656e-01 2.44614527e-01 1.98527500e-01 -1.29781594e-03 9.94796753e-01 -2.20379874e-01 5.37267148e-01 -3.19143504e-01 -1.72270811e+00 6.66086078e-01 1.08739269e+00 3.87692079e-02 5.84652722e-01 1.71548203e-01 8.70698690e-01 6.48582578e-02 -9.49125648e-01 3.02457482e-01 6.37506366e-01 -2.94875860e-01 1.02855384e+00 -4.31584626e-01 -6.65560439e-02 -4.34413344e-01 -4.31404531e-01 -1.19265294e+00 -5.45577526e-01 -6.26333237e-01 2.24827930e-01 1.31478345e+00 4.59070861e-01 -7.81871080e-01 7.48785734e-01 1.79452583e-01 5.16724944e-01 -6.80799961e-01 -1.09681392e+00 -8.52266192e-01 -5.78354478e-01 -2.75239289e-01 1.05780077e+00 1.13593113e+00 -5.59451133e-02 3.60362768e-01 -5.34271777e-01 2.62254268e-01 7.16194332e-01 -3.98768634e-01 5.54976583e-01 -1.48473060e+00 -1.33376122e-01 -6.40621960e-01 -5.33306122e-01 -7.98368514e-01 4.00641501e-01 -8.13776016e-01 -7.21210718e-01 -1.90426862e+00 -1.09660134e-01 -1.40307343e+00 -6.92134321e-01 6.00218847e-02 4.31685716e-01 2.17934418e-02 2.85930872e-01 1.88029647e-01 -5.08544683e-01 3.15322042e-01 1.34783113e+00 8.23089927e-02 -3.29461575e-01 1.71624914e-01 -4.97396499e-01 3.02982628e-01 1.17686474e+00 -4.35973585e-01 -1.03258216e+00 -4.67439324e-01 1.70220852e-01 7.21004605e-01 -1.84012622e-01 -7.57079184e-01 -1.39775667e-02 -6.61696911e-01 1.08252130e-01 -4.77797724e-02 2.51998007e-01 -1.73991907e+00 6.17927372e-01 5.17501652e-01 2.67905947e-02 1.00721426e-01 7.70304278e-02 4.01252478e-01 6.74022138e-02 -1.98257595e-01 7.36823916e-01 2.94539124e-01 -1.08907032e+00 6.60583794e-01 -7.15637028e-01 -1.12127103e-01 1.50084901e+00 -6.46715879e-01 1.01930879e-01 -4.01530594e-01 -8.52774560e-01 5.85904896e-01 1.36244208e-01 3.00487101e-01 8.56492877e-01 -1.33460605e+00 -4.72448200e-01 -1.77579224e-01 -1.48805037e-01 -6.75652981e-01 5.46597064e-01 7.02593327e-01 -9.50436950e-01 1.62578121e-01 -6.26500785e-01 -9.52119455e-02 -1.45014358e+00 3.20804596e-01 2.80637085e-01 -1.99213982e-01 3.84810902e-02 8.46683383e-01 -4.38091725e-01 -2.41615161e-01 2.93118745e-01 5.90754509e-01 -6.24116957e-01 -5.74515797e-02 9.34565291e-02 8.51353943e-01 -5.56070507e-01 -6.51684880e-01 -6.07542276e-01 4.67297286e-01 1.31731853e-01 -3.05124551e-01 1.19922984e+00 -6.36613011e-01 -6.21878862e-01 -3.57919753e-01 1.24981701e+00 -1.28567349e-02 -1.28534541e-01 -1.59444481e-01 -1.70622021e-01 -7.99957752e-01 3.74769807e-01 -5.16450107e-01 -1.47143376e+00 2.67262012e-01 8.69717240e-01 4.52491105e-01 1.40419745e+00 -5.27255356e-01 6.87463939e-01 -7.91351423e-02 8.89613390e-01 -1.29384720e+00 -1.07908390e-01 2.95903772e-01 2.60428429e-01 -9.73270893e-01 5.89936934e-02 -9.28354621e-01 -9.51060951e-02 1.17331123e+00 7.59467065e-01 1.94789574e-01 1.15037239e+00 3.90642695e-02 -1.32127672e-01 4.06458154e-02 -3.73227596e-01 -9.36766937e-02 -4.28176641e-01 6.16606832e-01 2.93364316e-01 2.79098004e-01 -7.66424060e-01 4.54199426e-02 3.09059381e-01 -2.24944875e-01 7.45686889e-01 6.95481658e-01 -7.69163966e-01 -1.71592748e+00 -1.14334434e-01 5.52858233e-01 -5.55443525e-01 1.94948632e-02 4.85509485e-01 9.35678780e-01 7.69162774e-02 1.20701396e+00 -1.19675256e-01 -5.37432432e-01 1.53977931e-01 -6.73795402e-01 6.84164986e-02 -3.05117965e-01 1.60831623e-02 -2.62427926e-01 1.07922271e-01 -2.45239392e-01 -6.31992340e-01 7.97884464e-02 -1.46858299e+00 -9.34274673e-01 -6.24936342e-01 7.22427011e-01 1.04045117e+00 6.64255023e-01 1.96931630e-01 1.08934939e+00 1.19737411e+00 -3.57680559e-01 6.43983036e-02 -2.90808946e-01 -1.03765512e+00 -2.06380263e-01 1.92048447e-03 -8.88528287e-01 -2.44378000e-01 -6.86944485e-01]
[5.881754398345947, 1.703986644744873]
3265d984-d497-4917-9d6b-f8a99ceaf5aa
detecting-attended-visual-targets-in-video
2003.02501
null
https://arxiv.org/abs/2003.02501v2
https://arxiv.org/pdf/2003.02501v2.pdf
Detecting Attended Visual Targets in Video
We address the problem of detecting attention targets in video. Our goal is to identify where each person in each frame of a video is looking, and correctly handle the case where the gaze target is out-of-frame. Our novel architecture models the dynamic interaction between the scene and head features and infers time-varying attention targets. We introduce a new annotated dataset, VideoAttentionTarget, containing complex and dynamic patterns of real-world gaze behavior. Our experiments show that our model can effectively infer dynamic attention in videos. In addition, we apply our predicted attention maps to two social gaze behavior recognition tasks, and show that the resulting classifiers significantly outperform existing methods. We achieve state-of-the-art performance on three datasets: GazeFollow (static images), VideoAttentionTarget (videos), and VideoCoAtt (videos), and obtain the first results for automatically classifying clinically-relevant gaze behavior without wearable cameras or eye trackers.
['James M. Rehg', 'Nataniel Ruiz', 'Eunji Chong', 'Yongxin Wang']
2020-03-05
detecting-attended-visual-targets-in-video-1
http://openaccess.thecvf.com/content_CVPR_2020/html/Chong_Detecting_Attended_Visual_Targets_in_Video_CVPR_2020_paper.html
http://openaccess.thecvf.com/content_CVPR_2020/papers/Chong_Detecting_Attended_Visual_Targets_in_Video_CVPR_2020_paper.pdf
cvpr-2020-6
['deep-attention', 'deep-attention']
['computer-vision', 'natural-language-processing']
[ 3.78705114e-01 2.24068109e-02 -4.01607901e-01 -3.87080789e-01 -3.43976259e-01 -3.34567398e-01 1.90101311e-01 -2.73784071e-01 -3.84486049e-01 4.01857138e-01 2.74648994e-01 9.89301130e-02 7.53187835e-02 3.14973384e-01 -7.42849410e-01 -6.40921533e-01 -2.32908860e-01 -1.26895830e-01 3.83325249e-01 3.23866874e-01 5.35448372e-01 1.64297611e-01 -2.18858099e+00 4.52431768e-01 3.90208572e-01 1.10290027e+00 2.44562626e-01 1.11419594e+00 5.66452980e-01 1.29622722e+00 -5.15987039e-01 1.14367856e-02 -2.62713104e-01 -5.96468866e-01 -8.79187286e-01 3.18780035e-01 1.31217718e+00 -4.74313498e-01 -2.75555283e-01 8.91132534e-01 2.87624866e-01 5.52245714e-02 3.74155521e-01 -1.65194190e+00 -5.77831805e-01 -5.67566901e-02 -8.87251437e-01 1.22351480e+00 1.06062698e+00 3.89266700e-01 6.95820570e-01 -6.18257821e-01 5.63494205e-01 1.22718716e+00 3.65251869e-01 9.40108299e-01 -8.56413662e-01 -7.91411221e-01 4.88343596e-01 9.30645943e-01 -9.90670204e-01 -1.05575812e+00 4.45322901e-01 -6.99449122e-01 8.94339740e-01 5.06352663e-01 5.87847471e-01 1.44239199e+00 4.03299481e-01 1.01080656e+00 7.93758392e-01 -4.69531804e-01 -1.18253611e-01 -2.58491963e-01 3.98775697e-01 6.56058013e-01 -6.41862825e-02 6.97086274e-04 -1.22211456e+00 9.95492339e-02 2.40314320e-01 3.53453159e-01 -8.50581944e-01 -5.93202114e-02 -1.30004036e+00 2.19162524e-01 4.00108784e-01 8.22043642e-02 -4.40006137e-01 1.64307550e-01 4.30757403e-02 1.45190492e-01 6.27689302e-01 1.31872073e-01 -2.89559811e-01 -5.48167527e-01 -9.60724831e-01 -6.30632788e-02 4.12470102e-01 1.01101887e+00 4.15428787e-01 -4.48056102e-01 -5.70784688e-01 4.42945600e-01 3.41370016e-01 8.15704525e-01 5.51433742e-01 -1.16304696e+00 2.30302498e-01 3.20560366e-01 1.73335433e-01 -9.37203407e-01 -6.77592278e-01 4.00008410e-01 2.25985292e-02 7.75071932e-03 5.79982102e-01 -3.69652182e-01 -6.88533604e-01 1.72651803e+00 4.09360051e-01 5.47219515e-01 -5.12370527e-01 1.04976416e+00 8.79308522e-01 1.38252124e-01 2.40508005e-01 -9.11620796e-01 1.55883217e+00 -1.07068241e+00 -1.10775506e+00 -4.05411214e-01 4.15818721e-01 -5.73348582e-01 9.18170154e-01 3.51616293e-01 -1.29896593e+00 -5.49176395e-01 -3.61719280e-01 7.05729052e-02 1.76306561e-01 -9.67041627e-02 2.65556931e-01 5.51333606e-01 -1.37663186e+00 -3.12185232e-02 -9.27629530e-01 -7.13360488e-01 6.75773680e-01 5.33427298e-01 -3.46390277e-01 5.49020655e-02 -5.52959204e-01 7.15982020e-01 -3.24965239e-01 4.95946640e-03 -8.00599277e-01 -7.29784191e-01 -9.04938936e-01 5.96319251e-02 6.08793914e-01 -5.49763024e-01 1.62443030e+00 -1.70562446e+00 -1.17179883e+00 1.02455175e+00 -1.14151561e+00 -1.09485351e-01 1.22573309e-01 -5.36423266e-01 -6.74764156e-01 6.31728590e-01 -5.20592779e-02 6.63446724e-01 1.39682198e+00 -5.90982497e-01 -1.08843708e+00 -6.36642933e-01 1.28114745e-01 2.71226525e-01 -4.32230562e-01 8.38585198e-01 -4.87255245e-01 -1.30060494e-01 -5.46812236e-01 -1.05571377e+00 6.61080182e-01 9.26765278e-02 -3.85245204e-01 -5.56538880e-01 1.02757025e+00 -3.85953307e-01 1.45305634e+00 -2.25799060e+00 1.84190214e-01 -2.71005720e-01 7.28049099e-01 -8.75268411e-03 1.07722379e-01 -3.57525826e-01 -4.46101338e-01 -2.00876713e-01 5.54626048e-01 -3.48549604e-01 -4.64533716e-01 -2.02806696e-01 -4.71088476e-02 7.06203938e-01 8.81103203e-02 9.05416310e-01 -1.06285870e+00 -7.53653347e-01 3.99364643e-02 3.51961017e-01 -7.20294952e-01 5.11399806e-01 1.19217122e-02 5.98599494e-01 -2.66377181e-01 6.90149367e-01 1.40290797e-01 -6.80512428e-01 -7.66576752e-02 -9.76095796e-02 -1.07586630e-01 -6.92087738e-03 -3.68956536e-01 1.41101789e+00 9.08659101e-02 1.37529147e+00 -1.16195478e-01 -2.88424969e-01 1.33074164e-01 3.43884945e-01 6.54950678e-01 -6.78720236e-01 4.40100908e-01 -4.68451053e-01 3.65771532e-01 -1.31032121e+00 2.69994229e-01 4.06976461e-01 3.62763196e-01 6.41805410e-01 1.97184160e-01 8.35076332e-01 2.55050063e-01 -6.42235577e-02 1.29239035e+00 3.60081382e-02 3.03293228e-01 -1.64420187e-01 5.88432550e-01 -3.40621710e-01 7.89088756e-02 6.23769403e-01 -8.50080907e-01 6.24679029e-01 6.77393675e-01 -3.68442774e-01 -2.92671710e-01 -5.56420863e-01 1.42618433e-01 1.95721495e+00 1.66201934e-01 -3.10618132e-01 -1.16247380e+00 -8.95458043e-01 -3.56901079e-01 4.56126243e-01 -1.28027523e+00 -6.28079847e-02 -5.29178798e-01 -3.30982000e-01 -4.11569811e-02 4.36369389e-01 -1.38731956e-01 -1.44474125e+00 -1.13832390e+00 -5.83031237e-01 -3.91284704e-01 -1.08816385e+00 -1.22702861e+00 -5.17364264e-01 -3.73244286e-01 -1.89127004e+00 -6.32986724e-01 -5.40770113e-01 7.33120203e-01 5.69866359e-01 1.04759073e+00 2.16530040e-01 -3.49141866e-01 1.03956902e+00 -1.93441212e-01 -5.26358545e-01 6.47428855e-02 7.82947894e-03 2.63097674e-01 5.57415545e-01 1.01547098e+00 5.96208163e-02 -9.15076435e-01 4.81869608e-01 -2.54148394e-01 -1.78069592e-01 5.68392277e-02 4.99763161e-01 1.72973752e-01 -7.36818016e-01 1.02727763e-01 -6.35731936e-01 3.72134089e-01 -6.18262589e-01 -3.99359584e-01 3.44485849e-01 -2.28493884e-01 -3.70342523e-01 -1.55483589e-01 -7.18667507e-01 -9.83143091e-01 1.74116176e-02 4.97733206e-01 -1.04241383e+00 -6.14181101e-01 -1.28809676e-01 3.62028688e-01 -2.31950600e-02 7.31510818e-01 -1.37918845e-01 2.41549253e-01 -1.06869198e-01 -1.16714470e-01 7.06332028e-01 5.39319396e-01 8.86324644e-02 -9.90972668e-03 5.96728265e-01 -7.42019862e-02 -8.34539235e-01 -1.33204937e+00 -6.62410557e-01 -8.57586443e-01 -9.04021442e-01 1.13565338e+00 -8.16719472e-01 -1.59481668e+00 4.74560052e-01 -1.04348886e+00 -3.48499924e-01 1.41623184e-01 4.90319371e-01 -7.20892191e-01 1.30788302e-02 -2.43452206e-01 -9.10942018e-01 -2.72869974e-01 -1.17285550e+00 1.39579475e+00 5.54346919e-01 -5.42559147e-01 -9.12027538e-01 1.16384722e-01 2.52294898e-01 1.38110101e-01 1.42861575e-01 1.93091348e-01 -3.85345697e-01 -5.93399525e-01 1.11931711e-01 -2.45145336e-01 -1.49860099e-01 3.50532681e-01 1.74170300e-01 -1.28186953e+00 -3.02832425e-01 -3.42330262e-02 -3.29998344e-01 4.62805361e-01 8.69685233e-01 1.43933594e+00 -2.52467185e-01 -7.65048027e-01 5.89666307e-01 5.85298598e-01 2.24981472e-01 5.09121060e-01 2.12286841e-02 7.87892342e-01 7.28020608e-01 7.88442194e-01 1.95006073e-01 4.89465088e-01 8.24854195e-01 4.98822778e-01 9.17657390e-02 -2.23654136e-01 2.78380662e-01 6.22126639e-01 1.99098825e-01 -4.21527475e-01 -4.81436104e-01 -1.02409327e+00 4.75200325e-01 -1.91322458e+00 -1.46854746e+00 -3.40175956e-01 2.05946279e+00 5.02129316e-01 -1.37145475e-01 6.32560253e-01 -2.54055679e-01 9.00333881e-01 -2.90506706e-02 -9.21613216e-01 -1.40640348e-01 4.95512813e-01 -1.11445010e-01 1.40324995e-01 3.09425741e-01 -1.28153479e+00 5.66671252e-01 7.68739939e+00 -1.88654646e-01 -1.34162104e+00 3.21170539e-01 5.33487976e-01 -1.02989554e+00 5.87441862e-01 -7.20282853e-01 -8.86835396e-01 8.28391314e-01 1.41451132e+00 -4.32669409e-02 4.76061791e-01 6.28032863e-01 3.36683720e-01 -3.94026041e-01 -1.61199713e+00 1.49310923e+00 1.04941118e+00 -8.73556077e-01 -6.65956795e-01 1.09248348e-01 2.06141189e-01 3.09511334e-01 3.16318721e-01 -1.69454385e-02 -2.95321196e-01 -1.00089598e+00 7.10708439e-01 1.12251246e+00 1.07861662e+00 -3.18834782e-01 2.70998329e-01 6.60826117e-02 -6.98345184e-01 -5.65030515e-01 2.30219409e-01 -1.52049422e-01 1.29521057e-01 -3.55330318e-01 -8.08829546e-01 -3.37193817e-01 1.31507301e+00 1.39427817e+00 -1.02348101e+00 1.22172856e+00 -6.29695728e-02 5.45193136e-01 -4.86737303e-02 -3.06441244e-02 -2.97321767e-01 5.84414184e-01 5.81118941e-01 9.80974376e-01 6.74288794e-02 2.46807694e-01 -1.60594702e-01 4.53099757e-01 -9.11621228e-02 -2.78266042e-01 -4.85931039e-01 3.35066527e-01 1.45233929e-01 1.02142012e+00 -3.41945499e-01 -1.41932055e-01 -6.52781188e-01 7.73197353e-01 3.93944830e-01 4.89095777e-01 -9.86586392e-01 -1.11750655e-01 1.08902729e+00 3.82205963e-01 4.09158260e-01 3.18434149e-01 3.96816641e-01 -1.25936759e+00 8.16600118e-03 -8.24200630e-01 5.73007882e-01 -1.42195356e+00 -1.08106506e+00 6.39909327e-01 1.94698736e-01 -1.27738607e+00 -5.95481694e-01 -6.42940819e-01 -5.51399767e-01 5.98408699e-01 -1.43146074e+00 -8.44458580e-01 -1.03690135e+00 1.07007813e+00 8.17707062e-01 -7.58150369e-02 5.85565150e-01 1.20210655e-01 -9.09184396e-01 7.10638702e-01 -4.25546795e-01 -2.88296659e-02 1.17025864e+00 -9.72952187e-01 -9.81605947e-02 7.64359653e-01 -3.81267108e-02 4.74708229e-01 7.20324457e-01 -2.62512654e-01 -1.29166472e+00 -7.52802849e-01 7.62130737e-01 -1.31534851e+00 6.37494266e-01 -4.56707984e-01 -6.71078503e-01 1.08380497e+00 6.19736254e-01 2.70100921e-01 8.41097891e-01 3.95262808e-01 -9.16265026e-02 2.56953132e-03 -7.87769139e-01 4.03489739e-01 1.23945653e+00 -5.30360699e-01 -7.89710224e-01 5.35683811e-01 4.37567085e-01 -7.50085831e-01 -4.27693456e-01 8.51551592e-02 9.02392685e-01 -1.18717587e+00 6.69645369e-01 -1.05619907e+00 3.11062962e-01 6.81098998e-02 4.32788879e-01 -9.40802336e-01 -3.24730963e-01 -8.35974753e-01 -8.63955498e-01 6.72783136e-01 -1.46159334e-02 -3.57214600e-01 6.17585182e-01 9.19289410e-01 1.02678716e-01 -3.93950909e-01 -7.22479105e-01 -6.80070519e-02 -7.02217400e-01 -1.53182492e-01 3.35989803e-01 7.60922134e-01 5.09322047e-01 3.81371886e-01 -3.87348622e-01 -5.35320118e-03 4.16348219e-01 -9.01477337e-02 8.75477970e-01 -1.18438315e+00 3.98405604e-02 -3.93055648e-01 -6.59478247e-01 -9.50316727e-01 4.89745975e-01 -7.20616803e-02 1.21638171e-01 -8.01697075e-01 5.19459784e-01 4.04401153e-01 -5.06924927e-01 5.98842680e-01 -5.29593706e-01 4.69863445e-01 2.26584196e-01 3.89263302e-01 -1.32957971e+00 -5.33477664e-02 1.21981871e+00 9.84534547e-02 -1.33299276e-01 1.22481823e-01 -7.74509013e-01 7.56324947e-01 2.77957737e-01 -4.00811464e-01 -4.19863433e-01 -4.44869608e-01 -3.53790671e-02 -2.94103529e-02 4.00137335e-01 -8.44051123e-01 5.62632263e-01 -2.23795533e-01 4.95477110e-01 -5.49283683e-01 4.01470065e-01 -7.49237776e-01 -4.89437819e-01 5.46167754e-02 -4.10114855e-01 2.90867895e-01 2.93325990e-01 6.73523843e-01 7.92729259e-02 1.88261792e-01 6.74606383e-01 3.56811136e-01 -6.01897895e-01 4.00888920e-01 -2.67951101e-01 3.40767860e-01 1.35717010e+00 -2.22704768e-01 -8.69639575e-01 -4.77938563e-01 -1.02249134e+00 3.54687572e-01 5.26493788e-01 9.27340508e-01 3.71491104e-01 -8.46106231e-01 -3.13608110e-01 4.70072120e-01 2.94439316e-01 -4.11913723e-01 4.38326329e-01 1.52357495e+00 -1.09108567e-01 6.12023294e-01 -3.67677063e-01 -1.28936207e+00 -1.91986048e+00 1.00645292e+00 5.07225752e-01 6.15714133e-01 -4.85472709e-01 9.42805111e-01 7.04611123e-01 6.10905588e-01 5.00234783e-01 -4.20214027e-01 -8.33684146e-01 1.65105894e-01 1.32890666e+00 3.05612445e-01 -6.61609769e-02 -1.17912245e+00 -5.33164322e-01 4.61280912e-01 -2.93941706e-01 5.09549141e-01 1.01642323e+00 -6.21285498e-01 -3.95280272e-02 6.93987131e-01 9.88183498e-01 -2.93967217e-01 -1.49927080e+00 -9.30080414e-02 -1.36099413e-01 -8.02269220e-01 9.95086730e-02 -6.06297016e-01 -1.08367443e+00 8.90762269e-01 9.98048067e-01 2.71690756e-01 1.45594597e+00 4.31682348e-01 3.77477974e-01 2.01363459e-01 1.08016573e-01 -5.72470903e-01 3.05070430e-01 2.19694316e-01 7.30500996e-01 -1.46998501e+00 -1.25583574e-01 -2.91849613e-01 -7.28175819e-01 8.73185813e-01 9.21230853e-01 1.15594544e-01 8.05813551e-01 -1.44011930e-01 9.19946060e-02 -5.63542187e-01 -1.20317996e+00 -3.99887234e-01 7.72444189e-01 6.89859807e-01 4.92116302e-01 -3.30621123e-01 4.25331801e-01 8.79670307e-02 -9.48940590e-03 2.60992706e-01 5.47248840e-01 6.87075555e-01 -2.78364420e-01 -4.01840687e-01 -5.71930587e-01 6.09257221e-01 -8.31530511e-01 6.02452010e-02 -3.67579311e-01 6.06747806e-01 -9.69051793e-02 1.10186315e+00 7.75118172e-01 -1.18975423e-01 3.66304398e-01 2.71791309e-01 7.05422103e-01 -6.60053909e-01 -4.26879019e-01 1.81911275e-01 -1.25282958e-01 -1.20129383e+00 -1.09145820e+00 -1.20647585e+00 -5.67704260e-01 -2.01048315e-01 -2.06421003e-01 -3.50458294e-01 7.40192533e-02 1.02916133e+00 6.49502754e-01 7.27653086e-01 4.34345990e-01 -1.16780436e+00 3.05572987e-01 -1.13845026e+00 -3.82177383e-01 6.54286087e-01 1.20418882e+00 -1.07407463e+00 -2.85541773e-01 7.47449398e-01]
[14.04182243347168, 0.08404454588890076]
af723e39-f07a-4be9-8907-64c9c9377f2a
gated-multimodal-units-for-information-fusion
1702.01992
null
http://arxiv.org/abs/1702.01992v1
http://arxiv.org/pdf/1702.01992v1.pdf
Gated Multimodal Units for Information Fusion
This paper presents a novel model for multimodal learning based on gated neural networks. The Gated Multimodal Unit (GMU) model is intended to be used as an internal unit in a neural network architecture whose purpose is to find an intermediate representation based on a combination of data from different modalities. The GMU learns to decide how modalities influence the activation of the unit using multiplicative gates. It was evaluated on a multilabel scenario for genre classification of movies using the plot and the poster. The GMU improved the macro f-score performance of single-modality approaches and outperformed other fusion strategies, including mixture of experts models. Along with this work, the MM-IMDb dataset is released which, to the best of our knowledge, is the largest publicly available multimodal dataset for genre prediction on movies.
['Fabio A. González', 'Manuel Montes-y-Gómez', 'Thamar Solorio', 'John Arevalo']
2017-02-07
null
null
null
null
['genre-classification']
['computer-vision']
[ 3.80044192e-01 -1.44846171e-01 -4.15644795e-01 -5.73979795e-01 -9.33774531e-01 -7.38714278e-01 6.02999926e-01 3.43739301e-01 -3.56594890e-01 6.63523734e-01 4.18030709e-01 -1.85035005e-01 7.08201677e-02 -3.63560230e-01 -7.92230189e-01 -7.68112838e-01 2.52713233e-01 4.43280786e-01 -2.18856812e-01 -2.18894437e-01 2.08118781e-02 4.66322936e-02 -1.54952085e+00 1.62043393e+00 1.93623841e-01 1.36735976e+00 1.34911053e-02 7.28530467e-01 2.24932373e-01 1.17304981e+00 -3.05536270e-01 -5.78838348e-01 -2.73582280e-01 -4.10288930e-01 -8.67249906e-01 -2.04900295e-01 6.60566092e-01 2.35557575e-02 -2.71434844e-01 7.44558394e-01 6.88308358e-01 4.36184108e-01 8.71493638e-01 -8.41234326e-01 -3.57231796e-01 1.32858205e+00 1.91405583e-02 -2.55437312e-03 6.97670639e-01 -2.69130886e-01 1.42144740e+00 -8.90884280e-01 9.43856895e-01 1.42234576e+00 4.95707572e-01 7.63865829e-01 -1.46565831e+00 -1.80435583e-01 2.25246042e-01 3.50664914e-01 -1.18401694e+00 -3.95555645e-01 4.75287199e-01 -3.82432222e-01 1.11726093e+00 4.24566597e-01 1.56765848e-01 1.52557755e+00 2.11327493e-01 1.12805450e+00 1.17941797e+00 -6.53219283e-01 1.98984668e-01 1.57285511e-01 4.09215167e-02 7.20177889e-01 -5.72392404e-01 -2.43947700e-01 -1.03655291e+00 -2.93291718e-01 4.84864771e-01 -2.37883016e-01 -2.37556115e-01 1.62063941e-01 -1.23326027e+00 7.64992893e-01 3.62931132e-01 3.52326065e-01 -1.49556786e-01 7.93224052e-02 4.19888467e-01 4.42111313e-01 3.47192973e-01 4.38410103e-01 -5.98009765e-01 9.63340029e-02 -1.01944101e+00 -6.83182999e-02 8.37004244e-01 4.33799416e-01 2.54202008e-01 -2.78866559e-01 -4.63849425e-01 9.36948180e-01 7.06771731e-01 1.83211699e-01 2.10679010e-01 -1.12256122e+00 5.14995813e-01 6.11979187e-01 -2.92662770e-01 -5.32523513e-01 -6.47292554e-01 -1.91441074e-01 -6.44675195e-01 1.80503562e-01 5.88048756e-01 -2.09813625e-01 -8.79653454e-01 1.78673816e+00 -1.81144059e-01 -9.59755406e-02 3.89525071e-02 1.01773775e+00 1.66208267e+00 6.49746180e-01 4.83814538e-01 -1.17507406e-01 1.24311805e+00 -8.47488105e-01 -6.16899490e-01 1.71218276e-01 5.60095251e-01 -8.84697378e-01 3.76080960e-01 9.26595747e-01 -1.12995172e+00 -8.19657445e-01 -9.98710275e-01 1.04068227e-01 -5.95387220e-01 6.10792100e-01 5.99033594e-01 5.59138536e-01 -1.15350568e+00 5.73485076e-01 -3.21749300e-01 -4.02699918e-01 2.98118293e-01 6.87069118e-01 -7.18298376e-01 -8.39058831e-02 -1.49885869e+00 8.04641604e-01 7.05033720e-01 2.86399554e-02 -1.23465967e+00 -1.50780797e-01 -8.10204744e-01 -5.98597266e-02 4.14689779e-02 -7.22392559e-01 1.16277158e+00 -1.36723971e+00 -1.41896284e+00 9.46581364e-01 -4.31092270e-03 -5.17907321e-01 -3.09996330e-03 2.71577895e-01 -5.37997067e-01 2.41340205e-01 -5.13576686e-01 1.18324995e+00 8.27970028e-01 -1.53620815e+00 -6.69676363e-01 -3.42847079e-01 2.41003618e-01 2.93776035e-01 -8.85830447e-02 2.02194929e-01 -3.77511710e-01 -5.79899311e-01 -9.39119607e-02 -9.57080483e-01 -8.88310820e-02 -7.84372985e-01 -3.08700025e-01 -4.70570207e-01 3.39345425e-01 -7.49826729e-01 1.35019553e+00 -1.90989172e+00 8.14094961e-01 2.03486025e-01 2.20856115e-01 -1.59049287e-01 -2.07439855e-01 5.88507414e-01 -8.65860283e-02 5.61654791e-02 1.99731156e-01 -8.54368746e-01 1.29808635e-01 1.75140426e-01 -1.83172509e-01 2.09703222e-01 -1.20935403e-01 7.09896147e-01 -4.26875174e-01 -4.24142420e-01 1.10012032e-01 5.94206214e-01 -4.16104496e-01 7.35134855e-02 -4.73131537e-01 5.64901710e-01 6.45098463e-02 1.08973086e+00 3.12797487e-01 -1.96112692e-01 5.70703268e-01 -5.30389011e-01 1.02226399e-01 -1.58020943e-01 -9.27180231e-01 2.20145583e+00 -1.65914685e-01 7.19898939e-01 5.88011481e-02 -7.03262150e-01 7.18271554e-01 7.14267969e-01 2.23539144e-01 -3.69245380e-01 7.27912009e-01 1.06889017e-01 -1.94508359e-02 -3.08636010e-01 6.17282510e-01 -2.35052571e-01 -4.39110100e-01 2.48731747e-01 1.12711203e+00 3.85722429e-01 5.00317335e-01 4.56999958e-01 8.31526101e-01 4.96200472e-01 -3.14545110e-02 2.75094897e-01 6.18926883e-01 -3.11728835e-01 2.66127169e-01 8.94730449e-01 1.14663817e-01 7.97851503e-01 6.72368348e-01 -2.48721138e-01 -4.94517773e-01 -9.29975212e-01 -6.44538552e-02 1.77286613e+00 -3.21499944e-01 -6.56642020e-01 -6.44275904e-01 -7.46731520e-01 -3.33298922e-01 7.19534814e-01 -1.03323328e+00 3.63739505e-02 -1.62918419e-02 -8.10211360e-01 7.24994659e-01 5.03706038e-01 -1.14586741e-01 -1.00167227e+00 -4.08570021e-02 -3.71850990e-02 -5.19177496e-01 -1.12890148e+00 8.76796339e-03 7.61679053e-01 -6.60219967e-01 -9.54648376e-01 -4.29877043e-01 -8.26752067e-01 5.28312624e-01 -5.02513587e-01 1.27376759e+00 -3.77181083e-01 2.30325520e-01 8.02134573e-01 -6.66291416e-01 -2.24905670e-01 -5.55340648e-01 1.68757766e-01 -9.02561247e-02 5.70949912e-01 2.60845810e-01 -1.06051728e-01 -3.47954124e-01 1.51941985e-01 -9.91238177e-01 6.44808710e-02 4.48325872e-01 1.00591648e+00 6.90696657e-01 -1.37996182e-01 3.16288799e-01 -6.46460414e-01 6.92267537e-01 -5.73070586e-01 -7.20986202e-02 5.64463556e-01 -1.26749218e-01 -2.20763370e-01 4.20162857e-01 -5.00507593e-01 -1.31434834e+00 3.96477878e-01 -8.72508660e-02 -4.10892516e-01 -5.38920939e-01 7.30813801e-01 -2.49599144e-01 -1.43417567e-01 5.75361490e-01 -2.60819107e-01 -4.57890451e-01 -6.43888533e-01 4.96424079e-01 6.89595103e-01 2.97763377e-01 -6.60321891e-01 -4.31379020e-01 2.64350474e-01 1.28941000e-01 -3.48060101e-01 -8.36540341e-01 -3.87805462e-01 -9.09097135e-01 -9.51420426e-01 1.27466607e+00 -1.00586283e+00 -7.23025560e-01 2.58822113e-01 -1.07039642e+00 -2.58326679e-01 1.16458386e-01 5.99992871e-01 -4.24724549e-01 3.22297178e-02 -8.21746171e-01 -7.96864569e-01 -5.16840965e-02 -1.04437566e+00 1.01182044e+00 3.42105716e-01 -2.96818405e-01 -1.39747155e+00 2.37409100e-01 8.11982810e-01 -5.24871163e-02 2.34231293e-01 8.63842249e-01 -1.06086087e+00 4.28460818e-03 -2.88251102e-01 3.15832049e-01 5.97731590e-01 -5.25429785e-01 5.99654950e-02 -1.46377242e+00 -2.13629857e-01 -5.26428103e-01 -8.54638994e-01 1.43016565e+00 4.68709648e-01 9.32749450e-01 2.50412732e-01 -1.35515064e-01 5.45512997e-02 1.33192861e+00 9.68328416e-02 4.18270707e-01 8.95784050e-02 6.35230482e-01 5.36512673e-01 2.38048673e-01 4.66286868e-01 2.01768503e-01 4.52391148e-01 6.42979503e-01 5.42863458e-02 -1.13300189e-01 -4.98269275e-02 5.75300038e-01 7.94610977e-01 -4.49707597e-01 -7.81901836e-01 -9.25126970e-01 -3.79661396e-02 -2.11769867e+00 -8.93248856e-01 -4.55475673e-02 2.14131117e+00 6.15780592e-01 1.60536710e-02 7.28321075e-02 -8.39730054e-02 5.01557052e-01 5.60226329e-02 -1.09234797e-02 -7.33479559e-01 -8.72064769e-01 1.87664032e-01 1.17562249e-01 3.96273047e-01 -1.60967267e+00 7.06227541e-01 6.79590321e+00 1.06438053e+00 -7.60774910e-01 4.14775580e-01 9.50962782e-01 -3.39368522e-01 -4.80317175e-02 -5.26646897e-02 -8.04706097e-01 2.57851481e-01 1.32781386e+00 8.06089103e-01 4.49109584e-01 2.46410578e-01 -1.02984518e-01 -4.33869809e-01 -1.17065203e+00 8.51913571e-01 5.98702788e-01 -1.07476664e+00 3.37823212e-01 -2.01272160e-01 1.02177906e+00 1.25738204e-01 2.72747546e-01 4.95175451e-01 1.18049093e-01 -1.35404515e+00 6.68462694e-01 1.17588794e+00 4.92787391e-01 -9.29337323e-01 1.08510494e+00 2.08005175e-01 -8.92829657e-01 -2.78395206e-01 -2.13613287e-01 4.53028716e-02 2.27336258e-01 -1.74949802e-02 -5.60650885e-01 6.84114695e-01 5.19895375e-01 5.51360846e-01 -8.36196601e-01 8.06435168e-01 -7.16772974e-02 8.35541725e-01 6.92524910e-02 3.72140780e-02 3.76102954e-01 1.27096280e-01 4.86626863e-01 1.59954655e+00 1.31055325e-01 -2.55801171e-01 2.66997933e-01 3.28569084e-01 -3.24810416e-01 4.61565137e-01 -2.99139589e-01 -1.46764427e-01 -1.58767015e-01 1.64861715e+00 -6.53298438e-01 -3.09733361e-01 -5.16204834e-01 7.33337998e-01 2.74006188e-01 4.29490089e-01 -7.85584688e-01 9.69121456e-02 -1.35502174e-01 -3.29721332e-01 3.49253058e-01 3.53935063e-01 -2.15372220e-01 -1.01761246e+00 -7.32065737e-01 -9.70370531e-01 9.34451878e-01 -1.16815364e+00 -1.41548574e+00 7.62839437e-01 -8.80426448e-03 -1.43707061e+00 -2.61780590e-01 -1.05565619e+00 -1.16220668e-01 7.31629848e-01 -4.74002779e-01 -1.69773602e+00 2.22027093e-01 7.10863233e-01 2.43556380e-01 -7.78955400e-01 1.30201805e+00 3.06614608e-01 -4.76152420e-01 6.26072407e-01 2.17352845e-02 6.98159188e-02 9.78551328e-01 -1.31542599e+00 -8.28776240e-01 3.90942603e-01 4.02173370e-01 6.42552197e-01 5.72304666e-01 -4.67589408e-01 -1.03118503e+00 -7.76059866e-01 8.69346976e-01 -6.31596625e-01 5.92856765e-01 -1.44014060e-01 -3.95124376e-01 5.90239823e-01 9.87146735e-01 -4.69893694e-01 1.54955971e+00 3.18593919e-01 -4.61640656e-01 9.26286057e-02 -9.98334467e-01 1.63817346e-01 3.26507628e-01 -6.49621844e-01 -4.22685832e-01 3.11485827e-01 2.45313108e-01 -3.69604677e-01 -1.51982832e+00 6.45946741e-01 7.90718079e-01 -8.15227985e-01 6.52968228e-01 -1.05424976e+00 8.32693696e-01 -2.49937519e-01 -8.94811928e-01 -1.31077063e+00 -3.40068221e-01 -3.52455080e-01 -5.09082556e-01 1.29862523e+00 9.30998147e-01 3.90964262e-02 2.61727870e-01 3.04418832e-01 1.79232117e-02 -7.64426351e-01 -1.18189895e+00 -3.31480708e-03 -1.20422244e-01 -8.04960310e-01 1.81174651e-02 8.64034712e-01 2.00600445e-01 7.73302734e-01 -7.27845371e-01 -2.99877107e-01 3.67244035e-01 -6.01655133e-02 1.26134470e-01 -1.04063940e+00 -4.93900239e-01 -5.79968870e-01 -4.41310942e-01 -7.67994940e-01 2.96385825e-01 -1.34647679e+00 -1.87880009e-01 -1.47722876e+00 4.46443945e-01 2.64153302e-01 -1.19158661e+00 6.57662570e-01 2.00202391e-01 8.05707991e-01 3.75350177e-01 -7.55679086e-02 -1.39088392e+00 6.29101768e-02 8.89014840e-01 -4.04410034e-01 2.70410161e-02 1.04529396e-01 -5.24372816e-01 7.45855391e-01 1.01820374e+00 -3.45065504e-01 -8.98545757e-02 -2.95726568e-01 6.99807048e-01 1.72091812e-01 2.45578557e-01 -8.58356953e-01 2.93236166e-01 3.99038553e-01 9.77746427e-01 -7.98593402e-01 7.01843619e-01 -6.31096721e-01 1.60820171e-01 -1.14839017e-01 -7.74238706e-01 -1.19399570e-01 2.28751615e-01 5.48959672e-01 -4.12236661e-01 -3.64677846e-01 3.36014479e-01 -1.75746217e-01 -7.29870081e-01 -1.36560842e-01 -8.85792255e-01 -6.37452543e-01 5.80611050e-01 2.08485290e-01 -2.52745003e-01 -5.03169000e-01 -1.59666705e+00 -8.48494843e-02 8.98112282e-02 6.74641550e-01 5.35843968e-01 -1.60030925e+00 -6.75458610e-01 -3.13514084e-01 2.02668190e-01 -9.30476367e-01 5.80309153e-01 1.12380779e+00 -5.31468727e-02 6.82456374e-01 -1.93325177e-01 -5.92908740e-01 -1.73168218e+00 2.97585696e-01 3.39152634e-01 -5.22974491e-01 5.71602881e-01 1.12116241e+00 -2.98869342e-01 -7.41105795e-01 5.06832361e-01 6.09018952e-02 -9.51866090e-01 7.80128181e-01 4.56632614e-01 3.50719184e-01 1.79908559e-01 -1.12126434e+00 -2.26427212e-01 1.92788675e-01 1.55623391e-01 -4.87045974e-01 1.10810399e+00 -6.20818846e-02 -4.62080568e-01 1.12850666e+00 1.01293695e+00 -1.19061261e-01 -7.21253514e-01 -1.18179709e-01 -5.09395242e-01 2.39017218e-01 2.76995063e-01 -1.50979173e+00 -1.02102208e+00 8.40673685e-01 8.27986062e-01 2.52657473e-01 1.29208505e+00 1.21124223e-01 1.51321679e-01 4.26343530e-01 8.92487019e-02 -1.33857644e+00 -1.21981472e-01 5.41713536e-01 7.42158711e-01 -1.31979930e+00 4.32380736e-02 1.92472056e-01 -1.04861259e+00 1.51707470e+00 5.91499567e-01 3.92504483e-01 6.08243763e-01 -1.32610127e-02 1.05357058e-01 -1.20851316e-01 -1.17370045e+00 -2.25689128e-01 1.00862455e+00 1.51357412e-01 8.27765882e-01 3.62833023e-01 -3.56409907e-01 1.09625053e+00 4.27703172e-01 -5.23144267e-02 -2.04387102e-02 7.72406459e-01 -2.31044218e-01 -1.16785979e+00 -5.59991241e-01 4.66755658e-01 -7.29293764e-01 -1.86337903e-01 -8.81729960e-01 3.28281939e-01 8.09262574e-01 1.38619626e+00 -2.21470520e-01 -9.30476546e-01 5.22579029e-02 4.00771767e-01 8.72309089e-01 -3.23117584e-01 -1.12775350e+00 4.77369189e-01 4.97661293e-01 -3.83550733e-01 -9.53902364e-01 -7.04814494e-01 -8.50420296e-01 3.82561646e-02 -1.76889598e-01 -4.07008175e-03 6.18539691e-01 8.04484308e-01 3.34516503e-02 6.40481472e-01 4.08086091e-01 -1.05040681e+00 6.01627491e-02 -1.28792322e+00 -5.51961184e-01 2.97752559e-01 -2.07097709e-01 -5.04291177e-01 -1.58827782e-01 2.19068319e-01]
[13.226664543151855, 5.112884521484375]
017681ce-1dd5-40ce-8c22-defe51d13312
aidroid-when-heterogeneous-information
1811.01027
null
https://arxiv.org/abs/1811.01027v2
https://arxiv.org/pdf/1811.01027v2.pdf
AiDroid: When Heterogeneous Information Network Marries Deep Neural Network for Real-time Android Malware Detection
The explosive growth and increasing sophistication of Android malware call for new defensive techniques that are capable of protecting mobile users against novel threats. In this paper, we first extract the runtime Application Programming Interface (API) call sequences from Android apps, and then analyze higher-level semantic relations within the ecosystem to comprehensively characterize the apps. To model different types of entities (i.e., app, API, IMEI, signature, affiliation) and the rich semantic relations among them, we then construct a structural heterogeneous information network (HIN) and present meta-path based approach to depict the relatedness over apps. To efficiently classify nodes (e.g., apps) in the constructed HIN, we propose the HinLearning method to first obtain in-sample node embeddings and then learn representations of out-of-sample nodes without rerunning/adjusting HIN embeddings at the first attempt. Afterwards, we design a deep neural network (DNN) classifier taking the learned HIN representations as inputs for Android malware detection. A comprehensive experimental study on the large-scale real sample collections from Tencent Security Lab is performed to compare various baselines. Promising experimental results demonstrate that our developed system AiDroid which integrates our proposed method outperforms others in real-time Android malware detection. AiDroid has already been incorporated into Tencent Mobile Security product that serves millions of users worldwide.
['Shifu Hou', 'Wenqiang Wan', 'Qi Xiong', 'Lingwei Chen', 'Fudong Shao', 'Yanfang Ye', 'Jiabin Wang', 'Jingwei Lei']
2018-11-02
null
null
null
null
['android-malware-detection', 'mobile-security']
['miscellaneous', 'miscellaneous']
[ 2.08482146e-01 -2.83526808e-01 -5.84844828e-01 -9.65575799e-02 -3.34443718e-01 -8.61165643e-01 4.80485678e-01 -9.39308405e-02 -1.55815324e-02 3.62234950e-01 6.47760034e-02 -8.41447234e-01 -2.59121120e-01 -8.57751787e-01 -7.07440078e-01 -2.57201880e-01 -3.74328732e-01 2.13068366e-01 5.18992543e-01 -5.02687134e-02 2.45371535e-01 3.48218471e-01 -1.51859534e+00 3.05418819e-01 7.29633152e-01 1.01928353e+00 -1.12574525e-01 7.55672693e-01 -1.67388618e-01 6.74309015e-01 -7.57851422e-01 -7.59874165e-01 1.34240434e-01 1.79496985e-02 -6.05059028e-01 -3.36666554e-01 8.76284391e-03 -3.31496239e-01 -2.97266126e-01 1.28606856e+00 1.84278831e-01 -1.74345046e-01 5.97464085e-01 -1.53042185e+00 -4.69530046e-01 7.62259543e-01 -7.38378942e-01 4.32076007e-01 3.40672910e-01 -1.07277110e-01 1.02709234e+00 -6.06022418e-01 6.52309835e-01 1.03746748e+00 9.41822529e-01 4.90871221e-01 -6.10988379e-01 -7.98891842e-01 1.32451177e-01 3.72377992e-01 -1.25367534e+00 -1.47484943e-01 8.77328813e-01 -4.32230443e-01 6.60631537e-01 3.70644808e-01 3.97355288e-01 1.66631269e+00 4.25511837e-01 4.46165860e-01 6.86386287e-01 2.97164768e-01 1.07279435e-01 3.69675130e-01 7.79946148e-01 8.99459600e-01 6.04341924e-01 -2.69051105e-01 -1.49482712e-01 -7.22198009e-01 -1.62865929e-02 5.60118377e-01 -3.94067504e-02 -1.97099537e-01 -5.69635451e-01 6.95130527e-01 4.86165524e-01 2.56133318e-01 -2.60064483e-01 -1.03878498e-01 8.27574313e-01 -1.27667055e-01 1.76461995e-01 1.55460030e-01 -6.67804897e-01 -3.70119750e-01 -4.82984215e-01 -5.33343591e-02 9.26262498e-01 8.11133444e-01 7.68199682e-01 -1.23088807e-01 3.46881896e-01 7.31617391e-01 4.94517833e-01 1.66509911e-01 9.92618084e-01 -5.43510437e-01 5.06197572e-01 1.16521311e+00 -4.65710521e-01 -1.64621282e+00 -8.92587285e-03 -3.94878417e-01 -6.36127174e-01 -3.84492815e-01 -1.80718869e-01 -1.87399179e-01 -5.65533161e-01 1.53573501e+00 4.56052184e-01 6.82479203e-01 -9.25992876e-02 9.25455242e-02 5.80422521e-01 6.08933330e-01 5.94702773e-02 -3.12782302e-02 1.62851357e+00 -7.39063025e-01 -3.71220618e-01 -2.82654092e-02 6.18129015e-01 -3.30120206e-01 1.09209907e+00 1.97180718e-01 -3.77748668e-01 -4.94843662e-01 -1.17502666e+00 1.26564592e-01 -9.88607466e-01 7.12342262e-02 4.34042692e-01 9.55245256e-01 -6.70207739e-01 3.22324097e-01 -9.13649201e-01 -2.17906445e-01 8.08066785e-01 5.97988784e-01 -3.23670149e-01 2.70826221e-01 -1.05272949e+00 8.67952779e-02 3.77555460e-01 -7.39824548e-02 -1.22030401e+00 -7.44830549e-01 -7.31873214e-01 2.05713242e-01 8.07848692e-01 -1.49058774e-01 9.39854980e-01 -7.00930417e-01 -1.16814148e+00 4.38501120e-01 4.35736583e-04 -2.92607576e-01 -2.61599183e-01 -1.79061353e-01 -8.08818579e-01 -4.95333858e-02 -1.81986988e-02 1.75983280e-01 9.46175516e-01 -1.37535644e+00 -6.00358903e-01 -5.55336654e-01 3.64491105e-01 -2.47960404e-01 -1.03253877e+00 9.84677970e-02 -5.81698775e-01 -2.75040507e-01 -4.58766967e-01 -1.10132372e+00 1.91767573e-01 -5.72006166e-01 -9.15424585e-01 -3.27486187e-01 1.45490980e+00 -9.20794904e-01 1.66717327e+00 -2.29912806e+00 -3.39021459e-02 3.95434588e-01 6.25570595e-01 6.62806332e-01 -7.87986144e-02 3.00457805e-01 -9.50043648e-02 5.53575456e-01 -3.44618410e-01 -2.66091347e-01 -1.21137597e-01 2.02408701e-01 -3.64972681e-01 8.53847787e-02 -5.54639734e-02 8.13502371e-01 -7.47554958e-01 -4.73996103e-01 -1.47017449e-01 6.12213910e-01 -6.76599205e-01 5.69996610e-02 -2.33478338e-01 -3.94470766e-02 -6.92609608e-01 1.03098691e+00 6.49831951e-01 -2.79132873e-01 6.10527039e-01 -3.28319460e-01 3.32011461e-01 4.76493537e-01 -6.20414674e-01 9.30503190e-01 -5.32689333e-01 5.22655606e-01 9.45044681e-02 -7.47350216e-01 6.40831888e-01 -1.58170253e-01 4.23107833e-01 -2.03623861e-01 3.01248193e-01 3.94697301e-02 3.28875065e-01 -5.85377276e-01 4.19591248e-01 6.83805585e-01 -2.81508446e-01 3.90252739e-01 -8.67788941e-02 9.13871884e-01 -2.77774572e-01 3.79733473e-01 1.47974753e+00 -2.62458891e-01 3.24425846e-01 -9.19584781e-02 1.01732349e+00 -2.50573337e-01 4.67373997e-01 2.16422662e-01 -2.83593357e-01 -3.08547407e-01 1.03583324e+00 -3.12427014e-01 -4.22085702e-01 -1.09992373e+00 2.30360031e-01 1.28974104e+00 2.14029357e-01 -9.10651147e-01 -1.09914768e+00 -1.63339412e+00 -1.01565048e-01 1.78818926e-01 -7.25150585e-01 -3.22503269e-01 -6.97848082e-01 -5.65196037e-01 9.22526360e-01 3.68370980e-01 7.03556061e-01 -9.27291572e-01 -2.76908189e-01 -2.22148344e-01 9.39911529e-02 -1.12776470e+00 -4.06679869e-01 -1.60927162e-01 -4.80381846e-01 -1.52037418e+00 8.38276744e-02 -7.58335829e-01 5.26091397e-01 2.87613630e-01 8.04649830e-01 2.39765316e-01 -2.86944747e-01 4.00020778e-01 -2.01430544e-01 5.35337022e-03 -2.53308862e-01 5.65863967e-01 2.41650313e-01 2.50787646e-01 5.13402939e-01 -6.41082287e-01 -4.16161180e-01 4.21798289e-01 -9.69750047e-01 -9.01117742e-01 4.01581496e-01 4.73954976e-01 3.96022409e-01 5.34023523e-01 4.44724590e-01 -1.26765060e+00 9.87427294e-01 -1.18874490e+00 -4.68333930e-01 1.02804616e-01 -4.47819382e-01 -4.12867278e-01 8.84568572e-01 -7.34798014e-01 -8.70934963e-01 -2.64887273e-01 -4.08209831e-01 -3.54058594e-01 4.59091030e-02 5.28253675e-01 -7.61811674e-01 -1.78103577e-02 4.86859232e-01 1.41786873e-01 -1.01796657e-01 -4.19577658e-01 3.81376773e-01 1.19010234e+00 3.48054767e-01 -4.03782994e-01 8.15241992e-01 2.40990251e-01 -1.56111047e-01 -9.52928901e-01 -3.32762778e-01 -2.24172980e-01 -1.15180351e-01 1.30982339e-01 8.20951819e-01 -4.81373668e-01 -1.05995774e+00 3.38159651e-01 -9.11426783e-01 3.12662981e-02 2.51726419e-01 -4.71382849e-02 8.65213051e-02 5.77197909e-01 -5.72478771e-01 -5.69790363e-01 -2.74439842e-01 -1.43720913e+00 1.07543337e+00 2.68385887e-01 -2.87881434e-01 -1.13097155e+00 1.27057791e-01 3.74740511e-01 2.00548157e-01 8.76764879e-02 1.24984002e+00 -1.27335930e+00 -5.73351860e-01 -3.99214536e-01 -2.82015294e-01 4.10538495e-01 5.21537721e-01 3.58962566e-01 -8.59043419e-01 -7.23346397e-02 1.75761834e-01 2.14814037e-01 4.69067931e-01 -1.75879315e-01 1.72571039e+00 -7.72601604e-01 -9.17064488e-01 7.21531928e-01 1.20313871e+00 5.06844223e-01 6.13778234e-01 1.17339134e-01 1.21960664e+00 7.34674156e-01 2.23443598e-01 1.30622759e-01 4.58116204e-01 6.05141580e-01 6.84361279e-01 5.75393260e-01 9.09461305e-02 -5.67630470e-01 7.39541173e-01 6.97579861e-01 2.05052540e-01 -3.38025749e-01 -7.43863881e-01 1.91938907e-01 -1.33724928e+00 -7.28205323e-01 2.70775650e-02 1.96767461e+00 5.48527896e-01 2.42985547e-01 4.97746497e-01 1.43004075e-01 8.07260573e-01 2.23606855e-01 -6.18436277e-01 -5.66837847e-01 5.13984442e-01 1.64452359e-01 4.12173659e-01 5.17093182e-01 -9.82341826e-01 9.66054738e-01 5.27473688e+00 1.05989408e+00 -1.10397506e+00 4.61791933e-01 5.22900641e-01 3.17703515e-01 -2.88426220e-01 -3.31242472e-01 -1.07741690e+00 9.22645092e-01 1.40036643e+00 1.39325485e-01 5.39190829e-01 1.08638728e+00 -2.60096580e-01 5.07183373e-01 -6.65710270e-01 7.69451201e-01 1.80971056e-01 -1.28452098e+00 6.10859282e-02 5.69194496e-01 2.88619816e-01 7.47864768e-02 3.26643318e-01 6.19931340e-01 1.86014861e-01 -7.20994771e-01 -1.62806749e-01 2.43144438e-01 6.57222748e-01 -9.71931636e-01 6.44745111e-01 2.98795223e-01 -1.33784783e+00 -6.46947086e-01 -1.07720293e-01 3.88693333e-01 -3.20734441e-01 2.42252812e-01 -9.57608640e-01 5.05934298e-01 7.79583037e-01 9.59200680e-01 -1.01227546e+00 3.97931010e-01 -7.54317567e-02 8.49008501e-01 -4.53541949e-02 -2.70141333e-01 4.43820953e-02 -1.11814871e-01 5.20271480e-01 9.56440210e-01 3.65913182e-01 -3.89682800e-01 -8.67874920e-02 6.80270553e-01 -4.92093652e-01 -3.76007939e-03 -1.01450193e+00 -6.70211315e-01 6.60236061e-01 1.74935675e+00 -6.36678398e-01 -3.95177543e-01 -2.74787366e-01 8.49986315e-01 1.89651802e-01 1.21798590e-01 -1.02858770e+00 -5.41387677e-01 1.13516045e+00 2.96708554e-01 2.59474009e-01 -2.98806816e-01 4.01474275e-02 -8.93182456e-01 8.08797684e-03 -1.27832425e+00 2.21092343e-01 -3.86105895e-01 -1.23141086e+00 9.01339412e-01 3.60714900e-03 -1.06446791e+00 -3.34607482e-01 -9.74637747e-01 -8.45141768e-01 3.92256469e-01 -9.10736322e-01 -1.09641612e+00 -2.34339774e-01 6.37229145e-01 4.88695413e-01 -5.70380569e-01 5.91298759e-01 5.57584763e-01 -1.09341264e+00 9.04990315e-01 -8.72918665e-02 2.64835954e-01 9.07076895e-02 -9.03427660e-01 5.41449606e-01 6.86610937e-01 1.16886914e-01 1.37183273e+00 4.55949940e-02 -1.23476005e+00 -1.57472849e+00 -1.27645171e+00 5.10236442e-01 -8.09396148e-01 1.18432176e+00 -7.37779438e-01 -9.83879983e-01 8.64699543e-01 1.49003536e-01 -1.23441359e-02 9.11847353e-01 5.19829392e-02 -7.40489721e-01 -2.72087336e-01 -1.19974530e+00 7.40844846e-01 1.24789882e+00 -7.35267460e-01 -1.15299881e-01 1.61270902e-01 1.09436691e+00 5.78281023e-02 -7.36932456e-01 4.87342983e-01 6.43600166e-01 -9.13604498e-01 1.04216707e+00 -7.74399400e-01 2.64910251e-01 -1.81841806e-01 -3.21837097e-01 -7.97828913e-01 1.29634291e-01 -7.54880071e-01 -7.86683917e-01 1.43603718e+00 4.02893156e-01 -8.33030045e-01 9.61008728e-01 -1.62930414e-01 -1.25222936e-01 -1.13401139e+00 -7.46813893e-01 -5.91195583e-01 -3.45063210e-01 -4.77798194e-01 9.17586088e-01 7.87022531e-01 -2.64526099e-01 4.38239127e-01 -2.31216386e-01 3.98097456e-01 5.09230435e-01 -4.68481809e-01 7.02631533e-01 -1.35998201e+00 -4.76144880e-01 -4.08045053e-01 -6.73084378e-01 -7.71951199e-01 5.20513296e-01 -8.39243412e-01 -4.26562846e-01 -8.10967624e-01 1.78185612e-01 -3.71186584e-01 -3.09852839e-01 3.18521410e-01 5.07709896e-03 3.28058720e-01 -1.78112745e-01 1.29165590e-01 -5.23973823e-01 1.94010451e-01 4.94894385e-01 -3.26733768e-01 -2.77323186e-01 3.34404737e-01 -1.09538090e+00 9.02291417e-01 8.64771962e-01 -4.37616438e-01 -8.53778780e-01 -3.13416161e-02 2.64125496e-01 -5.48547089e-01 3.27299714e-01 -8.98817837e-01 -2.82692518e-02 3.08687866e-01 8.08175951e-02 -5.28658032e-01 3.18321884e-01 -1.06700063e+00 5.09611331e-02 5.25819540e-01 -1.61760122e-01 3.90416831e-01 1.03395686e-01 9.09061313e-01 1.06969140e-01 -3.00753683e-01 3.78900766e-01 2.22962514e-01 -5.27362704e-01 4.39786136e-01 -4.27168667e-01 -1.55203313e-01 1.16150820e+00 -1.32257536e-01 -6.98436320e-01 -5.17172681e-04 -4.27981496e-01 -2.02738941e-01 3.34555715e-01 8.45384896e-01 5.90066671e-01 -1.18957877e+00 1.46002859e-01 4.19709861e-01 7.49824047e-02 -6.04627967e-01 1.40569940e-01 6.51712120e-01 -3.49211514e-01 6.80696443e-02 -1.02928832e-01 -4.34596866e-01 -1.65456533e+00 7.92120636e-01 2.13383242e-01 -4.34771359e-01 -2.63458937e-01 6.09261155e-01 -1.22800946e-01 -5.64999282e-01 2.36444890e-01 -1.49814934e-01 -5.18038630e-01 9.52206552e-02 5.40725768e-01 6.32926702e-01 -1.10732958e-01 -7.61767626e-01 -7.31457472e-01 5.82023025e-01 -3.56659681e-01 3.22896361e-01 8.58680606e-01 1.71044320e-01 -4.81697053e-01 8.31969678e-02 1.66145480e+00 6.40159070e-01 -6.92654192e-01 2.87743300e-01 2.76877224e-01 -3.78871560e-01 -3.87338758e-01 -5.59817910e-01 -1.15123665e+00 9.62785542e-01 8.06868374e-01 6.45684242e-01 1.00451279e+00 -1.13765160e-02 1.21403015e+00 4.58053082e-01 2.46040210e-01 -4.96579021e-01 2.93139964e-01 2.49723166e-01 2.96027035e-01 -8.87241006e-01 -2.51297563e-01 -6.02726221e-01 -3.83676767e-01 7.86401570e-01 8.29000533e-01 -1.41634420e-01 9.56392527e-01 3.35933924e-01 -3.55695516e-01 -3.15911472e-01 -3.72078091e-01 1.28334299e-01 2.38088027e-01 9.54170406e-01 9.72443670e-02 1.26979530e-01 -9.54385996e-02 9.74062741e-01 -4.48628105e-02 -4.84344244e-01 3.62511069e-01 6.15955055e-01 -3.50212455e-01 -1.41678894e+00 1.00839190e-01 6.21096909e-01 -6.84901297e-01 4.40745149e-03 -9.70027387e-01 8.03448081e-01 6.52513146e-01 8.56308281e-01 1.42438905e-02 -1.29694068e+00 1.37057215e-01 -1.03766441e-01 -9.66283232e-02 -7.17238069e-01 -7.96918571e-01 -3.42075348e-01 5.38962595e-02 -4.48552012e-01 3.07493836e-01 -3.42759758e-01 -1.09925830e+00 -1.99210197e-01 5.61334100e-03 6.51809052e-02 7.93298542e-01 7.06998229e-01 9.11354065e-01 7.01583087e-01 8.37907255e-01 -5.99596739e-01 -2.65867978e-01 -7.61552513e-01 -6.33254796e-02 1.45530865e-01 1.84097871e-01 -8.14557314e-01 -4.64085460e-01 -3.65806252e-01]
[14.414848327636719, 9.674386024475098]
8b92b8f9-b5b4-4cf3-9e10-a5992e4a8d19
fuzzy-labeling-semantics-for-quantitative
2207.07339
null
https://arxiv.org/abs/2207.07339v1
https://arxiv.org/pdf/2207.07339v1.pdf
Fuzzy Labeling Semantics for Quantitative Argumentation
The topic of evaluating argument strength in various quantitative argumentation systems has received increasing attention in the field of abstract argumentation. However, the existing gradual semantics on argument strength considers acceptability degree alone, which may be not sufficient to evaluate arguments in practical scenarios. To adopt a richer characterization for argument strength in real-world applications, we provide a novel quantitative method called fuzzy labeling for fuzzy argumentation systems. For fuzzy labeling, the argument strength is represented as a triple consisting of acceptability, rejectability, and undecidability degree. With a richer scale, it sheds new light on argument strength and gives us a deeper understanding into status of arguments. For the purpose of evaluating arguments, we provide a new way to establish gradual semantics by fuzzy labeling, which is crucial in the evaluation process. We first investigate the rationality postulates of fuzzy labeling, which are important for explaining the rationality of new semantics taking into account the acceptability, rejectability and undecidability degree together. We then propose a set of fuzzy labeling semantics and prove some important properties which are crucial for comparing, understanding and applying semantics.
['Yuping Shen', 'Zongshun Wang']
2022-07-15
null
null
null
null
['abstract-argumentation', 'abstract-argumentation']
['natural-language-processing', 'reasoning']
[ 2.24017836e-02 4.21625584e-01 -2.44728476e-01 -6.18298769e-01 5.76122552e-02 -8.45558524e-01 5.94601512e-01 7.95867920e-01 -3.46947461e-01 8.12863886e-01 7.27995560e-02 -5.97252131e-01 -7.48905241e-01 -1.28000164e+00 -2.05103010e-01 -3.93096864e-01 1.80989414e-01 3.14297765e-01 3.27236056e-01 -9.93628919e-01 4.03338581e-01 2.70946234e-01 -1.76591325e+00 4.23873156e-01 1.21743619e+00 9.32884157e-01 -3.68746787e-01 2.32121453e-01 -3.28350961e-01 8.90930057e-01 -5.70000947e-01 -6.25228763e-01 -9.26035270e-02 -3.98844808e-01 -1.10615623e+00 -3.30522984e-01 -2.90416151e-01 -8.33659694e-02 7.17695951e-01 1.65528274e+00 -1.31414652e-01 -4.11223099e-02 9.21904385e-01 -1.49685180e+00 -6.23168409e-01 1.20421267e+00 -3.17768365e-01 -2.89726425e-02 6.07559264e-01 -5.31362057e-01 1.19951451e+00 -3.75678599e-01 2.75852054e-01 1.39832342e+00 3.29567194e-01 3.48215193e-01 -6.36727631e-01 -5.20843826e-02 3.44509214e-01 3.45178276e-01 -7.70338535e-01 1.69643342e-01 9.33753669e-01 -3.04885715e-01 -1.24918642e-02 5.69072127e-01 8.14141035e-01 3.74285370e-01 -1.24808513e-01 6.06921077e-01 1.39422202e+00 -9.54318166e-01 4.03123081e-01 2.70743400e-01 9.54033852e-01 4.60399389e-01 7.19547212e-01 -2.33402252e-01 6.16169795e-02 -1.42668650e-01 4.48286980e-01 -2.35838205e-01 -3.38445485e-01 -2.88042158e-01 -9.82891321e-01 1.09918714e+00 2.69193500e-01 6.69950962e-01 -1.34976923e-01 -3.59898061e-01 4.34795886e-01 4.99006033e-01 1.73863143e-01 2.26195529e-01 -3.33130747e-01 1.31605893e-01 -3.08674961e-01 3.28787118e-01 8.49555910e-01 2.02563941e-01 6.34542704e-01 -3.52827370e-01 1.27201289e-01 4.58917677e-01 6.25185788e-01 6.45265222e-01 1.02781706e-01 -1.15257800e+00 2.28996664e-01 1.04088688e+00 4.95776772e-01 -1.05289257e+00 -3.41991782e-01 -4.88545820e-02 -5.36497831e-01 6.62112594e-01 6.48383915e-01 6.93279505e-02 -1.15309693e-01 1.88777542e+00 4.23745006e-01 -2.74021655e-01 4.30983931e-01 9.40015256e-01 7.69848943e-01 5.53357303e-01 6.49196804e-02 -4.86730546e-01 1.50774133e+00 -3.53181928e-01 -9.84399676e-01 6.68010652e-01 8.01092744e-01 -4.59942579e-01 1.38860548e+00 6.39232159e-01 -1.15577793e+00 -1.72803670e-01 -1.41409540e+00 7.86228552e-02 -6.58748925e-01 -6.73864037e-02 7.70414412e-01 9.21680510e-01 -8.51987243e-01 3.66128504e-01 -3.24930608e-01 -1.80299357e-01 -2.35726908e-01 2.05112159e-01 -1.01670615e-01 3.35899591e-01 -1.82332766e+00 1.15633988e+00 5.21234870e-01 3.58091354e-01 1.19742818e-01 5.00147939e-02 -7.58635640e-01 1.39447704e-01 6.43572569e-01 -3.59354615e-01 1.15283310e+00 -9.32206690e-01 -1.48993707e+00 6.99372113e-01 2.92381465e-01 -4.36411202e-01 7.41521358e-01 1.73832268e-01 -7.21483350e-01 2.57083476e-01 2.27762356e-01 9.21470821e-02 2.77098268e-01 -1.24992633e+00 -6.11701071e-01 -3.19486707e-01 1.00805533e+00 2.34578073e-01 -3.84810209e-01 7.31289685e-02 4.15200442e-01 -3.20394278e-01 2.43974119e-01 -5.55184543e-01 8.42898116e-02 -1.52640417e-01 -5.22115380e-02 -6.41680241e-01 4.52550441e-01 1.21220075e-01 1.28254640e+00 -1.86560309e+00 -9.99537185e-02 7.35750556e-01 2.49935120e-01 5.07053912e-01 3.29503000e-01 2.47884691e-01 -8.06353092e-02 4.54510301e-01 -4.02659833e-01 8.61401558e-01 3.91530931e-01 8.28757361e-02 -4.33460563e-01 8.22476894e-02 1.27176911e-01 5.52560925e-01 -1.06433153e+00 -5.51679552e-01 3.77603203e-01 2.53169090e-01 -4.18628365e-01 -3.34057301e-01 -1.59757882e-01 -1.80215463e-01 -9.96880054e-01 4.21243340e-01 8.65368605e-01 -1.05180321e-02 4.69783694e-01 -3.99445444e-01 -3.29972506e-01 1.47726722e-02 -1.57493079e+00 9.25273299e-01 -1.38564959e-01 1.58036649e-01 1.06785662e-01 -1.23447800e+00 9.02589142e-01 3.22697401e-01 2.19081402e-01 -6.80477142e-01 7.09285378e-01 4.77900982e-01 -5.90389967e-02 -4.33480889e-01 5.13531685e-01 -4.47660178e-01 -2.89812207e-01 8.90012324e-01 -4.72595632e-01 -3.34425688e-01 7.49681532e-01 1.86752394e-01 3.13451171e-01 -7.26017579e-02 5.40158927e-01 -7.93935776e-01 1.15145552e+00 -1.13654599e-01 3.24545979e-01 3.47496599e-01 -2.51516998e-01 -5.85081391e-02 9.21564996e-01 -7.63308406e-01 -5.71140826e-01 -1.13476455e+00 -3.57831120e-01 7.49583662e-01 6.67794883e-01 -2.85730809e-01 -1.04678750e+00 -6.65819287e-01 -1.20535888e-01 6.05127811e-01 -7.94748783e-01 -1.53906643e-01 -4.27374691e-01 -7.79261947e-01 5.04736841e-01 2.66820341e-01 5.69100320e-01 -9.63656962e-01 -8.31380308e-01 2.31937766e-02 -5.63983738e-01 -6.50337875e-01 3.60151708e-01 4.62988652e-02 -8.73912871e-01 -1.55866027e+00 -4.43447560e-01 -4.29538876e-01 5.66978812e-01 2.09437713e-01 9.86024141e-01 6.78167105e-01 4.42900211e-01 1.16652390e-03 -7.41699696e-01 -7.08152294e-01 -6.06858492e-01 -3.73837382e-01 8.83568972e-02 -2.66720176e-01 1.89581037e-01 -5.53690568e-02 -2.73333788e-01 3.62106115e-01 -1.25020885e+00 -2.78694808e-01 4.68988009e-02 5.30844688e-01 1.84822500e-01 3.46622109e-01 8.32615376e-01 -1.01424050e+00 1.30610824e+00 -1.98082954e-01 -5.90768695e-01 7.85647929e-01 -6.79034650e-01 4.31548446e-01 5.25134802e-01 -4.27783281e-02 -1.23725569e+00 -9.24830377e-01 -1.57640696e-01 5.95686793e-01 1.83213577e-01 7.55214870e-01 -4.39737618e-01 2.64896657e-02 6.67467415e-01 -5.46881318e-01 -3.43291201e-02 4.68343645e-02 6.67777419e-01 5.33738315e-01 2.23319620e-01 -1.18528998e+00 2.59452313e-01 6.38883531e-01 3.45838815e-01 -3.89334410e-01 -9.83941436e-01 -6.50623813e-02 -5.04492342e-01 -2.08475634e-01 4.99555618e-01 -3.63206148e-01 -1.36449754e+00 3.78133133e-02 -1.11499488e+00 2.52604514e-01 -4.48916525e-01 4.67488378e-01 -6.48550451e-01 8.00729752e-01 -5.62028050e-01 -1.07078111e+00 -1.71608046e-01 -1.17358291e+00 5.08792639e-01 1.47089779e-01 -3.15329194e-01 -1.23224688e+00 -1.00229517e-01 2.10503027e-01 6.41881824e-02 3.22915226e-01 1.29218543e+00 -4.96300280e-01 3.40159655e-01 8.02754238e-03 -2.85081714e-01 4.55165923e-01 3.61901522e-02 3.91205192e-01 -5.31033337e-01 3.02428216e-01 4.65772927e-01 -1.89816117e-01 6.01699114e-01 3.01487923e-01 4.92247701e-01 -1.09456666e-01 1.25595987e-01 -1.62604883e-01 1.27806675e+00 2.44695574e-01 6.91319406e-01 7.04815567e-01 1.09947687e-02 1.13731802e+00 1.19204307e+00 4.48255628e-01 4.76533383e-01 4.61687654e-01 5.23899317e-01 -5.31809367e-02 3.11676443e-01 2.67908007e-01 -8.81533488e-05 7.72217214e-01 -6.26469016e-01 -2.74290919e-01 -7.80245781e-01 1.10891499e-01 -1.98241580e+00 -1.15026414e+00 -5.24970055e-01 2.18400812e+00 8.42348695e-01 4.22652274e-01 2.30606645e-01 1.04590893e+00 1.15792370e+00 -1.48907393e-01 2.83892393e-01 -9.20003593e-01 -4.77327734e-01 1.18521117e-01 -1.53605655e-01 9.35767829e-01 -8.25082660e-01 6.36512220e-01 6.23688745e+00 6.87656403e-01 -9.83532012e-01 -1.93745464e-01 2.66506553e-01 5.45490026e-01 -8.94447267e-01 3.52957308e-01 -3.48849386e-01 2.96465367e-01 5.07533908e-01 -3.98207694e-01 4.83509377e-02 5.46245456e-01 6.35839552e-02 -2.80827492e-01 -7.74941325e-01 4.77214098e-01 -2.43897334e-01 -1.05539978e+00 3.27329159e-01 -2.01511055e-01 4.42389131e-01 -8.84016752e-01 -7.95276687e-02 -1.11153595e-01 2.63049453e-01 -7.05257714e-01 1.06805921e+00 4.46743935e-01 3.88619423e-01 -9.26135361e-01 1.26113164e+00 2.90273935e-01 -1.02891219e+00 3.89501825e-02 -3.16879272e-01 -7.05440104e-01 3.54089916e-01 8.38350654e-01 -2.63276547e-01 9.55715001e-01 2.36464635e-01 3.04576904e-01 -1.78321704e-01 4.77968305e-01 -5.61054230e-01 1.81255952e-01 -4.55912024e-01 -2.89061278e-01 2.22016007e-01 -4.73915458e-01 1.40708312e-01 6.63881898e-01 2.43242487e-01 4.38194722e-01 -2.08228678e-01 8.74352932e-01 1.76868424e-01 5.25449693e-01 -2.96784461e-01 1.11945681e-01 4.29467350e-01 1.01998806e+00 -1.24941730e+00 -5.02978563e-01 -1.80185914e-01 1.44648165e-01 9.54668224e-02 -1.28903780e-02 -9.10335302e-01 -4.49471444e-01 -5.22701144e-02 -2.57314686e-02 -5.46267271e-01 6.27911091e-02 -4.81412619e-01 -1.26111209e+00 1.83173642e-01 -7.39593089e-01 5.69224954e-01 -5.09591937e-01 -9.91611719e-01 6.17823303e-01 3.94896120e-01 -1.27035344e+00 -2.24576622e-01 -8.88181210e-01 -7.76797831e-01 7.02713311e-01 -1.60912979e+00 -9.36535299e-01 -1.37812316e-01 5.51202476e-01 -9.25443992e-02 3.42152655e-01 8.50110710e-01 -6.36930168e-02 -1.64725736e-01 1.27428934e-01 -3.65846992e-01 -2.99697161e-01 5.15856385e-01 -1.20876050e+00 -2.74119526e-01 5.02449751e-01 -1.76268354e-01 8.21120977e-01 9.06727612e-01 -4.55361515e-01 -7.29391515e-01 -1.24883436e-01 1.09253609e+00 -3.30687672e-01 7.20042765e-01 3.27001125e-01 -8.57852340e-01 7.15713054e-02 9.93489847e-02 -2.54397005e-01 7.37062931e-01 2.99961507e-01 -3.61339211e-01 -8.01531374e-02 -1.36155796e+00 6.63816154e-01 4.84406650e-01 -3.12963963e-01 -1.23532712e+00 -1.48252457e-01 4.01327461e-01 2.18572870e-01 -8.41457963e-01 8.32270741e-01 8.26127708e-01 -1.34417248e+00 8.50659251e-01 -3.94681871e-01 1.41458943e-01 -7.63568580e-01 -2.75573343e-01 -7.74599791e-01 -1.73382774e-01 -1.12732626e-01 3.30908567e-01 1.07733452e+00 4.84319448e-01 -8.25111330e-01 4.07803893e-01 4.14306521e-01 1.53973117e-01 -6.32405043e-01 -5.08910894e-01 -4.99415904e-01 4.34295058e-01 -4.22780275e-01 9.73427773e-01 1.21304977e+00 8.31933796e-01 1.27684444e-01 1.25895828e-01 -1.54541448e-01 4.62270856e-01 3.61644566e-01 3.71164978e-02 -1.96306205e+00 2.45962888e-01 -8.88599038e-01 -3.26860994e-01 -8.37178528e-02 2.51747936e-01 -6.47626162e-01 -3.54333609e-01 -1.67651975e+00 -1.78150579e-01 -5.22630632e-01 -6.42847598e-01 3.00858587e-01 -1.31223291e-01 1.85388938e-01 2.09103420e-01 2.35554457e-01 -5.77266574e-01 2.65807778e-01 1.24777758e+00 -6.86592385e-02 -2.06397288e-02 -5.63168079e-02 -9.39992309e-01 1.36293685e+00 8.31362307e-01 -7.08821490e-02 -7.87253439e-01 -1.74302980e-01 1.11938429e+00 5.15741706e-02 8.42573792e-02 -6.28094912e-01 1.05632488e-02 -6.44600570e-01 -2.25128487e-01 -2.58959591e-01 -1.00681372e-01 -6.96129560e-01 -1.03752524e-01 5.75726390e-01 -5.13849854e-01 4.53726873e-02 -3.08378190e-01 1.21038869e-01 -4.99668747e-01 -8.42093885e-01 5.72369933e-01 -2.80449670e-02 -5.66095293e-01 -3.82704973e-01 -1.85491323e-01 1.21946134e-01 1.01632833e+00 -3.87613982e-01 -4.37328011e-01 -1.86085910e-01 -9.37475979e-01 1.99936226e-01 6.53399467e-01 2.39291508e-03 3.86513323e-01 -1.29945147e+00 -6.68325722e-01 -1.05444916e-01 1.60854742e-01 -3.28948259e-01 1.24748878e-01 7.07707703e-01 -8.88511002e-01 2.26981819e-01 -5.33055604e-01 -1.89340726e-01 -1.09681416e+00 5.59672117e-01 -3.26574259e-02 -2.15147614e-01 1.97878201e-02 3.73000950e-01 2.43875325e-01 -2.81144738e-01 -7.36465454e-02 -6.50493026e-01 -8.91937494e-01 3.37620676e-01 4.36425507e-01 7.02055156e-01 6.84444904e-02 -7.32689202e-01 -4.48461950e-01 1.01215589e+00 2.62525827e-01 -4.18338299e-01 7.53015995e-01 -2.34706283e-01 -5.40575087e-01 5.46030641e-01 2.21660227e-01 3.19844186e-01 -5.63524246e-01 2.30595917e-01 1.94489866e-01 -7.02827349e-02 -1.87013790e-01 -7.70509660e-01 -4.25922900e-01 1.04937673e+00 1.62881494e-01 1.05101633e+00 1.10499907e+00 -8.79349485e-02 3.51323068e-01 5.77387869e-01 5.40934205e-01 -1.33332288e+00 -1.93229601e-01 5.15053034e-01 7.38065004e-01 -9.76394117e-01 3.55589436e-03 -9.99204218e-01 -6.39361143e-01 1.67087233e+00 6.24089390e-02 -2.72825342e-02 6.93819523e-01 2.69171774e-01 2.27128521e-01 -2.08812907e-01 -2.63861656e-01 -2.27829888e-01 2.92775840e-01 3.76906782e-01 7.22669482e-01 4.71230775e-01 -1.32302332e+00 1.00107956e+00 -3.16331148e-01 7.47047439e-02 8.55378032e-01 9.44332302e-01 -9.83664095e-01 -1.43827295e+00 -4.79920387e-01 -1.34228155e-01 -6.15420580e-01 8.30570757e-02 -4.72714961e-01 7.67324507e-01 3.25801104e-01 1.36582804e+00 -4.45588648e-01 -7.31004179e-02 3.69622648e-01 -3.92387420e-01 6.94172502e-01 -3.28029305e-01 -3.93557459e-01 -2.76319414e-01 2.77571291e-01 7.72896782e-02 -1.06977749e+00 -1.97299436e-01 -1.69410443e+00 -5.33121943e-01 -7.76321411e-01 1.12310851e+00 6.70479059e-01 1.25840509e+00 -4.67762798e-01 2.97745168e-01 4.39534068e-01 -1.26621187e-01 -6.73248291e-01 -6.62501037e-01 -7.95461416e-01 5.16724825e-01 -1.58312377e-02 -9.96451974e-01 -5.06640673e-01 -1.88717157e-01]
[8.8605318069458, 6.837864398956299]
14e1e875-16db-4b2f-bb49-6481f872367d
a-topological-nomenclature-for-3d-shape
1909.12887
null
https://arxiv.org/abs/1909.12887v2
https://arxiv.org/pdf/1909.12887v2.pdf
A Topological Nomenclature for 3D Shape Analysis in Connectomics
One of the essential tasks in connectomics is the morphology analysis of neurons and organelles like mitochondria to shed light on their biological properties. However, these biological objects often have tangled parts or complex branching patterns, which make it hard to abstract, categorize, and manipulate their morphology. In this paper, we develop a novel topological nomenclature system to name these objects like the appellation for chemical compounds to promote neuroscience analysis based on their skeletal structures. We first convert the volumetric representation into the topology-preserving reduced graph to untangle the objects. Next, we develop nomenclature rules for pyramidal neurons and mitochondria from the reduced graph and finally learn the feature embedding for shape manipulation. In ablation studies, we quantitatively show that graphs generated by our proposed method align with the perception of experts. On 3D shape retrieval and decomposition tasks, we qualitatively demonstrate that the encoded topological nomenclature features achieve better results than state-of-the-art shape descriptors. To advance neuroscience, we will release a 3D segmentation dataset of mitochondria and pyramidal neurons reconstructed from a 100um cube electron microscopy volume with their reduced graph and topological nomenclature annotations. Code is publicly available at https://github.com/donglaiw/ibexHelper.
['Won-Dong Jang', 'Abhimanyu Talwar', 'Zudi Lin', 'Xueying Wang', 'Jinglin Zhao', 'Jeff W. Lichtman', 'Donglai Wei', 'Bowen Zheng', 'Yuesong Wu', 'Hanspeter Pfister']
2019-09-27
null
null
null
null
['3d-shape-retrieval']
['computer-vision']
[-1.59347832e-01 4.51505631e-02 5.38402610e-02 -2.10817814e-01 -2.06480265e-01 -1.08762634e+00 3.33188295e-01 3.36123407e-01 -3.75363380e-01 6.95326567e-01 -7.56463856e-02 -1.88941374e-01 -1.58320576e-01 -6.17644012e-01 -6.47104263e-01 -8.53469670e-01 -6.48775920e-02 6.99745953e-01 1.06525280e-01 1.11520357e-01 5.05549431e-01 1.00375605e+00 -9.35974061e-01 1.61693379e-01 8.49501431e-01 9.28891361e-01 4.22024101e-01 2.99955845e-01 -2.15309516e-01 -1.94014326e-01 -3.67168605e-01 -4.00543064e-01 2.06186742e-01 -4.38704789e-02 -7.07592010e-01 -1.05585150e-01 4.32931006e-01 -1.59616433e-02 -3.47515613e-01 1.29773343e+00 4.31108296e-01 -3.29377711e-01 1.12947929e+00 -1.33444202e+00 -1.08962023e+00 3.65101010e-01 -2.96291500e-01 3.96482706e-01 1.02156959e-01 8.34621787e-02 1.06711364e+00 -8.63057375e-01 1.13180792e+00 1.23393798e+00 3.93523514e-01 5.09201169e-01 -1.70211089e+00 -7.36203969e-01 1.09696396e-01 1.48586094e-01 -1.22608709e+00 -3.88743311e-01 7.37417102e-01 -8.93722355e-01 7.25419521e-01 4.24131043e-02 1.06615853e+00 1.11899340e+00 4.98662412e-01 4.58905667e-01 1.09286940e+00 1.70667946e-01 3.07680100e-01 -3.79677027e-01 1.14854433e-01 9.96684670e-01 7.57713556e-01 -4.29443419e-01 -1.21784091e-01 -1.61597654e-02 1.08996534e+00 1.71605393e-01 -4.04033393e-01 -8.04062307e-01 -1.67066574e+00 4.87000883e-01 5.74059725e-01 2.99780160e-01 -1.41243413e-02 1.41971022e-01 3.24428737e-01 -3.05201802e-02 8.94522592e-02 6.60117209e-01 -3.39039356e-01 1.92590863e-01 -5.59649646e-01 2.02836260e-01 4.84477460e-01 9.26538408e-01 6.48784816e-01 -2.95874123e-02 1.38790727e-01 6.84616566e-01 3.86686534e-01 4.93525654e-01 3.07062060e-01 -1.23560631e+00 2.69526511e-01 9.77137208e-01 -2.94077694e-01 -1.14240921e+00 -7.75873899e-01 -2.13208064e-01 -1.10065746e+00 2.86455899e-01 6.69822156e-01 4.60439384e-01 -8.10152650e-01 1.67486393e+00 2.26308480e-01 -2.54489094e-01 -3.18315148e-01 8.18489194e-01 9.72585440e-01 7.44943693e-02 -1.15292504e-01 5.42913638e-02 1.49517083e+00 -4.09480065e-01 -4.95149195e-01 2.56192893e-01 8.07321727e-01 -1.79060921e-01 9.20045853e-01 2.19152108e-01 -9.54808176e-01 -3.96198221e-02 -1.01727092e+00 -3.84341180e-01 -7.03134894e-01 3.03991795e-01 6.18830800e-01 1.85921624e-01 -1.12090731e+00 8.45541298e-01 -1.04128027e+00 -5.98429918e-01 1.09871483e+00 3.91338646e-01 -7.89537549e-01 3.74214262e-01 -5.93202651e-01 8.16165924e-01 2.19041437e-01 -1.37167469e-01 -7.03633010e-01 -7.90288985e-01 -7.28317797e-01 -3.22540104e-02 -2.40747333e-01 -9.46902394e-01 5.54320753e-01 8.83055180e-02 -1.10672879e+00 1.30236125e+00 1.13987692e-01 -1.68438211e-01 1.15717903e-01 3.95957828e-01 1.20853625e-01 5.71205556e-01 2.38808766e-01 9.24278021e-01 5.13725817e-01 -1.16680706e+00 8.75009671e-02 -7.95668662e-01 2.62973811e-02 -7.12593496e-02 -2.79649675e-01 -3.89175773e-01 9.62800812e-03 -6.10952377e-01 6.11641228e-01 -8.34737182e-01 2.14439519e-02 8.08250666e-01 -5.21460056e-01 -1.64730437e-02 6.22217417e-01 -4.70332325e-01 7.30184257e-01 -2.05123949e+00 6.80104494e-01 5.18359691e-02 9.53838646e-01 -7.40136504e-02 -2.42284432e-01 3.33156556e-01 -2.07981959e-01 6.60488486e-01 -4.09443647e-01 -9.45743620e-02 7.23563880e-03 1.44754857e-01 -1.25258386e-01 8.63678694e-01 1.64912045e-01 1.14007747e+00 -1.02815390e+00 -6.08105242e-01 -1.67725459e-02 4.56096113e-01 -5.37707388e-01 -3.58341545e-01 3.72397974e-02 7.35736370e-01 -7.12017477e-01 8.96975994e-01 5.80369592e-01 -4.55816120e-01 1.80346489e-01 -7.39942431e-01 -8.70998278e-02 7.84638748e-02 -6.61724508e-01 1.93734264e+00 4.59182709e-02 6.27775848e-01 3.48983258e-01 -1.23073971e+00 6.81760550e-01 -1.08897472e-02 6.19692922e-01 -3.18270385e-01 3.76741230e-01 1.35142922e-01 2.44364351e-01 -1.89645439e-01 -3.63581747e-01 9.42313001e-02 3.74825522e-02 3.81652266e-01 2.38915443e-01 -5.61611295e-01 3.50892723e-01 3.64431560e-01 1.15748966e+00 1.00417748e-01 2.51064152e-01 -5.47227323e-01 2.33178362e-01 -2.02496663e-01 5.60331583e-01 3.97985205e-02 -5.05649567e-01 5.93687713e-01 8.30640912e-01 -3.95436913e-01 -1.28050923e+00 -1.52481043e+00 -5.43126047e-01 4.63685066e-01 1.23638302e-01 -3.43790084e-01 -7.44504213e-01 -2.94788182e-01 1.62264585e-01 2.80855428e-02 -7.29993701e-01 -1.72074571e-01 -2.26148292e-01 -5.80413342e-01 7.64428496e-01 3.65339398e-01 2.09560275e-01 -9.08418894e-01 -7.26026714e-01 -1.65433869e-01 -1.83084220e-01 -1.12930167e+00 -4.16197598e-01 1.86873168e-01 -1.04102004e+00 -1.18676114e+00 -5.62765539e-01 -8.06500375e-01 1.11143315e+00 8.93939063e-02 6.87482595e-01 2.40156263e-01 -6.78970635e-01 3.10823023e-01 -3.31148863e-01 -2.43193984e-01 1.05411768e-01 -7.10765719e-02 4.36202079e-01 -3.69199514e-01 -3.00668210e-01 -1.14653802e+00 -6.98069632e-01 1.92373961e-01 -8.62715602e-01 2.12941319e-01 5.04270673e-01 6.57302737e-01 1.06822836e+00 -3.51916671e-01 2.45755464e-01 -5.05518019e-01 4.67548579e-01 -1.60041064e-01 -5.44629335e-01 1.00168422e-01 -3.03540945e-01 2.86715001e-01 6.24980032e-01 -3.37061346e-01 -1.10296853e-01 5.09916805e-02 1.69779032e-01 -4.59509790e-01 -9.78959128e-02 2.33848512e-01 -4.55566227e-01 -3.56740534e-01 4.38895822e-01 3.11823994e-01 1.79362133e-01 -4.17269737e-01 3.77072722e-01 2.08932474e-01 4.40989286e-01 -8.05235922e-01 9.02786195e-01 8.44994187e-01 4.55738455e-01 -8.51929665e-01 -5.55736840e-01 -1.83367088e-01 -1.12234950e+00 -4.97142226e-02 1.10698569e+00 -4.24952447e-01 -9.77970779e-01 2.76202083e-01 -1.20627487e+00 -4.07527983e-01 -9.56486166e-02 3.25085908e-01 -8.80416095e-01 6.96797132e-01 -7.17349052e-01 -2.47972324e-01 -3.17668319e-01 -1.13976455e+00 1.18202138e+00 2.58642975e-02 -2.62264729e-01 -9.35787916e-01 6.49937615e-02 1.61380515e-01 9.50829387e-02 4.38928157e-01 1.53956020e+00 -5.58727443e-01 -6.35436118e-01 1.98122844e-01 -4.86212432e-01 -1.97835550e-01 9.74981561e-02 2.88395733e-01 -5.67378044e-01 -2.02884421e-01 -2.99438417e-01 -2.70572633e-01 8.33209693e-01 3.74230623e-01 1.41819870e+00 -1.07721381e-01 -5.15453875e-01 1.03845584e+00 1.08377457e+00 -8.84700660e-03 4.50653821e-01 2.37008438e-01 8.00255895e-01 6.88214779e-01 -3.05780862e-02 2.40557000e-01 2.80808061e-01 4.39815491e-01 7.01083422e-01 1.81516632e-01 -2.25672707e-01 3.88768017e-02 4.03509438e-02 1.04791510e+00 -2.69809097e-01 7.48182535e-02 -1.01704633e+00 3.65163296e-01 -1.54174042e+00 -8.88166726e-01 -1.24317370e-01 1.88137400e+00 8.82967591e-01 -1.05776312e-02 1.21371518e-03 -5.04148751e-02 6.92220092e-01 -8.99177641e-02 -7.07034290e-01 -4.08563763e-02 -2.37862185e-01 7.32520893e-02 2.79986292e-01 2.61011243e-01 -8.37733448e-01 9.22957301e-01 5.99680185e+00 5.14086545e-01 -1.02112770e+00 -3.56975347e-02 4.56708461e-01 3.77311334e-02 -4.50405687e-01 -5.94316795e-02 -4.90388125e-01 4.03987080e-01 4.75603670e-01 -2.21375406e-01 7.29718029e-01 4.05223906e-01 1.34187415e-01 2.12103665e-01 -1.35978544e+00 1.26733041e+00 -2.16936678e-01 -1.49361455e+00 4.67000425e-01 3.86634707e-01 3.20348501e-01 9.97934863e-02 2.03804061e-01 -2.10141197e-01 -1.99396331e-02 -1.11753428e+00 7.93874264e-01 8.33127141e-01 9.55561221e-01 -2.31746376e-01 2.02425331e-01 1.70821160e-01 -1.32474089e+00 1.34245560e-01 -7.87365973e-01 1.49401590e-01 -5.09767793e-03 6.42080247e-01 -5.39473951e-01 4.19996113e-01 6.33135736e-01 8.97108495e-01 -7.06843317e-01 1.30281687e+00 4.76762578e-02 1.07460238e-01 -2.14401066e-01 2.82065608e-02 -2.76936859e-01 -7.09864497e-01 7.00797915e-01 9.30461168e-01 3.88824910e-01 2.74303406e-01 -1.04114749e-02 1.57926321e+00 -3.14530164e-01 8.89182240e-02 -8.32607985e-01 -6.90182149e-01 5.17034471e-01 1.67864299e+00 -1.44479382e+00 -1.68270975e-01 -9.72539708e-02 7.19489038e-01 6.69314861e-01 4.61092681e-01 -7.09686399e-01 -4.32386130e-01 8.73034418e-01 2.01574475e-01 1.41310468e-01 -6.80808902e-01 -6.29119396e-01 -1.25488698e+00 7.73737021e-03 -2.71261454e-01 -1.75065055e-01 -1.01707625e+00 -1.36573815e+00 3.59992653e-01 -7.20391348e-02 -1.02690434e+00 6.40231371e-01 -1.11346364e+00 -5.42473197e-01 5.47680795e-01 -8.74350965e-01 -9.92057621e-01 -1.35387987e-01 1.87898576e-01 1.17767319e-01 -2.11694147e-02 8.74316871e-01 8.43177810e-02 -6.17955625e-01 1.97942764e-01 1.72434852e-01 3.44637364e-01 4.35862243e-01 -1.28805554e+00 2.03383967e-01 2.66378015e-01 2.98765689e-01 9.09885466e-01 4.26945120e-01 -6.81847930e-01 -1.44894218e+00 -1.10123777e+00 4.23596144e-01 -7.34484553e-01 9.54130709e-01 -7.32549429e-01 -9.17943716e-01 6.07740641e-01 -2.38503322e-01 3.11405629e-01 5.90253592e-01 -2.61245102e-01 -5.31571627e-01 8.49928707e-02 -1.21291590e+00 8.30000341e-01 1.56463706e+00 -5.37796974e-01 -6.58120632e-01 5.38395166e-01 5.86983740e-01 2.15218645e-02 -1.19566107e+00 3.50947440e-01 7.46655285e-01 -7.02439725e-01 1.10633254e+00 -6.47352934e-01 2.36044794e-01 -4.45028275e-01 -1.68138847e-01 -1.36718154e+00 -7.02677310e-01 -6.62021413e-02 1.33543191e-02 8.61245155e-01 4.28511590e-01 -7.30476081e-01 6.98430002e-01 1.99966431e-01 -3.99717510e-01 -9.95509505e-01 -1.09563398e+00 -8.99379253e-01 5.12583494e-01 6.43636584e-02 2.90534675e-01 9.49992239e-01 2.27077246e-01 2.58420199e-01 6.63583219e-01 -9.58549976e-02 7.76570916e-01 2.04950422e-01 2.91276693e-01 -1.71683753e+00 2.67729789e-01 -9.40522373e-01 -9.75758255e-01 -6.37617111e-01 4.12225813e-01 -1.50211799e+00 -4.28185403e-01 -1.80153716e+00 5.01868248e-01 -2.48491973e-01 -2.43434086e-01 6.70526922e-01 3.87862474e-01 3.50003421e-01 1.07013255e-01 2.59534061e-01 -6.59082532e-01 7.44870067e-01 1.78130162e+00 -4.33419019e-01 1.40362784e-01 -6.23194516e-01 -6.42407894e-01 7.26131082e-01 9.47097838e-01 -3.88353854e-01 -2.52902776e-01 -4.64321345e-01 3.76612127e-01 -2.99172103e-01 6.97814286e-01 -9.33205605e-01 7.42454529e-02 -1.44898042e-01 6.01120174e-01 -5.53515077e-01 4.43904668e-01 -7.37888515e-01 1.11050278e-01 5.15551031e-01 -1.14281677e-01 1.48017049e-01 1.35291770e-01 4.55997676e-01 3.07383060e-01 6.20791763e-02 8.11331809e-01 -2.23966479e-01 1.02697173e-02 6.31345809e-01 -5.34025609e-01 1.80482730e-01 9.32084858e-01 -4.37162519e-01 -6.85094535e-01 6.75836764e-03 -1.02371323e+00 8.19161609e-02 1.06149733e+00 -3.78051922e-02 8.59573543e-01 -1.57284880e+00 -3.73616636e-01 1.91853009e-02 1.80338994e-01 -3.09237614e-02 1.09913632e-01 1.27230227e+00 -7.28472054e-01 4.65760320e-01 -8.87733102e-01 -6.83334947e-01 -1.09789932e+00 6.40294135e-01 1.85829103e-01 5.84983192e-02 -6.12376213e-01 6.66191697e-01 7.24529028e-01 -6.39916897e-01 -4.27728705e-02 -9.72422004e-01 -2.57849485e-01 6.24301061e-02 1.25769198e-01 3.45258027e-01 -1.57890230e-01 -6.83798969e-01 -4.93292242e-01 1.00941980e+00 1.59028664e-01 1.65253088e-01 1.51653755e+00 -1.92640707e-01 -6.19160652e-01 5.94686627e-01 1.22647893e+00 -1.01509899e-01 -1.02160907e+00 2.17917472e-01 -1.52556285e-01 -1.91294655e-01 -3.27933788e-01 -5.25165915e-01 -1.10948789e+00 1.13500190e+00 3.42835128e-01 1.50230095e-01 7.77367711e-01 3.70102823e-01 6.34523571e-01 6.14586115e-01 6.10197127e-01 -6.52662575e-01 1.39501229e-01 5.07593513e-01 1.30063939e+00 -7.61392295e-01 8.78771767e-02 -6.02408826e-01 -1.33642063e-01 1.11562061e+00 3.62219959e-01 -2.41086021e-01 7.00874984e-01 1.46926492e-01 -2.79489428e-01 -5.88085592e-01 -6.14916205e-01 -1.31253332e-01 3.78113568e-01 9.54982400e-01 2.29203284e-01 7.65543059e-02 -2.62347996e-01 8.67756844e-01 -3.05018812e-01 -5.13929904e-01 5.75025201e-01 6.11143827e-01 -6.09107852e-01 -1.07146728e+00 -2.09536970e-01 8.18721712e-01 -3.50712031e-01 -1.40915170e-01 -8.29370022e-01 5.72223186e-01 6.14391863e-02 2.76298672e-01 -3.39019252e-03 -3.57996553e-01 1.63791925e-01 9.74631906e-02 9.95875955e-01 -8.76592398e-01 1.64464444e-01 -1.13684438e-01 -2.70865589e-01 -6.59675956e-01 -2.54474163e-01 -6.73536420e-01 -1.78699374e+00 -1.33274615e-01 3.30843069e-02 -1.19088516e-01 5.87637603e-01 6.50035143e-01 5.21275222e-01 3.76642406e-01 9.60701257e-02 -1.23375154e+00 -8.07027444e-02 -7.07488120e-01 -8.35788012e-01 4.65420872e-01 8.15380439e-02 -1.27005041e+00 -3.07395250e-01 2.69528300e-01]
[14.275208473205566, -3.13678240776062]
785fe652-9d5b-4c94-9d57-09726b020d2a
text-is-not-enough-integrating-visual
2109.05778
null
https://arxiv.org/abs/2109.05778v2
https://arxiv.org/pdf/2109.05778v2.pdf
Text is NOT Enough: Integrating Visual Impressions into Open-domain Dialogue Generation
Open-domain dialogue generation in natural language processing (NLP) is by default a pure-language task, which aims to satisfy human need for daily communication on open-ended topics by producing related and informative responses. In this paper, we point out that hidden images, named as visual impressions (VIs), can be explored from the text-only data to enhance dialogue understanding and help generate better responses. Besides, the semantic dependency between an dialogue post and its response is complicated, e.g., few word alignments and some topic transitions. Therefore, the visual impressions of them are not shared, and it is more reasonable to integrate the response visual impressions (RVIs) into the decoder, rather than the post visual impressions (PVIs). However, both the response and its RVIs are not given directly in the test process. To handle the above issues, we propose a framework to explicitly construct VIs based on pure-language dialogue datasets and utilize them for better dialogue understanding and generation. Specifically, we obtain a group of images (PVIs) for each post based on a pre-trained word-image mapping model. These PVIs are used in a co-attention encoder to get a post representation with both visual and textual information. Since the RVIs are not provided directly during testing, we design a cascade decoder that consists of two sub-decoders. The first sub-decoder predicts the content words in response, and applies the word-image mapping model to get those RVIs. Then, the second sub-decoder generates the response based on the post and RVIs. Experimental results on two open-domain dialogue datasets show that our proposed approach achieves superior performance over competitive baselines.
['Xiaofang Zhao', 'Yonghao Song', 'Xin Shen', 'Haolan Zhan', 'Lei Shen']
2021-09-13
null
null
null
null
['dialogue-understanding']
['natural-language-processing']
[ 3.89942229e-01 4.98242319e-01 1.38855159e-01 -5.94963849e-01 -8.86344135e-01 -5.13090134e-01 8.38873982e-01 -1.64870635e-01 -3.37486237e-01 7.22965479e-01 5.79685688e-01 -2.11040124e-01 7.29548693e-01 -8.49565744e-01 -5.71105897e-01 -4.38421398e-01 6.37463331e-01 6.46372736e-01 1.88243270e-01 -4.48538929e-01 1.39472961e-01 -2.73311824e-01 -9.78658497e-01 8.35197330e-01 1.01760435e+00 9.43119824e-01 7.14729071e-01 5.90287507e-01 -4.90034640e-01 7.51409292e-01 -9.29562271e-01 -6.84825480e-01 -5.20822816e-02 -9.83228207e-01 -9.44325626e-01 4.48093206e-01 -1.00666717e-01 -8.44582617e-01 -2.09456742e-01 9.33705389e-01 5.17439306e-01 2.91215390e-01 1.00317860e+00 -1.01252460e+00 -1.03902483e+00 5.10015547e-01 -6.62044883e-01 -2.10866064e-01 6.19572997e-01 5.03500819e-01 1.03968370e+00 -9.98856544e-01 7.22576618e-01 1.73176110e+00 -1.43426597e-01 8.17688882e-01 -9.09773469e-01 -4.96942073e-01 3.27907979e-01 1.87247530e-01 -1.00449955e+00 -2.06739947e-01 7.46532261e-01 -2.16614783e-01 6.61547005e-01 1.39466330e-01 5.16729474e-01 1.41638672e+00 -5.69244660e-02 1.26691067e+00 1.12492585e+00 -3.25051248e-01 -7.10597038e-02 5.72796583e-01 -1.78962052e-01 5.76827466e-01 -5.10110438e-01 -3.27321589e-01 -3.30060631e-01 1.85419425e-01 8.24501097e-01 -1.14022247e-01 -5.34154832e-01 5.81718087e-02 -1.19629264e+00 1.15296960e+00 5.53034365e-01 2.86941528e-02 -3.75262827e-01 -4.92567748e-01 3.00344229e-01 3.20760638e-01 4.89397436e-01 2.57192671e-01 -1.85475439e-01 -1.81107465e-02 -4.71055925e-01 1.47398710e-01 8.58149529e-01 1.05560541e+00 9.94689941e-01 -2.96209484e-01 -6.92818522e-01 1.23177266e+00 5.35062373e-01 4.79860693e-01 5.99015236e-01 -3.97758961e-01 9.27957535e-01 7.00855911e-01 6.25680983e-02 -9.64480340e-01 3.37910932e-03 1.99367777e-02 -9.12810385e-01 -1.48376405e-01 3.46148491e-01 -4.69975114e-01 -8.69593620e-01 1.68598473e+00 3.61089796e-01 -3.81388575e-01 4.77608204e-01 1.23337042e+00 1.29794991e+00 1.34525442e+00 -2.11089831e-02 -3.34023386e-01 1.56413627e+00 -1.39815855e+00 -8.61678660e-01 -5.21197796e-01 4.28009480e-01 -9.94557738e-01 1.43257332e+00 1.38622299e-02 -1.07213533e+00 -7.38822281e-01 -9.35100555e-01 -4.44470495e-01 -5.55679463e-02 2.81851441e-01 -7.03775287e-02 -1.10776208e-01 -8.74154508e-01 3.40133309e-02 -1.29242033e-01 -1.56532988e-01 9.71374810e-02 -4.53303717e-02 -7.43559301e-02 -1.57925203e-01 -1.53919971e+00 9.46229279e-01 1.90012723e-01 1.10889040e-01 -9.95615244e-01 -1.15583390e-01 -1.12755990e+00 -1.47913545e-01 4.66841251e-01 -6.58692241e-01 1.59047353e+00 -1.26772451e+00 -1.82818604e+00 8.47671866e-01 -3.30605507e-01 -5.88153340e-02 7.55833089e-01 -4.21084464e-02 -2.28689030e-01 3.21266055e-01 2.54875004e-01 1.10001898e+00 8.43534946e-01 -1.46364617e+00 -5.37661493e-01 -1.50206983e-01 4.10013288e-01 8.69295776e-01 -3.38951200e-01 -7.95841292e-02 -9.46762502e-01 -4.57632154e-01 -2.11037770e-01 -7.62759507e-01 -2.09670186e-01 -1.27073020e-01 -6.53329015e-01 -3.56441021e-01 6.26757324e-01 -9.32437718e-01 7.81615853e-01 -2.10293698e+00 2.54086316e-01 -1.17138438e-01 7.40465820e-02 2.08230123e-01 -4.19718176e-01 4.69270080e-01 2.51023054e-01 -1.23728357e-01 -2.23525852e-01 -3.41528624e-01 4.70320247e-02 1.42272890e-01 -3.96070629e-01 -1.14133291e-01 4.57088262e-01 1.11843681e+00 -8.74020815e-01 -7.07114935e-01 1.17073037e-01 2.38693595e-01 -2.93557227e-01 8.70797455e-01 -6.03080273e-01 6.65619910e-01 -6.74484015e-01 2.60289591e-02 6.48916900e-01 -3.39056373e-01 9.11535993e-02 -2.53032446e-01 6.75146729e-02 4.71714824e-01 -6.94011807e-01 1.54873741e+00 -6.16684675e-01 4.51339066e-01 -3.17459367e-02 -8.24257731e-01 1.23650801e+00 4.25310612e-01 -1.02162950e-01 -1.09811676e+00 1.99064016e-01 -8.68481491e-03 8.65200721e-03 -6.38743162e-01 5.34417033e-01 -2.09558830e-01 -2.43527323e-01 8.06737304e-01 9.12930816e-02 -3.11229318e-01 2.17670068e-01 4.45415616e-01 6.05874121e-01 1.03703871e-01 1.21005557e-01 2.56381810e-01 8.11913490e-01 7.46300146e-02 2.45237201e-01 4.80872005e-01 1.99481055e-01 8.87356877e-01 7.27591395e-01 -1.46944979e-02 -1.10583258e+00 -9.41719532e-01 2.97680706e-01 9.23498154e-01 3.31775516e-01 -7.80783668e-02 -9.45899546e-01 -7.88875878e-01 -4.96498942e-01 6.26714945e-01 -5.03038347e-01 -7.54759759e-02 -4.05136079e-01 -3.19492877e-01 2.35742778e-01 3.56409520e-01 7.95768857e-01 -1.62853777e+00 -2.13669389e-01 3.18646848e-01 -7.42410719e-01 -1.15596938e+00 -8.03957641e-01 -3.69826227e-01 -3.52203488e-01 -8.23476791e-01 -1.29911482e+00 -1.09021425e+00 9.51402783e-01 4.22689319e-01 9.46591854e-01 1.18720017e-01 1.41530275e-01 7.23675117e-02 -7.01777577e-01 -2.75819272e-01 -8.07341099e-01 -1.90038994e-01 -5.04307628e-01 3.75310212e-01 1.72155142e-01 -2.87642293e-02 -6.91998720e-01 4.03872132e-01 -7.83647656e-01 8.50051165e-01 8.44992638e-01 1.06120992e+00 4.44602609e-01 -5.82852840e-01 6.17468417e-01 -8.93770456e-01 1.16546547e+00 -3.75925153e-01 -2.64210910e-01 3.60502928e-01 -6.96929768e-02 -4.93586110e-03 7.57360458e-01 -5.31578004e-01 -1.57933354e+00 -1.93123654e-01 -3.74337167e-01 -2.29861572e-01 -1.31467968e-01 6.53140008e-01 -4.27263439e-01 5.35813570e-01 3.31966251e-01 4.00694340e-01 2.74456352e-01 -1.47883564e-01 5.18259406e-01 1.15343845e+00 4.85647738e-01 -4.26138759e-01 5.11782885e-01 -1.26320589e-02 -7.47183204e-01 -7.71435142e-01 -9.25824881e-01 -2.61684895e-01 -2.57232219e-01 -3.58138382e-01 1.19370341e+00 -9.96742189e-01 -3.62332791e-01 4.59719747e-01 -1.64919925e+00 -4.70810562e-01 -7.07500102e-03 2.89462000e-01 -5.16730130e-01 4.60403889e-01 -7.43749917e-01 -7.22702742e-01 -3.42752606e-01 -1.40504873e+00 1.13067114e+00 5.89301109e-01 -1.97836772e-01 -8.65881920e-01 -7.80747533e-02 6.64408624e-01 -3.70232351e-02 -2.18306392e-01 8.78651321e-01 -8.05985272e-01 -3.80962312e-01 -1.22344699e-02 -6.52991652e-01 4.85138774e-01 2.69708857e-02 -3.06315750e-01 -9.17517960e-01 -5.59367128e-02 1.46986067e-01 -7.65717983e-01 6.32267654e-01 -3.32961120e-02 9.18781459e-01 -6.84560239e-01 -1.87710449e-02 7.82076269e-02 9.02748764e-01 3.17611873e-01 8.40817332e-01 -4.72652279e-02 6.71607912e-01 1.04035735e+00 9.67127085e-01 3.95366967e-01 6.88645184e-01 6.55633986e-01 1.67480394e-01 -5.26750445e-01 -1.27194911e-01 -5.75282514e-01 5.95902085e-01 1.03781354e+00 2.20576778e-01 -6.16565049e-01 -5.12974620e-01 4.30033863e-01 -1.87376869e+00 -6.52927637e-01 -2.34972998e-01 1.88889134e+00 1.24250507e+00 -1.32477313e-01 -1.24020971e-01 -5.19451022e-01 8.34340870e-01 3.18468839e-01 -5.03831923e-01 -5.30132174e-01 -1.15707703e-03 -4.41058502e-02 -1.89001024e-01 4.99444127e-01 -6.43823862e-01 1.03756154e+00 4.75940657e+00 7.46172011e-01 -9.90340590e-01 1.65382192e-01 9.24033284e-01 2.97627121e-01 -5.08589923e-01 -4.82171550e-02 -7.52665937e-01 4.67710316e-01 6.24317884e-01 -1.13254264e-01 2.81056702e-01 6.49595559e-01 3.74560505e-01 3.14010046e-02 -1.00519073e+00 7.79909015e-01 3.20295781e-01 -1.03553987e+00 4.27857667e-01 -4.97336984e-02 7.00996995e-01 -4.08424586e-01 -3.20139378e-02 5.98572671e-01 3.37985277e-01 -9.13659692e-01 5.60797215e-01 3.51874083e-01 9.10129786e-01 -6.14766598e-01 8.82977605e-01 6.29764736e-01 -9.08836424e-01 2.58585274e-01 -5.00260949e-01 1.99286384e-03 3.92105013e-01 9.35545117e-02 -1.14793241e+00 5.47012508e-01 1.64998278e-01 4.67237443e-01 -3.21069062e-01 5.58866799e-01 -7.48666584e-01 2.61451244e-01 1.59203485e-01 -4.69774842e-01 5.67499042e-01 -3.58470738e-01 1.42982125e-01 1.10915065e+00 1.96579203e-01 4.00775373e-01 4.43476588e-01 1.02349496e+00 -1.76875323e-01 6.04270399e-01 -4.63124275e-01 -2.27717489e-01 2.71011442e-01 1.44340467e+00 -4.64656889e-01 -6.55111969e-01 -6.85784996e-01 1.39950633e+00 4.34962511e-01 4.61383343e-01 -7.92680681e-01 -4.21290457e-01 1.40436828e-01 -1.33986935e-01 1.35227352e-01 2.32912466e-01 7.76891410e-02 -1.22371626e+00 1.72901228e-02 -1.23130655e+00 2.22239241e-01 -1.16383076e+00 -1.32108009e+00 7.97059894e-01 -1.04046628e-01 -1.30116630e+00 -5.08611679e-01 -4.11815017e-01 -8.98693383e-01 1.25031614e+00 -1.41352999e+00 -1.14620674e+00 -3.86290699e-01 6.63400531e-01 1.18163145e+00 1.41797792e-02 7.93397248e-01 -3.71479392e-02 -4.57698137e-01 5.94339430e-01 -3.34892005e-01 3.24520081e-01 8.41133237e-01 -1.06854033e+00 3.85371685e-01 6.00618064e-01 -9.08247679e-02 3.79417986e-01 5.37053287e-01 -7.40415215e-01 -8.97549868e-01 -1.04041922e+00 8.79761457e-01 1.61739718e-02 3.98008794e-01 -3.66133153e-01 -1.08325541e+00 4.09506649e-01 8.20524514e-01 -6.31877303e-01 5.29869318e-01 -2.30976701e-01 3.57143432e-02 3.69442225e-01 -7.29388535e-01 9.01581764e-01 6.30868673e-01 -4.84566391e-01 -9.12530601e-01 4.49385971e-01 8.52522373e-01 -6.03294432e-01 -3.53902847e-01 1.28607318e-01 1.37985975e-01 -7.36968398e-01 7.06930041e-01 -3.90973300e-01 1.12104666e+00 -1.08626634e-01 2.03112081e-01 -1.57255912e+00 1.47143170e-01 -4.78910059e-01 3.07846636e-01 1.42700124e+00 6.02356553e-01 -3.30310196e-01 4.34230745e-01 4.51633960e-01 -3.28448325e-01 -6.91161990e-01 -3.73841554e-01 -1.56375721e-01 -1.42578945e-01 -2.03166902e-02 3.82793814e-01 6.97233498e-01 2.32133478e-01 1.22909105e+00 -5.99057794e-01 -1.49842054e-01 1.68540347e-02 3.06172639e-01 1.14914775e+00 -6.67325675e-01 -3.92325073e-01 -1.62224784e-01 2.98346490e-01 -1.84964573e+00 1.37448013e-01 -6.38000190e-01 5.41888475e-01 -1.98584759e+00 5.04291534e-01 -1.69639915e-01 2.11297095e-01 3.51106524e-01 -5.86369753e-01 1.93740577e-01 3.00047010e-01 2.48924315e-01 -6.07834578e-01 8.72779071e-01 2.04825068e+00 -3.01482290e-01 -2.85673738e-01 -7.29089007e-02 -7.60850310e-01 7.46691048e-01 4.90647197e-01 -8.20574313e-02 -6.45262003e-01 -4.96248633e-01 -1.90227136e-01 6.91827059e-01 2.57435232e-01 -4.50555623e-01 1.09307162e-01 -2.49360800e-01 5.21389008e-01 -7.47851074e-01 6.21169090e-01 -2.52495080e-01 -4.35363650e-01 1.69094682e-01 -6.39598489e-01 -1.04339868e-01 -1.82865277e-01 4.40569103e-01 -4.46120679e-01 -4.48295891e-01 5.99130273e-01 -4.73213583e-01 -6.13170624e-01 3.19912940e-01 -4.02071595e-01 1.90203801e-01 8.70149970e-01 -1.03130981e-01 -3.95817846e-01 -1.00942779e+00 -5.89330018e-01 6.85019374e-01 2.78619856e-01 6.73716843e-01 9.91912425e-01 -1.28257310e+00 -9.33028400e-01 9.03253406e-02 2.36010909e-01 2.96669811e-01 5.44775963e-01 5.40627480e-01 -3.06809455e-01 2.52958566e-01 -2.07523659e-01 -4.01167244e-01 -1.31281686e+00 3.91129941e-01 -5.56981340e-02 -5.26122093e-01 -5.67548394e-01 8.86895418e-01 1.00420141e+00 -5.11974156e-01 1.45622998e-01 1.71003446e-01 -7.05677629e-01 1.97384745e-01 8.21700990e-01 -4.23023194e-01 -5.00473022e-01 -8.19099724e-01 1.72521904e-01 2.52342880e-01 -4.17537093e-01 -6.48534179e-01 9.38105404e-01 -3.81004870e-01 -1.41698597e-02 3.94285530e-01 1.26519990e+00 -1.62093878e-01 -1.38520086e+00 -4.41232949e-01 -5.45281291e-01 -2.88557410e-01 -4.51700121e-01 -8.65019321e-01 -8.54521453e-01 1.34436798e+00 1.47405654e-01 5.08525372e-02 9.61997747e-01 1.15380570e-01 1.05539501e+00 2.68509477e-01 4.95657139e-02 -1.14697218e+00 7.10990310e-01 6.61591947e-01 1.32211888e+00 -1.17210627e+00 -3.61845195e-01 -4.91580933e-01 -1.51957798e+00 1.07658958e+00 1.05110061e+00 1.48124099e-01 -1.36211455e-01 -2.94205606e-01 5.00576138e-01 1.49328746e-02 -8.77110720e-01 -2.94192791e-01 2.40043879e-01 6.24306560e-01 4.34096903e-01 -3.69291902e-02 -4.60281581e-01 8.29854429e-01 -2.64572710e-01 -4.05533195e-01 5.33165455e-01 5.55310309e-01 -5.32967091e-01 -1.17579019e+00 -2.03711867e-01 3.24492097e-01 -1.10185742e-01 -1.56185880e-01 -8.14663589e-01 4.28245246e-01 -1.73417807e-01 1.18237901e+00 7.40291923e-02 -4.71701562e-01 3.21623832e-01 -7.22567812e-02 1.62567601e-01 -9.63700652e-01 -4.39943075e-01 2.72256613e-01 3.71935159e-01 -1.14021756e-01 -5.21989949e-02 -3.52309436e-01 -1.41570449e+00 -1.77413020e-02 -3.84098530e-01 3.89347672e-01 4.60415870e-01 1.21950841e+00 1.10053802e-02 6.35432005e-01 9.39700603e-01 -5.81936061e-01 -7.11332262e-01 -1.31333387e+00 -1.86254486e-01 6.68998599e-01 2.34532822e-02 -3.55793238e-01 1.12639852e-02 7.65056089e-02]
[10.993925094604492, 1.4120914936065674]
663e5b6d-9938-4ec7-9eca-a8ae4222f2d5
temporal-coherent-and-graph-optimized
1804.06253
null
http://arxiv.org/abs/1804.06253v1
http://arxiv.org/pdf/1804.06253v1.pdf
Temporal Coherent and Graph Optimized Manifold Ranking for Visual Tracking
Recently, weighted patch representation has been widely studied for alleviating the impact of background information included in bounding box to improve visual tracking results. However, existing weighted patch representation models generally exploit spatial structure information among patches in each frame separately which ignore (1) unary featureof each patch and (2) temporal correlation among patches in different frames. To address this problem, we propose a novel unified temporal coherence and graph optimized ranking model for weighted patch representation in visual tracking problem. There are three main contributions of this paper. First, we propose to employ a flexible graph ranking for patch weight computation which exploits both structure information among patches and unary feature of each patch simultaneously. Second, we propose a new more discriminative ranking model by further considering the temporal correlation among patches in different frames. Third, a neighborhood preserved, low-rank graph is learned and incorporated to build a unified optimized ranking model. Experiments on two benchmark datasets show the benefits of our model.
['Jin Tang', 'Bin Luo', 'Doudou Lin', 'Bo Jiang']
2018-04-17
null
null
null
null
['graph-ranking']
['graphs']
[ 1.00685477e-01 -3.33247334e-01 -5.38775802e-01 -1.19151697e-02 -3.87781858e-01 -4.67630208e-01 2.77721196e-01 2.54130870e-01 -1.27617776e-01 5.02841949e-01 3.10605347e-01 2.94367552e-01 -4.40771401e-01 -6.87277734e-01 -5.55174053e-01 -8.18733335e-01 -3.18390690e-02 -4.22152847e-01 1.00274181e+00 -7.95201585e-02 1.31831497e-01 4.16749448e-01 -1.30138731e+00 2.28365250e-02 8.04163337e-01 8.18528414e-01 4.31198984e-01 2.11414590e-01 1.86524630e-01 7.79141307e-01 -2.57726699e-01 -4.79724146e-02 2.58697897e-01 -2.77282625e-01 -2.45812461e-01 3.87128174e-01 8.98455501e-01 -1.41447142e-01 -6.63006127e-01 1.36307573e+00 5.25459230e-01 2.86983311e-01 2.40508199e-01 -1.06989694e+00 -4.78563040e-01 3.19312304e-01 -1.08431900e+00 5.38467884e-01 6.83263093e-02 3.92689221e-02 1.13245702e+00 -7.27385521e-01 7.47037888e-01 1.11294878e+00 5.71181893e-01 5.15357368e-02 -1.05254364e+00 -5.27233362e-01 7.92828798e-01 2.87296772e-01 -1.48175859e+00 -9.68398675e-02 1.26123285e+00 -5.17748952e-01 3.75957698e-01 3.40580970e-01 9.38151181e-01 3.31138521e-01 1.25787169e-01 7.64961541e-01 9.54314053e-01 -1.59247354e-01 -3.37940037e-01 -1.89362690e-01 3.97817612e-01 1.04605758e+00 4.65361923e-01 1.46202892e-01 -3.97669852e-01 -4.13945019e-01 6.64859176e-01 5.11703014e-01 -5.02061188e-01 -9.62785482e-01 -1.26539326e+00 5.77308536e-01 6.73896968e-01 4.91331756e-01 -2.50420362e-01 3.04450542e-01 2.48746321e-01 3.23031470e-02 5.59901237e-01 -2.83262402e-01 -1.55128941e-01 4.71321255e-01 -8.03902030e-01 1.76648065e-01 6.35046810e-02 9.21770990e-01 1.02547598e+00 1.80850387e-01 -6.88435256e-01 9.11988497e-01 5.20289421e-01 7.22108066e-01 2.67409384e-01 -5.74169874e-01 4.02444839e-01 8.67693663e-01 5.60519993e-02 -1.89382732e+00 -3.06411386e-01 -7.23371983e-01 -8.29430163e-01 -2.65666880e-02 1.54956371e-01 1.22514755e-01 -6.57561362e-01 1.83958006e+00 5.85637927e-01 7.00988293e-01 -4.57389444e-01 1.06538761e+00 9.68146205e-01 5.50113976e-01 9.88275185e-02 -4.43579227e-01 1.43670595e+00 -1.06404161e+00 -4.85091448e-01 1.05901748e-01 2.72683263e-01 -7.49598324e-01 6.24303937e-01 -1.00643799e-01 -7.85529554e-01 -7.91188717e-01 -1.13742685e+00 2.35758424e-01 -8.12952220e-02 2.07725480e-01 5.30615389e-01 3.54603469e-01 -8.74175727e-01 3.88195246e-01 -6.00529075e-01 -2.78620571e-01 3.34898412e-01 3.30429345e-01 -3.79809409e-01 -2.02803329e-01 -1.02155232e+00 2.54020184e-01 2.07417771e-01 1.41791925e-01 -6.22054338e-01 -6.60983384e-01 -7.82238305e-01 -7.57180527e-03 6.21431947e-01 -5.38017571e-01 6.30487561e-01 -8.82674754e-01 -9.38024879e-01 2.79060990e-01 -3.96553636e-01 -3.87344062e-02 2.10819304e-01 -1.73879657e-02 -4.80155140e-01 1.69445753e-01 1.37208149e-01 2.58010477e-01 8.76281321e-01 -1.24976921e+00 -6.92974329e-01 -4.00201619e-01 1.77999567e-02 2.68647641e-01 -3.41065496e-01 -2.27662116e-01 -8.60128343e-01 -1.03838539e+00 2.66034395e-01 -8.48401368e-01 -3.71219188e-01 5.73307164e-02 -2.46052012e-01 -2.42419690e-01 1.01890695e+00 -4.95881498e-01 1.72937465e+00 -2.13901019e+00 2.43776858e-01 4.34887379e-01 5.58923781e-01 4.01891530e-01 -3.42127889e-01 2.26968050e-01 3.23040574e-03 -1.76366135e-01 3.41063701e-02 9.22697484e-02 -2.49919668e-01 8.00229982e-02 -1.51338920e-01 5.66361129e-01 2.28212982e-01 8.13193262e-01 -1.15921092e+00 -8.79992247e-01 3.50709140e-01 7.07424641e-01 -4.22567546e-01 -1.01181492e-01 6.86146691e-02 4.66233015e-01 -8.99587154e-01 6.77255690e-01 1.00497115e+00 -1.62074685e-01 4.04986069e-02 -7.55187809e-01 -3.06017280e-01 -2.88663924e-01 -1.47289836e+00 1.64273822e+00 -2.21459847e-02 2.91407526e-01 -2.87448727e-02 -8.97934973e-01 7.99294531e-01 3.96869704e-02 9.05264020e-01 -4.96536583e-01 -7.82625005e-02 -8.19803178e-02 -1.93914518e-01 -3.38631004e-01 4.92122591e-01 1.36944190e-01 2.61520892e-01 2.05696061e-01 -9.32957008e-02 6.88476741e-01 2.26221859e-01 3.20607781e-01 1.01407647e+00 2.31243312e-01 2.39344969e-01 -4.75199521e-01 1.01167130e+00 -1.17026225e-01 1.15201771e+00 5.87355793e-01 -5.30094266e-01 6.45480514e-01 1.66532174e-01 -6.42416418e-01 -4.96086091e-01 -7.62870133e-01 5.68420999e-02 1.08902502e+00 9.00463402e-01 -7.17963040e-01 -2.52956063e-01 -8.88180315e-01 4.02857736e-02 -2.99024165e-01 -6.34318411e-01 -4.76089530e-02 -9.09126401e-01 -7.80901551e-01 4.51463042e-03 4.29049104e-01 5.32945454e-01 -5.88165104e-01 -3.64688814e-01 2.01443821e-01 -3.13810408e-01 -8.64039898e-01 -1.09131563e+00 -4.17960972e-01 -1.08762276e+00 -1.20182836e+00 -8.75522017e-01 -8.37156177e-01 7.77801216e-01 1.20397604e+00 9.27959800e-01 6.21981859e-01 -2.90993571e-01 6.11544490e-01 -4.49966580e-01 7.87267461e-02 5.96436024e-01 -1.49586007e-01 -3.22737843e-01 4.18163985e-01 1.95530392e-02 -4.38412637e-01 -9.51671481e-01 6.14807725e-01 -7.60526538e-01 -9.45308954e-02 6.76085711e-01 8.55761170e-01 1.02529848e+00 2.12360337e-01 1.68150097e-01 -8.16598713e-01 1.53013006e-01 -3.72226804e-01 -7.03432918e-01 5.95116258e-01 -2.93533593e-01 -1.47760073e-02 1.99518710e-01 -6.34655118e-01 -9.01410818e-01 3.09773922e-01 3.73506874e-01 -4.24148798e-01 3.44503194e-01 3.93577486e-01 -3.90134037e-01 -5.05224288e-01 8.09222311e-02 2.66065747e-01 -1.77523255e-01 -3.96767169e-01 2.35023573e-01 -1.07730582e-01 2.22688600e-01 -6.64358616e-01 1.18268800e+00 6.44928992e-01 2.93649256e-01 -7.44895577e-01 -7.57228196e-01 -7.49941349e-01 -6.14609957e-01 -4.15588468e-01 7.21262693e-01 -1.09951413e+00 -5.42741179e-01 8.73329416e-02 -9.45225596e-01 1.59033164e-01 -8.13473538e-02 5.84273040e-01 -1.91388428e-02 9.32946146e-01 -2.34742135e-01 -7.52376437e-01 -2.77155042e-01 -9.87438560e-01 1.17417240e+00 3.32809180e-01 3.62873405e-01 -8.82647395e-01 3.12106282e-01 3.01058460e-02 2.86097497e-01 4.97846663e-01 5.48033893e-01 8.00455883e-02 -9.67372954e-01 -2.84422398e-01 -6.16107106e-01 -1.13898572e-02 1.96270406e-01 1.03369489e-01 -4.84497219e-01 -3.55618149e-01 -2.81449735e-01 3.61367941e-01 1.17305589e+00 3.84701550e-01 8.50156188e-01 -2.33504251e-01 -7.53093779e-01 4.62366432e-01 1.51086521e+00 -1.99329369e-02 3.74205232e-01 7.85372313e-03 1.14165831e+00 4.49535817e-01 8.56709898e-01 3.75810355e-01 3.02682191e-01 1.02569866e+00 3.11770827e-01 -9.64452848e-02 -3.61633420e-01 -2.33265042e-01 4.23756957e-01 8.83427083e-01 -3.86933327e-01 1.42381057e-01 -4.08882022e-01 6.06299460e-01 -2.21135807e+00 -1.09346867e+00 -3.50115269e-01 2.22684622e+00 1.85271218e-01 -6.62987530e-02 3.36556166e-01 -1.54926986e-01 1.10260868e+00 6.56410158e-01 -2.90074080e-01 5.51624000e-01 -2.92889327e-01 -8.01662505e-02 4.28295523e-01 4.99501497e-01 -1.40222681e+00 7.75249839e-01 4.82422304e+00 1.06942165e+00 -9.61369276e-01 2.22293392e-01 2.07953930e-01 2.09826957e-02 -3.32422674e-01 2.15519935e-01 -7.87298739e-01 6.39204443e-01 -8.93573090e-02 -1.53866589e-01 1.52705804e-01 7.72283733e-01 3.56020071e-02 4.90604267e-02 -5.29999554e-01 9.63105202e-01 1.76974878e-01 -1.06935692e+00 1.88311130e-01 4.29444201e-02 7.12842464e-01 -1.62927687e-01 7.30371401e-02 8.89981687e-02 5.90289906e-02 -4.42065567e-01 5.81805348e-01 6.99153900e-01 2.18737245e-01 -7.12391794e-01 6.48458123e-01 -4.29215357e-02 -2.11257529e+00 1.64930522e-01 -7.40037620e-01 1.75386921e-01 -1.15583539e-02 7.31006265e-01 9.23090950e-02 1.08556759e+00 6.09098256e-01 1.34907508e+00 -8.39582086e-01 1.45138872e+00 5.64603060e-02 2.60636508e-01 -1.17642626e-01 2.76472986e-01 1.94657251e-01 -2.66635835e-01 8.75772953e-01 1.08174002e+00 1.85973436e-01 2.48335630e-01 6.15496278e-01 3.84480864e-01 3.79735142e-01 3.43418539e-01 -6.48480654e-01 4.02797312e-01 1.52701005e-01 1.50849414e+00 -8.79312217e-01 -1.46560565e-01 -7.15457559e-01 7.33283937e-01 2.53354579e-01 4.66073334e-01 -9.24491227e-01 -1.43853903e-01 5.74644864e-01 2.15599015e-01 5.98732591e-01 -3.44424218e-01 2.50958055e-01 -1.49097478e+00 1.83466241e-01 -5.25247693e-01 6.44040227e-01 -3.75825375e-01 -1.17835188e+00 5.86907864e-01 6.55797273e-02 -1.82437980e+00 4.62295651e-01 -3.43153715e-01 -6.59014106e-01 5.92164874e-01 -1.86818743e+00 -1.42947376e+00 -4.73939538e-01 7.79670298e-01 2.60902971e-01 2.28444944e-04 2.38337189e-01 5.35668194e-01 -7.78779566e-01 4.40738499e-01 -1.21608466e-01 1.79581210e-01 7.16102958e-01 -1.01234448e+00 -3.18319678e-01 9.92422998e-01 3.06107581e-01 7.52994061e-01 2.59163767e-01 -8.57872307e-01 -1.44468200e+00 -1.39554214e+00 7.25474715e-01 -2.14199692e-01 5.26483357e-01 4.77731563e-02 -9.04729366e-01 3.41863573e-01 2.04777509e-01 4.80500132e-01 6.03647292e-01 -6.45627268e-03 -5.29453218e-01 -4.83058959e-01 -7.65875876e-01 5.09846210e-01 1.15887165e+00 -3.05098921e-01 -3.83432120e-01 9.18195248e-02 5.77363193e-01 -1.74194902e-01 -7.61975646e-01 8.14815402e-01 7.44706869e-01 -9.11606550e-01 1.15827549e+00 -1.73147187e-01 -2.42532864e-01 -1.01864028e+00 -1.34272158e-01 -9.94863927e-01 -8.15634549e-01 -5.71812391e-01 -1.78760901e-01 1.27024996e+00 -2.00077742e-02 -5.15242517e-01 8.27426136e-01 2.36958668e-01 -2.43586767e-02 -6.27566934e-01 -7.87998080e-01 -8.18959594e-01 -4.36019748e-01 -1.66035052e-02 5.52477717e-01 9.12557721e-01 -2.24965125e-01 3.86688858e-01 -7.39183009e-01 3.58216763e-01 9.17438149e-01 5.30294180e-01 7.45005965e-01 -1.49361324e+00 -4.17327642e-01 -4.17319387e-01 -8.10696900e-01 -1.27431571e+00 -1.35861516e-01 -7.92244613e-01 -3.39816771e-02 -1.36419952e+00 6.79349244e-01 -3.50090057e-01 -8.37009192e-01 3.51030916e-01 -7.51324952e-01 5.61535895e-01 5.30070364e-01 3.93270999e-01 -1.11816800e+00 5.82442999e-01 1.44530034e+00 -2.65534520e-01 -2.94801258e-02 -2.12197289e-01 -6.24604285e-01 6.16995156e-01 4.85837936e-01 -4.89724934e-01 -5.03788531e-01 -4.87276256e-01 -6.44618943e-02 -5.84502891e-02 6.45462453e-01 -1.04740489e+00 2.39867598e-01 -2.01771542e-01 4.73345399e-01 -8.64205480e-01 1.15258768e-01 -1.18729758e+00 2.59487659e-01 4.97438550e-01 1.19751774e-01 1.19135745e-01 9.95561704e-02 1.18452203e+00 -4.18230146e-01 6.37584403e-02 7.60196209e-01 1.29998535e-01 -9.96212959e-01 6.84441447e-01 1.05838954e-01 -1.33199379e-01 1.12149692e+00 -1.82287291e-01 -3.00314724e-01 -8.98108035e-02 -6.59103215e-01 3.42405945e-01 4.86623764e-01 4.65896457e-01 6.23967230e-01 -1.74771774e+00 -6.07165158e-01 9.58301947e-02 3.10444176e-01 -3.61523479e-01 7.60991096e-01 1.07350504e+00 -1.38471648e-01 2.84384608e-01 -1.19436830e-01 -6.88051224e-01 -1.69506514e+00 6.39412761e-01 2.07922757e-01 -4.42641705e-01 -6.26988709e-01 6.49927318e-01 7.23724663e-01 9.73855630e-02 1.30296946e-01 -3.25891852e-01 -5.18909156e-01 5.67747727e-02 6.01700604e-01 3.54638547e-01 -2.90710896e-01 -1.20002270e+00 -5.31958878e-01 1.43833113e+00 -7.09358752e-02 4.23346162e-01 1.07759058e+00 -1.95632309e-01 -1.20282955e-01 1.31404772e-01 1.01993632e+00 3.65655154e-01 -1.27274609e+00 -5.40544510e-01 -6.40649945e-02 -9.26485658e-01 6.09438270e-02 -1.51185408e-01 -1.70560348e+00 6.82242215e-01 7.40264833e-01 -1.11774378e-03 1.17724895e+00 -2.52541542e-01 8.21266472e-01 8.25935788e-03 3.79676431e-01 -6.78448200e-01 3.42340350e-01 1.27987161e-01 6.86953187e-01 -1.17383754e+00 3.97750825e-01 -8.44525397e-01 -5.78463733e-01 8.44561934e-01 6.34743154e-01 -4.25952613e-01 9.32969809e-01 -3.89033526e-01 -1.76706567e-01 -3.64333481e-01 -3.85570586e-01 -6.31416142e-01 8.98260176e-01 5.93695819e-01 4.78136629e-01 -1.26201525e-01 -6.19122148e-01 5.12292981e-01 6.50904953e-01 -1.87768385e-01 -1.96374282e-01 1.06871092e+00 -5.75039506e-01 -1.26301098e+00 -2.21529543e-01 2.00501502e-01 -4.20907110e-01 -7.02144057e-02 -2.49500290e-01 6.49787486e-01 4.46969688e-01 9.22838449e-01 -3.44613522e-01 -5.42898715e-01 4.81363058e-01 -4.02668327e-01 4.65108633e-01 -5.51108003e-01 -5.26798606e-01 6.68844998e-01 -3.25904220e-01 -6.72396541e-01 -8.74917150e-01 -6.69384599e-01 -8.14744890e-01 7.98533112e-02 -5.21400750e-01 3.21105272e-01 -7.95696825e-02 4.72855240e-01 5.18974960e-01 4.97425377e-01 4.98008996e-01 -8.13984811e-01 3.17697134e-03 -6.27213776e-01 -8.02994251e-01 4.46736455e-01 3.60344321e-01 -1.19792235e+00 -1.02466986e-01 8.21161550e-03]
[6.423912525177002, -2.1602861881256104]
0d074735-3c2a-4e38-904d-428e12ffeeea
self-supervised-human-mesh-recovery-with
2209.04596
null
https://arxiv.org/abs/2209.04596v1
https://arxiv.org/pdf/2209.04596v1.pdf
Self-supervised Human Mesh Recovery with Cross-Representation Alignment
Fully supervised human mesh recovery methods are data-hungry and have poor generalizability due to the limited availability and diversity of 3D-annotated benchmark datasets. Recent progress in self-supervised human mesh recovery has been made using synthetic-data-driven training paradigms where the model is trained from synthetic paired 2D representation (e.g., 2D keypoints and segmentation masks) and 3D mesh. However, on synthetic dense correspondence maps (i.e., IUV) few have been explored since the domain gap between synthetic training data and real testing data is hard to address for 2D dense representation. To alleviate this domain gap on IUV, we propose cross-representation alignment utilizing the complementary information from the robust but sparse representation (2D keypoints). Specifically, the alignment errors between initial mesh estimation and both 2D representations are forwarded into regressor and dynamically corrected in the following mesh regression. This adaptive cross-representation alignment explicitly learns from the deviations and captures complementary information: robustness from sparse representation and richness from dense representation. We conduct extensive experiments on multiple standard benchmark datasets and demonstrate competitive results, helping take a step towards reducing the annotation effort needed to produce state-of-the-art models in human mesh estimation.
['Ziyan Wu', 'David Doermann', 'Terrence Chen', 'Srikrishna Karanam', 'Benjamin Planche', 'Meng Zheng', 'Xuan Gong']
2022-09-10
null
null
null
null
['human-mesh-recovery']
['computer-vision']
[ 3.90083581e-01 4.51710165e-01 -4.53473210e-01 -1.76514953e-01 -1.23300958e+00 -1.33029088e-01 3.73280942e-01 6.00472800e-02 2.49526259e-02 6.48361802e-01 2.85348028e-01 5.39793134e-01 9.55659151e-02 -7.57509530e-01 -1.00293720e+00 -4.28867549e-01 6.91685230e-02 9.98313367e-01 2.91763902e-01 -3.58166665e-01 -1.30520836e-02 5.18291175e-01 -1.61588633e+00 1.79434627e-01 7.36289144e-01 9.49457884e-01 5.43004982e-02 2.01482669e-01 -1.54728472e-01 2.58502990e-01 -2.51546234e-01 1.16318148e-02 5.24668455e-01 -4.06130016e-01 -8.84357035e-01 2.49156594e-01 8.53519082e-01 -2.55872011e-01 -5.19030988e-01 9.50097263e-01 7.04812884e-01 -9.29617509e-02 8.85070682e-01 -1.00696683e+00 -4.14452344e-01 1.94082782e-01 -1.08276105e+00 -2.86760002e-01 6.61037266e-01 1.66061148e-01 6.02613509e-01 -1.30835235e+00 1.32605863e+00 1.42133284e+00 9.90678072e-01 6.58823252e-01 -1.60104489e+00 -6.74872041e-01 -1.34213921e-02 -4.37498689e-01 -1.49242187e+00 -4.07865703e-01 1.06347394e+00 -7.13105440e-01 6.54394686e-01 1.33924559e-01 7.96256602e-01 1.32473803e+00 -1.44243129e-02 7.39722908e-01 9.27715421e-01 -2.35959873e-01 -1.55033227e-02 -2.58787498e-02 -4.88735467e-01 8.96437228e-01 1.75937057e-01 1.63622528e-01 -4.64667499e-01 -2.16356933e-01 1.30760086e+00 -1.42725930e-01 -2.67910779e-01 -9.61143672e-01 -1.30834150e+00 6.38733625e-01 5.16358376e-01 -1.27405211e-01 -5.36685824e-01 1.03479706e-01 5.01511753e-01 2.26932973e-01 8.58962774e-01 4.63919669e-01 -4.11292344e-01 1.71734914e-01 -1.17557764e+00 6.10295773e-01 5.06725192e-01 1.00912511e+00 8.14547181e-01 2.30125353e-01 -4.21879953e-03 1.18249989e+00 1.48462623e-01 6.37256861e-01 3.22990358e-01 -9.48366702e-01 5.70379734e-01 7.87957907e-01 -2.19428226e-01 -1.25118971e+00 -3.24842393e-01 -2.37007380e-01 -1.00083625e+00 3.48958373e-01 5.91948628e-01 1.96798190e-01 -1.16924953e+00 1.59263885e+00 6.92336738e-01 2.49925017e-01 -2.68585503e-01 9.28299189e-01 1.14938080e+00 2.61893988e-01 -1.09567724e-01 1.03362791e-01 1.03900027e+00 -8.40995789e-01 -2.95780391e-01 -2.89619803e-01 4.76504385e-01 -7.87308335e-01 1.01432228e+00 8.08982253e-02 -1.38366330e+00 -6.76416099e-01 -7.83920348e-01 -7.60938674e-02 -9.15245786e-02 -4.14720736e-02 3.13566834e-01 -8.01748708e-02 -6.55197918e-01 7.13484943e-01 -8.50452483e-01 -3.01921278e-01 7.91008413e-01 3.39643985e-01 -8.53312016e-01 -2.70197183e-01 -9.66905892e-01 7.21017241e-01 1.94179684e-01 -1.37074705e-04 -7.41672277e-01 -9.63684320e-01 -1.27785242e+00 -5.02890408e-01 3.10876966e-01 -8.13679099e-01 7.53048360e-01 -7.01295972e-01 -1.23922431e+00 1.56278884e+00 1.17753193e-01 -1.42218202e-01 7.77832270e-01 -7.11484104e-02 3.74310799e-02 2.24197552e-01 5.64167917e-01 1.00605547e+00 1.09761643e+00 -1.62927437e+00 -3.01970363e-01 -4.45749074e-01 -5.05526125e-01 -6.36323318e-02 2.03768000e-01 -3.92259657e-01 -4.11373109e-01 -1.23936331e+00 6.03347659e-01 -1.15631521e+00 -3.55776668e-01 4.96012449e-01 -5.88972092e-01 1.29794627e-01 7.30270684e-01 -7.97615647e-01 9.02501822e-01 -1.97347629e+00 6.44831181e-01 4.01424021e-01 3.37117136e-01 -3.13256197e-02 -2.25674704e-01 2.30753481e-01 -2.07495317e-01 -1.24672912e-01 -6.15834832e-01 -5.77539325e-01 -2.74677187e-01 2.45651454e-01 -1.75877005e-01 6.99684978e-01 5.05436540e-01 9.44770753e-01 -9.96409178e-01 -7.96397209e-01 3.82351071e-01 5.62075019e-01 -6.70950830e-01 4.41983461e-01 -2.26754665e-01 8.71441662e-01 -4.05727655e-01 1.08644331e+00 8.39067936e-01 -3.43014777e-01 -1.79795891e-01 -4.90567178e-01 2.87215590e-01 -5.07693700e-02 -1.28494871e+00 2.34257197e+00 -5.01695335e-01 3.40930402e-01 1.76986396e-01 -8.63832653e-01 9.07910287e-01 2.66717702e-01 8.35840344e-01 -7.01551735e-01 1.06388994e-01 4.66310084e-01 -4.35053825e-01 -4.07370120e-01 1.20855361e-01 -1.21464580e-01 -1.38427280e-02 2.88612962e-01 1.89843372e-01 -6.69284463e-01 -2.44572192e-01 8.52408260e-02 8.59157026e-01 5.55438280e-01 3.21009755e-01 -1.10123925e-01 2.26816460e-01 2.58351713e-01 5.83094656e-01 2.23385036e-01 6.14365116e-02 1.37129712e+00 2.81693786e-01 -4.45154309e-01 -1.22033620e+00 -1.08467913e+00 -3.72731060e-01 4.99255985e-01 4.19925541e-01 -4.64813501e-01 -7.03137517e-01 -7.84883857e-01 4.52905834e-01 1.31744012e-01 -1.03013992e+00 -2.07684606e-01 -9.66510236e-01 -2.27735251e-01 4.35349107e-01 6.26030564e-01 2.82767117e-01 -1.03907073e+00 -4.21449035e-01 2.03330979e-01 -1.38752833e-01 -1.12808192e+00 -3.65124881e-01 -2.68561821e-02 -1.00153589e+00 -1.23242140e+00 -9.96903658e-01 -8.33036542e-01 1.00312865e+00 7.04840720e-02 1.41761374e+00 2.51532733e-01 -6.11123443e-01 3.27398092e-01 -1.29591331e-01 -7.01973513e-02 -4.28958505e-01 6.00759052e-02 -3.29193771e-02 -2.14189485e-01 -4.59746331e-01 -7.42816985e-01 -7.07903802e-01 5.34459591e-01 -5.40740371e-01 2.95110852e-01 6.25518203e-01 1.11912131e+00 1.28434384e+00 -3.86548609e-01 4.19552684e-01 -1.15829420e+00 9.22717080e-02 -7.35808134e-01 -1.46266386e-01 -4.41998132e-02 -3.42639357e-01 -5.10929525e-02 2.91550577e-01 -5.90022326e-01 -9.08593893e-01 3.34544897e-01 -2.61142313e-01 -9.54041481e-01 -1.16342925e-01 8.62829760e-02 -5.89604825e-02 -2.31900677e-01 8.31544578e-01 -2.78226882e-02 2.98066169e-01 -5.92540145e-01 3.04235727e-01 2.99307734e-01 6.97111309e-01 -8.93230081e-01 9.32230294e-01 5.58146715e-01 1.09402478e-01 -7.57771194e-01 -8.25765431e-01 -3.40116978e-01 -9.02627051e-01 -1.88809365e-01 7.68236816e-01 -1.14242625e+00 2.53101550e-02 2.86407977e-01 -9.58291769e-01 -4.34082717e-01 -7.24242449e-01 1.96715638e-01 -8.44795763e-01 2.78073788e-01 -5.86233974e-01 -3.86031002e-01 -3.62713933e-01 -1.35022998e+00 1.89716959e+00 -2.54111111e-01 -7.07656324e-01 -8.87189031e-01 2.01015547e-01 3.86366248e-01 1.09812811e-01 1.04885828e+00 8.48291039e-01 -8.47015902e-02 -2.85585701e-01 -4.43617493e-01 -1.53806701e-01 2.57295907e-01 2.67052174e-01 -1.24569856e-01 -8.93419921e-01 -3.34312916e-01 -3.31084639e-01 -7.17849255e-01 6.71448231e-01 3.47759455e-01 1.30235565e+00 -4.25838167e-03 -5.96653759e-01 8.02951753e-01 1.05764580e+00 -6.48261368e-01 4.84312981e-01 4.55424190e-02 1.15078175e+00 8.20709527e-01 8.62501621e-01 2.72748411e-01 2.33680531e-01 9.39826012e-01 4.19834465e-01 -4.40271676e-01 -8.05842757e-01 -7.54629552e-01 -3.18396032e-01 6.79573476e-01 -1.70147747e-01 2.54164249e-01 -1.03303802e+00 6.05519533e-01 -1.79034865e+00 -5.56018293e-01 -8.07091072e-02 2.18589878e+00 9.71857667e-01 2.50140548e-01 1.16855025e-01 1.02337219e-01 4.88113105e-01 1.95873126e-01 -6.61022782e-01 3.02042902e-01 -1.03162967e-01 4.20059741e-01 2.20046639e-01 4.05744672e-01 -1.07182312e+00 1.06942821e+00 5.86373663e+00 1.04510939e+00 -9.56078053e-01 -7.09601771e-03 4.23618466e-01 -1.21820429e-02 -3.09851557e-01 -2.04528078e-01 -3.67951423e-01 4.47451890e-01 3.01508069e-01 2.25386873e-01 1.67784579e-02 9.87257183e-01 -7.17159584e-02 -2.72372086e-02 -1.38090837e+00 1.53273487e+00 -2.44936775e-02 -1.50409245e+00 1.95782647e-01 1.71944246e-01 1.11788380e+00 -1.08347118e-01 -1.10967204e-01 1.43693924e-01 2.43372619e-02 -1.14621508e+00 7.69597411e-01 5.56617498e-01 1.41736293e+00 -4.92034078e-01 5.21355093e-01 2.84689635e-01 -1.39944506e+00 4.25129771e-01 -2.19621807e-01 2.54803240e-01 2.10094169e-01 4.99989688e-01 -7.35393167e-01 5.37656367e-01 5.47595799e-01 1.03939676e+00 -3.96268785e-01 6.68222189e-01 2.55354606e-02 2.30004668e-01 -3.47597241e-01 8.07687223e-01 -5.82135394e-02 -1.06052734e-01 6.26099348e-01 1.01952362e+00 3.45034972e-02 5.97952791e-02 4.38859254e-01 9.26884770e-01 -2.50700295e-01 2.24677548e-01 -8.06159735e-01 2.93584704e-01 4.67846870e-01 8.75515580e-01 -9.39680994e-01 -3.02270204e-01 -1.50597826e-01 9.94574845e-01 5.12759984e-01 1.12437278e-01 -6.04859710e-01 1.45103633e-01 7.88076043e-01 8.92778575e-01 8.89625517e-05 -8.32492709e-02 -5.23896754e-01 -1.03546429e+00 2.00524163e-02 -8.52138162e-01 1.20981075e-01 -5.33075750e-01 -1.36825264e+00 4.90213841e-01 8.39726627e-02 -1.45979226e+00 -2.86388159e-01 -9.04819593e-02 -3.22032958e-01 7.36358106e-01 -1.11436188e+00 -1.29714561e+00 -5.38418055e-01 4.52900946e-01 9.06580448e-01 -1.24161929e-01 9.65680480e-01 4.39953744e-01 -3.59878480e-01 7.50882328e-01 -3.11583936e-01 1.63435742e-01 8.46733809e-01 -9.86190438e-01 6.65122211e-01 2.33379185e-01 1.70128495e-02 2.36859620e-01 4.68640327e-01 -1.13798416e+00 -1.42766881e+00 -1.18491316e+00 2.81898260e-01 -6.66814089e-01 1.06700882e-01 -4.37700689e-01 -1.16474903e+00 4.65052664e-01 -5.76805353e-01 6.67815208e-01 3.17666113e-01 -2.06187829e-01 -3.84287357e-01 3.06628495e-01 -1.38124955e+00 4.47994530e-01 1.50917721e+00 -5.05374849e-01 -4.43160325e-01 4.09145653e-01 5.67515135e-01 -9.61108029e-01 -1.25488865e+00 8.63040268e-01 6.89026594e-01 -6.48059785e-01 1.36184907e+00 -6.21533453e-01 8.27770531e-01 -1.45426273e-01 -3.12546700e-01 -1.16922605e+00 -6.47464320e-02 -2.79237837e-01 -3.18814397e-01 1.01533461e+00 1.33504704e-01 -1.12490460e-01 1.13427222e+00 4.69926596e-01 -1.45868883e-01 -1.32295036e+00 -8.48017633e-01 -5.12300074e-01 4.38776389e-02 -2.44259775e-01 4.60693836e-01 1.06782663e+00 -3.84467393e-01 1.21896200e-01 -4.68385994e-01 -1.01793796e-01 8.02974224e-01 1.63164183e-01 1.11210763e+00 -1.44875622e+00 -2.50981911e-03 -2.27885827e-01 -6.29433334e-01 -1.10171998e+00 3.82220209e-01 -9.46115017e-01 2.03417912e-02 -1.49873507e+00 -7.14190379e-02 -8.11447859e-01 2.35141858e-01 3.63473713e-01 -3.03852744e-02 7.58435786e-01 -1.23478927e-01 4.75049794e-01 -2.41705045e-01 8.05537581e-01 1.60008466e+00 -2.80043632e-01 -1.23511612e-01 -1.43478855e-01 -3.33802611e-01 9.93926287e-01 3.53418291e-01 -6.26953006e-01 -3.64314765e-01 -3.88777405e-01 3.42142321e-02 2.93360174e-01 4.95130748e-01 -1.02286351e+00 -2.51176618e-02 -1.59886852e-02 5.83749294e-01 -5.89968264e-01 5.54234982e-01 -8.31028581e-01 3.03912669e-01 1.23434983e-01 -2.68932045e-01 -1.48277894e-01 5.31880781e-02 8.52977812e-01 -2.31532782e-01 2.55858511e-01 1.01260066e+00 -3.26823950e-01 -5.40988982e-01 7.09230602e-01 4.09808844e-01 6.58488691e-01 1.01298654e+00 -6.03813589e-01 4.24494833e-01 -3.21176276e-02 -9.28095579e-01 1.59330636e-01 9.61706400e-01 4.53672349e-01 1.07537973e+00 -1.61369145e+00 -8.91963065e-01 5.07795811e-01 2.44097799e-01 7.58156955e-01 3.80216986e-01 6.79010749e-01 -5.41938007e-01 -1.99190184e-01 -4.08347487e-01 -9.62861300e-01 -1.16946590e+00 2.98848093e-01 3.49185169e-01 -1.44489259e-01 -1.17931509e+00 9.39222038e-01 2.49173254e-01 -6.82698727e-01 4.27359819e-01 -2.41536796e-01 1.85515672e-01 -9.00201034e-03 1.77802831e-01 4.27097797e-01 1.86685428e-01 -9.69232440e-01 -3.17052484e-01 1.23322916e+00 5.21843508e-02 2.67496407e-01 1.39647102e+00 1.10119097e-01 2.29805671e-02 3.92911941e-01 1.49994540e+00 1.16323195e-01 -1.37379324e+00 -4.33601856e-01 -2.70723701e-01 -8.93620133e-01 -1.92568630e-01 -2.44408876e-01 -1.41949379e+00 6.44378603e-01 6.11223221e-01 -3.55506599e-01 6.30386531e-01 3.57698232e-01 1.10228395e+00 -1.12009868e-01 5.88939071e-01 -9.53805029e-01 2.31428772e-01 8.13317895e-02 1.31893063e+00 -1.58133328e+00 3.61077279e-01 -9.02378023e-01 -4.28716063e-01 8.52151632e-01 7.64561892e-01 -5.82280993e-01 8.56690824e-01 1.85042307e-01 -8.91049281e-02 -8.09495986e-01 -2.28415236e-01 1.20016292e-03 4.96377796e-01 7.55589843e-01 2.84116268e-01 5.39032333e-02 7.11581632e-02 2.29831427e-01 -3.14224601e-01 -3.49056214e-01 -9.35946777e-02 8.58344257e-01 -1.97345931e-02 -9.66335356e-01 -4.06360537e-01 6.23118401e-01 -5.39748594e-02 3.08479548e-01 -4.18236703e-01 1.06920183e+00 2.25660726e-01 3.34952086e-01 1.66021977e-02 -3.71010870e-01 8.19933951e-01 -3.14498127e-01 6.84810817e-01 -9.30531979e-01 -2.96778411e-01 2.06032246e-02 -1.40544306e-02 -8.73461962e-01 -3.27894837e-01 -5.22716701e-01 -1.33108282e+00 -2.02618301e-01 -1.17797315e-01 -2.17849061e-01 3.56384814e-01 7.12314069e-01 3.42348456e-01 4.90277231e-01 4.54707772e-01 -1.66984379e+00 -3.02307874e-01 -7.60331571e-01 -4.85023677e-01 8.29670966e-01 2.97313541e-01 -1.30999279e+00 -2.12923408e-01 -8.16400126e-02]
[7.171933650970459, -1.325011134147644]
7a833aa3-e2d1-45d5-8242-0854d8e65cea
neural-kaleidoscopic-space-sculpting
null
null
http://openaccess.thecvf.com//content/CVPR2023/html/Ahn_Neural_Kaleidoscopic_Space_Sculpting_CVPR_2023_paper.html
http://openaccess.thecvf.com//content/CVPR2023/papers/Ahn_Neural_Kaleidoscopic_Space_Sculpting_CVPR_2023_paper.pdf
Neural Kaleidoscopic Space Sculpting
We introduce a method that recovers full-surround 3D reconstructions from a single kaleidoscopic image using a neural surface representation. Full-surround 3D reconstruction is critical for many applications, such as augmented and virtual reality. A kaleidoscope, which uses a single camera and multiple mirrors, is a convenient way of achieving full-surround coverage, as it redistributes light directions and thus captures multiple viewpoints in a single image. This enables single-shot and dynamic full-surround 3D reconstruction. However, using a kaleidoscopic image for multi-view stereo is challenging, as we need to decompose the image into multi-view images by identifying which pixel corresponds to which virtual camera, a process we call labeling. To address this challenge, pur approach avoids the need to explicitly estimate labels, but instead "sculpts" a neural surface representation through the careful use of silhouette, background, foreground, and texture information present in the kaleidoscopic image. We demonstrate the advantages of our method in a range of simulated and real experiments, on both static and dynamic scenes.
['Aswin C. Sankaranarayanan', 'Ioannis Gkioulekas', 'Michael De Zeeuw', 'Byeongjoo Ahn']
2023-01-01
null
null
null
cvpr-2023-1
['3d-reconstruction']
['computer-vision']
[ 4.5046493e-01 -3.4640020e-01 1.8918566e-01 -2.1667333e-01 -3.4058321e-01 -6.8882507e-01 3.6328146e-01 -4.1839263e-01 -1.7582750e-01 4.8745605e-01 2.2645822e-02 -1.3265924e-01 4.1549993e-01 -6.9237828e-01 -7.2272766e-01 -7.7203000e-01 8.2452136e-01 2.8382051e-01 4.0484017e-01 -3.8716570e-02 1.7368859e-01 7.8176606e-01 -1.7814496e+00 2.6588342e-01 2.4267612e-01 7.2948086e-01 7.1501243e-01 5.3565747e-01 -1.1820902e-01 5.3864586e-01 -2.5706351e-01 -2.2411266e-01 4.0912288e-01 -3.5978368e-01 -1.7587055e-01 6.5452492e-01 7.9168433e-01 -6.2182552e-01 -1.9616544e-01 1.1447071e+00 2.5308555e-01 1.0340752e-01 4.3958691e-01 -7.8556269e-01 -2.2162777e-01 -3.4788054e-01 -1.0707549e+00 8.4886469e-02 6.1625159e-01 -1.9981360e-01 5.3023291e-01 -9.5995837e-01 9.0929514e-01 1.0460080e+00 3.8830328e-01 4.4356915e-01 -1.2400280e+00 -5.4945564e-01 6.5545939e-02 -2.5724965e-01 -1.3769237e+00 -6.2331384e-01 1.1539903e+00 -4.3392032e-01 5.7136327e-01 1.8162240e-01 7.5555634e-01 7.9910952e-01 2.6081061e-01 6.2678009e-01 1.3505367e+00 -6.2512851e-01 1.8706469e-01 3.4595421e-01 -1.0520698e-01 7.4676025e-01 1.9873020e-01 1.9652185e-01 -5.2437854e-01 -4.5391724e-02 1.3237094e+00 4.8723009e-01 -7.0054978e-01 -1.0004940e+00 -1.0148114e+00 2.4021754e-01 8.8922240e-02 1.2377457e-01 -2.0684545e-01 -8.9814506e-02 -1.7026065e-01 9.3326420e-02 6.5114343e-01 1.7772697e-01 -4.5432597e-02 2.1040243e-01 -7.8709060e-01 -2.2219254e-01 5.0661016e-01 7.9024458e-01 9.7720271e-01 5.4964293e-02 5.7802409e-01 8.0085105e-01 2.1324770e-01 8.6872339e-01 -6.8953402e-02 -1.3580770e+00 1.5991659e-01 5.4670233e-01 3.5523996e-01 -8.5674876e-01 -2.0478906e-01 -2.3328133e-01 -6.1817175e-01 7.6909596e-01 3.1446743e-01 2.3532385e-02 -8.6080849e-01 1.2686441e+00 6.9352615e-01 5.8513653e-02 -7.1060829e-02 1.1071993e+00 4.7098136e-01 4.7396421e-01 -7.6357156e-01 -4.3996325e-01 1.1575308e+00 -7.0341557e-01 -6.7184937e-01 -2.9164258e-01 1.7501451e-01 -9.0027207e-01 9.4632119e-01 5.9857064e-01 -1.1908894e+00 -2.1847560e-01 -1.0135850e+00 -1.6845755e-01 7.0001200e-02 1.4743134e-01 2.3181237e-01 4.9238834e-01 -8.6298323e-01 2.2521882e-01 -8.5853279e-01 -2.9320392e-01 -7.6163612e-02 7.8745715e-02 -7.2487849e-01 -3.9038855e-01 -3.7359768e-01 8.1565064e-01 -1.3881946e-01 -2.0240974e-01 -5.1733601e-01 -3.6380598e-01 -8.7316960e-01 3.3218876e-02 6.6747844e-01 -7.1881711e-01 9.2750645e-01 -8.1545979e-01 -1.5584116e+00 1.2458235e+00 -5.0913370e-01 2.6281148e-01 1.9839057e-01 -1.6501810e-01 -2.0145683e-01 4.1878948e-01 -1.8791364e-01 1.1366096e-01 9.4425344e-01 -1.7069416e+00 -3.2321903e-01 -7.8958976e-01 2.1103290e-01 5.6560689e-01 2.7482877e-02 3.7735593e-02 -6.6991389e-01 -2.9185247e-01 6.3031077e-01 -8.6520022e-01 -1.1578642e-01 3.2992941e-01 -2.7306908e-01 6.1619174e-01 1.0756521e+00 -2.3925373e-01 5.3224492e-01 -2.2684472e+00 -1.6466759e-02 -8.4350467e-02 3.5389706e-01 1.2670413e-01 2.1198422e-01 2.7896473e-01 1.2164559e-01 -5.9516823e-01 -9.4158144e-04 -6.1227465e-01 -6.4159262e-01 5.9713591e-03 -1.9147003e-01 8.1190246e-01 -4.8705348e-01 2.9376951e-01 -7.4799824e-01 -3.8225248e-01 8.5886550e-01 8.0599707e-01 -4.4895697e-01 1.8001282e-01 -3.2732219e-02 5.3435946e-01 -2.8746721e-01 5.7605278e-01 9.3765187e-01 -3.9213103e-01 1.9932391e-01 -2.2382547e-01 -3.7615672e-01 -1.4504239e-01 -1.3344654e+00 1.7149487e+00 -6.0157144e-01 7.2913784e-01 6.4337766e-01 -5.4549938e-01 1.0085044e+00 1.8215564e-01 4.3407980e-01 -6.8198353e-01 1.4132304e-01 2.4228187e-01 -7.3225003e-01 -2.3149244e-01 5.1742941e-01 -4.6631607e-01 2.4629198e-01 7.3844004e-01 -2.7381361e-01 -4.8547038e-01 -1.4309120e-01 1.5399234e-01 5.8715868e-01 2.3038787e-01 3.7122932e-01 -2.6152149e-02 4.5068946e-01 -1.7985559e-01 5.2893639e-01 3.2884669e-01 1.1301550e-01 1.1416540e+00 1.0615813e-01 -5.5180103e-01 -1.0407528e+00 -1.1620511e+00 -4.6331018e-02 4.4888824e-01 7.0577341e-01 -4.8393246e-02 -5.8003551e-01 -2.5076798e-01 -1.7901532e-01 5.7769412e-01 -4.2287824e-01 2.6788792e-01 -5.3075165e-01 -6.5076172e-02 -2.8645471e-01 1.8097463e-01 4.0232331e-01 -5.0737846e-01 -1.1240396e+00 -2.2481361e-01 -3.1851831e-01 -1.3575250e+00 -5.6272680e-01 7.7131256e-02 -9.0066260e-01 -1.2736558e+00 -9.0110737e-01 -5.5335134e-01 8.5532671e-01 1.3493568e+00 1.0241991e+00 -2.1061024e-01 -2.4364690e-01 6.8370366e-01 -8.6599916e-02 6.1456494e-02 -6.5384671e-02 -5.8259642e-01 -2.8845070e-02 2.5359884e-01 6.1087150e-02 -7.7052367e-01 -6.4594841e-01 4.8507732e-01 -7.4270207e-01 5.7214314e-01 3.0179742e-01 6.4154643e-01 8.9364457e-01 -3.4663248e-01 -1.9665383e-01 -9.9165756e-01 -6.2095597e-02 1.7950607e-02 -1.0592521e+00 2.6106304e-01 -1.6900019e-01 -2.7913061e-01 5.9252959e-01 -1.6551377e-01 -1.3948772e+00 3.7483868e-01 2.4467164e-01 -1.0703908e+00 -2.0676318e-01 -9.3625367e-02 -8.4997855e-02 -2.4497804e-01 6.1872274e-01 4.0863886e-01 2.9404631e-01 -6.1502564e-01 2.1096040e-01 6.8804783e-01 5.2959192e-01 -2.0201799e-01 3.7858564e-01 1.2622921e+00 6.2717483e-03 -1.3341793e+00 -9.2507273e-01 -6.7182380e-01 -7.6419771e-01 -4.7998488e-01 7.7484745e-01 -9.4483882e-01 -8.4531498e-01 4.0669367e-01 -1.2087420e+00 -2.2083052e-01 -2.4817991e-01 6.6496855e-01 -6.3347417e-01 5.0962991e-01 -4.3853045e-01 -9.0326142e-01 3.4200316e-03 -1.1396579e+00 1.4633040e+00 3.4431151e-01 1.1391311e-01 -9.1228580e-01 1.0646641e-01 4.9158019e-01 9.4229534e-02 1.8833147e-01 6.5041977e-01 3.6780810e-01 -9.0548098e-01 -8.9055948e-02 -2.7570993e-01 3.3206832e-01 2.6329890e-01 -1.6215238e-01 -1.3153454e+00 -1.2586148e-01 6.8960613e-01 -1.6435657e-01 6.0945398e-01 6.1584014e-01 9.0570670e-01 3.4081417e-01 -3.4868878e-01 1.0035447e+00 1.6623175e+00 3.2667395e-01 4.7289857e-01 5.0103594e-02 7.6030421e-01 5.6475157e-01 3.6335295e-01 2.5978097e-01 3.3792555e-01 8.8688850e-01 5.5431181e-01 -3.4436744e-01 -3.7614360e-01 -2.5527185e-01 1.4032513e-01 6.5013123e-01 -2.7429011e-02 -2.7854028e-01 -5.8611339e-01 3.9699900e-01 -1.3541435e+00 -8.8939041e-01 -1.5748863e-01 2.3907721e+00 2.3933843e-01 -4.3567270e-01 -2.6324415e-01 -7.3788702e-02 8.0008256e-01 2.8767741e-01 -7.0593613e-01 1.2825312e-02 -3.2883129e-01 -2.3009695e-01 4.2302552e-01 6.7635787e-01 -7.4496353e-01 8.7638098e-01 6.5751519e+00 3.7261713e-01 -1.4529246e+00 4.5735195e-02 2.8950661e-01 -3.0307931e-01 -5.4611593e-01 5.6819316e-02 -6.6305840e-01 3.0274168e-01 4.9601115e-02 1.2068525e-01 5.5989617e-01 6.4990598e-01 1.7261019e-01 -7.2967863e-01 -9.1661775e-01 1.3532747e+00 5.8317727e-01 -1.3451886e+00 -1.0788171e-01 2.2482631e-01 9.4633186e-01 1.6340560e-01 -9.2127919e-02 -6.6497564e-01 2.1105081e-01 -4.5178100e-01 5.6414497e-01 3.5277861e-01 1.0458616e+00 -3.8831592e-01 1.6514134e-01 5.8796436e-01 -9.2981625e-01 1.0159766e-01 -4.0326256e-01 -6.9620565e-02 5.4475462e-01 7.2278404e-01 -3.9567566e-01 3.6686307e-01 5.3535122e-01 6.2862593e-01 4.9977377e-02 8.1049514e-01 2.2022940e-02 -2.4258082e-01 -3.1520593e-01 3.1329864e-01 -2.7857241e-01 -6.7451799e-01 6.0597426e-01 4.9667543e-01 5.1621091e-01 3.9460674e-01 1.4077531e-01 6.8023413e-01 1.8908794e-01 -2.2361524e-01 -1.0987867e+00 3.9181203e-01 1.7423016e-01 1.0805411e+00 -9.1884106e-01 -2.8410318e-01 -6.2539387e-01 1.0980101e+00 8.1026167e-02 5.9556472e-01 -3.4844378e-01 -2.2608732e-01 3.9646837e-01 5.1528198e-01 3.9311522e-01 -5.1738250e-01 -7.5492397e-02 -1.7192186e+00 7.5741015e-02 -5.0653207e-01 -2.7098253e-02 -1.4548540e+00 -7.1005440e-01 6.3431460e-01 6.5884441e-02 -1.4535534e+00 -1.0403466e-01 -5.8776385e-01 -2.4017793e-01 7.7065849e-01 -1.4966271e+00 -1.0251298e+00 -6.4634800e-01 8.7274587e-01 4.4547960e-01 2.0030707e-01 6.4638114e-01 2.2991361e-01 -1.9301660e-01 -1.4331919e-01 3.6368611e-01 -4.0401736e-01 6.7983902e-01 -8.7424690e-01 1.1659413e-01 6.8321085e-01 2.5105646e-01 4.4341940e-01 6.4292991e-01 -4.2697150e-01 -1.5347595e+00 -5.5508983e-01 5.5574995e-01 -3.2540914e-01 -1.1144739e-03 -3.7222803e-01 -7.4679905e-01 6.9796985e-01 6.1475776e-02 2.8103560e-01 4.3504366e-01 -6.3977323e-02 -1.3951732e-01 -2.0808795e-01 -1.0484676e+00 5.7857263e-01 1.0837715e+00 -8.4159040e-01 -2.9076439e-01 1.7818189e-01 4.2323634e-01 -7.4898237e-01 -2.9940838e-01 3.7912983e-02 8.1925476e-01 -1.6118371e+00 9.1518462e-01 2.9062772e-01 2.3742257e-01 -5.1126528e-01 -4.1579884e-01 -1.3300252e+00 8.8388622e-02 -5.5582321e-01 1.0734984e-01 6.5554142e-01 -4.3159943e-02 -8.1024766e-01 1.0041332e+00 4.8903662e-01 -1.4160052e-01 -5.5158103e-01 -6.5180141e-01 -5.7236451e-01 -5.4734224e-01 -1.7587915e-01 3.2890043e-01 8.2917637e-01 -3.5948598e-01 3.7632692e-01 -4.7071588e-01 1.8856120e-01 9.3542171e-01 6.6992450e-01 1.0232859e+00 -1.2220837e+00 -2.9009077e-01 1.0110340e-01 -2.1653168e-01 -1.3862535e+00 -6.2711820e-02 -5.2648687e-01 -3.8308524e-02 -1.2798038e+00 2.2917809e-01 -3.3422756e-01 2.8491819e-01 -2.3610507e-01 3.2077444e-01 4.0411648e-01 2.3564170e-01 4.0007296e-01 -2.9881158e-01 5.9710234e-01 1.4760983e+00 2.7174786e-01 -2.5590998e-01 4.2440467e-02 -4.7003549e-01 1.0752146e+00 2.5507793e-01 -3.6098310e-01 -5.8072931e-01 -8.4751135e-01 1.7525685e-01 6.4606714e-01 3.7710068e-01 -7.9569900e-01 4.0200981e-01 -2.7458945e-01 4.1670200e-01 -7.6668811e-01 1.1886635e+00 -9.4633484e-01 4.3541053e-01 1.3484524e-01 1.2084035e-01 -1.3124809e-02 -1.2273173e-01 7.5562572e-01 -1.9588643e-01 -2.2701029e-01 9.4717354e-01 -4.8806033e-01 -5.0758511e-01 8.7769106e-02 -8.2854658e-02 -1.6615357e-01 1.0894996e+00 -6.6094685e-01 -2.9158509e-01 -4.4715124e-01 -5.4956359e-01 -2.0549396e-01 1.4363940e+00 1.4658249e-03 1.0204078e+00 -1.0132431e+00 -1.0924839e-01 8.4780443e-01 -4.1628294e-03 8.0702908e-02 6.1026055e-01 6.5124917e-01 -8.1045145e-01 3.0657798e-01 -2.0042746e-01 -1.0388074e+00 -1.5143986e+00 5.2698481e-01 4.6766052e-01 1.6688170e-01 -1.2389213e+00 4.6060094e-01 9.7144300e-01 -3.7132266e-01 -1.1041807e-02 -1.6338436e-01 -8.3915982e-03 -3.3296806e-01 6.4925265e-01 1.8543831e-01 2.7890164e-02 -6.1761081e-01 -5.4463334e-02 1.1788177e+00 3.8614504e-02 -3.8551676e-01 1.2020693e+00 -5.9455836e-01 2.5947267e-02 7.2465622e-01 1.1050233e+00 3.7887228e-01 -1.5618894e+00 -3.5808665e-01 -8.8482511e-01 -1.1496663e+00 3.2568410e-01 -1.4546262e-01 -1.0750458e+00 1.2333250e+00 2.6250944e-01 -3.0785350e-02 1.2051893e+00 -1.2496617e-01 6.4031160e-01 3.3445469e-01 7.6898748e-01 -8.9425516e-01 1.2599717e-01 3.4654120e-01 6.1459637e-01 -1.0599781e+00 1.8539262e-01 -5.8833027e-01 -6.8555349e-01 1.0843630e+00 5.5223042e-01 -1.5479377e-01 7.5727475e-01 3.3834144e-01 1.6481765e-01 -4.5783550e-01 -6.7865497e-01 -9.1616459e-02 -9.3060240e-02 4.4035882e-01 4.7020130e-02 -7.9418644e-02 3.9649424e-01 -2.5226864e-01 1.7839053e-01 -6.7354992e-02 9.1373050e-01 8.2215285e-01 -4.3066451e-01 -8.1858653e-01 -4.8085204e-01 1.2122671e-01 -1.8877441e-01 2.7592224e-01 -2.7789602e-01 6.7223471e-01 -2.0378561e-01 7.1736485e-01 3.6426112e-01 -2.2642805e-01 2.8944463e-01 -3.0793175e-01 8.4388232e-01 -8.0821985e-01 -3.3163331e-02 4.7966132e-01 -1.0190251e-01 -5.1332408e-01 -6.2780523e-01 -5.9244812e-01 -8.5307914e-01 -1.3450843e-01 -5.1216638e-01 -1.7886858e-01 9.3179387e-01 7.9543895e-01 4.0519589e-01 3.1526180e-03 8.8403291e-01 -1.0983232e+00 -6.3939430e-02 -2.8248599e-01 -1.0556200e+00 2.6666012e-01 6.8526882e-01 -8.8899827e-01 -5.1475835e-01 1.7029820e-02]
[9.427943229675293, -2.7291018962860107]
342b373d-0152-455c-afce-acbf87141c07
motifretro-exploring-the-combinability
2305.15153
null
https://arxiv.org/abs/2305.15153v1
https://arxiv.org/pdf/2305.15153v1.pdf
MotifRetro: Exploring the Combinability-Consistency Trade-offs in retrosynthesis via Dynamic Motif Editing
Is there a unified framework for graph-based retrosynthesis prediction? Through analysis of full-, semi-, and non-template retrosynthesis methods, we discovered that they strive to strike an optimal balance between combinability and consistency: \textit{Should atoms be combined as motifs to simplify the molecular editing process, or should motifs be broken down into atoms to reduce the vocabulary and improve predictive consistency?} Recent works have studied several specific cases, while none of them explores different combinability-consistency trade-offs. Therefore, we propose MotifRetro, a dynamic motif editing framework for retrosynthesis prediction that can explore the entire trade-off space and unify graph-based models. MotifRetro comprises two components: RetroBPE, which controls the combinability-consistency trade-off, and a motif editing model, where we introduce a novel LG-EGAT module to dynamiclly add motifs to the molecule. We conduct extensive experiments on USPTO-50K to explore how the trade-off affects the model performance and finally achieve state-of-the-art performance.
['Stan Z. Li', 'Cheng Tan', 'Xingran Chen', 'Zhangyang Gao']
2023-05-20
null
null
null
null
['retrosynthesis']
['medical']
[ 2.35663876e-01 -1.42269686e-01 -6.13052964e-01 -7.13948533e-02 -1.69475988e-01 -1.07880557e+00 5.12400806e-01 2.01885968e-01 -1.35627002e-01 6.05338931e-01 2.87547242e-02 -7.62886345e-01 -2.11235182e-03 -8.70729685e-01 -7.12577045e-01 -6.40812695e-01 1.40015855e-01 2.77971834e-01 6.61083817e-01 -3.89059097e-01 5.52705526e-01 4.44422275e-01 -1.13760316e+00 3.28634918e-01 1.06739068e+00 4.67552423e-01 3.12114000e-01 2.51355112e-01 -1.02889612e-01 3.43315393e-01 -2.53403425e-01 -5.33635855e-01 3.94935817e-01 -6.32328331e-01 -5.64050555e-01 -3.81798685e-01 6.68268055e-02 1.57778442e-01 -2.18614921e-01 6.74508810e-01 5.55545688e-01 -5.31700961e-02 8.59488487e-01 -1.11914468e+00 -2.67897666e-01 7.52910018e-01 -5.89590788e-01 7.00351074e-02 2.08272904e-01 2.14670405e-01 1.12052882e+00 -6.78327501e-01 7.22963214e-01 1.03945124e+00 6.17410421e-01 3.48201722e-01 -1.38772213e+00 -9.08088386e-01 2.21550807e-01 7.03142658e-02 -1.35986805e+00 -4.29693401e-01 5.69359779e-01 -2.26895452e-01 1.18236732e+00 3.86017531e-01 8.55787635e-01 8.86331201e-01 7.59760082e-01 6.80129230e-02 7.42654204e-01 -2.96255141e-01 2.63206482e-01 -3.43744218e-01 -1.75128490e-01 9.44285214e-01 3.72204304e-01 -2.09396798e-03 -6.85198009e-01 -5.18141806e-01 8.22683275e-01 -1.70490704e-02 -8.01637992e-02 -4.68916446e-01 -1.22062778e+00 7.68728971e-01 2.20203713e-01 1.58165783e-01 5.84486537e-02 2.43476719e-01 4.05406266e-01 1.51674405e-01 -4.52702940e-02 9.43464577e-01 -4.63171870e-01 5.86663373e-02 -9.24311280e-01 2.33898446e-01 7.99646139e-01 1.18366337e+00 7.72357166e-01 -8.04422572e-02 -1.41130656e-01 7.16654301e-01 1.24350689e-01 -1.98838487e-01 2.78326750e-01 -6.23587728e-01 3.85997146e-01 6.53581500e-01 -1.16569065e-01 -7.54464865e-01 -5.63662767e-01 -2.39637762e-01 -5.62381387e-01 -5.24621904e-01 3.42493981e-01 -1.32603675e-01 -9.59223092e-01 1.83089042e+00 5.18353224e-01 2.89570466e-02 -5.02011120e-01 6.69907749e-01 6.79888666e-01 8.45331669e-01 3.50110561e-01 -4.70273376e-01 1.24976087e+00 -1.16865897e+00 -3.61987084e-01 1.54304698e-01 8.75862896e-01 -9.27942753e-01 8.56388688e-01 4.51106340e-01 -1.11953855e+00 -3.75068516e-01 -1.25613391e+00 -1.27264783e-01 -4.44954187e-01 1.79536231e-02 8.15213859e-01 7.11391509e-01 -6.87874615e-01 1.00627410e+00 -6.83989942e-01 -3.08098465e-01 -2.00598270e-01 6.21875346e-01 -2.72237003e-01 2.96064228e-01 -1.02833629e+00 7.82177448e-01 7.98706949e-01 -4.15137969e-02 -9.62553263e-01 -8.21856856e-01 -4.15394396e-01 1.24114618e-01 9.35974538e-01 -1.05701613e+00 9.62777436e-01 -7.11209357e-01 -1.69050241e+00 3.14983785e-01 1.73654743e-02 -6.18053973e-02 2.46916801e-01 4.07465905e-01 -2.57588238e-01 -3.64108354e-01 -2.38035336e-01 6.53030694e-01 7.37367392e-01 -9.05748248e-01 -2.90263653e-01 -9.85989571e-02 2.24101424e-01 7.24623278e-02 -1.41715989e-01 -1.21870771e-01 -7.07654059e-01 -1.05477917e+00 1.58749878e-01 -1.42074156e+00 -3.54369134e-01 -2.50975102e-01 -5.49147964e-01 -4.52959314e-02 3.84766817e-01 -2.78940350e-01 1.66534650e+00 -1.83144534e+00 5.43885708e-01 5.28677464e-01 9.22545642e-02 1.73550755e-01 -4.68073189e-01 1.09522855e+00 -3.74546468e-01 5.23291767e-01 5.78670949e-02 2.92239994e-01 -1.02916218e-01 7.85528719e-02 -1.91363037e-01 1.68146864e-01 1.15578361e-02 7.63266206e-01 -8.43825877e-01 -3.61575514e-01 -7.29294717e-02 1.62746608e-01 -1.06598389e+00 -4.16440703e-02 -8.34696889e-01 2.74247289e-01 -5.32943428e-01 6.55352712e-01 5.49520373e-01 -2.42875814e-01 1.00743663e+00 -4.02228951e-01 -4.49455768e-01 3.97715718e-01 -1.15267730e+00 1.83172524e+00 -1.16821818e-01 -1.51848942e-01 -3.18295926e-01 -4.72688913e-01 8.85666847e-01 9.75004211e-02 3.78033549e-01 -5.10277867e-01 4.02555168e-02 1.94008961e-01 2.17373192e-01 1.45420790e-01 7.57604420e-01 -1.55106828e-01 7.03129843e-02 3.63250375e-01 -1.49340197e-01 -1.48934796e-01 5.39596736e-01 2.39868298e-01 1.08690238e+00 3.92742664e-01 5.65182984e-01 -5.86273491e-01 4.31298375e-01 1.09234169e-01 8.40781450e-01 7.00065732e-01 3.32469314e-01 4.48777914e-01 8.76562715e-01 -4.79005635e-01 -1.12068522e+00 -7.12128699e-01 2.85303205e-01 1.37099457e+00 1.62699953e-01 -1.35108066e+00 -7.92709708e-01 -6.91906154e-01 -2.04893962e-01 5.32907367e-01 -3.94800156e-01 -4.26889151e-01 -4.42160219e-01 -9.41358387e-01 6.07903838e-01 3.22902769e-01 1.73329309e-01 -4.45851266e-01 -4.68399152e-02 4.06502604e-01 -7.31343701e-02 -5.98109603e-01 -1.10849357e+00 5.07010698e-01 -7.73869574e-01 -1.11530054e+00 -2.60925591e-01 -6.20227873e-01 5.39008796e-01 5.36200166e-01 9.42467868e-01 3.59735191e-01 -8.26170817e-02 -1.55199960e-01 -4.43035364e-01 -2.97508240e-01 -4.88260031e-01 4.90183324e-01 -5.58153503e-02 -5.40593624e-01 -2.82381147e-01 -8.43880832e-01 -7.94632375e-01 7.92792022e-01 -1.07933021e+00 4.08444434e-01 4.98869240e-01 8.54982018e-01 8.81300747e-01 1.06745008e-02 3.56388986e-01 -7.37138271e-01 3.94245297e-01 -2.82976925e-01 -7.50232816e-01 6.96252465e-01 -7.84594774e-01 4.27415609e-01 8.14660132e-01 -5.82212985e-01 -5.54535031e-01 2.42285624e-01 -1.83420882e-01 -1.86104268e-01 5.73719025e-01 6.21525824e-01 -6.12781048e-01 -2.06521586e-01 2.81375647e-01 2.05060959e-01 -2.02942654e-01 -4.03236538e-01 5.75621188e-01 8.88957083e-02 -2.57954951e-02 -9.67224658e-01 5.67307115e-01 -1.66661646e-02 3.52984518e-01 -7.14412630e-01 -3.22715849e-01 -2.11959913e-01 -4.97788876e-01 1.74937218e-01 6.53572142e-01 -7.66062260e-01 -8.66066992e-01 7.20264912e-02 -9.73204195e-01 -3.47980052e-01 2.46413469e-01 2.96301935e-02 -5.92744350e-01 5.55235684e-01 -5.13937354e-01 -1.75659463e-01 -3.30702871e-01 -1.60770297e+00 8.93756092e-01 -8.19995478e-02 -2.99906373e-01 -5.08166313e-01 3.77319962e-01 2.14466631e-01 1.35666355e-01 5.84276691e-02 1.67350328e+00 -6.87823057e-01 -7.06794918e-01 5.81836641e-01 -5.74967116e-02 -2.97197253e-01 1.98942915e-01 2.44503587e-01 -3.13768983e-01 -3.62345517e-01 -7.69621015e-01 6.58372045e-02 8.62984657e-01 4.33290005e-02 1.48035169e+00 -4.02299672e-01 -6.06687069e-01 6.62692964e-01 1.18932939e+00 4.24979538e-01 9.08316195e-01 2.41418734e-01 7.75086880e-01 1.53008878e-01 6.56166792e-01 4.49321717e-01 1.32362127e-01 1.08911693e+00 4.48005825e-01 3.06498051e-01 2.22858354e-01 -6.60794735e-01 4.49921012e-01 9.28199947e-01 -3.33620876e-01 -6.42567039e-01 -7.76194096e-01 -2.08896622e-01 -1.85097980e+00 -8.21007729e-01 1.49104550e-01 2.23783398e+00 1.00147510e+00 1.17771305e-01 5.71980000e-01 5.26151694e-02 4.57957596e-01 2.85216182e-01 -3.55963767e-01 -5.59796512e-01 3.08853626e-01 2.05096394e-01 7.21954286e-01 3.39748353e-01 -7.48313844e-01 1.31174564e+00 6.93455362e+00 1.43385458e+00 -1.43588662e+00 -1.29055426e-01 4.79566097e-01 -1.03748336e-01 -6.00594103e-01 6.20243549e-01 -8.89738142e-01 4.74744737e-01 8.69301558e-01 -2.08235998e-02 5.43311298e-01 6.72637522e-01 3.21000189e-01 2.15814725e-01 -1.22978067e+00 6.38627708e-01 -4.02073801e-01 -1.87347937e+00 6.91696584e-01 2.39330813e-01 3.97602797e-01 -2.98586100e-01 -2.49877125e-01 2.36180320e-01 2.26211846e-01 -8.76036584e-01 8.43139708e-01 5.17310619e-01 4.67670858e-01 -9.68649864e-01 -7.62526458e-03 1.88653037e-01 -1.47722650e+00 -7.71331042e-02 -2.66586602e-01 3.33132446e-01 -8.37189779e-02 3.01705182e-01 -1.00158060e+00 9.75643158e-01 2.20727384e-01 5.65705359e-01 -4.97727066e-01 8.08787704e-01 -3.85704963e-03 3.05218875e-01 -1.42924562e-01 -1.51026875e-01 2.77083486e-01 -6.20468140e-01 3.66106182e-01 1.13396442e+00 2.84084558e-01 2.20216662e-01 4.12385225e-01 7.69554019e-01 -1.42557979e-01 2.31795207e-01 -2.16922641e-01 -4.34377372e-01 6.83127224e-01 9.69520032e-01 -1.05786896e+00 1.28520513e-03 -2.15081781e-01 7.79142261e-01 2.44869098e-01 1.67557493e-01 -1.21860588e+00 -1.74994573e-01 5.13806224e-01 5.32831967e-01 5.43089867e-01 -5.06524682e-01 9.76227522e-02 -9.64869738e-01 -4.18512553e-01 -1.32984054e+00 3.97345245e-01 -5.96919477e-01 -8.64136577e-01 1.87879384e-01 -8.76136273e-02 -8.42117846e-01 3.97583425e-01 -4.65293616e-01 -6.88122272e-01 3.38113815e-01 -1.08993697e+00 -1.11288679e+00 1.62859678e-01 1.71910092e-01 4.17180508e-01 1.81390226e-01 5.46238482e-01 2.61716038e-01 -9.59653974e-01 7.58345366e-01 2.01667055e-01 -5.21134675e-01 8.58700275e-01 -8.29342365e-01 3.75470042e-01 6.25236094e-01 -4.77409177e-02 1.19584680e+00 7.01793849e-01 -8.43311250e-01 -2.03674197e+00 -1.22971845e+00 4.66222286e-01 -4.73450601e-01 6.37775838e-01 -5.02051055e-01 -5.57006598e-01 4.44711655e-01 -1.64776132e-01 -3.99496108e-01 8.91731381e-01 7.58856460e-02 -6.31853163e-01 -9.68488008e-02 -7.38422275e-01 9.69020784e-01 1.39419663e+00 -1.86803430e-01 -2.28739604e-01 1.83706567e-01 1.05311394e+00 -2.77130187e-01 -1.12203407e+00 4.88960087e-01 7.93554246e-01 -8.62352431e-01 1.13398802e+00 -6.07780814e-01 2.78282940e-01 -4.60935980e-01 -9.20749456e-02 -1.04475713e+00 -6.93658948e-01 -1.05360639e+00 -3.03398333e-02 1.07663822e+00 7.33782947e-01 -4.51876283e-01 5.89038491e-01 1.98311344e-01 -3.55192810e-01 -9.30863798e-01 -7.55197883e-01 -9.24935997e-01 1.88019291e-01 -1.43112354e-02 1.05311763e+00 8.71279120e-01 2.55581319e-01 5.04916072e-01 -5.78319728e-01 -9.29693654e-02 -1.19911820e-01 1.79793254e-01 9.70912755e-01 -7.72946835e-01 -5.67625463e-01 -4.92317438e-01 -3.06503233e-02 -1.11045849e+00 -1.19244941e-01 -1.04547131e+00 -2.93457389e-01 -1.13258862e+00 5.23543417e-01 -5.03997862e-01 -1.82875991e-01 6.69321239e-01 -1.13141701e-01 -1.29481316e-01 3.37392151e-01 2.24783555e-01 -7.25912035e-01 5.97324550e-01 1.36584949e+00 -9.03560221e-02 -6.27526045e-01 -5.08057952e-01 -8.54193270e-01 3.26616943e-01 5.28975844e-01 -5.78955352e-01 -5.86095095e-01 -1.34627298e-01 7.22718537e-01 2.64902294e-01 -1.18078716e-01 -5.80564380e-01 1.02520555e-01 -6.95123732e-01 4.55146693e-02 -7.70593047e-01 2.24070147e-01 -6.89675093e-01 8.30193043e-01 4.70368445e-01 -1.13791138e-01 2.84238338e-01 2.65750706e-01 7.54404545e-01 3.43421042e-01 -5.04257791e-02 5.17086029e-01 -2.26100817e-01 -3.77789140e-01 6.49990022e-01 -6.38413310e-01 -3.18666816e-01 9.63745415e-01 -2.70844132e-01 -3.75497669e-01 1.90907940e-01 -5.42958319e-01 9.19746086e-02 9.47042704e-01 4.37700927e-01 8.78892690e-02 -1.13804662e+00 -4.99495417e-02 -5.29486239e-02 2.25515768e-01 -3.52750272e-01 -9.89512503e-02 7.05411434e-01 -6.74601734e-01 5.22760570e-01 6.26801234e-03 -3.30828428e-01 -1.21051848e+00 8.14671576e-01 2.55970061e-01 -5.12399077e-01 2.95962077e-02 4.28350419e-01 2.03392655e-01 -6.37845516e-01 -1.80088580e-02 -5.06601274e-01 1.60282880e-01 6.40320703e-02 1.59844175e-01 3.21298391e-01 2.45956659e-01 -2.37511754e-01 -4.76421952e-01 5.04868686e-01 -3.57600123e-01 4.07497197e-01 1.16357911e+00 2.04351649e-01 -3.30925405e-01 -5.08950800e-02 6.80017650e-01 6.55542389e-02 -9.29024696e-01 2.69987077e-01 -1.85212903e-02 -3.26680213e-01 -2.20292106e-01 -6.74691200e-01 -6.80697620e-01 2.81940699e-01 1.50002018e-01 -5.10829464e-02 9.20226336e-01 -3.55314344e-01 6.93481922e-01 6.06900871e-01 5.10089815e-01 -1.06154788e+00 4.07487243e-01 6.29193962e-01 7.28896081e-01 -6.11308873e-01 3.79443675e-01 -8.03202152e-01 -3.86289835e-01 1.09899747e+00 5.23712456e-01 1.15760453e-01 5.32448709e-01 6.64823689e-04 -8.27130914e-01 -9.33975577e-02 -9.47302818e-01 1.82052463e-01 1.36720255e-01 1.51773036e-01 5.92517257e-01 -3.03354114e-02 -7.78170288e-01 7.12484717e-01 -2.00710505e-01 -1.08971983e-01 2.18906268e-01 1.14990294e+00 -5.55129051e-01 -1.87618959e+00 -2.56958723e-01 2.17527270e-01 -2.43786484e-01 -3.33257973e-01 -8.25691223e-01 8.58968675e-01 3.76314759e-01 7.37082958e-01 -4.31104839e-01 -6.05027199e-01 3.21602225e-01 6.21890575e-02 7.25850523e-01 -6.77876711e-01 -9.29945529e-01 3.89371067e-01 3.33712071e-01 -5.35019934e-01 -1.52603105e-01 -2.17421323e-01 -1.07093072e+00 -7.40572453e-01 -7.80254960e-01 5.08708477e-01 4.54400241e-01 7.72823930e-01 6.88807964e-01 5.37642479e-01 6.81733310e-01 -6.84390485e-01 -2.85041869e-01 -3.99959654e-01 -3.18820387e-01 -3.64537507e-01 -2.28393197e-01 -6.26383126e-01 3.28024253e-02 -2.20307067e-01]
[4.495567321777344, 6.109785079956055]
6e582e6c-4634-493e-91bd-9562abba163a
using-large-pre-trained-language-model-to
2212.01217
null
https://arxiv.org/abs/2212.01217v1
https://arxiv.org/pdf/2212.01217v1.pdf
Using Large Pre-Trained Language Model to Assist FDA in Premarket Medical Device
This paper proposes a possible method using natural language processing that might assist in the FDA medical device marketing process. Actual device descriptions are taken and matched with the device description in FDA Title 21 of CFR to determine their corresponding device type. Both pre-trained word embeddings such as FastText and large pre-trained sentence embedding models such as sentence transformers are evaluated on their accuracy in characterizing a piece of device description. An experiment is also done to test whether these models can identify the devices wrongly classified in the FDA database. The result shows that sentence transformer with T5 and MPNet and GPT-3 semantic search embedding show high accuracy in identifying the correct classification by narrowing down the correct label to be contained in the first 15 most likely results, as compared to 2585 types of device descriptions that must be manually searched through. On the other hand, all methods demonstrate high accuracy in identifying completely incorrectly labeled devices, but all fail to identify false device classifications that are wrong but closely related to the true label.
['Zongzhe Xu']
2022-11-03
null
null
null
null
['marketing']
['miscellaneous']
[ 4.92977738e-01 4.05275464e-01 -3.99047405e-01 -4.68977362e-01 -9.69144285e-01 -9.17575717e-01 3.21284592e-01 9.71193671e-01 -4.04110521e-01 4.41783726e-01 3.95678818e-01 -8.60909820e-01 -3.02750319e-01 -6.11685336e-01 -5.13840914e-01 -1.00465693e-01 2.69029409e-01 6.50864840e-01 -2.25644067e-01 2.95197695e-01 2.03098461e-01 3.97712260e-01 -9.88461852e-01 6.24975741e-01 5.50840497e-01 1.08213770e+00 6.69128597e-02 4.61340874e-01 -4.89606500e-01 5.98900199e-01 -7.77561843e-01 -6.83994114e-01 1.65133566e-01 -8.30754563e-02 -9.10724461e-01 -1.13975123e-01 4.35989499e-01 -3.42749864e-01 5.56888878e-02 1.37317252e+00 6.39302909e-01 -5.00121117e-01 8.69109571e-01 -5.81327796e-01 -8.14856648e-01 6.83135092e-01 7.93779269e-02 2.11113065e-01 1.03196537e+00 8.24390575e-02 1.00281870e+00 -1.01631200e+00 7.06260502e-01 1.26267171e+00 7.91401327e-01 6.40750527e-01 -1.09052765e+00 -7.72271514e-01 -2.79572725e-01 -1.98206007e-01 -1.42842710e+00 -2.50334114e-01 4.28515255e-01 -8.14583719e-01 1.36592174e+00 2.34250516e-01 2.47155488e-01 1.20465887e+00 6.15673780e-01 1.57166943e-01 6.77450180e-01 -3.35726440e-01 3.71618897e-01 7.36197293e-01 5.83133578e-01 6.86534107e-01 8.43046606e-01 3.21356095e-02 5.57772815e-03 -6.82888925e-01 4.62697953e-01 5.63460514e-02 -2.51915127e-01 2.39237882e-02 -9.36355293e-01 8.90702784e-01 2.14212015e-01 6.82374120e-01 -6.26004994e-01 -1.65698931e-01 6.52132392e-01 1.57991890e-02 3.71214390e-01 1.21390605e+00 -8.75517070e-01 6.21321648e-02 -7.84669340e-01 1.90077648e-02 7.19455957e-01 1.15727699e+00 1.91560909e-01 -1.80484816e-01 -3.32059890e-01 5.75218260e-01 2.42332056e-01 3.37613016e-01 7.00897515e-01 -4.85393703e-01 4.40020531e-01 8.77763808e-01 2.88706601e-01 -8.76425803e-01 -4.04226691e-01 -5.32089710e-01 -2.49881744e-01 -4.92714047e-02 5.14836833e-02 -1.96335409e-02 -1.27374899e+00 1.08056986e+00 -1.75142944e-01 -1.84140921e-01 3.02343875e-01 4.65577990e-01 1.03351545e+00 4.96622980e-01 7.41223931e-01 -3.15468274e-02 1.85548079e+00 -3.41433167e-01 -1.01471758e+00 -6.13298751e-02 9.48786199e-01 -1.08399010e+00 8.61556411e-01 3.03394794e-01 -7.73215055e-01 -8.04662108e-01 -1.24530721e+00 1.71260819e-01 -7.97860563e-01 2.71335930e-01 4.49148953e-01 8.72213304e-01 -5.24622381e-01 8.38048756e-01 -3.72423142e-01 -4.05114740e-01 5.75991273e-01 5.99416912e-01 -3.31929117e-01 -1.35386899e-01 -1.34235001e+00 1.25726223e+00 5.06864429e-01 -2.30202571e-01 -6.89515531e-01 -1.05406857e+00 -1.32094085e+00 1.94035307e-01 -1.34750620e-01 -7.07337558e-01 1.18442810e+00 -6.68067694e-01 -7.91544676e-01 9.29388583e-01 -1.40455127e-01 -5.81923723e-01 1.24028802e-01 -4.01593186e-03 -1.08348298e+00 8.04127306e-02 4.40315694e-01 7.41199255e-01 5.51814318e-01 -7.91056991e-01 -6.40544116e-01 -3.61011326e-01 2.36312486e-02 -2.46865377e-01 -3.25954050e-01 5.24273850e-02 2.52985179e-01 -5.63203394e-01 -1.04404278e-01 -6.48964763e-01 -2.18028262e-01 -2.33038917e-01 -6.86588883e-01 -5.88017404e-01 2.92746097e-01 -7.28902519e-01 1.54793167e+00 -2.14278316e+00 -7.95099676e-01 2.35200644e-01 3.80733192e-01 6.05860472e-01 -1.42415464e-01 4.37022209e-01 -6.08078182e-01 8.60005915e-01 1.88584358e-01 7.12554753e-02 -4.43369076e-02 9.96060204e-03 -4.28406417e-01 2.56272823e-01 5.43460608e-01 9.14629877e-01 -1.07749259e+00 -2.73897260e-01 3.12710255e-01 2.37639368e-01 -1.41864538e-01 -1.99641079e-01 -7.23172203e-02 7.81332999e-02 -7.57222414e-01 6.72932863e-01 6.49787426e-01 -3.77953589e-01 1.17473863e-01 -6.39828503e-01 2.13171288e-01 5.41225493e-01 -5.64018667e-01 1.48979712e+00 -3.05832595e-01 2.76228696e-01 -7.85310030e-01 -4.30662781e-01 6.48089886e-01 6.93121910e-01 5.12864053e-01 -6.62877858e-01 3.47107679e-01 4.16592509e-01 5.55531904e-02 -9.21575725e-01 4.90943909e-01 -3.33599210e-01 -1.87408566e-01 6.40592277e-02 1.77604795e-01 4.96334434e-02 -2.20956281e-01 -1.18123554e-01 1.29689074e+00 -3.57921422e-01 3.14412177e-01 -2.05014706e-01 4.92041409e-01 3.43563974e-01 4.22553420e-01 8.05360615e-01 -2.28206933e-01 4.73188192e-01 2.20178202e-01 -6.37286186e-01 -1.12807822e+00 -1.03287125e+00 -4.83485371e-01 5.78047223e-02 -7.03892671e-03 -6.21469796e-01 -5.63656628e-01 -1.00693989e+00 1.73857674e-01 1.12011874e+00 -6.55270278e-01 -3.53134573e-01 1.15571124e-02 -2.60984063e-01 6.27886713e-01 5.43259144e-01 -4.56785969e-02 -9.47054863e-01 -5.02790809e-01 5.33900261e-01 3.30372721e-01 -1.11554956e+00 -5.22325516e-01 2.55294263e-01 -8.17154109e-01 -1.51570237e+00 -6.08344734e-01 -1.11453962e+00 8.38985801e-01 -5.76519132e-01 8.36596847e-01 -3.63737583e-01 -3.73119801e-01 2.30691358e-01 -3.88549566e-01 -6.41103446e-01 -7.49033093e-01 -2.55243778e-01 8.54230858e-03 -4.13434088e-01 1.03542721e+00 3.84567797e-01 -6.25168681e-01 -4.23408821e-02 -9.55136895e-01 -6.34737194e-01 4.23514873e-01 6.48028672e-01 5.05245388e-01 2.16522828e-01 6.64676309e-01 -1.25875092e+00 1.15816331e+00 -3.48880082e-01 -2.12369636e-01 5.15949070e-01 -1.13366652e+00 2.92530745e-01 5.13027012e-01 -6.38091087e-01 -4.11164701e-01 3.39264899e-01 -6.09902382e-01 -3.70428950e-01 -2.25588813e-01 5.04364014e-01 2.03009844e-01 6.70864359e-02 8.06289494e-01 -8.96048993e-02 -1.07725471e-01 -5.13508320e-01 6.94338754e-02 8.52372527e-01 5.13010025e-02 8.00874233e-02 3.52461815e-01 -3.42121348e-02 -4.42095965e-01 -3.47250462e-01 -7.11677372e-01 -6.78358495e-01 8.09419528e-03 3.13204437e-01 1.11099148e+00 -6.62739217e-01 -5.97640812e-01 -2.45823592e-01 -1.51649070e+00 6.20450556e-01 -5.15174270e-01 7.99838245e-01 -1.32471368e-01 3.55953932e-01 -5.22557914e-01 -6.30999506e-01 -5.24562478e-01 -1.28887606e+00 1.20857692e+00 -2.49283060e-01 -1.09759986e+00 -9.45844769e-01 -3.27493727e-01 3.25554051e-02 2.39572048e-01 -6.95568975e-03 1.62342715e+00 -1.38265467e+00 1.74602181e-01 -8.45643878e-01 -2.48180330e-02 5.55028737e-01 7.17965603e-01 -3.04811597e-01 -9.84785676e-01 -1.15304358e-01 2.13345483e-01 3.11438143e-01 5.79819977e-01 6.98026896e-01 9.94124293e-01 -2.67170936e-01 -7.44324267e-01 -8.93786773e-02 1.61437249e+00 9.73611414e-01 6.21986091e-01 -5.05787022e-02 3.58895957e-01 5.07902980e-01 5.59470177e-01 7.15706870e-02 -2.06279740e-01 4.69067991e-01 1.72964960e-01 -1.37637794e-01 -1.63683176e-01 -5.75838327e-01 1.47347689e-01 3.71849597e-01 9.10787761e-01 -3.90196353e-01 -7.93489337e-01 5.27767241e-01 -1.25532818e+00 -6.55773818e-01 -1.45311370e-01 2.17525840e+00 6.97978854e-01 4.91318047e-01 -2.12288350e-01 2.16919661e-01 7.98121154e-01 -3.21511537e-01 -4.84500557e-01 -8.79053116e-01 8.53173621e-03 6.66935682e-01 9.40227628e-01 4.80168492e-01 -1.05687642e+00 7.37482965e-01 7.24039984e+00 8.82361531e-01 -9.09633815e-01 1.32876918e-01 6.72676742e-01 4.32755381e-01 -6.67467475e-01 -3.34519595e-01 -8.25628400e-01 6.13238513e-01 1.14558172e+00 -1.23190135e-01 -3.58183026e-01 6.98275864e-01 2.37412363e-01 2.13349819e-01 -1.39924848e+00 1.20083070e+00 1.09220177e-01 -1.63534284e+00 6.19008899e-01 -4.36226130e-02 3.86307746e-01 -6.44515216e-01 1.93761975e-01 1.67000011e-01 -1.35535672e-01 -1.13477349e+00 4.78429437e-01 2.07827464e-01 1.02213252e+00 -3.30291599e-01 1.21466029e+00 -9.28410068e-02 -8.84178996e-01 -2.97448844e-01 -3.61797243e-01 2.41897389e-01 1.66350305e-01 5.30257523e-01 -1.48138547e+00 4.52129096e-01 4.05228674e-01 7.73093104e-01 -5.06011963e-01 1.11728716e+00 -5.99305257e-02 3.60886633e-01 -4.19304753e-03 -2.58185804e-01 2.26556718e-01 3.97482395e-01 3.54723662e-01 9.75539088e-01 4.61499512e-01 -1.73470303e-01 -8.03017169e-02 9.23584878e-01 6.27887482e-03 1.31097525e-01 -9.51181591e-01 -7.82404542e-01 3.77459347e-01 5.02547622e-01 -6.14459574e-01 -5.91417015e-01 -3.11716646e-01 8.95528734e-01 -7.15945303e-01 2.29606256e-01 -8.55508626e-01 -5.08654833e-01 6.60035372e-01 2.70357102e-01 1.73734948e-01 4.52826560e-01 -5.14245510e-01 -5.76304615e-01 3.80261876e-02 -5.13950825e-01 4.01077956e-01 -9.69184399e-01 -1.55764890e+00 1.03189135e+00 -3.36747944e-01 -1.69034600e+00 -5.46160787e-02 -9.69375849e-01 -6.97503537e-02 9.82593656e-01 -1.15507638e+00 -7.15299904e-01 4.00301427e-01 3.05856224e-02 9.23402786e-01 -1.77315444e-01 1.46148169e+00 7.63457894e-01 1.02274595e-02 7.12180018e-01 -3.24253917e-01 2.62374878e-01 5.70935071e-01 -1.05440342e+00 2.87099957e-01 2.10032031e-01 -6.44468963e-02 1.01741767e+00 7.36519933e-01 -1.07878101e+00 -7.97443151e-01 -1.18525314e+00 1.66704929e+00 -7.15805829e-01 5.52970290e-01 -6.82368875e-02 -5.41849732e-01 3.21501583e-01 7.67488107e-02 -4.51448351e-01 8.52665544e-01 -2.47502849e-01 -3.02779078e-01 1.18679501e-01 -1.48637319e+00 3.55760038e-01 6.85035050e-01 -8.37884903e-01 -1.01036882e+00 1.02746737e+00 1.08843935e+00 -1.97034821e-01 -1.14796495e+00 4.87753510e-01 4.01331246e-01 9.83181596e-02 9.84999239e-01 -1.10394466e+00 4.40308273e-01 -2.66324043e-01 5.15673831e-02 -8.57649684e-01 -2.14686021e-01 -1.47854760e-01 5.50353169e-01 1.07894957e+00 1.17884219e+00 -6.29630685e-01 7.13348866e-01 8.01652491e-01 -1.19225845e-01 -8.19721997e-01 -8.01579714e-01 -9.37068224e-01 -2.16868490e-01 -3.66409004e-01 5.22289455e-01 9.97542143e-01 4.12745774e-03 3.83954644e-01 3.51480907e-03 1.96908951e-01 1.59280345e-01 -2.66698033e-01 -2.76697606e-01 -1.14910984e+00 8.04353654e-02 -3.28503907e-01 -9.20978069e-01 -6.90774441e-01 6.43502027e-02 -1.02962685e+00 -1.11285448e-01 -1.65665793e+00 3.25239785e-02 -5.04778087e-01 -4.77234483e-01 4.12825584e-01 2.34269336e-01 -1.05159044e-01 -3.12684566e-01 -9.55497921e-02 -9.33061615e-02 -1.89293921e-01 8.81145954e-01 -8.25192153e-01 -2.42477264e-02 1.80260465e-01 -9.19906914e-01 5.17341852e-01 6.70168579e-01 -1.01394212e+00 -4.86645609e-01 -4.22847033e-01 2.61280090e-01 3.79245579e-02 2.20747158e-01 -7.93333292e-01 -2.56263223e-02 4.20610726e-01 4.59864587e-01 -2.56401211e-01 1.36520907e-01 -1.43785322e+00 3.76565754e-01 8.34146798e-01 -9.37991738e-01 4.30273414e-01 4.33505177e-01 7.19049215e-01 -1.43872470e-01 -7.42498636e-01 1.51247814e-01 -1.29132792e-01 -5.36305130e-01 1.35088548e-01 -6.08648300e-01 -3.27205241e-01 8.32250714e-01 -5.86321235e-01 -1.46283239e-01 -8.11592862e-02 -1.23358107e+00 -1.67977754e-02 1.15833700e-01 7.32732177e-01 8.07834566e-01 -1.21053219e+00 -1.94576606e-01 1.34693041e-01 5.05879641e-01 -6.97257936e-01 2.28746398e-03 3.29771936e-01 -6.29740775e-01 8.25379074e-01 1.78729221e-01 -2.70338058e-01 -1.22246087e+00 9.28109944e-01 2.77456045e-01 -2.16812626e-01 -4.26112711e-01 7.61569321e-01 -2.59939563e-02 -1.03901900e-01 4.58790928e-01 -1.06436646e+00 -3.76990408e-01 -5.79334125e-02 5.32430708e-01 -2.34186754e-01 4.45614338e-01 -3.85054171e-01 -7.61121690e-01 7.18880832e-01 -9.87719893e-02 3.21653068e-01 1.07080984e+00 3.54155958e-01 1.85436815e-01 2.79354751e-01 1.74280739e+00 -9.81157050e-02 -1.69756666e-01 6.40162677e-02 4.13068116e-01 -2.49262184e-01 -6.97792843e-02 -1.29571867e+00 -7.07943559e-01 7.39405096e-01 1.35671830e+00 3.66079420e-01 7.37832248e-01 5.38234524e-02 9.53840196e-01 2.78613921e-02 3.66156995e-01 -9.76621568e-01 -3.52216601e-01 -5.12053892e-02 4.51850861e-01 -1.11962175e+00 -8.23855177e-02 -6.82904363e-01 -6.12785697e-01 1.29364049e+00 1.10366940e-01 -4.61443588e-02 8.33134472e-01 -2.82351524e-02 -3.19756418e-02 -7.86822796e-01 -1.09637447e-01 8.79852995e-02 4.20042396e-01 5.18382072e-01 4.83202040e-01 2.32876837e-01 -6.12098098e-01 8.40349853e-01 1.69871837e-01 9.42951068e-02 3.09000522e-01 8.28661919e-01 -3.43324728e-02 -1.19483662e+00 1.67046160e-01 9.11639571e-01 -8.92204225e-01 -4.35528964e-01 -5.00172377e-01 5.37619472e-01 5.13946533e-01 1.17751753e+00 -7.52613544e-02 -6.72088861e-01 6.84347153e-01 3.62394035e-01 3.12323809e-01 -1.08038676e+00 -1.02323675e+00 -8.55017528e-02 4.38242853e-01 -4.09386128e-01 -2.42003903e-01 -2.13523641e-01 -1.13272810e+00 2.09789500e-01 -6.39641285e-01 3.01037669e-01 7.42629111e-01 8.54375541e-01 5.97028255e-01 7.49570012e-01 1.92937553e-01 7.63603486e-03 -7.16208100e-01 -8.14798474e-01 -4.63947445e-01 7.33367324e-01 3.73406053e-01 -4.93427008e-01 -2.81597376e-01 8.21375921e-02]
[8.365416526794434, 8.657524108886719]
f412f1ea-e5ee-476e-93ac-60536b11e10c
an-effective-approach-to-unsupervised-machine
1902.01313
null
https://arxiv.org/abs/1902.01313v2
https://arxiv.org/pdf/1902.01313v2.pdf
An Effective Approach to Unsupervised Machine Translation
While machine translation has traditionally relied on large amounts of parallel corpora, a recent research line has managed to train both Neural Machine Translation (NMT) and Statistical Machine Translation (SMT) systems using monolingual corpora only. In this paper, we identify and address several deficiencies of existing unsupervised SMT approaches by exploiting subword information, developing a theoretically well founded unsupervised tuning method, and incorporating a joint refinement procedure. Moreover, we use our improved SMT system to initialize a dual NMT model, which is further fine-tuned through on-the-fly back-translation. Together, we obtain large improvements over the previous state-of-the-art in unsupervised machine translation. For instance, we get 22.5 BLEU points in English-to-German WMT 2014, 5.5 points more than the previous best unsupervised system, and 0.5 points more than the (supervised) shared task winner back in 2014.
['Mikel Artetxe', 'Gorka Labaka', 'Eneko Agirre']
2019-02-04
an-effective-approach-to-unsupervised-machine-1
https://aclanthology.org/P19-1019
https://aclanthology.org/P19-1019.pdf
acl-2019-7
['unsupervised-machine-translation']
['natural-language-processing']
[ 4.93254572e-01 1.43248335e-01 -4.58261430e-01 -3.17986459e-01 -1.56058061e+00 -7.86879063e-01 1.05609000e+00 -3.17275785e-02 -6.49921000e-01 1.17031562e+00 2.11264208e-01 -7.94821739e-01 3.35731447e-01 -3.36097956e-01 -9.52350259e-01 -4.52255726e-01 4.69466835e-01 1.32089365e+00 -1.98769912e-01 -5.55059433e-01 1.46793142e-01 5.33863232e-02 -8.38236630e-01 3.76554459e-01 1.25419879e+00 2.70011902e-01 1.42515421e-01 5.01374424e-01 -2.08936676e-01 2.56001651e-01 -4.78064448e-01 -7.59808898e-01 3.58611971e-01 -6.28716826e-01 -9.94859874e-01 -1.61241591e-01 4.22388464e-01 9.44151506e-02 1.05766729e-01 8.38029981e-01 6.77930176e-01 -1.38470531e-01 5.53940594e-01 -6.33189678e-01 -8.64473641e-01 1.17380273e+00 -6.26326501e-01 -5.57802282e-02 7.76644647e-02 1.54091179e-01 1.02329981e+00 -1.18878257e+00 8.95590067e-01 9.88961041e-01 5.72089314e-01 7.39034891e-01 -1.49903429e+00 -7.24137723e-01 -3.20744604e-01 -9.60356072e-02 -1.12928712e+00 -7.00673878e-01 3.22790444e-01 -2.99114972e-01 1.55109978e+00 1.98095500e-01 2.28731930e-01 1.40922904e+00 2.66264647e-01 7.01195598e-01 1.52746582e+00 -9.93989170e-01 -1.70170248e-01 2.86491424e-01 -3.05182159e-01 2.93799639e-01 -2.87152845e-02 2.23049030e-01 -4.95347887e-01 1.38910990e-02 6.42418981e-01 -5.39016187e-01 2.37420201e-01 -1.52854668e-02 -1.80837619e+00 7.61059165e-01 -1.30882531e-01 5.06837785e-01 -2.66325653e-01 -3.13965380e-02 4.13854688e-01 8.05684566e-01 9.24392462e-01 7.58964121e-01 -9.41046000e-01 -5.30769467e-01 -1.31294584e+00 -7.35922530e-02 8.44440579e-01 1.09904003e+00 8.84551346e-01 -6.34072125e-02 -1.70169160e-01 1.08118141e+00 -3.93293984e-02 8.05352747e-01 5.66106319e-01 -6.98791504e-01 1.04097116e+00 2.00448051e-01 -1.01089478e-01 -2.33671889e-01 -8.38148966e-02 -6.74203336e-01 -7.64128327e-01 -2.28673682e-01 4.05551493e-01 -4.03068990e-01 -9.68836725e-01 1.76762414e+00 -8.09862651e-03 -1.89148307e-01 1.45532519e-01 7.04444826e-01 1.42134950e-01 7.02642858e-01 -1.04813710e-01 -4.75100011e-01 9.44787323e-01 -1.40913582e+00 -7.03992724e-01 -3.40244263e-01 1.00492573e+00 -1.51410949e+00 1.03692889e+00 3.19260478e-01 -1.38256848e+00 -5.75405121e-01 -8.74732256e-01 -1.55588120e-01 -3.19251657e-01 2.61886865e-01 5.74277461e-01 5.03145337e-01 -1.25291264e+00 7.98868179e-01 -7.29036212e-01 -6.47258997e-01 -1.40172448e-02 7.69450605e-01 -4.17651922e-01 8.09335485e-02 -1.37261569e+00 1.46457350e+00 2.95580804e-01 2.61613051e-03 -7.10045576e-01 -4.03845817e-01 -5.04696131e-01 -4.90358800e-01 3.00006509e-01 -9.35716808e-01 1.53470314e+00 -1.22412038e+00 -2.11485982e+00 1.11873841e+00 -3.52538913e-01 -5.33152223e-01 7.00099885e-01 -3.78140956e-01 -3.75143468e-01 -3.52508903e-01 3.02282631e-01 7.67704248e-01 5.55571973e-01 -8.54946256e-01 -5.22876859e-01 -1.52670518e-01 -3.56969237e-01 2.41913542e-01 -2.96988875e-01 5.79529285e-01 -6.30110621e-01 -7.23831594e-01 -1.44681916e-01 -1.21284294e+00 -4.09880608e-01 -9.33699489e-01 -3.91553640e-01 -1.22444190e-01 1.29680514e-01 -8.77889633e-01 1.13634849e+00 -1.67303646e+00 7.09445536e-01 -4.34933677e-02 -1.44955873e-01 2.05467522e-01 -5.09718359e-01 7.44132876e-01 -1.24051254e-02 1.07720055e-01 -3.92675966e-01 -7.01449335e-01 8.65099877e-02 3.43563110e-01 -2.47881696e-01 1.92968100e-01 3.75524223e-01 1.24006915e+00 -8.59914303e-01 -4.93281126e-01 -8.67344514e-02 1.09134756e-01 -4.14458990e-01 -8.82451683e-02 -2.52322257e-01 8.33590806e-01 -1.46560773e-01 4.77876216e-01 4.85702634e-01 1.45020243e-02 4.49499071e-01 3.29422832e-01 -4.40308899e-01 8.00100148e-01 -4.73465174e-01 2.15506315e+00 -6.24254048e-01 6.65072262e-01 -2.90430576e-01 -8.68203282e-01 9.95670199e-01 5.14164686e-01 4.47041839e-01 -7.90518522e-01 1.31825149e-01 9.70747232e-01 1.46185964e-01 -2.14629725e-01 6.21556222e-01 -2.81961262e-01 -4.64586094e-02 6.89668536e-01 4.05663133e-01 -1.58944070e-01 2.88610518e-01 -9.50186774e-02 8.13476741e-01 5.68718791e-01 -2.45225374e-02 -5.52737713e-01 3.92102182e-01 4.15701151e-01 4.23821896e-01 5.44043422e-01 1.53020039e-01 6.80464745e-01 5.49285822e-02 -2.67480552e-01 -1.55970955e+00 -9.08323407e-01 -2.98346169e-02 1.11482692e+00 -4.77872908e-01 -4.52403605e-01 -1.03206158e+00 -7.26151228e-01 -3.94985497e-01 8.30006897e-01 -4.84674811e-01 1.06147461e-01 -1.12549746e+00 -1.00807631e+00 7.82868564e-01 3.12654972e-01 2.40112916e-01 -1.00999868e+00 2.78226495e-01 4.96832699e-01 -5.66949606e-01 -1.10961270e+00 -6.98525071e-01 3.45661134e-01 -1.10530245e+00 -2.85212338e-01 -8.69302928e-01 -8.95497978e-01 3.32314879e-01 -3.77295651e-02 1.55917668e+00 -3.13511103e-01 3.87633234e-01 -3.49656045e-01 -2.27445275e-01 -4.48485434e-01 -8.84169161e-01 9.64246988e-01 3.49715501e-01 -2.15968400e-01 7.23033190e-01 -6.42462790e-01 -7.23921433e-02 2.87010640e-01 -5.44691026e-01 3.94936770e-01 1.04899073e+00 9.41680372e-01 5.07896960e-01 -7.92144775e-01 4.94212955e-01 -9.76803720e-01 6.38511956e-01 -1.94360495e-01 -4.66608733e-01 3.85370970e-01 -1.03014505e+00 2.21832335e-01 4.75723952e-01 -5.49386561e-01 -8.45233023e-01 -8.13744515e-02 -1.49733350e-01 -1.17989004e-01 5.14633954e-02 4.70479816e-01 4.12176140e-02 4.22021076e-02 6.70259356e-01 3.89002144e-01 -2.12012623e-02 -5.63338459e-01 5.84421873e-01 8.66427600e-01 3.24247599e-01 -8.51204991e-01 1.04588807e+00 -9.21057761e-02 -3.86448532e-01 -2.71793395e-01 -6.46788657e-01 -2.85736650e-01 -1.05365849e+00 1.29710689e-01 7.59535670e-01 -9.06981468e-01 -3.22593860e-02 2.78259486e-01 -1.20876884e+00 -6.38909221e-01 -5.82951233e-02 7.39219844e-01 -8.23456764e-01 2.33051345e-01 -9.67558980e-01 -2.88162440e-01 -7.40643024e-01 -1.36284304e+00 1.26693392e+00 -4.07676488e-01 -5.88022709e-01 -1.04572475e+00 5.52031279e-01 6.53396964e-01 6.01140559e-01 -1.87708855e-01 7.75675118e-01 -6.23398542e-01 -3.30066770e-01 1.23315074e-01 -1.17385358e-01 4.25733089e-01 6.05217963e-02 -1.27193034e-01 -6.53918564e-01 -3.35936755e-01 -1.56015784e-01 -1.94334358e-01 7.13417172e-01 1.98164299e-01 2.37711787e-01 -1.88930690e-01 -2.12274462e-01 6.39211237e-01 1.18503845e+00 -8.22318718e-02 5.66798270e-01 8.10156941e-01 5.52086413e-01 4.58591938e-01 5.10041356e-01 -1.79188862e-01 3.39012951e-01 9.36688960e-01 -1.85655698e-01 -3.82752717e-01 -1.57709792e-01 -2.40947261e-01 7.10965455e-01 1.71714473e+00 -5.61598301e-01 4.48409393e-02 -1.07530069e+00 5.52032948e-01 -1.89547646e+00 -5.68945408e-01 -2.69914031e-01 2.18097210e+00 1.45616817e+00 2.09064886e-01 4.76166010e-02 -3.97646755e-01 6.13984525e-01 -3.14026862e-01 -1.13476887e-01 -8.46185267e-01 -3.76975477e-01 6.97967529e-01 8.69567931e-01 5.98815382e-01 -8.26926410e-01 1.62452817e+00 7.09446812e+00 1.00209260e+00 -1.05975544e+00 4.62525457e-01 3.92498374e-01 -1.07638821e-01 -3.76897007e-01 2.06571102e-01 -7.45305955e-01 1.95931852e-01 1.42991376e+00 -3.06883212e-02 7.92696059e-01 2.58695990e-01 3.00338268e-01 4.05922532e-01 -1.08748925e+00 6.50158048e-01 -1.97164994e-02 -1.25839198e+00 1.50225937e-01 3.47570539e-01 1.32992017e+00 6.38562739e-01 -1.28393881e-02 6.22952521e-01 5.08353233e-01 -9.35314536e-01 5.88554800e-01 2.00032383e-01 1.06859374e+00 -7.00029790e-01 7.78646469e-01 4.58601207e-01 -6.02868557e-01 4.39783216e-01 -3.41780245e-01 -1.01067759e-01 4.53369953e-02 5.45262814e-01 -8.35494757e-01 9.18348908e-01 2.66841292e-01 6.48058116e-01 -4.27115947e-01 5.26985943e-01 -4.01444465e-01 9.70476806e-01 -2.33147159e-01 5.32708131e-02 6.47659302e-01 -5.42951703e-01 5.37192583e-01 1.47626770e+00 4.25210506e-01 -4.48480785e-01 6.91284016e-02 4.79885072e-01 -4.62532073e-01 5.11168838e-01 -4.05025423e-01 -1.55544728e-01 9.92202982e-02 1.07946992e+00 -3.29890847e-01 -5.90636730e-01 -3.84556770e-01 1.44384718e+00 5.62521160e-01 3.86552244e-01 -8.34083617e-01 -1.01678647e-01 5.18504083e-01 -1.77571937e-01 1.51963249e-01 -5.30195117e-01 -5.62723041e-01 -1.50658298e+00 1.09442577e-01 -1.39186704e+00 -1.02711871e-01 -4.25569326e-01 -1.21269393e+00 9.14001524e-01 -3.56111884e-01 -1.27279103e+00 -6.80657923e-01 -4.90503669e-01 -2.46698335e-01 1.25703585e+00 -1.41952145e+00 -1.39446735e+00 6.36577129e-01 2.49090120e-01 8.60088944e-01 -4.10958350e-01 1.22932136e+00 5.70009291e-01 -5.61568677e-01 9.20621812e-01 6.26817882e-01 6.99457526e-02 1.41519928e+00 -1.29083395e+00 1.10642350e+00 8.13834369e-01 4.07927126e-01 9.08726215e-01 6.85398638e-01 -6.17146730e-01 -1.47077191e+00 -1.02737415e+00 1.80108905e+00 -9.95621145e-01 1.02596307e+00 -5.91107965e-01 -5.61745882e-01 8.58925462e-01 8.83263350e-01 -6.29569054e-01 5.33831000e-01 4.79915828e-01 -4.13925439e-01 1.60557047e-01 -6.35270834e-01 7.80451238e-01 1.00609481e+00 -6.32230520e-01 -7.14931726e-01 5.62338233e-01 7.94327080e-01 -4.22672778e-01 -9.81919527e-01 4.75350529e-01 5.56055129e-01 -4.22368169e-01 7.26156592e-01 -8.37506652e-01 6.94388092e-01 -7.58910403e-02 -9.18659195e-02 -1.61571264e+00 -3.00084233e-01 -1.17778587e+00 1.93260372e-01 1.21153486e+00 1.26482511e+00 -6.85018957e-01 5.88370264e-01 6.92136213e-02 -2.90595263e-01 -7.12733328e-01 -9.72287476e-01 -9.82992351e-01 8.19584906e-01 -3.39944154e-01 4.62315232e-01 1.17618966e+00 1.26956835e-01 9.43937182e-01 -7.73322642e-01 -2.69924939e-01 5.49902439e-01 3.74711044e-02 9.10596550e-01 -8.87409210e-01 -5.94434738e-01 -6.80935323e-01 2.16764629e-01 -1.14426160e+00 9.49776247e-02 -1.24865270e+00 9.31084827e-02 -1.42322266e+00 4.49648649e-01 -1.52458087e-01 -4.29748893e-01 5.12065887e-01 -1.95300072e-01 6.23830378e-01 5.80375753e-02 5.99578321e-01 -3.12525004e-01 2.77180135e-01 1.33081937e+00 -1.42536849e-01 -1.03613839e-01 -1.40951693e-01 -5.55779338e-01 1.86432332e-01 1.00197279e+00 -6.53615892e-01 6.60840571e-02 -1.19666851e+00 2.84011602e-01 -9.17925984e-02 -2.50875652e-01 -5.55677891e-01 1.31715581e-01 -1.47452384e-01 1.33516565e-01 -4.52853620e-01 1.15622990e-01 -4.25252885e-01 5.30870073e-02 1.98601395e-01 -4.06979233e-01 3.99767965e-01 3.75449032e-01 -4.28029932e-02 -2.23754302e-01 -8.21640342e-02 4.43849385e-01 -1.28736049e-01 -1.38784170e-01 1.35549694e-01 -4.34043765e-01 -3.61292213e-02 4.66172099e-01 2.79186387e-03 -1.35663420e-01 -4.02073897e-02 -5.25859654e-01 4.19265479e-02 4.45844054e-01 6.15533352e-01 -2.48616636e-02 -1.25686538e+00 -1.39680910e+00 8.81498829e-02 1.52785614e-01 -5.17058671e-01 -3.42976242e-01 1.22781026e+00 -3.31850857e-01 6.68806076e-01 -2.14941755e-01 -4.79898483e-01 -8.85195613e-01 3.62131029e-01 6.21675923e-02 -7.24137783e-01 -3.26329738e-01 5.89317203e-01 -3.60429645e-01 -1.11523521e+00 -2.68903673e-01 -5.29117286e-02 3.53830457e-01 -1.47369578e-01 1.91351026e-01 1.24552935e-01 3.90572935e-01 -5.59861243e-01 -1.54360622e-01 7.47067750e-01 -2.34270811e-01 -8.20957065e-01 1.28250766e+00 -1.02673545e-01 -4.62388217e-01 5.06145239e-01 1.04339647e+00 2.96136707e-01 -5.52655220e-01 -5.26971400e-01 3.87684911e-01 -6.78377897e-02 -1.74067706e-01 -1.23126841e+00 -5.29635012e-01 8.95020664e-01 1.95938915e-01 -2.98498541e-01 9.40466642e-01 -2.35508472e-01 1.05751848e+00 6.38224840e-01 5.52932441e-01 -1.32882690e+00 -5.03913164e-01 9.32513714e-01 6.39222801e-01 -1.28587770e+00 -3.32421392e-01 -2.24689350e-01 -4.18966800e-01 1.06825733e+00 1.87028304e-01 -2.15046876e-03 -1.10134557e-02 3.64224374e-01 4.92543042e-01 2.82827556e-01 -9.76349175e-01 -1.11456029e-01 4.00900453e-01 2.78868139e-01 8.60088348e-01 3.01716745e-01 -7.07739949e-01 2.14264423e-01 -4.27641779e-01 -1.26719913e-02 2.77606361e-02 5.73710024e-01 -1.80277228e-01 -1.91608834e+00 -2.14318842e-01 9.57296975e-03 -6.69093966e-01 -7.43798494e-01 -7.34077394e-01 8.25151861e-01 -1.03704475e-01 9.93468404e-01 -2.55161762e-01 -5.48121572e-01 2.54581243e-01 3.58660907e-01 7.22856164e-01 -7.60532737e-01 -1.06012523e+00 5.24428308e-01 3.94892454e-01 -2.89461195e-01 -4.66697812e-01 -6.85541391e-01 -7.99109817e-01 -3.82772952e-01 -4.57765698e-01 3.97851765e-01 9.02010143e-01 1.16437030e+00 3.44508737e-01 2.49997705e-01 5.63914120e-01 -7.07938612e-01 -6.75878108e-01 -1.60840154e+00 1.94518402e-01 1.21798493e-01 -1.05184421e-01 -1.10111825e-01 -1.13560550e-01 4.68406864e-02]
[11.589447975158691, 10.348690032958984]
4771548f-1b92-4f97-a83f-89b222fdfaf0
sold2-self-supervised-occlusion-aware-line
2104.03362
null
https://arxiv.org/abs/2104.03362v2
https://arxiv.org/pdf/2104.03362v2.pdf
SOLD2: Self-supervised Occlusion-aware Line Description and Detection
Compared to feature point detection and description, detecting and matching line segments offer additional challenges. Yet, line features represent a promising complement to points for multi-view tasks. Lines are indeed well-defined by the image gradient, frequently appear even in poorly textured areas and offer robust structural cues. We thus hereby introduce the first joint detection and description of line segments in a single deep network. Thanks to a self-supervised training, our method does not require any annotated line labels and can therefore generalize to any dataset. Our detector offers repeatable and accurate localization of line segments in images, departing from the wireframe parsing approach. Leveraging the recent progresses in descriptor learning, our proposed line descriptor is highly discriminative, while remaining robust to viewpoint changes and occlusions. We evaluate our approach against previous line detection and description methods on several multi-view datasets created with homographic warps as well as real-world viewpoint changes. Our full pipeline yields higher repeatability, localization accuracy and matching metrics, and thus represents a first step to bridge the gap with learned feature points methods. Code and trained weights are available at https://github.com/cvg/SOLD2.
['Marc Pollefeys', 'Martin R. Oswald', 'Viktor Larsson', 'Juan-Ting Lin', 'Rémi Pautrat']
2021-04-07
null
http://openaccess.thecvf.com//content/CVPR2021/html/Pautrat_SOLD2_Self-Supervised_Occlusion-Aware_Line_Description_and_Detection_CVPR_2021_paper.html
http://openaccess.thecvf.com//content/CVPR2021/papers/Pautrat_SOLD2_Self-Supervised_Occlusion-Aware_Line_Description_and_Detection_CVPR_2021_paper.pdf
cvpr-2021-1
['line-detection', 'wireframe-parsing']
['computer-vision', 'computer-vision']
[-7.97843486e-02 -3.51091921e-01 -2.67311871e-01 -3.85444820e-01 -9.86281335e-01 -1.02182281e+00 7.64336109e-01 4.80318367e-01 -3.22390646e-01 3.47409844e-01 -1.45345420e-01 2.14304954e-01 4.89985757e-02 -6.48595393e-01 -8.16776752e-01 -3.15041482e-01 -8.13995898e-02 5.60393929e-01 5.36860883e-01 -2.84679830e-01 4.40326393e-01 1.02984476e+00 -1.37002194e+00 2.21700296e-02 2.20477223e-01 1.10638833e+00 -1.00008033e-01 5.08059382e-01 1.37556538e-01 1.58190295e-01 -2.13048860e-01 -3.83702606e-01 3.69518965e-01 1.15483306e-01 -5.48574388e-01 3.60031188e-01 1.24434018e+00 -5.04712939e-01 -3.67436260e-01 4.91079450e-01 6.66112304e-01 -1.31100163e-01 4.68584299e-01 -1.06178939e+00 -3.87969643e-01 1.42452955e-01 -8.27154338e-01 7.37807341e-03 8.39254618e-01 -1.39963970e-01 1.21220517e+00 -1.25906086e+00 1.09884381e+00 7.66582310e-01 1.11886394e+00 2.13190317e-01 -1.27235448e+00 -1.98640138e-01 1.40291802e-03 5.23903826e-03 -1.33044147e+00 -6.02672338e-01 9.73001301e-01 -4.96346653e-01 9.86006558e-01 7.11381715e-03 7.72770882e-01 1.24256968e+00 1.07274197e-01 9.10560608e-01 9.75830615e-01 -4.98017520e-01 -1.29057929e-01 -6.95907697e-02 -1.55123901e-02 9.31701005e-01 2.78940529e-01 1.91670097e-02 -6.55829191e-01 8.49830266e-03 1.15548778e+00 2.23253593e-01 -4.16864693e-01 -1.28549302e+00 -1.51768184e+00 6.09730363e-01 3.64948809e-01 1.95663646e-01 -1.14862457e-01 3.70383084e-01 4.02644515e-01 -2.19935328e-02 3.95779788e-01 4.24527764e-01 -4.09177303e-01 -2.69881189e-01 -1.14743781e+00 2.02437773e-01 6.22239113e-01 1.11367106e+00 9.32182610e-01 -2.56822050e-01 2.34233886e-01 8.37494254e-01 3.27680111e-02 6.53142333e-01 1.37600034e-01 -8.75719845e-01 4.63329643e-01 4.79955524e-01 1.43702701e-01 -1.20640182e+00 -9.00936484e-01 -6.02980494e-01 -3.67359966e-01 3.05272818e-01 5.68024397e-01 1.01185821e-01 -5.79986036e-01 1.17448020e+00 2.05988839e-01 -3.08411807e-01 -5.59529781e-01 7.13860154e-01 6.34943068e-01 1.92424972e-02 -6.95079088e-01 2.32487693e-01 1.39013720e+00 -9.94710803e-01 -2.60408640e-01 -2.57533997e-01 6.84185266e-01 -1.14553392e+00 8.12440872e-01 4.74531621e-01 -1.13543630e+00 -4.42327023e-01 -1.11206603e+00 -3.65813404e-01 -3.50874394e-01 2.52507538e-01 8.30258667e-01 3.35971862e-01 -1.08590055e+00 6.63826883e-01 -9.52808142e-01 -6.51487350e-01 5.51535070e-01 3.84855628e-01 -8.44842732e-01 -1.32263809e-01 -4.93430734e-01 6.93640649e-01 5.05508222e-02 1.33789524e-01 -2.33046681e-01 -5.69509506e-01 -1.10904419e+00 -1.16307348e-01 3.33854318e-01 -8.26816082e-01 1.07145751e+00 -4.54261959e-01 -1.24781573e+00 1.28248143e+00 -2.61184543e-01 -3.43547523e-01 9.43665385e-01 -4.89071369e-01 -9.22692269e-02 5.23580611e-01 1.36883914e-01 7.37572849e-01 6.88796341e-01 -1.43945575e+00 -5.41747808e-01 -3.81399244e-01 -4.24636565e-02 -2.57099848e-02 1.26159906e-01 -2.04742789e-01 -5.36139429e-01 -3.39783728e-01 5.28316855e-01 -9.19656277e-01 -2.65764631e-02 4.81267899e-01 -4.62883651e-01 -1.43589899e-01 8.30196977e-01 -2.41920620e-01 8.29224527e-01 -1.94053435e+00 -1.37668610e-01 2.84442723e-01 2.99936384e-01 -4.77071144e-02 2.86013018e-02 8.92204821e-01 3.85055027e-04 -1.21556744e-01 -5.62979886e-03 -4.40152258e-01 1.24455802e-02 -1.30861878e-01 -2.25856557e-01 9.87071574e-01 1.86454341e-01 9.42777634e-01 -8.05817962e-01 -4.02316630e-01 5.22953391e-01 5.71253121e-01 -3.75780046e-01 -1.01454370e-01 5.46254627e-02 2.31901631e-01 -2.20181033e-01 8.47875655e-01 6.47082746e-01 -2.95498312e-01 -3.21670800e-01 -4.82891023e-01 -3.82564247e-01 -1.90102711e-01 -1.17941105e+00 2.27061605e+00 -5.22892177e-01 9.23873782e-01 -3.04106325e-01 -5.91796815e-01 1.25702989e+00 -3.62329707e-02 7.56208301e-01 -5.59772551e-01 3.78789790e-02 5.72225213e-01 -4.28225219e-01 -3.16578656e-01 5.51643729e-01 3.65160227e-01 -2.75873858e-02 1.14186808e-01 2.84651130e-01 -5.22057176e-01 2.82302767e-01 5.25362752e-02 9.39399004e-01 7.39225507e-01 5.29644847e-01 1.15167782e-01 3.42369199e-01 -7.99104124e-02 3.51130307e-01 6.07591093e-01 -1.05722427e-01 1.16500330e+00 4.65350121e-01 -7.62060106e-01 -1.39195144e+00 -1.12799585e+00 -5.54022968e-01 5.96806884e-01 2.80916691e-01 -4.35076565e-01 -3.40195477e-01 -4.78417903e-01 2.51390815e-01 -8.23596306e-03 -4.42786366e-01 3.26604545e-01 -7.52571404e-01 -7.23862201e-02 3.83853406e-01 5.93960941e-01 4.66605812e-01 -4.60658103e-01 -7.22184539e-01 1.45812124e-01 3.08540314e-01 -1.42152464e+00 -4.68250811e-01 1.56715646e-01 -5.98109245e-01 -1.31222689e+00 -9.40907001e-01 -7.10931778e-01 6.57907665e-01 5.35184860e-01 1.12864578e+00 1.58095863e-02 -4.63074625e-01 9.42876637e-01 -2.76585013e-01 -7.37809464e-02 2.14815542e-01 1.65038645e-01 1.13833474e-03 -1.62296295e-01 2.91222960e-01 -5.82002163e-01 -1.04798520e+00 3.41182709e-01 -4.56605464e-01 -1.00842096e-01 3.78906339e-01 8.69522989e-01 7.45528936e-01 -7.98993647e-01 -4.88650948e-02 -6.42565072e-01 3.29017639e-01 -2.84554251e-02 -8.65882277e-01 1.21632010e-01 -4.28227156e-01 -1.72843605e-01 4.46292967e-01 1.72177196e-01 -4.65235859e-01 5.12610495e-01 -1.31743953e-01 -2.73090303e-01 -5.39362431e-01 5.76032735e-02 2.90261596e-01 -5.27162015e-01 6.71870410e-01 5.92511892e-02 -1.29380658e-01 -3.24968725e-01 6.06378913e-01 2.41080180e-01 5.54853082e-01 -3.91740173e-01 9.36340392e-01 1.12905908e+00 2.60534167e-01 -8.88990879e-01 -6.79949284e-01 -9.37920570e-01 -1.46738112e+00 -2.66118139e-01 6.48127973e-01 -8.19212079e-01 -7.35709310e-01 4.33033794e-01 -1.36010563e+00 -9.95927397e-03 -1.72210321e-01 5.10380566e-01 -9.46555495e-01 7.37908483e-01 -5.28980076e-01 -3.92325252e-01 -1.92817718e-01 -1.15770829e+00 1.64579225e+00 8.46623704e-02 -1.91855252e-01 -1.25104630e+00 2.60847092e-01 2.46634126e-01 1.08748317e-01 8.03308427e-01 4.03111309e-01 -6.07417822e-01 -7.19433188e-01 -6.99930012e-01 -2.22980633e-01 6.20163940e-02 -1.07712582e-01 4.25893694e-01 -1.05276859e+00 -3.49902689e-01 -4.45868701e-01 -4.70773846e-01 9.07597184e-01 2.62079835e-01 8.61950517e-01 2.18151763e-01 -5.14515996e-01 1.08266127e+00 1.58994079e+00 -1.17166869e-01 2.22704008e-01 9.08037245e-01 8.17720950e-01 4.97128129e-01 3.87577474e-01 5.73006570e-01 4.57399935e-01 1.07068145e+00 3.23006034e-01 -4.33832794e-01 -1.85473084e-01 -3.37973744e-01 8.17547664e-02 5.95596015e-01 -9.07020867e-02 -1.50709331e-01 -1.15499556e+00 3.12204570e-01 -1.62149048e+00 -9.69961405e-01 -4.61295068e-01 2.20557332e+00 3.53565693e-01 1.79958373e-01 2.94745177e-01 8.97200257e-02 3.57389122e-01 3.08351249e-01 -5.01757681e-01 -3.23841393e-01 -4.63024825e-01 2.49039959e-02 7.92387187e-01 3.72609705e-01 -1.41211593e+00 9.53040123e-01 6.40685940e+00 4.97097105e-01 -1.41441178e+00 -2.55362004e-01 2.22965330e-01 9.49995518e-02 -3.82340521e-01 5.75029701e-02 -8.79641116e-01 -5.55401556e-02 3.01802576e-01 2.70672321e-01 5.13248853e-02 8.53853226e-01 8.80398303e-02 -9.91707891e-02 -1.33779216e+00 1.24259591e+00 3.65612537e-01 -1.53433168e+00 -3.01615030e-01 -3.28650549e-02 6.96867168e-01 5.41342616e-01 1.25238985e-01 -3.54848832e-01 -1.73173964e-01 -7.68101156e-01 7.12921739e-01 4.32494462e-01 8.07896376e-01 -4.89259511e-01 5.37366450e-01 -5.88518679e-02 -1.15762424e+00 2.20511690e-01 -3.62622023e-01 2.11084813e-01 3.53302956e-01 4.51854825e-01 -9.56198215e-01 6.87510490e-01 3.33619833e-01 1.14021337e+00 -8.73029649e-01 1.34795940e+00 -1.00363627e-01 3.45960036e-02 -4.27937537e-01 3.55458736e-01 2.80769259e-01 -2.06824139e-01 3.87439042e-01 1.37347877e+00 3.03263426e-01 -6.10681474e-01 2.45383188e-01 6.04281008e-01 2.31838286e-01 3.42730314e-01 -9.35273767e-01 3.10265362e-01 4.07661617e-01 1.51714540e+00 -1.06820154e+00 9.64630842e-02 -8.04683626e-01 1.14180064e+00 5.72123408e-01 3.56520712e-01 -5.66781282e-01 -5.13812423e-01 3.06761414e-01 2.08200499e-01 5.04340649e-01 -9.10491049e-01 -1.60349041e-01 -1.37742972e+00 3.33670259e-01 -4.38048929e-01 -7.43661600e-04 -7.08821535e-01 -1.07589030e+00 6.61744714e-01 -1.79475203e-01 -1.68993986e+00 -3.33351165e-01 -9.41599011e-01 -3.97680819e-01 3.62740278e-01 -1.76987183e+00 -1.65929937e+00 -6.03629291e-01 4.28904742e-01 4.92961138e-01 -6.64739385e-02 8.77059042e-01 1.33386180e-01 -3.54986578e-01 6.04145586e-01 4.65729952e-01 2.45879367e-01 9.54515755e-01 -1.30742788e+00 6.19454026e-01 6.81891680e-01 6.03010356e-01 6.90754473e-01 5.01927257e-01 -1.87699482e-01 -1.48225605e+00 -4.91379917e-01 5.06366968e-01 -6.88757479e-01 7.94000447e-01 -7.46393204e-01 -7.37538934e-01 6.93753004e-01 1.39717713e-01 4.17703003e-01 6.73695922e-01 1.57000214e-01 -7.13246584e-01 -1.12594388e-01 -7.49140441e-01 3.16831678e-01 1.19845629e+00 -5.29464543e-01 -2.86501080e-01 5.71309447e-01 1.93506733e-01 -8.14985454e-01 -1.07750094e+00 1.86645076e-01 7.82126427e-01 -1.48932219e+00 1.20845020e+00 -1.19867317e-01 3.35330129e-01 -1.46446019e-01 -1.51541680e-02 -1.11088324e+00 -2.09570557e-01 -7.13026941e-01 2.11333081e-01 1.18656945e+00 4.31688458e-01 -6.08127892e-01 8.25215936e-01 2.59407520e-01 -3.28872859e-01 -9.02865469e-01 -1.04391062e+00 -8.43853414e-01 -1.61420375e-01 -3.34823012e-01 2.66944945e-01 8.51760864e-01 2.75182277e-02 -7.50276446e-02 -3.12979430e-01 -8.87344871e-03 7.36864567e-01 4.33539569e-01 1.06278706e+00 -1.20116448e+00 -1.66836366e-01 -6.03821516e-01 -9.18407142e-01 -1.23385608e+00 5.30789904e-02 -9.95084882e-01 -2.12190166e-01 -1.60151672e+00 -1.23684630e-01 -4.44002837e-01 1.55067638e-01 3.06265861e-01 3.16014618e-01 4.87414032e-01 2.37644777e-01 4.11126465e-01 -7.39379227e-01 3.94275188e-01 1.10001898e+00 2.21748471e-01 7.89663717e-02 -1.42980022e-02 -7.62786865e-02 1.00997853e+00 6.39935791e-01 -8.20818394e-02 4.20942940e-02 -5.99212408e-01 3.40653628e-01 -2.34652162e-02 5.28067946e-01 -9.90750968e-01 4.74415302e-01 2.20426276e-01 6.06140614e-01 -7.20721722e-01 6.54344440e-01 -8.39191198e-01 -3.00227106e-01 2.50031918e-01 -1.24354064e-01 3.72084349e-01 5.16711958e-02 4.97934401e-01 -2.08822966e-01 -1.26367003e-01 5.63537836e-01 -2.52682418e-01 -1.05802560e+00 4.64073509e-01 1.85821950e-01 -1.85089469e-01 1.11425924e+00 -7.53041029e-01 -5.73798358e-01 -2.98476338e-01 -6.17033720e-01 1.20272964e-01 1.01769543e+00 4.30958480e-01 8.42795610e-01 -1.24028718e+00 -5.32866359e-01 3.64865392e-01 5.43923438e-01 1.46212414e-01 -9.67690349e-03 9.56581295e-01 -1.11289120e+00 6.50936067e-01 -3.57412696e-01 -9.89116848e-01 -1.11819816e+00 4.02004510e-01 3.97960126e-01 2.69557416e-01 -8.53648722e-01 7.19896317e-01 -1.26103004e-02 -2.75567114e-01 1.44895494e-01 -3.65162581e-01 -3.64289805e-02 2.25571752e-01 1.33704245e-01 4.04201984e-01 6.07670769e-02 -7.72062898e-01 -3.52899820e-01 1.38854825e+00 -1.26726761e-01 5.34084067e-03 1.51708758e+00 -2.13879615e-01 3.92164253e-02 5.79611003e-01 1.41768682e+00 5.14448822e-01 -1.38767767e+00 -2.07447380e-01 2.42041856e-01 -6.01713777e-01 -2.02371895e-01 -2.81573892e-01 -7.54889250e-01 1.00029087e+00 4.83954668e-01 1.10588066e-01 5.01667321e-01 8.66078213e-02 7.76288688e-01 2.03181028e-01 6.00567579e-01 -9.02380407e-01 1.83276936e-01 2.95976400e-01 9.92556870e-01 -1.50658953e+00 2.15577841e-01 -5.28209507e-01 -3.68115336e-01 1.80701673e+00 4.42022592e-01 -4.56494838e-01 3.44692588e-01 2.69003421e-01 2.61474758e-01 -3.80375057e-01 -3.19826156e-01 -2.35501796e-01 3.69428039e-01 7.82053053e-01 6.64950967e-01 -2.78747499e-01 -4.98927012e-02 -2.40351647e-01 -2.10224018e-01 -2.77334571e-01 5.43925285e-01 8.35163772e-01 -4.61343139e-01 -1.21240008e+00 -2.88966179e-01 9.97013673e-02 -2.46150002e-01 1.54248178e-01 -2.43898854e-01 1.18270147e+00 -1.63195506e-01 4.58835512e-01 1.32138848e-01 -1.74808413e-01 6.39338255e-01 -1.42195150e-01 6.19383752e-01 -5.05400419e-01 -2.81793088e-01 2.61524588e-01 4.38871160e-02 -8.96733880e-01 -5.19965112e-01 -9.73263204e-01 -1.02999568e+00 -4.37193401e-02 -4.26243126e-01 -2.81151682e-01 7.94884264e-01 5.80089211e-01 4.17458594e-01 6.44467324e-02 6.95601761e-01 -1.20078123e+00 -4.68896121e-01 -3.80395591e-01 -4.59586591e-01 5.81615567e-01 5.02557039e-01 -7.43808150e-01 -3.04383337e-01 -2.17566594e-01]
[8.158283233642578, -2.0482399463653564]
5db111f8-9d9a-4bb1-a833-2430fa441863
analyzing-and-improving-the-robustness-of
null
null
https://ieeexplore.ieee.org/document/9679972
https://ieeexplore.ieee.org/document/9679972
Analyzing and Improving the Robustness of Tabular Classifiers using Counterfactual Explanations
Recent studies have revealed that Machine Learning (ML) models are vulnerable to adversarial perturbations. Such perturbations can be intentionally or accidentally added to the original inputs, evading the classifier's behavior to misclassify the crafted samples. A widely-used solution is to retrain the model using data points generated by various attack strategies. However, this creates a classifier robust to some particular evasions and can not defend unknown or universal perturbations. Counterfactual explanations are a specific class of post-hoc explanation methods that provide minimal modification to the input features in order to obtain a particular outcome from the model. In addition to the resemblance of counterfactual explanations to the universal perturbations, the possibility of generating instances from specific classes makes such approaches suitable for analyzing and improving the model's robustness. Rather than explaining the model's decisions in the deployment phase, we utilize the distance information obtained from counterfactuals and propose novel metrics to analyze the robustness of tabular classifiers. Further, we introduce a decision boundary modification approach using customized counterfactual data points to improve the robustness of the models without compromising their accuracy. Our framework addresses the robustness of black-box classifiers in the tabular setting, which is considered an under-explored research area. Through several experiments and evaluations, we demonstrate the efficacy of our approach in analyzing and improving the robustness of black-box tabular classifiers.
['Ingrid Chieh Yu', 'Peyman Rasouli']
2021-12-13
null
null
null
20th-ieee-international-conference-on-machine-1
['counterfactual-explanation']
['miscellaneous']
[ 4.01622444e-01 4.87845570e-01 -3.33573312e-01 -2.23434046e-01 -3.61090243e-01 -9.98548329e-01 8.03000689e-01 2.42689952e-01 -9.52536985e-02 1.02361107e+00 -1.66376844e-01 -6.90936625e-01 -2.81054586e-01 -8.89910221e-01 -1.13785553e+00 -9.07708347e-01 -7.82141834e-02 1.53194815e-01 -4.36550565e-02 -7.03517720e-02 5.13952374e-01 7.83206701e-01 -1.70211875e+00 4.97481048e-01 8.45448077e-01 8.45524251e-01 -5.84367752e-01 4.47674930e-01 9.17752162e-02 5.05425632e-01 -1.09397984e+00 -8.01924527e-01 5.96558154e-01 -2.74446100e-01 -5.28819442e-01 3.71450000e-02 4.14331436e-01 -3.80821079e-02 -7.08334818e-02 1.21453416e+00 1.85675785e-01 -3.21678892e-02 6.15387917e-01 -1.88058341e+00 -4.27952617e-01 1.01284385e+00 -3.28616709e-01 5.85025223e-03 2.98386723e-01 4.73245621e-01 6.69341147e-01 -1.64097711e-01 2.91572899e-01 1.41190076e+00 6.66631222e-01 7.21723139e-01 -1.36726189e+00 -1.03853261e+00 4.18866783e-01 1.08924828e-01 -9.73353386e-01 -2.90390730e-01 7.55880117e-01 -3.34461600e-01 2.45772332e-01 7.97443688e-01 1.31345481e-01 1.53595555e+00 5.22309661e-01 5.20294666e-01 1.28739321e+00 -3.05512518e-01 6.51197433e-01 5.57077229e-01 -1.43317496e-02 2.28060469e-01 8.63297701e-01 6.62167370e-01 -1.46378294e-01 -7.54184842e-01 -8.17319751e-02 4.77687567e-02 -4.54030842e-01 -5.91340721e-01 -9.12324786e-01 9.19866920e-01 4.27973688e-01 6.46441206e-02 -2.83495754e-01 -1.27691301e-02 5.83432317e-01 2.75176942e-01 3.06794792e-01 9.45270836e-01 -7.16512203e-01 1.41630366e-01 -5.23892164e-01 3.61727744e-01 6.68546021e-01 5.91489434e-01 5.13821900e-01 5.21401465e-02 -4.03538793e-01 1.43405676e-01 -8.20987821e-02 5.34063935e-01 5.95729411e-01 -4.73703533e-01 4.29117411e-01 6.76859260e-01 1.65006742e-01 -1.17896092e+00 -8.96793902e-02 -7.19616711e-01 -6.72261834e-01 5.21866262e-01 6.73944771e-01 -1.55052170e-01 -7.01641321e-01 1.83394563e+00 3.66580456e-01 1.33050114e-01 5.11965081e-02 6.50760829e-01 2.61082407e-02 1.74767420e-01 3.66477706e-02 -2.42610499e-01 9.82118666e-01 -4.17709112e-01 -5.39647877e-01 -1.46739870e-01 7.50592232e-01 -2.32428551e-01 1.05434132e+00 4.71057236e-01 -5.77684522e-01 -3.13330621e-01 -1.36414194e+00 9.28728461e-01 -6.83696210e-01 -5.76564252e-01 4.69933420e-01 1.33309412e+00 -3.43052834e-01 8.41511488e-01 -3.96191955e-01 -4.78487760e-02 5.99020600e-01 3.24752688e-01 -4.89795893e-01 -5.08787902e-03 -1.35872078e+00 9.04823661e-01 4.78827327e-01 -7.03251585e-02 -1.08912563e+00 -8.64415109e-01 -5.34324944e-01 2.26243734e-01 7.10903406e-01 -5.18619537e-01 9.48182762e-01 -1.14263892e+00 -1.03585625e+00 3.09754819e-01 2.06824943e-01 -9.46890235e-01 9.72593009e-01 1.17016599e-01 -5.23600578e-01 -2.42764503e-01 -8.45481083e-02 9.39120799e-02 1.22620320e+00 -1.56016374e+00 -6.09574735e-01 -5.81806839e-01 3.90614152e-01 -2.85401553e-01 -2.56668568e-01 -3.71600956e-01 5.10711789e-01 -7.47503221e-01 -1.87653467e-01 -1.06968296e+00 -3.79428595e-01 -4.60080236e-01 -8.82537901e-01 3.14404488e-01 1.06029165e+00 -1.35509297e-01 1.13522923e+00 -1.98860645e+00 -3.75446260e-01 3.59309167e-01 -4.64777537e-02 3.46578747e-01 8.56677294e-02 2.32930750e-01 -4.91666406e-01 7.92167842e-01 -3.44557166e-01 9.33917761e-02 1.22781843e-02 1.01262927e-01 -8.52159142e-01 5.86774528e-01 1.21909007e-01 6.08574867e-01 -7.42171884e-01 1.13642983e-01 1.43668503e-01 -1.97074022e-02 -5.79758227e-01 1.03298284e-01 -2.40613058e-01 2.98801661e-01 -2.92913079e-01 4.39629316e-01 8.15976441e-01 3.63781780e-01 1.81377336e-01 4.16132361e-02 2.50017107e-01 1.15086444e-01 -1.21755040e+00 7.14606702e-01 -3.86607170e-01 3.84280831e-01 -3.80653709e-01 -1.14577067e+00 8.59251618e-01 2.42630109e-01 2.61344053e-02 -1.85867861e-01 3.14579606e-01 2.35490412e-01 2.10972264e-01 -2.05764085e-01 1.90342665e-01 -3.33584696e-01 -3.51207942e-01 5.59563458e-01 -4.10523802e-01 1.41669199e-01 -2.07285613e-01 -6.53958693e-02 1.08308601e+00 -2.59151578e-01 5.30380905e-01 -2.76568890e-01 6.19215727e-01 1.19893990e-01 6.78668082e-01 1.23491561e+00 -2.51827389e-01 3.70675206e-01 6.80253804e-01 -5.98453224e-01 -8.30623388e-01 -9.54482257e-01 -3.39701176e-01 4.52460200e-01 -1.90639980e-02 -1.38185732e-02 -8.28512192e-01 -1.62505782e+00 5.96397698e-01 1.36534286e+00 -1.19203985e+00 -8.78658175e-01 -1.16782255e-01 -9.45036113e-01 8.18946540e-01 2.20183820e-01 4.81787980e-01 -7.79963493e-01 -7.49962747e-01 1.38755590e-02 4.31755967e-02 -7.19807982e-01 -1.56239569e-01 2.33157545e-01 -8.91952515e-01 -1.49752033e+00 2.35835724e-02 2.77787119e-01 8.21560144e-01 1.37691140e-01 7.86365986e-01 8.35738331e-02 -4.15024683e-02 1.26592175e-03 -4.66064900e-01 -8.34846437e-01 -8.74070823e-01 -2.37972010e-02 4.75535035e-01 3.58239323e-01 2.10489050e-01 -4.71738160e-01 -2.18702942e-01 5.08607447e-01 -1.19233620e+00 -4.70764369e-01 4.80101109e-01 8.97802949e-01 1.39567211e-01 5.23985267e-01 8.57021928e-01 -1.34947562e+00 7.62683690e-01 -6.83485985e-01 -5.34679770e-01 4.38071489e-01 -8.92780006e-01 4.21722740e-01 1.23914981e+00 -9.29192483e-01 -8.83537054e-01 -1.80985808e-01 4.43093538e-01 -5.71999192e-01 -4.41162765e-01 2.34330937e-01 -6.30514443e-01 1.41549660e-02 9.74262774e-01 -6.57972461e-03 6.85189664e-02 -3.22753459e-01 4.05566961e-01 5.55113316e-01 4.54896331e-01 -5.99775732e-01 1.39317763e+00 5.57276785e-01 1.70911714e-01 -1.89908147e-01 -8.25788200e-01 2.71322310e-01 -4.71501112e-01 -1.30161956e-01 1.52867481e-01 -1.74333885e-01 -7.90621102e-01 1.69418275e-01 -8.07626367e-01 4.58936691e-02 -3.59519690e-01 1.13196470e-01 -3.85578901e-01 2.54041731e-01 4.12529051e-01 -1.10525250e+00 -3.64803672e-02 -1.17872787e+00 5.98179817e-01 1.40129462e-01 -3.48171085e-01 -9.47243154e-01 -1.48924589e-01 2.64659405e-01 2.82981604e-01 8.16560566e-01 1.12038958e+00 -1.40799129e+00 -3.05403948e-01 -8.20921898e-01 3.32970053e-01 3.54478121e-01 2.43042424e-01 1.51261136e-01 -1.19270957e+00 -4.90567774e-01 2.31065229e-01 7.22631589e-02 5.14999688e-01 1.08958967e-01 1.41005886e+00 -9.87228155e-01 -5.40456414e-01 5.49859524e-01 1.12627792e+00 4.63867962e-01 6.38941407e-01 7.84496844e-01 2.16963619e-01 8.04123998e-01 9.90101397e-01 4.68608320e-01 -2.62806267e-01 5.87183595e-01 1.00755346e+00 8.77098441e-02 5.08869171e-01 -4.62855846e-01 4.19132084e-01 -4.49918956e-01 4.64012355e-01 -2.23954409e-01 -6.95502579e-01 3.97141695e-01 -1.78837574e+00 -1.30147529e+00 2.21305508e-02 2.66791940e+00 4.07348275e-01 5.20824134e-01 7.90162235e-02 6.91998601e-01 8.95754457e-01 -3.83591540e-02 -7.63476789e-01 -8.14706445e-01 -1.52928352e-01 -3.70792486e-02 7.04437375e-01 2.66776234e-01 -1.02267563e+00 5.24221718e-01 5.79622030e+00 7.76069105e-01 -1.20283067e+00 -8.96823853e-02 8.67583573e-01 -3.10699314e-01 -4.33425754e-01 2.61533737e-01 -5.56546807e-01 7.32141197e-01 9.89944577e-01 -6.15136504e-01 3.40317756e-01 1.05165577e+00 3.22945267e-01 1.18583582e-01 -1.33998609e+00 3.75881404e-01 -8.96119326e-02 -1.39502370e+00 4.16556418e-01 2.12392092e-01 6.17289305e-01 -6.91347897e-01 3.77307951e-01 3.76758724e-01 3.77093077e-01 -1.13595462e+00 9.29744601e-01 3.61656845e-01 5.86493373e-01 -1.20994675e+00 1.14678776e+00 6.40783012e-01 -4.10732538e-01 -6.15154386e-01 -2.43256867e-01 -2.15322688e-01 -4.81299490e-01 5.54828942e-01 -1.17500877e+00 7.31424809e-01 6.35323107e-01 7.69967660e-02 -8.63348424e-01 7.71690130e-01 -3.20038557e-01 6.59191012e-01 -2.33855676e-02 -3.32383625e-02 1.76980235e-02 1.94082364e-01 8.62033963e-01 7.49673247e-01 4.88259755e-02 -2.59164006e-01 -2.36435205e-01 9.82570946e-01 3.50762829e-02 -1.04505420e-01 -1.06690013e+00 1.94024369e-01 8.05326581e-01 9.75459933e-01 -3.85093689e-01 -1.21531896e-01 1.43986970e-01 4.84357774e-01 5.51664596e-03 2.89994329e-01 -9.97601330e-01 -4.92912292e-01 8.88922334e-01 2.87156790e-01 -1.13586195e-01 4.58993584e-01 -6.55824661e-01 -1.05054116e+00 1.19522192e-01 -1.37724113e+00 5.31427681e-01 -4.14722234e-01 -1.26713109e+00 5.51735282e-01 2.29996279e-01 -1.37034738e+00 -3.02819401e-01 -4.01089907e-01 -8.03443134e-01 6.14760399e-01 -1.00809109e+00 -8.60933602e-01 -6.61673546e-02 3.59488070e-01 1.37056455e-01 -3.83698821e-01 6.32768452e-01 -4.50131238e-01 -7.59545803e-01 1.10102165e+00 6.55373782e-02 -1.17387377e-01 5.11192024e-01 -1.15430391e+00 3.02029073e-01 1.27260363e+00 -3.76599164e-05 8.55411053e-01 1.24631846e+00 -6.59007668e-01 -1.02532041e+00 -1.15730178e+00 4.62362885e-01 -6.79783881e-01 5.65866351e-01 -4.18892503e-01 -9.97703314e-01 6.36249363e-01 -2.20278502e-01 -1.59875944e-01 5.66214442e-01 1.66472822e-01 -5.37789702e-01 -2.53951043e-01 -1.76543057e+00 9.11070466e-01 7.29710996e-01 -2.03227296e-01 -7.66429543e-01 1.29345944e-02 6.86395347e-01 7.37346988e-03 -5.12445390e-01 6.76149547e-01 5.62463045e-01 -1.00316381e+00 7.66049325e-01 -1.20377219e+00 2.51088053e-01 -3.70370805e-01 -1.64542586e-01 -1.59811139e+00 1.64678134e-02 -6.93829179e-01 -4.61214595e-02 1.37158227e+00 5.51892936e-01 -1.12619960e+00 7.52790272e-01 8.57107580e-01 2.55794048e-01 -6.40921891e-01 -1.16900718e+00 -1.04474354e+00 2.85710245e-01 -5.42166889e-01 1.27835965e+00 1.14232111e+00 7.18764365e-02 -3.23706597e-01 -2.67199337e-01 5.38921714e-01 7.84355760e-01 1.22744426e-01 1.13521254e+00 -1.11150181e+00 -1.10539310e-01 -3.94836307e-01 -4.21289116e-01 -4.39325534e-02 4.38026220e-01 -8.10876787e-01 -1.89919576e-01 -6.34513140e-01 2.45650932e-02 -4.69732672e-01 -2.76778489e-01 5.96923590e-01 -4.06044364e-01 -6.43599480e-02 3.38910609e-01 4.36273701e-02 9.56899822e-02 4.04435635e-01 6.93008542e-01 -2.37853810e-01 -6.59711584e-02 4.05280024e-01 -1.12537575e+00 6.52372420e-01 9.74244475e-01 -1.01444447e+00 -3.49440157e-01 2.90905595e-01 1.83010567e-02 -1.84446111e-01 6.24311090e-01 -8.98891091e-01 -7.00892210e-02 -4.50512618e-01 2.18560323e-01 -2.02272102e-01 -3.02850515e-01 -1.18395650e+00 4.21809107e-01 9.10184801e-01 -5.89134753e-01 9.12923664e-02 3.65048856e-01 7.93631911e-01 4.79230694e-02 -4.18331802e-01 8.48788142e-01 9.25069153e-02 -1.74998378e-04 5.91805577e-02 -4.29304540e-01 -2.39281386e-01 1.45595789e+00 -3.81368399e-01 -6.78425789e-01 -3.08327764e-01 -4.17634219e-01 3.02879978e-03 6.84884012e-01 6.79546952e-01 3.18904072e-01 -1.17024612e+00 -5.75697422e-01 2.17111349e-01 2.67965168e-01 -3.89885813e-01 1.02749199e-01 4.95552063e-01 -8.65242109e-02 4.33993876e-01 -3.23030949e-01 -2.07922310e-01 -1.17736638e+00 1.13347518e+00 6.56001389e-01 -2.99777627e-01 -1.42873496e-01 3.46003175e-01 2.10017085e-01 -5.54130733e-01 2.96110809e-02 -1.31188437e-01 -2.04560012e-02 -3.47211547e-02 3.51747721e-01 4.29509223e-01 1.98938310e-01 -1.83129370e-01 -4.89904881e-01 -2.51349628e-01 -1.03003845e-01 4.54567373e-02 9.87731934e-01 1.61676899e-01 1.52135178e-01 3.19510311e-01 8.40323210e-01 3.01468879e-01 -1.10974276e+00 1.77459344e-01 3.25857848e-01 -8.32449079e-01 -3.05428803e-01 -1.12313807e+00 -8.53892624e-01 7.08488405e-01 4.18249518e-01 6.07101858e-01 1.08594489e+00 -4.83865142e-01 1.02373563e-01 1.21944264e-01 4.99708623e-01 -6.68505609e-01 -2.16523856e-01 -6.83897361e-02 9.41201329e-01 -1.16656458e+00 -8.38470235e-02 -2.77133137e-01 -5.75336397e-01 9.33916807e-01 6.50593221e-01 6.65286332e-02 2.84750640e-01 1.57086611e-01 -6.32654503e-03 2.50710875e-01 -1.03830528e+00 3.82051677e-01 2.31414199e-01 8.05545211e-01 -1.29539713e-01 2.60945141e-01 -3.06670010e-01 9.85121608e-01 -4.64512408e-01 -4.53525484e-01 9.52005029e-01 6.01550758e-01 -1.02750942e-01 -1.04140067e+00 -9.64106619e-01 5.55533290e-01 -5.33930779e-01 1.22249290e-01 -8.35153222e-01 1.18436253e+00 7.49978647e-02 1.32076478e+00 -2.70965248e-01 -7.66946793e-01 6.32208765e-01 3.46970022e-01 -1.30410790e-01 -4.93874848e-01 -9.61753726e-01 -7.60993302e-01 3.23763750e-02 -4.81919527e-01 1.95451841e-01 -8.51789653e-01 -7.31523693e-01 -5.47413290e-01 -4.70518053e-01 4.12585497e-01 6.40357614e-01 1.05893791e+00 4.04691547e-01 5.30539155e-01 1.07092369e+00 -4.28652316e-01 -1.27707648e+00 -8.32512438e-01 -3.98218095e-01 7.83908844e-01 3.55401546e-01 -8.94024670e-01 -1.01813853e+00 -4.28680569e-01]
[5.753055572509766, 7.646125316619873]
9a83d95a-5354-42d3-9f2b-e306c57188fe
towards-listening-to-10-people-simultaneously
2010.11871
null
https://arxiv.org/abs/2010.11871v2
https://arxiv.org/pdf/2010.11871v2.pdf
Towards Listening to 10 People Simultaneously: An Efficient Permutation Invariant Training of Audio Source Separation Using Sinkhorn's Algorithm
In neural network-based monaural speech separation techniques, it has been recently common to evaluate the loss using the permutation invariant training (PIT) loss. However, the ordinary PIT requires to try all $N!$ permutations between $N$ ground truths and $N$ estimates. Since the factorial complexity explodes very rapidly as $N$ increases, a PIT-based training works only when the number of source signals is small, such as $N = 2$ or $3$. To overcome this limitation, this paper proposes a SinkPIT, a novel variant of the PIT losses, which is much more efficient than the ordinary PIT loss when $N$ is large. The SinkPIT is based on Sinkhorn's matrix balancing algorithm, which efficiently finds a doubly stochastic matrix which approximates the best permutation in a differentiable manner. The author conducted an experiment to train a neural network model to decompose a single-channel mixture into 10 sources using the SinkPIT, and obtained promising results.
['Hideyuki Tachibana']
2020-10-22
null
null
null
null
['audio-source-separation']
['audio']
[ 8.48346129e-02 -1.25932127e-01 2.48955805e-02 -2.43126214e-01 -1.12309039e+00 -2.56380498e-01 -9.31237191e-02 -4.55959380e-01 -5.42055488e-01 6.98811233e-01 -1.78215444e-01 -3.92078012e-01 -3.75090748e-01 -6.07719183e-01 -8.61113906e-01 -8.27813447e-01 -3.22215736e-01 2.07873672e-01 -1.94398686e-03 -1.90571751e-02 -8.69431440e-03 2.38144472e-01 -1.69779921e+00 -1.40710339e-01 7.81331003e-01 1.25716889e+00 2.72017568e-01 6.96883321e-01 -1.42408058e-01 4.59268034e-01 -7.66051948e-01 -4.95794564e-01 8.17200840e-01 -9.22421277e-01 -3.48457903e-01 -4.00180221e-01 3.09922487e-01 1.36504546e-01 -4.97572720e-01 1.54750240e+00 8.75090361e-01 3.15747261e-01 4.68551934e-01 -1.34636080e+00 -6.73355535e-02 1.08227193e+00 -5.37755609e-01 2.35383123e-01 -2.70117015e-01 -1.05349012e-01 1.01110721e+00 -6.76734746e-01 -5.80227971e-02 9.94456589e-01 1.01990759e+00 3.32681507e-01 -1.10876346e+00 -1.24410164e+00 5.92595898e-02 1.58625260e-01 -1.60622025e+00 -6.30419850e-01 1.07855296e+00 -1.33209035e-01 7.81316340e-01 4.05240804e-01 6.17585242e-01 3.79439622e-01 -3.14412653e-01 8.25763464e-01 9.95072842e-01 -6.41165972e-01 3.19431812e-01 1.54157728e-01 -9.24902484e-02 7.30707586e-01 1.26757726e-01 1.37331024e-01 -7.40996122e-01 -8.59386548e-02 6.25057101e-01 -4.56669986e-01 -4.89361852e-01 -2.88883179e-01 -7.20636189e-01 7.27727354e-01 2.12892488e-01 2.90737957e-01 -1.11717373e-01 4.09244657e-01 1.53586015e-01 5.82704604e-01 2.53599733e-01 5.67879498e-01 -4.85506743e-01 -4.54799205e-01 -1.31588948e+00 -8.37314408e-03 1.01032257e+00 8.05243373e-01 7.73584545e-01 5.96680999e-01 3.55407178e-01 1.35319650e+00 3.59780222e-01 7.22806275e-01 4.61960524e-01 -1.18794465e+00 4.24316853e-01 8.63567665e-02 -2.12987915e-01 -9.17786539e-01 -1.48862824e-01 -8.59624863e-01 -9.20547307e-01 1.93515196e-01 6.50821686e-01 -4.09257382e-01 -7.68634379e-01 2.19844151e+00 4.02848981e-02 1.71805799e-01 -1.58096388e-01 8.54522943e-01 5.06665885e-01 6.35861516e-01 -5.39069474e-01 -4.37060982e-01 8.91000688e-01 -7.95681417e-01 -7.37846971e-01 -3.17880422e-01 2.42702842e-01 -8.80649328e-01 9.79706049e-01 5.97626507e-01 -1.40608847e+00 -3.84573191e-01 -1.18209827e+00 5.57178617e-01 -9.29875895e-02 1.35240287e-01 6.86606407e-01 1.23404133e+00 -1.07419109e+00 7.15982914e-01 -6.27650797e-01 2.37093896e-01 2.29937106e-01 6.00176990e-01 -1.17492201e-02 4.49244827e-02 -1.12128317e+00 4.60198551e-01 1.91580474e-01 4.29854125e-01 -8.31708908e-01 -6.60705030e-01 -7.05037594e-01 2.66481131e-01 2.43985608e-01 -2.92791039e-01 1.22163069e+00 -1.19643211e+00 -1.89097726e+00 4.25603300e-01 -3.31694573e-01 -6.18024826e-01 3.36147785e-01 1.56222507e-01 -4.79004294e-01 2.39803195e-02 -1.75423771e-01 4.44985032e-01 1.08512139e+00 -1.00203156e+00 -5.22164166e-01 -1.60340458e-01 -1.44762814e-01 6.32612482e-02 -1.59125164e-01 -1.21013895e-02 -5.32548666e-01 -7.54265308e-01 6.01509452e-01 -8.90855312e-01 -2.58006752e-01 -3.58644783e-01 -4.08338010e-01 -4.47956808e-02 2.37710297e-01 -5.62926173e-01 1.14218295e+00 -2.34940171e+00 -1.12692989e-01 6.30567610e-01 2.29977183e-02 3.26596528e-01 -2.08655849e-01 9.52608585e-02 -3.02750230e-01 -1.21965848e-01 -5.22612512e-01 -3.40427846e-01 1.83717921e-01 -5.24172336e-02 -9.75215137e-02 2.94059485e-01 -3.29183787e-01 2.81429619e-01 -5.01097977e-01 -2.07324550e-01 -1.36508957e-01 2.25253478e-01 -8.52718472e-01 4.45441566e-02 6.82836026e-02 -1.85926825e-01 1.46304205e-01 4.21599746e-01 9.22149479e-01 -6.22443594e-02 1.21072002e-01 -2.43115053e-01 -4.95360158e-02 4.29501295e-01 -1.59693420e+00 1.43822670e+00 -6.15884185e-01 1.04117942e+00 4.11809385e-01 -1.30884027e+00 9.09764826e-01 3.69902164e-01 4.48536307e-01 -4.77837831e-01 3.13252568e-01 5.81555665e-01 2.28174850e-01 -3.60169709e-02 2.33216435e-01 -5.77476263e-01 2.23822482e-02 5.49031377e-01 6.72483593e-02 -2.78211176e-01 1.90079644e-01 -2.74576787e-02 1.06557238e+00 -3.05924505e-01 -1.34379594e-02 -1.85765773e-01 3.66774142e-01 -3.54038984e-01 7.87871718e-01 9.08104181e-01 -3.09679598e-01 6.48189127e-01 5.58303356e-01 1.27904877e-01 -5.35207331e-01 -1.32082808e+00 1.72560051e-01 7.20404148e-01 3.44142281e-02 -4.96342033e-02 -9.58469629e-01 -3.81105602e-01 -3.11267644e-01 5.59876084e-01 -1.55515060e-01 -1.95678592e-01 -5.37668228e-01 -9.70930576e-01 7.55696833e-01 2.93240100e-01 9.07130778e-01 -7.73447156e-01 -1.99056461e-01 1.54936790e-01 -3.16767871e-01 -8.43814731e-01 -6.69827163e-01 7.39239454e-01 -7.13841975e-01 -8.23834479e-01 -8.26703131e-01 -1.03364086e+00 6.52288496e-01 1.62775442e-01 7.42619693e-01 -4.64580923e-01 7.73273408e-03 -4.64900630e-03 -5.34281842e-02 -4.41798210e-01 -2.54961640e-01 -1.17652146e-02 2.25087330e-01 2.05367319e-02 2.40483314e-01 -9.94184196e-01 -4.92826015e-01 3.29021335e-01 -6.24643207e-01 -3.76628906e-01 6.67138577e-01 8.71175528e-01 6.42883122e-01 7.72552252e-01 6.95933342e-01 -5.24898589e-01 7.26348400e-01 6.67879432e-02 -7.53587663e-01 -6.34987354e-02 -4.31854934e-01 2.94363260e-01 7.67539859e-01 -7.20317006e-01 -6.36045992e-01 -1.69422552e-02 -5.11459231e-01 -7.04124391e-01 3.24491978e-01 5.67924619e-01 -3.42325926e-01 -3.80708694e-01 5.97146809e-01 4.10777837e-01 4.69214516e-03 -5.22430956e-01 1.57939762e-01 7.09347725e-01 4.29765821e-01 -3.04526836e-01 7.11476803e-01 8.01893324e-02 -9.80895609e-02 -1.01611602e+00 -5.35138011e-01 -3.41297805e-01 1.38704002e-01 -2.52215937e-02 2.10036948e-01 -6.85350716e-01 -1.02204216e+00 5.22665083e-01 -7.06739306e-01 -4.16693091e-01 -6.32776380e-01 1.14669371e+00 -6.06608808e-01 3.24113786e-01 -3.17678899e-01 -1.02168536e+00 -1.99820146e-01 -1.13086832e+00 1.99408159e-01 3.64118554e-02 4.93256636e-02 -4.00785178e-01 7.64092356e-02 1.10766523e-01 4.96461391e-01 -3.75311255e-01 8.89491141e-01 -4.21443582e-01 -6.46920025e-01 -3.49317372e-01 -1.43873468e-01 9.18612480e-01 5.12618311e-02 -2.74799794e-01 -9.59407210e-01 -2.66665876e-01 5.20478487e-01 3.55732888e-02 7.02882886e-01 8.03538620e-01 1.18408287e+00 -3.54160070e-01 3.21046114e-02 9.05984879e-01 1.38030005e+00 8.12940359e-01 6.16170585e-01 -3.65237631e-02 1.91139653e-01 2.98767000e-01 5.68895740e-03 2.12283760e-01 3.66041698e-02 4.14652169e-01 2.09156618e-01 -2.74731126e-02 -1.95428953e-01 -1.67918146e-01 4.51826215e-01 1.44854832e+00 1.77111879e-01 -1.43469632e-01 -5.67445815e-01 4.92692888e-01 -1.20338511e+00 -9.06698406e-01 4.77202445e-01 2.30614352e+00 8.93096030e-01 3.37973475e-01 3.79264466e-02 6.50625944e-01 7.36720622e-01 2.39915267e-01 -5.45814574e-01 -4.62800652e-01 -1.02467984e-01 5.42671621e-01 7.10915804e-01 5.66495001e-01 -8.62542808e-01 6.29869819e-01 6.88619995e+00 1.45053482e+00 -1.18930864e+00 1.05705284e-01 5.16846061e-01 -4.49963152e-01 -2.62070954e-01 -2.21095562e-01 -6.94409430e-01 6.04455829e-01 9.67007756e-01 -2.23892331e-02 7.75432825e-01 7.38623142e-01 5.09501323e-02 -1.84932455e-01 -8.50269735e-01 1.47744012e+00 1.82078212e-01 -1.06102812e+00 -2.11430982e-01 -1.20966822e-01 4.34705675e-01 -1.51509317e-02 3.02908868e-01 2.93447971e-01 3.64366889e-01 -8.76512825e-01 5.69417536e-01 1.19025342e-01 5.95251620e-01 -1.04810250e+00 4.30397749e-01 4.19351459e-01 -1.15461266e+00 -4.66229767e-02 -3.23647380e-01 6.22863807e-02 1.52587205e-01 8.30698371e-01 -7.13238299e-01 2.23148555e-01 6.87385678e-01 4.89124954e-02 1.77711263e-01 1.34747553e+00 -1.08190745e-01 7.75478661e-01 -7.87544191e-01 -2.90022016e-01 9.84591022e-02 -2.45598719e-01 8.14721584e-01 1.05531240e+00 7.21791744e-01 -2.77952310e-02 -2.90221810e-01 6.18028760e-01 -3.15621883e-01 1.17176156e-02 -3.47599536e-01 -1.95925131e-01 5.55446923e-01 8.35711420e-01 -7.79682040e-01 9.22065005e-02 -9.00725201e-02 6.23270035e-01 -5.07860817e-02 3.96134645e-01 -8.87861252e-01 -1.07927525e+00 7.10554600e-01 -8.50223675e-02 5.53460777e-01 -2.92711049e-01 -9.31587592e-02 -8.26037288e-01 9.49170515e-02 -1.02958167e+00 6.88999519e-02 -6.27053618e-01 -1.05003572e+00 8.11923742e-01 -2.13109523e-01 -1.39220309e+00 -1.04797423e-01 -4.12767708e-01 -3.34258944e-01 8.17418337e-01 -1.30928302e+00 -2.99895525e-01 3.31690282e-01 6.01169109e-01 3.49531949e-01 -3.48025799e-01 5.43930709e-01 7.93972313e-01 -7.51195371e-01 1.28261733e+00 2.80093461e-01 5.65025732e-02 3.48859876e-01 -8.45526099e-01 -2.20386218e-02 1.09003317e+00 3.96182984e-01 7.38705218e-01 8.74862611e-01 2.22050697e-02 -1.07376420e+00 -7.15740919e-01 8.45180988e-01 3.90519351e-01 3.96394193e-01 -3.12191308e-01 -5.38447261e-01 3.26990157e-01 5.75418323e-02 -2.52878845e-01 8.34429860e-01 6.94184154e-02 -3.38458538e-01 -7.19348490e-01 -1.05240655e+00 7.02327907e-01 1.08874285e+00 -6.04716003e-01 -3.10760617e-01 8.56986195e-02 7.26630449e-01 -3.84770632e-01 -5.19306719e-01 3.65896434e-01 5.91562986e-01 -1.00939643e+00 9.88622665e-01 -2.00564824e-02 2.48617008e-02 -3.26632500e-01 -4.97703254e-01 -1.43885124e+00 1.57500297e-01 -1.05171168e+00 9.47402269e-02 9.20225918e-01 8.52217495e-01 -9.77089167e-01 9.86373067e-01 2.65915841e-01 -3.28527600e-01 -6.43783510e-01 -1.33543408e+00 -1.10769498e+00 -1.26135617e-01 -8.68168950e-01 6.01600230e-01 6.82518780e-01 -9.70302597e-02 -2.02599331e-03 -3.61831278e-01 1.87822714e-01 5.90358198e-01 5.70842735e-02 6.23149216e-01 -8.36865187e-01 -8.26250374e-01 -7.37393796e-01 -3.10616910e-01 -1.51156509e+00 8.20673257e-02 -8.81117940e-01 4.49872553e-01 -1.05550313e+00 -2.28932127e-01 -8.77477705e-01 -6.19568825e-01 1.64006516e-01 2.06631064e-01 2.67492294e-01 2.23354295e-01 -1.68907732e-01 -2.50482708e-01 6.10218525e-01 9.97316599e-01 -2.38683194e-01 -3.83257955e-01 4.49976563e-01 -8.10853362e-01 9.48058188e-01 8.37288618e-01 -7.23422706e-01 -5.30866146e-01 -3.48539591e-01 3.25013727e-01 2.41894618e-01 -1.05552390e-01 -1.41312516e+00 3.97461504e-01 1.84353918e-01 -6.38738573e-02 -6.90043807e-01 8.09611738e-01 -8.61175597e-01 2.93109387e-01 4.63696063e-01 -3.51890832e-01 -2.17845976e-01 1.87859163e-01 4.41109031e-01 -5.31133711e-01 -4.95444536e-01 8.41410041e-01 7.17893839e-02 -2.74515510e-01 2.29658023e-01 -2.45195866e-01 2.36074761e-01 5.45647562e-01 -2.62728840e-01 9.86403376e-02 -7.25543141e-01 -6.90653622e-01 -1.75233483e-01 -1.78422943e-01 -1.23085648e-01 5.40878356e-01 -1.20535016e+00 -3.07823002e-01 6.03696883e-01 -5.34980655e-01 -1.43126771e-01 3.09212744e-01 9.16739941e-01 -4.75688040e-01 2.53389776e-01 2.03083590e-01 -5.16495943e-01 -1.13941669e+00 1.64624795e-01 6.72827423e-01 -1.45371601e-01 -1.74124584e-01 1.52328861e+00 -2.88504232e-02 -4.24699932e-01 5.40329278e-01 -2.86413610e-01 1.13647066e-01 8.43955763e-03 3.55304331e-01 5.03487706e-01 1.38386667e-01 -2.56561667e-01 -2.33046234e-01 6.62487209e-01 3.22677732e-01 -6.90623581e-01 1.10354328e+00 -7.36261951e-03 -2.44866610e-01 3.32791090e-01 1.54456139e+00 3.56625408e-01 -1.16708159e+00 -3.34675908e-01 -4.09634024e-01 -6.25239193e-01 1.35855062e-03 -4.54241484e-01 -1.49008799e+00 9.34073091e-01 7.30314493e-01 3.18380564e-01 1.57767558e+00 -2.71293193e-01 9.85031068e-01 5.16311646e-01 4.28500950e-01 -1.06032646e+00 -6.03692085e-02 6.07292831e-01 5.15497446e-01 -7.95171797e-01 -3.45381021e-01 -2.38726899e-01 -2.98234552e-01 7.42393374e-01 3.48225594e-01 -1.53262138e-01 1.03533220e+00 4.96439457e-01 1.53273627e-01 1.65110275e-01 -3.40046436e-01 -1.13777258e-01 1.17605506e-02 4.28001136e-01 1.01506017e-01 1.45083755e-01 -2.73714185e-01 5.86334825e-01 -7.43859947e-01 -8.68663341e-02 2.05120757e-01 5.96200109e-01 -5.33750951e-01 -1.15686440e+00 -3.12269628e-01 5.89142680e-01 -5.94654143e-01 -3.40072215e-01 7.46548474e-02 5.38822055e-01 1.12366699e-01 1.04826093e+00 6.12451360e-02 -5.73061407e-01 3.47471565e-01 7.30541274e-02 4.92247373e-01 -1.30074024e-01 -4.69849706e-01 2.23497152e-01 3.53027396e-02 -5.07307053e-01 -5.56792021e-01 -5.42400777e-01 -1.09711492e+00 -3.64830554e-01 -4.75442737e-01 6.45994604e-01 8.20895672e-01 8.05179298e-01 8.83559734e-02 5.49758673e-01 8.85789156e-01 -4.64392364e-01 -5.11668026e-01 -7.91406512e-01 -9.32375193e-01 -1.48039460e-01 3.58751446e-01 -4.22326952e-01 -8.07180882e-01 -1.50038093e-01]
[15.242488861083984, 5.64982795715332]
bbf3e2fa-4961-4ebe-b393-712fa277e6fd
geometric-scene-parsing-with-hierarchical
1604.01931
null
http://arxiv.org/abs/1604.01931v2
http://arxiv.org/pdf/1604.01931v2.pdf
Geometric Scene Parsing with Hierarchical LSTM
This paper addresses the problem of geometric scene parsing, i.e. simultaneously labeling geometric surfaces (e.g. sky, ground and vertical plane) and determining the interaction relations (e.g. layering, supporting, siding and affinity) between main regions. This problem is more challenging than the traditional semantic scene labeling, as recovering geometric structures necessarily requires the rich and diverse contextual information. To achieve these goals, we propose a novel recurrent neural network model, named Hierarchical Long Short-Term Memory (H-LSTM). It contains two coupled sub-networks: the Pixel LSTM (P-LSTM) and the Multi-scale Super-pixel LSTM (MS-LSTM) for handling the surface labeling and relation prediction, respectively. The two sub-networks provide complementary information to each other to exploit hierarchical scene contexts, and they are jointly optimized for boosting the performance. Our extensive experiments show that our model is capable of parsing scene geometric structures and outperforming several state-of-the-art methods by large margins. In addition, we show promising 3D reconstruction results from the still images based on the geometric parsing.
['Xiaobai Liu', 'Ruimao Zhang', 'Liang Lin', 'Zhanglin Peng', 'Xiaodan Liang']
2016-04-07
null
null
null
null
['scene-labeling']
['computer-vision']
[ 6.19683623e-01 3.91607761e-01 4.54214513e-02 -6.38736188e-01 -1.07373261e+00 -3.02381903e-01 2.88116157e-01 4.60859202e-02 -5.37433242e-03 1.24331065e-01 2.14661345e-01 -3.24384391e-01 8.96443203e-02 -1.10067821e+00 -1.16086626e+00 -7.10890114e-01 -1.16453230e-01 4.68771428e-01 5.45549572e-01 1.75592210e-02 4.13019776e-01 5.88538110e-01 -1.59665990e+00 5.67934513e-01 7.57149458e-01 1.26655900e+00 7.59993017e-01 5.78070939e-01 -6.32335424e-01 1.00207484e+00 -1.51211426e-01 1.62849054e-02 -8.69917870e-02 1.45401740e-02 -1.00822198e+00 2.40910128e-01 7.18501210e-01 -2.90461421e-01 -8.20671991e-02 1.06465614e+00 1.58652648e-01 8.94956961e-02 3.14557105e-01 -5.78272223e-01 -6.13676906e-01 5.18902481e-01 -7.75340796e-01 -3.39567512e-01 1.43586129e-01 -5.40872753e-01 1.17549527e+00 -1.02316368e+00 4.70653385e-01 1.51080239e+00 4.77094710e-01 2.60301381e-01 -7.39844859e-01 -3.95903856e-01 7.82369256e-01 2.59695381e-01 -1.25225139e+00 -1.15963824e-01 9.99837637e-01 -4.40172672e-01 1.09946299e+00 2.70457962e-03 2.26458833e-01 8.00799787e-01 3.61818001e-02 9.98698473e-01 8.01057398e-01 -3.26036602e-01 -2.44833007e-02 -3.03777218e-01 3.57326448e-01 1.08778214e+00 -2.46171936e-01 -2.24600002e-01 -3.99979532e-01 1.25052363e-01 1.10971570e+00 -7.56697357e-02 -1.35812825e-02 -4.81420159e-01 -9.72156584e-01 5.89729190e-01 7.60983706e-01 2.49572605e-01 -3.51700932e-01 3.07168752e-01 1.53443739e-01 -2.35912830e-01 7.39008069e-01 1.16737716e-01 -6.84152722e-01 5.71443558e-01 -6.17699981e-01 -2.61591505e-02 2.92500496e-01 1.06826270e+00 9.55513299e-01 -4.84140851e-02 -3.17184217e-02 9.88179684e-01 5.43710291e-01 4.30256039e-01 -2.66365428e-02 -9.46239769e-01 7.78380275e-01 6.97225749e-01 -7.16295615e-02 -1.05776620e+00 -6.74583793e-01 -4.40313041e-01 -9.06488478e-01 -8.28930214e-02 1.05661638e-01 1.56255275e-01 -1.23450446e+00 1.74493742e+00 5.32001913e-01 5.33630133e-01 1.20980935e-02 8.09599876e-01 1.18860102e+00 9.97324467e-01 4.29858029e-01 1.33021593e-01 1.40875602e+00 -1.37986672e+00 -5.47564447e-01 -6.82253420e-01 6.96931243e-01 -7.61715174e-01 9.19365168e-01 1.40945837e-01 -1.22904956e+00 -7.75051057e-01 -9.44407225e-01 -6.54062629e-01 -4.85420018e-01 2.30977923e-01 7.03075767e-01 -2.16848888e-02 -1.09667051e+00 4.91560131e-01 -7.89424360e-01 -2.64955550e-01 3.22742879e-01 3.97169441e-01 -1.98680833e-01 -1.26970828e-01 -1.02181304e+00 4.50568587e-01 3.12827677e-01 5.20736754e-01 -6.73879266e-01 -2.76721805e-01 -1.21300185e+00 1.83122799e-01 4.83933300e-01 -7.55550504e-01 1.14145195e+00 -8.04727912e-01 -1.38594615e+00 1.14772320e+00 -4.34238702e-01 -2.10276216e-01 2.66290139e-02 -4.51734722e-01 -1.01988308e-01 -3.06233088e-03 1.62196040e-01 8.80681932e-01 6.95354581e-01 -1.73652911e+00 -7.84715235e-01 -6.81341946e-01 3.35430391e-02 5.09050429e-01 4.33432423e-02 7.19381089e-04 -8.44225526e-01 -6.10722542e-01 7.71931887e-01 -5.99502683e-01 -4.22572285e-01 -3.37733418e-01 -9.79577899e-01 -2.33858973e-01 8.75581682e-01 -9.13647830e-01 8.53918433e-01 -2.08077836e+00 2.88933545e-01 1.17221370e-01 5.37596494e-02 -4.27526757e-02 -2.81276673e-01 1.17218427e-01 -5.37910052e-02 7.80799389e-02 -4.39327478e-01 -7.16586351e-01 -1.15836121e-01 2.76486576e-01 -4.69564199e-01 2.30146185e-01 2.13396743e-01 9.96807694e-01 -5.82225800e-01 -6.51930392e-01 5.14301836e-01 5.08407652e-01 -2.76916951e-01 4.60464805e-01 -6.75275028e-01 5.94285548e-01 -8.43735695e-01 7.58588314e-01 8.30762565e-01 -5.75330973e-01 1.86126918e-01 -3.48818839e-01 -3.37742120e-01 4.26976532e-01 -1.09027326e+00 2.09107137e+00 -4.85728472e-01 1.96410641e-01 2.47209027e-01 -9.15143192e-01 1.02755523e+00 -5.46744978e-03 1.69155300e-01 -9.41876829e-01 1.08589485e-01 8.55469033e-02 -5.83635867e-01 -5.60006678e-01 5.59835851e-01 1.80041850e-01 -1.72766626e-01 1.69227868e-01 -3.05993091e-02 -8.13026875e-02 -2.60367215e-01 -7.69315138e-02 5.95317900e-01 5.11377454e-01 -1.93842277e-02 -2.56618142e-01 6.78116381e-01 -1.62960410e-01 6.57118738e-01 6.23799920e-01 2.60319024e-01 6.79616451e-01 3.02581161e-01 -4.84178960e-01 -8.70977223e-01 -9.04812217e-01 1.28108054e-01 1.23122108e+00 5.81579626e-01 -4.73569892e-02 -1.07109118e+00 -4.59260821e-01 -3.24740261e-01 6.96951330e-01 -6.18353546e-01 1.89165965e-01 -9.62174833e-01 -5.14846444e-01 3.22435945e-02 7.74562597e-01 7.34661162e-01 -1.22177517e+00 -4.30074871e-01 9.18889344e-02 -1.87961578e-01 -1.51422513e+00 -1.31659612e-01 2.40306526e-01 -9.04193521e-01 -1.01000345e+00 -3.33896846e-01 -1.30032885e+00 6.47414565e-01 5.07745564e-01 1.34420919e+00 6.44114614e-02 9.27197710e-02 2.17292234e-02 -2.07143098e-01 -1.12359628e-01 1.42219543e-01 2.79002368e-01 -5.51528454e-01 1.82096601e-01 -3.02877277e-02 -6.25820875e-01 -4.13257867e-01 2.21053466e-01 -7.54807293e-01 8.74301255e-01 7.15256453e-01 6.12265110e-01 1.19250882e+00 -5.38152568e-02 3.95704620e-02 -1.34387791e+00 -1.64285749e-01 -3.41152579e-01 -7.49065161e-01 6.17915571e-01 -6.44287989e-02 6.39832690e-02 6.11858606e-01 2.31911272e-01 -1.54901016e+00 2.66936570e-01 -4.78586346e-01 -3.74583215e-01 -4.93486673e-01 3.58333409e-01 -5.62642574e-01 -9.72502083e-02 -1.53260365e-01 1.57861829e-01 -8.87082040e-01 -7.63677299e-01 5.26691675e-01 2.79938847e-01 6.41321003e-01 -5.80617785e-01 5.61330557e-01 5.38535535e-01 1.25540093e-01 -8.22864473e-01 -1.50685275e+00 -2.21272439e-01 -9.70904768e-01 -4.50630449e-02 1.50807905e+00 -9.64465797e-01 -4.65053409e-01 7.52106845e-01 -1.52873516e+00 -7.10041940e-01 7.82987922e-02 9.69398171e-02 -4.59990323e-01 1.89702883e-01 -9.25152481e-01 -7.67200053e-01 -2.84880608e-01 -1.12449467e+00 1.77620411e+00 1.86595306e-01 3.77618641e-01 -1.11725819e+00 -2.23894492e-01 5.96811652e-01 4.68169898e-02 4.80817914e-01 1.39142263e+00 -1.63058594e-01 -1.15853024e+00 1.96491584e-01 -7.39521503e-01 -1.73009038e-02 -8.33207816e-02 -9.39653516e-02 -1.24026144e+00 1.47440165e-01 1.02795303e-01 -3.10914725e-01 1.16753805e+00 7.45335519e-01 1.65243626e+00 -5.40274149e-03 -3.18729013e-01 9.87013698e-01 1.37280190e+00 2.17299983e-01 6.21944427e-01 2.17484683e-01 1.46275902e+00 1.08202386e+00 6.75080121e-01 3.80489752e-02 8.45118403e-01 5.39869964e-01 6.92619801e-01 -7.05591798e-01 -9.90446135e-02 -5.47480702e-01 -1.81335118e-02 9.14495051e-01 2.16740832e-01 -4.06180799e-01 -9.50479627e-01 2.68384337e-01 -2.09506059e+00 -4.51153159e-01 -4.66848582e-01 1.79184031e+00 1.44258186e-01 2.35383719e-01 -5.32000721e-01 -2.66174167e-01 8.05345297e-01 7.02693641e-01 -6.69753551e-01 -3.06706548e-01 -3.38498592e-01 7.86671788e-02 6.79630101e-01 8.68730307e-01 -1.45854044e+00 1.49782681e+00 5.68577290e+00 9.21377480e-01 -8.63992989e-01 5.91157414e-02 1.17661142e+00 3.67247492e-01 -3.74596208e-01 -6.98823482e-02 -8.18960547e-01 3.78874689e-03 4.88094568e-01 7.49703884e-01 2.06674978e-01 7.40243733e-01 1.83975980e-01 -5.47986589e-02 -8.07909310e-01 8.78085196e-01 9.12866667e-02 -1.30661201e+00 2.66397178e-01 -1.97490871e-01 6.76357448e-01 2.79101491e-01 -9.32729840e-02 1.12413973e-01 3.19453239e-01 -9.87961590e-01 9.26761627e-01 6.14036143e-01 7.09609747e-01 -6.23613894e-01 5.83843231e-01 3.59883100e-01 -1.73070049e+00 1.03484094e-01 -2.74049252e-01 1.76752642e-01 5.28295815e-01 6.80192173e-01 -2.91498099e-03 7.79040933e-01 8.24541450e-01 9.44373786e-01 -4.95526522e-01 4.17718470e-01 -4.70555812e-01 2.36716524e-01 -1.56423643e-01 3.88324827e-01 5.41245937e-01 -4.56381321e-01 6.58542812e-02 1.01634526e+00 2.45176345e-01 2.55024642e-01 4.62000728e-01 9.64632571e-01 -1.31467029e-01 3.23229022e-02 -5.13334274e-01 2.52664536e-01 1.69749156e-01 1.28616464e+00 -9.21294749e-01 -4.54712540e-01 -5.55116117e-01 9.74391699e-01 6.29197836e-01 4.62586790e-01 -6.90117359e-01 7.45326281e-02 3.71366680e-01 -9.39531699e-02 4.35730457e-01 -4.00855064e-01 -6.35238409e-01 -9.74093139e-01 5.06165735e-02 -2.84505248e-01 4.49215829e-01 -1.04145563e+00 -1.07910490e+00 7.60158718e-01 -2.13047042e-01 -8.54262888e-01 2.22246721e-01 -6.47178888e-01 -6.33670688e-01 7.14030087e-01 -1.93030953e+00 -1.68279314e+00 -4.09081072e-01 5.03876567e-01 7.80584514e-01 4.49159354e-01 6.68205976e-01 3.10209781e-01 -7.98719585e-01 4.86805402e-02 -2.62516171e-01 4.90587391e-02 1.61253855e-01 -1.25835311e+00 8.34216893e-01 8.53258848e-01 1.51012555e-01 1.92496613e-01 2.74622232e-01 -7.18272567e-01 -1.31099510e+00 -1.47816896e+00 1.07767284e+00 -1.27480656e-04 3.12173039e-01 -7.29619205e-01 -9.19055402e-01 8.51185918e-01 -8.64023492e-02 -2.09697515e-01 4.51397508e-01 1.78004041e-01 -3.71493846e-01 6.89143166e-02 -8.02128315e-01 4.97565478e-01 1.39257526e+00 -6.14807844e-01 -2.18163654e-01 4.78687227e-01 1.23382890e+00 -6.26827121e-01 -6.39799953e-01 8.32389534e-01 1.22215137e-01 -1.21031404e+00 1.18836260e+00 -2.10041717e-01 4.44550574e-01 -2.24032894e-01 -5.39008141e-01 -6.13719821e-01 -4.02238518e-01 -5.94048761e-02 3.55790593e-02 1.26760423e+00 4.29235607e-01 -3.36004257e-01 1.05713344e+00 4.96017486e-01 -7.00798094e-01 -1.00924623e+00 -6.23695135e-01 -2.62632847e-01 -5.96252531e-02 -6.43951237e-01 5.93959928e-01 8.57562959e-01 -7.99296200e-01 6.23251617e-01 -4.43706751e-01 7.19242454e-01 7.46767759e-01 7.56999552e-01 5.40047765e-01 -1.03797507e+00 -2.57923186e-01 -4.72771794e-01 3.84686925e-02 -1.69967091e+00 3.41354907e-01 -6.71588719e-01 2.17129499e-01 -2.06643510e+00 1.29328936e-01 -5.27574122e-01 -2.85083652e-01 4.45134401e-01 -2.32679784e-01 -5.97979054e-02 -9.65197459e-02 -5.92859536e-02 -9.16494668e-01 6.90362930e-01 1.29080129e+00 -1.90198004e-01 -2.33851179e-01 -1.80555329e-01 -5.37745118e-01 1.11470747e+00 6.74330294e-01 -1.76051676e-01 -3.25372338e-01 -1.12641895e+00 1.23655304e-01 3.33901912e-01 3.47350806e-01 -8.03946197e-01 2.90112019e-01 -8.00502524e-02 2.60718346e-01 -1.17256975e+00 5.48104942e-01 -5.62297881e-01 -7.24826008e-02 1.23853989e-01 -2.97102004e-01 -4.74255234e-02 1.29712060e-01 4.62960005e-01 -4.12809968e-01 5.08434139e-02 6.44409060e-01 -3.89385551e-01 -1.12531734e+00 5.57356715e-01 4.49307635e-02 -1.62159890e-01 7.03286052e-01 -5.52313738e-02 -2.25181594e-01 -9.53699574e-02 -6.47080600e-01 4.34944719e-01 3.35977137e-01 3.73840392e-01 7.03318834e-01 -1.15215743e+00 -3.73378992e-01 2.77456284e-01 -3.59124690e-02 8.51534963e-01 7.78417706e-01 3.97522926e-01 -6.66490376e-01 5.00430822e-01 1.91032782e-01 -7.56372750e-01 -1.19384384e+00 4.57564503e-01 3.29056263e-01 -4.74812746e-01 -7.18055546e-01 1.29776013e+00 1.00036883e+00 -7.18915403e-01 4.56000775e-01 -6.27844274e-01 -3.33169967e-01 -2.35157982e-01 1.62128821e-01 2.04699207e-02 3.13014537e-02 -7.80208707e-01 -2.22983330e-01 1.34042704e+00 3.70924436e-02 2.83653945e-01 1.44790113e+00 -2.68530369e-01 -5.35518527e-01 5.86007535e-01 1.16178179e+00 -3.56717646e-01 -1.32568359e+00 -5.28760731e-01 1.88076764e-01 -2.44961992e-01 1.25367299e-01 -4.77522999e-01 -1.23687291e+00 1.25141096e+00 3.91278088e-01 1.70734882e-01 1.23587286e+00 2.44890153e-01 1.11167479e+00 3.45525324e-01 4.37143117e-01 -8.85927796e-01 -1.11512192e-01 8.37565303e-01 6.17339015e-01 -1.12147927e+00 -1.87714174e-01 -1.05562401e+00 -2.80879438e-01 1.00329304e+00 7.49128461e-01 -1.83395639e-01 6.45320773e-01 1.03732504e-01 -1.22553448e-03 -6.46927059e-01 -5.74416757e-01 -3.72352868e-01 4.89831746e-01 9.49785560e-02 3.73984665e-01 1.90070510e-01 3.07422847e-01 4.19011623e-01 -4.87795025e-02 -6.05685532e-01 -5.66333830e-02 6.53360367e-01 -6.37180746e-01 -1.02958548e+00 -2.51252055e-01 2.74952263e-01 -2.60603815e-01 -1.71344280e-01 -4.48339343e-01 5.45061886e-01 4.13641930e-01 7.43987799e-01 3.63898993e-01 -4.32891726e-01 3.01756501e-01 -1.89244092e-01 3.65310460e-01 -7.49136806e-01 -3.42366070e-01 3.77984375e-01 1.96166828e-01 -9.02325273e-01 -4.70284075e-01 -4.31880414e-01 -1.50389361e+00 3.20789553e-02 -2.72617966e-01 -1.24065235e-01 6.30129158e-01 1.18952167e+00 3.28051209e-01 8.43667209e-01 6.48605585e-01 -1.18826938e+00 3.11449021e-01 -7.03391969e-01 -5.68804204e-01 2.61431992e-01 2.42323369e-01 -5.78415096e-01 3.25845033e-02 -1.32788137e-01]
[9.581010818481445, 0.3583662807941437]
d8d38ee5-358d-4334-bd9d-7f3b96fa6f8d
interactive-fashion-content-generation-using
2306.05182
null
https://arxiv.org/abs/2306.05182v1
https://arxiv.org/pdf/2306.05182v1.pdf
Interactive Fashion Content Generation Using LLMs and Latent Diffusion Models
Fashionable image generation aims to synthesize images of diverse fashion prevalent around the globe, helping fashion designers in real-time visualization by giving them a basic customized structure of how a specific design preference would look in real life and what further improvements can be made for enhanced customer satisfaction. Moreover, users can alone interact and generate fashionable images by just giving a few simple prompts. Recently, diffusion models have gained popularity as generative models owing to their flexibility and generation of realistic images from Gaussian noise. Latent diffusion models are a type of generative model that use diffusion processes to model the generation of complex data, such as images, audio, or text. They are called "latent" because they learn a hidden representation, or latent variable, of the data that captures its underlying structure. We propose a method exploiting the equivalence between diffusion models and energy-based models (EBMs) and suggesting ways to compose multiple probability distributions. We describe a pipeline on how our method can be used specifically for new fashionable outfit generation and virtual try-on using LLM-guided text-to-image generation. Our results indicate that using an LLM to refine the prompts to the latent diffusion model assists in generating globally creative and culturally diversified fashion styles and reducing bias.
['Nevasini Sasikumar', 'Krishna Sri Ipsit Mantri']
2023-05-15
null
null
null
null
['virtual-try-on']
['computer-vision']
[ 1.43163234e-01 2.88701300e-02 2.37122215e-02 -4.12698686e-01 -3.80123019e-01 -7.74553895e-01 7.24650800e-01 -1.42127842e-01 2.42571846e-01 4.03868556e-01 7.33399749e-01 2.31490582e-02 6.17207475e-02 -9.45963085e-01 -5.87116599e-01 -4.70685333e-01 3.88864160e-01 5.10215878e-01 -2.82892555e-01 -3.17971021e-01 3.69622223e-02 1.54159978e-01 -1.69152796e+00 6.09160542e-01 6.61219597e-01 7.06154943e-01 4.78268564e-01 8.49245071e-01 -3.52180541e-01 6.48526371e-01 -6.38348103e-01 -5.06683469e-01 1.06470913e-01 -7.34824836e-01 -2.56009966e-01 2.08965257e-01 4.83856261e-01 -3.00027966e-01 -6.92353537e-03 8.33518267e-01 6.57956541e-01 1.90672174e-01 6.99873865e-01 -1.17829645e+00 -1.61386526e+00 7.54532456e-01 -2.75637448e-01 -3.22064847e-01 7.79918671e-01 6.39739037e-01 5.61053336e-01 -7.04768121e-01 1.14899290e+00 1.63700747e+00 5.96393764e-01 6.42011225e-01 -1.85066640e+00 -3.88228834e-01 1.81224585e-01 -4.98951599e-02 -9.80273664e-01 -4.63095307e-01 1.06510854e+00 -5.28380811e-01 6.79347575e-01 7.73698032e-01 1.34183013e+00 1.70711887e+00 2.95439720e-01 8.02608848e-01 1.33294058e+00 -3.76006067e-01 4.22828704e-01 3.32087815e-01 -4.63291824e-01 5.76606870e-01 -2.50833273e-01 1.35444254e-01 -5.28162241e-01 -1.95344687e-01 1.18380368e+00 -1.26730219e-01 2.62107402e-01 -1.80761889e-01 -1.12749040e+00 1.02375889e+00 3.60145241e-01 1.99901260e-04 -5.66497028e-01 3.81151587e-01 1.70426201e-02 2.25781426e-01 7.14730442e-01 6.39412463e-01 -8.51957276e-02 -2.84562111e-01 -8.21966827e-01 6.66333318e-01 6.36192024e-01 1.01082134e+00 3.55519742e-01 3.05214703e-01 -4.97810066e-01 8.45787704e-01 3.81441921e-01 6.29715502e-01 1.86670259e-01 -1.21911812e+00 -3.00863534e-01 3.88315856e-01 1.76030129e-01 -1.10996127e+00 -7.34598860e-02 -3.16180795e-01 -6.53174818e-01 4.27410394e-01 1.65119946e-01 -2.08446383e-01 -1.24012828e+00 1.67763019e+00 2.35530168e-01 -1.67123973e-01 -4.48596627e-01 1.01891637e+00 1.00772619e+00 9.33622658e-01 3.67367089e-01 8.00519343e-03 1.32561517e+00 -7.09316075e-01 -8.65872204e-01 -2.66149198e-03 -4.05778401e-02 -1.23195231e+00 1.42042220e+00 3.39279592e-01 -1.34762251e+00 -8.25027883e-01 -6.76922381e-01 -1.11201212e-01 -3.04588586e-01 -9.55262631e-02 8.77241313e-01 6.88167989e-01 -1.22193503e+00 5.47647417e-01 -5.79953432e-01 -5.18784225e-01 3.04405481e-01 -1.39702111e-01 1.46716405e-02 2.37553590e-03 -8.76444757e-01 5.72146833e-01 -2.10435674e-01 -8.20530951e-02 -6.95740879e-01 -1.01375794e+00 -8.09481502e-01 -2.43933544e-01 -7.48847574e-02 -1.34450197e+00 1.01106322e+00 -1.21140122e+00 -1.74998784e+00 6.94237173e-01 -3.00710332e-02 4.43220064e-02 6.10303462e-01 -5.77070154e-02 -3.91304165e-01 -1.63796648e-01 4.52325977e-02 1.21846068e+00 1.16792774e+00 -1.56187880e+00 -6.15364313e-02 -2.68380567e-02 -4.02347185e-02 1.77927509e-01 7.21664131e-02 -5.92159145e-02 -2.24382266e-01 -1.15136111e+00 -8.74623507e-02 -1.01055062e+00 -5.19555032e-01 2.19084322e-01 -6.15826726e-01 1.03509083e-01 9.26711619e-01 -8.62337530e-01 1.05015957e+00 -2.15983295e+00 2.43391737e-01 3.05090547e-01 1.75739065e-01 -2.36121073e-01 -2.59291589e-01 5.42872310e-01 1.68774143e-01 3.12937111e-01 2.75292665e-01 -4.74275142e-01 3.74154150e-01 2.14377150e-01 -2.11770296e-01 -2.02021658e-01 1.21995900e-02 1.24252009e+00 -1.01282680e+00 -1.98691413e-01 5.46889126e-01 9.58699822e-01 -8.20100427e-01 4.80861291e-02 -6.35095298e-01 8.14601958e-01 -1.83376074e-02 3.52066398e-01 7.38857508e-01 -1.30237892e-01 2.45963693e-01 -5.78578293e-01 -1.34561986e-01 -7.89979100e-02 -1.34924316e+00 1.83265650e+00 -6.06625319e-01 6.74052298e-01 -9.76809189e-02 1.39125437e-03 1.08012021e+00 -3.82490717e-02 2.49821931e-01 -8.42348576e-01 -8.63111168e-02 -1.78898439e-01 -4.17142391e-01 -5.75870335e-01 8.70258927e-01 -2.78516024e-01 -1.69327840e-01 5.86746335e-01 -6.02634363e-02 -4.64291126e-01 4.39999364e-02 2.44760826e-01 6.77969217e-01 5.43410838e-01 -2.36257970e-01 -6.55856952e-02 -5.21565676e-01 1.05634693e-03 1.43545136e-01 7.60788143e-01 4.93966997e-01 7.80329704e-01 1.06226265e-01 -6.89773023e-01 -1.34697402e+00 -1.52339411e+00 3.44661057e-01 9.66540635e-01 -4.21672314e-03 -7.23433077e-01 -6.63986981e-01 -2.02633560e-01 -9.48534161e-02 1.16005850e+00 -7.56538868e-01 -2.27670923e-01 -2.43526235e-01 -6.74982309e-01 -1.87456235e-02 4.46404874e-01 2.58334398e-01 -1.19963884e+00 -4.64092404e-01 2.95550972e-01 -2.28392258e-01 -4.65109974e-01 -7.29595423e-01 -1.64870515e-01 -7.56129801e-01 -3.61011297e-01 -8.45642865e-01 -7.00427651e-01 7.50738204e-01 9.62300077e-02 1.36663032e+00 -3.42505962e-01 -6.48055017e-01 5.42673290e-01 -2.30049014e-01 -2.57383227e-01 -8.82581294e-01 -3.56750846e-01 -1.84430659e-01 -6.50039911e-02 3.78568657e-02 -6.99214876e-01 -1.06168389e+00 2.88372487e-01 -1.02899718e+00 7.82639384e-01 2.41795599e-01 4.43738371e-01 7.29287744e-01 4.51016091e-02 2.34630048e-01 -7.69783378e-01 1.21090209e+00 -4.48823512e-01 -8.38636756e-02 4.08808626e-02 -4.45830762e-01 1.97191700e-01 4.18702126e-01 -9.75998759e-01 -1.23455346e+00 -3.25852148e-02 -1.91021770e-01 -3.62755716e-01 -2.50988126e-01 3.69310051e-01 -5.88892139e-02 3.37037385e-01 6.16601288e-01 -2.92307008e-02 -6.61739409e-02 -7.23121345e-01 1.20061707e+00 1.71265990e-01 3.42239261e-01 -5.78274608e-01 6.06823087e-01 4.84533906e-01 -2.93293774e-01 -6.30417585e-01 -2.30420008e-01 2.46895701e-01 -2.58867145e-01 -6.74979568e-01 9.97088671e-01 -5.89447975e-01 -6.85466766e-01 3.72023195e-01 -9.00729299e-01 -6.64099872e-01 -1.09140480e+00 1.61778972e-01 -6.65101051e-01 -2.52930880e-01 -6.71274900e-01 -6.87520623e-01 -4.36805077e-02 -9.82180774e-01 1.14648879e+00 4.43681389e-01 -1.01428974e+00 -1.17909360e+00 1.51876867e-01 8.02288428e-02 9.75603938e-01 5.25833309e-01 1.12554634e+00 2.45956853e-01 -5.29107451e-01 4.82588373e-02 3.79633792e-02 5.13578020e-02 3.74099076e-01 2.83333987e-01 -7.63578892e-01 6.02356941e-02 -3.25067043e-01 5.02034947e-02 5.40773034e-01 9.09996808e-01 9.60622787e-01 -6.49786770e-01 -2.28425801e-01 5.81177235e-01 1.34205794e+00 9.35100913e-02 8.09311569e-01 -9.17008519e-03 7.62972891e-01 5.37431896e-01 5.02293296e-02 6.03962660e-01 4.75412160e-01 8.18556130e-01 5.07273190e-02 -5.33422112e-01 -7.40166605e-01 -9.61371124e-01 2.09707171e-01 7.15102911e-01 -2.98560649e-01 -1.47067562e-01 -1.45413637e-01 3.71724278e-01 -1.64499831e+00 -1.33408427e+00 -3.51482838e-01 1.87901473e+00 8.35793674e-01 3.29498649e-02 4.70828444e-01 -1.92848027e-01 5.01623333e-01 7.79682994e-02 -5.90912700e-01 -8.48437071e-01 -3.37699920e-01 2.60872394e-01 2.00975895e-01 5.59178889e-01 -7.14821935e-01 8.27979743e-01 7.48838854e+00 9.21885014e-01 -1.11020494e+00 4.75733802e-02 8.57842386e-01 -3.04688066e-01 -1.25685561e+00 1.02681639e-02 -4.91657495e-01 5.92291653e-01 7.03415275e-01 -2.59621125e-02 6.82518899e-01 7.67912865e-01 7.01828778e-01 -1.37418136e-01 -8.54735851e-01 1.16139388e+00 -2.78249439e-02 -1.76472807e+00 4.67236727e-01 3.49834263e-02 9.91894901e-01 -8.04032743e-01 7.73203373e-01 -6.91024512e-02 8.86443853e-01 -1.10243165e+00 1.24305689e+00 1.05604243e+00 9.59862530e-01 -5.28443873e-01 -3.89609113e-02 -9.43531916e-02 -1.07594323e+00 1.44206211e-01 -1.25873923e-01 -5.58369756e-02 7.15538442e-01 6.15251005e-01 -6.77620530e-01 3.16866208e-03 5.06340802e-01 4.36067402e-01 -5.60432792e-01 8.28759611e-01 -1.41887978e-01 5.31450868e-01 -2.36394495e-01 8.65565892e-03 -2.00617045e-01 -2.47409239e-01 7.42364526e-01 1.35421968e+00 6.66623056e-01 -2.76000500e-01 6.19030744e-02 1.45667088e+00 2.40317971e-01 1.17377363e-01 -5.54522455e-01 -1.83779076e-01 1.53972730e-01 1.32924628e+00 -9.88741040e-01 -1.75457776e-01 1.30094901e-01 1.46899343e+00 -3.46705794e-01 5.59264123e-01 -7.88906932e-01 1.07670799e-01 5.98117709e-01 5.37153959e-01 1.52377501e-01 -3.75071615e-01 -3.72679621e-01 -8.44899237e-01 -3.70844603e-01 -9.03526545e-01 -7.74108320e-02 -1.48922265e+00 -1.60110950e+00 5.01503706e-01 1.57749921e-01 -8.26925039e-01 -1.88762918e-01 -8.43024105e-02 -6.04473472e-01 9.27452981e-01 -6.15945160e-01 -1.60912776e+00 -4.52701330e-01 3.44249308e-01 8.28425288e-01 1.85683668e-01 9.88487005e-01 1.23702288e-01 -6.45218929e-03 2.47553170e-01 -6.55879751e-02 -4.25401032e-01 6.83874190e-01 -1.34916341e+00 8.32908988e-01 2.98037142e-01 2.02455193e-01 6.39985681e-01 1.16319001e+00 -1.00057840e+00 -1.36559761e+00 -8.16269040e-01 5.59529603e-01 -4.82457370e-01 2.26891294e-01 -5.99651337e-01 -2.83581614e-01 4.62865233e-01 5.06795406e-01 -6.53834105e-01 8.45321953e-01 4.78924327e-02 -8.35521147e-02 1.42937332e-01 -1.20070744e+00 1.00164890e+00 1.30912113e+00 -3.27724040e-01 -2.32989043e-02 1.27472728e-01 7.24176168e-01 -2.80769974e-01 -7.72225618e-01 -1.64931715e-01 8.85728061e-01 -8.79298508e-01 1.06888986e+00 -4.75312442e-01 5.71503401e-01 -2.31699914e-01 -1.26104370e-01 -1.84747612e+00 -8.49866033e-01 -1.11170065e+00 -1.17077142e-01 1.44713795e+00 3.24348181e-01 -6.23063259e-02 7.19108224e-01 7.93248355e-01 9.42751020e-02 -3.71109486e-01 -3.49034369e-01 -4.13349360e-01 -3.74280006e-01 -6.07731104e-01 8.94779861e-01 8.60197246e-01 -4.63476449e-01 2.29560792e-01 -6.63878024e-01 -3.88492137e-01 6.45321786e-01 2.00602204e-01 8.53038669e-01 -1.16360343e+00 -4.86929059e-01 -4.35581267e-01 -2.15402916e-01 -1.04127717e+00 -4.88005072e-01 -7.98300683e-01 -1.90225229e-01 -1.98913693e+00 2.00394824e-01 -6.72161341e-01 2.41800159e-01 1.02862217e-01 -7.10052699e-02 5.89495242e-01 5.29278159e-01 -1.20578590e-03 -3.50039124e-01 3.87819260e-01 1.70583785e+00 -2.10949630e-01 -6.17958784e-01 -2.16459736e-01 -1.10726535e+00 6.18280947e-01 5.58263481e-01 -3.56423855e-01 -6.32853568e-01 -4.29592520e-01 7.97298372e-01 -1.89350575e-01 7.22928286e-01 -7.84144402e-01 -1.71648517e-01 -3.90013576e-01 8.07173073e-01 -3.28129232e-01 4.39024001e-01 -5.76558352e-01 1.29003704e+00 1.95409819e-01 -4.71954048e-01 1.82806626e-01 1.89657941e-01 4.61678624e-01 3.29938859e-01 2.27589533e-01 4.82040644e-01 -3.47482145e-01 -6.72403932e-01 -4.78075221e-02 -6.55966282e-01 -2.89875060e-01 7.31278181e-01 -4.01879072e-01 -1.93714276e-01 -1.02771866e+00 -1.29755735e+00 -3.42081398e-01 7.58607507e-01 8.38510633e-01 7.32112467e-01 -1.82059014e+00 -6.65290773e-01 4.39442337e-01 -2.02108026e-01 -4.94167715e-01 6.98901236e-01 3.39879870e-01 -3.82253796e-01 -2.02783257e-01 -4.30684298e-01 -4.92241532e-01 -1.15163064e+00 5.54942012e-01 -9.36929137e-02 6.67068816e-04 -6.35586202e-01 8.79630446e-01 2.54492730e-01 -2.99867958e-01 -2.57204622e-01 -4.19761717e-01 1.14250399e-01 1.23602979e-01 4.43393499e-01 3.75115246e-01 -4.11002278e-01 -4.64044988e-01 1.26816109e-01 4.70430821e-01 2.20758319e-01 -4.32725042e-01 1.46616554e+00 -3.82858098e-01 2.34662697e-01 6.42405272e-01 7.51904368e-01 3.68646801e-01 -1.69903255e+00 2.43254066e-01 -7.00886607e-01 -7.32987761e-01 -8.69894493e-03 -1.24993062e+00 -9.46086347e-01 4.97019827e-01 8.54095936e-01 4.32689488e-01 1.04437470e+00 2.42211282e-01 9.51301515e-01 -6.09749973e-01 2.50888586e-01 -1.14382637e+00 4.17083293e-01 -1.80637851e-01 1.22970855e+00 -7.25510836e-01 -1.52390093e-01 -3.03509623e-01 -9.08230662e-01 9.82276738e-01 2.35090107e-01 -8.52037370e-02 6.14087343e-01 4.44789231e-01 1.79767266e-01 -2.80593783e-01 -6.72713339e-01 8.20216443e-03 5.17329872e-01 1.01405358e+00 4.16090876e-01 3.90183479e-01 -2.72680819e-02 5.22691786e-01 -5.03271639e-01 2.12033793e-01 1.44793391e-01 7.59464443e-01 -1.75681978e-01 -1.33771861e+00 -4.27939802e-01 3.63362849e-01 2.56819520e-02 -4.92466390e-02 -3.22678179e-01 3.18271220e-01 4.29802388e-01 9.67503369e-01 2.41798967e-01 -4.71535832e-01 3.16562146e-01 -4.89369072e-02 8.21715057e-01 -3.84709537e-01 -5.09688199e-01 3.36349308e-01 1.80483907e-01 -5.53438127e-01 -3.23377401e-01 -5.56611300e-01 -5.88888288e-01 -7.75008976e-01 9.73289683e-02 -1.95905700e-01 9.80507195e-01 3.98801267e-01 7.07243860e-01 7.75031626e-01 3.25288385e-01 -1.16379714e+00 2.05362767e-01 -7.51349986e-01 -4.17262673e-01 6.75629795e-01 -8.64579007e-02 -4.17813629e-01 4.69721444e-02 6.24602854e-01]
[11.45038890838623, -0.2499590516090393]
e7f3e41d-469b-4020-81c2-8b346020b7bc
analyzing-input-and-output-representations-1
1903.03369
null
https://arxiv.org/abs/1903.03369v4
https://arxiv.org/pdf/1903.03369v4.pdf
Analyzing Input and Output Representations for Speech-Driven Gesture Generation
This paper presents a novel framework for automatic speech-driven gesture generation, applicable to human-agent interaction including both virtual agents and robots. Specifically, we extend recent deep-learning-based, data-driven methods for speech-driven gesture generation by incorporating representation learning. Our model takes speech as input and produces gestures as output, in the form of a sequence of 3D coordinates. Our approach consists of two steps. First, we learn a lower-dimensional representation of human motion using a denoising autoencoder neural network, consisting of a motion encoder MotionE and a motion decoder MotionD. The learned representation preserves the most important aspects of the human pose variation while removing less relevant variation. Second, we train a novel encoder network SpeechE to map from speech to a corresponding motion representation with reduced dimensionality. At test time, the speech encoder and the motion decoder networks are combined: SpeechE predicts motion representations based on a given speech signal and MotionD then decodes these representations to produce motion sequences. We evaluate different representation sizes in order to find the most effective dimensionality for the representation. We also evaluate the effects of using different speech features as input to the model. We find that mel-frequency cepstral coefficients (MFCCs), alone or combined with prosodic features, perform the best. The results of a subsequent user study confirm the benefits of the representation learning.
['Hedvig Kjellström', 'Gustav Eje Henter', 'Taras Kucherenko', 'Dai Hasegawa', 'Naoshi Kaneko']
2019-03-08
analyzing-input-and-output-representations
null
null
arxiv-2019-3
['gesture-generation']
['robots']
[ 2.77790397e-01 2.81748086e-01 2.11357772e-02 -2.24235058e-01 -6.95705414e-01 -4.20420051e-01 8.72254193e-01 -6.41167641e-01 -4.77059901e-01 3.65622848e-01 8.91805053e-01 9.22408104e-02 2.26504341e-01 -5.62757552e-01 -5.98263264e-01 -9.28781331e-01 -5.75309508e-02 4.48015869e-01 1.96374655e-01 -3.65373552e-01 1.83884263e-01 4.30197567e-01 -1.65201747e+00 4.23782945e-01 1.95586428e-01 5.00016928e-01 5.81338406e-01 1.18916452e+00 6.18963726e-02 6.72014654e-01 -9.01410639e-01 3.59711707e-01 1.34474635e-01 -8.34257245e-01 -8.43674958e-01 1.20009899e-01 -2.51233280e-01 -7.58884311e-01 -4.76165712e-01 4.79218632e-01 7.20612466e-01 6.55213654e-01 9.16714549e-01 -8.63596797e-01 -5.22123098e-01 5.11688769e-01 -6.71119541e-02 -1.68415353e-01 7.67819285e-01 4.76269901e-01 7.52154469e-01 -8.36584866e-01 1.10033000e+00 1.62974215e+00 3.31161469e-01 1.16399109e+00 -1.09594309e+00 -8.40158239e-02 -5.31581752e-02 -5.60711063e-02 -1.08098435e+00 -4.81618166e-01 8.35074902e-01 -5.02815485e-01 1.34481800e+00 1.69062585e-01 7.35678077e-01 1.66578233e+00 1.39390856e-01 7.71188855e-01 3.83953005e-01 -5.66976368e-01 4.76686746e-01 -2.39311606e-01 -3.29073519e-01 4.30330157e-01 -3.92674088e-01 4.60897386e-01 -6.13799751e-01 -5.35999089e-02 8.92003715e-01 -3.13170671e-01 -2.56990701e-01 -3.55900943e-01 -1.37873447e+00 9.68872845e-01 3.57753098e-01 6.20486081e-01 -7.31019080e-01 4.46219057e-01 3.31797361e-01 1.44025043e-01 2.51101971e-01 1.63802072e-01 -2.50405341e-01 -5.08921027e-01 -7.29306102e-01 3.99303019e-01 8.58862400e-01 4.94852662e-01 1.35093406e-01 2.69077927e-01 -1.22208618e-01 7.60005355e-01 6.04767323e-01 4.80175316e-01 1.04853988e+00 -1.14718032e+00 4.47205752e-01 4.86241840e-02 8.94558728e-02 -8.37389827e-01 -5.84503591e-01 3.48459750e-01 -6.25619829e-01 4.59811240e-01 3.35497797e-01 -5.30236840e-01 -1.15270627e+00 1.89689064e+00 2.07580939e-01 1.90749362e-01 4.92549092e-01 1.36052823e+00 8.11262012e-01 1.00818002e+00 1.04576357e-01 -1.69132203e-01 1.04438674e+00 -8.72377813e-01 -7.86177039e-01 -9.06477794e-02 9.06768143e-01 -5.99968195e-01 9.47756588e-01 2.05961987e-01 -1.00948429e+00 -7.19662547e-01 -1.08082891e+00 -1.82509236e-02 -1.90046057e-01 1.40871018e-01 1.64441273e-01 4.54600573e-01 -1.10530901e+00 7.25496709e-01 -1.08987713e+00 -6.00108802e-01 -1.65966734e-01 3.91096681e-01 -2.75094926e-01 4.88253504e-01 -1.05563116e+00 9.32271481e-01 4.42746639e-01 6.16178848e-02 -9.07186031e-01 7.75597841e-02 -1.20919061e+00 -2.54260957e-01 -1.62594676e-01 -8.39500010e-01 1.55365419e+00 -9.32319105e-01 -2.26826978e+00 2.66921341e-01 -2.52576023e-01 -4.20484543e-01 3.93903255e-01 -1.90465778e-01 -1.58660144e-01 2.93599516e-01 -1.69833183e-01 1.24529552e+00 1.03664613e+00 -1.34221232e+00 -4.24860120e-01 -9.64562446e-02 -2.42008790e-01 6.12187207e-01 -1.34377301e-01 -2.37808228e-01 -2.80879885e-01 -8.33436608e-01 5.06543703e-02 -1.17925048e+00 -1.83305800e-01 -3.43147010e-01 -2.45735615e-01 -4.74033738e-03 9.95271325e-01 -8.75981927e-01 1.04435980e+00 -2.17791057e+00 8.06986690e-01 2.13360652e-01 -2.89870411e-01 2.07686797e-01 -5.52785456e-01 4.82138097e-01 2.77360398e-02 -6.51879907e-02 -3.04090798e-01 -5.97356200e-01 -8.36403295e-02 4.40310001e-01 -2.30090812e-01 1.78683981e-01 3.44549745e-01 9.17280436e-01 -8.96382928e-01 -6.51405379e-02 5.79031825e-01 9.55503345e-01 -7.22613156e-01 2.89733827e-01 -3.41779232e-01 9.83180225e-01 -3.54834735e-01 3.94194648e-02 8.46296698e-02 2.57127017e-01 2.46005654e-01 5.88219203e-02 3.02779805e-02 5.93442261e-01 -1.08122241e+00 1.99743676e+00 -5.92818260e-01 8.73056889e-01 5.33419810e-02 -7.73780882e-01 1.09223366e+00 6.73504651e-01 6.08895421e-01 -5.64163983e-01 2.20835134e-01 1.23453222e-01 7.92127699e-02 -9.21759605e-01 5.79282701e-01 -2.59358704e-01 -1.29129723e-01 5.50631821e-01 6.13679811e-02 -3.92485619e-01 -1.12925261e-01 -1.44041717e-01 1.06645823e+00 5.94258606e-01 1.64985195e-01 3.63968968e-01 3.28228205e-01 -5.79562690e-03 5.62758856e-02 2.96795458e-01 -2.68697422e-02 8.90207112e-01 1.99920878e-01 -3.78673494e-01 -1.16842270e+00 -8.27423632e-01 5.26868045e-01 1.26050401e+00 3.30311921e-03 -1.50684148e-01 -8.81728888e-01 -5.01850665e-01 -2.95680135e-01 9.80532765e-01 -5.78905523e-01 -3.31620723e-01 -1.01245868e+00 -4.83999997e-01 4.78426367e-01 6.83200657e-01 1.05162263e-01 -1.79613817e+00 -1.18489027e+00 2.30337828e-01 -2.52349585e-01 -8.59079301e-01 -5.12597382e-01 1.79173768e-01 -7.77979255e-01 -5.75204372e-01 -1.07169402e+00 -1.08779275e+00 3.77599388e-01 -8.28328263e-03 5.54109156e-01 -1.05638616e-01 -1.05053790e-01 4.73856807e-01 -6.99714065e-01 -9.15083364e-02 -8.59611332e-01 -7.31655443e-03 8.40097964e-02 -1.86047152e-01 2.24277154e-01 -4.54108983e-01 -4.28527445e-01 -1.37524847e-02 -8.37010086e-01 -5.76481521e-02 4.87626672e-01 8.99087489e-01 1.83920369e-01 -3.48732203e-01 4.23438370e-01 -1.19335636e-01 9.03602362e-01 -3.48196536e-01 -6.77426979e-02 -3.19247037e-01 2.75560498e-01 4.29243714e-01 3.44306558e-01 -7.60177493e-01 -1.16879702e+00 5.87616324e-01 -5.73018074e-01 -4.56402957e-01 -5.22372603e-01 3.39709073e-01 -4.58548553e-02 4.83723104e-01 8.12229455e-01 2.88927794e-01 3.39365840e-01 -4.11825210e-01 7.25989044e-01 8.27353060e-01 5.75982332e-01 -1.12145849e-01 4.91074950e-01 3.50759953e-01 -3.43542397e-01 -1.30275393e+00 1.35226503e-01 -2.27746308e-01 -9.79506016e-01 -2.17592552e-01 1.39299166e+00 -5.17120779e-01 -6.65211916e-01 3.87146205e-01 -1.64242291e+00 -7.46813655e-01 -4.99185592e-01 8.56125176e-01 -1.09348202e+00 7.90788457e-02 -6.18134677e-01 -1.00456297e+00 -1.24234639e-01 -1.36357594e+00 1.39953196e+00 7.04873726e-02 -7.38584042e-01 -8.21743488e-01 4.78166997e-01 4.80125099e-03 1.21429376e-01 3.93054247e-01 6.56992733e-01 -5.29503465e-01 -1.92511141e-01 1.13950744e-01 5.01016140e-01 1.15909271e-01 2.20317006e-01 -1.21987306e-01 -8.74045491e-01 -1.49789006e-01 -1.57556266e-01 -1.84401453e-01 7.74372101e-01 7.95024037e-01 5.00105500e-01 -3.64787698e-01 -2.51648158e-01 4.36803639e-01 7.72909462e-01 8.17371190e-01 6.67137802e-01 2.81289101e-01 6.55677676e-01 8.33757162e-01 4.03746933e-01 4.28076923e-01 2.25615650e-01 1.01761782e+00 3.53060156e-01 1.98438540e-01 -3.28080297e-01 -3.12161744e-01 7.81400383e-01 8.92760813e-01 -2.42787153e-01 -2.77618736e-01 -7.44030833e-01 5.47006965e-01 -1.87370181e+00 -1.07393909e+00 3.27598929e-01 1.98230696e+00 3.49428117e-01 -1.63458586e-01 5.77164650e-01 2.02883452e-01 4.36282635e-01 2.70763397e-01 -4.26683128e-01 -7.76337981e-01 2.06252500e-01 1.90794259e-01 3.09546739e-02 7.12553382e-01 -1.08002305e+00 1.19165063e+00 6.13862085e+00 1.55537561e-01 -1.33805168e+00 -1.32605761e-01 9.82611924e-02 -2.10352853e-01 -1.34696558e-01 -3.74988139e-01 -2.59840131e-01 2.99599588e-01 1.08510768e+00 2.04962820e-01 5.72430015e-01 7.03860641e-01 6.93753958e-01 1.15935706e-01 -1.10774553e+00 9.12547290e-01 9.35024321e-02 -1.28642213e+00 1.96112335e-01 1.26836628e-01 4.04593825e-01 -1.88448504e-01 -2.24005356e-02 2.62137920e-01 2.73516327e-01 -1.13740218e+00 9.38422084e-01 6.04646444e-01 4.79242861e-01 -7.25522876e-01 7.16144800e-01 5.06733894e-01 -1.33439016e+00 -1.14686169e-01 2.28902418e-02 -1.97318137e-01 6.72909200e-01 -4.05656040e-01 -1.22870350e+00 1.91735998e-01 4.08006698e-01 5.40342569e-01 8.48926604e-02 5.02215445e-01 -2.68790483e-01 3.10429603e-01 -1.67058870e-01 -3.29985499e-01 3.53280604e-01 -4.43457775e-02 8.90598893e-01 1.36683953e+00 5.59637904e-01 2.41221458e-01 7.14594722e-02 6.31712198e-01 3.15888584e-01 -4.85641733e-02 -1.01909554e+00 -1.20505273e-01 4.83510584e-01 6.24522686e-01 -7.09603846e-01 -3.26923490e-01 -4.87609915e-02 1.47946680e+00 -1.11256458e-01 5.87068558e-01 -4.67273474e-01 -1.96425810e-01 7.81722128e-01 -2.77007163e-01 5.66648364e-01 -7.07471192e-01 -5.22459075e-02 -7.37791061e-01 -3.01502228e-01 -8.24031293e-01 3.40671875e-02 -7.08232880e-01 -6.12103403e-01 7.35358477e-01 6.74641207e-02 -1.38318598e+00 -1.42813623e+00 -5.52076101e-01 -5.86495340e-01 8.31041873e-01 -6.46050751e-01 -8.90316427e-01 -2.01253653e-01 3.43460679e-01 1.05854475e+00 -2.81926632e-01 1.07748234e+00 -1.19204499e-01 -3.24874490e-01 2.16986164e-01 -1.86747447e-01 2.27112144e-01 2.48544961e-01 -1.06027818e+00 8.10766757e-01 4.66556311e-01 3.82009178e-01 4.20424461e-01 7.02976525e-01 -7.01772213e-01 -1.21440744e+00 -8.32632005e-01 7.77580440e-01 -4.44408506e-01 2.55489558e-01 -2.93888479e-01 -7.13101268e-01 5.68149149e-01 1.81047708e-01 -3.66873175e-01 3.08919400e-01 -5.24614215e-01 1.49553597e-01 7.18757808e-01 -9.75967526e-01 7.43407130e-01 1.12636566e+00 -4.43244278e-01 -8.86001706e-01 -4.55895998e-02 8.98972571e-01 -5.43974102e-01 -5.90113521e-01 8.80770385e-02 7.78728843e-01 -6.82613790e-01 9.66981769e-01 -5.49996614e-01 3.86142910e-01 -1.80486917e-01 -2.91314721e-01 -1.76302111e+00 -1.23531148e-01 -6.88977122e-01 -1.87655210e-01 6.86029017e-01 4.71418053e-01 -1.76079988e-01 9.45369303e-01 2.33290598e-01 -3.36991698e-02 -3.95986468e-01 -1.00360036e+00 -6.47524059e-01 2.57619172e-02 -4.86673832e-01 4.17341232e-01 5.55896997e-01 2.69949883e-01 5.20501256e-01 -5.64539015e-01 9.28681046e-02 8.84079859e-02 -1.93838075e-01 9.54044759e-01 -8.32384288e-01 -4.56712395e-01 -5.36641121e-01 -4.55608636e-01 -1.38694000e+00 1.85380355e-01 -6.17372513e-01 6.63406491e-01 -1.71547878e+00 -2.41799295e-01 1.73079640e-01 4.04109865e-01 2.96484828e-01 1.31995931e-01 -4.71178666e-02 6.00666702e-01 2.66562521e-01 4.58365642e-02 8.52747262e-01 1.22728574e+00 -4.65987362e-02 -1.01616085e+00 4.24220506e-03 4.69618365e-02 8.20285201e-01 7.64838457e-01 -2.21600518e-01 -6.25097990e-01 -4.98173624e-01 -5.26337922e-01 3.53260279e-01 3.37060422e-01 -9.20362711e-01 4.06608619e-02 -7.70637915e-02 4.10594463e-01 -4.89714921e-01 8.02584589e-01 -6.24075830e-01 1.38548743e-02 8.26296747e-01 -4.98535335e-01 3.31593119e-02 1.28239900e-01 3.96923900e-01 -1.38290152e-01 -1.84607774e-01 4.22391504e-01 -6.54217675e-02 -8.93698931e-01 -2.79479206e-01 -8.63854885e-01 -4.17867064e-01 8.24348271e-01 -3.81372929e-01 2.30412990e-01 -9.64490116e-01 -1.25141001e+00 -2.27270141e-01 1.32454664e-01 8.48861814e-01 8.97585988e-01 -1.47337914e+00 -6.25491381e-01 3.46442223e-01 -2.95936108e-01 -6.70899684e-03 -1.20320385e-02 4.42234069e-01 -5.55273592e-01 4.97062683e-01 -3.11994523e-01 -7.08871722e-01 -1.22128975e+00 3.20169628e-01 3.03421468e-01 1.40844941e-01 -9.75509524e-01 7.86667287e-01 2.35497653e-02 -5.96263885e-01 3.88616830e-01 -5.95442414e-01 -4.36314404e-01 4.19369861e-02 5.24485469e-01 2.91841149e-01 -2.73127109e-01 -1.17051172e+00 -3.96355689e-01 6.57392979e-01 4.31467086e-01 -1.08340406e+00 1.32589638e+00 -2.89609246e-02 6.15311682e-01 4.76862997e-01 1.39995790e+00 -3.63497853e-01 -1.62140477e+00 2.58468717e-01 1.41161129e-01 -3.98735069e-02 -2.00145662e-01 -4.86559123e-01 -7.98434258e-01 8.90769541e-01 6.38945222e-01 1.15924627e-02 9.30047393e-01 1.03843883e-01 8.70011210e-01 3.13895851e-01 2.66338110e-01 -1.13562167e+00 5.77029467e-01 7.60858417e-01 1.19403100e+00 -8.88726890e-01 -6.37189627e-01 -4.27227356e-02 -1.03266811e+00 1.18700731e+00 3.01774353e-01 -2.07699805e-01 4.04970437e-01 2.79143631e-01 3.77543747e-01 -6.11308031e-02 -7.00096488e-01 -4.37266290e-01 3.62492651e-01 1.15112150e+00 5.39621174e-01 1.35088533e-01 -2.55994380e-01 3.08110595e-01 -7.16886640e-01 -1.67929418e-02 3.52704853e-01 1.03021955e+00 -5.20702899e-01 -9.89855468e-01 -6.19491935e-01 -1.55017659e-01 2.89537191e-01 2.39006042e-01 -6.75563514e-01 8.60707998e-01 1.49519108e-02 1.05773556e+00 2.99072236e-01 -8.33687901e-01 4.74567503e-01 3.03621739e-01 4.04635400e-01 -5.98367155e-01 -3.66004497e-01 4.19087857e-01 1.09329045e-01 -6.57220364e-01 -3.33115399e-01 -7.44268775e-01 -1.81316137e+00 7.94767365e-02 2.34969035e-01 -1.81128725e-01 9.01237190e-01 1.01835239e+00 2.57740200e-01 7.37207413e-01 4.39141154e-01 -1.80082512e+00 -5.62302232e-01 -1.24671412e+00 -1.98461443e-01 6.87409461e-01 6.20344758e-01 -6.23318434e-01 -2.60538965e-01 3.91771644e-01]
[5.610669136047363, -0.10460688918828964]
597ee8b3-d5f3-4ad1-a8c0-07a0f6b58255
low-resource-learning-with-knowledge-graphs-a
2112.10006
null
https://arxiv.org/abs/2112.10006v6
https://arxiv.org/pdf/2112.10006v6.pdf
Zero-shot and Few-shot Learning with Knowledge Graphs: A Comprehensive Survey
Machine learning especially deep neural networks have achieved great success but many of them often rely on a number of labeled samples for supervision. As sufficient labeled training data are not always ready due to e.g., continuously emerging prediction targets and costly sample annotation in real world applications, machine learning with sample shortage is now being widely investigated. Among all these studies, many prefer to utilize auxiliary information including those in the form of Knowledge Graph (KG) to reduce the reliance on labeled samples. In this survey, we have comprehensively reviewed over 90 papers about KG-aware research for two major sample shortage settings -- zero-shot learning (ZSL) where some classes to be predicted have no labeled samples, and few-shot learning (FSL) where some classes to be predicted have only a small number of labeled samples that are available. We first introduce KGs used in ZSL and FSL as well as their construction methods, and then systematically categorize and summarize KG-aware ZSL and FSL methods, dividing them into different paradigms such as the mapping-based, the data augmentation, the propagation-based and the optimization-based. We next present different applications, including not only KG augmented prediction tasks such as image classification, question answering, text classification and knowledge extraction, but also KG completion tasks, and some typical evaluation resources for each task. We eventually discuss some challenges and open problems from different perspectives.
['Huajun Chen', 'Wen Zhang', 'Jeff Z. Pan', 'Zhuo Chen', 'Yuan He', 'Ian Horrocks', 'Yuxia Geng', 'Jiaoyan Chen']
2021-12-18
null
null
null
null
['knowledge-graph-embeddings', 'knowledge-graph-embeddings']
['graphs', 'methodology']
[ 3.42763603e-01 3.18728000e-01 -7.19639182e-01 -5.91193616e-01 -4.61022377e-01 4.98096049e-02 2.94004321e-01 2.51411855e-01 -4.99133587e-01 9.44936275e-01 -8.54490697e-02 -1.47555903e-01 -3.22529525e-01 -1.08881545e+00 -5.17472923e-01 -6.95954323e-01 2.78685719e-01 6.16852343e-01 1.97455481e-01 -5.52460924e-02 -8.42416957e-02 3.41885127e-02 -1.79278111e+00 1.86428115e-01 9.57826614e-01 1.33990514e+00 1.93480596e-01 2.79828846e-01 -6.11394167e-01 8.45058799e-01 -4.93831962e-01 -4.34337884e-01 1.48120090e-01 -1.89788073e-01 -8.96270454e-01 9.11162943e-02 3.24597299e-01 -2.33639002e-01 -3.88531417e-01 1.17893100e+00 7.08398223e-01 5.35442710e-01 5.90757906e-01 -1.71149659e+00 -8.31825256e-01 7.32617199e-01 -3.90386403e-01 6.95465803e-02 -2.30159864e-01 1.97745506e-02 9.22864556e-01 -1.19266903e+00 6.69491291e-01 9.17563856e-01 7.15580821e-01 7.91809201e-01 -7.55895138e-01 -5.95837474e-01 2.00704709e-01 9.94973183e-01 -1.24751222e+00 -5.02005875e-01 9.53601658e-01 -3.71464849e-01 8.94573629e-01 -7.78837577e-02 5.46194613e-01 9.47288394e-01 -4.42090839e-01 1.13576770e+00 8.26412499e-01 -7.09493160e-01 4.31567609e-01 4.18964863e-01 8.36953223e-01 7.68935919e-01 1.94905758e-01 -1.13228060e-01 -8.55530977e-01 -7.90279955e-02 2.47347444e-01 2.02432051e-02 -3.79840910e-01 -4.27515715e-01 -9.47116673e-01 9.68219399e-01 3.02020013e-01 4.19368595e-02 -2.19487116e-01 -3.03150743e-01 5.77720344e-01 3.60014707e-01 8.42041790e-01 2.43958920e-01 -8.96153271e-01 1.41984820e-01 -8.00574660e-01 4.66839820e-02 7.83949018e-01 1.37025321e+00 1.06706405e+00 3.51165920e-01 -3.41644168e-01 1.20324922e+00 -1.04817310e-02 1.05939403e-01 7.20658123e-01 -3.54238153e-01 5.76660872e-01 8.03950548e-01 -1.41534939e-01 -7.45158374e-01 -3.81707639e-01 -3.89075577e-01 -1.25910223e+00 -2.90381253e-01 1.63564429e-01 -3.62774521e-01 -1.05974197e+00 1.77513909e+00 5.51376700e-01 4.59947616e-01 1.05797410e-01 6.04614496e-01 1.49565923e+00 3.95565540e-01 1.05406426e-01 -4.86069798e-01 1.19242156e+00 -1.22960579e+00 -1.00587738e+00 -3.91569018e-01 1.02993822e+00 -2.66003042e-01 1.26591814e+00 2.73663908e-01 -5.05006313e-01 -5.65795064e-01 -1.15496802e+00 -1.15746379e-01 -9.27038610e-01 6.66546673e-02 7.61215091e-01 6.43864274e-01 -8.33488464e-01 7.66036570e-01 -5.24351180e-01 -3.05129379e-01 8.75899732e-01 2.65992194e-01 -4.46612179e-01 -5.09996951e-01 -1.67547536e+00 7.41376102e-01 6.93742394e-01 -4.37281802e-02 -6.57046676e-01 -8.83469701e-01 -1.00585914e+00 1.23197928e-01 7.11206794e-01 -6.32532299e-01 1.23902810e+00 -6.79814637e-01 -1.15608859e+00 7.77899325e-01 -7.82256573e-02 -3.51701587e-01 2.46166766e-01 -1.18781410e-01 -5.48896909e-01 -1.62033558e-01 2.72729620e-02 5.85682571e-01 6.67949200e-01 -8.83131266e-01 -6.80913150e-01 -4.80849028e-01 -1.63931306e-02 2.32351273e-01 -9.99518931e-01 -2.07699284e-01 -3.75784218e-01 -4.16994125e-01 -2.03809232e-01 -4.46178913e-01 -3.04928660e-01 1.16870627e-01 -5.70484996e-01 -6.86356783e-01 1.07409799e+00 -3.71419877e-01 1.31829607e+00 -1.87811148e+00 -2.50227153e-01 -2.45675713e-01 3.45743865e-01 6.61387503e-01 -3.38599294e-01 4.55452293e-01 -1.26408234e-01 -6.44467548e-02 -2.87694305e-01 -4.29950595e-01 -1.40101925e-01 4.20784980e-01 -2.42593735e-01 2.31125712e-01 2.17675328e-01 1.11573017e+00 -1.12917149e+00 -6.32021248e-01 2.30526835e-01 1.85430825e-01 -1.05385184e-01 3.80069693e-03 -3.94812196e-01 1.13134235e-01 -3.95436466e-01 8.29446733e-01 5.12854397e-01 -6.00308359e-01 -1.66207105e-01 -3.95709574e-01 2.18058303e-01 8.63743126e-02 -1.28016937e+00 1.55526471e+00 -3.31895053e-01 4.12447870e-01 -2.54545718e-01 -1.48882699e+00 8.72646272e-01 3.83221358e-01 3.85249168e-01 -2.61800647e-01 1.95980251e-01 -1.29679553e-04 -1.00033596e-01 -6.16142690e-01 3.56189191e-01 -3.11621338e-01 2.81436771e-01 4.19495374e-01 6.33972287e-01 2.91067541e-01 4.35978800e-01 2.60423124e-01 1.03934789e+00 -2.38571674e-01 6.87740505e-01 -5.22960313e-02 3.85882407e-01 1.62718847e-01 8.24711263e-01 7.71668434e-01 -4.49172020e-01 3.52542698e-01 2.89313525e-01 -5.02557158e-01 -9.17332292e-01 -4.89953309e-01 -1.34020522e-01 1.35540450e+00 9.99387354e-02 -4.00391847e-01 -6.10455096e-01 -9.00200844e-01 -1.69885606e-01 9.16061819e-01 -7.81003714e-01 -6.06969476e-01 2.90261395e-03 -1.15076697e+00 3.52127135e-01 7.43122518e-01 7.27676213e-01 -1.28144085e+00 -1.57932863e-01 7.85295069e-02 -3.20869945e-02 -1.03956974e+00 -1.53322592e-01 5.18565536e-01 -9.21320617e-01 -1.31781542e+00 -5.14575064e-01 -8.85250807e-01 8.20127487e-01 3.29508513e-01 1.19103658e+00 3.66869532e-02 -4.44958568e-01 2.01324880e-01 -7.90510237e-01 -9.43065345e-01 7.77761787e-02 1.73798144e-01 2.55692482e-01 7.89019242e-02 1.01930261e+00 -5.09348452e-01 -3.17630559e-01 1.86610609e-01 -7.82382846e-01 2.56201357e-01 4.75674629e-01 1.12253320e+00 7.54429221e-01 5.23369499e-02 1.09934914e+00 -1.39773262e+00 6.63642645e-01 -6.70986593e-01 -1.63257897e-01 5.71554482e-01 -8.49355340e-01 -2.40723729e-01 7.85629809e-01 -5.58991015e-01 -8.76986861e-01 -3.48604247e-02 -1.55108526e-01 -7.36235082e-01 -1.31939813e-01 6.96512461e-01 -4.81163442e-01 -5.34963757e-02 9.03751671e-01 2.42594436e-01 2.10488643e-02 -5.20428896e-01 5.15306056e-01 8.65401924e-01 1.08018667e-01 -2.53158420e-01 5.55666208e-01 2.23322168e-01 -1.58170179e-01 -8.63283038e-01 -1.57445872e+00 -6.43522620e-01 -8.32373798e-01 -2.20853522e-01 5.17552972e-01 -5.93750834e-01 -2.48614892e-01 5.67716122e-01 -9.35641408e-01 -3.81394625e-01 -8.72614741e-01 4.07159448e-01 -3.93877596e-01 1.53302550e-01 -5.04869461e-01 -6.85217261e-01 -7.02634096e-01 -8.00777733e-01 5.98057091e-01 2.24312127e-01 1.57883123e-01 -1.03791666e+00 -2.53855474e-02 3.11281919e-01 3.92513424e-01 -1.25610694e-01 1.24570847e+00 -1.17938948e+00 -1.97000533e-01 -4.97747332e-01 -3.03883076e-01 5.50053537e-01 3.67949158e-01 -4.11078066e-01 -1.18535829e+00 -2.80732512e-01 -5.63354827e-02 -8.42846513e-01 9.27264869e-01 3.08095485e-01 1.50434494e+00 -1.96926191e-01 -4.21786129e-01 4.79096740e-01 1.26975501e+00 8.36859941e-02 3.38755339e-01 -4.63234335e-02 8.72444987e-01 7.35361457e-01 9.30384696e-01 3.86294842e-01 2.01961800e-01 2.88867950e-01 3.56732011e-01 1.51573703e-01 -2.19578549e-01 -1.65547684e-01 -4.29987684e-02 1.21287394e+00 4.50313538e-02 -3.53200465e-01 -9.40058529e-01 5.51671147e-01 -2.07022786e+00 -7.14015365e-01 -8.72000158e-02 2.12570095e+00 9.50378001e-01 -9.83390436e-02 -2.04686105e-01 5.21817029e-01 9.36667621e-01 1.90661579e-01 -1.08259547e+00 1.37571633e-01 -5.34520149e-02 2.13319436e-01 3.56549442e-01 1.56223729e-01 -1.32179475e+00 1.03845549e+00 5.86473417e+00 1.35767591e+00 -6.18815005e-01 4.33732569e-01 8.27258229e-01 -2.72654239e-02 1.48046790e-02 -1.39954090e-01 -1.14663470e+00 2.60265678e-01 7.50926256e-01 -2.53164023e-01 1.39801681e-01 1.30997217e+00 -2.29484692e-01 1.86355822e-02 -1.12250805e+00 1.21976852e+00 2.25011259e-01 -1.57564509e+00 2.56909907e-01 -2.17080399e-01 6.92178786e-01 1.93704873e-01 -2.45131850e-01 8.26946855e-01 2.45941430e-01 -8.03381145e-01 8.99047554e-02 4.95076567e-01 1.12593472e+00 -6.68502927e-01 8.24796140e-01 9.16950285e-01 -1.10173237e+00 6.63470477e-03 -8.99879217e-01 -2.25284472e-01 8.17735642e-02 1.13678253e+00 -6.95617855e-01 7.62104571e-01 5.95548153e-01 9.51785147e-01 -4.01067346e-01 1.11068559e+00 -2.92016834e-01 7.46162534e-01 -1.70686841e-01 -2.11588383e-01 -4.84412070e-03 -1.53557226e-01 8.97237882e-02 7.79268682e-01 1.80586845e-01 2.64951944e-01 3.96390736e-01 5.72013080e-01 -3.36651266e-01 4.71045703e-01 -5.93651116e-01 -1.23802640e-01 6.58781052e-01 1.40087974e+00 -5.78467607e-01 -7.71159708e-01 -8.16491187e-01 5.28415501e-01 6.02339327e-01 2.28857413e-01 -3.88684720e-01 -6.43127680e-01 5.20199716e-01 -1.11899585e-01 -1.82128504e-01 1.69888198e-01 -2.40613431e-01 -1.29893804e+00 -1.53012633e-01 -4.86708850e-01 4.88715589e-01 -5.65245986e-01 -1.72082675e+00 3.95533562e-01 -2.52958108e-02 -1.23429596e+00 -1.29021648e-02 -7.96161354e-01 -4.19381589e-01 8.69840622e-01 -1.70051193e+00 -9.15823519e-01 -5.34064412e-01 5.91530383e-01 7.79086590e-01 -4.68752205e-01 9.37925398e-01 6.40883744e-01 -9.35955644e-01 6.11297905e-01 2.16971979e-01 2.32601494e-01 6.66911662e-01 -1.00556326e+00 1.93230346e-01 4.86914098e-01 3.29989254e-01 3.98061454e-01 2.89722353e-01 -7.28718698e-01 -1.08816135e+00 -1.49000371e+00 8.11481297e-01 -1.66836828e-01 4.90723938e-01 -2.93804735e-01 -1.19970059e+00 6.28293753e-01 -1.87334329e-01 6.25686109e-01 9.76656973e-01 2.40993291e-01 -1.90730989e-01 -2.16371283e-01 -1.09087968e+00 4.01392549e-01 1.20516038e+00 -3.15410763e-01 -3.78036559e-01 6.34583473e-01 8.21103275e-01 -9.25010741e-02 -6.84863746e-01 6.49899662e-01 9.04843211e-02 -6.25592113e-01 7.36836672e-01 -1.13309753e+00 3.01280290e-01 8.27439129e-02 -3.59155051e-02 -1.33364618e+00 -2.97905236e-01 1.03372961e-01 -5.64505160e-01 1.23558211e+00 2.99414098e-01 -5.64645290e-01 1.32010829e+00 6.77459240e-01 -3.70090365e-01 -1.28026462e+00 -8.21395218e-01 -8.21017861e-01 -2.00644419e-01 -5.51474154e-01 4.60652322e-01 1.42946804e+00 -6.73801601e-02 6.96495712e-01 -5.36158323e-01 -3.19324076e-01 6.31086230e-01 2.07126476e-02 6.69079304e-01 -1.67181826e+00 -1.31697908e-01 -1.10321544e-01 -2.82579601e-01 -7.94016361e-01 1.15344316e-01 -1.18057847e+00 -8.11009319e-04 -1.75856745e+00 4.71288651e-01 -5.30447602e-01 -4.57541734e-01 1.02362442e+00 -3.24999094e-01 8.48114491e-02 -1.72673270e-01 1.29995540e-01 -7.62164593e-01 9.23410773e-01 1.32373762e+00 -2.25238994e-01 -8.97039846e-02 9.83329862e-02 -6.04169607e-01 9.01587069e-01 7.74208784e-01 -3.73424113e-01 -9.42683280e-01 -1.31076947e-01 3.52073401e-01 -2.82644242e-01 4.21435200e-02 -8.80038619e-01 3.47430497e-01 -2.23798200e-01 3.27057272e-01 -6.40175402e-01 4.42655265e-01 -5.61672628e-01 -3.55936229e-01 2.18114018e-01 -3.83045852e-01 -5.48959851e-01 -5.53247109e-02 7.59630799e-01 -3.70481670e-01 -6.97234273e-01 8.03133011e-01 -2.63322800e-01 -1.15757430e+00 1.04600215e+00 3.48510928e-02 3.45156401e-01 1.13249123e+00 -1.74969390e-01 -4.66123134e-01 -2.23208100e-01 -8.93001854e-01 4.74273622e-01 -2.01294079e-01 3.94191951e-01 7.70188868e-01 -1.46570742e+00 -4.06090260e-01 1.69690385e-01 5.47818840e-01 3.85671258e-01 5.58609307e-01 8.64677608e-01 -3.68592180e-02 4.04666096e-01 7.24450126e-02 -1.17311709e-01 -1.21745682e+00 9.01081324e-01 -8.04920569e-02 -4.46016133e-01 -6.01436436e-01 1.17159021e+00 1.60429865e-01 -7.98907876e-01 4.76047367e-01 7.60880634e-02 -5.49011409e-01 2.91888565e-01 7.54045665e-01 4.05344844e-01 2.80225605e-01 -1.94338650e-01 -1.07856989e-01 7.16613755e-02 -3.50663334e-01 6.13184571e-01 1.26636159e+00 9.46613774e-02 -8.87772292e-02 8.22250485e-01 1.03621292e+00 -8.09622943e-01 -8.98855865e-01 -8.88787508e-01 1.66859217e-02 -2.34867975e-01 5.63866533e-02 -7.15858817e-01 -1.18691778e+00 1.28699780e+00 3.26147705e-01 1.84955671e-01 1.03017735e+00 -2.94405501e-02 6.86057925e-01 8.56727839e-01 5.70646822e-01 -1.44182694e+00 5.93954064e-02 6.36170387e-01 5.48087120e-01 -1.45077789e+00 -2.27888767e-02 -6.80896342e-01 -4.75679159e-01 8.54930282e-01 9.64619935e-01 2.28044346e-01 1.12411630e+00 -1.39837917e-02 -2.77767241e-01 -1.15394946e-02 -9.16482508e-01 -3.25443745e-01 2.50124842e-01 8.72434080e-01 2.14984953e-01 -1.78399831e-01 -1.00512765e-01 9.70573306e-01 -9.63086337e-02 2.06892490e-01 1.51331559e-01 8.53387117e-01 -6.73755348e-01 -1.16544116e+00 5.70136635e-03 1.44635391e+00 1.61568392e-02 -3.77508432e-01 -1.28136277e-01 6.96437180e-01 3.39551747e-01 7.58377314e-01 -1.62735712e-02 -4.84807193e-01 1.83132991e-01 5.63202083e-01 2.81200767e-01 -1.19821084e+00 -4.86181751e-02 -5.78805208e-01 1.30281344e-01 -2.12518066e-01 -4.74224478e-01 -1.28786549e-01 -9.22991335e-01 -1.15939640e-01 -8.42717648e-01 4.06184942e-02 4.28787887e-01 1.10665095e+00 2.80558616e-01 5.35303354e-01 2.83360034e-01 -6.42168760e-01 -5.56608796e-01 -1.29050624e+00 -1.03896904e+00 2.10969806e-01 -5.88119254e-02 -9.08711493e-01 -2.60507792e-01 -5.45829125e-02]
[10.027558326721191, 3.219343423843384]
321a7b78-0040-49bd-823d-e83861aac801
cross-modal-search-method-of-technology-video
2210.05243
null
https://arxiv.org/abs/2210.05243v1
https://arxiv.org/pdf/2210.05243v1.pdf
Cross-modal Search Method of Technology Video based on Adversarial Learning and Feature Fusion
Technology videos contain rich multi-modal information. In cross-modal information search, the data features of different modalities cannot be compared directly, so the semantic gap between different modalities is a key problem that needs to be solved. To address the above problems, this paper proposes a novel Feature Fusion based Adversarial Cross-modal Retrieval method (FFACR) to achieve text-to-video matching, ranking and searching. The proposed method uses the framework of adversarial learning to construct a video multimodal feature fusion network and a feature mapping network as generator, a modality discrimination network as discriminator. Multi-modal features of videos are obtained by the feature fusion network. The feature mapping network projects multi-modal features into the same semantic space based on semantics and similarity. The modality discrimination network is responsible for determining the original modality of features. Generator and discriminator are trained alternately based on adversarial learning, so that the data obtained by the feature mapping network is semantically consistent with the original data and the modal features are eliminated, and finally the similarity is used to rank and obtain the search results in the semantic space. Experimental results demonstrate that the proposed method performs better in text-to-video search than other existing methods, and validate the effectiveness of the method on the self-built datasets of technology videos.
['Ang Li', 'Meiyu Liang', 'Junping Du', 'Xiangbin Liu']
2022-10-11
null
null
null
null
['text-to-video-search']
['natural-language-processing']
[ 1.80060759e-01 -6.97979510e-01 -1.32209823e-01 -4.71627451e-02 -9.62273657e-01 -7.09308803e-01 7.13032305e-01 -2.94856727e-01 -2.66495168e-01 3.43776494e-01 3.72766674e-01 3.40334564e-01 -5.39923847e-01 -7.55575538e-01 -4.02442962e-01 -8.48343134e-01 2.59187698e-01 8.87774676e-02 2.80123800e-01 -3.14762890e-01 4.62772250e-01 3.33555937e-01 -1.80484784e+00 7.48694062e-01 7.52914310e-01 1.50036263e+00 1.05884977e-01 2.99730539e-01 -2.67248899e-01 6.70850754e-01 -4.33570504e-01 -1.72635734e-01 4.63178486e-01 -5.38222909e-01 -7.01703012e-01 -2.16553450e-01 3.13963592e-01 -1.94691539e-01 -7.59705424e-01 1.29661369e+00 7.64458895e-01 2.77373612e-01 8.53231013e-01 -1.68479025e+00 -7.41542995e-01 1.12312742e-01 -2.42316693e-01 -5.34064956e-02 9.19651330e-01 -2.29285106e-01 6.16013825e-01 -8.73641193e-01 4.82674837e-01 1.53276789e+00 4.57448095e-01 6.28428638e-01 -6.16519749e-01 -8.35110545e-01 -2.53381401e-01 5.43623209e-01 -1.79024708e+00 -3.55010241e-01 9.92837191e-01 -4.88745421e-01 3.92965257e-01 4.82509434e-01 4.85903174e-01 9.15509939e-01 2.91670442e-01 7.29134560e-01 8.09110641e-01 -3.16596210e-01 -2.97807932e-01 2.88254380e-01 -4.09077466e-01 7.66703546e-01 -3.68985265e-01 1.86384082e-01 -5.40981829e-01 -3.53476822e-01 6.17717087e-01 5.50109863e-01 -4.26992744e-01 -4.18970913e-01 -1.26320827e+00 8.77558529e-01 4.61437881e-01 7.08406448e-01 -9.91475508e-02 -1.49241239e-01 6.08369708e-01 6.06201887e-01 -4.97632399e-02 1.16201527e-01 -9.81654003e-02 1.82459757e-01 -7.20051169e-01 9.26249474e-02 3.98239911e-01 9.93372738e-01 8.28689814e-01 -9.96572450e-02 -2.68548459e-01 7.81983197e-01 5.99713802e-01 8.42379332e-01 1.12886095e+00 -8.13073874e-01 5.10480881e-01 7.41659403e-01 -1.58392042e-01 -1.54862869e+00 6.75408915e-02 3.82901222e-01 -6.60109401e-01 1.27451839e-02 -4.56928648e-02 4.99030985e-02 -8.26373816e-01 1.56270587e+00 1.79175064e-01 1.80892199e-01 4.41735029e-01 1.10252738e+00 1.38626564e+00 7.97426760e-01 7.22617656e-02 -1.89592585e-01 1.14933777e+00 -7.71020353e-01 -7.10728049e-01 3.22947055e-01 1.78568274e-01 -1.15374625e+00 6.06640935e-01 4.73658293e-02 -8.17882419e-01 -8.71939063e-01 -9.99429107e-01 2.78128356e-01 -7.52437234e-01 -4.29020301e-02 1.83167234e-01 3.47915977e-01 -7.93549299e-01 4.12211329e-01 -1.43813297e-01 -3.68404359e-01 9.94501635e-02 3.15738618e-01 -7.76518583e-01 -2.25591153e-01 -1.70811200e+00 6.57866001e-01 9.49883461e-01 -8.26930478e-02 -7.95485198e-01 -4.31311905e-01 -8.48121822e-01 -1.32022902e-01 2.43906081e-01 -6.60167694e-01 5.27968407e-01 -1.39549804e+00 -1.00966537e+00 6.19436085e-01 1.31078824e-01 2.41892070e-01 4.17820811e-01 2.06937402e-01 -1.14519465e+00 5.36844313e-01 3.58628213e-01 5.66806793e-01 1.15811610e+00 -1.36230123e+00 -9.79712486e-01 -4.41023469e-01 -3.99124548e-02 4.66883391e-01 -4.85790193e-01 1.32864624e-01 -6.77284002e-01 -6.11213326e-01 2.87878186e-01 -7.80730128e-01 4.87758011e-01 -2.11080432e-01 -9.07134563e-02 -2.44956762e-01 1.35426116e+00 -7.41720915e-01 1.27422774e+00 -2.29890037e+00 4.63210046e-01 5.70029080e-01 -1.43320262e-01 1.43531218e-01 -1.52001947e-01 4.92736220e-01 -2.67558753e-01 6.67752773e-02 1.48061335e-01 4.36909139e-01 -2.01917261e-01 -7.36672059e-02 -4.50124368e-02 2.86059916e-01 -1.55210122e-01 7.98548400e-01 -9.76544797e-01 -1.07716334e+00 3.74169320e-01 2.40841255e-01 -9.72072110e-02 3.28888774e-01 2.47736827e-01 3.62422645e-01 -9.13250923e-01 1.06934416e+00 5.86518049e-01 2.59733826e-01 -1.62062407e-01 -8.53404224e-01 1.41426697e-01 -8.42288256e-01 -1.35341358e+00 1.92564178e+00 -2.64310151e-01 2.34036729e-01 -2.56458730e-01 -1.01945460e+00 8.19626451e-01 4.27465588e-01 7.48476863e-01 -9.53572690e-01 4.21418458e-01 3.50297838e-01 -3.86399955e-01 -1.10638571e+00 4.26928848e-01 -2.12916389e-01 -4.26945299e-01 1.23123519e-01 1.31105661e-01 -5.82838356e-02 -8.13883096e-02 2.80695289e-01 6.72964334e-01 7.29527101e-02 -3.13530028e-01 8.28257650e-02 1.14793408e+00 3.97968292e-02 5.18780053e-01 3.92007440e-01 -1.50172142e-02 5.10378480e-01 -8.15316737e-02 -2.16021433e-01 -7.56927133e-01 -1.12856960e+00 1.00701237e-02 8.44644964e-01 8.18780541e-01 -1.16986990e-01 -5.10521531e-01 -1.04609454e+00 1.78660154e-01 1.07439376e-01 -6.24045134e-01 -7.80644834e-01 -1.46439552e-01 1.24638319e-01 7.44925439e-01 2.54223704e-01 8.35188150e-01 -1.05184579e+00 3.52272987e-02 -8.18238631e-02 -4.82220173e-01 -7.43416965e-01 -8.22880685e-01 -6.04274392e-01 -4.68709618e-01 -1.30472457e+00 -8.08631957e-01 -1.24079001e+00 5.71942329e-01 4.25978035e-01 4.97588158e-01 1.02432437e-01 -2.65607148e-01 8.73737872e-01 -6.17721558e-01 1.72806785e-01 -5.70230484e-01 -4.71088946e-01 1.84956670e-01 4.62924331e-01 3.33997428e-01 -1.79496706e-01 -6.01155519e-01 5.51703095e-01 -1.43451250e+00 -4.02315527e-01 4.27382559e-01 1.03496790e+00 5.12964487e-01 4.45268035e-01 5.90868533e-01 -2.70487040e-01 6.52365267e-01 -6.69123828e-01 -2.77659893e-01 6.97611928e-01 -3.82982701e-01 -3.23731154e-02 5.86088419e-01 -5.70335627e-01 -1.10102224e+00 3.59768122e-02 3.18532437e-01 -1.07740998e+00 4.29467531e-03 6.29601538e-01 -6.36736214e-01 -5.52547395e-01 3.26907158e-01 6.06954813e-01 2.24426165e-01 -2.31424376e-01 3.68593782e-01 8.45358849e-01 5.73167801e-01 -2.97458619e-01 8.94852817e-01 2.97756195e-01 9.84228849e-02 -3.26947719e-01 -1.24546170e-01 -4.68482196e-01 -3.60586911e-01 -5.29678345e-01 1.15148640e+00 -9.86686826e-01 -6.07954025e-01 4.69529539e-01 -9.46002185e-01 8.31186235e-01 1.63963847e-02 6.68460786e-01 -5.55379987e-01 7.21611083e-01 -1.80386558e-01 -4.08756286e-01 -3.38909030e-01 -1.17150033e+00 9.91068661e-01 5.61219752e-01 3.42189014e-01 -1.05452180e+00 -7.52979144e-02 4.68668252e-01 2.28833288e-01 2.46522874e-01 1.02679884e+00 -7.85698712e-01 -4.10880685e-01 -6.19903505e-01 -1.36288598e-01 4.93377686e-01 3.23963344e-01 1.54660754e-02 -5.86884856e-01 -3.58020961e-01 -1.59896761e-02 -2.99239814e-01 7.26012707e-01 -8.03491771e-02 1.14816570e+00 -2.36538053e-01 -5.26011467e-01 4.47216719e-01 1.64468086e+00 6.55375898e-01 7.32741654e-01 2.15818644e-01 7.44352579e-01 4.00733501e-01 1.07687569e+00 1.06096856e-01 5.82490899e-02 5.42311788e-01 3.30458909e-01 1.57517776e-01 -8.00788030e-02 -5.27765393e-01 2.62393475e-01 8.31650138e-01 2.10734606e-02 -1.64982960e-01 -4.49794322e-01 4.79546368e-01 -1.98150074e+00 -1.44795156e+00 3.13933134e-01 2.14333200e+00 5.29436767e-01 -3.83605778e-01 -1.39741739e-02 5.24675325e-02 1.09541583e+00 -8.93186703e-02 -3.76050323e-01 -3.38674686e-03 -3.22457314e-01 -2.49377728e-01 2.41723329e-01 1.75308958e-01 -1.20210767e+00 6.11628473e-01 5.53940773e+00 1.57103539e+00 -1.11845791e+00 1.17498502e-01 1.64407492e-03 2.42824599e-01 -6.29491866e-01 -3.54550709e-03 -2.58793056e-01 7.39752412e-01 1.92193285e-01 -3.23339731e-01 5.89599609e-01 6.73111320e-01 -5.70344217e-02 7.47259259e-02 -1.00331211e+00 1.51898623e+00 6.03005409e-01 -1.04649019e+00 6.34463072e-01 -2.69911259e-01 8.35040927e-01 -4.47100967e-01 1.59815565e-01 2.54018873e-01 -8.70042071e-02 -8.53120089e-01 6.08176410e-01 1.17289352e+00 9.60946381e-01 -7.35717297e-01 7.87468672e-01 7.17463642e-02 -1.57038569e+00 -3.20098460e-01 -2.80496269e-01 7.00387180e-01 -9.24254283e-02 -1.14820756e-01 -2.19357446e-01 1.33476651e+00 6.48819625e-01 7.70409584e-01 -6.47849619e-01 9.74666655e-01 4.20983255e-01 -1.98714793e-01 -1.06113657e-01 1.33212134e-01 1.02301173e-01 -2.38368407e-01 7.27754593e-01 9.17983115e-01 6.04613662e-01 -1.77481666e-01 3.80263627e-01 5.27942836e-01 1.12133082e-02 4.41151142e-01 -7.94573963e-01 -1.46690041e-01 7.02762663e-01 1.14352477e+00 -2.66373098e-01 -4.09481317e-01 -4.65709686e-01 1.15366936e+00 -3.06291163e-01 3.82129669e-01 -9.63472366e-01 -8.64689350e-01 1.39605507e-01 -2.46382147e-01 1.49647355e-01 4.26058084e-01 3.66608381e-01 -1.21428263e+00 1.18460350e-01 -8.94988954e-01 8.09696913e-01 -1.14141655e+00 -1.61440432e+00 5.62689602e-01 4.98978570e-02 -1.89641964e+00 -3.15116584e-01 -2.77225286e-01 -5.29916823e-01 1.03679836e+00 -1.27131414e+00 -1.41683245e+00 -6.13666594e-01 1.55154026e+00 2.96471357e-01 -9.74616408e-01 7.55263329e-01 6.82491124e-01 -1.07079465e-02 8.02390039e-01 4.00842458e-01 2.29818106e-01 8.12723637e-01 -5.18855512e-01 -8.37891579e-01 5.16763091e-01 -2.44445220e-01 4.14232105e-01 1.89540535e-01 -8.50029707e-01 -1.69325531e+00 -9.91041005e-01 4.31185007e-01 -7.68248066e-02 5.90016127e-01 4.31701005e-01 -5.89324474e-01 1.62123963e-01 1.29471973e-01 -1.12901255e-01 7.42896438e-01 -6.77122056e-01 -4.47168171e-01 -4.46972966e-01 -1.52343452e+00 3.41061652e-01 8.28305781e-01 -8.71881247e-01 -8.95078480e-01 2.80006289e-01 8.27037394e-01 -3.13354353e-03 -1.18348527e+00 6.28282309e-01 8.45566332e-01 -6.18081450e-01 1.09110212e+00 -6.72469318e-01 5.10624051e-01 -6.98394001e-01 -6.16735280e-01 -1.17337155e+00 -3.06726813e-01 -3.49746943e-02 4.52924013e-01 1.48072147e+00 1.09262571e-01 -5.41585505e-01 3.24181795e-01 2.92661518e-01 9.15952842e-04 -2.40951717e-01 -1.14855516e+00 -8.48667383e-01 -2.33106822e-01 1.22646466e-01 7.91575909e-01 1.17695022e+00 5.71095943e-03 8.26253146e-02 -3.84823412e-01 -4.23560366e-02 4.72235829e-01 2.78403848e-01 3.11424732e-01 -1.02500534e+00 1.23782200e-03 -3.74924958e-01 -7.93518782e-01 -5.60535967e-01 3.59432340e-01 -1.08735275e+00 -1.25223309e-01 -1.26817393e+00 4.15578842e-01 -4.22033012e-01 -6.39786720e-01 3.04311782e-01 3.31901908e-02 3.95511061e-01 3.76761913e-01 4.59912956e-01 -7.54503727e-01 7.09322393e-01 1.35639608e+00 -6.00486517e-01 1.32019222e-02 -2.32419953e-01 -4.74663258e-01 4.76756006e-01 5.10546625e-01 -2.25437582e-01 -4.34318542e-01 -3.00327897e-01 -1.41157642e-01 5.53141892e-01 4.11314994e-01 -1.02699614e+00 4.39631194e-01 -3.21448237e-01 6.14429057e-01 -5.27843535e-01 4.58390117e-01 -1.54166663e+00 4.36627775e-01 4.35217112e-01 -2.55004942e-01 2.91553233e-02 9.97477490e-03 7.06075788e-01 -6.26109183e-01 -4.25599098e-01 4.21742380e-01 -2.10854128e-01 -1.12639785e+00 4.15274411e-01 -1.02030136e-01 -9.58191976e-03 1.22406828e+00 -5.87580383e-01 -2.13430583e-01 -4.55887705e-01 -9.03148413e-01 3.49984556e-01 3.74327660e-01 8.90818954e-01 1.05383980e+00 -2.14290428e+00 -5.33727884e-01 2.15234533e-01 4.39358592e-01 -6.18137538e-01 5.77229798e-01 4.21745121e-01 -3.28807086e-01 2.46043339e-01 -4.68933403e-01 -5.11456072e-01 -1.50092745e+00 8.89010847e-01 3.89028758e-01 9.76736248e-02 2.15634137e-01 4.69642580e-01 -4.44624387e-02 -2.38730893e-01 8.74926336e-03 6.53808236e-01 -5.15554190e-01 3.11125129e-01 3.09108555e-01 4.08462405e-01 -1.48697838e-01 -1.23960531e+00 -4.52383965e-01 1.10095024e+00 9.27251652e-02 -1.97085723e-01 8.53099942e-01 -2.66558051e-01 -3.89010221e-01 2.22185299e-01 1.81486082e+00 4.49562967e-02 -4.21966732e-01 -3.83105308e-01 -6.58442140e-01 -8.50567758e-01 -5.85000999e-02 -7.98931181e-01 -1.34079683e+00 4.70998168e-01 1.19714165e+00 2.65378326e-01 1.49212015e+00 -1.39161255e-02 7.86184311e-01 1.07635379e-01 3.07998061e-01 -1.21224904e+00 3.16358119e-01 1.95396513e-01 9.36962783e-01 -1.19629300e+00 -1.62654221e-01 -3.38663012e-01 -5.80548942e-01 1.16217959e+00 6.87406600e-01 -7.72882327e-02 8.24837029e-01 -3.61412555e-01 -4.96367887e-02 -1.30302921e-01 -1.17579989e-01 -1.02922164e-01 6.32805228e-01 6.73869312e-01 -2.20245376e-01 -2.75107354e-01 -4.69999373e-01 5.78754723e-01 2.54407734e-01 -1.34763122e-01 -1.80211827e-01 9.75159645e-01 -4.59077626e-01 -1.18921554e+00 -6.97183251e-01 3.07253808e-01 -2.97801137e-01 -8.00941736e-02 -2.80454069e-01 5.10692418e-01 5.52151620e-01 1.23878419e+00 -1.23969717e-02 -1.16168237e+00 2.75361508e-01 1.23089984e-01 4.31470484e-01 7.60579184e-02 -4.37998116e-01 3.49020064e-02 -3.19608957e-01 -4.88113523e-01 -6.97272480e-01 -4.60706502e-01 -1.16737628e+00 -2.15715662e-01 -5.36981046e-01 4.74479258e-01 5.80441296e-01 8.49056065e-01 3.79465550e-01 3.29324216e-01 1.26253331e+00 -5.59591770e-01 -3.91720504e-01 -6.71308637e-01 -5.20818174e-01 1.06189978e+00 7.00783730e-02 -8.79621446e-01 -3.15550357e-01 2.30415821e-01]
[11.177681922912598, 1.1961500644683838]
324e1360-d736-49a8-82b6-d573526d336b
wearable-seld-dataset-dataset-for-sound-event
2202.08458
null
https://arxiv.org/abs/2202.08458v1
https://arxiv.org/pdf/2202.08458v1.pdf
Wearable SELD dataset: Dataset for sound event localization and detection using wearable devices around head
Sound event localization and detection (SELD) is a combined task of identifying the sound event and its direction. Deep neural networks (DNNs) are utilized to associate them with the sound signals observed by a microphone array. Although ambisonic microphones are popular in the literature of SELD, they might limit the range of applications due to their predetermined geometry. Some applications (including those for pedestrians that perform SELD while walking) require a wearable microphone array whose geometry can be designed to suit the task. In this paper, for the development of such a wearable SELD, we propose a dataset named Wearable SELD dataset. It consists of data recorded by 24 microphones placed on a head and torso simulators (HATS) with some accessories mimicking wearable devices (glasses, earphones, and headphones). We also provide experimental results of SELD using the proposed dataset and SELDNet to investigate the effect of microphone configuration.
['Yasuhiro Oikawa', 'Shoichiro Saito', 'Kohei Yatabe', 'Masahiro Yasuda', 'Kento Nagatomo']
2022-02-17
null
null
null
null
['sound-event-localization-and-detection']
['audio']
[-6.63609281e-02 -4.82322037e-01 4.16288704e-01 -3.47330123e-01 -2.67297715e-01 -2.08991334e-01 -6.48519099e-02 -2.09321886e-01 -3.28452349e-01 4.23607051e-01 4.73856151e-01 -1.48821145e-01 2.51367778e-01 -7.50232041e-01 -5.92915595e-01 -5.81516325e-01 -5.67630753e-02 -3.59300047e-01 3.45838010e-01 2.39218339e-01 -1.14600874e-01 1.85643867e-01 -1.77705979e+00 -2.63326950e-02 4.35692757e-01 1.19770610e+00 3.62726301e-01 4.97501582e-01 2.23383859e-01 2.32402548e-01 -8.44992995e-01 -8.06392133e-02 4.69978079e-02 -3.96420509e-01 2.94513255e-01 -1.80971369e-01 2.12639064e-01 -4.08552349e-01 -2.65332699e-01 8.49291265e-01 1.24622428e+00 2.04478264e-01 3.61578852e-01 -1.01094985e+00 -2.13634372e-01 4.38343465e-01 -3.73173296e-01 4.21811938e-01 6.27212465e-01 7.00080991e-02 4.86463219e-01 -8.38296294e-01 -9.53486040e-02 7.71463633e-01 8.95518482e-01 5.45565546e-01 -5.86154521e-01 -8.18878651e-01 -1.04485966e-01 3.73302966e-01 -1.51846147e+00 -7.12365866e-01 1.14494312e+00 -2.70205170e-01 7.42358983e-01 3.49848837e-01 6.81536794e-01 1.45027256e+00 1.78476959e-01 2.23658189e-01 7.34377146e-01 -2.48440072e-01 7.29097307e-01 2.45104000e-01 6.42299578e-02 1.50639966e-01 3.26917648e-01 -1.12091333e-01 -7.12429702e-01 -1.41791061e-01 7.52700567e-01 9.89306048e-02 -3.91288012e-01 5.69367185e-02 -1.09321940e+00 3.67160231e-01 3.50002676e-01 3.21570098e-01 -6.92474008e-01 -2.81280354e-02 5.46957910e-01 -1.88339233e-01 1.80254519e-01 2.43136212e-02 -9.00280401e-02 -1.58143401e-01 -6.37992799e-01 4.07371968e-02 9.66821849e-01 7.06993282e-01 -5.30162416e-02 2.74351448e-01 8.62865224e-02 1.00506210e+00 4.15619224e-01 4.38124746e-01 7.13760197e-01 -4.77722824e-01 4.53381509e-01 1.92240998e-01 2.20242113e-01 -1.28393841e+00 -6.98665142e-01 -5.59392571e-01 -9.18684423e-01 -3.18746954e-01 1.10077135e-01 -6.43871129e-01 -3.95542175e-01 1.66752458e+00 4.63344395e-01 7.00947046e-01 -1.58590168e-01 1.26675022e+00 1.23052049e+00 5.61839879e-01 1.62696674e-01 -1.62627682e-01 1.56272495e+00 -4.19127256e-01 -8.71283650e-01 -2.14862019e-01 -7.62366951e-02 -4.85015780e-01 1.19835734e+00 5.17936647e-01 -7.27847576e-01 -8.84217322e-01 -1.05289829e+00 3.07778746e-01 -2.66058594e-01 3.30549479e-01 2.13436216e-01 1.03594029e+00 -7.33223319e-01 -8.22668076e-02 -8.39848399e-01 -5.39880812e-01 -1.01944819e-01 3.17735314e-01 -5.55632934e-02 5.74104369e-01 -1.27587402e+00 3.40631545e-01 -1.84411734e-01 5.74347973e-01 -9.42382455e-01 -2.39971906e-01 -6.29831731e-01 2.38645181e-01 -1.02019712e-01 -7.34555602e-01 1.17986035e+00 -5.29503584e-01 -1.82849431e+00 4.37516242e-01 -4.53880988e-02 -3.90125513e-01 1.97697788e-01 -5.51724911e-01 -9.82970178e-01 1.97904393e-01 2.88291704e-02 4.98420782e-02 7.69848764e-01 -7.55732059e-01 -4.82216954e-01 -4.41276461e-01 -1.07046463e-01 6.77186474e-02 -8.13143432e-01 2.89614230e-01 1.26349494e-01 -6.02681696e-01 1.66838750e-01 -6.15269601e-01 1.90492552e-02 -2.01393917e-01 -7.17651308e-01 -1.68451265e-01 5.75440288e-01 -8.05775762e-01 1.39048541e+00 -2.31088829e+00 -4.84415978e-01 6.69697300e-02 -7.12994486e-02 3.61954302e-01 2.02574849e-01 3.16153347e-01 1.11198360e-02 -2.73640215e-01 1.35919049e-01 -4.30488139e-01 -5.74414618e-02 -2.57904679e-02 -2.69282252e-01 5.48205256e-01 -2.56163895e-01 1.42698199e-01 -4.46601361e-01 -2.42583215e-01 3.46909553e-01 5.73755085e-01 -4.64205712e-01 6.34943068e-01 2.60552198e-01 5.38701117e-01 -4.96557802e-01 5.71196377e-01 6.35807395e-01 4.23745781e-01 -1.15621783e-01 -3.78812879e-01 -2.14469492e-01 4.00960475e-01 -1.56120563e+00 1.20576549e+00 -6.52487099e-01 4.63486373e-01 2.97202080e-01 -9.57932293e-01 1.16566074e+00 7.68868625e-01 2.30463624e-01 -4.56370592e-01 4.26689237e-01 1.12186410e-01 -1.76697858e-02 -1.25296354e+00 6.72650188e-02 -6.80100694e-02 1.21600226e-01 1.08141527e-01 -2.25818887e-01 4.76565242e-01 -2.87147284e-01 -5.14928639e-01 1.10651612e+00 -2.13201642e-01 2.17682466e-01 -1.06066711e-01 5.14430106e-01 -6.51429415e-01 7.27969468e-01 6.21770322e-01 -2.67028421e-01 7.56168604e-01 -1.02058619e-01 -4.29870039e-01 -5.21285594e-01 -1.14468944e+00 -1.38825327e-01 1.13925111e+00 1.77206248e-01 -1.80469155e-01 -7.46165216e-01 -1.65243894e-01 -3.05961579e-01 6.28822327e-01 -1.08537056e-01 -7.57615045e-02 -5.39872527e-01 -7.48172760e-01 8.72665644e-01 7.78886378e-01 8.72012675e-01 -1.17878127e+00 -9.73406196e-01 5.15537858e-01 -1.66298822e-01 -1.31269550e+00 -4.01948690e-01 2.19144985e-01 -4.05331165e-01 -7.82599568e-01 -6.59531236e-01 -9.85999048e-01 2.68770128e-01 2.00460017e-01 6.66844904e-01 -6.11712098e-01 -1.01697862e-01 5.22074819e-01 -2.55741656e-01 -7.33018637e-01 2.33374685e-01 -1.74503639e-01 5.13693094e-01 4.04749453e-01 4.17052716e-01 -1.29732728e+00 -1.12122011e+00 5.17635822e-01 -6.02207243e-01 -3.80772829e-01 1.75169453e-01 2.45522454e-01 1.74368694e-01 3.71807307e-01 1.00637102e+00 -1.05383210e-01 9.51847732e-01 -7.49666452e-01 -1.86657324e-01 -1.12661712e-01 3.34153414e-01 -9.24810350e-01 7.63610899e-01 -6.50853217e-01 -1.06994081e+00 1.35625189e-04 -5.97415864e-01 -8.86052176e-02 -7.31485605e-01 3.37626934e-01 -7.54541337e-01 1.83385357e-01 5.64453542e-01 1.54256493e-01 -4.28140819e-01 -8.61510634e-01 -2.32589200e-01 1.38045669e+00 7.93441534e-01 -3.82958084e-01 6.03385605e-02 1.94984108e-01 -2.51166880e-01 -1.05824387e+00 -3.06995213e-01 -4.76553619e-01 -3.68602760e-02 -4.62621778e-01 9.78209794e-01 -9.91964936e-01 -9.98522162e-01 7.17186213e-01 -1.20663631e+00 1.62729576e-01 2.71218866e-01 9.39758301e-01 -9.62776542e-02 -2.56128628e-02 -5.11066973e-01 -1.21139872e+00 -4.88874227e-01 -8.61847103e-01 8.67208481e-01 5.43562174e-01 -3.91707093e-01 -6.70398355e-01 2.55235713e-02 2.75597960e-01 4.81737554e-01 3.64159465e-01 5.28698862e-01 -5.82128048e-01 9.16609764e-02 -3.51696849e-01 3.29776675e-01 5.16313732e-01 3.37820351e-01 -3.49973977e-01 -1.34737515e+00 9.19996127e-02 6.73340917e-01 5.12751862e-02 3.54044497e-01 7.48668492e-01 1.35129809e+00 -2.57174581e-01 -3.35582346e-01 5.12901247e-01 1.15391803e+00 7.23707438e-01 5.24721086e-01 5.14131226e-02 6.61190987e-01 5.21525502e-01 1.30603882e-02 5.90788364e-01 4.78410482e-01 6.57809377e-01 4.23663646e-01 1.20515697e-01 -1.03729971e-01 -2.54777908e-01 6.21538222e-01 9.75767195e-01 -4.05284762e-03 -6.81371689e-01 -7.28444636e-01 5.84453642e-01 -1.34406614e+00 -5.62787950e-01 -2.63764828e-01 2.15480137e+00 4.75094587e-01 4.00315113e-02 3.83513004e-01 4.46637243e-01 1.10627675e+00 -3.91746461e-02 -5.82947552e-01 -3.69389981e-01 6.94371238e-02 3.45954686e-01 -7.07167666e-04 5.92622533e-03 -1.06833744e+00 3.67342271e-02 5.84762430e+00 3.21515203e-01 -1.40913081e+00 2.07535952e-01 2.55490184e-01 -1.52091742e-01 1.46449849e-01 -7.20756352e-01 -7.83570349e-01 8.14474344e-01 1.04444957e+00 4.75543141e-01 7.98910186e-02 9.15033937e-01 6.83080316e-01 -1.96426868e-01 -1.17200422e+00 1.27122259e+00 -7.73809552e-02 -5.96718132e-01 -4.06225383e-01 -3.19693804e-01 2.36301258e-01 -2.63504684e-01 6.91051036e-02 -8.80433396e-02 -5.66299200e-01 -6.24730647e-01 6.81206107e-01 3.71479273e-01 4.86584872e-01 -5.14848471e-01 6.71739280e-01 2.97787786e-01 -1.19373071e+00 -1.62774175e-01 -2.63132632e-01 -5.95721006e-01 3.18012863e-01 7.19644964e-01 -7.61619925e-01 1.34707704e-01 9.54119503e-01 4.28894550e-01 -2.62982279e-01 1.44394398e+00 -1.29372850e-01 1.13909054e+00 -7.06619501e-01 -3.83151263e-01 -2.89279938e-01 4.19284739e-02 8.89313877e-01 1.10893703e+00 7.60099471e-01 1.81137305e-02 -1.12389296e-01 7.77469635e-01 2.53595691e-02 1.21678889e-01 -7.48218596e-01 4.43734467e-01 7.55975962e-01 1.08369625e+00 -5.02207160e-01 1.93547476e-02 -4.23703492e-01 5.40894806e-01 -4.37777907e-01 5.24723291e-01 -8.77230346e-01 -6.93209529e-01 6.93423212e-01 3.46265614e-01 2.08030745e-01 -2.11679459e-01 -5.08776426e-01 -6.94640458e-01 5.20226359e-01 -5.24187207e-01 8.94108638e-02 -9.89171743e-01 -1.39975739e+00 5.99555790e-01 -1.30961299e-01 -1.33528888e+00 1.56521127e-01 -3.08658957e-01 -1.11106086e+00 7.29819238e-01 -8.58377397e-01 -5.89265764e-01 -5.71184695e-01 6.34531975e-01 4.04765159e-01 -4.68894765e-02 7.69697845e-01 8.36231053e-01 -9.35880423e-01 5.81290841e-01 -1.00425370e-01 4.03754003e-02 5.77902317e-01 -7.30346322e-01 2.49781728e-01 7.20479369e-01 -7.67294392e-02 7.93047845e-01 9.08101261e-01 -4.54165995e-01 -1.24435520e+00 -1.05777860e+00 6.00083411e-01 -1.52384455e-03 2.82247961e-01 -6.18108928e-01 -6.84970021e-01 4.29344177e-01 2.38125443e-01 3.10079474e-02 8.85420918e-01 -1.41862825e-01 1.33067727e-01 -6.23214066e-01 -1.22055531e+00 5.60501158e-01 1.19948959e+00 -4.50285465e-01 -6.52728260e-01 1.23444036e-01 3.59411359e-01 -2.93147713e-01 -6.89426839e-01 3.00637275e-01 7.93802321e-01 -1.06026971e+00 8.77367079e-01 -2.93309484e-02 1.09086536e-01 -3.35805058e-01 -2.59881586e-01 -1.26353490e+00 -4.34735417e-03 -4.93407160e-01 -4.04452486e-03 1.58857822e+00 -3.07301939e-01 -8.96957994e-01 5.46320796e-01 4.80679274e-01 -2.00091243e-01 -5.43415129e-01 -1.17395973e+00 -6.29350781e-01 -6.57270193e-01 -6.71133220e-01 7.87580550e-01 3.91638130e-01 -4.44441140e-02 4.39630687e-01 -3.88769805e-01 5.82983851e-01 4.29694295e-01 -4.62854952e-01 3.96377742e-01 -1.09331369e+00 -1.36645019e-01 3.54283117e-02 -6.68266296e-01 -1.09942222e+00 -2.17641890e-01 -1.37071937e-01 2.53234863e-01 -1.25736570e+00 -2.66155005e-01 -2.33867154e-01 -3.80315274e-01 2.76003461e-02 -1.64945759e-02 2.16265365e-01 -3.18161249e-01 -3.79633278e-01 -3.67035776e-01 5.90459526e-01 5.87136984e-01 7.20573813e-02 -4.37328547e-01 4.46995527e-01 -4.98887628e-01 1.05372679e+00 8.70346844e-01 -3.80444914e-01 -4.98700351e-01 -6.15156591e-01 -6.53628036e-02 2.65071988e-01 8.01140249e-01 -1.51258981e+00 6.51024640e-01 4.18246806e-01 4.20452774e-01 -4.26925898e-01 4.39904869e-01 -9.80316818e-01 3.54957819e-01 2.80882776e-01 -1.95482299e-01 1.41894504e-01 1.31090850e-01 4.12797630e-01 -1.86848700e-01 -4.52630669e-02 4.29184705e-01 6.18922189e-02 -4.11566854e-01 -6.18487857e-02 -7.62835324e-01 -2.73940176e-01 8.14860821e-01 -3.10601890e-01 -9.83309671e-02 -5.24733186e-01 -8.86251211e-01 -2.08833441e-01 -2.61348546e-01 4.24565613e-01 8.44024420e-01 -1.42081678e+00 -2.05773637e-01 4.52586472e-01 -2.37973496e-01 -1.33237578e-02 3.30533415e-01 6.71123445e-01 -1.38589770e-01 3.27937037e-01 -2.35388905e-01 -5.05577445e-01 -1.15213382e+00 1.64225772e-01 5.74812949e-01 4.23925579e-01 -6.86379731e-01 8.54410589e-01 2.53405273e-01 -1.07904531e-01 7.86694586e-01 -5.71487725e-01 -4.86884832e-01 -1.57300621e-01 6.27323151e-01 6.79516196e-01 1.71653897e-01 -2.21518472e-01 -6.60060108e-01 5.59930801e-01 6.33557737e-01 -5.20901829e-02 1.31143081e+00 -4.23972011e-01 1.39386803e-01 4.24146026e-01 1.02109754e+00 1.36963591e-01 -7.48836696e-01 -3.54129560e-02 -2.80657917e-01 -9.38983783e-02 2.89380141e-02 -4.99078304e-01 -1.18297684e+00 9.92980599e-01 1.06466305e+00 3.14554513e-01 1.34266019e+00 -3.03301543e-01 1.19660914e+00 2.90704697e-01 4.98354882e-01 -8.40098500e-01 1.57379191e-02 4.48995568e-02 7.69227803e-01 -7.59782374e-01 -6.08027816e-01 -2.09113955e-01 -2.50586003e-01 1.06184566e+00 7.73334980e-01 -2.66809642e-01 9.68464255e-01 3.75913441e-01 1.00036308e-01 -7.75464848e-02 -2.76471764e-01 -5.40115014e-02 -8.02499503e-02 9.42443907e-01 3.34135920e-01 7.55559951e-02 -2.76293755e-01 1.46151555e+00 -4.14661497e-01 2.07671657e-01 2.19505638e-01 7.39654720e-01 -4.02383476e-01 -4.65407789e-01 -6.62239134e-01 4.08536077e-01 -6.98775113e-01 1.34034246e-01 1.95784811e-02 1.67830557e-01 7.15983689e-01 1.53555202e+00 8.03649053e-02 -6.96304142e-01 7.29632020e-01 -9.19798017e-02 1.17876925e-01 -4.70350146e-01 -6.82966292e-01 7.25113377e-02 1.48573071e-01 -2.58998305e-01 -3.76749337e-01 -5.09754598e-01 -1.03654528e+00 6.72225878e-02 -8.01461190e-02 1.22603461e-01 7.21307337e-01 6.88119173e-01 3.99280727e-01 8.66431355e-01 6.94672346e-01 -7.51209617e-01 -3.04202408e-01 -1.13511014e+00 -9.52705562e-01 7.10963085e-02 3.78181964e-01 -7.52596140e-01 -3.31299901e-01 1.80300251e-01]
[15.020191192626953, 5.622372627258301]
437088a7-11c3-4a71-9560-4f70e6825a67
towards-selection-of-text-to-speech-data-to
2306.00998
null
https://arxiv.org/abs/2306.00998v1
https://arxiv.org/pdf/2306.00998v1.pdf
Towards Selection of Text-to-speech Data to Augment ASR Training
This paper presents a method for selecting appropriate synthetic speech samples from a given large text-to-speech (TTS) dataset as supplementary training data for an automatic speech recognition (ASR) model. We trained a neural network, which can be optimised using cross-entropy loss or Arcface loss, to measure the similarity of a synthetic data to real speech. We found that incorporating synthetic samples with considerable dissimilarity to real speech, owing in part to lexical differences, into ASR training is crucial for boosting recognition performance. Experimental results on Librispeech test sets indicate that, in order to maintain the same speech recognition accuracy as when using all TTS data, our proposed solution can reduce the size of the TTS data down below its $30\,\%$, which is superior to several baseline methods.
['Ozlem Kalinli', 'Jay Mahadeokar', 'Yuan Shangguan', 'Gil Keren', 'Chunyang Wu', 'Leda Sari', 'Shuo Liu']
2023-05-30
null
null
null
null
['automatic-speech-recognition']
['speech']
[ 7.04251945e-01 3.64930063e-01 -7.98061192e-02 -6.52490437e-01 -1.30551767e+00 -4.47624862e-01 6.86867654e-01 -2.33020157e-01 -5.04041672e-01 6.63963318e-01 3.09576392e-01 -6.86218917e-01 2.90266186e-01 -1.99293941e-01 -5.75851381e-01 -5.12755156e-01 1.88643768e-01 4.28318977e-01 -4.06189635e-02 -2.79102325e-01 8.23250934e-02 6.56666696e-01 -1.60378075e+00 5.08541703e-01 8.84146810e-01 1.07014978e+00 4.16060656e-01 7.15430081e-01 -3.08122009e-01 6.33562207e-01 -1.22596824e+00 -3.16382408e-01 3.86654675e-01 -4.36717123e-01 -6.77333951e-01 1.87194198e-01 4.51000988e-01 -5.60924299e-02 -6.31028116e-01 1.09734333e+00 7.56199121e-01 3.74945551e-01 6.52638614e-01 -8.31021786e-01 -3.46260101e-01 8.66371453e-01 1.78780276e-02 4.44300175e-01 4.42761511e-01 8.78327638e-02 7.82612920e-01 -1.13253105e+00 2.91998655e-01 1.42488408e+00 3.93671691e-01 6.47663534e-01 -1.25905526e+00 -7.61218727e-01 2.14386955e-02 -1.06147029e-01 -1.47793245e+00 -1.39538741e+00 6.88482881e-01 -6.12464594e-03 1.33846867e+00 7.19333708e-01 3.12164217e-01 1.47641015e+00 -3.30764174e-01 9.46922660e-01 9.89006162e-01 -8.32195044e-01 1.86976731e-01 3.64998013e-01 -8.27043429e-02 2.03541175e-01 -2.68736541e-01 2.78934926e-01 -5.21656334e-01 -1.68422177e-01 2.52961665e-01 -8.07420194e-01 -5.36778629e-01 1.80002719e-01 -1.11582208e+00 6.38127685e-01 6.96256161e-02 3.01019728e-01 -3.46326649e-01 -2.71833897e-01 4.65870768e-01 6.58010423e-01 5.15325725e-01 4.86916333e-01 -4.49591249e-01 -3.98409098e-01 -9.61593807e-01 -9.87205505e-02 6.69722617e-01 9.29880738e-01 3.15684527e-01 8.35732877e-01 -1.55433893e-01 1.45652843e+00 1.95207044e-01 7.48780191e-01 1.02496958e+00 -6.67887092e-01 1.03118253e+00 1.97580084e-01 -1.35544300e-01 -5.78891158e-01 2.90789604e-01 -3.51873487e-01 -6.92813277e-01 -2.66002268e-02 2.55189955e-01 -1.06247365e-01 -1.07090545e+00 1.59202039e+00 5.91184087e-02 -2.31507733e-01 4.58909273e-01 6.24731302e-01 6.33888721e-01 9.87279236e-01 -1.99047506e-01 -4.90308195e-01 9.01500642e-01 -7.00937808e-01 -7.37833619e-01 -4.90855217e-01 7.29909122e-01 -1.04729605e+00 1.32074523e+00 1.51450053e-01 -1.04361582e+00 -6.46803856e-01 -1.15079331e+00 4.68228340e-01 -3.85540694e-01 2.14976251e-01 -1.10390536e-01 1.11212218e+00 -1.07948494e+00 4.20502305e-01 -4.03126270e-01 -2.18269005e-01 4.29244302e-02 2.43390739e-01 -2.94909775e-01 4.63668555e-02 -1.39783907e+00 9.16656137e-01 6.36622071e-01 4.91459295e-02 -6.05868399e-01 -3.12521487e-01 -9.68190193e-01 -5.53390495e-02 2.04996139e-01 1.26838014e-01 1.47842813e+00 -1.15412784e+00 -1.72551703e+00 6.25675619e-01 -3.32301974e-01 -6.81427479e-01 4.33773696e-01 2.98480809e-01 -1.03545284e+00 4.13298458e-02 -2.81185716e-01 4.64098364e-01 1.30620039e+00 -1.16847289e+00 -4.72149462e-01 -1.64114729e-01 -5.30162692e-01 4.31467175e-01 -2.82723069e-01 4.28322762e-01 -2.26532817e-01 -1.12050653e+00 1.51636124e-01 -9.03806388e-01 -1.11397043e-01 -6.44047201e-01 -5.25056958e-01 -2.10679486e-01 7.75489628e-01 -8.69450748e-01 1.32692420e+00 -2.28305650e+00 -6.03430867e-02 5.24535894e-01 -3.13345164e-01 8.78328741e-01 -3.61489236e-01 5.24136126e-01 -4.83075380e-01 1.91859692e-01 -1.82772279e-01 -3.08577478e-01 -8.18628147e-02 1.42766178e-01 -6.56939209e-01 2.30763361e-01 2.52702504e-01 4.58993107e-01 -5.84160745e-01 -1.98209882e-01 2.02641636e-01 3.24108779e-01 3.75200175e-02 4.60717261e-01 -5.97113324e-03 3.40530202e-02 -4.87388484e-02 3.48770887e-01 4.76324558e-01 2.78912336e-01 8.56565405e-03 1.61866665e-01 2.40607530e-01 9.86379802e-01 -1.10603261e+00 9.62806582e-01 -6.21995270e-01 9.32406664e-01 1.44138306e-01 -9.72554564e-01 1.26093554e+00 6.10394001e-01 5.23461662e-02 -9.01451766e-01 -6.98327571e-02 4.31590080e-01 2.29793683e-01 -9.06236693e-02 5.72308540e-01 -1.48754254e-01 2.43736595e-01 3.93048942e-01 -2.20252559e-01 -3.85278165e-01 8.63834191e-03 5.05984612e-02 9.55366671e-01 -5.25182784e-01 9.86832157e-02 -6.07203171e-02 6.81440234e-01 -2.28996143e-01 3.22471291e-01 8.22255671e-01 -1.70124948e-01 5.80288649e-01 2.09540147e-02 -3.86692435e-02 -1.25403249e+00 -1.03130412e+00 -1.69032678e-01 7.16773450e-01 -4.40172076e-01 -3.75904590e-01 -9.38233733e-01 -6.08642220e-01 -2.29480281e-01 1.21887553e+00 -1.01380497e-01 -2.27835283e-01 -7.89469361e-01 -5.84229767e-01 1.04667485e+00 3.13729972e-01 1.90232590e-01 -1.22617698e+00 7.77162910e-02 3.41393948e-01 -2.35264897e-01 -1.26584482e+00 -9.06435311e-01 4.97944295e-01 -6.02770448e-01 -5.59231102e-01 -7.93228984e-01 -8.16293716e-01 5.63757777e-01 2.84956515e-01 7.66480684e-01 -1.46393672e-01 4.15303595e-02 -2.78318990e-02 -5.70712090e-01 -1.07127905e-01 -1.41481757e+00 7.21852630e-02 3.33082259e-01 9.14430842e-02 2.79746234e-01 -1.32081255e-01 -8.42450187e-02 4.51396286e-01 -6.97471321e-01 -1.25471845e-01 5.81663132e-01 1.01380706e+00 2.91895747e-01 1.90063760e-01 7.78399646e-01 -3.95270854e-01 1.00075233e+00 -5.59214912e-02 -3.30719829e-01 2.89807826e-01 -6.60814106e-01 5.74750081e-02 9.63849247e-01 -6.37420356e-01 -9.52140450e-01 -1.40690625e-01 -5.36306798e-01 -5.16847193e-01 -2.75891215e-01 4.26403373e-01 -2.90433645e-01 5.47519065e-02 7.61210084e-01 6.77038372e-01 3.77248526e-01 -4.08651203e-01 1.71009541e-01 1.58249485e+00 1.44348487e-01 -3.53087872e-01 6.23826087e-01 -3.00711423e-01 -7.24089026e-01 -1.43509126e+00 -4.17583019e-01 -4.23220336e-01 -3.50075394e-01 9.55635458e-02 1.93988577e-01 -8.28526676e-01 -2.32312486e-01 4.15157765e-01 -1.06997013e+00 -3.61106873e-01 -4.19073343e-01 6.52818501e-01 -4.96248275e-01 4.51318115e-01 -4.73133326e-01 -1.19053674e+00 -3.69089067e-01 -1.30040479e+00 9.72255647e-01 -2.35180482e-01 -3.10433000e-01 -6.42225444e-01 -1.42348364e-01 5.35585701e-01 4.43962932e-01 -4.71121281e-01 7.51668870e-01 -1.21536946e+00 -1.82830572e-01 -3.97842497e-01 -5.81862777e-02 1.05438125e+00 4.00271118e-01 6.42260909e-02 -1.07667542e+00 -5.60347199e-01 1.11879006e-01 -3.55074078e-01 5.71951807e-01 1.06993869e-01 1.08948195e+00 -8.06899011e-01 -7.08250403e-02 3.63920540e-01 7.38736510e-01 8.02802384e-01 7.38992512e-01 1.04651488e-01 3.36875170e-01 6.75062776e-01 5.20771563e-01 1.28687575e-01 -1.36268035e-01 8.81640136e-01 -2.82586008e-01 1.34032732e-02 -4.43002731e-01 -2.68134862e-01 7.97078133e-01 1.42351866e+00 6.88572109e-01 -4.87562329e-01 -9.51002002e-01 4.71892476e-01 -1.20042813e+00 -1.09672725e+00 2.53271669e-01 2.39266229e+00 9.61351693e-01 4.48086143e-01 1.49757221e-01 4.37165767e-01 1.04643595e+00 4.57845718e-01 -3.79812241e-01 -7.09727645e-01 -4.00963157e-01 2.73917407e-01 4.25108910e-01 5.72612286e-01 -7.34454036e-01 9.48470891e-01 7.19278574e+00 1.52956986e+00 -1.36233342e+00 -2.59946197e-01 7.86820769e-01 1.70555096e-02 -3.68878871e-01 -3.34826589e-01 -8.05725813e-01 5.88924766e-01 1.66414666e+00 -2.88195819e-01 7.33975053e-01 6.77879632e-01 4.64748323e-01 4.16929871e-01 -8.80276620e-01 1.04656363e+00 2.14290395e-01 -1.05618203e+00 2.77435869e-01 -1.17917927e-02 2.89927512e-01 3.23938251e-01 1.96585670e-01 3.26050311e-01 3.01520020e-01 -1.17322767e+00 7.89142191e-01 -2.44655743e-01 1.09353888e+00 -8.08435440e-01 5.06264269e-01 4.51721370e-01 -1.03098297e+00 9.07687545e-02 -2.12519497e-01 3.73825133e-01 -5.39990291e-02 2.39933744e-01 -1.64598250e+00 3.73051614e-01 3.26012224e-01 1.74719989e-01 -4.51366752e-01 6.19875968e-01 2.10745871e-01 8.69077921e-01 -3.94123226e-01 -3.02024186e-01 2.71190286e-01 -4.48594615e-02 6.98981941e-01 1.39220583e+00 3.73182863e-01 -1.72457814e-01 -6.33338690e-02 4.47647452e-01 -1.54984161e-01 3.90765786e-01 -8.20825219e-01 -4.98618037e-01 9.28144038e-01 4.31005001e-01 -3.68913502e-01 -4.79000956e-01 -1.47798955e-01 8.53779554e-01 2.69643396e-01 5.43960154e-01 -3.95333260e-01 -8.11731160e-01 6.59185231e-01 -8.87077302e-03 2.83017874e-01 -8.92333984e-02 -1.66355655e-01 -1.11124170e+00 2.98003972e-01 -1.52832866e+00 -1.04939416e-01 -6.45071030e-01 -1.05407786e+00 1.28147185e+00 -3.09245586e-01 -1.23572588e+00 -7.57479131e-01 -4.93303329e-01 -3.09093028e-01 1.30711198e+00 -1.08733559e+00 -5.91880977e-01 3.05021048e-01 3.40797514e-01 1.06617260e+00 -8.39179277e-01 8.03623796e-01 1.32754445e-01 -5.15085399e-01 1.07097602e+00 4.49437350e-01 2.27766111e-01 4.15813535e-01 -1.03818226e+00 1.07703173e+00 9.49400961e-01 4.49220777e-01 5.97000420e-01 7.48727560e-01 -4.89135176e-01 -1.12245119e+00 -9.79143798e-01 9.95664954e-01 -3.68825436e-01 5.59451401e-01 -4.62305099e-01 -1.11434782e+00 5.05164027e-01 1.66102991e-01 -3.99497122e-01 5.50374210e-01 -1.18833482e-01 -4.32905942e-01 -2.31913924e-01 -1.09433436e+00 8.43215287e-01 8.59475076e-01 -1.00530434e+00 -8.06157649e-01 1.74536169e-01 1.01431656e+00 -2.92859614e-01 -8.03251624e-01 3.55249465e-01 3.06687355e-01 -6.23652816e-01 9.71429467e-01 -6.52889132e-01 -1.91865921e-01 -9.93569866e-02 -5.29821217e-01 -1.77085054e+00 1.52775183e-01 -9.33204830e-01 6.51794896e-02 1.17051673e+00 7.21499622e-01 -6.18642449e-01 7.34225094e-01 4.58437264e-01 -3.23002309e-01 -5.03470421e-01 -1.36014259e+00 -1.21520770e+00 -5.24234883e-02 -7.44761705e-01 6.90455616e-01 8.21793735e-01 -1.47007272e-01 3.41021448e-01 -2.39562988e-01 7.99063221e-02 5.11920452e-01 -5.05296052e-01 6.62772655e-01 -6.58124328e-01 -2.28756759e-02 -5.25106788e-01 -3.37856680e-01 -1.25293279e+00 5.17289221e-01 -8.89491796e-01 3.11227113e-01 -1.01266277e+00 -6.00029886e-01 -6.36356711e-01 -2.70623714e-01 1.72967479e-01 -4.86390218e-02 -6.35234639e-02 2.47168854e-01 -9.83478595e-03 1.33822486e-02 8.91416252e-01 9.51571465e-01 -3.50609124e-01 -2.00470850e-01 3.52528691e-01 -2.86712289e-01 3.06271553e-01 7.77790844e-01 -4.64129359e-01 -4.87373024e-01 -1.28145888e-01 -5.65563262e-01 4.17838395e-01 -2.15791181e-01 -8.64196599e-01 9.51288491e-02 -8.26023594e-02 1.20879136e-01 -5.63375652e-01 6.37540698e-01 -6.80694520e-01 1.45652354e-01 3.52681845e-01 -7.67965317e-01 5.38941892e-03 2.12205112e-01 4.12485510e-01 -5.28836012e-01 -3.99012238e-01 8.97423804e-01 1.28075451e-01 -4.45461512e-01 -3.54563110e-02 -6.58179224e-01 2.19786763e-01 4.65172768e-01 -4.86879736e-01 -5.60325123e-02 -5.90165854e-01 -5.55616200e-01 -2.36417085e-01 2.65623808e-01 7.54344821e-01 8.41458619e-01 -1.38093507e+00 -7.68220782e-01 5.93129933e-01 2.79372722e-01 -3.89169723e-01 -1.30255461e-01 2.33977035e-01 -2.93150514e-01 6.26728833e-01 2.56640613e-01 -3.62862080e-01 -1.44552135e+00 3.84534478e-01 4.83672082e-01 9.08099040e-02 -4.29859936e-01 7.66511559e-01 -1.65520251e-01 -6.76990807e-01 6.59277678e-01 -2.02163354e-01 1.92925513e-01 -1.63364545e-01 5.35719335e-01 1.91608533e-01 5.27078450e-01 -9.83081520e-01 -3.86813611e-01 -2.70284880e-02 -2.72612810e-01 -6.10470533e-01 8.78021955e-01 -2.44600937e-01 4.06557471e-01 4.96905178e-01 1.37664950e+00 2.08599374e-01 -9.38495934e-01 -3.39058667e-01 1.74182117e-01 -6.55822754e-01 3.73943038e-02 -7.44472206e-01 -7.88436055e-01 6.09880030e-01 4.69694257e-01 6.22975111e-01 9.14079309e-01 -2.14232937e-01 7.65618682e-01 6.77765548e-01 2.42784455e-01 -1.37111306e+00 -8.89395252e-02 6.41596198e-01 1.03783512e+00 -1.32600379e+00 -5.37693620e-01 -2.59242535e-01 -8.22318554e-01 1.00318158e+00 4.15065497e-01 1.90418065e-01 5.25845826e-01 2.12530211e-01 3.21459711e-01 4.37207967e-01 -9.05530334e-01 -7.52469525e-02 3.24364573e-01 7.02974975e-01 3.95714134e-01 4.72577140e-02 8.21909606e-02 5.68441786e-02 -5.58205128e-01 -6.38643146e-01 4.14949954e-01 5.49673915e-01 -5.55049777e-01 -1.19446254e+00 -5.20660281e-01 6.19554579e-01 -3.71824026e-01 -4.20146614e-01 -5.52556574e-01 5.12799203e-01 -5.32018840e-01 1.25576723e+00 1.12571903e-01 -5.56058288e-01 5.68734884e-01 2.72380650e-01 1.41088024e-01 -6.16272032e-01 -6.75380468e-01 1.78669557e-01 6.28092885e-01 -2.03203931e-01 3.62889543e-02 -6.18537545e-01 -8.69108200e-01 -2.09358260e-01 -4.64882821e-01 4.86315638e-01 9.23354626e-01 9.02162671e-01 2.85136223e-01 3.06473821e-01 1.10648489e+00 -4.45464760e-01 -1.18170738e+00 -1.24077713e+00 -4.85116243e-01 3.53383631e-01 5.63041151e-01 -2.25444466e-01 -6.32587671e-01 -8.58781114e-02]
[14.49475383758545, 6.61788272857666]
da0f7e55-0c58-4a91-9c82-f31f1bffdea7
adapting-neural-link-predictors-for-complex
2301.12313
null
https://arxiv.org/abs/2301.12313v3
https://arxiv.org/pdf/2301.12313v3.pdf
Adapting Neural Link Predictors for Data-Efficient Complex Query Answering
Answering complex queries on incomplete knowledge graphs is a challenging task where a model needs to answer complex logical queries in the presence of missing knowledge. Prior work in the literature has proposed to address this problem by designing architectures trained end-to-end for the complex query answering task with a reasoning process that is hard to interpret while requiring data and resource-intensive training. Other lines of research have proposed re-using simple neural link predictors to answer complex queries, reducing the amount of training data by orders of magnitude while providing interpretable answers. The neural link predictor used in such approaches is not explicitly optimised for the complex query answering task, implying that its scores are not calibrated to interact together. We propose to address these problems via CQD$^{\mathcal{A}}$, a parameter-efficient score \emph{adaptation} model optimised to re-calibrate neural link prediction scores for the complex query answering task. While the neural link predictor is frozen, the adaptation component -- which only increases the number of model parameters by $0.03\%$ -- is trained on the downstream complex query answering task. Furthermore, the calibration component enables us to support reasoning over queries that include atomic negations, which was previously impossible with link predictors. In our experiments, CQD$^{\mathcal{A}}$ produces significantly more accurate results than current state-of-the-art methods, improving from $34.4$ to $35.1$ Mean Reciprocal Rank values averaged across all datasets and query types while using $\leq 30\%$ of the available training query types. We further show that CQD$^{\mathcal{A}}$ is data-efficient, achieving competitive results with only $1\%$ of the training complex queries, and robust in out-of-domain evaluations.
['Isabelle Augenstein', 'Michael Cochez', 'Daniel Daza', 'Pasquale Minervini', 'Erik Arakelyan']
2023-01-29
null
null
null
null
['complex-query-answering']
['knowledge-base']
[ 8.96200314e-02 6.43708527e-01 -1.89202636e-01 -5.29644012e-01 -1.11209846e+00 -6.21889472e-01 2.29991719e-01 3.16755027e-01 -5.54994941e-01 7.72219718e-01 -2.51573473e-01 -7.92422533e-01 -4.34111089e-01 -1.24819934e+00 -1.22053325e+00 1.49089977e-01 -1.02554247e-01 1.02392900e+00 5.48832834e-01 -6.16917074e-01 -6.89966828e-02 1.69627711e-01 -1.61278248e+00 3.78704131e-01 9.00652826e-01 1.24252546e+00 -2.90054560e-01 7.96234846e-01 -3.65145892e-01 9.35196638e-01 -6.25801325e-01 -9.45643485e-01 1.33790806e-01 -2.05710977e-01 -1.22723401e+00 -7.99116135e-01 6.91511095e-01 -2.55723745e-01 -1.92613602e-01 9.02743220e-01 2.80769169e-01 -8.99458677e-02 3.96220207e-01 -1.16441679e+00 -5.97040474e-01 6.92674577e-01 8.94178171e-03 1.96819842e-01 5.11703372e-01 -3.21904644e-02 1.62628520e+00 -8.18028271e-01 5.36548495e-01 1.15779364e+00 5.40588856e-01 4.52126980e-01 -1.33212948e+00 -7.25374341e-01 1.66654244e-01 3.27387571e-01 -1.47695720e+00 -3.94657314e-01 4.95283574e-01 -1.71785161e-01 1.50160182e+00 3.54796797e-01 2.80255318e-01 4.81283575e-01 -1.96663529e-01 6.04447365e-01 5.25002241e-01 -3.67885023e-01 1.20092206e-01 2.92941574e-02 3.77523035e-01 1.08602309e+00 3.72969657e-01 -9.42745432e-02 -5.66371918e-01 -2.93580532e-01 5.14916539e-01 -2.91237205e-01 -9.35752392e-02 -2.36103922e-01 -7.71203399e-01 8.33131850e-01 6.40094817e-01 8.97153765e-02 -2.43813634e-01 5.39578974e-01 3.15413088e-01 7.49991715e-01 1.29263416e-01 8.34326386e-01 -9.51586962e-01 -9.43677798e-02 -5.74165821e-01 3.91724676e-01 1.35842788e+00 1.18967116e+00 1.08579624e+00 -2.23010078e-01 -2.84952074e-01 7.12580204e-01 1.95902333e-01 4.85528886e-01 -2.66624719e-01 -1.34173119e+00 8.12820971e-01 1.18015826e+00 8.83483514e-02 -7.39300728e-01 -3.75914514e-01 -6.84184670e-01 -4.36860591e-01 6.56121671e-02 6.85567439e-01 -8.92254859e-02 -7.30541468e-01 2.02255416e+00 1.54014975e-01 -2.26258352e-01 2.13022709e-01 5.95695734e-01 7.42574334e-01 3.98456693e-01 1.76447198e-01 2.02061180e-02 1.55439532e+00 -7.47801900e-01 -3.01678568e-01 -4.26283360e-01 1.07683897e+00 -5.15038252e-01 1.50261176e+00 4.27818924e-01 -1.34745133e+00 -4.19282138e-01 -1.09361041e+00 -4.16137338e-01 -4.85089928e-01 -1.58066675e-01 7.93552220e-01 4.63791490e-01 -1.20799506e+00 2.82600194e-01 -4.68473047e-01 -4.09217365e-02 2.89454013e-01 7.98493505e-01 -1.49517164e-01 -4.41589445e-01 -1.55407846e+00 1.01243353e+00 4.96676564e-01 -1.89622417e-02 -5.96048951e-01 -8.93817306e-01 -7.51668453e-01 4.29612875e-01 9.64456975e-01 -1.08925760e+00 1.38388121e+00 -4.03124422e-01 -1.22572911e+00 6.73248112e-01 -4.54178870e-01 -4.13631648e-01 3.51339191e-01 -3.29042017e-01 -4.35917228e-01 5.21749295e-02 3.34385186e-02 6.57116592e-01 3.83475840e-01 -1.01818538e+00 -5.39494336e-01 -2.93909490e-01 7.85333276e-01 2.97526661e-02 -1.55556962e-01 -2.83109516e-01 -8.87313068e-01 -2.39298612e-01 1.72862530e-01 -7.94422209e-01 8.66956115e-02 -4.39473055e-03 -1.56036660e-01 -5.73277295e-01 5.87622344e-01 -4.65262890e-01 1.43113065e+00 -1.75778210e+00 4.54093106e-02 4.70890552e-01 3.23134243e-01 2.58217216e-01 -1.31286234e-01 3.22703749e-01 4.84000295e-02 2.79167145e-01 -3.03659409e-01 -1.51382804e-01 2.59330809e-01 5.77185094e-01 -3.50694478e-01 -2.58137614e-01 4.04751599e-01 1.17211401e+00 -7.00823545e-01 -5.51509023e-01 -4.21133786e-01 1.88505098e-01 -1.04132521e+00 2.06360757e-01 -1.10041380e+00 -1.40168130e-01 -4.39464122e-01 7.65483022e-01 3.24030995e-01 -6.44986749e-01 2.45657966e-01 -2.54576415e-01 4.66857761e-01 5.62967181e-01 -1.27791297e+00 1.78864706e+00 -5.29528737e-01 2.65807152e-01 1.95246205e-01 -9.26906765e-01 8.13723683e-01 2.35102206e-01 2.25520119e-01 -1.00393260e+00 -1.90427378e-01 4.92102265e-01 5.36489040e-02 -5.79535723e-01 3.34496021e-01 -1.61578119e-01 -2.78201312e-01 4.20072198e-01 4.39903699e-02 -4.37680274e-01 4.58223224e-01 3.91697675e-01 1.61751318e+00 9.79622528e-02 -2.20804110e-01 5.58118559e-02 7.53425717e-01 2.20632970e-01 4.84747171e-01 1.00055194e+00 2.24733070e-01 4.60234322e-02 8.65378320e-01 -3.73491794e-01 -7.72180855e-01 -9.98248577e-01 1.96345821e-01 1.55628872e+00 -8.60502571e-03 -5.11351228e-01 -6.59998953e-01 -7.37800419e-01 6.73791096e-02 9.89104271e-01 -1.97834671e-01 -2.65078098e-01 -8.20293367e-01 -3.68678093e-01 9.90957379e-01 6.20102644e-01 5.45102775e-01 -9.68833566e-01 -2.78241068e-01 2.64262140e-01 -3.17777634e-01 -1.01496148e+00 8.27085376e-02 2.99600422e-01 -8.22373509e-01 -1.32444537e+00 -7.20821507e-03 -4.57716525e-01 5.77984989e-01 -3.77221227e-01 1.71974254e+00 4.65318829e-01 5.78648085e-03 3.98193210e-01 -5.84204905e-02 -5.59228003e-01 -3.48419160e-01 3.42685044e-01 -3.19568247e-01 -4.64205235e-01 5.67978978e-01 -5.11215806e-01 -5.30820072e-01 3.54973584e-01 -1.17049575e+00 -2.95557618e-01 8.19568574e-01 8.44533205e-01 6.34772062e-01 -5.28259985e-02 6.48956954e-01 -1.25923789e+00 6.99560583e-01 -4.13561583e-01 -7.96320677e-01 5.12938559e-01 -9.98414397e-01 5.68974435e-01 7.39425242e-01 -9.94222313e-02 -8.11588824e-01 -2.49656856e-01 -2.11080581e-01 -2.36888930e-01 1.23879895e-01 9.14880395e-01 -1.21406071e-01 1.04051016e-01 1.07261693e+00 -1.33170187e-01 -1.42031848e-01 -3.10316890e-01 5.18435299e-01 1.49766132e-01 6.42846823e-01 -1.01905143e+00 9.06814396e-01 1.81022301e-01 2.12113068e-01 -2.93542117e-01 -9.97489572e-01 -2.75765210e-01 -1.81440845e-01 2.92159677e-01 5.50021172e-01 -8.52129459e-01 -1.22243118e+00 -8.87295231e-02 -1.16773236e+00 -3.97683531e-01 -2.49372423e-01 2.17498332e-01 -4.54849929e-01 2.42803499e-01 -5.22847235e-01 -5.81557631e-01 -4.30966884e-01 -1.07704210e+00 8.68193507e-01 -2.81949788e-02 -5.48426211e-01 -9.03266013e-01 -1.82039380e-01 7.59826005e-01 5.93898952e-01 -4.57555838e-02 1.66935229e+00 -8.85312200e-01 -9.56787765e-01 -4.30188477e-01 -4.82338727e-01 3.86502326e-01 -3.64522338e-01 -4.97837633e-01 -9.58514571e-01 -3.70683707e-02 -2.06086770e-01 -5.76391160e-01 6.97550416e-01 -2.07848802e-01 1.09612215e+00 -4.59527820e-01 -1.85834974e-01 2.26102874e-01 1.23641527e+00 -5.69599457e-02 4.72886950e-01 2.61909571e-02 4.66421843e-01 4.81605738e-01 3.43730360e-01 -2.62779109e-02 7.21275270e-01 5.26546717e-01 6.98409736e-01 1.16980799e-01 7.06757754e-02 -3.08165014e-01 2.59789359e-02 1.79332063e-01 7.60878995e-02 -1.93112195e-01 -1.23278344e+00 4.84272748e-01 -1.89598620e+00 -6.99410260e-01 -1.51729733e-01 2.09051633e+00 1.19060922e+00 6.25657380e-01 -2.44747937e-01 1.65520072e-01 4.81875911e-02 -2.07921609e-01 -7.46039748e-01 -3.09912711e-01 -2.90785599e-02 8.36058915e-01 2.84142524e-01 8.81502271e-01 -5.94952643e-01 9.74098444e-01 5.75615788e+00 5.72569966e-01 -8.08267832e-01 -1.13806352e-01 1.38438910e-01 -1.15590610e-01 -7.01312482e-01 3.91589910e-01 -8.14974904e-01 6.20606020e-02 1.15924180e+00 1.01073392e-01 5.99170923e-01 8.17187965e-01 -2.86760539e-01 -2.74401933e-01 -1.68694675e+00 6.17711246e-01 -1.19484775e-01 -1.42178178e+00 5.97090125e-02 -1.91084951e-01 2.70231754e-01 -3.55219729e-02 -6.27297238e-02 9.64203477e-01 5.58470309e-01 -1.24479413e+00 4.38591957e-01 7.91260242e-01 6.64252937e-01 -5.40404439e-01 7.87380040e-01 4.92869049e-01 -1.11869049e+00 -2.96139449e-01 -4.96984087e-03 -2.12530524e-01 -1.03471488e-01 6.24853671e-01 -1.06175601e+00 6.32681131e-01 8.03481698e-01 4.63528484e-02 -7.55921066e-01 5.15467465e-01 -3.54700476e-01 5.66737592e-01 -7.10985243e-01 -1.01809487e-01 -1.48607045e-02 2.80230045e-01 1.43706352e-01 9.89916801e-01 9.20852944e-02 3.25270116e-01 6.36084899e-02 1.00890493e+00 -5.07660091e-01 -1.59018472e-01 -3.90293300e-01 -2.89122071e-02 4.71683532e-01 7.28738964e-01 -7.61995837e-02 -5.04872084e-01 -3.16984028e-01 5.80275595e-01 6.71472371e-01 6.25363886e-01 -8.33452225e-01 -6.07315004e-01 4.74929452e-01 3.05364311e-01 2.76283413e-01 -2.51456589e-01 -3.68110240e-01 -9.89379883e-01 6.59313440e-01 -8.35573614e-01 9.28917527e-01 -8.02379668e-01 -1.12952602e+00 3.76439422e-01 2.42765248e-01 -5.00127435e-01 -6.49010718e-01 -6.63308561e-01 -1.98950812e-01 1.17700469e+00 -1.79237652e+00 -1.02684879e+00 -3.21709812e-01 6.84026599e-01 -4.36703973e-02 1.31981643e-02 1.05241215e+00 4.10224080e-01 -2.89077401e-01 8.27318192e-01 -4.77203667e-01 1.87949404e-01 5.54140925e-01 -1.18058193e+00 6.49332479e-02 5.10776043e-01 -6.94787204e-02 8.89108360e-01 5.76821566e-01 -4.53117222e-01 -1.67363644e+00 -8.40269387e-01 1.33461463e+00 -7.05805182e-01 6.25726700e-01 -2.28019863e-01 -1.16354942e+00 7.32916772e-01 -7.57926181e-02 1.21669821e-01 7.92740166e-01 4.30553496e-01 -8.52562189e-01 -4.89284456e-01 -1.14573967e+00 5.48230469e-01 1.23718965e+00 -8.62554491e-01 -6.97192311e-01 2.52109021e-01 9.97777581e-01 -3.55888546e-01 -1.28384590e+00 6.72703922e-01 3.20741117e-01 -9.02618349e-01 1.02547216e+00 -8.20865452e-01 2.97855467e-01 -4.14927483e-01 -3.83973718e-01 -5.95464826e-01 -5.49745634e-02 -4.83988076e-01 -6.44041896e-01 8.66287112e-01 1.02199888e+00 -7.88890243e-01 1.07905173e+00 1.13580358e+00 -1.92130208e-01 -9.33400095e-01 -9.22546089e-01 -3.53219062e-01 4.34055217e-02 -7.53499806e-01 5.16128957e-01 6.54755056e-01 -9.21310708e-02 7.04133511e-01 3.46123815e-01 3.27828377e-01 3.47818404e-01 3.06651089e-02 8.58242989e-01 -1.40328443e+00 -6.45641267e-01 -4.55554545e-01 2.82514971e-02 -1.14079738e+00 2.71308661e-01 -9.03600633e-01 -2.88599908e-01 -1.72895133e+00 -3.10348600e-01 -7.60969818e-01 -1.71840996e-01 1.02346563e+00 -1.37933418e-01 -1.65313616e-01 -6.72954395e-02 1.04116991e-01 -6.45242989e-01 2.58337140e-01 9.78555620e-01 -3.59651774e-01 1.74776986e-02 6.44395426e-02 -8.97243083e-01 4.92843986e-01 4.48349148e-01 -3.76824260e-01 -8.44880700e-01 -8.17020237e-01 1.10439372e+00 2.72446692e-01 6.23798430e-01 -9.19032156e-01 6.07563317e-01 -2.30420977e-02 6.75784722e-02 -4.45269674e-01 6.70013666e-01 -9.43215191e-01 -1.43261738e-02 4.21239465e-01 -3.93334508e-01 2.85570979e-01 3.09755027e-01 5.05284131e-01 -3.20995599e-01 -2.09123001e-01 4.31881577e-01 -2.74116904e-01 -6.04110122e-01 9.52668563e-02 2.77988583e-01 5.21253467e-01 6.99769199e-01 2.24634446e-03 -5.87173879e-01 -4.84110296e-01 -6.81981683e-01 6.67670786e-01 9.08733308e-02 1.23516105e-01 5.18626571e-01 -9.19970214e-01 -5.36385357e-01 8.82762447e-02 2.66822726e-01 6.51353955e-01 -1.66756511e-02 6.80920124e-01 -5.89068592e-01 6.67874932e-01 3.08286905e-01 -4.03116077e-01 -8.21987391e-01 4.01559412e-01 5.16559362e-01 -7.80978203e-01 1.13494448e-01 1.07393491e+00 -3.75227302e-01 -7.88524866e-01 4.22126919e-01 -5.67984641e-01 1.49710774e-01 -1.13148436e-01 5.48025817e-02 3.42152894e-01 3.84887725e-01 7.08367750e-02 -3.45131516e-01 2.95356244e-01 -1.04371294e-01 -1.14468761e-01 1.18277431e+00 3.04733962e-01 -3.23852718e-01 -9.58409812e-03 1.17397404e+00 -2.11169526e-01 -7.23528385e-01 -6.06723428e-01 2.29095981e-01 4.88156080e-02 -3.70320767e-01 -1.30575502e+00 -7.05272257e-01 8.15927505e-01 2.67344236e-01 2.98852950e-01 1.01959538e+00 1.87450498e-01 6.21060789e-01 1.22210562e+00 3.76179904e-01 -1.00872838e+00 2.00672284e-01 8.29844356e-01 8.33444118e-01 -1.23694921e+00 -7.96580780e-03 -4.13893878e-01 -1.09999076e-01 8.08948159e-01 8.40178311e-01 1.99029267e-01 5.10976732e-01 8.26510414e-02 -1.07484162e-01 -5.84993005e-01 -1.02989733e+00 -1.12987287e-01 3.02045226e-01 2.64238864e-01 4.03168768e-01 -2.37968564e-01 -7.39225969e-02 6.03377461e-01 -3.37014794e-01 7.33793303e-02 8.84731188e-02 1.10101092e+00 -3.21729153e-01 -1.30587208e+00 -1.80276453e-01 6.15347624e-01 -2.93130577e-01 -2.08994374e-01 -4.54075575e-01 8.61617029e-01 6.78435788e-02 1.08524156e+00 -1.59701407e-01 -2.76359409e-01 7.92951405e-01 6.13259137e-01 3.33641052e-01 -5.97244203e-01 -7.04241753e-01 -4.79915738e-01 5.43132126e-01 -5.57072461e-01 -6.32334352e-02 -1.14468545e-01 -1.69107878e+00 -3.58539164e-01 -1.18956074e-01 3.61518472e-01 4.23646748e-01 9.85959589e-01 6.72560811e-01 4.87940460e-01 -8.17324370e-02 1.98062230e-03 -8.29083025e-01 -7.71105051e-01 9.61618423e-02 3.29096287e-01 1.58975631e-01 -4.92745727e-01 -5.02408892e-02 -1.25815839e-01]
[9.478941917419434, 7.714141845703125]
c1ca1bfc-8e76-4f38-818e-fde4f8af0b7b
doubly-contrastive-end-to-end-semantic
2211.11131
null
https://arxiv.org/abs/2211.11131v1
https://arxiv.org/pdf/2211.11131v1.pdf
Doubly Contrastive End-to-End Semantic Segmentation for Autonomous Driving under Adverse Weather
Road scene understanding tasks have recently become crucial for self-driving vehicles. In particular, real-time semantic segmentation is indispensable for intelligent self-driving agents to recognize roadside objects in the driving area. As prior research works have primarily sought to improve the segmentation performance with computationally heavy operations, they require far significant hardware resources for both training and deployment, and thus are not suitable for real-time applications. As such, we propose a doubly contrastive approach to improve the performance of a more practical lightweight model for self-driving, specifically under adverse weather conditions such as fog, nighttime, rain and snow. Our proposed approach exploits both image- and pixel-level contrasts in an end-to-end supervised learning scheme without requiring a memory bank for global consistency or the pretraining step used in conventional contrastive methods. We validate the effectiveness of our method using SwiftNet on the ACDC dataset, where it achieves up to 1.34%p improvement in mIoU (ResNet-18 backbone) at 66.7 FPS (2048x1024 resolution) on a single RTX 3080 Mobile GPU at inference. Furthermore, we demonstrate that replacing image-level supervision with self-supervision achieves comparable performance when pre-trained with clear weather images.
['Jong-Hwan Kim', 'Jongoh Jeong']
2022-11-21
null
null
null
null
['road-scene-understanding', 'real-time-semantic-segmentation']
['computer-vision', 'computer-vision']
[ 2.90067703e-01 -1.40675828e-01 -6.84900442e-03 -6.25559747e-01 -3.55377853e-01 -2.40571856e-01 2.43558303e-01 -1.38356969e-01 -8.48273635e-01 5.99949956e-01 -6.69564903e-01 -7.82926917e-01 4.78546917e-01 -9.11794364e-01 -9.46062803e-01 -6.08267248e-01 3.27455997e-03 4.06422168e-01 9.01319146e-01 -4.48914647e-01 1.38913706e-01 3.11184436e-01 -2.11446357e+00 -1.03556983e-01 1.36345541e+00 1.06867051e+00 5.34727752e-01 1.00655472e+00 1.51341185e-02 5.46608150e-01 -4.91213650e-01 -2.13146627e-01 4.91064370e-01 1.37616647e-02 -5.24353981e-01 1.68973908e-01 9.00850058e-01 -5.96696138e-01 -3.68175715e-01 1.01405942e+00 3.82881522e-01 1.67204514e-01 1.21574752e-01 -1.16063201e+00 3.31533492e-01 -2.42454167e-02 -7.24168420e-01 6.74562395e-01 -4.32913005e-01 5.06473303e-01 4.92909938e-01 -4.50855076e-01 2.26040050e-01 1.01402247e+00 4.62096632e-01 3.16485971e-01 -6.95499480e-01 -8.27110112e-01 1.57966793e-01 5.74619114e-01 -1.16876769e+00 -5.19455791e-01 3.48822057e-01 5.23421988e-02 1.00515056e+00 1.59368560e-01 6.94588184e-01 4.51731145e-01 2.79750407e-01 6.96079612e-01 1.31047809e+00 7.30425119e-02 2.87108541e-01 7.58716837e-02 2.04093233e-01 7.71971583e-01 5.25821984e-01 1.53973132e-01 -4.06055063e-01 4.16665435e-01 5.80215216e-01 -1.89310029e-01 9.62522402e-02 8.34403634e-02 -7.94502914e-01 6.39116287e-01 6.21406257e-01 -2.46052399e-01 -2.22468466e-01 4.50396597e-01 4.09324527e-01 1.56919554e-01 7.45745063e-01 -2.31276408e-01 -4.92994636e-01 -2.68194586e-01 -1.05785716e+00 1.35782301e-01 5.78215480e-01 9.94317472e-01 1.12832260e+00 3.86894822e-01 3.34916055e-01 4.79980111e-01 1.39358088e-01 1.08451581e+00 1.80544510e-01 -1.02895105e+00 6.03185713e-01 1.73004061e-01 1.31267026e-01 -7.71804690e-01 -5.10798931e-01 -6.57696187e-01 -7.61804163e-01 4.52885300e-01 2.15510651e-01 -2.83033162e-01 -1.31888247e+00 1.16189635e+00 5.88058829e-01 8.44005466e-01 2.76920706e-01 1.12283683e+00 7.38093674e-01 6.81174159e-01 2.04600632e-01 1.16519287e-01 1.40458357e+00 -1.44823372e+00 -4.92106944e-01 -8.25512648e-01 7.18424618e-01 -6.40093446e-01 1.14308989e+00 2.42251039e-01 -9.64045942e-01 -8.22396815e-01 -1.34829855e+00 -1.79062575e-01 -3.35712403e-01 -1.03267357e-01 7.45091259e-01 9.09494758e-01 -1.13166511e+00 3.74637812e-01 -1.06622910e+00 -2.42406532e-01 5.94739914e-01 3.85005027e-01 1.16101705e-01 -1.78764820e-01 -1.16904724e+00 7.71031976e-01 3.09353441e-01 2.97778279e-01 -8.21534991e-01 -8.08911562e-01 -6.88195884e-01 -1.06257945e-01 4.18485016e-01 -3.90240520e-01 1.14090657e+00 -9.56472218e-01 -1.58780694e+00 7.22300589e-01 -4.88712341e-01 -9.76058841e-01 6.62433684e-01 -4.22434062e-01 -3.98367941e-01 3.14865112e-01 1.14983700e-01 1.00138175e+00 9.06342983e-01 -1.21306431e+00 -1.23529398e+00 -3.54776233e-01 1.22146107e-01 4.79586989e-01 -6.83486313e-02 -2.19837844e-01 -7.22303510e-01 -1.53174281e-01 4.31374349e-02 -1.19863498e+00 -5.40694714e-01 1.57957356e-02 2.91669127e-02 2.14672625e-01 1.47455752e+00 -5.47108948e-01 6.21540010e-01 -2.08961844e+00 -7.37391829e-01 1.03643760e-01 3.33856232e-02 8.28626394e-01 1.61038905e-01 -4.33126986e-01 4.34951067e-01 -2.66643316e-01 -5.24969757e-01 -4.18672204e-01 -3.36649507e-01 6.55548096e-01 -3.06760281e-01 5.51376283e-01 1.64652660e-01 8.86021614e-01 -7.39445746e-01 -6.35996878e-01 6.59539998e-01 6.17946327e-01 -3.87334675e-01 7.03981221e-02 -1.07491665e-01 4.56870079e-01 -3.27843219e-01 4.80895251e-01 1.18013322e+00 -6.66595027e-02 -1.23355061e-01 7.22021535e-02 -3.51455539e-01 3.04521024e-01 -9.81778741e-01 1.37593973e+00 -7.21828699e-01 9.55935597e-01 3.13798875e-01 -9.26876783e-01 7.05147922e-01 -1.82356223e-01 7.16428459e-02 -1.27251351e+00 1.80629253e-01 3.23813766e-01 -1.19720176e-01 -3.47004414e-01 8.55236232e-01 1.31737113e-01 2.01072603e-01 6.86326846e-02 -4.60936904e-01 -2.87492692e-01 3.47780317e-01 -1.13705238e-02 8.42537344e-01 -2.58689541e-02 -3.30828547e-01 -3.78356695e-01 4.76869017e-01 4.14648533e-01 4.84953403e-01 6.97319090e-01 -4.48703915e-01 4.76038426e-01 7.46742589e-03 -4.49446321e-01 -1.01777613e+00 -9.44729090e-01 -3.21984768e-01 1.07983673e+00 7.40456998e-01 9.72990245e-02 -9.16542590e-01 -4.76089895e-01 -1.69770882e-01 6.72336042e-01 -1.16267920e-01 7.79255182e-02 -8.37242603e-01 -9.86806095e-01 5.72967291e-01 6.31184518e-01 1.36929643e+00 -6.79386735e-01 -1.31204009e+00 3.30518454e-01 -1.24625601e-01 -1.73241961e+00 -1.72605589e-01 1.02379277e-01 -1.10227811e+00 -8.79986644e-01 -4.54348594e-01 -7.99956024e-01 5.21499157e-01 9.86065865e-01 1.03483069e+00 2.31289670e-01 -2.93806583e-01 -9.15503651e-02 -1.29724458e-01 -5.24531126e-01 -1.40997395e-01 1.30869433e-01 -2.40938947e-01 -1.64869204e-01 1.81319833e-01 -3.96376401e-01 -9.18373048e-01 5.64554095e-01 -7.53073990e-01 3.51956189e-01 5.17681062e-01 4.79501545e-01 5.53603888e-01 1.34824529e-01 3.37642878e-01 -8.77108932e-01 -1.14009485e-01 -1.70067072e-01 -9.67301011e-01 -1.83344290e-01 -6.66818857e-01 -2.73121506e-01 5.59323311e-01 8.44940320e-02 -1.31135905e+00 2.42733628e-01 -4.21629190e-01 -1.78964987e-01 -2.38727361e-01 4.61418033e-02 -1.87486177e-03 -4.00326341e-01 5.15130341e-01 2.86414146e-01 1.14321895e-01 1.10889673e-01 3.81333292e-01 8.19300354e-01 9.16560411e-01 -2.98485637e-01 8.08786213e-01 1.09612918e+00 4.73095812e-02 -1.18309021e+00 -7.46164560e-01 -6.93751991e-01 -3.88100117e-01 -2.85190463e-01 9.92423534e-01 -1.43344212e+00 -7.36850560e-01 7.71370709e-01 -8.25727582e-01 -7.55153894e-01 5.89192100e-02 3.73003960e-01 -4.61019754e-01 4.50843036e-01 -3.49009722e-01 -5.95154047e-01 -5.90525448e-01 -1.13823116e+00 1.06744254e+00 6.40925586e-01 4.26192760e-01 -8.03122997e-01 -5.79819858e-01 7.54976571e-01 6.97157741e-01 -4.99963090e-02 2.64577329e-01 4.09983024e-02 -8.41456413e-01 5.21087423e-02 -7.53071070e-01 3.46696794e-01 -1.35166690e-01 -1.71168849e-01 -1.23212993e+00 -2.23521546e-01 -1.93724576e-02 -1.99767575e-01 1.12178290e+00 4.58011717e-01 1.23508871e+00 9.40061882e-02 -2.90670276e-01 9.08730209e-01 1.37040877e+00 2.02199727e-01 8.56777132e-01 4.32810843e-01 9.57020104e-01 4.47114855e-01 1.07299304e+00 1.95250273e-01 8.39365780e-01 4.57987398e-01 7.10105538e-01 -6.67914689e-01 -3.22525203e-01 1.93877250e-01 2.06317618e-01 3.72435272e-01 -1.36786297e-01 -3.15531582e-01 -9.13955450e-01 6.40115678e-01 -1.75365734e+00 -6.38626873e-01 -5.14609277e-01 2.23308063e+00 5.47701061e-01 5.63276172e-01 -1.32947579e-01 5.25003625e-03 5.57581306e-01 3.40534836e-01 -8.91993582e-01 -4.48481351e-01 -1.93352252e-01 4.06114608e-01 1.42833805e+00 5.80095530e-01 -1.22398782e+00 1.41336787e+00 5.58525419e+00 8.28749776e-01 -1.44820499e+00 2.72429943e-01 8.89419675e-01 4.85627539e-02 6.98281825e-02 -6.63635209e-02 -1.02471399e+00 4.84911442e-01 1.30517006e+00 1.83938384e-01 5.63863143e-02 8.93609107e-01 6.19004726e-01 -6.85845554e-01 -3.77942443e-01 8.60211670e-01 -1.12406418e-01 -1.23795426e+00 -4.05245095e-01 -4.83653955e-02 8.12864423e-01 6.17090523e-01 1.13634214e-01 1.42202064e-01 2.22194299e-01 -8.47944140e-01 6.26298428e-01 -1.80004314e-01 5.72215259e-01 -8.15677047e-01 8.49244118e-01 2.46267423e-01 -1.26310730e+00 2.11093992e-01 -3.97888541e-01 -2.63518870e-01 3.41838092e-01 6.86917782e-01 -8.94019961e-01 3.30501467e-01 9.86825168e-01 6.18793249e-01 -5.39161384e-01 8.47983539e-01 -2.48418614e-01 1.01018655e+00 -7.12756276e-01 1.36120647e-01 6.31717324e-01 -1.32523254e-01 3.41418654e-01 1.27489936e+00 8.45061839e-02 1.05324455e-01 1.90589935e-01 2.01430038e-01 2.24078298e-01 -3.13499272e-01 -1.10827871e-01 5.26402891e-01 9.04690847e-02 1.20851779e+00 -1.16724634e+00 -7.07932949e-01 -3.74055386e-01 9.85293329e-01 -1.67236716e-01 3.59683603e-01 -1.36769044e+00 -3.68969113e-01 1.08875477e+00 3.03524792e-01 6.39161110e-01 -6.62993431e-01 -6.97544634e-01 -8.82880390e-01 1.81942612e-01 -4.49078590e-01 -6.18498139e-02 -6.10769391e-01 -5.45375049e-01 5.98943770e-01 -1.61597908e-01 -1.07896423e+00 1.24167822e-01 -4.52174872e-01 -5.67496717e-01 6.80603087e-01 -2.50222659e+00 -1.01358044e+00 -9.14957583e-01 5.38393080e-01 7.46561527e-01 3.18563193e-01 2.10505351e-01 5.73544621e-01 -5.32437801e-01 4.75506067e-01 1.07096724e-01 -2.15469837e-01 5.42175949e-01 -1.01969314e+00 1.01372182e+00 1.17360628e+00 -2.76980549e-01 -5.46391234e-02 8.49248707e-01 -5.54858565e-01 -1.37029290e+00 -1.58962131e+00 4.42396551e-01 1.24043792e-01 3.05977762e-01 -3.11400533e-01 -8.93597364e-01 2.54078478e-01 1.46779194e-01 4.31345195e-01 2.36768611e-02 -3.30952138e-01 -4.18885238e-02 -4.54494625e-01 -1.08015704e+00 6.04462504e-01 1.15628433e+00 -1.21250980e-01 3.19328755e-02 4.56621379e-01 8.07846963e-01 -9.49964464e-01 -2.73437798e-01 4.44522440e-01 1.88115194e-01 -1.06112385e+00 8.55854809e-01 2.42575966e-02 1.27963990e-01 -6.81198180e-01 1.14314973e-01 -8.15371871e-01 2.45231539e-01 -5.18487751e-01 1.66623399e-01 5.91806352e-01 3.13657969e-01 -9.65640187e-01 1.23564231e+00 3.56218547e-01 -5.62652409e-01 -6.07302725e-01 -1.06654322e+00 -8.26829970e-01 -2.63687968e-01 -8.01394522e-01 3.77433538e-01 3.96334738e-01 -7.23659992e-01 5.76973408e-02 -2.04840109e-01 5.97609639e-01 7.60585964e-01 9.48576555e-02 1.04667127e+00 -8.70193303e-01 9.36863199e-02 -1.05171390e-01 -6.21783614e-01 -1.50755024e+00 1.18497156e-01 -4.40776676e-01 4.61370438e-01 -1.50212681e+00 -1.98496416e-01 -7.62202919e-01 1.82234533e-02 4.04071271e-01 -3.16564202e-01 6.60007834e-01 1.85179934e-02 1.42353475e-01 -7.30854511e-01 4.83487785e-01 1.25941753e+00 -1.26914084e-01 -2.25089967e-01 9.58001316e-02 -4.51282710e-01 7.78793454e-01 9.99211431e-01 -4.01039600e-01 -6.45551682e-01 -7.00710952e-01 -1.51700571e-01 -2.91276395e-01 6.60186410e-01 -1.52505493e+00 3.57636392e-01 -5.39116301e-02 1.32060777e-02 -8.04606736e-01 4.21341300e-01 -7.00141191e-01 -2.10871965e-01 7.46671557e-01 3.70982438e-01 -3.94682847e-02 6.17285013e-01 5.94130397e-01 -2.10534126e-01 1.15629725e-01 8.94641638e-01 1.26914516e-01 -1.42797601e+00 2.95795649e-01 -4.45301712e-01 1.30774066e-01 1.15529478e+00 -6.14170730e-01 -4.71447468e-01 -1.87194303e-01 -1.37500748e-01 6.26106799e-01 4.22744870e-01 3.68830711e-01 5.33709049e-01 -5.31231642e-01 -6.91966236e-01 3.10886770e-01 -1.28789738e-01 5.26589811e-01 6.69881165e-01 8.29764068e-01 -1.20093739e+00 3.72426003e-01 -1.98857784e-01 -1.06105709e+00 -1.29306650e+00 -3.63849476e-03 2.74973363e-01 1.94651671e-02 -8.51542711e-01 8.12367797e-01 1.86977118e-01 -3.42712253e-02 -8.04086700e-02 -4.79541928e-01 1.56099096e-01 -3.11469376e-01 5.07658958e-01 4.47008580e-01 4.59722072e-01 -6.56099677e-01 -3.22636604e-01 4.82199073e-01 -1.63053557e-01 -1.12802267e-01 1.10375643e+00 -3.62061203e-01 2.64194012e-01 -4.21910733e-02 1.00606799e+00 -4.98940080e-01 -1.83208787e+00 -1.00484252e-01 -3.93237859e-01 -5.72082937e-01 5.91228306e-01 -4.22055006e-01 -1.48200202e+00 8.83897483e-01 1.02718282e+00 -1.88379034e-01 1.22117996e+00 -3.34254473e-01 1.43729162e+00 4.55675691e-01 4.09713238e-01 -1.19007421e+00 -3.18177938e-01 6.48803830e-01 1.20259508e-01 -1.59094524e+00 1.18851848e-01 -7.57307112e-01 -6.78494096e-01 7.35813737e-01 7.84853458e-01 -1.78007558e-01 5.73257327e-01 4.30977523e-01 4.22427773e-01 4.65894537e-03 -4.48970348e-01 -6.24591529e-01 -3.89730968e-02 5.41594923e-01 -1.06846161e-01 1.61006093e-01 -2.06471235e-01 -2.51856357e-01 -3.40195715e-01 -7.31751323e-02 6.24054193e-01 1.11025202e+00 -7.95163095e-01 -5.70869386e-01 -2.00162232e-01 4.33112741e-01 -2.51096636e-01 -7.34401196e-02 4.18149829e-01 6.44498408e-01 2.68979937e-01 1.24779379e+00 5.61300933e-01 -2.26591185e-01 2.62927979e-01 -4.65335697e-01 1.47619054e-01 -3.35381031e-01 -3.51406783e-01 -1.63912818e-01 1.33285880e-01 -6.54692292e-01 -6.43208921e-01 -5.47667265e-01 -1.66036999e+00 -5.61968565e-01 -1.91354781e-01 -1.63282260e-01 1.17482114e+00 1.05714703e+00 5.30112684e-01 5.51725388e-01 6.10727906e-01 -1.03243756e+00 4.68380637e-02 -5.23333848e-01 -3.05274844e-01 -8.07841867e-02 2.82827854e-01 -5.41155875e-01 -2.25326329e-01 1.39500201e-01]
[8.990496635437012, -0.8834424614906311]
371728d0-ff24-44b4-a19a-ec3bd073f4e8
leveraging-patient-similarity-and-time-series
1704.07498
null
http://arxiv.org/abs/1704.07498v3
http://arxiv.org/pdf/1704.07498v3.pdf
Leveraging Patient Similarity and Time Series Data in Healthcare Predictive Models
Patient time series classification faces challenges in high degrees of dimensionality and missingness. In light of patient similarity theory, this study explores effective temporal feature engineering and reduction, missing value imputation, and change point detection methods that can afford similarity-based classification models with desirable accuracy enhancement. We select a piecewise aggregation approximation method to extract fine-grain temporal features and propose a minimalist method to impute missing values in temporal features. For dimensionality reduction, we adopt a gradient descent search method for feature weight assignment. We propose new patient status and directional change definitions based on medical knowledge or clinical guidelines about the value ranges for different patient status levels, and develop a method to detect change points indicating positive or negative patient status changes. We evaluate the effectiveness of the proposed methods in the context of early Intensive Care Unit mortality prediction. The evaluation results show that the k-Nearest Neighbor algorithm that incorporates methods we select and propose significantly outperform the relevant benchmarks for early ICU mortality prediction. This study makes contributions to time series classification and early ICU mortality prediction via identifying and enhancing temporal feature engineering and reduction methods for similarity-based time series classification.
['Samir AbdelRahman', 'Mohammad Amin Morid', 'Olivia R. Liu Sheng']
2017-04-25
null
null
null
null
['icu-mortality']
['medical']
[ 1.69754058e-01 -6.24023557e-01 -2.74764925e-01 -4.63101417e-01 -5.70937514e-01 -2.52078772e-01 5.13272248e-02 6.72205389e-01 -2.25855976e-01 8.53281200e-01 6.33700132e-01 -3.29531491e-01 -1.30554569e+00 -5.39122701e-01 4.65055071e-02 -6.22026205e-01 -7.85412371e-01 5.24489880e-01 -4.01633084e-01 -2.61226296e-01 3.89901698e-01 6.70951545e-01 -1.37062442e+00 7.20675230e-01 8.97927344e-01 1.21831763e+00 -5.74149013e-01 5.25539517e-01 1.09737791e-01 9.72502828e-01 -2.94877350e-01 1.73100069e-01 4.45158213e-01 -6.89589202e-01 -6.25348270e-01 -4.47300851e-01 -8.30555335e-02 -3.36869389e-01 -4.03858840e-01 1.60330757e-01 8.72414589e-01 1.87324867e-01 9.27051961e-01 -1.46553874e+00 -1.81727871e-01 2.94252813e-01 -1.05466209e-01 4.98271137e-01 3.71333420e-01 -2.53798813e-02 3.48913252e-01 -8.98592770e-01 2.63283342e-01 5.98480761e-01 1.48347712e+00 5.17499864e-01 -1.12817597e+00 -3.64553154e-01 -1.32523969e-01 4.18698281e-01 -1.35516024e+00 -3.75042915e-01 8.78898561e-01 -7.58052409e-01 1.09510851e+00 7.77532220e-01 8.11789155e-01 8.19646180e-01 7.54748523e-01 1.63400874e-01 9.66842830e-01 -3.71443748e-01 3.82791370e-01 -3.78347307e-01 5.02160132e-01 4.19557750e-01 6.40510172e-02 4.36133385e-01 -6.01310551e-01 -9.56722498e-01 4.71486211e-01 1.13846517e+00 -3.73096883e-01 -4.35017586e-01 -1.93883801e+00 5.71928680e-01 -1.55950457e-01 1.83177531e-01 -8.37429583e-01 -1.85488075e-01 7.70505250e-01 7.21975923e-01 4.30696517e-01 4.88577008e-01 -1.03470194e+00 -4.45392013e-01 -1.02091706e+00 4.67509963e-02 7.59653091e-01 7.84668803e-01 2.70045757e-01 -2.68144459e-01 -7.11412609e-01 7.72218049e-01 -3.05260569e-01 9.75714847e-02 9.25269604e-01 -1.21270669e+00 3.60182196e-01 7.82791674e-01 3.23989242e-01 -1.01854300e+00 -8.01677823e-01 -3.18288207e-01 -1.09769177e+00 -2.07684278e-01 1.01614170e-01 -1.81957021e-01 -3.82996738e-01 1.42030871e+00 2.56348282e-01 5.46136558e-01 1.74421087e-01 4.56449211e-01 3.37452203e-01 8.81390795e-02 4.75303456e-03 -1.01679313e+00 1.12414610e+00 -3.57200325e-01 -7.70884156e-01 6.33324504e-01 1.26203239e+00 -3.49900603e-01 4.70141768e-01 2.65254915e-01 -8.67040396e-01 -1.19007036e-01 -4.11679178e-01 3.96195769e-01 -4.11518440e-02 -5.88448420e-02 6.75535202e-01 2.39760667e-01 -7.01436222e-01 1.27850986e+00 -1.21875906e+00 -3.65268677e-01 4.04063255e-01 4.87854362e-01 -3.16236407e-01 5.43352924e-02 -1.10839963e+00 8.07704329e-01 -8.34844913e-03 7.57145584e-02 -2.89459497e-01 -1.41796327e+00 -5.99097311e-01 -1.99878395e-01 -3.50581735e-01 -1.42905498e+00 6.84485078e-01 -3.63889188e-01 -8.57226968e-01 5.73500633e-01 -3.80718768e-01 -5.76587975e-01 6.14549398e-01 -1.45791084e-01 -6.06866300e-01 7.72868469e-02 -4.72331187e-03 -3.85493855e-04 7.00422943e-01 -4.15112615e-01 -5.74629307e-01 -7.21393824e-01 -6.73335671e-01 1.88769296e-01 -6.18131697e-01 -3.60092781e-02 5.87486625e-01 -1.00304687e+00 4.32086855e-01 -6.87819242e-01 -6.60874188e-01 -5.42719029e-02 1.36926562e-01 -8.94843936e-02 6.87483311e-01 -6.47356331e-01 1.91661227e+00 -1.95830023e+00 -1.47324577e-01 3.05368990e-01 4.88846779e-01 -4.93297368e-01 1.71138316e-01 7.88051784e-01 -4.01500404e-01 -1.62282243e-01 -4.95149434e-01 -1.28730372e-01 -3.40119243e-01 3.19777638e-01 -4.15565491e-01 5.39398909e-01 -1.09748617e-01 8.79975319e-01 -7.71225333e-01 -3.75974327e-01 4.25265878e-01 4.00166363e-01 -7.07088113e-01 2.71148413e-01 6.64065540e-01 6.51007414e-01 -4.43895012e-01 6.00929856e-01 3.32677126e-01 -1.90540835e-01 -5.14864326e-02 -2.59695768e-01 -1.70264065e-01 2.82057207e-02 -8.59211028e-01 1.67214465e+00 -9.97615904e-02 1.87623128e-01 -7.07211077e-01 -1.11687076e+00 1.05560231e+00 4.91492271e-01 1.64532745e+00 -2.93809593e-01 9.71031040e-02 2.47459933e-01 -1.14266485e-01 -7.94081569e-01 -2.93734614e-02 -1.91398844e-01 -3.55872512e-02 4.14497763e-01 -6.07444823e-01 2.61585236e-01 -2.02501327e-01 -2.13704631e-01 1.70539284e+00 -2.28298768e-01 5.97210228e-01 -4.71686482e-01 3.86447221e-01 2.27122709e-01 9.99540985e-01 6.46305799e-01 -5.31158328e-01 7.46300697e-01 2.14456096e-01 -1.20728850e+00 -8.85241687e-01 -1.00696933e+00 -5.00876486e-01 7.72342563e-01 -2.80617416e-01 -2.90103197e-01 -1.50541291e-01 -4.89056915e-01 4.21676546e-01 4.71784502e-01 -9.19223011e-01 -6.42088056e-01 -9.27480280e-01 -1.22869956e+00 2.66418755e-01 7.18639433e-01 -2.94658959e-01 -7.80484080e-01 -9.82982993e-01 7.17163920e-01 -1.96080863e-01 -3.28616261e-01 -5.66470444e-01 3.49021286e-01 -1.74875510e+00 -1.14739609e+00 -6.65896595e-01 -8.21954787e-01 7.64081538e-01 -4.24572714e-02 8.99698496e-01 -2.12079644e-01 -5.54580092e-01 6.45927370e-01 -3.56496096e-01 -2.11348221e-01 -1.86893985e-01 -3.41826588e-01 5.85439205e-01 -1.07070088e-01 7.26861835e-01 -8.66117060e-01 -1.34102511e+00 3.79288763e-01 -5.84330738e-01 -2.49893218e-01 1.63135350e-01 1.15111256e+00 6.65346742e-01 -8.27815682e-02 9.89240468e-01 -5.28931379e-01 8.80197227e-01 -9.00785267e-01 1.10086031e-01 2.15436295e-01 -1.47186649e+00 -5.79996891e-02 5.35363734e-01 -5.19171596e-01 -3.41255248e-01 1.11733325e-01 3.65897924e-01 -7.30345011e-01 -1.55884787e-01 3.65597814e-01 5.67334712e-01 2.13578239e-01 7.71974266e-01 4.09848958e-01 2.77455568e-01 -5.76646149e-01 -7.11146072e-02 5.60261071e-01 4.48930621e-01 -5.38136244e-01 1.91186607e-01 5.71557581e-01 3.54655802e-01 -2.57983804e-01 -2.64415294e-01 -9.00729597e-01 -1.08177757e+00 8.31611082e-02 3.94788653e-01 -9.27588224e-01 -7.48327017e-01 1.76992118e-01 -8.64903092e-01 7.41067231e-02 -9.28225875e-01 8.66467953e-01 -1.01918960e+00 3.58339429e-01 -5.87369323e-01 -5.47325015e-01 -6.78423345e-01 -6.65358245e-01 7.89882600e-01 -5.96413136e-01 -7.75841951e-01 -1.17996669e+00 5.38680792e-01 -1.94225967e-01 4.41742688e-01 9.54741895e-01 1.19079602e+00 -7.86097050e-01 1.06114216e-01 -3.50422502e-01 2.64906973e-01 8.19200575e-02 7.25134909e-01 -2.87767768e-01 -2.93870568e-01 -4.61170912e-01 3.84861857e-01 5.36171257e-01 5.47089696e-01 6.95383549e-01 1.31951666e+00 -6.35438859e-01 -5.61994672e-01 8.27733815e-01 1.16402149e+00 4.12018806e-01 3.54097009e-01 2.73344636e-01 4.77067113e-01 7.12465584e-01 5.27297318e-01 1.36070681e+00 4.61258799e-01 3.51314098e-01 1.99680664e-02 8.36826339e-02 2.48455584e-01 2.19475135e-01 1.97805278e-02 1.23746002e+00 -1.77982420e-01 3.94074082e-01 -1.21515179e+00 7.10390925e-01 -2.11380696e+00 -1.11845565e+00 -3.33236694e-01 2.30781579e+00 9.36480522e-01 -4.70132262e-01 2.16058627e-01 5.03092468e-01 4.95995402e-01 -5.37862957e-01 -6.53400660e-01 -4.08577383e-01 -2.17655860e-02 1.50115147e-01 3.00475419e-01 3.61486375e-02 -9.35673773e-01 -1.04068398e-01 7.08113670e+00 1.34277165e-01 -7.52027929e-01 2.82499008e-02 5.85381389e-01 -3.54858458e-01 -2.56149583e-02 -2.49298126e-01 -2.46443793e-01 5.31212151e-01 1.18753850e+00 -5.06371617e-01 2.32058689e-01 5.98633647e-01 8.76065135e-01 6.65140152e-01 -1.53469157e+00 1.44838393e+00 -1.43937424e-01 -1.45852363e+00 4.89800610e-02 -1.80788949e-01 9.08738196e-01 -2.23757982e-01 -1.08228587e-01 5.17522283e-02 -2.29973421e-01 -8.81989479e-01 9.40334201e-02 1.39815438e+00 8.45590115e-01 -6.98929727e-01 8.29845011e-01 1.36412904e-01 -1.16413534e+00 -6.92684352e-01 7.85668660e-03 -4.13288087e-01 1.06182270e-01 7.64581501e-01 -7.56282449e-01 4.84778166e-01 7.41435826e-01 1.14043725e+00 -2.00713456e-01 1.29540110e+00 7.93765783e-01 4.90675032e-01 -2.45098799e-01 4.41830993e-01 -3.13492447e-01 -7.29786158e-02 5.79428732e-01 9.34953690e-01 8.78138244e-01 5.67359090e-01 1.73908785e-01 3.37055027e-01 4.10289705e-01 3.00756365e-01 -7.88721502e-01 4.30959523e-01 5.42149901e-01 6.42881393e-01 -2.81848580e-01 -2.98142433e-01 -3.73118967e-01 9.29306149e-01 -3.27545226e-01 2.65273899e-01 -5.59874713e-01 -3.98429692e-01 8.93250763e-01 4.05268490e-01 -2.32840702e-01 -1.09672561e-01 -8.61589491e-01 -1.14127755e+00 3.16572585e-03 -8.00482452e-01 1.09265602e+00 -2.22485468e-01 -1.73229182e+00 4.70514923e-01 2.34064478e-02 -2.12399268e+00 -6.43744528e-01 -2.50537217e-01 -6.14755571e-01 7.91443169e-01 -1.18456650e+00 -4.84958142e-01 -3.08390856e-01 1.06224132e+00 3.56691092e-01 -2.82873869e-01 1.13134980e+00 3.41132432e-01 -3.67125034e-01 4.70910639e-01 6.29500091e-01 -1.01992667e-01 6.80959344e-01 -8.46611381e-01 -9.79420915e-02 2.69988835e-01 -7.26203442e-01 9.42718089e-01 4.60954964e-01 -7.24476278e-01 -1.39028943e+00 -1.42543209e+00 1.10147417e+00 -1.00302470e+00 5.25977969e-01 3.35678786e-01 -9.42538679e-01 4.00737375e-01 -4.36472952e-01 2.78255820e-01 1.21553910e+00 -1.36039108e-01 3.14354450e-02 -3.75456214e-01 -1.43118203e+00 3.09491336e-01 1.27886593e+00 -4.03650790e-01 -1.03510988e+00 4.12239790e-01 4.88013834e-01 4.69543450e-02 -1.72564125e+00 1.16090298e+00 8.14795554e-01 -6.59313858e-01 1.11440682e+00 -1.11935937e+00 9.52919424e-02 -1.69008553e-01 -2.39657015e-01 -1.00252616e+00 -7.46564806e-01 -8.70610118e-01 -5.47253430e-01 6.12495780e-01 1.02674417e-01 -8.20648730e-01 6.51477873e-01 1.07905078e+00 -2.28483841e-01 -9.68318641e-01 -1.23706222e+00 -8.36529434e-01 1.56903550e-01 -2.54355252e-01 9.11290050e-01 1.40528131e+00 4.80182111e-01 -2.34190062e-01 -4.28825110e-01 -1.10750869e-01 6.69052124e-01 3.67124528e-01 2.37100929e-01 -1.56646943e+00 6.42434284e-02 -4.06979203e-01 -5.47445118e-01 -1.25357643e-01 -2.28424415e-01 -9.12422299e-01 -4.59017456e-01 -1.43582928e+00 2.09580272e-01 -6.07261062e-01 -1.12813377e+00 5.94844937e-01 -3.29225399e-02 -3.66161652e-02 -3.85476261e-01 6.68813050e-01 -2.24758729e-01 5.56423128e-01 8.09263229e-01 4.30847406e-02 -7.29871094e-01 2.57233977e-01 -3.34488571e-01 4.46694940e-01 8.06862652e-01 -7.99334824e-01 -4.70046490e-01 -5.03941579e-03 -8.38228837e-02 6.78073227e-01 3.61683339e-01 -1.06900847e+00 2.27362826e-01 -5.10395467e-01 6.07201099e-01 -5.98845243e-01 -1.51430853e-02 -1.18564200e+00 4.76406902e-01 9.11226034e-01 -4.52649146e-01 7.90398240e-01 1.74569890e-01 5.45332611e-01 -8.78057927e-02 4.09332693e-01 4.76462930e-01 2.09746450e-01 -5.24897754e-01 7.02772379e-01 -3.16748202e-01 -2.94853114e-02 1.10665500e+00 -5.68346977e-01 -1.98459718e-02 6.42421544e-02 -1.21515822e+00 1.92734689e-01 1.90284908e-01 4.39014673e-01 1.07256556e+00 -1.75441885e+00 -9.54604268e-01 3.90033066e-01 4.63190824e-01 -6.89587176e-01 4.52913165e-01 1.53783524e+00 -2.85175234e-01 3.16927940e-01 -2.76904225e-01 -6.87272489e-01 -1.12845409e+00 8.28291893e-01 3.30969214e-01 -4.89319302e-02 -9.50386405e-01 2.74224728e-01 -8.24507624e-02 -2.45115936e-01 8.09916258e-02 -5.50268769e-01 -1.94867641e-01 1.97092772e-01 5.16109526e-01 9.57564592e-01 2.94136316e-01 -3.21318686e-01 -7.08526969e-01 6.68565512e-01 2.16596454e-01 4.13861722e-01 1.51175475e+00 -3.44597965e-01 -2.71432161e-01 6.00004852e-01 1.45482612e+00 -5.41328549e-01 -9.46183860e-01 -1.06201582e-01 1.63544223e-01 -4.13118273e-01 -1.92123055e-01 -7.57548511e-01 -4.36183751e-01 4.75183636e-01 1.16411006e+00 -5.14449626e-02 1.48168242e+00 -3.80002141e-01 8.97607803e-01 4.86441314e-01 4.06214178e-01 -7.90798545e-01 -3.92120510e-01 4.40044910e-01 8.56168509e-01 -1.20100284e+00 -7.31322393e-02 2.19568431e-01 -4.08773035e-01 1.17630851e+00 9.81884524e-02 -1.13277212e-01 1.14334738e+00 1.33960471e-01 9.53756645e-02 -3.78178954e-02 -1.13842356e+00 4.09398824e-01 3.26757580e-01 5.16848683e-01 4.21421498e-01 2.90827900e-01 -8.44598830e-01 7.72759318e-01 -1.36278018e-01 3.90629977e-01 1.35645002e-01 1.08289504e+00 -2.10429266e-01 -9.52500165e-01 -3.30054283e-01 1.14992154e+00 -4.10444528e-01 -2.31940225e-01 8.33071321e-02 2.94602334e-01 -6.45253286e-02 8.75432730e-01 1.42567456e-01 -3.65003288e-01 6.82721913e-01 4.42148775e-01 3.46643329e-01 -2.40638375e-01 -7.50928164e-01 -2.63872683e-01 -2.24089637e-01 -7.16426611e-01 -1.99124038e-01 -9.42923784e-01 -1.38563871e+00 -1.30514935e-01 2.36020684e-01 4.85175461e-01 3.39244664e-01 6.44068837e-01 9.72323179e-01 5.90925872e-01 1.10139918e+00 -3.57511789e-01 -8.05402100e-01 -7.02636719e-01 -5.21204472e-01 7.42422640e-01 6.82031691e-01 -4.91565108e-01 -4.89494145e-01 1.56650409e-01]
[7.952169418334961, 6.130927562713623]
e2868c59-b422-4d11-a26a-bf3603d6fa16
open-retrieval-conversational-question
2005.11364
null
https://arxiv.org/abs/2005.11364v1
https://arxiv.org/pdf/2005.11364v1.pdf
Open-Retrieval Conversational Question Answering
Conversational search is one of the ultimate goals of information retrieval. Recent research approaches conversational search by simplified settings of response ranking and conversational question answering, where an answer is either selected from a given candidate set or extracted from a given passage. These simplifications neglect the fundamental role of retrieval in conversational search. To address this limitation, we introduce an open-retrieval conversational question answering (ORConvQA) setting, where we learn to retrieve evidence from a large collection before extracting answers, as a further step towards building functional conversational search systems. We create a dataset, OR-QuAC, to facilitate research on ORConvQA. We build an end-to-end system for ORConvQA, featuring a retriever, a reranker, and a reader that are all based on Transformers. Our extensive experiments on OR-QuAC demonstrate that a learnable retriever is crucial for ORConvQA. We further show that our system can make a substantial improvement when we enable history modeling in all system components. Moreover, we show that the reranker component contributes to the model performance by providing a regularization effect. Finally, further in-depth analyses are performed to provide new insights into ORConvQA.
['W. Bruce Croft', 'Cen Chen', 'Liu Yang', 'Mohit Iyyer', 'Minghui Qiu', 'Chen Qu']
2020-05-22
null
null
null
null
['conversational-search']
['natural-language-processing']
[-4.79586460e-02 1.47599965e-01 -2.43412703e-01 -2.10954174e-01 -1.68670154e+00 -9.85312462e-01 9.32728410e-01 2.25180492e-01 -4.12316799e-01 6.25869215e-01 8.63585532e-01 -5.48571825e-01 -1.99977770e-01 -6.01444304e-01 -4.33133692e-01 -3.54975194e-01 1.42028660e-01 8.72386813e-01 4.65077043e-01 -6.86089754e-01 3.83112133e-01 -1.37791842e-01 -1.30311847e+00 6.49540484e-01 8.63308907e-01 7.34348595e-01 2.08964199e-01 1.07046437e+00 -1.41822994e-01 1.14425230e+00 -5.62709510e-01 -5.56471169e-01 -4.09017392e-02 -6.22666597e-01 -1.68065012e+00 -3.00232291e-01 3.57110232e-01 -4.59880620e-01 -4.13755685e-01 4.90182698e-01 5.46828747e-01 3.51052374e-01 6.89861298e-01 -1.04275239e+00 -6.06700063e-01 8.74510705e-01 1.67230323e-01 3.12711686e-01 1.03001606e+00 1.37189358e-01 1.65347087e+00 -8.28403771e-01 5.96981108e-01 1.40394390e+00 1.21720515e-01 6.71114743e-01 -9.79442537e-01 -1.72552601e-01 2.14000955e-01 4.73484397e-01 -1.12147796e+00 -5.84642231e-01 6.97357237e-01 -1.48728639e-01 1.09659767e+00 8.09265614e-01 3.98737788e-01 1.00924766e+00 -1.70389012e-01 1.22748125e+00 6.82919502e-01 -5.13632476e-01 1.55884907e-01 7.97941014e-02 8.19629431e-01 5.68701208e-01 -3.93654734e-01 -1.80224255e-01 -7.10948944e-01 -8.16497087e-01 9.35763344e-02 -1.48190528e-01 -6.50412142e-01 -6.28826618e-02 -6.86035573e-01 9.52219784e-01 2.60539114e-01 1.71473622e-01 -2.65250295e-01 1.13910466e-01 3.40586543e-01 9.74607348e-01 2.68614560e-01 8.23995948e-01 -4.68180120e-01 -3.84353846e-01 -4.42444026e-01 3.94584805e-01 1.53140056e+00 8.62755001e-01 6.37784243e-01 -7.98153102e-01 -6.15933359e-01 1.14882123e+00 1.19140550e-01 5.24002135e-01 4.37052727e-01 -1.33138251e+00 5.21860182e-01 6.65627837e-01 4.27011013e-01 -7.21519530e-01 -6.80210739e-02 -1.75336391e-01 -2.61880189e-01 -6.99019432e-01 2.21154556e-01 1.48215070e-01 -3.28657717e-01 1.75975478e+00 1.89521924e-01 -2.21146926e-01 2.99004585e-01 9.14512932e-01 9.83426690e-01 7.28538215e-01 -1.68239251e-01 -3.79336447e-01 1.66557372e+00 -1.38229382e+00 -5.27435303e-01 -3.00285488e-01 9.14401591e-01 -6.52590632e-01 1.38624299e+00 2.28592515e-01 -1.21614254e+00 1.12775341e-01 -6.80821359e-01 -4.84636158e-01 -7.54710287e-02 -3.06358010e-01 4.68928069e-01 2.60450393e-01 -1.25418627e+00 -3.25090401e-02 -3.93895298e-01 -4.39345360e-01 -3.18483114e-01 -2.74436027e-02 4.02549841e-02 -2.06033915e-01 -1.50801969e+00 9.93831694e-01 -2.93082833e-01 -9.44724977e-02 -9.66690779e-01 -3.16460222e-01 -5.59785664e-01 3.07452708e-01 7.24095643e-01 -9.98842001e-01 2.06170774e+00 -4.81971532e-01 -1.42332888e+00 7.28368223e-01 -5.54070771e-01 -4.09921169e-01 2.13109076e-01 -2.33243227e-01 -8.40746909e-02 5.50544679e-01 -4.01965678e-02 1.23298690e-01 4.19358850e-01 -1.07738924e+00 -6.14658535e-01 -1.09203041e-01 6.55304670e-01 5.06813705e-01 -1.86506361e-01 6.81485981e-02 -8.49798739e-01 -1.32670745e-01 -9.72584933e-02 -9.79876697e-01 1.08805545e-01 -5.98760843e-01 -3.99819523e-01 -1.12429607e+00 3.93698514e-01 -6.33248925e-01 1.60259509e+00 -1.60661793e+00 2.23477691e-01 1.64717093e-01 3.76147300e-01 1.39704704e-01 -6.61381185e-01 1.24825191e+00 5.28730094e-01 3.05262685e-01 -2.40962319e-02 -3.18010986e-01 1.27693027e-01 3.98530103e-02 -6.31010532e-01 -5.79318441e-02 -1.47425652e-01 1.25854182e+00 -1.17256939e+00 -5.54449797e-01 -3.33571970e-01 -1.46715334e-02 -6.63514733e-01 4.60137248e-01 -6.82237029e-01 4.78981249e-02 -8.81915748e-01 5.01645803e-01 -8.81289765e-02 -6.21054292e-01 2.16727167e-01 3.11621904e-01 3.82854015e-01 1.10111797e+00 -5.14947832e-01 1.58306038e+00 -7.50118971e-01 5.88671863e-01 4.09327716e-01 -6.77360058e-01 4.71785158e-01 5.09567320e-01 2.31282055e-01 -1.01394331e+00 -5.66678494e-02 1.03218772e-01 -1.74259782e-01 -6.40676200e-01 8.65839541e-01 3.05335015e-01 -2.02689976e-01 1.13161349e+00 -1.78010300e-01 -3.09278995e-01 3.09760302e-01 1.09626269e+00 1.54201126e+00 -6.09888196e-01 -1.22124767e-02 -1.65296525e-01 7.22127676e-01 1.48327157e-01 6.65336549e-02 1.34170163e+00 -7.32811764e-02 2.96501189e-01 5.80154479e-01 3.20815928e-02 -6.42048180e-01 -6.87740386e-01 2.03291729e-01 1.64131248e+00 2.05520630e-01 -7.35117674e-01 -5.42470694e-01 -7.39105582e-01 -7.21593127e-02 7.45908558e-01 -3.29755217e-01 -3.21892053e-01 -7.47693837e-01 -1.90425888e-01 4.78857994e-01 2.53817201e-01 1.52541906e-01 -9.68982339e-01 -1.43267274e-01 1.54344603e-01 -1.02815533e+00 -7.94313788e-01 -9.96165574e-01 -2.18701795e-01 -5.72223544e-01 -1.40590286e+00 -5.41184008e-01 -8.10711682e-01 7.35630020e-02 8.24619234e-01 1.71215987e+00 6.35069966e-01 2.60480672e-01 1.13651323e+00 -6.00737274e-01 4.77003772e-03 -5.47887385e-01 6.10040605e-01 -2.88739771e-01 -2.70070225e-01 5.25931239e-01 -6.40895784e-01 -9.43783104e-01 5.32808661e-01 -7.35803425e-01 -2.34320357e-01 3.22751790e-01 8.77047360e-01 3.96459512e-02 -6.99790776e-01 8.92241120e-01 -7.98914969e-01 1.52347219e+00 -6.20916665e-01 -3.25145930e-01 6.46866024e-01 -7.31115162e-01 3.61750066e-01 2.81855822e-01 -3.24248105e-01 -9.78518307e-01 -5.91667235e-01 -2.29479507e-01 9.89548564e-02 5.21924734e-01 8.18255424e-01 1.52493834e-01 3.06241482e-01 8.76300812e-01 4.75240707e-01 -4.16107997e-02 -4.76453632e-01 7.11545050e-01 1.01158237e+00 3.60966384e-01 -8.02707732e-01 4.81295675e-01 1.56073496e-01 -6.93607748e-01 -7.64816582e-01 -9.65469539e-01 -1.10337150e+00 1.10973433e-01 -1.92823946e-01 3.87078375e-01 -9.31370616e-01 -1.33478284e+00 -1.01063237e-01 -1.32872784e+00 -4.81628478e-01 -6.34801760e-02 1.19327560e-01 -3.87870759e-01 5.12823820e-01 -1.04512203e+00 -1.16779697e+00 -8.18285406e-01 -9.12929773e-01 1.04419112e+00 6.38820678e-02 -4.47953969e-01 -9.04373169e-01 4.02795970e-01 9.13698733e-01 4.60141957e-01 -7.82439947e-01 1.02687669e+00 -1.13539207e+00 -9.37726974e-01 -2.79549509e-01 -5.31551689e-02 1.16920069e-01 -2.46573195e-01 -4.99474913e-01 -8.45994651e-01 -3.26372385e-01 -5.47274575e-02 -8.40577841e-01 1.06630683e+00 -1.79499716e-01 7.42298841e-01 -6.40586138e-01 -3.15812588e-01 -2.64239430e-01 8.49148333e-01 1.07238300e-01 4.57601905e-01 9.92062986e-02 1.76421672e-01 7.14782357e-01 3.80514801e-01 2.03956515e-01 9.11920726e-01 6.43762529e-01 1.13533422e-01 6.24822736e-01 -1.87426820e-01 -3.76555204e-01 3.93356740e-01 1.43540239e+00 2.74971515e-01 -6.56825602e-01 -9.01268065e-01 7.67063260e-01 -2.04912400e+00 -9.28239644e-01 1.04237944e-01 1.97039044e+00 1.22168434e+00 -1.72099575e-01 8.81287307e-02 -2.88548946e-01 2.64318049e-01 2.67424494e-01 -5.48655093e-01 -3.21386456e-01 -1.58228651e-02 6.44681156e-02 -2.41927892e-01 1.23588574e+00 -5.01995623e-01 9.45452034e-01 6.54175758e+00 7.29709268e-01 -6.33567572e-01 8.47585946e-02 2.36389250e-01 -9.08875689e-02 -7.67214179e-01 2.26733461e-01 -6.71816945e-01 1.97638437e-01 9.54011083e-01 -5.17853379e-01 7.38724530e-01 8.07682931e-01 -3.63115445e-02 -1.95726499e-01 -1.39084578e+00 7.94799209e-01 2.98271298e-01 -1.29129815e+00 1.87611505e-01 -1.78157538e-01 3.16848636e-01 1.50802195e-01 -3.64012957e-01 8.22922111e-01 8.85526836e-01 -7.57030785e-01 1.39318839e-01 6.32217586e-01 2.42504016e-01 -4.19200182e-01 6.07971609e-01 7.94995427e-01 -9.36625242e-01 -1.34190679e-01 -1.64807633e-01 -1.20164864e-01 2.01139346e-01 1.79554924e-01 -9.99648869e-01 5.09837866e-01 3.44687074e-01 1.93636388e-01 -3.93078297e-01 7.95059443e-01 -3.61149549e-01 9.54282522e-01 -3.94051820e-01 -6.09105527e-01 2.06296712e-01 -1.12669989e-01 6.69537902e-01 1.06439257e+00 -1.76421240e-01 5.87511420e-01 3.21372598e-01 4.23749208e-01 -5.00848889e-01 1.64552644e-01 -3.36792409e-01 -6.73383102e-02 9.00211334e-01 1.06588972e+00 9.29683372e-02 -5.62489033e-01 -2.51528442e-01 8.62331092e-01 6.54822052e-01 5.81944406e-01 -2.01209292e-01 -3.21244687e-01 5.07584989e-01 -1.79777578e-01 -5.12088127e-02 -4.88689467e-02 3.73558313e-01 -1.41653049e+00 3.07322353e-01 -1.28827190e+00 8.74625504e-01 -8.85759056e-01 -1.46027792e+00 6.13260925e-01 -1.39842751e-02 -7.24885881e-01 -9.72901702e-01 -1.28625873e-02 -5.85165381e-01 9.02379870e-01 -1.55256641e+00 -7.87178159e-01 -1.63035914e-01 5.56330621e-01 8.34111273e-01 9.66435373e-02 8.57237577e-01 1.70870423e-01 -2.10451052e-01 4.89514500e-01 2.15047002e-02 1.01448990e-01 7.65222788e-01 -1.08956981e+00 9.93986204e-02 5.30393779e-01 2.52658665e-01 1.01667249e+00 5.70554376e-01 -3.68534386e-01 -1.89221644e+00 -4.98726994e-01 1.37350249e+00 -1.02884436e+00 7.02167749e-01 -3.68261576e-01 -9.97646868e-01 7.27988720e-01 5.37099540e-01 -6.60750866e-01 6.45475507e-01 6.76251233e-01 -5.38882911e-01 3.00453436e-02 -4.78554159e-01 9.14866209e-01 1.03796434e+00 -1.22998452e+00 -1.11265898e+00 6.24246061e-01 1.30902946e+00 -1.41245067e-01 -4.77465987e-01 1.20337240e-01 4.66503322e-01 -6.06214285e-01 9.46311772e-01 -8.69350553e-01 3.41937631e-01 6.55074716e-02 -8.63507688e-02 -1.22686934e+00 -9.31508690e-02 -1.06055772e+00 -5.01349032e-01 1.08354485e+00 6.81921780e-01 -6.33413792e-01 6.35165393e-01 9.29126620e-01 -8.02363083e-02 -6.34588540e-01 -8.80667686e-01 -5.33389568e-01 2.19609246e-01 -5.17552853e-01 4.72777694e-01 7.01313138e-01 6.20427489e-01 1.12594318e+00 -2.93012351e-01 -9.56883803e-02 8.27260390e-02 4.49125737e-01 9.53013361e-01 -1.07968080e+00 -3.57953399e-01 -4.44469899e-01 5.58602393e-01 -2.03817892e+00 1.28017828e-01 -8.99534583e-01 2.76227295e-01 -1.72306204e+00 4.58094388e-01 -2.77781963e-01 -1.45864263e-01 2.11023927e-01 -3.75639796e-01 -2.81898767e-01 -2.13803649e-02 6.10567808e-01 -1.15453327e+00 7.35814393e-01 1.38379967e+00 -3.33381325e-01 -3.22470456e-01 1.34462297e-01 -9.98254359e-01 3.02072972e-01 4.26647037e-01 -3.26671034e-01 -7.53662467e-01 -4.07299966e-01 6.19483411e-01 8.13499629e-01 1.34324387e-01 -8.61182138e-02 8.18857193e-01 -6.98630586e-02 -7.55314171e-01 -5.80648601e-01 5.48691452e-01 -4.08465117e-01 -5.93109667e-01 1.41774803e-01 -1.00816417e+00 2.18909189e-01 -3.55925053e-01 6.65558815e-01 -3.61764431e-01 -1.99097216e-01 -2.19405070e-02 -1.97968885e-01 -3.29904318e-01 1.30139381e-01 -6.49791658e-01 4.78221506e-01 1.92617238e-01 4.85826224e-01 -7.81736493e-01 -1.28366840e+00 -4.21369970e-01 1.01096046e+00 5.61560690e-02 4.50914055e-01 4.81780648e-01 -1.06364119e+00 -9.04370725e-01 -4.86977965e-01 4.49065745e-01 -2.36048669e-01 1.00026838e-01 7.08303034e-01 1.09008176e-03 8.84751499e-01 7.35032976e-01 -3.69346589e-01 -1.38604128e+00 4.55853075e-01 2.30657890e-01 -6.62408531e-01 -3.13122392e-01 7.97184408e-01 8.15456063e-02 -6.84121132e-01 4.76333171e-01 -2.00512305e-01 -3.16755325e-01 4.12490368e-02 7.57262349e-01 1.25622898e-01 1.02910116e-01 -8.70748460e-02 -2.16002703e-01 -6.92367107e-02 -3.22072774e-01 -6.25811517e-01 8.44343722e-01 -5.32457054e-01 -3.65365982e-01 3.59929800e-01 1.22595382e+00 2.46111929e-01 -4.73400980e-01 -8.25936317e-01 1.88657269e-01 -6.28277063e-02 -1.57084838e-01 -1.05954790e+00 -4.22522008e-01 5.33258557e-01 -1.63843602e-01 6.17010415e-01 8.94028783e-01 5.15069962e-01 9.93467748e-01 1.28221595e+00 3.38579088e-01 -8.62270176e-01 4.26386267e-01 9.35950637e-01 1.28329861e+00 -1.18450141e+00 -2.85565287e-01 -2.65057147e-01 -4.90535885e-01 7.94009387e-01 4.15542960e-01 2.62005001e-01 4.65762317e-01 -3.11179966e-01 2.24683285e-01 -4.85473692e-01 -1.56157410e+00 -3.53492439e-01 4.17256027e-01 4.52637896e-02 2.69180030e-01 -2.07483023e-01 -5.68666577e-01 6.90409541e-01 -3.71084213e-01 -2.50071019e-01 3.55718553e-01 9.90360677e-01 -4.89282548e-01 -1.22916341e+00 1.75298937e-02 3.69942695e-01 -2.51590371e-01 -4.83364493e-01 -8.86371136e-01 4.12555605e-01 -1.06228113e+00 1.66228914e+00 -1.32401943e-01 -4.49841291e-01 2.80051231e-01 4.97621268e-01 9.56401899e-02 -4.59608853e-01 -9.11488175e-01 -2.43093327e-01 7.72740006e-01 -6.30664051e-01 -1.74636498e-01 -3.17611575e-01 -8.01598370e-01 -3.29708934e-01 -6.34184003e-01 1.16292894e+00 2.78892338e-01 9.39869404e-01 7.38343418e-01 1.54485134e-02 8.01522732e-01 4.75621447e-02 -9.31844234e-01 -1.17302966e+00 -1.48293957e-01 4.43795174e-01 4.85295445e-01 -7.78052136e-02 -5.61856449e-01 -2.27694482e-01]
[12.060586929321289, 7.835334300994873]
a28e5852-74bd-4a38-9549-856de2d7ab2b
structured-training-for-neural-network
1506.06158
null
http://arxiv.org/abs/1506.06158v1
http://arxiv.org/pdf/1506.06158v1.pdf
Structured Training for Neural Network Transition-Based Parsing
We present structured perceptron training for neural network transition-based dependency parsing. We learn the neural network representation using a gold corpus augmented by a large number of automatically parsed sentences. Given this fixed network representation, we learn a final layer using the structured perceptron with beam-search decoding. On the Penn Treebank, our parser reaches 94.26% unlabeled and 92.41% labeled attachment accuracy, which to our knowledge is the best accuracy on Stanford Dependencies to date. We also provide in-depth ablative analysis to determine which aspects of our model provide the largest gains in accuracy.
['Slav Petrov', 'Chris Alberti', 'Michael Collins', 'David Weiss']
2015-06-19
structured-training-for-neural-network-1
https://aclanthology.org/P15-1032
https://aclanthology.org/P15-1032.pdf
ijcnlp-2015-7
['transition-based-dependency-parsing']
['natural-language-processing']
[ 1.66201994e-01 1.06386983e+00 -5.30434966e-01 -9.51034904e-01 -1.03393507e+00 -5.29802203e-01 3.28931101e-02 1.75538868e-01 -7.51332700e-01 1.04108334e+00 4.12322849e-01 -9.10409033e-01 3.87103707e-01 -6.72391236e-01 -6.96208596e-01 -3.06522667e-01 -2.74110317e-01 7.64172614e-01 7.55190700e-02 -2.56400496e-01 -2.89199501e-01 3.52270663e-01 -4.89485890e-01 4.98735338e-01 4.77691382e-01 5.11986911e-01 1.39174670e-01 7.85994768e-01 -1.60158321e-01 1.25664675e+00 -5.17202258e-01 -6.84100211e-01 -9.92281660e-02 -3.93497288e-01 -1.38077748e+00 -5.20741999e-01 2.47001603e-01 -4.69001144e-01 -3.15656841e-01 9.51410413e-01 4.39844914e-02 9.95670855e-02 1.68832615e-01 -4.42954540e-01 -8.89858902e-01 1.69462717e+00 -1.51129887e-01 8.18204284e-01 8.00820664e-02 -3.00078869e-01 1.55408609e+00 -6.68388247e-01 5.72815955e-01 1.29163671e+00 5.54228663e-01 8.74461710e-01 -1.25585949e+00 -6.53451562e-01 3.58639926e-01 -1.32021368e-01 -5.54872274e-01 -6.79532886e-01 6.78380072e-01 -1.42377764e-01 1.77077413e+00 -2.86584973e-01 3.84628356e-01 1.01462662e+00 3.36547673e-01 6.48831964e-01 6.98483884e-01 -9.89776909e-01 2.13209376e-01 -2.46992365e-01 1.03777277e+00 9.56597626e-01 2.44397968e-01 1.19255461e-01 -2.99026132e-01 -1.95510276e-02 7.09974885e-01 -3.85545462e-01 2.03731120e-01 3.36041719e-01 -5.08818388e-01 1.14237964e+00 7.31943905e-01 4.66647089e-01 -3.02359074e-01 2.87180364e-01 2.54347503e-01 3.75737011e-01 4.30106580e-01 5.63204646e-01 -1.12138557e+00 -1.36687115e-01 -4.11806256e-01 -3.52962703e-01 1.05490088e+00 7.79686511e-01 7.68376291e-01 3.81091386e-01 4.17282373e-01 9.71213520e-01 4.05790329e-01 2.18316108e-01 3.79858792e-01 -1.07338142e+00 9.88014877e-01 2.84001708e-01 -3.08640540e-01 -1.89657375e-01 -7.19088972e-01 -3.17319840e-01 -4.69274133e-01 1.12150602e-01 5.08844733e-01 -9.81728852e-01 -1.03046560e+00 1.86484802e+00 -1.71082154e-01 -1.96491882e-01 6.94825768e-01 5.09911597e-01 8.81771326e-01 6.89041018e-01 5.96782267e-01 -1.64649025e-01 1.29289961e+00 -9.97689307e-01 -7.37656355e-01 -9.35324073e-01 1.17543602e+00 -2.27554858e-01 5.68521261e-01 4.87981364e-02 -1.29106331e+00 -2.67110854e-01 -1.15152562e+00 -1.27617061e-01 6.83186948e-02 -8.79046135e-03 1.00281382e+00 7.79133201e-01 -1.25545573e+00 7.33872533e-01 -1.51056969e+00 -2.61905372e-01 4.31046426e-01 6.18242443e-01 -5.22466302e-01 -2.61551086e-02 -1.44423473e+00 1.29941487e+00 8.18925738e-01 1.82595104e-01 -4.73902255e-01 -6.30141124e-02 -1.17243350e+00 2.09951565e-01 -2.35077739e-02 -2.69161761e-01 1.81049097e+00 -9.34993505e-01 -1.71213484e+00 8.34886491e-01 -2.48025641e-01 -8.22278857e-01 -6.49861872e-01 -2.64609277e-01 -4.14256543e-01 1.40213326e-01 -5.60135171e-02 8.20349038e-01 2.02829242e-02 -7.21400976e-01 -5.12007058e-01 -4.72933412e-01 3.08384061e-01 7.46248811e-02 -2.46314138e-01 5.95873535e-01 -4.32596691e-02 -3.21225792e-01 4.76078212e-01 -9.28741574e-01 -5.74421823e-01 -1.03899395e+00 -5.24882257e-01 -5.00993848e-01 1.16062298e-01 -1.02762532e+00 1.10059476e+00 -1.84489453e+00 -1.81931213e-01 -1.60205409e-01 -2.28569712e-02 2.08482876e-01 -3.54164183e-01 2.54583508e-01 -4.50779140e-01 2.96436757e-01 -1.62961885e-01 -5.26212275e-01 -3.61377180e-01 7.08209097e-01 -1.15190603e-01 1.14924334e-01 6.27535641e-01 1.04927135e+00 -8.13143969e-01 -5.81295729e-01 -2.77207673e-01 1.31189510e-01 -6.53282881e-01 2.39209428e-01 -2.36905292e-01 3.36137444e-01 -6.14620030e-01 7.08626807e-01 2.11872756e-01 -2.40593880e-01 7.68199801e-01 3.45031351e-01 1.34292513e-01 1.04947591e+00 -3.54633659e-01 1.94785893e+00 -4.61601257e-01 6.78476274e-01 2.11142704e-01 -1.01854825e+00 1.08383906e+00 5.78710854e-01 -2.19261155e-01 -4.19822574e-01 2.69337177e-01 7.05954209e-02 3.66970271e-01 -2.75773048e-01 3.38706374e-01 -5.98082364e-01 -4.03421015e-01 4.53619868e-01 5.37007868e-01 2.93177158e-01 2.29064524e-02 3.44159126e-01 1.63698244e+00 1.24419808e-01 3.50639135e-01 -1.79200590e-01 7.54220504e-03 3.25308859e-01 9.03702617e-01 7.23923385e-01 -2.14583322e-01 3.84755909e-01 6.58888161e-01 -6.34661615e-01 -7.68491983e-01 -9.88703012e-01 -1.64090618e-01 1.46850157e+00 -5.98397315e-01 -2.75516719e-01 -8.45145285e-01 -9.39738572e-01 -5.85753620e-01 9.83002067e-01 -5.29038370e-01 2.56914169e-01 -1.16350734e+00 -8.37128937e-01 8.74277115e-01 1.09910882e+00 1.07362397e-01 -1.77075553e+00 -2.57106245e-01 5.44966519e-01 4.99765053e-02 -1.38335192e+00 2.40484282e-01 1.22335589e+00 -1.44507444e+00 -7.69738257e-01 4.25116047e-02 -1.48080826e+00 6.84315741e-01 -6.46714926e-01 1.45626175e+00 7.14853033e-02 3.91670167e-01 -5.25041163e-01 -4.85657960e-01 -1.26382962e-01 -9.23807383e-01 5.43979287e-01 -1.26487494e-01 -9.53628540e-01 7.07463622e-01 -6.90108001e-01 1.23829521e-01 -5.27222514e-01 -2.29453415e-01 -1.38214350e-01 8.48417163e-01 9.87442315e-01 1.61955193e-01 -1.89032629e-01 6.00754797e-01 -1.47755218e+00 5.15851378e-01 -5.33861279e-01 -4.63003248e-01 6.90786839e-02 -3.47940475e-01 4.63683814e-01 7.44267285e-01 1.14277512e-01 -1.54265177e+00 4.40979362e-01 -8.17842245e-01 2.10439295e-01 -5.44019818e-01 7.80179679e-01 -7.85486549e-02 4.27699655e-01 8.67029727e-01 -4.90566730e-01 -5.60917675e-01 -8.31261694e-01 5.13535857e-01 5.54006398e-01 8.34734380e-01 -6.58932984e-01 2.48696610e-01 -2.05988035e-01 -2.93964505e-01 -2.40194887e-01 -1.31904912e+00 -1.12282678e-01 -1.08728468e+00 6.26652002e-01 1.26614916e+00 -8.89395416e-01 -3.65890384e-01 1.25271589e-01 -1.54719269e+00 -7.22572088e-01 -1.56047568e-01 6.29245698e-01 -2.69507974e-01 1.21232316e-01 -1.58117020e+00 -7.06052780e-01 -5.65771759e-01 -8.28379512e-01 4.21640337e-01 1.10686779e-01 -3.87324691e-01 -1.22155380e+00 5.77960253e-01 2.26039037e-01 -1.22397169e-01 -1.72307834e-01 1.03255856e+00 -1.25674510e+00 -1.29098650e-02 -1.31139696e-01 -7.32715875e-02 3.81852329e-01 2.45793108e-02 -2.85000056e-01 -1.10661924e+00 -1.15464829e-01 -1.86541364e-01 -6.71165168e-01 8.17558169e-01 5.11528909e-01 4.78471667e-01 -2.12112293e-01 -3.78328204e-01 4.41285521e-01 1.28918159e+00 4.72736508e-01 3.43133360e-01 2.68970668e-01 5.30086637e-01 5.23010433e-01 2.34986290e-01 -7.21321926e-02 3.23383689e-01 3.27257626e-02 4.02682036e-01 1.38582885e-01 -1.15315933e-02 -3.45992416e-01 4.14324194e-01 1.06446600e+00 9.87640992e-02 -2.17852399e-01 -1.32338500e+00 5.20461261e-01 -1.65666330e+00 -6.51982546e-01 6.76381439e-02 1.37617218e+00 1.18420637e+00 8.18992317e-01 -3.41145486e-01 -1.55492589e-01 8.47558737e-01 8.71960372e-02 -3.55494976e-01 -1.09711599e+00 1.10212736e-01 9.60673928e-01 6.77271545e-01 9.73874271e-01 -1.15968764e+00 1.56447160e+00 7.61154842e+00 2.79011969e-02 -6.06288552e-01 3.21080804e-01 7.09588468e-01 3.74048203e-02 -4.26691696e-02 2.88593888e-01 -1.19504607e+00 -3.19567062e-02 1.89229012e+00 5.56341171e-01 1.71178862e-01 1.05669856e+00 -3.78103167e-01 2.20818549e-01 -1.19843459e+00 3.46854985e-01 -2.98759907e-01 -1.40538299e+00 -4.40895438e-01 -2.85945445e-01 4.12037402e-01 6.63916826e-01 -5.38888574e-01 6.72202229e-01 1.56175601e+00 -1.21059263e+00 2.55353510e-01 -1.64764151e-01 7.01904953e-01 -6.10853732e-01 8.02072883e-01 4.66144621e-01 -8.09127152e-01 -1.14791274e-01 -5.54079592e-01 -6.60769463e-01 3.96618515e-01 3.37574124e-01 -9.85520422e-01 4.23584320e-02 5.87927699e-01 5.86428285e-01 -2.79967934e-01 2.47811109e-01 -1.00030887e+00 1.34928560e+00 -3.81850153e-01 -1.24077164e-01 3.81557792e-01 1.15390874e-01 7.17509165e-02 1.47768736e+00 -3.95275116e-01 5.95780134e-01 1.29817724e-01 3.59498292e-01 -4.10894215e-01 -1.03807241e-01 -3.54062915e-01 -3.69561324e-03 8.00756216e-01 1.08881915e+00 -6.65099382e-01 -2.48993576e-01 -5.66545784e-01 6.54657125e-01 1.35929322e+00 3.39032322e-01 -2.13392794e-01 -3.50058615e-01 3.37083191e-01 -5.21984398e-01 3.96100253e-01 -3.47569138e-01 -5.57654798e-01 -1.06391275e+00 -2.79656529e-01 -3.44686717e-01 6.98446512e-01 -6.93733990e-01 -1.21886003e+00 1.07889044e+00 -1.31329879e-01 -2.09224522e-01 -8.05372477e-01 -9.42482352e-01 -7.88155973e-01 9.66481030e-01 -1.32924020e+00 -9.16742742e-01 5.57004213e-01 1.83144718e-01 5.90648413e-01 -2.44161025e-01 1.50748551e+00 -1.64747134e-01 -9.91411746e-01 7.05591381e-01 -1.62226230e-01 1.02037942e+00 3.19420934e-01 -1.42883992e+00 1.28623152e+00 1.08419859e+00 2.62012303e-01 7.07336605e-01 4.36756641e-01 -5.68763614e-01 -1.02176428e+00 -8.67964625e-01 1.31192851e+00 -6.47988737e-01 8.07159424e-01 -3.43225032e-01 -9.54267681e-01 1.66957688e+00 4.10116583e-01 5.41996397e-02 8.19343209e-01 8.02099764e-01 -3.46789479e-01 3.50146294e-01 -9.22471106e-01 1.73695520e-01 1.00036240e+00 -4.15541500e-01 -1.37622237e+00 8.64423588e-02 1.08831716e+00 -4.37593967e-01 -9.76061463e-01 2.01348081e-01 2.61410892e-01 -6.29629314e-01 5.43733060e-01 -1.20745289e+00 5.72151363e-01 5.61281800e-01 -2.13391393e-01 -1.38398087e+00 -7.18943059e-01 -4.54848677e-01 -1.48968309e-01 1.09378862e+00 1.11317229e+00 -6.31144881e-01 1.32657111e+00 1.03327882e+00 -5.22740006e-01 -7.12120414e-01 -1.03616619e+00 -4.67483521e-01 7.57151365e-01 -5.47236681e-01 2.85343051e-01 1.02129114e+00 6.90607071e-01 1.13005722e+00 2.70902347e-02 2.48784542e-01 4.07888293e-01 -6.92169368e-02 2.86333822e-02 -1.30753124e+00 -4.68278140e-01 2.49364581e-02 -5.03260605e-02 -1.20210373e+00 8.55325043e-01 -1.06640506e+00 3.79967391e-01 -1.63382494e+00 1.60732135e-01 -6.32385850e-01 -4.61989284e-01 1.14117825e+00 -2.13680342e-01 6.22789226e-02 -5.82797863e-02 1.57922044e-01 -4.49922353e-01 -9.25883651e-03 7.46353567e-01 2.11123973e-01 -2.48834476e-01 -2.05246195e-01 -9.49429870e-01 9.33718860e-01 1.32640731e+00 -1.02790022e+00 -1.06966943e-01 -9.74390745e-01 1.62670776e-01 6.23250723e-01 -4.02841687e-01 -6.76879585e-01 1.64318651e-01 -6.36637658e-02 3.58772278e-01 -4.34564531e-01 4.46771294e-01 -3.43865216e-01 -5.27574658e-01 5.55276752e-01 -7.44608521e-01 1.12951681e-01 3.01413924e-01 2.84632593e-01 -1.13361031e-01 -7.47794807e-01 5.91431677e-01 -3.94778371e-01 -6.91784799e-01 -7.45058060e-02 -7.02937365e-01 1.81019694e-01 3.19096923e-01 1.85706958e-01 -3.79682273e-01 -6.60055727e-02 -1.24604118e+00 6.92066848e-02 7.25360364e-02 1.49644792e-01 2.87383974e-01 -8.14014316e-01 -7.83665836e-01 1.65956259e-01 -3.49748135e-01 1.30116358e-01 -2.40434512e-01 8.46805330e-03 -5.10800242e-01 6.20112360e-01 -2.73567080e-01 -6.85236529e-02 -1.08034956e+00 3.01838011e-01 2.40839481e-01 -6.82108819e-01 -6.72559857e-01 1.14963520e+00 -1.80682525e-01 -6.03106916e-01 1.06460318e-01 -5.19095182e-01 -4.47832555e-01 -4.09288257e-01 4.15630609e-01 -2.04477653e-01 2.18966708e-01 -3.86920512e-01 -5.39830923e-01 -9.19652954e-02 -3.91770929e-01 -3.36559564e-01 1.43628561e+00 2.71138996e-01 -5.62421419e-02 3.25841904e-01 1.11395776e+00 -4.10176776e-02 -1.18793726e+00 -2.41062000e-01 3.95767152e-01 3.40298086e-01 1.10356472e-01 -8.96123528e-01 -9.51811850e-01 9.70921755e-01 3.92293893e-02 9.52673852e-02 7.34173417e-01 4.34043586e-01 8.93803298e-01 1.10005009e+00 1.61199614e-01 -7.91544080e-01 -3.11314583e-01 1.15942574e+00 1.92956313e-01 -1.21715486e+00 -1.67065814e-01 -3.13656658e-01 -3.85904193e-01 1.19093549e+00 7.74218678e-01 -5.41432202e-01 7.33958602e-01 8.92219007e-01 4.09771472e-01 -1.69330537e-01 -1.07280135e+00 -6.79033250e-02 -4.37324882e-01 5.85813642e-01 9.05461490e-01 1.85867891e-01 -4.53700498e-02 1.02829230e+00 -5.51769018e-01 -2.46510252e-01 5.45300424e-01 1.11393404e+00 -7.72661090e-01 -1.40466201e+00 -1.61097050e-01 3.22283387e-01 -9.58628237e-01 -6.44068539e-01 -2.84675598e-01 5.76767743e-01 -5.29179633e-01 1.25457335e+00 4.15534914e-01 -2.95125544e-01 1.56166270e-01 6.40076637e-01 5.35002947e-01 -1.30731118e+00 -8.07384729e-01 -3.59419495e-01 8.98113668e-01 -1.64913878e-01 -3.21319044e-01 -4.62587178e-01 -1.93295181e+00 -1.02665853e-02 -5.48597038e-01 4.60363328e-01 5.98881662e-01 1.12101769e+00 3.55854593e-02 3.90042633e-01 3.33270520e-01 -6.45376325e-01 -5.93448400e-01 -1.41859639e+00 -3.12925518e-01 -1.37696834e-02 2.29148895e-01 -6.26582727e-02 -2.41312727e-01 1.26262844e-01]
[10.330048561096191, 9.664861679077148]
346cddde-2b34-4b92-bb15-b1a0b5ae389f
optimizing-drug-design-by-merging-generative
2305.06334
null
https://arxiv.org/abs/2305.06334v1
https://arxiv.org/pdf/2305.06334v1.pdf
Optimizing Drug Design by Merging Generative AI With Active Learning Frameworks
Traditional drug discovery programs are being transformed by the advent of machine learning methods. Among these, Generative AI methods (GM) have gained attention due to their ability to design new molecules and enhance specific properties of existing ones. However, current GM methods have limitations, such as low affinity towards the target, unknown ADME/PK properties, or the lack of synthetic tractability. To improve the applicability domain of GM methods, we have developed a workflow based on a variational autoencoder coupled with active learning steps. The designed GM workflow iteratively learns from molecular metrics, including drug likeliness, synthesizability, similarity, and docking scores. In addition, we also included a hierarchical set of criteria based on advanced molecular modeling simulations during a final selection step. We tested our GM workflow on two model systems, CDK2 and KRAS. In both cases, our model generated chemically viable molecules with a high predicted affinity toward the targets. Particularly, the proportion of high-affinity molecules inferred by our GM workflow was significantly greater than that in the training data. Notably, we also uncovered novel scaffolds significantly dissimilar to those known for each target. These results highlight the potential of our GM workflow to explore novel chemical space for specific targets, thereby opening up new possibilities for drug discovery endeavors.
['Victor Guallar', 'Soumya Ray', 'Ajay S Yekkirala', 'Laura Malo', 'Julia Vilalta Mor', 'Yang Ming Zhu', 'Lucía Díaz', 'Marek Orzechowski', 'Alexis Molina', 'Isaac Filella-Merce']
2023-05-04
null
null
null
null
['drug-discovery']
['medical']
[ 3.56323868e-01 1.59416676e-01 -2.32732937e-01 -2.28261463e-02 -7.90747881e-01 -8.54770482e-01 5.28432965e-01 4.62094665e-01 -1.82675377e-01 1.46157634e+00 9.07551348e-02 -4.09796000e-01 -2.21946508e-01 -8.27232420e-01 -8.07017207e-01 -1.14182055e+00 5.10778874e-02 6.30268455e-01 -6.10977784e-02 -1.70508415e-01 3.28741133e-01 6.70913994e-01 -1.00188375e+00 -7.94723332e-02 1.63526440e+00 4.44111705e-01 3.03291976e-01 -4.22224663e-02 2.39778440e-02 4.71979797e-01 -3.58497858e-01 -3.77691597e-01 1.75843313e-02 -5.04828215e-01 -4.84378934e-01 -3.45650345e-01 -1.58320200e-02 -1.46816820e-01 -7.52885342e-02 7.67798066e-01 6.81336641e-01 2.81211436e-01 8.83277118e-01 -6.15389109e-01 -8.58389497e-01 4.47620690e-01 -1.12339847e-01 -1.64576635e-01 3.03039372e-01 4.92948949e-01 1.02200365e+00 -1.22559857e+00 7.21097231e-01 7.50735044e-01 3.73003095e-01 6.54476285e-01 -1.52999997e+00 -7.89116621e-01 1.75274722e-02 2.59754479e-01 -1.35020852e+00 -3.45824063e-01 5.96258879e-01 -6.26161337e-01 1.00633359e+00 1.68335944e-01 8.33631516e-01 1.21086812e+00 2.93615758e-01 5.06175637e-01 8.03306341e-01 -9.99806821e-02 9.36750889e-01 1.44014835e-01 -4.62758452e-01 6.23009086e-01 3.12986076e-01 1.76202446e-01 -5.43159842e-01 -5.60838759e-01 5.36450624e-01 1.27661109e-01 -3.44008684e-01 -4.74886566e-01 -9.58366036e-01 1.18201292e+00 4.33663458e-01 2.23066926e-01 -5.37700057e-01 -1.44065946e-01 -1.27295583e-01 -6.01517260e-01 2.17461288e-01 1.22658277e+00 -4.91133928e-01 1.86187059e-01 -7.53853261e-01 5.30992225e-02 7.24019766e-01 3.97949696e-01 7.24406064e-01 1.08389661e-01 9.00487974e-02 2.56986320e-01 4.55903381e-01 1.32514447e-01 7.57883638e-02 -5.66171825e-01 -9.88095179e-02 6.42058432e-01 -9.68028605e-03 -7.12891340e-01 -3.50912690e-01 -6.73949599e-01 -5.39649725e-01 1.47721004e-02 9.11557078e-02 -1.41254872e-01 -8.23799610e-01 1.81508088e+00 5.12964010e-01 7.44797066e-02 1.74733892e-01 7.37306833e-01 8.21248949e-01 7.69576669e-01 5.71622729e-01 -6.15153551e-01 6.79121137e-01 -5.86833239e-01 -5.35877585e-01 3.43088746e-01 5.51629782e-01 -4.52757418e-01 6.91475987e-01 4.40036446e-01 -9.40145373e-01 -1.45627350e-01 -1.10622537e+00 3.82926702e-01 -3.85618925e-01 -2.05983698e-01 1.06464362e+00 6.43729448e-01 -7.11111665e-01 1.02845430e+00 -7.72812724e-01 3.35371345e-02 7.73417950e-01 6.98361278e-01 -1.78789973e-01 2.03638181e-01 -1.24556017e+00 9.30399537e-01 6.43316984e-01 8.22490454e-02 -1.33157444e+00 -1.06037903e+00 -5.74729681e-01 2.01970056e-01 4.30750608e-01 -9.82633173e-01 7.46110737e-01 -7.09077239e-01 -1.93465722e+00 1.13444082e-01 1.08497456e-01 -2.68793285e-01 1.73708618e-01 1.34057194e-01 -2.81272560e-01 7.44518638e-02 -1.98199600e-01 6.02030277e-01 2.91645646e-01 -1.03336692e+00 -1.82820141e-01 -3.47856939e-01 1.32505214e-02 1.86327100e-01 -2.87280738e-01 -3.69383931e-01 2.34665185e-01 -3.41126949e-01 -1.86454266e-01 -9.06348646e-01 -6.41444206e-01 -9.38102081e-02 -5.19569397e-01 -9.04608965e-02 3.48945767e-01 -3.48966509e-01 1.02772832e+00 -1.86325622e+00 6.56162500e-01 4.20309901e-01 4.74629253e-01 4.59848940e-01 9.83832311e-03 9.08614397e-01 -1.09473810e-01 4.43334758e-01 -2.58945942e-01 4.13019150e-01 -4.13887531e-01 -2.53401905e-01 -1.75896272e-01 3.41599971e-01 3.49917054e-01 9.50733066e-01 -1.08017921e+00 -2.10393205e-01 -1.69875706e-03 6.35576606e-01 -7.75500655e-01 6.30190670e-02 -7.94021904e-01 9.62623537e-01 -7.49216080e-01 7.29477346e-01 4.23292994e-01 -4.99576956e-01 4.93356824e-01 -1.45728262e-02 -2.62058973e-01 1.35178447e-01 -5.47959924e-01 1.47042608e+00 -9.24478322e-02 3.31115499e-02 -7.51745462e-01 -6.83673859e-01 8.63453865e-01 3.14865500e-01 6.74568653e-01 -5.50677896e-01 2.88359076e-02 4.06790137e-01 3.13669086e-01 -2.43294537e-01 -3.16190794e-02 -3.07498872e-01 4.82862055e-01 3.50619517e-02 3.75392437e-02 3.29576991e-02 1.85215399e-01 1.03590831e-01 8.17627788e-01 3.83475572e-01 3.82976055e-01 -1.49249345e-01 4.72727895e-01 2.43879527e-01 7.26517558e-01 3.53663892e-01 7.85040781e-02 3.66926938e-01 3.87780875e-01 -2.98990130e-01 -1.10554600e+00 -9.86755252e-01 -3.17887217e-01 5.74529886e-01 3.78071517e-02 -5.12002110e-01 -5.04757941e-01 -6.18606150e-01 -2.62269646e-01 8.21428299e-01 -6.26618564e-01 -5.78524232e-01 -2.60258555e-01 -1.00442672e+00 4.78944778e-01 1.84337392e-01 1.46006979e-02 -7.17861295e-01 -1.34956958e-02 5.93570828e-01 2.59879321e-01 -6.62723482e-01 -2.40211785e-01 2.33332396e-01 -7.71596432e-01 -1.17295027e+00 -8.18645656e-01 -6.64938331e-01 6.77940726e-01 -1.52270928e-01 5.90126216e-01 -2.48104706e-01 -2.88116634e-01 -1.63934484e-01 -1.51224270e-01 -6.02778971e-01 -6.14752233e-01 -7.75033534e-02 1.80408001e-01 -2.05008268e-01 -7.38967815e-03 -7.52253532e-01 -9.54570889e-01 2.44439036e-01 -6.75016165e-01 9.14397985e-02 7.76163101e-01 9.45966840e-01 9.62285519e-01 -4.71541286e-02 9.43179727e-01 -7.48103499e-01 5.71423173e-01 -7.67821670e-01 -5.89010835e-01 3.14873338e-01 -7.76270866e-01 2.09795207e-01 6.51955128e-01 -7.81990528e-01 -1.08163524e+00 3.91557157e-01 -3.80865306e-01 5.48969060e-02 6.58454597e-02 7.82736599e-01 -5.31728148e-01 -1.77949235e-01 7.35975027e-01 3.85744780e-01 1.92159668e-01 -2.56911010e-01 1.71262860e-01 2.17195243e-01 -1.53161079e-01 -5.49655318e-01 5.88155389e-01 1.16844594e-01 2.14735895e-01 -8.07410240e-01 -3.00239414e-01 4.53983154e-03 -2.20310658e-01 7.65615106e-02 7.38274813e-01 -8.37762713e-01 -1.10859346e+00 4.16044444e-02 -8.83274496e-01 -3.30284089e-01 -8.70446414e-02 7.82410979e-01 -4.06916857e-01 4.87225443e-01 -2.72342265e-01 -5.77778220e-01 -5.73503375e-01 -1.65733671e+00 6.16565764e-01 3.60742450e-01 -4.53535438e-01 -9.63303447e-01 4.92086142e-01 4.55967367e-01 3.76125693e-01 5.52360713e-01 1.44051790e+00 -8.53229761e-01 -9.39337611e-01 9.74532068e-02 2.24306792e-01 -1.03272900e-01 2.28614748e-01 3.66786450e-01 -7.00563550e-01 -1.82460040e-01 -5.15432298e-01 -2.41884321e-01 6.90556526e-01 4.76449013e-01 1.01421344e+00 -3.86844158e-01 -5.87672353e-01 5.43119729e-01 1.32353997e+00 9.86694813e-01 7.00855613e-01 2.53158033e-01 6.08053088e-01 1.35129586e-01 3.63001645e-01 5.49428165e-01 -1.39388219e-01 6.78461790e-01 6.28349304e-01 -3.41677427e-01 3.34130287e-01 -4.16748941e-01 2.49822855e-01 1.39288962e-01 -3.17796975e-01 -4.34370309e-01 -8.72593701e-01 1.32482439e-01 -1.55773604e+00 -8.59405041e-01 -5.23062050e-02 2.28141069e+00 1.17201018e+00 6.22509420e-02 2.45341480e-01 -3.33508253e-01 5.04237771e-01 -4.56008255e-01 -1.02617645e+00 -2.43028492e-01 -3.16468596e-01 5.29830337e-01 1.13195017e-01 2.98188686e-01 -5.55125654e-01 1.00346339e+00 6.37746811e+00 8.96561623e-01 -1.18440759e+00 -4.43655938e-01 7.52744734e-01 -1.90038569e-02 -7.53114164e-01 4.63894904e-01 -8.17613065e-01 4.11705136e-01 8.09396625e-01 -4.09671038e-01 1.68209642e-01 7.12293684e-01 3.63934040e-01 1.32491276e-01 -1.14121211e+00 4.63351816e-01 -3.19888830e-01 -2.02956939e+00 3.57000977e-01 4.49011654e-01 7.79096782e-01 -3.16602677e-01 1.29836038e-01 7.01164454e-02 1.88347057e-01 -1.29186893e+00 3.37636083e-01 4.77843761e-01 6.98232532e-01 -9.78117168e-01 3.55040073e-01 2.68355250e-01 -4.24524426e-01 4.38522398e-02 -1.40949845e-01 2.40711913e-01 3.09431814e-02 5.83601177e-01 -1.23492646e+00 4.53361720e-01 -2.10379213e-02 5.05200326e-01 -3.72887135e-01 1.20267010e+00 -1.24676332e-01 3.72487307e-01 -1.97740525e-01 -5.24653673e-01 1.50844276e-01 -5.02524734e-01 4.35509741e-01 6.18172169e-01 1.27218723e-01 2.60766506e-01 5.92988878e-02 1.28485847e+00 -5.56436740e-02 3.35414588e-01 -4.23270524e-01 -5.94574571e-01 5.39430976e-01 1.00142694e+00 -5.70143104e-01 -1.06935307e-01 4.77531180e-03 6.31976247e-01 1.19882949e-01 4.95543391e-01 -1.02031827e+00 -1.33637667e-01 6.35691524e-01 3.89892533e-02 1.19622678e-01 -9.08535421e-02 -5.48832268e-02 -9.53931570e-01 -4.66086775e-01 -9.69232261e-01 1.57404408e-01 -4.00829226e-01 -9.57266569e-01 5.67635298e-01 -1.92207456e-01 -9.28242445e-01 1.74657643e-01 -2.54134893e-01 -4.92430896e-01 7.88272023e-01 -9.06521082e-01 -8.28329444e-01 8.57860297e-02 1.98985517e-01 4.77778196e-01 -3.65935117e-01 1.04035985e+00 2.41880700e-01 -9.05306160e-01 3.67389411e-01 4.76410478e-01 -6.75154507e-01 5.42387605e-01 -9.30753708e-01 2.94447243e-02 4.32076216e-01 -4.29681409e-03 9.32278633e-01 8.96689892e-01 -8.47621202e-01 -1.33248687e+00 -7.82031775e-01 3.77811551e-01 -3.00376594e-01 5.24391294e-01 -2.26692125e-01 -7.49318898e-01 2.97146618e-01 -1.24568567e-01 -6.21168733e-01 1.26025331e+00 5.41330278e-02 -3.70365381e-02 2.26077721e-01 -1.06503737e+00 9.11748230e-01 7.98256814e-01 -7.97581226e-02 -6.07695524e-03 5.87759137e-01 6.74208283e-01 -3.33246648e-01 -1.10838473e+00 5.56277394e-01 5.83177924e-01 -6.28768265e-01 1.05871618e+00 -1.00374126e+00 5.22400618e-01 -3.66721720e-01 5.09013198e-02 -1.27255583e+00 -6.36395216e-01 -5.71422100e-01 -2.52321661e-01 9.31893706e-01 1.04006350e+00 -5.67432165e-01 8.94962311e-01 7.13249385e-01 -1.35364935e-01 -1.27732396e+00 -6.88678265e-01 -5.74978590e-01 2.61913776e-01 2.03722954e-01 6.59723043e-01 8.28743577e-01 2.68072844e-01 6.31403506e-01 -2.72807837e-01 -5.65157197e-02 3.24071199e-01 1.65478513e-02 4.11851823e-01 -1.24889159e+00 -6.02543950e-01 -4.85723585e-01 -1.06508017e-01 -6.20522618e-01 1.84056014e-02 -9.43342268e-01 -5.00185370e-01 -1.42840731e+00 5.03237724e-01 -4.44301367e-01 -5.62552065e-02 4.75516915e-01 -2.42977560e-01 -2.00062022e-01 -2.08736464e-01 2.88096070e-01 -2.75396496e-01 8.91500473e-01 1.35014153e+00 -3.62467051e-01 -7.26624727e-01 7.10104555e-02 -1.12898171e+00 2.98485100e-01 9.31375980e-01 -4.00696278e-01 -5.01814723e-01 1.41669303e-01 6.21781051e-01 1.25835642e-01 4.70782630e-04 -6.11897826e-01 -1.02248276e-02 -5.98807752e-01 5.83670259e-01 -2.11655110e-01 3.11490178e-01 -5.40794730e-01 8.63310814e-01 6.36510491e-01 -1.87700704e-01 -6.89341009e-01 1.74789175e-01 7.85708964e-01 -3.31384409e-03 -5.36804646e-02 7.17349052e-01 4.13095672e-03 -4.06375706e-01 6.22427344e-01 -7.68069804e-01 -4.80897576e-01 1.12984669e+00 -5.12355030e-01 -9.71769914e-02 -7.57732429e-03 -8.99982929e-01 -1.19397044e-01 4.51050818e-01 1.28346384e-01 7.60413706e-01 -9.12552655e-01 -5.39903343e-01 -1.13828242e-01 2.84231037e-01 3.13078426e-02 1.98727757e-01 9.14872408e-01 -6.45359278e-01 5.24395227e-01 -1.43344864e-01 -3.99857134e-01 -1.11427820e+00 8.49227250e-01 4.85346198e-01 -7.31159821e-02 -2.40954697e-01 5.96206367e-01 5.47947884e-01 -1.43405944e-01 -1.16980597e-02 1.73417300e-01 -1.48934186e-01 -2.21587159e-02 2.54321605e-01 2.49960274e-01 1.47897482e-01 -2.09282964e-01 -4.92417902e-01 2.20582724e-01 -3.48182321e-01 3.79332423e-01 1.73068297e+00 4.00161594e-01 2.00214922e-01 -7.61922449e-02 8.58924150e-01 1.64777115e-02 -1.28686810e+00 1.42996892e-01 -7.39014521e-02 -1.19686149e-01 1.36315897e-01 -1.02976918e+00 -7.03722239e-01 3.21552604e-01 5.02505600e-01 -4.50181305e-01 8.02390754e-01 3.13736647e-02 5.42960823e-01 4.01364148e-01 3.52745712e-01 -9.05736148e-01 2.71518886e-01 1.77329510e-01 8.33306551e-01 -1.17191350e+00 1.94882393e-01 -3.21585834e-01 -5.02828240e-01 1.10972250e+00 4.24991250e-01 4.78259385e-01 2.35075414e-01 -2.04075247e-01 -3.67037028e-01 -3.56457800e-01 -7.08860457e-01 -1.26423500e-02 1.80745095e-01 5.73958397e-01 5.85754454e-01 -1.20272398e-01 -6.31854057e-01 4.20952678e-01 2.41449147e-01 6.13057474e-03 4.56190675e-01 7.61376739e-01 -3.48850787e-01 -1.32277250e+00 2.92626861e-02 6.61689192e-02 -3.19543362e-01 -3.57748985e-01 -8.48790109e-01 6.03258491e-01 9.76023525e-02 7.81575799e-01 -5.63002348e-01 -2.70423681e-01 1.65665329e-01 -1.13189466e-01 6.08216941e-01 -6.22398257e-01 -3.99708331e-01 2.55396307e-01 3.86543125e-02 -1.40676498e-01 -3.83623764e-02 -3.88083935e-01 -1.33340764e+00 -3.06383669e-01 -9.40253854e-01 5.47110498e-01 7.25634456e-01 8.84136379e-01 7.18394578e-01 4.25241768e-01 8.06011558e-01 -5.38836241e-01 -2.17218146e-01 -5.21078885e-01 -4.07354951e-01 -7.90870786e-02 -5.02952747e-02 -6.62613988e-01 -7.86718130e-02 -5.89041710e-02]
[4.981445789337158, 5.707146644592285]
89b727df-d4a4-485d-94e1-6f60db729759
learning-and-memorizing-representative
2001.01349
null
https://arxiv.org/abs/2001.01349v1
https://arxiv.org/pdf/2001.01349v1.pdf
Learning and Memorizing Representative Prototypes for 3D Point Cloud Semantic and Instance Segmentation
3D point cloud semantic and instance segmentation is crucial and fundamental for 3D scene understanding. Due to the complex structure, point sets are distributed off balance and diversely, which appears as both category imbalance and pattern imbalance. As a result, deep networks can easily forget the non-dominant cases during the learning process, resulting in unsatisfactory performance. Although re-weighting can reduce the influence of the well-classified examples, they cannot handle the non-dominant patterns during the dynamic training. In this paper, we propose a memory-augmented network to learn and memorize the representative prototypes that cover diverse samples universally. Specifically, a memory module is introduced to alleviate the forgetting issue by recording the patterns seen in mini-batch training. The learned memory items consistently reflect the interpretable and meaningful information for both dominant and non-dominant categories and cases. The distorted observations and rare cases can thus be augmented by retrieving the stored prototypes, leading to better performances and generalization. Exhaustive experiments on the benchmarks, i.e. S3DIS and ScanNetV2, reflect the superiority of our method on both effectiveness and efficiency. Not only the overall accuracy but also nondominant classes have improved substantially.
['Chunhua Shen', 'Zhi Tian', 'Dong Gong', 'Tong He']
2020-01-06
null
https://www.ecva.net/papers/eccv_2020/papers_ECCV/html/3039_ECCV_2020_paper.php
https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123630545.pdf
eccv-2020-8
['3d-instance-segmentation-1']
['computer-vision']
[-1.52523980e-01 -2.48804912e-01 -7.45589808e-02 -5.96047044e-01 -2.11077631e-01 -2.37023920e-01 1.81438908e-01 2.62005568e-01 -3.09621960e-01 5.21184385e-01 -1.51500821e-01 4.50928546e-02 -3.00171852e-01 -8.85718048e-01 -7.16641605e-01 -7.62672961e-01 8.46203789e-03 6.78234577e-01 4.43744630e-01 5.43703958e-02 3.36675107e-01 7.05931365e-01 -2.06774521e+00 4.33707386e-01 1.23951614e+00 1.17203712e+00 3.73538017e-01 -1.46010548e-01 -6.56912804e-01 3.10485780e-01 -7.14182436e-01 -1.10118471e-01 2.00391278e-01 2.56677885e-02 -3.46731633e-01 2.64137864e-01 4.07020479e-01 -2.53823996e-01 -3.39231431e-01 1.13984394e+00 4.07348067e-01 2.36012220e-01 4.93713170e-01 -1.13759756e+00 -4.94145066e-01 2.63242662e-01 -8.64609063e-01 2.06775054e-01 -3.30597423e-02 2.40323871e-01 6.14498138e-01 -1.12378979e+00 4.94466960e-01 1.33382440e+00 4.62208003e-01 3.58057439e-01 -8.47515702e-01 -8.46109211e-01 6.62993014e-01 6.80962086e-01 -1.56580579e+00 -1.94583789e-01 8.98707688e-01 -2.86730558e-01 6.17508888e-01 4.01989430e-01 8.41133058e-01 7.23121405e-01 -1.30869210e-01 9.16928768e-01 6.54234707e-01 3.05603351e-02 2.70764589e-01 3.23741555e-01 4.06591892e-01 2.89940506e-01 6.54764771e-01 -1.44156709e-01 -3.36005747e-01 2.39989698e-01 5.41430473e-01 6.83706105e-01 -3.57579112e-01 -4.83671278e-01 -1.04580069e+00 3.38818938e-01 7.16101408e-01 2.10622892e-01 -4.31893378e-01 -3.55789244e-01 3.38661313e-01 2.02024370e-01 5.37491262e-01 1.14901878e-01 -5.54339290e-01 8.21330696e-02 -9.26240027e-01 2.65802115e-01 3.24806362e-01 1.14142120e+00 1.05291808e+00 -5.07071428e-02 -2.01385751e-01 1.10432577e+00 1.17298469e-01 5.13566613e-01 5.48020780e-01 -4.27576900e-01 6.61679983e-01 1.35714650e+00 -7.47954473e-02 -1.26273608e+00 -4.71253932e-01 -9.96997893e-01 -1.28306019e+00 1.24931827e-01 1.69500649e-01 4.48073089e-01 -1.36856949e+00 1.55439067e+00 4.80185926e-01 1.45196021e-01 -3.03093642e-01 1.04057479e+00 9.30544257e-01 6.83257401e-01 3.34824510e-02 -2.73828596e-01 1.06822932e+00 -7.05867767e-01 -4.23016131e-01 -3.22624117e-01 2.49651238e-01 -6.07865870e-01 1.12132645e+00 2.96792567e-01 -9.18001592e-01 -7.70082712e-01 -1.08044004e+00 9.34665725e-02 -3.26029837e-01 -5.06886281e-02 4.54622716e-01 1.55507445e-01 -5.02939939e-01 7.11069167e-01 -7.24217057e-01 1.03543550e-01 1.05797064e+00 1.68797404e-01 -1.31028265e-01 -5.21407664e-01 -9.01329815e-01 4.95636165e-01 6.50900900e-01 3.07073116e-01 -6.54350221e-01 -9.27520871e-01 -4.65090036e-01 1.12346314e-01 4.65328038e-01 -4.25657481e-01 8.95923138e-01 -7.86475599e-01 -8.44010353e-01 6.41138077e-01 -2.47691065e-01 -3.82780330e-03 6.92352295e-01 -1.35051429e-01 -5.37393928e-01 -9.11527276e-02 7.33605176e-02 6.07314467e-01 8.65042031e-01 -1.35344601e+00 -8.61716509e-01 -7.00122476e-01 -1.74872473e-01 3.40792775e-01 -2.65113741e-01 -8.56853962e-01 -6.41831279e-01 -6.20432079e-01 9.50634181e-01 -7.48095632e-01 -5.67400781e-03 1.34864494e-01 -3.02239150e-01 -3.49724501e-01 8.25004518e-01 -3.47217023e-01 1.14595485e+00 -2.30334663e+00 3.39089856e-02 3.32214803e-01 3.19293708e-01 3.00515264e-01 -2.93538533e-02 -9.86279547e-02 -5.19205146e-02 -2.87070964e-02 -2.40997404e-01 -1.15177467e-01 -1.34701833e-01 4.72525179e-01 -3.45761210e-01 3.14280659e-01 3.52266073e-01 5.87349355e-01 -8.17044079e-01 -2.78608203e-01 2.70402402e-01 1.69786483e-01 -5.50809622e-01 -4.94295396e-02 -3.55071396e-01 4.38496441e-01 -4.17161942e-01 8.70752513e-01 1.23625290e+00 -2.90399104e-01 -2.48302802e-01 -3.39123666e-01 1.00818932e-01 2.23270595e-01 -1.48277414e+00 1.72863185e+00 -4.26171005e-01 2.07524359e-01 -2.75618106e-01 -8.80078077e-01 1.03248584e+00 -2.10568920e-01 1.70319736e-01 -1.00395441e+00 -5.46937473e-02 3.87452871e-01 -4.95275632e-02 -6.29349709e-01 5.80450892e-01 -5.96702956e-02 3.02662373e-01 3.15118879e-01 -2.19076768e-01 2.62587577e-01 8.33285823e-02 -9.73305013e-03 5.23815632e-01 -4.17584069e-02 1.42586408e-02 -1.77796707e-01 4.60084438e-01 1.59975797e-01 8.67549777e-01 5.97707033e-01 -2.67444160e-02 6.05839491e-01 2.93178380e-01 -8.04732025e-01 -8.15772474e-01 -1.08219719e+00 -2.62710243e-01 9.04217839e-01 7.90848374e-01 -7.40881488e-02 -2.03170002e-01 -5.72896659e-01 2.16280371e-01 8.47696662e-01 -4.77823228e-01 -5.85988462e-01 -5.85433364e-01 -8.80777955e-01 1.73098911e-02 5.54613471e-01 5.25028944e-01 -1.10502982e+00 -5.70616901e-01 1.11020260e-01 -3.19032408e-02 -5.71778834e-01 -1.63230285e-01 7.25645795e-02 -1.16085064e+00 -1.08698678e+00 -6.95895016e-01 -7.15423107e-01 1.01352787e+00 7.23386884e-01 1.20789993e+00 3.76903296e-01 -1.29054159e-01 -2.70668834e-01 -3.31476986e-01 -5.80742657e-01 -8.70604720e-03 1.27575561e-01 6.74932310e-03 -3.41864377e-02 6.17882133e-01 -7.16626585e-01 -7.39524364e-01 4.77743924e-01 -9.45640266e-01 1.31371677e-01 6.66391432e-01 7.47591972e-01 1.04166651e+00 3.74762386e-01 5.74724674e-01 -9.43253279e-01 2.02972382e-01 -5.77898085e-01 -4.12408531e-01 1.93026274e-01 -6.66180074e-01 -9.52861533e-02 5.38293898e-01 -5.94759583e-01 -9.39559996e-01 -3.06883872e-01 1.99153945e-02 -7.68943548e-01 -2.04828680e-01 2.92513072e-01 -5.48574626e-01 1.98752105e-01 4.41651344e-01 3.42739999e-01 -8.92235190e-02 -7.11748183e-01 5.47547154e-02 4.73056585e-01 3.21357071e-01 -5.33879340e-01 8.12144279e-01 5.23827791e-01 -1.19002700e-01 -5.81976652e-01 -7.54658818e-01 -4.18389589e-01 -4.44737792e-01 -1.69321463e-01 2.40435049e-01 -9.71398056e-01 -3.83279711e-01 7.11133778e-01 -1.13987565e+00 2.08638221e-01 -5.23457646e-01 2.60666251e-01 4.36109714e-02 1.66654915e-01 -2.76142418e-01 -6.85965240e-01 -1.46048889e-01 -1.08349884e+00 9.23146188e-01 5.94564497e-01 4.09137048e-02 -4.29982394e-01 -2.97782034e-01 -4.97979298e-02 3.56040120e-01 1.01313926e-01 1.21973729e+00 -5.80375373e-01 -1.00306857e+00 -3.55520427e-01 -4.51458067e-01 2.31623948e-01 1.14378840e-01 -9.32382122e-02 -1.00917602e+00 -2.51855820e-01 8.42696652e-02 -1.17214490e-02 1.08743143e+00 5.73379099e-02 1.57398677e+00 -1.57732174e-01 -4.50732976e-01 6.26584232e-01 1.22110915e+00 2.93367326e-01 6.51874840e-01 2.42772698e-01 8.75944197e-01 5.83158731e-01 7.67321050e-01 4.02176529e-01 3.12735856e-01 3.39535862e-01 6.38770759e-01 9.97498706e-02 -6.02035597e-02 -3.00999224e-01 -3.26614499e-01 9.68175113e-01 1.05248839e-01 -1.24416776e-01 -9.48843181e-01 5.76256514e-01 -1.79138207e+00 -8.10083807e-01 -9.42763016e-02 2.23220539e+00 7.06231594e-01 4.67834860e-01 -1.73052326e-01 3.63021135e-01 8.67435813e-01 1.08965486e-01 -9.92269039e-01 2.50689119e-01 -2.51944005e-01 5.36768883e-03 1.63352355e-01 -1.67037528e-02 -8.37308109e-01 5.26552200e-01 4.57900429e+00 1.11652303e+00 -1.17929113e+00 3.28802653e-02 7.96634495e-01 -5.20784259e-01 -5.02187312e-01 -3.91630560e-01 -6.96012735e-01 6.54405773e-01 1.63560510e-01 -5.64258993e-02 2.49360666e-01 9.90850925e-01 -1.34149194e-01 -3.41628417e-02 -1.02922761e+00 1.40981972e+00 1.15069244e-02 -1.38738155e+00 4.45263565e-01 -1.59561202e-01 6.86169267e-01 -6.67810515e-02 1.22308642e-01 3.68443996e-01 -4.40379173e-01 -7.38171637e-01 9.86304700e-01 6.32186234e-01 7.36218274e-01 -1.03447652e+00 8.08485627e-01 6.18651748e-01 -9.24664497e-01 -1.88141927e-01 -8.82244825e-01 1.20711394e-01 -1.38055280e-01 1.09795249e+00 -4.18585300e-01 6.31361544e-01 9.98700023e-01 1.00472248e+00 -6.17345035e-01 1.31744885e+00 3.62235270e-02 2.48700678e-01 -4.84318495e-01 3.04755066e-02 1.42114505e-01 -3.00820887e-01 6.80687070e-01 7.32240081e-01 4.05875683e-01 1.16218753e-01 1.73296947e-02 8.32706332e-01 -1.49299368e-01 -4.27416302e-02 -2.13137627e-01 3.01376760e-01 6.87402368e-01 9.60711896e-01 -8.62827122e-01 -4.84917283e-01 -1.60170421e-01 5.89086473e-01 3.50995511e-01 3.38848710e-01 -5.45357108e-01 -2.60964304e-01 7.19073057e-01 2.59988815e-01 4.20134664e-01 -3.76344919e-02 -7.24220693e-01 -1.21346438e+00 4.79120612e-01 -7.50099003e-01 5.08825123e-01 -3.82570773e-01 -1.40214789e+00 6.86610579e-01 -5.91096804e-02 -1.67872977e+00 1.89293414e-01 -2.60090768e-01 -5.72845221e-01 7.19087541e-01 -1.54983521e+00 -6.47484004e-01 -8.99400532e-01 2.82730728e-01 6.75047338e-01 -7.44250044e-02 3.07733566e-01 7.88543224e-01 -7.67924845e-01 6.36976242e-01 1.33846775e-01 -3.65469486e-01 5.23176789e-01 -9.38621402e-01 2.07212418e-01 6.03635192e-01 3.67798507e-02 5.41180432e-01 5.14774859e-01 -7.00929165e-01 -1.05233800e+00 -1.36787534e+00 4.64482784e-01 -2.28859544e-01 4.97995988e-02 -2.35691369e-01 -1.58096957e+00 2.14729756e-01 -5.76067746e-01 -6.01343177e-02 4.03267503e-01 5.59550859e-02 -4.26910460e-01 -5.58679283e-01 -1.11064076e+00 4.98587847e-01 1.45325601e+00 -1.61004364e-01 -6.61584437e-01 2.25142807e-01 7.65594542e-01 -5.93136549e-01 -3.67835671e-01 7.05187798e-01 2.98032433e-01 -1.10306740e+00 9.97270942e-01 -4.93970126e-01 4.10678387e-01 -5.59785426e-01 -6.77312016e-02 -1.24469864e+00 -3.90242726e-01 1.12055957e-01 -2.36336052e-01 1.26728809e+00 2.42247656e-01 -6.21786773e-01 9.12164330e-01 4.10782188e-01 -3.47649664e-01 -9.26312983e-01 -1.08727288e+00 -6.82587504e-01 -9.20215324e-02 -4.01590347e-01 1.18500352e+00 9.07738090e-01 -6.80845499e-01 9.03680697e-02 -1.31811619e-01 1.33258745e-01 5.93702972e-01 7.29945660e-01 6.12450540e-01 -1.44722295e+00 4.02877331e-02 -5.52853584e-01 -6.09133840e-01 -1.21430123e+00 -2.27251425e-01 -8.70785654e-01 4.37099338e-02 -1.28043067e+00 1.57486886e-01 -1.10275102e+00 -5.14237046e-01 2.77490586e-01 -4.61972237e-01 1.13307871e-01 2.90487092e-02 6.07445180e-01 -8.36813033e-01 7.38713741e-01 1.44994104e+00 -3.98988724e-01 -4.32972133e-01 2.22258985e-01 -6.82749867e-01 7.89588869e-01 7.90232420e-01 -6.10439837e-01 -6.17455840e-01 -7.37754524e-01 2.36691371e-01 -4.43173587e-01 3.86639386e-01 -1.16875052e+00 3.17704380e-01 -1.06855415e-01 8.54000032e-01 -1.36782861e+00 1.67003959e-01 -1.07795227e+00 3.57277334e-01 3.61935914e-01 -2.78176963e-02 -7.54363313e-02 3.38902920e-01 7.86352336e-01 -2.19740018e-01 -1.73528016e-01 6.75638974e-01 -2.42370129e-01 -8.36525500e-01 6.23878241e-01 3.10269564e-01 1.56668931e-01 9.68206763e-01 -5.75335324e-01 -3.34889084e-01 5.02771772e-02 -5.94177425e-01 5.06285787e-01 6.57844901e-01 5.89251399e-01 8.39236677e-01 -1.38349605e+00 -5.09423375e-01 6.36241376e-01 3.28745425e-01 9.32775021e-01 8.52562368e-01 5.68187177e-01 -4.18969452e-01 7.84439296e-02 -2.79145330e-01 -1.00366378e+00 -7.17433870e-01 6.42499387e-01 2.24096999e-01 3.17575634e-02 -7.40527451e-01 7.83697665e-01 3.09913844e-01 -3.02115858e-01 4.70041931e-01 -3.62728953e-01 -1.67765915e-01 2.99685091e-01 7.00511515e-01 4.96994972e-01 3.04520786e-01 -3.39232892e-01 -5.03307402e-01 5.59836745e-01 -4.31906015e-01 8.10701370e-01 1.52327788e+00 -1.90923169e-01 -1.45950407e-01 6.57707393e-01 1.06814623e+00 -2.58824706e-01 -1.25105989e+00 -3.73571992e-01 -3.68142873e-02 -7.53643811e-01 -1.74877092e-01 -6.38024151e-01 -1.37278652e+00 1.01784313e+00 7.63681769e-01 1.21107042e-01 1.19586205e+00 -5.52477539e-02 8.35220277e-01 4.31812912e-01 3.92452270e-01 -9.71638322e-01 -8.51822346e-02 4.10023153e-01 8.43340933e-01 -1.15630746e+00 1.08449474e-01 -5.23413479e-01 -4.00595188e-01 9.07198191e-01 8.98854733e-01 -1.61177322e-01 6.27074897e-01 -2.50471026e-01 -9.39403102e-03 -3.11913878e-01 -5.34222305e-01 1.67802483e-01 3.83212000e-01 4.14836466e-01 -2.12385118e-01 1.07611395e-01 2.28510313e-02 7.38057613e-01 -3.57068814e-02 -2.00281769e-01 1.27989367e-01 7.72153854e-01 -4.89285856e-01 -5.38879871e-01 -2.81040847e-01 9.09232497e-01 1.13727733e-01 1.93929866e-01 -2.65043061e-02 7.77786911e-01 5.72065592e-01 4.84788179e-01 4.57331479e-01 -4.87859279e-01 8.15086484e-01 -6.52718991e-02 2.57640481e-01 -4.50560778e-01 -2.61199474e-01 -1.70175523e-01 -4.02913779e-01 -5.02309859e-01 -2.71715671e-01 -4.37586010e-01 -1.28613663e+00 -4.42904890e-01 -3.96463484e-01 4.15032580e-02 3.88613671e-01 7.25978136e-01 5.60245097e-01 8.40156496e-01 6.88015103e-01 -8.85805488e-01 -6.30519152e-01 -7.99725056e-01 -5.79818964e-01 7.41552711e-01 2.20304295e-01 -1.02098656e+00 -4.58787858e-01 -4.42192972e-01]
[7.895294666290283, -3.2336583137512207]
476b5c6e-7c82-4dd5-95a1-799fba1622b8
oops-did-i-just-say-that-testing-and
2305.02626
null
https://arxiv.org/abs/2305.02626v1
https://arxiv.org/pdf/2305.02626v1.pdf
"Oops, Did I Just Say That?" Testing and Repairing Unethical Suggestions of Large Language Models with Suggest-Critique-Reflect Process
As the popularity of large language models (LLMs) soars across various applications, ensuring their alignment with human values has become a paramount concern. In particular, given that LLMs have great potential to serve as general-purpose AI assistants in daily life, their subtly unethical suggestions become a serious and real concern. Tackling the challenge of automatically testing and repairing unethical suggestions is thus demanding. This paper introduces the first framework for testing and repairing unethical suggestions made by LLMs. We first propose ETHICSSUITE, a test suite that presents complex, contextualized, and realistic moral scenarios to test LLMs. We then propose a novel suggest-critic-reflect (SCR) process, serving as an automated test oracle to detect unethical suggestions. We recast deciding if LLMs yield unethical suggestions (a hard problem; often requiring human expertise and costly to decide) into a PCR task that can be automatically checked for violation. Moreover, we propose a novel on-the-fly (OTF) repairing scheme that repairs unethical suggestions made by LLMs in real-time. The OTF scheme is applicable to LLMs in a black-box API setting with moderate cost. With ETHICSSUITE, our study on seven popular LLMs (e.g., ChatGPT, GPT-4) uncovers in total 109,824 unethical suggestions. We apply our OTF scheme on two LLMs (Llama-13B and ChatGPT), which generates valid repair to a considerable amount of unethical ones, paving the way for more ethically conscious LLMs.
['Shuai Wang', 'Ao Sun', 'Zongjie Li', 'Pingchuan Ma']
2023-05-04
null
null
null
null
['moral-scenarios']
['miscellaneous']
[ 1.39297202e-01 5.74021697e-01 1.59600880e-02 -3.58049631e-01 -6.85583770e-01 -7.90185452e-01 5.07903636e-01 -7.41795078e-02 -4.68481541e-01 7.56015837e-01 -2.53384739e-01 -9.52742040e-01 4.61887419e-02 -4.48018044e-01 -7.03447163e-01 -3.34964454e-01 4.09061044e-01 3.03376287e-01 -3.56818289e-02 -1.40042603e-01 6.89941764e-01 1.14953846e-01 -1.27433932e+00 1.84344783e-01 1.33887732e+00 4.56792682e-01 -1.05772652e-01 6.65819347e-01 3.13954324e-01 8.42265248e-01 -8.28914165e-01 -1.27762294e+00 1.19563304e-01 -3.12897652e-01 -1.01150572e+00 -2.75321096e-01 3.96843106e-01 -6.88164234e-01 5.17667472e-01 1.28664923e+00 5.54954529e-01 -1.27728909e-01 5.37268698e-01 -1.88642871e+00 -5.52605093e-01 8.72837543e-01 -3.70735943e-01 -3.45716506e-01 4.99065340e-01 5.84136844e-01 9.54019845e-01 -5.02925694e-01 4.58993107e-01 1.48186469e+00 4.26577449e-01 8.76228452e-01 -9.88997996e-01 -7.03462064e-01 -2.69817084e-01 1.55559570e-01 -1.00757039e+00 -4.31253225e-01 4.25664127e-01 -5.28739095e-01 9.05368745e-01 8.36179137e-01 3.30037117e-01 1.73106635e+00 3.12370002e-01 6.15746260e-01 1.15743113e+00 -4.82298464e-01 5.42204976e-01 7.62289584e-01 9.61900800e-02 5.78044593e-01 5.54987252e-01 -2.31422722e-01 -2.83900261e-01 -1.00583959e+00 2.21348003e-01 -1.84331208e-01 7.95762911e-02 1.56511739e-01 -8.37733626e-01 8.92117023e-01 -3.65342736e-01 7.10104182e-02 -4.03458416e-01 7.23935142e-02 4.27346110e-01 2.70868838e-01 1.22902088e-01 7.58714139e-01 -4.71047401e-01 -8.54887009e-01 -5.43750882e-01 4.44134831e-01 1.11146033e+00 6.57579660e-01 5.22465229e-01 -2.68421650e-01 -1.72526315e-01 8.56803060e-01 2.54191458e-01 5.53735793e-01 2.53272086e-01 -1.44594479e+00 3.67418259e-01 6.30941749e-01 5.00374854e-01 -1.06715870e+00 9.48218033e-02 -1.33676022e-01 -4.27451491e-01 2.42805451e-01 2.51335114e-01 -1.77341551e-01 5.28965071e-02 1.69162953e+00 3.46261233e-01 -1.51494548e-01 -2.58641750e-01 6.82583272e-01 3.13357823e-03 3.22704703e-01 -2.12087017e-02 -3.26207489e-01 1.08600342e+00 -7.27617443e-01 -5.26243150e-01 -2.24904463e-01 1.19043040e+00 -4.97613162e-01 1.91536677e+00 9.24871683e-01 -9.95011866e-01 2.19610140e-01 -5.27747035e-01 1.21864118e-01 1.05544012e-02 -1.11968979e-01 4.79761273e-01 1.16292620e+00 -1.07274902e+00 4.42125022e-01 -2.84611851e-01 -4.82368141e-01 2.36970752e-01 9.26550329e-02 -3.29447299e-01 -7.28506818e-02 -1.03101778e+00 8.90954673e-01 -1.31522372e-01 6.27625212e-02 -8.26816618e-01 -2.59098828e-01 -5.92263758e-01 -9.18053910e-02 9.54853594e-01 -1.34123415e-01 1.48853326e+00 -6.18672848e-01 -1.48006308e+00 8.32796812e-01 -3.66088115e-02 -2.50190496e-01 9.85565007e-01 -2.90221214e-01 -2.64598638e-01 -2.09401309e-01 4.26691145e-01 4.80425000e-01 9.53311324e-01 -8.97738755e-01 -2.65898526e-01 5.34494892e-02 3.17478150e-01 -3.12322170e-01 -7.21021235e-01 5.67958534e-01 2.94146746e-01 -3.52298081e-01 -7.52333999e-01 -1.00594199e+00 -4.96901833e-02 -3.36460531e-01 -9.01523709e-01 -6.81463540e-01 4.86849278e-01 -4.43678379e-01 1.65165603e+00 -2.08138442e+00 -1.35858938e-01 1.99721396e-01 1.38162643e-01 5.64854920e-01 -2.25377306e-01 3.70973915e-01 3.71725768e-01 8.05393100e-01 -3.60268056e-01 -1.15457900e-01 5.99509537e-01 1.62829701e-02 -4.06139463e-01 1.81327388e-01 2.74545819e-01 9.19025958e-01 -8.14012706e-01 -5.11945009e-01 -2.81244487e-01 -1.58419982e-01 -1.05469286e+00 5.18246293e-01 -2.98966527e-01 5.89605346e-02 -3.70878249e-01 9.06100929e-01 5.34504831e-01 3.78027223e-02 -5.30470610e-02 5.29908955e-01 -1.53661832e-01 3.44467759e-01 -7.36590862e-01 1.01351452e+00 -1.79492131e-01 1.22740149e-01 8.76127407e-02 -5.07513463e-01 1.01101923e+00 1.50364414e-01 -2.55643606e-01 -6.69131994e-01 -1.80433635e-02 5.88830769e-01 2.09954634e-01 -1.02998614e+00 5.27653158e-01 1.18812978e-01 -5.48703671e-01 1.11104274e+00 -5.25148451e-01 -1.06349520e-01 4.61743884e-02 3.61738592e-01 1.66432106e+00 -1.58912446e-02 2.88320899e-01 -1.46799117e-01 6.56314790e-01 -2.81041384e-01 7.97370851e-01 1.05651391e+00 -6.48358703e-01 -3.76724862e-02 9.18350816e-01 -2.88142353e-01 -9.55494821e-01 -4.94924754e-01 2.21672997e-01 9.15457845e-01 -5.56468308e-01 -5.74208319e-01 -1.17581010e+00 -1.21224535e+00 -1.59915350e-02 1.34151995e+00 -2.74428010e-01 -4.76516813e-01 -2.36460730e-01 -2.83574402e-01 1.01792014e+00 -2.45857581e-01 5.50346553e-01 -1.45293820e+00 -1.12524998e+00 2.24380028e-02 -3.42495054e-01 -9.12749231e-01 -3.01027358e-01 -1.76743969e-01 -2.64612138e-01 -1.11621332e+00 -2.81205386e-01 4.86730039e-02 5.92986584e-01 8.94833636e-03 7.91412711e-01 5.03939748e-01 -3.22799891e-01 2.24293038e-01 -5.31753540e-01 -4.34899837e-01 -9.30791914e-01 -1.72260746e-01 3.85766447e-01 -1.36708111e-01 6.01603746e-01 -4.43270564e-01 -9.40467119e-02 4.96577978e-01 -8.25288475e-01 -2.62985915e-01 4.25472647e-01 5.43102682e-01 -2.99716622e-01 -1.02210052e-01 7.41905987e-01 -1.15095043e+00 1.19463611e+00 -5.44656873e-01 -4.75824505e-01 3.65114838e-01 -6.90463185e-01 -1.74969375e-01 6.73674464e-01 -4.97068614e-01 -7.65113473e-01 -6.60541952e-01 -1.77877188e-01 -1.32091530e-02 -4.90254462e-01 3.60983431e-01 -1.81274876e-01 7.73322955e-03 7.86991239e-01 -1.65606409e-01 -7.23034441e-02 -3.93834651e-01 1.17829271e-01 1.18022120e+00 2.69085914e-01 -8.88426304e-01 8.72017622e-01 -2.17546269e-01 -3.68306458e-01 -5.89453518e-01 -5.19234836e-01 -1.33539289e-02 1.00477599e-01 -2.32692301e-01 4.17666912e-01 -1.02299221e-01 -1.32671165e+00 4.37426716e-01 -1.35369623e+00 -6.53887928e-01 2.47513428e-01 -1.94889009e-02 -4.57305551e-01 8.35844219e-01 -5.14167547e-01 -1.43753433e+00 -5.82561970e-01 -1.41260803e+00 7.82790899e-01 -1.95506856e-01 -1.00829303e+00 -4.88111913e-01 -2.93124486e-02 7.50461400e-01 5.00044644e-01 -6.53409585e-02 1.25863099e+00 -8.56190801e-01 -1.73684940e-01 -3.40674609e-01 8.54594782e-02 5.81549883e-01 -1.19491532e-01 1.69426620e-01 -7.81856120e-01 -1.40834853e-01 2.47125119e-01 -8.32556427e-01 -9.18984637e-02 -3.17170620e-01 1.11596370e+00 -1.27054954e+00 2.08944455e-01 -2.30199490e-02 6.93225265e-01 2.42424011e-01 7.23182619e-01 5.15215099e-01 4.04894531e-01 8.22626948e-01 9.98253286e-01 8.14604342e-01 2.55052537e-01 6.57402635e-01 5.54085076e-01 4.70899016e-01 6.64325476e-01 -1.54860571e-01 1.20400906e+00 4.49177384e-01 3.09688002e-01 -3.36581081e-01 -9.65332747e-01 1.69717342e-01 -1.93145573e+00 -9.63777363e-01 -2.12239519e-01 2.30066895e+00 1.04435956e+00 2.19602942e-01 2.00100064e-01 2.69887507e-01 5.46827137e-01 -3.09345037e-01 -4.21193630e-01 -1.36090672e+00 3.76314014e-01 -1.70786977e-01 -9.29305181e-02 4.25411046e-01 -6.12103403e-01 8.63721669e-01 5.41426468e+00 7.40646541e-01 -8.67402136e-01 1.65270060e-01 8.30593705e-01 8.69633853e-02 -7.49192536e-01 1.53149620e-01 -7.13512301e-01 7.74727166e-01 1.22558737e+00 -2.18010202e-01 6.86488450e-01 1.10833943e+00 5.79421163e-01 -3.05213183e-01 -1.15276313e+00 7.59843409e-01 1.54690549e-01 -8.24139535e-01 -1.48841992e-01 2.90660441e-01 2.76267141e-01 -2.80137271e-01 1.12495407e-01 4.34331208e-01 3.80812734e-01 -7.76891291e-01 8.86804998e-01 2.66664028e-01 5.48413932e-01 -7.20667958e-01 8.25866699e-01 1.00581539e+00 1.45382574e-02 -2.80437678e-01 -4.65341568e-01 -3.29277515e-01 -1.20935433e-01 7.77097106e-01 -1.09199345e+00 -3.38370278e-02 7.47806847e-01 3.73484880e-01 -6.48400128e-01 7.63255596e-01 -6.22077584e-01 7.65286863e-01 -5.16933165e-02 -2.94791549e-01 1.25868723e-01 -2.51252651e-01 6.43634439e-01 9.13296223e-01 3.77268910e-01 -1.39777422e-01 -1.14200212e-01 1.40554798e+00 -2.13416126e-02 7.34526813e-02 -7.86434412e-01 -5.99382937e-01 6.24459326e-01 1.48200619e+00 -1.67419091e-01 -1.80409774e-01 1.34852707e-01 1.08063066e+00 3.96409690e-01 1.85917974e-01 -8.34734082e-01 -3.97745728e-01 5.23373306e-01 1.34169146e-01 -4.98685449e-01 2.04488665e-01 -5.39320782e-02 -1.11067116e+00 3.75268459e-01 -1.47045958e+00 1.94182042e-02 -8.00663948e-01 -1.23503006e+00 4.93192822e-01 -2.78700311e-02 -8.37544739e-01 -6.06303453e-01 -3.62746775e-01 -8.08437467e-01 6.27151370e-01 -1.11349928e+00 -7.64442980e-01 1.12583824e-02 2.17129663e-01 3.20729524e-01 -4.63748723e-02 9.09411788e-01 1.90953076e-01 -8.70604932e-01 9.06729281e-01 -7.47866452e-01 -4.14553136e-01 9.53115463e-01 -7.96365917e-01 4.74147916e-01 9.80260253e-01 -3.69770348e-01 1.11177170e+00 6.73575222e-01 -9.36430275e-01 -1.42283964e+00 -8.67000818e-01 1.53362262e+00 -6.64233804e-01 8.26918900e-01 -7.24671245e-01 -9.26504314e-01 5.20216882e-01 -2.82322727e-02 -6.03879154e-01 6.06963575e-01 -9.37131047e-02 -4.86461073e-01 2.19069257e-01 -1.47094381e+00 8.26444924e-01 8.81745934e-01 -6.88935697e-01 -3.48946452e-01 5.97278118e-01 8.59745741e-01 2.34577194e-01 -2.97914714e-01 -3.94995064e-02 3.93614411e-01 -1.21133518e+00 2.11708501e-01 -5.68513691e-01 5.25102973e-01 4.83543426e-02 9.04899687e-02 -9.55695212e-01 5.84427640e-02 -1.36141503e+00 2.33223438e-01 1.67293310e+00 2.64259279e-01 -8.40489924e-01 3.47037494e-01 1.42704225e+00 -1.83820501e-02 -5.77077568e-01 -9.10198212e-01 -7.54755318e-01 -1.73891842e-01 -8.54108989e-01 5.12342811e-01 1.16859007e+00 5.70683837e-01 -2.27604374e-01 -6.10200703e-01 -2.51305491e-01 4.91343081e-01 -6.01256967e-01 1.06119704e+00 -1.05967045e+00 -5.25759876e-01 -3.06730539e-01 -1.62624624e-02 -2.02336922e-01 4.19942319e-01 -8.00864816e-01 8.97432715e-02 -9.05980587e-01 4.90621030e-01 -5.80815852e-01 6.90292865e-02 1.12263870e+00 -2.28758737e-01 -1.51618049e-01 1.09614179e-01 5.37783578e-02 -6.44314229e-01 2.06059232e-01 4.64586884e-01 9.32795405e-02 8.57598558e-02 1.81098189e-02 -9.12241161e-01 8.21262777e-01 8.24912488e-01 -7.35830545e-01 -3.38202268e-01 -1.89334497e-01 8.99174452e-01 -3.34770322e-01 7.39260137e-01 -7.15332925e-01 -1.80757884e-02 -6.03887439e-01 -7.74119139e-01 -4.37431224e-02 -2.18629777e-01 -6.48420393e-01 3.77347097e-02 5.17414808e-01 -5.22379577e-01 2.48306856e-01 -5.57560101e-02 9.69634503e-02 3.42505991e-01 -7.83887923e-01 2.59200364e-01 -1.98734716e-01 -1.62413090e-01 -3.78778353e-02 -6.22005224e-01 -1.00684606e-01 9.99090970e-01 1.87942591e-02 -7.46857166e-01 -3.48896563e-01 8.48364159e-02 2.08841756e-01 7.14105666e-01 3.99494171e-01 7.65044808e-01 -9.13905799e-01 -4.91527110e-01 1.99373156e-01 1.99476659e-01 -5.32040894e-01 -1.97438691e-02 9.98261690e-01 -1.74314722e-01 3.53383482e-01 -1.19062483e-01 -2.10185722e-01 -1.30491364e+00 5.00390947e-01 -7.95465857e-02 1.09404124e-01 -2.93547541e-01 9.35264587e-01 -1.04776815e-01 -7.21530497e-01 5.01502216e-01 -8.43204632e-02 1.33288294e-01 -3.41106296e-01 5.72192192e-01 5.73791206e-01 -4.62094471e-02 -1.57306045e-01 -4.37450051e-01 -8.26328099e-02 -1.46983922e-01 -3.18036735e-01 1.29766989e+00 6.46697506e-02 -7.60286570e-01 4.19032007e-01 7.65322089e-01 2.87451178e-01 -7.54669070e-01 3.39637339e-01 4.03282672e-01 -6.37810647e-01 -4.49661255e-01 -1.12679160e+00 -3.36440742e-01 9.95791912e-01 -9.94117111e-02 3.37471932e-01 7.19285488e-01 -2.76533008e-01 7.56335318e-01 7.97017395e-01 8.38290691e-01 -1.20538521e+00 2.28767008e-01 2.62978941e-01 8.99414241e-01 -1.16280723e+00 -3.90218645e-01 -2.01523498e-01 -8.41327488e-01 8.29400063e-01 8.48982394e-01 4.30572838e-01 -1.46669954e-01 2.60056015e-02 -6.72901198e-02 -3.26010324e-02 -1.09446943e+00 8.89790654e-01 -3.15863580e-01 3.81424904e-01 3.73353839e-01 2.78643042e-01 -6.20678604e-01 8.64723742e-01 -2.44706050e-01 2.17173040e-01 1.12878740e+00 9.32357907e-01 -4.16002810e-01 -1.45955539e+00 -4.06248212e-01 3.54349107e-01 -4.41972762e-01 -2.23083854e-01 -9.81163979e-01 2.68136114e-01 1.27413854e-01 1.31391478e+00 -5.76196730e-01 -7.07837880e-01 1.23628512e-01 2.70026922e-01 -2.93691754e-01 -6.58117890e-01 -1.04797053e+00 -1.20400138e-01 5.13600409e-01 -8.97429943e-01 6.40022606e-02 -6.40799105e-01 -9.59027767e-01 -7.09579706e-01 -2.51738220e-01 3.62299323e-01 5.54636896e-01 1.04579401e+00 6.13151193e-01 -2.01303065e-01 6.83122396e-01 -3.52067888e-01 -1.14785671e+00 -9.06359136e-01 -2.59764761e-01 5.02943575e-01 2.34102309e-02 -4.87151384e-01 -5.17994404e-01 -4.39199477e-01]
[10.29927921295166, 7.6158246994018555]
07d54cee-d462-4509-bd0f-35beb13de91e
effects-of-human-dynamics-on-epidemic
1605.00899
null
http://arxiv.org/abs/1605.00899v1
http://arxiv.org/pdf/1605.00899v1.pdf
Effects of human dynamics on epidemic spreading in C\^{o}te d'Ivoire
Understanding and predicting outbreaks of contagious diseases are crucial to the development of society and public health, especially for underdeveloped countries. However, challenging problems are encountered because of complex epidemic spreading dynamics influenced by spatial structure and human dynamics (including both human mobility and human interaction intensity). We propose a systematical model to depict nationwide epidemic spreading in C\^{o}te d'Ivoire, which integrates multiple factors, such as human mobility, human interaction intensity, and demographic features. We provide insights to aid in modeling and predicting the epidemic spreading process by data-driven simulation and theoretical analysis, which is otherwise beyond the scope of local evaluation and geometrical views. We show that the requirement that the average local basic reproductive number to be greater than unity is not necessary for outbreaks of epidemics. The observed spreading phenomenon can be roughly explained as a heterogeneous diffusion-reaction process by redefining mobility distance according to the human mobility volume between nodes, which is beyond the geometrical viewpoint. However, the heterogeneity of human dynamics still poses challenges to precise prediction.
[]
2016-04-30
null
null
null
null
['human-dynamics']
['computer-vision']
[-1.51952282e-01 -1.08335465e-01 -9.97836217e-02 2.46127173e-01 5.39806545e-01 -3.24508816e-01 4.66839671e-01 1.96811602e-01 -4.62699682e-01 6.83966279e-01 8.42556804e-02 -6.35949194e-01 -4.76012975e-01 -9.81723905e-01 -7.91377574e-02 -9.80498314e-01 -8.37623477e-01 6.01988196e-01 2.76784956e-01 -5.65332592e-01 -1.09352894e-01 7.33843327e-01 -7.98138618e-01 -4.25049037e-01 1.10063219e+00 3.25855166e-01 1.43697590e-01 7.03149080e-01 -2.19854824e-02 2.61575550e-01 -6.14297569e-01 -5.66791445e-02 3.59316021e-02 -5.43303251e-01 -4.42198992e-01 5.36380820e-02 -1.28638816e+00 -1.61603332e-01 -2.99087584e-01 6.23268306e-01 2.50723839e-01 -3.09562266e-01 1.11522830e+00 -1.30383229e+00 -6.66153729e-01 2.33860001e-01 -7.70459056e-01 5.20940542e-01 1.65973246e-01 5.31526171e-02 3.91745567e-01 -2.28597492e-01 6.90758944e-01 8.29964340e-01 7.54006386e-01 2.03505039e-01 -7.54537761e-01 -3.77967805e-01 1.08433828e-01 -2.28885487e-01 -1.64388180e+00 8.03173557e-02 6.57538891e-01 -6.02323294e-01 6.32133543e-01 3.34384829e-01 1.07389271e+00 7.35059977e-01 7.05763698e-01 1.25310108e-01 4.80533034e-01 -2.28449270e-01 1.13448374e-01 -4.28820960e-02 -1.89866364e-01 5.29029727e-01 7.46093869e-01 -9.00838990e-03 1.73984900e-01 -4.68535632e-01 1.24320674e+00 3.21649551e-01 -3.72134835e-01 -2.53955543e-01 -1.03253317e+00 8.57887387e-01 5.17447114e-01 6.09491050e-01 -6.44109190e-01 -1.81666851e-01 -6.08079210e-02 4.82283533e-01 7.80231595e-01 2.69005895e-01 -5.50531209e-01 1.17876455e-01 -6.17084026e-01 2.21161842e-01 6.52555525e-01 5.48181534e-01 4.57163006e-01 -1.56140402e-01 4.87447977e-01 4.06948388e-01 3.63338381e-01 1.15348482e+00 -1.69181183e-01 -6.20077491e-01 3.40247601e-01 9.16329086e-01 3.10250431e-01 -1.26043868e+00 -8.29335272e-01 -4.44990277e-01 -1.53544235e+00 -3.44737858e-01 4.46887553e-01 -6.66797161e-01 -4.73958403e-01 1.54422987e+00 4.69896138e-01 3.54658403e-02 -9.08521935e-02 5.82430720e-01 3.57910454e-01 9.11938965e-01 2.61022240e-01 -9.06991243e-01 1.03237998e+00 -2.79074073e-01 -7.38178372e-01 2.50599682e-01 1.00182772e+00 -3.65590870e-01 3.21865469e-01 -2.64425516e-01 -1.04709578e+00 5.24018258e-02 -3.05887908e-01 8.62894535e-01 -3.14620852e-01 -3.05430740e-01 4.06163901e-01 5.97572088e-01 -1.18975258e+00 1.02593414e-01 -9.86272812e-01 -8.20650458e-01 1.42355949e-01 3.08991909e-01 -1.29715309e-01 2.43564516e-01 -1.43459773e+00 7.32987225e-01 -2.07717910e-01 2.89051533e-01 -1.63787663e-01 -4.51912522e-01 -5.68571031e-01 -8.76515880e-02 -1.59244575e-02 -8.59729767e-01 5.22174478e-01 -4.26715195e-01 -8.33574057e-01 3.70699525e-01 -2.42553145e-01 -1.22777328e-01 7.03487515e-01 3.03563267e-01 -7.32644796e-01 2.50633299e-01 -3.58095020e-02 9.01298374e-02 2.41860509e-01 -1.30777943e+00 -4.79273379e-01 -5.31535089e-01 -2.76870400e-01 3.15959007e-01 -3.04942399e-01 3.16613913e-01 -2.71552533e-01 -6.10617578e-01 1.24960296e-01 -9.34050739e-01 -7.97342002e-01 -2.12130949e-01 -2.27141559e-01 5.25416341e-03 7.89598584e-01 -6.28690600e-01 1.60265243e+00 -1.71442878e+00 1.74056366e-01 6.46364868e-01 4.90803331e-01 1.71149179e-01 6.51564002e-02 9.15622473e-01 3.53387892e-01 3.13589394e-01 -4.35851544e-01 2.06525400e-01 -5.47923386e-01 -3.48323248e-02 -1.40362129e-01 8.12088847e-01 1.70163512e-01 8.36820364e-01 -9.81335223e-01 -3.30005616e-01 5.44265211e-02 7.06839681e-01 -5.15198767e-01 4.84359711e-02 2.08001107e-01 6.91869378e-01 -9.17986929e-01 3.21425527e-01 8.01414967e-01 -4.74275410e-01 4.96398926e-01 5.14540553e-01 -4.13588017e-01 -3.00300658e-01 -8.59258115e-01 4.69213784e-01 -2.51990885e-01 2.75858670e-01 4.14005727e-01 -7.05775201e-01 7.77776599e-01 4.45864677e-01 9.30611193e-01 -3.33052009e-01 2.23627672e-01 2.23114211e-02 1.94481045e-01 -4.82607394e-01 3.42440605e-01 -1.14857703e-01 3.34768482e-02 9.83292580e-01 -6.43380880e-01 3.54221433e-01 -1.43989539e-02 4.16065305e-01 1.08152294e+00 -5.80626845e-01 4.68909740e-01 -6.10791564e-01 2.28758141e-01 1.31192371e-01 3.63873482e-01 3.92986804e-01 -2.33096749e-01 -1.48994744e-01 7.64424741e-01 -4.10389036e-01 -1.10650182e+00 -1.14872074e+00 -7.35802799e-02 7.65121818e-01 6.01212382e-01 -9.47113410e-02 -7.48254955e-01 -1.35768250e-01 -1.21857606e-01 3.05393338e-02 -7.71838546e-01 -1.04583651e-01 -7.62337029e-01 -1.51372516e+00 4.22569960e-01 1.18024960e-01 3.35527599e-01 -1.00059664e+00 -5.93371630e-01 2.16489479e-01 -1.22480206e-01 -8.66718948e-01 -1.90654159e-01 -3.06990743e-01 -8.71967196e-01 -1.29512119e+00 -1.22168911e+00 -6.64256215e-01 1.04405725e+00 5.08453786e-01 6.28027916e-01 6.04837596e-01 -2.32689157e-01 3.75346720e-01 -3.26637238e-01 -1.70352876e-01 -7.05527127e-01 2.97000140e-01 2.66018271e-01 -2.69652903e-01 1.79357752e-01 -5.11819780e-01 -1.11343622e+00 7.45783925e-01 -8.79538834e-01 -1.42745003e-01 2.96156108e-01 1.90177843e-01 -5.00058532e-02 4.01450932e-01 8.04229796e-01 -5.01023412e-01 8.87118399e-01 -1.16011298e+00 -3.90285432e-01 3.84871006e-01 -2.61624694e-01 -4.06881928e-01 5.34194469e-01 -3.51940036e-01 -1.07231021e+00 -3.75349015e-01 -2.80715595e-03 4.08237159e-01 -9.92602110e-02 4.17384773e-01 1.06988505e-01 1.80834293e-01 2.43777990e-01 2.89443016e-01 1.68739051e-01 -2.99190491e-01 7.02668428e-02 8.16342235e-01 -2.03388229e-01 1.51218707e-02 7.55280614e-01 7.40290701e-01 4.54120897e-02 -1.50059199e+00 3.15371454e-01 -4.56004471e-01 -7.57924855e-01 -4.99083251e-01 6.73686087e-01 -7.05492914e-01 -1.05617774e+00 7.74305046e-01 -1.29049265e+00 -3.49465728e-01 1.38727799e-01 5.46233892e-01 -2.09800750e-01 3.37235570e-01 -9.40759242e-01 -1.29500818e+00 -3.51519175e-02 -8.24474275e-01 6.81825101e-01 -1.05833568e-01 -1.93608999e-01 -1.64315498e+00 4.76962239e-01 -1.04104951e-01 7.49990284e-01 3.15184802e-01 1.08408082e+00 -3.02488476e-01 -4.54315364e-01 -7.14350790e-02 -3.09038490e-01 -4.53183860e-01 4.86899912e-01 2.42744058e-01 2.34884359e-02 -2.10805252e-01 -2.37545408e-02 5.68626821e-01 5.32846510e-01 7.19118297e-01 3.92733544e-01 -4.65519220e-01 -9.31311667e-01 2.88235754e-01 1.13052201e+00 4.05493438e-01 5.76651394e-01 3.07456069e-02 5.07767498e-01 9.85383570e-01 1.16840757e-01 7.94214547e-01 6.34519637e-01 5.32780468e-01 1.56071603e-01 -4.46700811e-01 5.25955021e-01 1.08239278e-01 6.27795383e-02 1.21640313e+00 -7.78760910e-01 -6.09170377e-01 -1.39450073e+00 5.86056232e-01 -1.53997958e+00 -1.16143203e+00 -6.01335168e-01 2.10956264e+00 4.19078052e-01 -8.34631398e-02 5.95801651e-01 -1.15185613e-02 9.33838487e-01 1.41779110e-01 -2.58865684e-01 -1.99351206e-01 -1.60778925e-01 -4.42620218e-01 6.14891350e-01 7.71262169e-01 -6.07323110e-01 5.86469769e-01 7.51148701e+00 4.26977217e-01 -1.00107920e+00 1.32521065e-02 6.08428657e-01 2.31016770e-01 -6.57626927e-01 -3.56402397e-01 -5.06649375e-01 5.11330187e-01 8.72138560e-01 -3.10931146e-01 2.28946611e-01 5.67247905e-02 7.00693488e-01 1.02778682e-02 -1.04142574e-03 2.49755517e-01 -3.24266195e-01 -1.00321507e+00 -7.08004786e-03 5.17860353e-01 6.74433172e-01 -1.16868513e-02 -6.30331412e-02 -1.77280962e-01 3.88503373e-01 -8.06405365e-01 -1.70825005e-01 4.74994391e-01 7.30356574e-01 -7.09789634e-01 6.45117819e-01 7.97835648e-01 -1.51382184e+00 6.51659891e-02 -9.23507065e-02 -2.23646641e-01 8.44285488e-01 7.52374232e-01 -6.10230267e-01 4.45756495e-01 4.51350123e-01 5.51664710e-01 -5.40672615e-02 8.58455777e-01 1.59284353e-01 5.21767139e-01 -5.04419267e-01 -1.07349053e-01 1.19629122e-01 -4.70855713e-01 5.54667473e-01 9.87994432e-01 5.79030573e-01 5.13574421e-01 -1.77876726e-01 5.78950644e-01 3.18920642e-01 1.77444384e-01 -9.09547925e-01 2.15286240e-02 3.63209814e-01 7.03860164e-01 -1.09753406e+00 1.55015901e-01 -1.07005849e-01 7.04818964e-01 -1.16426475e-01 7.05596507e-01 -8.53433788e-01 -2.73339540e-01 6.21484101e-01 8.23085964e-01 1.38622783e-02 -6.07633650e-01 5.27706891e-02 -1.08571100e+00 -2.49290228e-01 -1.62298232e-01 1.84798446e-02 -7.14579225e-02 -1.04975736e+00 8.46710920e-01 2.04783440e-01 -9.81851339e-01 -3.96142304e-01 -3.95369798e-01 -9.56602693e-01 5.86726129e-01 -1.13787460e+00 -8.32605660e-01 4.94981594e-02 6.01852775e-01 -6.47575781e-02 4.83655035e-02 5.67680776e-01 2.81665117e-01 -7.47683048e-01 1.57001931e-02 4.72364902e-01 -4.90151271e-02 -4.41342071e-02 -5.94770253e-01 6.00976706e-01 4.23519373e-01 -6.94831192e-01 6.73108041e-01 8.30556929e-01 -1.16404283e+00 -9.68114436e-01 -1.08999002e+00 1.35819066e+00 -3.76115829e-01 9.97918725e-01 -2.88268596e-01 -8.01258981e-01 3.98303539e-01 1.10911213e-01 -2.64874101e-01 6.58505559e-01 -1.12073012e-01 3.06213588e-01 3.60722512e-01 -1.04676485e+00 9.43476856e-01 1.30655670e+00 -1.54840961e-01 -4.36854549e-02 3.36531311e-01 6.07136667e-01 4.95440245e-01 -8.85561824e-01 5.00005364e-01 4.63945150e-01 -7.38216698e-01 8.81752610e-01 -5.26891708e-01 5.70084974e-02 -1.82636902e-02 2.84727335e-01 -1.28202164e+00 -3.46609294e-01 -6.11492515e-01 8.92655477e-02 6.60564065e-01 5.18608391e-01 -9.43332434e-01 6.13650441e-01 1.95102900e-01 7.00594306e-01 -1.02579308e+00 -9.43320572e-01 -7.81417191e-01 3.49147588e-01 -3.35681848e-02 6.48049295e-01 1.06793630e+00 1.78454176e-01 -2.20433678e-02 -5.33793867e-01 2.03655362e-01 4.85743284e-01 -2.98652858e-01 4.84921664e-01 -1.51399493e+00 2.05953613e-01 -3.55783969e-01 -1.66372314e-01 -1.01029670e+00 -2.57503450e-01 -1.52099937e-01 -2.22071901e-01 -1.65054655e+00 1.96314733e-02 -8.07525039e-01 -2.21420322e-02 -2.86327153e-01 1.31162666e-02 -3.28837149e-02 -1.95610225e-01 5.88745892e-01 -5.13642073e-01 6.58924222e-01 1.42396498e+00 3.32596749e-01 -6.89962029e-01 3.29807580e-01 -3.71501803e-01 7.84626663e-01 1.04858184e+00 -4.67348725e-01 -5.43331504e-01 -1.57110408e-01 5.04837751e-01 6.24974728e-01 2.84422278e-01 -5.47105730e-01 3.20828646e-01 -5.88885248e-01 -6.13953993e-02 -3.83784682e-01 1.57264769e-01 -8.38627219e-01 4.82763618e-01 1.01485050e+00 1.10298067e-01 5.56567907e-01 -8.68818611e-02 7.68260241e-01 1.58852547e-01 4.65933442e-01 4.20747310e-01 1.04339242e-01 2.13314110e-04 4.90249932e-01 -1.00924921e+00 -9.88518596e-02 1.46090508e+00 -2.99584061e-01 -5.52110612e-01 -5.69364309e-01 -6.30571127e-01 2.90480584e-01 6.49387002e-01 1.01567373e-01 4.71294641e-01 -1.07628632e+00 -7.97590077e-01 2.33420864e-01 -1.77232951e-01 -4.91249561e-01 6.37303293e-01 1.20879555e+00 -1.00303972e+00 6.02682471e-01 -2.24809214e-01 -2.63488948e-01 -1.02468765e+00 6.28569543e-01 2.05996990e-01 -4.43597466e-01 -2.41443217e-01 3.00652295e-01 2.61020631e-01 -5.09389520e-01 -1.74320236e-01 -1.13737760e-02 -2.64963359e-01 -1.63213927e-02 4.05653059e-01 6.98098540e-01 -5.31583011e-01 -1.09655714e+00 -4.97011721e-01 7.58545935e-01 1.67293891e-01 5.80075569e-02 1.22645962e+00 -5.37572920e-01 -2.66880155e-01 1.24658972e-01 8.90013218e-01 1.01857647e-01 -8.86036932e-01 1.52462244e-01 -1.95675313e-01 -5.19052781e-02 -4.30245668e-01 -3.33578974e-01 -1.10799682e+00 6.14770710e-01 4.49486673e-02 9.69648361e-01 9.54409182e-01 3.07704866e-01 8.26361597e-01 1.35069981e-01 5.01764596e-01 -6.62971020e-01 -2.69701630e-01 3.50653768e-01 6.69430256e-01 -9.35308456e-01 -1.88143894e-01 -7.92085111e-01 -4.63952065e-01 4.68093932e-01 3.70135754e-01 -1.02670461e-01 1.51772499e+00 3.79168093e-01 6.00771718e-02 -2.28340894e-01 -6.92635775e-01 -6.26844913e-02 -2.20046509e-02 7.16618299e-01 3.32910746e-01 2.56515920e-01 -7.31818974e-01 4.61817056e-01 3.34720880e-01 -9.09795463e-02 3.14302146e-01 7.35127866e-01 -6.80488527e-01 -9.62644339e-01 -3.70043218e-01 4.49560881e-01 -2.73610830e-01 -2.89285257e-02 -3.36416483e-01 1.03816211e+00 5.92741631e-02 9.51660633e-01 4.41134125e-01 -1.93423301e-01 2.05370951e-02 -6.29167736e-01 -9.16204788e-03 -1.18918262e-01 -3.51812512e-01 -6.34861067e-02 -3.80557925e-01 5.31402938e-02 -4.55229044e-01 -3.77400577e-01 -1.13431239e+00 -1.06114221e+00 -3.50582987e-01 4.98198271e-01 4.47144479e-01 8.76977623e-01 4.25701261e-01 1.71000689e-01 8.13301027e-01 -6.21296644e-01 5.21312058e-02 -6.89907968e-01 -1.06535256e+00 3.85514051e-02 3.22063327e-01 -4.58054513e-01 -6.14797533e-01 -3.04453105e-01]
[5.955801486968994, 4.436418533325195]
158dd09f-b5b6-4884-b8d1-355c5ef05ec3
efficient-liver-segmentation-with-3d-cnn
2208.13271
null
https://arxiv.org/abs/2208.13271v1
https://arxiv.org/pdf/2208.13271v1.pdf
Efficient liver segmentation with 3D CNN using computed tomography scans
The liver is one of the most critical metabolic organs in vertebrates due to its vital functions in the human body, such as detoxification of the blood from waste products and medications. Liver diseases due to liver tumors are one of the most common mortality reasons around the globe. Hence, detecting liver tumors in the early stages of tumor development is highly required as a critical part of medical treatment. Many imaging modalities can be used as aiding tools to detect liver tumors. Computed tomography (CT) is the most used imaging modality for soft tissue organs such as the liver. This is because it is an invasive modality that can be captured relatively quickly. This paper proposed an efficient automatic liver segmentation framework to detect and segment the liver out of CT abdomen scans using the 3D CNN DeepMedic network model. Segmenting the liver region accurately and then using the segmented liver region as input to tumors segmentation method is adopted by many studies as it reduces the false rates resulted from segmenting abdomen organs as tumors. The proposed 3D CNN DeepMedic model has two pathways of input rather than one pathway, as in the original 3D CNN model. In this paper, the network was supplied with multiple abdomen CT versions, which helped improve the segmentation quality. The proposed model achieved 94.36%, 94.57%, 91.86%, and 93.14% for accuracy, sensitivity, specificity, and Dice similarity score, respectively. The experimental results indicate the applicability of the proposed method.
['Mohammed Sallah', 'Mohammed Elmogy', 'Ahmed Elgarayhi', 'Nabila Eladawi', 'Yasmeen Al-Saeed', 'Khaled Humady']
2022-08-28
null
null
null
null
['liver-segmentation']
['medical']
[-6.35582805e-01 -1.19261369e-01 -2.17561364e-01 -1.48510128e-01 1.50698617e-01 -4.17011648e-01 1.11930780e-01 2.20025599e-01 -4.53077883e-01 5.21593571e-01 1.71053559e-01 -4.13581759e-01 2.45151162e-01 -7.93186784e-01 -2.40871832e-02 -8.94738913e-01 -1.58521980e-01 5.41691303e-01 8.40816647e-02 3.90393734e-01 -2.29539815e-02 1.06594408e+00 -6.83221519e-01 -9.09685045e-02 1.03587198e+00 1.01172280e+00 5.50594144e-02 4.01843071e-01 -2.94460982e-01 4.63327616e-01 -3.03539902e-01 -3.26110013e-02 5.30665040e-01 -8.32569838e-01 -5.37913442e-01 1.90231055e-01 -3.67380083e-01 -6.86949849e-01 -9.41825211e-02 1.09573591e+00 4.95444059e-01 -1.40839055e-01 6.99795067e-01 -8.84723127e-01 -2.09058315e-01 5.84386587e-01 -7.16001153e-01 2.26470426e-01 1.06404331e-02 2.37814263e-01 4.05309984e-04 -7.51666605e-01 1.09846115e-01 8.00845444e-01 5.03064036e-01 4.34130579e-01 -5.62341511e-01 -8.49953294e-01 -6.90370500e-01 -6.40654704e-03 -1.28887069e+00 -6.66207401e-03 2.18920857e-01 -6.49967432e-01 4.90243137e-01 2.25634143e-01 1.28219640e+00 1.66080937e-01 6.94833577e-01 5.42979062e-01 1.14417374e+00 -1.32119343e-01 -4.30757627e-02 8.86814296e-02 -2.39924453e-02 1.14763284e+00 6.10786557e-01 7.40493461e-02 2.71961212e-01 4.41959761e-02 1.14680684e+00 3.81141573e-01 -4.91724283e-01 -2.00148523e-01 -1.29942572e+00 9.11921859e-01 8.84311140e-01 4.57036823e-01 -5.26796520e-01 -2.60995571e-02 5.01818478e-01 -1.56410038e-01 1.32810324e-01 8.21688250e-02 -3.10238332e-01 1.52425617e-01 -8.22187901e-01 -3.77450258e-01 7.64870405e-01 5.82246602e-01 1.12724043e-01 1.39904037e-01 -5.79716973e-02 4.39248830e-01 5.63894808e-01 4.46776986e-01 9.86240387e-01 -3.51173192e-01 -1.56054541e-01 9.52802181e-01 -1.86805859e-01 -7.95722604e-01 -6.76454782e-01 -6.54371560e-01 -1.51563990e+00 -6.94934130e-02 4.47503567e-01 -1.74212918e-01 -1.18994176e+00 1.17328787e+00 6.19726002e-01 -7.56360739e-02 -5.55279776e-02 1.28635371e+00 1.07103992e+00 5.25396705e-01 3.28366399e-01 -1.04536772e-01 1.60958219e+00 -8.86744916e-01 -4.89775956e-01 4.37412262e-01 6.09834731e-01 -9.08550084e-01 5.45920014e-01 3.37177031e-02 -7.41939008e-01 -1.05243586e-01 -7.59692192e-01 2.60257632e-01 -1.80135086e-01 3.63585293e-01 7.96681523e-01 4.58842188e-01 -8.01623225e-01 2.89868236e-01 -1.19528508e+00 -7.15975583e-01 5.19674897e-01 4.51007485e-01 -4.02378172e-01 -1.29897237e-01 -8.93195689e-01 1.02327478e+00 5.84665060e-01 1.41229495e-01 -9.07035828e-01 -5.35354018e-01 -7.51027822e-01 1.76138133e-01 4.04992811e-02 -6.92683339e-01 1.04068959e+00 -5.52611470e-01 -1.39866674e+00 6.97046876e-01 1.55086264e-01 -3.11310589e-01 5.56060255e-01 2.96405286e-01 2.10990191e-01 3.26942593e-01 -8.63799155e-02 5.97200692e-01 -4.69567552e-02 -6.45094395e-01 -4.31500733e-01 -3.54863912e-01 -4.27342445e-01 2.78612942e-01 2.25175709e-01 1.07018605e-01 -2.94527739e-01 -3.77670199e-01 3.18596870e-01 -9.63889480e-01 -2.48045847e-01 3.50876153e-01 -4.72625822e-01 -4.85051982e-02 7.54593074e-01 -9.87425208e-01 6.91704452e-01 -1.86620045e+00 -1.24890558e-01 3.70093316e-01 3.20095539e-01 3.36276323e-01 4.75843728e-01 -5.51831014e-02 -1.12338752e-01 4.39615101e-01 -2.40274921e-01 2.85719544e-01 -3.87763411e-01 4.88111861e-02 7.03523695e-01 8.48589122e-01 -2.41150811e-01 8.19171906e-01 -7.96376348e-01 -7.25870728e-01 4.63265091e-01 6.18103564e-01 -7.27858692e-02 2.68113524e-01 3.08487982e-01 7.38183320e-01 -4.24828649e-01 9.56824958e-01 5.77090263e-01 -3.20032209e-01 -1.42992526e-01 -2.87048638e-01 -1.78844914e-01 -3.96190017e-01 -7.16704071e-01 1.45285571e+00 -8.84257108e-02 3.35839033e-01 1.32489383e-01 -8.96321774e-01 6.80711925e-01 8.87883961e-01 7.59188592e-01 -5.34522831e-01 6.77537739e-01 3.24365854e-01 5.26816785e-01 -8.03437948e-01 -4.83145893e-01 -4.48768616e-01 3.43574613e-01 3.14943582e-01 -2.52677888e-01 -1.95141077e-01 7.13986084e-02 -7.03659803e-02 6.14624977e-01 -2.87393957e-01 9.43107188e-01 -4.33372766e-01 6.47801697e-01 2.23096073e-01 6.78578556e-01 -2.41378583e-02 -5.71190298e-01 3.92844528e-01 2.43976593e-01 -7.45095015e-01 -8.88364017e-01 -7.60800958e-01 -3.05142164e-01 -4.75666970e-02 2.20797718e-01 4.72504318e-01 -7.58961856e-01 -7.29208589e-01 -7.42807016e-02 1.22193828e-01 -4.89847839e-01 8.18638578e-02 -2.70280927e-01 -8.35795999e-01 6.52624726e-01 4.30540472e-01 1.09470761e+00 -8.45531583e-01 -9.63790298e-01 2.61607859e-02 -1.30738586e-01 -5.92962384e-01 -3.31168562e-01 4.93424423e-02 -1.15139103e+00 -1.59888685e+00 -1.22191620e+00 -9.60268259e-01 9.43187177e-01 2.41848767e-01 6.44093037e-01 6.54040456e-01 -6.64463103e-01 -2.05318555e-01 -7.23363608e-02 -4.21133488e-01 -3.88497710e-01 -9.30546969e-02 7.59992078e-02 -3.73677343e-01 1.56332329e-01 -5.88164441e-02 -9.88011658e-01 2.89984614e-01 -7.78812706e-01 3.60890687e-01 9.76797879e-01 7.61854291e-01 5.38555026e-01 8.72022435e-02 2.11011380e-01 -5.96942484e-01 5.45086443e-01 -6.11999035e-01 -6.18443489e-01 1.27321007e-02 -3.79120529e-01 -2.91577131e-01 6.79781735e-01 -2.96120733e-01 -6.61623180e-01 2.96369910e-01 -2.44387332e-02 -3.25195938e-01 -3.02210152e-01 6.58636570e-01 2.18752086e-01 -2.69190043e-01 2.00949937e-01 2.79739082e-01 4.23148483e-01 -1.80214033e-01 -3.00294161e-01 4.99896437e-01 1.81811512e-01 -1.50863856e-01 4.79270369e-01 9.05740708e-02 4.03436720e-01 -7.50972748e-01 -1.94267303e-01 -6.50074482e-01 -5.18565953e-01 -3.40827286e-01 1.17463577e+00 -8.24020565e-01 -8.52020860e-01 4.82711852e-01 -9.38353479e-01 -1.38936877e-01 1.01843797e-01 1.12033892e+00 1.39235899e-01 4.66537774e-01 -1.11251259e+00 -4.26814228e-01 -9.53642547e-01 -1.65455937e+00 1.88907489e-01 6.88998342e-01 4.02002931e-02 -1.13182676e+00 -3.80867869e-01 -8.87130052e-02 8.96830201e-01 6.32201016e-01 1.15242946e+00 -7.79251933e-01 -5.82198203e-01 -4.34249818e-01 -4.18306530e-01 1.46914020e-01 4.01184559e-01 8.79395977e-02 -4.14119661e-01 -1.54097468e-01 1.77617043e-01 7.36294240e-02 4.28348571e-01 6.91918731e-01 1.02940786e+00 -3.91267508e-01 -2.37386346e-01 8.18391860e-01 1.61299300e+00 7.85919011e-01 4.07703161e-01 3.06447782e-02 4.44514871e-01 2.08194882e-01 2.17560947e-01 3.43759596e-01 6.16748258e-02 -9.83175859e-02 7.61931658e-01 -6.45444453e-01 -1.27235085e-01 1.79705083e-01 -1.58100814e-01 1.06745470e+00 -7.27902353e-02 -5.87436184e-02 -1.09263241e+00 5.90693116e-01 -1.23255634e+00 -7.68003047e-01 -4.66591179e-01 2.05099702e+00 7.05188334e-01 -5.19323528e-01 7.44078541e-03 -3.15765589e-02 7.50939846e-01 -7.31989980e-01 -5.13254464e-01 -1.32282078e-01 4.78316009e-01 2.93761075e-01 6.19454980e-01 2.37963915e-01 -8.50183666e-01 3.14628094e-01 5.24609137e+00 1.85348406e-01 -1.53628957e+00 -9.90539789e-02 9.14192379e-01 3.61924052e-01 3.10558617e-01 -2.91422665e-01 -2.52531677e-01 5.87070584e-01 2.59949923e-01 -1.10830143e-01 1.45049721e-01 7.98462331e-01 6.15047872e-01 -5.83864033e-01 -9.08839464e-01 8.65516782e-01 -2.04192311e-01 -1.03829396e+00 -2.27856450e-02 1.58377603e-01 4.50336546e-01 -1.38598710e-01 -1.96471572e-01 -8.59056041e-02 -5.22811152e-02 -1.14484787e+00 4.78977636e-02 3.29430878e-01 8.37823153e-01 -7.37349033e-01 1.52162051e+00 5.58348596e-01 -1.13555288e+00 3.53059977e-01 -1.91713825e-01 1.29213661e-01 -7.23003149e-02 7.25932717e-01 -1.48020518e+00 3.50904316e-01 4.72790271e-01 4.05779511e-01 -3.66084546e-01 1.77019525e+00 -1.96026981e-01 5.57192922e-01 -5.92787206e-01 -1.89107239e-01 2.50671536e-01 -4.20117915e-01 1.45075664e-01 1.05170345e+00 7.15426445e-01 3.73548210e-01 4.29804325e-01 7.73325205e-01 -2.57845074e-01 5.07329762e-01 -5.72175860e-01 2.98741221e-01 2.34735847e-01 1.44735634e+00 -1.33490574e+00 -5.37210643e-01 -2.08653688e-01 7.08373129e-01 -3.58676702e-01 1.32070919e-02 -8.92214775e-01 -3.71387869e-01 1.55938640e-01 1.35339215e-01 -4.58137870e-01 1.05311081e-01 -3.56796622e-01 -1.07681918e+00 -5.86137533e-01 -6.46675348e-01 3.56821269e-01 -4.67037916e-01 -8.88958573e-01 4.40697640e-01 -7.60914907e-02 -1.13917041e+00 1.82257265e-01 -3.89032841e-01 -6.72337294e-01 1.20424688e+00 -1.44866765e+00 -1.01481104e+00 -9.11037326e-01 5.08811653e-01 5.97183526e-01 1.21848494e-01 8.76791179e-01 2.78600812e-01 -7.01404452e-01 2.21089825e-01 -1.76932931e-01 6.79418802e-01 5.35563171e-01 -1.22684669e+00 -5.18631160e-01 9.50884223e-01 -6.98953390e-01 7.59245157e-01 1.94685400e-01 -7.21022308e-01 -1.16778719e+00 -1.06223047e+00 5.42756855e-01 4.79769319e-01 -8.21264368e-03 3.14966410e-01 -5.60052633e-01 4.78378505e-01 4.41611618e-01 3.34708691e-01 8.51714253e-01 -8.26974690e-01 4.62395340e-01 7.24602938e-02 -1.69949913e+00 3.61596465e-01 -1.42771649e-04 1.46457359e-01 -2.19867289e-01 3.15362275e-01 2.35220313e-01 -7.96285391e-01 -1.07990229e+00 4.20176983e-01 6.62998021e-01 -9.02292848e-01 7.47353852e-01 -8.53685588e-02 3.87462616e-01 -4.63864774e-01 3.61922354e-01 -1.40926421e+00 -2.39227816e-01 1.73644900e-01 4.78085242e-02 7.13120103e-01 2.40385979e-01 -6.47228956e-01 8.80492032e-01 5.86098015e-01 -2.42060632e-01 -9.81130838e-01 -5.84141374e-01 -2.76716650e-01 7.35499412e-02 4.05526608e-01 5.55238664e-01 1.08231926e+00 -1.34217337e-01 -2.81119570e-02 2.07489341e-01 9.01231393e-02 8.20336401e-01 -1.75136268e-01 4.42174613e-01 -1.24332416e+00 2.97097951e-01 -5.14650941e-01 -4.19880331e-01 -6.33485198e-01 -5.10290802e-01 -1.04548395e+00 -6.85158670e-02 -2.00175643e+00 5.11640370e-01 -5.16178727e-01 -1.40226930e-01 7.34684467e-01 3.77969220e-02 2.45091289e-01 1.10428549e-01 2.81256050e-01 1.76280603e-01 1.20644376e-01 1.68114388e+00 -1.71430305e-01 -1.21563934e-01 7.25978613e-02 -3.99587333e-01 7.14633763e-01 1.22726095e+00 -4.18916911e-01 -2.22149178e-01 -3.15416187e-01 -4.40270096e-01 4.92460042e-01 1.14605129e-01 -1.02404726e+00 4.11727756e-01 -2.51747906e-01 1.14156795e+00 -6.14537299e-01 -2.14868575e-01 -1.21732926e+00 5.47783315e-01 1.42897987e+00 5.44981174e-02 1.68576673e-01 1.39671013e-01 5.41434661e-02 -2.10491031e-01 -1.76858649e-01 1.29071379e+00 -8.31997454e-01 -3.52225542e-01 5.23832262e-01 -6.13095582e-01 -3.75323147e-01 1.53728676e+00 -1.88733637e-01 1.88542493e-02 -2.39019677e-01 -5.72995067e-01 3.34254146e-01 2.99296081e-01 -2.94113249e-01 7.94610679e-01 -1.15856314e+00 -6.53411567e-01 2.46070608e-01 -2.29122296e-01 5.07358491e-01 2.82543097e-02 1.53265071e+00 -1.53544474e+00 5.86170495e-01 -3.25670451e-01 -6.94303751e-01 -1.30206716e+00 3.45589250e-01 7.92045295e-01 -3.40312034e-01 -5.34695268e-01 6.05583906e-01 6.26711324e-02 -1.50180638e-01 2.17281669e-01 -8.34056675e-01 -1.40160754e-01 -3.19780380e-01 2.57761478e-01 3.35442930e-01 6.34339452e-02 -4.34237808e-01 -3.70632172e-01 5.32024801e-01 1.49550319e-01 3.49995911e-01 1.03527498e+00 2.15670150e-02 -5.47573745e-01 -1.28457531e-01 9.72416461e-01 -1.60046756e-01 -5.41988552e-01 8.40142593e-02 -3.77816558e-01 -3.66135210e-01 2.24097282e-01 -1.33137012e+00 -1.42988396e+00 1.03448677e+00 8.58341813e-01 4.17157682e-03 1.02805042e+00 -2.82587558e-01 8.67393076e-01 -1.24447644e-01 3.09230804e-01 -2.87768573e-01 -2.89954275e-01 2.10787401e-01 5.06048679e-01 -1.39535499e+00 2.14038208e-01 -3.36718947e-01 -4.20382500e-01 1.48390818e+00 7.17938900e-01 -6.09530695e-02 6.74084723e-01 2.15037346e-01 3.25239122e-01 -1.92824632e-01 -5.08356169e-02 7.70326331e-02 2.92971227e-02 2.36212164e-01 7.18341410e-01 1.76669657e-01 -6.31941438e-01 3.44631046e-01 2.57370353e-01 1.32436782e-01 4.34932709e-01 9.16097045e-01 -5.44216752e-01 -5.12658477e-01 -4.71034586e-01 6.86779857e-01 -9.63662028e-01 -3.56759019e-02 -2.00989306e-01 1.12880290e+00 1.71608597e-01 6.29662156e-01 -2.01067597e-01 5.92991523e-02 -9.57132131e-02 9.98905897e-02 2.30177343e-01 -4.77182955e-01 -7.95794487e-01 1.29487574e-01 -5.55305064e-01 -2.45460972e-01 -3.83823276e-01 -1.37684584e-01 -1.63208210e+00 -4.05466259e-01 -4.40798283e-01 4.09237504e-01 1.05985832e+00 7.91871846e-01 3.55465487e-02 4.37568456e-01 5.21066070e-01 -5.80558479e-01 -4.21397418e-01 -1.13935006e+00 -5.03734291e-01 3.00414175e-01 2.26288766e-01 -4.74476010e-01 -3.16056371e-01 7.41376653e-02]
[14.493050575256348, -2.7156295776367188]
3d9fd0fc-13e4-4d28-8257-2a88666cee36
robust-face-alignment-using-a-mixture-of
1511.04404
null
http://arxiv.org/abs/1511.04404v2
http://arxiv.org/pdf/1511.04404v2.pdf
Robust Face Alignment Using a Mixture of Invariant Experts
Face alignment, which is the task of finding the locations of a set of facial landmark points in an image of a face, is useful in widespread application areas. Face alignment is particularly challenging when there are large variations in pose (in-plane and out-of-plane rotations) and facial expression. To address this issue, we propose a cascade in which each stage consists of a mixture of regression experts. Each expert learns a customized regression model that is specialized to a different subset of the joint space of pose and expressions. The system is invariant to a predefined class of transformations (e.g., affine), because the input is transformed to match each expert's prototype shape before the regression is applied. We also present a method to include deformation constraints within the discriminative alignment framework, which makes our algorithm more robust. Our algorithm significantly outperforms previous methods on publicly available face alignment datasets.
['Tim K. Marks', 'Salil Tambe', 'Oncel Tuzel']
2015-11-13
null
null
null
null
['robust-face-alignment']
['computer-vision']
[ 8.87231752e-02 -1.99576840e-01 -3.32509995e-01 -7.95711696e-01 -5.29234111e-01 -6.66448295e-01 4.86721188e-01 -5.33275962e-01 -2.52944201e-01 1.43578231e-01 -4.15758137e-03 3.52959603e-01 2.16470286e-01 -4.04994190e-01 -6.99581981e-01 -7.38376796e-01 2.80603796e-01 4.98715907e-01 -1.78339064e-01 -2.37774640e-01 2.50412434e-01 1.04352844e+00 -1.27967072e+00 -1.96371019e-01 3.83343309e-01 8.95943940e-01 -2.40106791e-01 3.41559201e-01 2.84262985e-01 7.65851066e-02 -4.10397500e-01 -5.77842593e-01 5.59836268e-01 -3.29160333e-01 -5.92843831e-01 3.82350773e-01 1.13316131e+00 -2.12735519e-01 -1.31642982e-01 1.27719545e+00 5.54043710e-01 1.34526893e-01 5.93607128e-01 -1.48552072e+00 -3.09914142e-01 -1.20815597e-01 -9.39990699e-01 -1.92632064e-01 4.09554005e-01 -1.59142673e-01 8.36741507e-01 -1.27449751e+00 5.57315350e-01 1.28880775e+00 6.49673104e-01 7.13187099e-01 -1.33107126e+00 -8.22342336e-01 2.93479145e-01 5.71396351e-02 -1.62442172e+00 -8.03385973e-01 9.90944386e-01 -4.42343771e-01 5.53202271e-01 1.13258615e-01 5.27290285e-01 8.41515541e-01 2.51676857e-01 2.19599485e-01 7.86676407e-01 -3.52727681e-01 4.51259091e-02 -2.28277713e-01 -2.99581826e-01 6.93947732e-01 -2.03358885e-02 -2.31392652e-01 -5.47907889e-01 -3.13717604e-01 1.08091414e+00 2.98569910e-02 -7.38270432e-02 -6.50134146e-01 -8.55889022e-01 7.05148041e-01 3.06579381e-01 1.41905308e-01 -2.80190140e-01 3.76232341e-02 -9.01435688e-02 2.30337679e-01 3.91984493e-01 3.14746976e-01 -4.45027858e-01 3.52138102e-01 -1.01795852e+00 2.79496908e-01 6.69129848e-01 1.00328219e+00 1.06351948e+00 -1.89124944e-03 5.79114854e-02 8.75694633e-01 5.45969725e-01 6.76161945e-01 2.65303373e-01 -1.05876291e+00 3.76164258e-01 4.70313698e-01 5.67656802e-03 -1.35774779e+00 -4.26276475e-01 -1.94171265e-01 -7.33149350e-01 2.13465765e-01 4.35897112e-01 -9.67229903e-02 -7.89628327e-01 2.06879878e+00 7.54298389e-01 4.88269478e-01 -3.39718699e-01 9.79651749e-01 6.07829750e-01 2.40571648e-01 -1.37497097e-01 -3.50453198e-01 1.26599169e+00 -9.34478223e-01 -5.79039574e-01 -4.59105343e-01 3.94944027e-02 -1.11507702e+00 6.62913382e-01 1.33077260e-02 -9.69388127e-01 -5.28758228e-01 -8.04275751e-01 -9.74520966e-02 -9.83862877e-02 3.85502994e-01 2.76237667e-01 4.23862398e-01 -1.25773489e+00 3.44060987e-01 -7.85365582e-01 -3.06032270e-01 3.32233131e-01 9.07963693e-01 -8.97357941e-01 2.01224133e-01 -6.48870170e-01 9.38427150e-01 -3.46450299e-01 4.92392093e-01 -5.10976374e-01 -4.37100202e-01 -1.00318551e+00 -2.13848054e-01 3.66382748e-01 -5.92032254e-01 1.25499249e+00 -1.38783920e+00 -1.79454064e+00 1.21528697e+00 -6.78933799e-01 4.14578438e-01 5.04164994e-01 -2.96975970e-01 -2.29874656e-01 -1.05171725e-02 6.78931326e-02 6.94564342e-01 1.51079535e+00 -1.03908193e+00 -1.33528218e-01 -8.74611616e-01 -2.41679862e-01 1.65687874e-01 -2.72731990e-01 5.15661776e-01 -9.96814489e-01 -5.86989820e-01 3.06511939e-01 -1.30625081e+00 -3.37254882e-01 2.32017532e-01 -3.49529862e-01 -1.36908129e-01 8.61799955e-01 -7.01844692e-01 9.34053063e-01 -2.02432919e+00 4.68741864e-01 6.73254669e-01 5.40517047e-02 -2.68052947e-02 -3.03545326e-01 1.07749980e-02 -3.83967727e-01 -1.49251878e-01 -4.59218062e-02 -4.79093194e-01 -2.80105591e-01 1.20500907e-01 -1.57748669e-01 8.94645452e-01 4.07713145e-01 7.57598519e-01 -5.88837802e-01 -4.74837661e-01 -8.90480876e-02 5.70448220e-01 -6.41224384e-01 3.33272278e-01 3.25737819e-02 8.50920439e-01 -5.72393239e-01 7.36291528e-01 8.12542796e-01 4.83615510e-02 2.52991229e-01 -3.84617716e-01 6.00938424e-02 -5.33789098e-02 -1.23599517e+00 1.64178061e+00 -2.87910789e-01 4.20755833e-01 5.99157661e-02 -8.44079733e-01 9.87380922e-01 3.87537301e-01 7.51934230e-01 -1.96829110e-01 3.28860283e-01 3.64767872e-02 -4.26512808e-02 -2.54048169e-01 1.45385966e-01 2.04876587e-01 1.69210866e-01 5.22151113e-01 2.15543285e-01 -2.36536205e-01 -7.03945979e-02 -3.36840868e-01 5.68652332e-01 2.67614007e-01 5.70134699e-01 -2.07004443e-01 6.78787649e-01 -4.89948034e-01 8.50634336e-01 1.05373114e-01 -2.06857651e-01 6.98062837e-01 4.66784805e-01 -6.02085710e-01 -1.14359891e+00 -7.31393933e-01 -1.03796102e-01 9.85738218e-01 1.03414588e-01 -3.70351613e-01 -1.02169120e+00 -7.91044295e-01 1.14210054e-01 -1.67790353e-01 -7.27071226e-01 -7.59579018e-02 -9.53118801e-01 -5.15092075e-01 3.06692332e-01 6.35594308e-01 3.59253705e-01 -7.41449118e-01 -2.63696522e-01 -1.42854556e-01 -5.28215722e-04 -1.19594073e+00 -1.07218957e+00 -3.34595919e-01 -7.40900278e-01 -1.19398212e+00 -5.50551534e-01 -8.88433218e-01 1.47709990e+00 2.34611511e-01 8.63349438e-01 1.70642838e-01 -1.80464953e-01 5.46187103e-01 6.40605241e-02 -3.54039431e-01 -7.45987147e-02 1.98029038e-02 3.65903974e-01 5.10274470e-01 3.98720056e-01 -5.09002626e-01 -5.39483070e-01 8.01512659e-01 -4.92649108e-01 -2.29356289e-01 1.71445906e-01 7.80040801e-01 7.04966128e-01 -3.61991316e-01 1.82845771e-01 -6.25760615e-01 3.03611875e-01 -1.29537493e-01 -8.10372233e-01 3.59169990e-01 -1.39642894e-01 -1.20811060e-01 4.28487241e-01 -6.01646245e-01 -7.14598358e-01 7.27720916e-01 -9.97142959e-03 -6.57549024e-01 8.78112391e-02 1.66088104e-01 -5.14687121e-01 -7.29830265e-01 4.39656436e-01 -2.83900142e-01 2.53524214e-01 -2.72436500e-01 2.79668957e-01 3.50503981e-01 5.95391512e-01 -6.79239213e-01 1.22344124e+00 3.56407255e-01 3.19674909e-01 -7.49159157e-01 -6.47859633e-01 -3.99542898e-01 -1.43800163e+00 -3.53897303e-01 6.23959184e-01 -7.94645190e-01 -7.13973999e-01 5.61056852e-01 -1.26403356e+00 -5.35805128e-04 1.95582554e-01 3.85976255e-01 -5.93235433e-01 3.24217439e-01 -1.59053758e-01 -4.48096812e-01 -2.21526906e-01 -1.49881470e+00 1.40278459e+00 5.03284276e-01 -4.61173713e-01 -7.95053124e-01 4.08687927e-02 6.34599850e-02 1.47171006e-01 2.13403136e-01 5.46527088e-01 -5.53200126e-01 -2.81835049e-01 -3.55544299e-01 2.41413310e-01 3.10799509e-01 4.86337066e-01 5.48546374e-01 -8.97358298e-01 -4.95425552e-01 7.08465874e-02 -1.60235479e-01 3.53347898e-01 2.67521083e-01 1.27329767e+00 -3.83801550e-01 -3.79155219e-01 8.93695176e-01 9.33315337e-01 3.16915475e-02 4.25210029e-01 1.32808223e-01 7.84903944e-01 6.69499934e-01 5.47019005e-01 2.92128325e-01 2.38972649e-01 1.20612574e+00 2.30585873e-01 -3.23443621e-01 2.04847395e-01 -1.00599371e-01 3.73619407e-01 5.80017626e-01 -4.64927793e-01 3.80545080e-01 -7.21821606e-01 2.28332933e-02 -1.72662759e+00 -8.66577983e-01 4.63318557e-01 2.39192653e+00 8.51499021e-01 -5.66663325e-01 2.61416942e-01 -1.46689191e-01 9.18240607e-01 1.31233007e-01 -6.37398064e-01 -2.33608097e-01 6.71014339e-02 3.82134467e-01 2.34540731e-01 6.04611933e-01 -1.22194195e+00 1.26299047e+00 6.96283627e+00 2.38377228e-01 -1.56780601e+00 -2.22987339e-01 6.12950504e-01 1.00433469e-01 1.19487904e-02 -9.81353223e-02 -1.04798353e+00 2.28626743e-01 2.88024426e-01 -9.58336424e-03 4.56904590e-01 9.23606336e-01 6.55493140e-02 2.40630254e-01 -1.31116295e+00 1.01770890e+00 5.38525283e-01 -8.62030864e-01 -1.37706220e-01 5.44665791e-02 8.08772862e-01 -2.90694863e-01 2.90258348e-01 -2.25893795e-01 7.77270347e-02 -1.21242702e+00 4.79242623e-01 3.74025941e-01 9.41920459e-01 -7.52646148e-01 4.08303201e-01 8.84182900e-02 -1.14054132e+00 1.95045605e-01 -3.68584841e-01 2.80763626e-01 -1.22871742e-01 1.06886484e-01 -8.47842872e-01 2.07008854e-01 5.80884278e-01 6.63165569e-01 -4.14412647e-01 7.55695939e-01 -4.59655136e-01 1.49413183e-01 -4.08982247e-01 4.82808322e-01 -2.28233770e-01 -5.61085522e-01 3.40650171e-01 8.09896708e-01 5.15412807e-01 -3.70242894e-02 3.42783719e-01 5.06015420e-01 -2.65190452e-01 3.36399108e-01 -5.46934307e-01 3.13023269e-01 5.46798050e-01 1.63241136e+00 -4.82827306e-01 1.68825954e-01 -5.29206157e-01 1.02149856e+00 4.50522810e-01 3.86980027e-01 -8.23062479e-01 -1.40599329e-02 9.70729053e-01 2.28059459e-02 2.54268795e-01 -3.19118977e-01 4.93747555e-02 -1.10686207e+00 2.46750772e-01 -1.17910624e+00 1.89816698e-01 -5.37707508e-01 -1.15817130e+00 6.49770319e-01 -8.71528015e-02 -1.12892079e+00 -4.78565484e-01 -6.37821198e-01 -8.33852232e-01 9.75987613e-01 -1.34729230e+00 -1.43079472e+00 -5.21995246e-01 8.86934817e-01 1.57500029e-01 -3.19182903e-01 9.06330407e-01 1.28524244e-01 -7.34972477e-01 1.02015853e+00 -2.85307825e-01 4.33238268e-01 1.16913915e+00 -6.91298485e-01 4.75507766e-01 7.21192777e-01 3.59467238e-01 9.91034746e-01 5.80170214e-01 -5.32875538e-01 -1.45578086e+00 -1.10449648e+00 9.44744766e-01 -5.35816967e-01 3.92092228e-01 -4.24255580e-01 -7.49858916e-01 1.14581299e+00 -1.45514995e-01 4.17315304e-01 6.89148307e-01 1.64168049e-02 -6.56600475e-01 -4.07152712e-01 -1.20008767e+00 8.19514453e-01 9.72330928e-01 -5.41230798e-01 -2.33524427e-01 1.77306250e-01 3.98215167e-02 -9.00301218e-01 -8.29772532e-01 5.73518693e-01 8.84756267e-01 -5.41192353e-01 9.78860259e-01 -7.39393532e-01 1.37207836e-01 -5.44275403e-01 -5.33249937e-02 -1.31542861e+00 -3.42708260e-01 -8.75341654e-01 1.06522888e-01 1.18423736e+00 1.16232887e-01 -6.53883636e-01 7.55798697e-01 6.61773026e-01 2.71156698e-01 -8.07996690e-01 -1.11308539e+00 -3.34834725e-01 -1.40058041e-01 9.92441624e-02 8.42295051e-01 9.74169672e-01 -2.65660852e-01 2.33557135e-01 -4.26474899e-01 4.66431528e-01 3.06614369e-01 1.56012699e-01 1.29817629e+00 -1.21908438e+00 4.62886542e-02 -4.53636765e-01 -7.58358479e-01 -9.52362835e-01 7.61130929e-01 -7.84128964e-01 7.15459371e-03 -6.45498991e-01 1.83366403e-01 -2.61545539e-01 -1.04798777e-02 8.24915707e-01 -2.24644199e-01 4.45980400e-01 1.04740463e-01 3.42620522e-01 -1.17404275e-01 3.76584023e-01 1.25595760e+00 4.15431801e-03 -8.26510340e-02 3.74260247e-01 -6.14754558e-01 8.97684216e-01 7.03595817e-01 -4.14904982e-01 -1.96412951e-01 -4.45998460e-01 -4.07328680e-02 -2.29663312e-01 1.26771986e-01 -5.49977303e-01 3.86215806e-01 -4.06257123e-01 8.89213622e-01 -2.76134610e-01 5.25826156e-01 -9.38139021e-01 1.88144386e-01 1.26057357e-01 -1.02736220e-01 4.36056525e-01 1.19943544e-01 1.14425145e-01 -2.84000844e-01 -3.36937569e-02 1.04606640e+00 1.59888685e-01 -3.65732849e-01 8.61461103e-01 2.11752504e-01 -2.54760146e-01 1.19977832e+00 -3.05206656e-01 7.91231319e-02 -4.31050599e-01 -4.83022094e-01 1.75334871e-01 7.32266843e-01 5.38445771e-01 4.19944793e-01 -1.57320213e+00 -6.33091688e-01 6.07800961e-01 1.19145110e-01 1.56207904e-01 -1.16837062e-02 9.27853107e-01 -2.77611375e-01 7.45094121e-02 -4.30587381e-01 -7.76206672e-01 -1.82035851e+00 2.04128176e-01 6.21779799e-01 7.87553489e-02 -2.65942186e-01 8.55547428e-01 1.94651559e-01 -5.20170093e-01 1.63585350e-01 1.24933474e-01 -3.63857001e-01 -1.69928446e-02 4.83744204e-01 9.03049260e-02 6.47037700e-02 -1.44684863e+00 -4.99101937e-01 1.49551833e+00 -6.58613443e-02 1.49707664e-02 1.18409586e+00 1.92744255e-01 -4.85794932e-01 -9.02428012e-03 1.24816024e+00 4.39317286e-01 -1.35329676e+00 -2.91465729e-01 -8.42057243e-02 -6.59882128e-01 -2.98963338e-01 -3.06357056e-01 -1.47273386e+00 6.20894194e-01 3.33062977e-01 -6.30147398e-01 1.17697895e+00 -3.10927555e-02 3.20900291e-01 3.24172199e-01 3.42516780e-01 -9.86446440e-01 1.83196083e-01 4.63242173e-01 1.22023666e+00 -1.23852015e+00 1.85513139e-01 -6.67723954e-01 -4.99450088e-01 1.33226752e+00 8.49522710e-01 -1.58574030e-01 7.38065779e-01 2.66497582e-01 3.40561271e-01 -1.89017672e-02 -3.77232134e-01 1.35633752e-01 6.92018330e-01 5.50865054e-01 5.55500507e-01 -1.05536267e-01 -6.42700270e-02 3.76783967e-01 -4.43256319e-01 -3.22961479e-01 -1.03391804e-01 7.65032709e-01 -1.12966023e-01 -1.48944974e+00 -5.79946160e-01 -1.37316529e-02 -3.96514028e-01 3.00988615e-01 -6.14964604e-01 7.43735969e-01 1.91070922e-02 6.12487137e-01 2.22124159e-01 -2.62066275e-01 3.48261267e-01 1.18796363e-01 8.66067290e-01 -5.80098927e-01 -4.33699787e-01 3.83734584e-01 -4.76581246e-01 -7.97584891e-01 -5.69881618e-01 -8.43056381e-01 -1.12830412e+00 -2.19743341e-01 -2.96438515e-01 -1.32039741e-01 7.38951623e-01 9.59173381e-01 3.58883142e-01 -9.07791704e-02 1.10659862e+00 -1.06558347e+00 -6.04973257e-01 -7.61638284e-01 -3.80170435e-01 4.91109669e-01 3.00267249e-01 -8.46821129e-01 -1.64393410e-01 2.74065346e-01]
[13.358819007873535, 0.29944318532943726]
f0d53d93-219c-4b9c-8cb7-206dda7ef046
fine-grained-noise-control-for-multispeaker
2204.0507
null
https://arxiv.org/abs/2204.05070v2
https://arxiv.org/pdf/2204.05070v2.pdf
Fine-grained Noise Control for Multispeaker Speech Synthesis
A text-to-speech (TTS) model typically factorizes speech attributes such as content, speaker and prosody into disentangled representations.Recent works aim to additionally model the acoustic conditions explicitly, in order to disentangle the primary speech factors, i.e. linguistic content, prosody and timbre from any residual factors, such as recording conditions and background noise.This paper proposes unsupervised, interpretable and fine-grained noise and prosody modeling. We incorporate adversarial training, representation bottleneck and utterance-to-frame modeling in order to learn frame-level noise representations. To the same end, we perform fine-grained prosody modeling via a Fully Hierarchical Variational AutoEncoder (FVAE) which additionally results in more expressive speech synthesis.
['Pirros Tsiakoulis', 'Aimilios Chalamandaris', 'Gunu Jho', 'June Sig Sung', 'Spyros Raptis', 'Konstantinos Markopoulos', 'Konstantinos Klapsas', 'Nikolaos Ellinas', 'Georgios Vamvoukakis', 'Karolos Nikitaras']
2022-04-11
null
null
null
null
['expressive-speech-synthesis']
['speech']
[-6.73372820e-02 1.80620208e-01 5.54400356e-03 -3.16083491e-01 -9.63603020e-01 -5.64922869e-01 6.86419904e-01 -1.59072995e-01 -3.75656858e-02 6.33541346e-01 9.88098145e-01 -1.17419057e-01 2.77480841e-01 -3.98724645e-01 -6.07497752e-01 -8.30910385e-01 3.43977243e-01 1.11379586e-01 -4.04819280e-01 -2.13821709e-01 -4.96418357e-01 4.11865950e-01 -1.57819629e+00 1.68700635e-01 6.42188728e-01 6.22205675e-01 6.51221797e-02 1.16171110e+00 -6.31479546e-02 1.15844071e+00 -8.59959900e-01 -3.43287677e-01 -5.65796010e-02 -5.00760019e-01 -1.61236614e-01 1.99317053e-01 9.69358534e-02 -4.67860222e-01 -5.69654822e-01 1.02151036e+00 4.76138651e-01 3.28227192e-01 7.16199398e-01 -8.91073823e-01 -5.58143020e-01 9.22543764e-01 -1.12786844e-01 -3.68244797e-02 1.45829335e-01 3.01181585e-01 1.13500643e+00 -8.06539714e-01 2.86423177e-01 1.57253933e+00 3.51657510e-01 5.84920228e-01 -1.77061081e+00 -6.11354589e-01 1.04250431e-01 -3.54229510e-02 -1.18026423e+00 -8.15153360e-01 1.24373865e+00 -5.67868173e-01 7.23728836e-01 4.63255376e-01 3.79248589e-01 1.80730307e+00 -9.45636109e-02 7.37864256e-01 9.37909186e-01 -6.49031460e-01 1.88866228e-01 5.88958040e-02 9.72827058e-03 9.82515365e-02 -4.12401855e-01 4.28190768e-01 -3.74255568e-01 -7.51063004e-02 7.20794797e-01 -2.15717927e-01 -2.46265143e-01 9.65166558e-03 -8.23889852e-01 7.03128278e-01 -1.41610876e-01 3.09820950e-01 -5.59594810e-01 3.41673046e-01 5.30188799e-01 1.76393688e-01 4.63086367e-01 2.21225709e-01 -4.96296674e-01 -2.00738281e-01 -9.33675110e-01 3.23009044e-01 7.98657835e-01 7.47022986e-01 7.19664156e-01 1.08931696e+00 -4.11124557e-01 1.14915812e+00 6.01138532e-01 6.95073605e-01 5.49949229e-01 -1.05013013e+00 3.10884565e-01 -1.57712176e-01 -2.17799637e-02 -5.57243764e-01 -4.40783687e-02 -3.90367776e-01 -8.45651686e-01 4.64729369e-01 -1.75909791e-02 -3.00835133e-01 -1.15869296e+00 2.21291947e+00 1.21076331e-01 4.51501071e-01 1.72890335e-01 8.83669138e-01 6.41087294e-01 9.15275693e-01 3.55474323e-01 -4.82332945e-01 1.45164096e+00 -9.32465553e-01 -1.45215297e+00 -1.37307554e-01 -1.00363031e-01 -9.94924366e-01 1.06187856e+00 3.43582481e-01 -1.41379714e+00 -1.10996175e+00 -1.07499778e+00 -2.58480281e-01 -1.09020181e-01 1.26126677e-01 -8.11011642e-02 8.05949926e-01 -6.34880900e-01 3.96699637e-01 -9.72582698e-01 4.46138173e-01 -2.10507199e-01 2.21429840e-01 -7.96887204e-02 4.37934667e-01 -1.53560078e+00 7.79814780e-01 3.73440862e-01 -7.18381852e-02 -1.30394447e+00 -8.78167689e-01 -1.12089884e+00 2.68283248e-01 3.85795206e-01 -7.11242259e-01 1.34682834e+00 -1.04367721e+00 -2.13337350e+00 3.88790548e-01 -2.75286257e-01 -5.16013384e-01 2.64451444e-01 -4.00538534e-01 -6.78836405e-01 -8.96934196e-02 -5.64637303e-01 2.13705167e-01 1.55650723e+00 -1.28514969e+00 -6.81456029e-02 -2.06781819e-01 -2.42710963e-01 2.98799127e-01 -1.46826088e-01 1.41242698e-01 -3.54916930e-01 -1.26727259e+00 -2.38767490e-01 -7.29302168e-01 -1.20920287e-02 -4.66108859e-01 -4.22080129e-01 8.74883961e-03 7.77608514e-01 -1.13421392e+00 1.21590304e+00 -2.31610155e+00 8.58852208e-01 -2.48562202e-01 2.66922683e-01 3.16502243e-01 1.86700858e-02 3.33228201e-01 -2.27541149e-01 4.75879572e-02 1.84816152e-01 -1.14413631e+00 2.90364176e-01 5.82932472e-01 -4.75199372e-01 1.80603921e-01 3.20279360e-01 7.62845397e-01 -5.77369153e-01 -2.48642594e-01 8.41935635e-01 8.92327666e-01 -7.64265001e-01 6.30092859e-01 -3.90515536e-01 8.27836335e-01 -1.59495533e-01 1.11662157e-01 4.68546569e-01 6.76498473e-01 3.40934703e-03 -4.60756183e-01 -7.03005195e-02 5.06985426e-01 -1.08127904e+00 1.58257687e+00 -8.41189742e-01 5.29189765e-01 6.89425528e-01 -6.18396521e-01 9.24465597e-01 9.32025313e-01 3.12783539e-01 -3.55944604e-01 2.85275280e-01 -1.43566346e-02 3.31917219e-02 -3.42941195e-01 5.47563136e-01 -5.99090755e-01 -4.71442342e-02 -1.69072270e-01 5.97259700e-01 -4.38808233e-01 -4.42890733e-01 -5.27400374e-02 5.81068218e-01 1.70176774e-01 3.49302351e-01 -1.05062887e-01 5.53535223e-01 -8.00015688e-01 6.37261868e-01 4.17550445e-01 -2.46072128e-01 8.06897819e-01 2.53775239e-01 1.00248486e-01 -1.43587458e+00 -1.38157570e+00 6.53558746e-02 1.19683158e+00 -4.47169602e-01 -3.95095974e-01 -9.83975947e-01 6.40238300e-02 -1.83068797e-01 1.28882444e+00 -6.19803488e-01 -2.29591534e-01 -6.90517426e-01 -7.22012967e-02 8.21408689e-01 4.86413658e-01 -3.01924884e-01 -9.35187638e-01 9.89917442e-02 4.73233253e-01 -3.07664536e-02 -1.22589767e+00 -6.54157996e-01 4.01362151e-01 -4.09902722e-01 -4.12939847e-01 -6.08652592e-01 -1.68987513e-01 -2.20098272e-01 -3.90756220e-01 9.41864133e-01 -5.81870019e-01 -9.79232043e-02 1.77416086e-01 -3.93893212e-01 -4.94995028e-01 -1.08806705e+00 -3.76504600e-01 3.82352710e-01 2.60885477e-01 -5.26410490e-02 -8.76424015e-01 -3.26180249e-01 8.57635662e-02 -1.01735258e+00 2.81871110e-02 2.08875939e-01 9.25811470e-01 7.23471403e-01 3.10123228e-02 5.33699274e-01 -7.07935035e-01 6.61019027e-01 -4.20944661e-01 -4.32746738e-01 -6.77965879e-02 -1.13484107e-01 1.10488176e-01 1.00475001e+00 -6.69361353e-01 -1.54063475e+00 -1.86823860e-01 -7.55118668e-01 -1.17093396e+00 -6.69979393e-01 1.78141236e-01 -7.50320673e-01 6.85042381e-01 7.40455925e-01 2.63566613e-01 -1.44176871e-01 -5.94156802e-01 8.06834519e-01 6.62902296e-01 7.34605849e-01 -7.46240675e-01 8.76882374e-01 1.02810524e-01 -4.69469100e-01 -1.13426232e+00 -7.09464312e-01 -2.51599997e-01 -5.65464139e-01 1.72381520e-01 1.15778804e+00 -1.30446553e+00 -4.77171868e-01 2.87940979e-01 -1.30210638e+00 -2.09824830e-01 -7.37741947e-01 7.71414042e-01 -9.21603560e-01 1.94139183e-01 -8.58709514e-01 -1.31820190e+00 -1.31596178e-01 -1.31733632e+00 1.12306476e+00 -1.40526826e-02 -3.11401933e-01 -9.29170847e-01 1.11342020e-01 5.47584653e-01 3.98866981e-01 3.30349624e-01 7.93597758e-01 -6.33433700e-01 -3.40089858e-01 1.56885713e-01 3.35268885e-01 9.22731221e-01 2.23567858e-01 1.84293926e-01 -1.47954834e+00 -3.36952582e-02 4.81492728e-01 -1.76335379e-01 5.63090503e-01 7.14714885e-01 7.77692378e-01 -4.84520286e-01 3.64629805e-01 6.93041027e-01 9.05221760e-01 3.82344723e-01 7.04461575e-01 -4.57849681e-01 9.52233374e-01 4.98414725e-01 1.64695978e-02 4.80870634e-01 3.79186757e-02 7.76996613e-01 2.73409933e-01 -1.05564728e-01 -6.34856164e-01 -5.28586984e-01 4.24267173e-01 1.36501265e+00 2.02317443e-03 -4.59824532e-01 -3.58629227e-01 3.56785178e-01 -1.40558267e+00 -9.84330952e-01 7.30646700e-02 1.91398597e+00 7.80495465e-01 1.40432805e-01 5.42530492e-02 3.99520278e-01 6.51184499e-01 5.09701550e-01 -5.78861475e-01 -6.27787769e-01 -2.65878290e-01 3.43416274e-01 9.38119739e-02 9.24130380e-01 -8.49883616e-01 1.12197566e+00 6.13666821e+00 1.06544471e+00 -1.03491604e+00 3.16343039e-01 3.49947780e-01 -1.88958690e-01 -8.46743226e-01 -2.19413802e-01 -5.56923032e-01 2.17275679e-01 1.18959713e+00 5.79495504e-02 9.54675674e-01 7.10956454e-01 6.51733041e-01 6.54983044e-01 -1.08070064e+00 8.00362468e-01 1.99739616e-02 -8.15553427e-01 6.52815923e-02 -2.13825423e-02 4.34562534e-01 -8.68602395e-02 2.28361726e-01 5.92239499e-01 4.12106812e-01 -1.15823257e+00 1.27844846e+00 4.94108200e-01 6.98992610e-01 -7.44350910e-01 2.79804289e-01 4.56484079e-01 -1.07069337e+00 1.36503085e-01 4.16639186e-02 -3.27138826e-02 5.11356771e-01 3.20623308e-01 -6.54881835e-01 5.79108655e-01 2.18639791e-01 1.90615326e-01 2.25826144e-01 9.71938893e-02 -4.11407679e-01 1.09726763e+00 -8.41552690e-02 3.44629019e-01 -7.46313930e-02 -1.03890702e-01 9.60365415e-01 1.28111291e+00 9.26251709e-02 2.55571753e-01 1.67560689e-02 1.22714937e+00 -6.43864498e-02 -5.28083965e-02 -2.68166840e-01 -3.85412753e-01 5.65805256e-01 9.36388731e-01 1.08861988e-02 -1.36936352e-01 -4.79003429e-01 9.38159943e-01 8.75814855e-02 7.61569738e-01 -8.33156109e-01 1.18901744e-01 1.41059089e+00 -8.45205858e-02 3.12628329e-01 -4.13206369e-01 2.54060496e-02 -1.38000464e+00 -2.99635768e-01 -1.09422350e+00 -1.50631756e-01 -8.79317760e-01 -1.21117401e+00 7.49702990e-01 -1.53242409e-01 -8.05906236e-01 -7.62129486e-01 -5.44044614e-01 -4.77950215e-01 1.39903247e+00 -1.28433049e+00 -1.25709248e+00 2.54751951e-01 8.08157682e-01 1.05805159e+00 -1.75549448e-01 1.02180576e+00 3.50861281e-01 -6.96073890e-01 6.48850918e-01 1.30848616e-01 1.14843890e-01 4.94147122e-01 -1.33019853e+00 4.21698123e-01 7.69487500e-01 3.83841634e-01 5.55010557e-01 1.23308587e+00 -3.41921866e-01 -1.30727744e+00 -9.42557931e-01 4.20228988e-01 -4.18098122e-01 8.74296129e-01 -7.88164496e-01 -1.01406527e+00 6.25755668e-01 2.93108255e-01 7.09883645e-02 9.36735809e-01 1.91235468e-01 -5.15062451e-01 -3.95578407e-02 -8.67813289e-01 7.04684973e-01 4.84234899e-01 -1.22738886e+00 -8.82860005e-01 -2.25069076e-01 1.46821737e+00 -4.63724017e-01 -8.93876493e-01 2.18993559e-01 4.52348977e-01 -9.94559228e-01 1.16517115e+00 -5.47394574e-01 2.76965678e-01 -1.78082079e-01 -6.35057390e-01 -1.53177285e+00 -3.32662791e-01 -9.82979715e-01 -5.23683190e-01 1.69335103e+00 2.20814019e-01 -1.84711814e-01 4.24461991e-01 4.76121843e-01 -4.05084461e-01 -1.94254860e-01 -9.55848455e-01 -5.49758255e-01 1.90524265e-01 -9.08919394e-01 6.52630448e-01 7.85990179e-01 -3.36887330e-01 5.79929709e-01 -1.03133011e+00 5.23307979e-01 3.80445987e-01 -3.29238743e-01 6.12699032e-01 -1.05892146e+00 -7.46244907e-01 -3.36002558e-01 -1.38044953e-01 -9.01118696e-01 4.13725883e-01 -3.20145339e-01 2.06175879e-01 -8.58015716e-01 -3.26631784e-01 5.56916632e-02 -5.05202591e-01 -7.40909055e-02 -2.46516377e-01 -2.30807751e-01 4.04458880e-01 -1.31471470e-01 9.66205820e-02 1.06712413e+00 1.20086098e+00 -4.30116653e-02 -2.72464305e-01 3.66977565e-02 -4.42347825e-01 8.64177525e-01 5.74281931e-01 -3.34744632e-01 -5.79537094e-01 -4.85943168e-01 -5.43574333e-01 8.34602118e-01 2.24794284e-01 -7.15232432e-01 -4.96162288e-02 -1.30438134e-01 1.34038821e-01 -2.72014588e-01 1.02462566e+00 -8.99232328e-01 2.66430944e-01 3.67604382e-02 -5.70410848e-01 -5.31183600e-01 4.51944619e-01 6.99881077e-01 -6.02896690e-01 -1.90226495e-01 7.51836598e-01 6.75042868e-02 -1.28941461e-01 1.31664500e-01 -5.63335240e-01 3.69527973e-02 6.55196488e-01 8.50108936e-02 9.91540700e-02 -5.72442114e-01 -1.30072820e+00 -3.87487322e-01 5.33816479e-02 6.22437716e-01 2.20968470e-01 -1.22877443e+00 -1.06024039e+00 4.84577835e-01 -1.63846627e-01 -2.90728092e-01 7.32756793e-01 1.90928325e-01 -4.55873013e-02 3.26891065e-01 1.20685071e-01 -3.43778521e-01 -1.12105453e+00 7.84204245e-01 3.23709786e-01 -2.15724781e-01 -2.93811262e-01 8.42196524e-01 5.03283560e-01 -4.22670782e-01 3.04042786e-01 -4.56022173e-01 -6.64049834e-02 1.94971282e-02 4.07582283e-01 3.39928508e-01 -2.19345197e-01 -1.05522144e+00 -2.17300504e-02 3.31485808e-01 3.20654839e-01 -4.96690691e-01 1.03203058e+00 -4.34438437e-01 3.98974568e-01 7.39518881e-01 1.36185610e+00 6.57876313e-01 -1.53904712e+00 -1.63014010e-01 -4.30721402e-01 -2.73303948e-02 4.32218999e-01 -7.15393722e-01 -7.69528449e-01 1.15267074e+00 3.86531383e-01 3.61677051e-01 1.19223046e+00 -9.73544642e-02 6.67968690e-01 -2.65345931e-01 -6.26247749e-02 -9.30158019e-01 4.72439229e-02 5.09801626e-01 1.08731425e+00 -9.17403400e-01 -4.96930748e-01 -2.35893875e-01 -8.29013109e-01 9.02994752e-01 2.26815388e-01 -1.24982893e-01 7.61694610e-01 8.98729920e-01 2.41181657e-01 4.00882214e-01 -7.53828108e-01 -2.02793479e-01 6.05201364e-01 5.25751948e-01 4.66604352e-01 3.94737154e-01 3.00287992e-01 1.03744149e+00 -4.52112526e-01 -3.87650400e-01 1.57646596e-01 5.60734086e-02 -3.35127115e-01 -1.20029414e+00 -6.15426838e-01 -1.79278716e-01 -6.76493526e-01 -3.21013421e-01 -5.62282978e-03 3.62780541e-01 2.75395185e-01 1.05483937e+00 3.82331721e-02 -4.64097828e-01 5.83913743e-01 3.57563466e-01 1.67847142e-01 -6.18970752e-01 -6.86782539e-01 8.21008384e-01 1.65423110e-01 -3.40123266e-01 -1.52371153e-01 -5.60018957e-01 -1.01083076e+00 -4.51265611e-02 -3.02927971e-01 1.27495244e-01 7.56647229e-01 9.45865095e-01 2.18677241e-02 1.27902436e+00 5.57236373e-01 -9.92809236e-01 -7.26409018e-01 -1.14692020e+00 -6.94883347e-01 6.49644852e-01 8.40385795e-01 -3.69093925e-01 -5.47944844e-01 2.54535735e-01]
[14.988808631896973, 6.527825832366943]
25b7747c-7643-416e-953a-7cc96ae90285
automatic-context-window-composition-for
1805.10498
null
http://arxiv.org/abs/1805.10498v1
http://arxiv.org/pdf/1805.10498v1.pdf
Automatic context window composition for distant speech recognition
Distant speech recognition is being revolutionized by deep learning, that has contributed to significantly outperform previous HMM-GMM systems. A key aspect behind the rapid rise and success of DNNs is their ability to better manage large time contexts. With this regard, asymmetric context windows that embed more past than future frames have been recently used with feed-forward neural networks. This context configuration turns out to be useful not only to address low-latency speech recognition, but also to boost the recognition performance under reverberant conditions. This paper investigates on the mechanisms occurring inside DNNs, which lead to an effective application of asymmetric contexts.In particular, we propose a novel method for automatic context window composition based on a gradient analysis. The experiments, performed with different acoustic environments, features, DNN architectures, microphone settings, and recognition tasks show that our simple and efficient strategy leads to a less redundant frame configuration, which makes DNN training more effective in reverberant scenarios.
['Mirco Ravanelli', 'Maurizio Omologo']
2018-05-26
null
null
null
null
['distant-speech-recognition']
['speech']
[ 2.71758139e-02 -4.23646688e-01 3.56218100e-01 -2.42442802e-01 -3.15998077e-01 -2.45663404e-01 6.82094038e-01 -1.01343296e-01 -6.68103993e-01 6.89430892e-01 3.68340373e-01 -4.74411249e-01 -4.57274579e-02 -4.56415653e-01 -2.86235332e-01 -9.82313454e-01 -1.17034100e-01 -1.40221968e-01 3.29949409e-01 -1.39563635e-01 1.52903661e-01 7.72200882e-01 -1.91086614e+00 2.14669108e-01 3.79788131e-01 8.30796242e-01 9.04527009e-01 8.83961678e-01 -3.00970584e-01 6.35729373e-01 -9.49366987e-01 -2.20751092e-01 -2.15526279e-02 -4.41697270e-01 -4.27400619e-01 -1.98373124e-01 2.83119027e-02 -2.14694098e-01 -2.25086331e-01 8.44315529e-01 9.77470160e-01 3.97341698e-01 2.46872112e-01 -3.99679869e-01 -3.68658192e-02 6.86898708e-01 1.16733007e-01 5.66027343e-01 2.28523135e-01 -7.22103333e-03 7.50031471e-01 -8.39463949e-01 3.03222656e-01 1.11520374e+00 4.39081103e-01 7.04819858e-01 -9.53121245e-01 -4.29433554e-01 3.25081140e-01 7.98964679e-01 -1.03141606e+00 -6.64131045e-01 7.50537574e-01 6.84553012e-02 1.33626401e+00 5.15571535e-01 9.21409667e-01 1.46934175e+00 8.74304548e-02 5.34298122e-01 9.24760103e-01 -8.92668903e-01 4.44197267e-01 -1.17429318e-02 -1.15348332e-01 -2.23817229e-02 -9.96985659e-02 4.67095040e-02 -4.30865437e-01 2.25300506e-01 4.49550003e-01 -4.11191322e-02 -4.45730120e-01 1.92347333e-01 -8.97045672e-01 5.58491290e-01 -3.90463956e-02 1.03670347e+00 -3.84692430e-01 5.22357523e-02 5.38832903e-01 2.88536578e-01 5.49267530e-01 3.27478617e-01 -4.36041355e-01 -6.35547996e-01 -8.40400994e-01 -1.47638945e-02 9.14720476e-01 3.06012243e-01 3.33666682e-01 5.94978213e-01 1.71386048e-01 1.18984520e+00 2.30289057e-01 4.22809750e-01 8.01977456e-01 -4.89208937e-01 4.36349064e-01 -2.50818674e-02 -2.39471018e-01 -6.64544284e-01 -3.74759793e-01 -8.47490966e-01 -6.77489221e-01 1.28745213e-01 2.69121438e-01 -2.43458271e-01 -7.10241079e-01 1.68785822e+00 3.28565806e-01 3.99221897e-01 1.17644750e-01 7.97109425e-01 3.61330271e-01 8.40151131e-01 -1.00121558e-01 -3.62308890e-01 1.21007574e+00 -5.32845080e-01 -9.89059746e-01 -6.41572699e-02 3.84405732e-01 -1.18381894e+00 1.02155769e+00 7.56675065e-01 -7.49361336e-01 -7.23550797e-01 -1.04857671e+00 2.90914118e-01 -5.45148611e-01 6.19291253e-02 2.13426396e-01 1.02325356e+00 -1.17636096e+00 5.82799375e-01 -7.13615239e-01 -3.10206741e-01 -5.54884672e-02 2.83555061e-01 -7.36389682e-02 2.33021259e-01 -1.18638492e+00 9.16716397e-01 4.92979527e-01 5.70800722e-01 -5.76598406e-01 -2.75457472e-01 -4.46957111e-01 2.68672198e-01 2.67378479e-01 -2.70733207e-01 1.37713385e+00 -9.26443279e-01 -1.95188344e+00 3.21925223e-01 -3.20935369e-01 -9.01265085e-01 3.94287735e-01 -5.86966455e-01 -7.43470669e-01 3.75699289e-02 -6.07792556e-01 2.95083463e-01 9.15311396e-01 -9.49602783e-01 -7.15306461e-01 -1.75327584e-01 -2.58638263e-01 1.85153544e-01 -6.22205794e-01 1.13579750e-01 -1.48798779e-01 -6.87345982e-01 -2.72947606e-02 -7.93620765e-01 -1.81499839e-01 -7.00113833e-01 -1.32682458e-01 -2.50879079e-01 1.04373944e+00 -7.92767227e-01 1.27340889e+00 -2.12161899e+00 -1.92980375e-02 1.36881813e-01 -2.45894954e-01 9.74430084e-01 1.59763008e-01 5.45123160e-01 -3.99749391e-02 -1.28129765e-01 2.36058354e-01 -6.59232855e-01 -1.74247921e-01 4.66275573e-01 -4.88728046e-01 4.74072099e-02 3.17779519e-02 1.81846455e-01 -5.14026105e-01 -2.31051758e-01 7.49243557e-01 8.09338272e-01 -3.26001137e-01 2.58781224e-01 -1.25702843e-01 4.78675365e-01 1.91350728e-02 1.92660257e-01 4.48490947e-01 4.81089592e-01 2.39935860e-01 2.61433665e-02 -5.32628000e-01 5.61359584e-01 -1.16562533e+00 1.30451739e+00 -8.84946048e-01 1.09737790e+00 1.49695292e-01 -1.03171253e+00 1.25494349e+00 6.35317743e-01 1.25988945e-01 -8.91856432e-01 1.91605926e-01 4.92913246e-01 8.85397270e-02 -5.29013216e-01 5.76276481e-01 -5.68912625e-02 6.21909559e-01 -1.89884499e-01 3.35246921e-02 7.81460032e-02 6.97353855e-02 -4.29815203e-01 7.07219422e-01 -8.59329179e-02 1.89295173e-01 5.70601411e-02 8.06498766e-01 -8.12174320e-01 5.03382742e-01 6.41772807e-01 -6.50544837e-02 5.23139298e-01 1.94886446e-01 -4.71340328e-01 -8.24589431e-01 -7.26981461e-01 -2.94884518e-02 9.23769653e-01 -2.52015322e-01 -1.74378023e-01 -8.10620606e-01 -7.78043196e-02 -5.89788377e-01 8.11765850e-01 -4.76137400e-02 -4.26940471e-02 -1.11598587e+00 -4.86084133e-01 3.69410366e-01 3.86820704e-01 5.92371762e-01 -1.16509879e+00 -6.59984052e-01 5.75074017e-01 -1.41648769e-01 -1.25255442e+00 -3.01528978e-03 4.30467933e-01 -1.18152308e+00 -6.06052935e-01 -9.51330423e-01 -7.45027959e-01 -9.50849652e-02 2.21800163e-01 7.31220901e-01 -1.63504526e-01 -1.90994278e-01 3.09651643e-01 -7.11369574e-01 -4.01391745e-01 -5.65940499e-01 2.93552011e-01 1.29494637e-01 1.38329446e-01 2.24870309e-01 -9.09359753e-01 -5.92002213e-01 3.86582136e-01 -8.47081125e-01 -1.49480850e-01 6.94292247e-01 7.57620990e-01 3.08142990e-01 5.54002374e-02 7.83433139e-01 -4.08194989e-01 5.49877167e-01 -1.88291550e-01 -7.50716209e-01 1.82783797e-01 -5.07946074e-01 -5.85126393e-02 8.53449643e-01 -5.02910018e-01 -1.44422209e+00 -3.30323994e-01 -9.38227177e-01 -2.12369323e-01 -4.09720331e-01 1.59742668e-01 -3.91624719e-01 1.85917512e-01 5.19047201e-01 2.53400713e-01 -2.67347783e-01 -9.28996384e-01 1.63726524e-01 7.56427288e-01 1.98986009e-01 -4.20123577e-01 2.38535166e-01 2.99600154e-01 -8.13884214e-02 -1.41999519e+00 6.75795898e-02 -4.31664854e-01 -3.10627729e-01 -4.26586747e-01 7.65075982e-01 -6.09506190e-01 -7.07536042e-01 6.35026276e-01 -1.27301300e+00 -3.36125225e-01 -1.09701999e-01 9.50457752e-01 -2.49741748e-01 4.05130893e-01 -5.71298003e-01 -1.29204404e+00 -3.09533089e-01 -1.29723227e+00 5.00027478e-01 3.67555410e-01 6.71183467e-02 -9.55354214e-01 3.70156579e-02 8.82215723e-02 7.50935376e-01 -2.82110840e-01 7.00173080e-01 -8.62430036e-01 -6.28827155e-01 -4.25806828e-02 1.59388646e-01 9.37914848e-01 1.18789174e-01 1.21419527e-01 -1.49308336e+00 -1.26569197e-01 4.02380913e-01 3.53591681e-01 6.47345603e-01 4.70688194e-01 1.11926413e+00 -2.03335956e-01 -1.59741566e-01 4.37034518e-01 1.12582791e+00 8.82880270e-01 7.59583473e-01 2.12226376e-01 2.39972636e-01 4.70304012e-01 4.95749503e-01 3.98468256e-01 -1.96330830e-01 9.50558007e-01 3.97310823e-01 -9.00743008e-02 -5.14450312e-01 1.59938887e-01 3.70089442e-01 1.38245142e+00 -2.74108618e-01 -4.01942700e-01 -7.68306911e-01 4.99471813e-01 -1.55030882e+00 -1.07133007e+00 1.71614721e-01 2.22813058e+00 4.80992556e-01 3.87149453e-01 5.85873611e-02 4.34176713e-01 8.17277372e-01 3.43960404e-01 -1.11323483e-01 -7.24876404e-01 -2.46281222e-01 5.52131653e-01 3.48401256e-02 5.93631566e-01 -7.64081001e-01 7.56621599e-01 5.52075338e+00 1.00516188e+00 -1.65137064e+00 2.10692346e-01 4.19366091e-01 -2.75571495e-01 -5.70906401e-02 -1.91654995e-01 -9.25051451e-01 5.39574802e-01 1.30104816e+00 2.67537981e-01 1.56912848e-01 9.55999076e-01 4.41716582e-01 -1.42337680e-01 -9.32306468e-01 1.12271893e+00 -8.88705626e-02 -1.25078356e+00 -1.31528094e-01 1.72683358e-01 2.30860591e-01 9.88404825e-02 2.61229336e-01 2.35750362e-01 -3.86703014e-01 -6.20821178e-01 8.36402833e-01 3.45953256e-01 3.75850886e-01 -8.94299507e-01 9.55553770e-01 1.10563934e-01 -1.20348394e+00 -2.71003813e-01 -4.15153891e-01 -1.55941874e-01 2.34952614e-01 9.01028335e-01 -1.37298298e+00 5.19888282e-01 7.32744694e-01 2.00639769e-01 -1.37904003e-01 1.28348958e+00 -1.11001752e-01 8.83570373e-01 -4.08238053e-01 -5.92667878e-01 9.96450260e-02 -4.66627441e-02 8.14955235e-01 1.55211771e+00 6.11013174e-01 -3.86089236e-01 -3.93160820e-01 3.85147452e-01 3.49615932e-01 2.26652920e-01 -6.15547121e-01 2.21383810e-01 6.81796730e-01 9.03205097e-01 -7.02122450e-01 -1.14763953e-01 -3.11566204e-01 7.48351038e-01 -5.25332130e-02 5.88710248e-01 -8.25114906e-01 -4.90425110e-01 7.59963334e-01 -1.72892883e-01 5.99001706e-01 -7.07385719e-01 -4.10790853e-02 -9.22894955e-01 1.75370753e-01 -7.63904512e-01 -3.71231474e-02 -4.92958814e-01 -6.62510037e-01 9.68604028e-01 4.23366129e-02 -1.05608571e+00 -5.57045639e-01 -8.80699813e-01 -7.28942871e-01 9.52554047e-01 -1.48132610e+00 -6.51800632e-01 1.06208831e-01 4.52133179e-01 9.65464056e-01 -1.75547361e-01 8.36221218e-01 7.09404290e-01 -6.29317224e-01 4.65213507e-01 2.55632639e-01 -1.57489479e-01 5.23149014e-01 -1.02054346e+00 4.39000994e-01 1.07669163e+00 5.14121830e-01 6.44955635e-01 9.66330707e-01 -2.51340508e-01 -9.87336576e-01 -5.30846953e-01 8.78674567e-01 1.37439862e-01 3.62723827e-01 -6.46466434e-01 -9.12549317e-01 4.74698842e-02 3.93997669e-01 -4.54388618e-01 6.99517250e-01 2.87304521e-01 -7.30142593e-02 -7.77970076e-01 -7.09908187e-01 8.65185440e-01 8.94234717e-01 -6.98035479e-01 -7.28096783e-01 -2.20455136e-02 7.10637987e-01 -1.86404720e-01 -2.56321281e-01 2.43298769e-01 5.54028451e-01 -1.52080953e+00 9.72800970e-01 -3.18225659e-02 -2.29065850e-01 -2.67626345e-02 -3.83070469e-01 -1.26137280e+00 1.94161698e-01 -8.79387558e-01 -2.42275834e-01 1.33714497e+00 3.82842511e-01 -1.07331002e+00 6.37478292e-01 1.30018771e-01 -4.63513225e-01 -7.06477463e-01 -1.47058141e+00 -1.02153695e+00 -2.03235939e-01 -6.81284487e-01 5.83004951e-01 4.06867743e-01 -4.19655353e-01 -7.61580542e-02 -2.80552596e-01 9.36299562e-03 5.36125600e-02 -2.77598470e-01 4.94078636e-01 -9.75007534e-01 -5.64231575e-01 -6.58411205e-01 -5.32065034e-01 -1.20188057e+00 -8.01084563e-02 -1.19567066e-01 7.83356950e-02 -1.10528016e+00 -6.81787670e-01 -4.41712379e-01 -3.59943241e-01 6.00644834e-02 3.25346701e-02 -3.69028710e-02 3.25532466e-01 -8.79134312e-02 -1.47681072e-01 4.80448872e-01 7.32082784e-01 2.54697949e-01 -2.72379518e-01 4.60996270e-01 1.74884442e-02 5.95710576e-01 1.02554142e+00 -2.18293950e-01 -4.99900192e-01 -4.59041864e-01 1.25339389e-01 -1.94611788e-01 2.31319383e-01 -1.63994884e+00 1.98082969e-01 2.73379117e-01 2.91801721e-01 -6.53416574e-01 8.24712038e-01 -1.00366700e+00 2.96536326e-01 5.09747624e-01 -3.64020723e-03 1.15947008e-01 3.75725418e-01 4.01161581e-01 -4.53661531e-01 -5.76251507e-01 5.44173896e-01 6.47593895e-03 -7.29418457e-01 -4.17831779e-01 -7.53942251e-01 -5.20744622e-01 6.19761109e-01 -4.05216783e-01 4.64964211e-02 -4.27566737e-01 -8.98532391e-01 -5.78302443e-01 -1.28725171e-01 4.60942268e-01 4.74435657e-01 -9.25409794e-01 -2.54656643e-01 2.54066199e-01 -5.29997885e-01 -2.79458761e-01 5.27956247e-01 6.96132481e-01 -4.56339240e-01 7.02770472e-01 7.54270405e-02 -6.77263737e-01 -1.60173702e+00 2.77494639e-01 4.08788294e-01 -5.04096150e-02 -5.58498859e-01 9.46794569e-01 -3.14951055e-02 1.83793828e-01 7.14318752e-01 -8.89150143e-01 -2.45012820e-01 2.07122222e-01 7.15326130e-01 5.35966873e-01 5.94675660e-01 -2.46626616e-01 -2.15283766e-01 4.30512339e-01 -2.65414208e-01 -3.88661504e-01 1.21401846e+00 -3.02562565e-01 2.26023316e-01 5.98583817e-01 1.06664693e+00 2.28742763e-01 -1.14057851e+00 -7.36772493e-02 1.87009588e-01 -4.41634893e-01 1.17622048e-01 -7.03226566e-01 -7.92201936e-01 1.24407852e+00 9.05500889e-01 7.04466403e-01 1.39363635e+00 -3.36068511e-01 8.67537796e-01 4.90487963e-01 4.06205177e-01 -1.20055866e+00 1.27831146e-01 7.22385406e-01 7.14420319e-01 -8.28768075e-01 -4.56976175e-01 1.83023121e-02 -1.43919572e-01 1.53659344e+00 2.98071653e-01 4.07899290e-01 5.66676974e-01 2.49827638e-01 3.03315878e-01 4.03640985e-01 -5.63856900e-01 -3.82397681e-01 -1.61320195e-01 5.64541042e-01 4.51289177e-01 1.25975357e-02 -3.16973597e-01 2.88491935e-01 -1.73797145e-01 -2.08869711e-01 1.72158107e-01 8.00418377e-01 -6.33737743e-01 -1.48857796e+00 -6.93586111e-01 -6.73348606e-02 -5.56784213e-01 -2.81321760e-02 -7.50284120e-02 7.32814133e-01 2.50521541e-01 1.13903868e+00 -1.05938695e-01 -4.04611021e-01 3.51452500e-01 4.78697151e-01 3.34603667e-01 -3.10824722e-01 -6.84479535e-01 3.85125726e-01 1.92259356e-01 -3.56485754e-01 -3.29495221e-01 -5.46306729e-01 -1.02488184e+00 -9.43526998e-02 -6.12615049e-01 2.92028040e-01 1.25624454e+00 1.04069650e+00 3.09632659e-01 7.94027746e-01 6.33466065e-01 -9.86735642e-01 -3.94107252e-01 -1.15901411e+00 -2.78855681e-01 -4.45217639e-02 5.47364116e-01 -4.36191946e-01 -5.26969969e-01 -1.82615995e-01]
[14.842227935791016, 5.873052597045898]
13ad7a38-a4b2-44d2-8808-a9af2c40f830
appt-asymmetric-parallel-point-transformer
2303.17815
null
https://arxiv.org/abs/2303.17815v1
https://arxiv.org/pdf/2303.17815v1.pdf
APPT : Asymmetric Parallel Point Transformer for 3D Point Cloud Understanding
Transformer-based networks have achieved impressive performance in 3D point cloud understanding. However, most of them concentrate on aggregating local features, but neglect to directly model global dependencies, which results in a limited effective receptive field. Besides, how to effectively incorporate local and global components also remains challenging. To tackle these problems, we propose Asymmetric Parallel Point Transformer (APPT). Specifically, we introduce Global Pivot Attention to extract global features and enlarge the effective receptive field. Moreover, we design the Asymmetric Parallel structure to effectively integrate local and global information. Combined with these designs, APPT is able to capture features globally throughout the entire network while focusing on local-detailed features. Extensive experiments show that our method outperforms the priors and achieves state-of-the-art on several benchmarks for 3D point cloud understanding, such as 3D semantic segmentation on S3DIS, 3D shape classification on ModelNet40, and 3D part segmentation on ShapeNet.
['Deng Cai', 'Binbin Lin', 'Boxi Wu', 'Wenxiao Wang', 'Zheng Yang', 'Zhihao Chi', 'Tu Zheng', 'Hengjia Li']
2023-03-31
null
null
null
null
['3d-shape-retrieval', '3d-part-segmentation']
['computer-vision', 'computer-vision']
[-3.04107428e-01 -3.31204012e-02 -3.02714091e-02 -3.92894864e-01 -4.66685504e-01 -6.56364501e-01 4.98431504e-01 9.24763307e-02 -6.59620464e-02 2.12630332e-02 -1.34216189e-01 -1.05409905e-01 -2.55794406e-01 -1.06091547e+00 -8.82723093e-01 -4.94356990e-01 1.34495586e-01 7.59521186e-01 5.75956702e-01 -1.64066240e-01 2.97791302e-01 1.10495412e+00 -1.33346450e+00 8.76583382e-02 8.88953626e-01 1.23906267e+00 4.44589764e-01 8.21698681e-02 -5.27054727e-01 1.84975833e-01 -3.64827096e-01 -7.69497231e-02 3.32912058e-01 3.81393939e-01 -7.12126613e-01 1.10492215e-01 4.87060875e-01 -3.74829859e-01 -2.67398298e-01 9.18339312e-01 4.36741024e-01 -1.28578895e-03 6.33405983e-01 -1.10013151e+00 -5.53237557e-01 3.27002227e-01 -6.10194385e-01 -6.59137592e-02 -4.85930592e-02 2.10253313e-01 9.25935268e-01 -1.04401362e+00 4.82120961e-01 1.53188145e+00 6.04497433e-01 2.15918034e-01 -1.06648779e+00 -7.43943989e-01 7.39302814e-01 -3.45984678e-04 -1.30038917e+00 -3.77789102e-02 1.06513453e+00 -2.33771488e-01 1.29588187e+00 5.13906069e-02 8.68089497e-01 6.37481689e-01 -6.35742471e-02 1.08511603e+00 8.46144617e-01 1.21244118e-01 7.93327466e-02 -3.45620632e-01 1.03479885e-01 6.09384894e-01 -1.77975029e-01 -6.78240648e-03 -2.98955649e-01 1.14641033e-01 1.36031604e+00 2.29282320e-01 -9.89258289e-03 -5.59905946e-01 -1.13143098e+00 6.13191366e-01 8.79469335e-01 2.51730263e-01 -4.16706800e-01 1.23537548e-01 1.52185023e-01 6.57898411e-02 8.05234373e-01 2.58968860e-01 -7.62695789e-01 4.77198064e-02 -7.13928461e-01 3.57876211e-01 5.50391793e-01 1.25984800e+00 1.19001567e+00 -8.75851959e-02 -1.03989653e-01 9.53948438e-01 3.73433381e-01 8.18028629e-01 -1.73898757e-01 -1.12632108e+00 5.22722304e-01 1.09977019e+00 -1.65166095e-01 -1.13273144e+00 -5.90957701e-01 -6.39133990e-01 -8.73394072e-01 7.86252245e-02 2.12031499e-01 4.04696673e-01 -1.42305136e+00 1.45085013e+00 3.74479473e-01 2.73425192e-01 -3.81926537e-01 1.04477835e+00 1.15574157e+00 6.31592631e-01 -1.47114545e-01 4.47362453e-01 1.16128445e+00 -1.12221754e+00 -1.74259901e-01 -3.70297939e-01 2.42817611e-01 -6.65184915e-01 8.55743110e-01 7.72669315e-02 -1.13833952e+00 -7.02079237e-01 -7.11997092e-01 -3.15751165e-01 -3.30032349e-01 8.08051899e-02 7.60845363e-01 1.38701051e-01 -1.11630404e+00 5.34446180e-01 -1.10484660e+00 -1.18177481e-01 8.26878190e-01 6.84960723e-01 -2.72007376e-01 -2.94775039e-01 -6.71834409e-01 4.88250494e-01 1.66662276e-01 9.46087986e-02 -7.69466937e-01 -1.10629916e+00 -8.33738744e-01 2.56136954e-01 2.91664779e-01 -1.03253222e+00 1.03422284e+00 -3.21249813e-01 -1.59471917e+00 7.67276347e-01 -2.39667833e-01 -1.21978447e-01 2.80973285e-01 -1.24555707e-01 2.57148772e-01 4.13147360e-01 8.98959562e-02 1.27538931e+00 8.41702938e-01 -1.35999143e+00 -4.86234695e-01 -7.88736403e-01 2.44555920e-01 4.05583173e-01 -1.23664297e-01 -3.85960758e-01 -8.94333184e-01 -5.29614866e-01 7.69626260e-01 -8.12427223e-01 -3.56227219e-01 2.96895877e-02 -3.50067973e-01 -5.38652003e-01 1.08691621e+00 -2.65768260e-01 4.70158219e-01 -2.22474718e+00 1.68058574e-01 3.71374279e-01 5.17198801e-01 1.84281796e-01 -2.77753294e-01 6.58393055e-02 1.80148467e-01 2.28802592e-01 -2.43492305e-01 -5.78892708e-01 1.95915237e-01 5.33552468e-01 -3.19237113e-01 2.27658197e-01 3.41773033e-01 1.27181792e+00 -6.51384234e-01 -3.71994644e-01 7.04661906e-01 7.30802894e-01 -7.43963480e-01 -1.19781300e-01 -2.98214644e-01 6.60840213e-01 -1.10698271e+00 8.03895295e-01 1.32315981e+00 -4.20884043e-01 -4.47664410e-01 -2.05793947e-01 -1.88159510e-01 3.52672547e-01 -7.82895446e-01 2.01975417e+00 -5.57682216e-01 1.19231038e-01 3.02087456e-01 -9.49798763e-01 1.18486762e+00 -2.87605226e-02 6.12177134e-01 -6.76346600e-01 1.18502647e-01 2.50235915e-01 -2.71422386e-01 -1.71556298e-04 1.61126912e-01 1.45934984e-01 2.09732980e-01 8.34872276e-02 5.56356423e-02 -6.48562968e-01 -2.36169487e-01 -3.48612517e-02 9.53469396e-01 2.69298047e-01 -3.11227709e-01 -2.23444283e-01 5.48330307e-01 -6.63646236e-02 5.90905309e-01 5.50941408e-01 -8.91923532e-03 9.03438270e-01 4.57931072e-01 -5.18299639e-01 -7.59419620e-01 -1.16596723e+00 -2.31092751e-01 6.98254943e-01 7.32385755e-01 -3.61627281e-01 -5.35936594e-01 -8.18899751e-01 2.27271602e-01 3.21746737e-01 -3.38201791e-01 1.00497037e-01 -7.12029517e-01 -4.01138037e-01 2.20289454e-01 9.33920860e-01 8.68085384e-01 -9.76113141e-01 -4.12639648e-01 2.15180323e-01 -1.10884339e-01 -1.27183914e+00 -3.38172883e-01 6.28844276e-02 -1.26504600e+00 -8.42181861e-01 -6.32871985e-01 -7.03999102e-01 7.66103446e-01 5.86050391e-01 1.17269564e+00 -6.34890124e-02 4.27417755e-02 3.53908807e-01 -3.27648193e-01 -2.38773569e-01 2.95521170e-01 4.39003587e-01 -4.54651654e-01 -1.99414283e-01 3.74982446e-01 -8.77764225e-01 -7.67933667e-01 7.07267821e-01 -6.71746016e-01 1.14064135e-01 7.52039194e-01 5.79532325e-01 9.46970463e-01 -7.41842538e-02 2.56909788e-01 -5.15746653e-01 1.41241238e-01 -1.80297256e-01 -5.54937720e-01 8.63083228e-02 -1.82234257e-01 -1.79372549e-01 4.94278431e-01 -1.53631195e-01 -9.93896008e-01 7.22983200e-03 -5.58623552e-01 -8.66334200e-01 -4.43165720e-01 2.13424087e-01 -4.67434287e-01 -4.29404020e-01 -1.82799697e-02 2.57651925e-01 -1.10232189e-01 -7.18274236e-01 2.64146090e-01 2.94709980e-01 1.51039168e-01 -8.32857490e-01 7.62074113e-01 7.91130483e-01 1.21042468e-01 -8.76754940e-01 -6.74622655e-01 -5.92581511e-01 -8.42791617e-01 1.36543363e-02 8.69246960e-01 -9.07141447e-01 -8.42726886e-01 7.95474470e-01 -1.28400612e+00 -2.86440998e-01 -2.95257211e-01 2.60136545e-01 -5.82248271e-01 1.40704915e-01 -6.08005702e-01 -3.81370664e-01 -3.05184603e-01 -1.45589042e+00 1.67967045e+00 2.87455708e-01 2.42974550e-01 -9.28164780e-01 -3.85073632e-01 3.76850486e-01 5.05638778e-01 4.53484468e-02 1.06024611e+00 -3.01702559e-01 -1.19431973e+00 -3.06364484e-02 -7.11399674e-01 2.96186119e-01 7.79061839e-02 -2.66042292e-01 -1.01715100e+00 -1.10973366e-01 -3.90807865e-03 -1.00004688e-01 1.09297025e+00 6.96115255e-01 1.53445971e+00 2.54841447e-01 -6.21666193e-01 1.04619420e+00 1.09621215e+00 -8.06676298e-02 4.81987000e-01 -1.09178133e-01 1.09140646e+00 5.42183816e-01 5.13928592e-01 2.79698074e-01 6.32633805e-01 5.53470314e-01 8.30669463e-01 -2.29447722e-01 -8.61732215e-02 -4.22498554e-01 -2.18769729e-01 9.13772583e-01 -3.75597924e-01 -2.00042561e-01 -8.70462179e-01 5.56854725e-01 -1.82805014e+00 -2.87690639e-01 -1.36472866e-01 1.72387755e+00 2.16311172e-01 1.37252301e-01 -2.61703759e-01 -1.93236798e-01 5.27244985e-01 4.06541407e-01 -7.17364311e-01 3.74998935e-02 -1.94535404e-01 4.33064640e-01 6.22407198e-01 2.72983521e-01 -1.18934083e+00 1.35207987e+00 5.57270432e+00 1.24400282e+00 -1.19678545e+00 1.25840828e-01 6.46473825e-01 1.36210680e-01 -6.00206494e-01 8.46920609e-02 -9.07170892e-01 1.60284504e-01 7.45260417e-02 3.76327515e-01 2.09964991e-01 8.24402094e-01 -5.44344336e-02 1.56935677e-01 -8.87064993e-01 1.13658440e+00 -2.00379461e-01 -1.25229573e+00 2.37524301e-01 2.93732703e-01 8.26217949e-01 5.56999028e-01 7.39232674e-02 2.34146535e-01 -6.49707168e-02 -1.01850760e+00 6.86886966e-01 3.96593273e-01 5.19118786e-01 -8.20334196e-01 5.98065615e-01 5.17324626e-01 -1.52087808e+00 1.54938981e-01 -6.65230691e-01 5.15974425e-02 1.23139709e-01 8.55053544e-01 -4.67367351e-01 6.49023473e-01 8.33105326e-01 1.14177096e+00 -3.66318375e-01 1.05883932e+00 -2.89171696e-01 2.77853489e-01 -8.76738787e-01 1.59360021e-01 6.06664836e-01 -2.01194927e-01 7.78728068e-01 7.09072053e-01 2.93949962e-01 7.82352909e-02 4.61094916e-01 1.29016936e+00 -9.06479210e-02 -3.75976153e-02 -5.28217375e-01 1.08278804e-01 3.66109699e-01 1.25625634e+00 -1.05760622e+00 -1.08932242e-01 -4.20323133e-01 6.17638409e-01 2.73274213e-01 4.84816670e-01 -5.95534027e-01 -1.46670625e-01 8.32569778e-01 1.90255269e-01 6.97137415e-01 -6.17834866e-01 -5.46476364e-01 -1.10425210e+00 2.06775710e-01 -3.59560430e-01 -9.71775427e-02 -6.06957793e-01 -1.26802981e+00 5.30781150e-01 1.54881448e-01 -1.21482301e+00 4.18163091e-01 -7.36357391e-01 -5.77431679e-01 7.79828429e-01 -1.81158757e+00 -1.60721207e+00 -5.16564548e-01 6.28928363e-01 5.55564463e-01 1.67687625e-01 3.03430706e-01 2.64098912e-01 -2.38162428e-01 2.48964563e-01 -3.81551921e-01 -5.29473387e-02 4.32775110e-01 -1.14237821e+00 7.80602694e-01 2.85063714e-01 -1.47812562e-02 6.03980064e-01 -9.64248702e-02 -6.75694823e-01 -1.38478041e+00 -1.09191585e+00 4.45157409e-01 -4.10837561e-01 3.19218993e-01 -4.12347138e-01 -1.06428027e+00 4.81834173e-01 -1.23463057e-01 3.75693023e-01 2.80189924e-02 1.39548108e-01 -3.73074949e-01 -1.52303457e-01 -1.02908862e+00 2.90287286e-01 1.64879549e+00 -3.63879263e-01 -5.11163473e-01 2.52913088e-01 9.40798402e-01 -8.12295437e-01 -9.67109919e-01 8.21674466e-01 2.52631247e-01 -9.31886494e-01 1.30527067e+00 -1.24193184e-01 4.21750635e-01 -3.75424802e-01 -8.76590014e-02 -1.26568353e+00 -4.55428123e-01 -3.35899323e-01 7.19717145e-02 1.10999334e+00 2.17250273e-01 -8.34654450e-01 9.92575526e-01 3.09504330e-01 -6.35815561e-01 -1.00708246e+00 -9.29944038e-01 -6.96874499e-01 2.72897691e-01 -5.77922165e-01 1.09588194e+00 6.67460501e-01 -6.99749768e-01 1.45897910e-01 2.22009286e-01 2.44230255e-01 6.43991947e-01 5.57799757e-01 7.42554009e-01 -1.45169806e+00 3.89552899e-02 -7.32667685e-01 -5.15966237e-01 -1.90799654e+00 3.01929206e-01 -9.85617101e-01 -1.16380513e-01 -1.67014551e+00 5.70078082e-02 -9.54536200e-01 -2.86554277e-01 6.84860051e-01 1.20211929e-01 2.94832915e-01 2.63839096e-01 1.97700575e-01 -5.92963457e-01 8.70907187e-01 1.92429030e+00 -3.89815897e-01 -3.33776325e-01 6.91892505e-02 -4.05202895e-01 8.21102500e-01 6.64156497e-01 -2.50419647e-01 -5.31291187e-01 -8.89676452e-01 1.61030605e-01 -1.38582170e-01 6.90457046e-01 -8.64416063e-01 3.10019791e-01 -1.82294309e-01 4.69567597e-01 -1.30429208e+00 5.81474721e-01 -8.39538634e-01 -2.51648545e-01 2.46446393e-03 2.70068914e-01 -2.40145102e-01 3.70453596e-01 5.63602686e-01 -4.93373036e-01 1.22921906e-01 5.88503361e-01 -3.33361506e-01 -6.92607343e-01 8.97711933e-01 2.22766683e-01 -2.17240863e-02 8.05780649e-01 -2.63992131e-01 -3.02368313e-01 -2.41613761e-03 -6.81679964e-01 6.72644377e-01 6.68859720e-01 5.31715810e-01 9.09146130e-01 -1.20530701e+00 -5.17157733e-01 5.51112711e-01 -1.85375754e-02 9.10133779e-01 6.51728988e-01 8.96916211e-01 -6.80870533e-01 5.84319413e-01 -1.68693215e-01 -1.21296513e+00 -8.42492640e-01 2.65832126e-01 3.71743977e-01 -8.44030678e-02 -9.96872604e-01 9.75200534e-01 8.10693622e-01 -9.04026151e-01 6.87853545e-02 -7.67279327e-01 -1.63043857e-01 -2.14078367e-01 1.01857847e-02 1.50180250e-01 1.68365702e-01 -5.39492428e-01 -5.77237189e-01 1.40477383e+00 -4.52345610e-02 3.14008176e-01 1.53425694e+00 -2.72488613e-02 -3.10439497e-01 -3.82327996e-02 1.24355972e+00 -1.07078888e-01 -1.56702554e+00 -2.61092126e-01 -4.97295707e-01 -6.13517284e-01 3.15134108e-01 -5.37863135e-01 -1.54172015e+00 1.24870801e+00 2.86827356e-01 8.66473988e-02 1.09294391e+00 3.70164841e-01 1.08863866e+00 3.97266179e-01 6.03238344e-01 -7.77918816e-01 -1.39968783e-01 9.37201262e-01 8.86005223e-01 -1.10540855e+00 -1.24004483e-01 -9.91776168e-01 -2.63355851e-01 1.04888165e+00 7.75001764e-01 -3.87262642e-01 9.34439301e-01 2.05020711e-01 -2.20842466e-01 -3.98969114e-01 -5.51814377e-01 -2.70224720e-01 4.11707371e-01 5.68883657e-01 -1.04065366e-01 6.77039176e-02 2.21935377e-01 4.63433892e-01 -1.48935542e-01 -2.69193113e-01 -1.83104008e-01 6.77820265e-01 -2.43632674e-01 -1.18897271e+00 -2.61835575e-01 3.54556769e-01 -1.51625916e-01 8.29393417e-02 -4.37766492e-01 8.87984276e-01 2.54802167e-01 5.94376445e-01 4.22871262e-01 -3.38657647e-01 4.51162368e-01 -3.42607528e-01 6.81916535e-01 -4.99157101e-01 -4.72977161e-01 4.45928425e-01 -3.16657782e-01 -7.38095462e-01 -4.73204255e-01 -4.67798263e-01 -1.40704751e+00 -2.60505468e-01 -2.62893021e-01 -1.29195675e-01 8.00686479e-01 1.07568157e+00 7.40555346e-01 6.82564795e-01 5.31681716e-01 -1.27832484e+00 -1.49137080e-01 -8.97287190e-01 -3.99919719e-01 5.88631965e-02 1.83553457e-01 -9.65512097e-01 -2.57671326e-01 -5.23155034e-01]
[7.970794677734375, -3.461402177810669]
a6314247-80ef-4fef-8022-cf29fa05cb5c
leveraging-sequence-embedding-and
2112.00344
null
https://arxiv.org/abs/2112.00344v1
https://arxiv.org/pdf/2112.00344v1.pdf
Leveraging Sequence Embedding and Convolutional Neural Network for Protein Function Prediction
The capability of accurate prediction of protein functions and properties is essential in the biotechnology industry, e.g. drug development and artificial protein synthesis, etc. The main challenges of protein function prediction are the large label space and the lack of labeled training data. Our method leverages unsupervised sequence embedding and the success of deep convolutional neural network to overcome these challenges. In contrast, most of the existing methods delete the rare protein functions to reduce the label space. Furthermore, some existing methods require additional bio-information (e.g., the 3-dimensional structure of the proteins) which is difficult to be determined in biochemical experiments. Our proposed method significantly outperforms the other methods on the publicly available benchmark using only protein sequences as input. This allows the process of identifying protein functions to be sped up.
['Min Sun', 'Jia-Hua Wu', 'Po-Han Chi', 'Wei-Cheng Tseng']
2021-12-01
null
null
null
null
['protein-function-prediction']
['medical']
[ 3.74067307e-01 -1.05393670e-01 -2.08215430e-01 -2.60915369e-01 -3.20608079e-01 -9.21531320e-01 2.68856548e-02 3.87036264e-01 -4.65597957e-01 1.28192115e+00 -6.30735140e-03 -4.57687765e-01 8.61673132e-02 -5.64719856e-01 -7.77548254e-01 -9.76880252e-01 1.10310018e-01 3.35904658e-01 1.45494074e-01 -4.87909503e-02 1.67696759e-01 6.51661515e-01 -1.25813413e+00 1.09200869e-02 9.14017200e-01 9.25930798e-01 4.71070290e-01 3.37293774e-01 -3.59167576e-01 3.85251522e-01 -3.36409301e-01 -1.61489457e-01 1.26264304e-01 -3.01349252e-01 -7.18835413e-01 -1.23379685e-01 -2.27512315e-01 -1.01303875e-01 -1.53915048e-01 1.04250646e+00 5.63422084e-01 9.73746553e-02 6.14661813e-01 -9.49748874e-01 -8.66468132e-01 -1.68337915e-02 -2.63122290e-01 -1.07763432e-01 1.58120260e-01 3.09191465e-01 1.11849308e+00 -1.04480517e+00 6.74278677e-01 9.90017295e-01 4.93793100e-01 4.31434333e-01 -1.41240478e+00 -4.69131470e-01 -1.25917614e-01 3.01958084e-01 -1.15155101e+00 -1.88439876e-01 7.94249833e-01 -6.14991665e-01 1.09760356e+00 -1.21915437e-01 4.19870526e-01 9.04889703e-01 9.82564464e-02 5.30334234e-01 8.92992854e-01 -1.52857944e-01 3.46010268e-01 -2.40665168e-01 2.44050354e-01 7.52439618e-01 1.85710683e-01 -7.86196440e-02 -2.03589842e-01 -4.00024980e-01 4.47853863e-01 4.78192240e-01 -4.43128258e-01 -6.76194012e-01 -1.34936678e+00 8.04695308e-01 3.23318183e-01 2.08411977e-01 -4.76845652e-01 -1.19639657e-01 5.22874415e-01 2.09095359e-01 4.34976012e-01 6.94292367e-01 -1.17555118e+00 2.17628609e-02 -5.30651212e-01 6.37543807e-03 8.44725609e-01 6.95012927e-01 1.00603044e+00 -1.89526528e-01 2.42436528e-01 7.37506568e-01 1.43424913e-01 -2.98768170e-02 5.61198890e-01 -5.67091286e-01 4.79819030e-02 9.23525453e-01 4.14694399e-01 -6.70309186e-01 -5.00256717e-01 -5.73404655e-02 -5.72294831e-01 1.81423038e-01 6.77769065e-01 -5.99938445e-02 -9.39138412e-01 1.71169662e+00 3.60252470e-01 1.67130530e-01 3.25558662e-01 9.00112629e-01 7.28935897e-01 6.71898723e-01 1.32528976e-01 -2.37431377e-01 1.18784499e+00 -9.87589836e-01 -6.73689485e-01 1.93667814e-01 7.79655993e-01 -7.22823441e-01 1.04238832e+00 8.27126130e-02 -5.57598948e-01 -3.71396244e-01 -1.24315417e+00 -2.90647686e-01 -7.34046161e-01 2.98621595e-01 9.83794808e-01 2.78088629e-01 -5.88807702e-01 8.14013481e-01 -8.63224149e-01 -3.79519612e-01 4.92260456e-01 7.50425220e-01 -7.97626793e-01 -1.28803164e-01 -1.21141171e+00 8.25631320e-01 7.53624618e-01 -6.94629550e-02 -6.82813168e-01 -6.90955281e-01 -8.06732178e-01 3.43878686e-01 4.44220930e-01 -2.00785309e-01 9.61532235e-01 -7.33470738e-01 -1.53462410e+00 7.53129125e-01 -2.60507762e-01 -2.89993376e-01 1.97008178e-01 -1.72304779e-01 -6.15046546e-02 2.03321621e-01 -1.23751268e-01 5.77215850e-01 2.69543946e-01 -7.63613403e-01 -3.53550017e-01 -3.42004836e-01 2.79575646e-01 -7.41640478e-02 -2.79586077e-01 -1.38483495e-01 -1.23781107e-01 -3.94449413e-01 9.38685760e-02 -9.18870270e-01 -4.76583004e-01 2.52516985e-01 -1.71984509e-01 -4.35781091e-01 8.28832030e-01 -5.83269954e-01 6.99008644e-01 -2.03510308e+00 3.92956495e-01 -1.38550118e-01 5.45327365e-01 3.66182089e-01 -5.98042086e-02 6.62104726e-01 -3.60747635e-01 1.22345667e-02 -1.66368008e-01 3.34461778e-01 -1.13071963e-01 1.31916329e-01 -1.42794866e-02 6.35827363e-01 3.97005379e-01 9.76011395e-01 -8.99465919e-01 -3.29997361e-01 2.02646971e-01 4.97030199e-01 -3.24674726e-01 2.55771935e-01 -4.15528208e-01 6.54693723e-01 -6.11246765e-01 6.15223527e-01 6.07448816e-01 -7.69873321e-01 5.41305542e-01 -4.66798007e-01 -7.86130577e-02 4.34515297e-01 -6.02701008e-01 1.61288321e+00 -1.99926645e-03 1.63921744e-01 -2.04567388e-01 -1.32139039e+00 9.10340786e-01 5.35218060e-01 1.01369715e+00 -5.15147969e-02 6.13222830e-02 1.87071085e-01 3.33420843e-01 -4.07855809e-01 1.07235312e-02 -4.58044969e-02 2.94221282e-01 4.88215595e-01 1.20501027e-01 6.72838986e-01 2.11796194e-01 -7.15609416e-02 1.15209579e+00 5.41250050e-01 5.47753453e-01 -2.90874571e-01 7.39619553e-01 2.35200286e-01 1.16080785e+00 3.01631689e-02 -3.57299417e-01 3.54302436e-01 7.15117574e-01 -7.21985638e-01 -1.38686109e+00 -6.07608795e-01 -2.66253836e-02 1.15481389e+00 1.45920306e-01 -4.47561562e-01 -6.87553644e-01 -9.85837936e-01 1.40151352e-01 2.80034300e-02 -4.82060522e-01 -3.37042779e-01 -5.83381176e-01 -9.17030334e-01 3.65853488e-01 4.28868115e-01 -5.52948490e-02 -1.08215988e+00 -3.80372852e-01 5.77710032e-01 -5.06884269e-02 -9.69750583e-01 -4.20271397e-01 5.98684609e-01 -8.28389347e-01 -1.31119251e+00 -6.74437106e-01 -1.10364938e+00 7.16668546e-01 2.24426076e-01 7.42436230e-01 9.38500986e-02 -3.24999720e-01 -4.34815109e-01 -3.18696439e-01 -3.77209634e-01 -2.64146090e-01 1.48293406e-01 1.74284071e-01 -6.46347404e-02 8.32111359e-01 -7.27835000e-01 -8.85163665e-01 4.52118486e-01 -8.99801910e-01 1.37438446e-01 6.18506551e-01 1.27152562e+00 8.28977108e-01 9.23779309e-02 1.00804651e+00 -1.04972529e+00 4.96456534e-01 -4.73747402e-01 -6.53676629e-01 4.98128176e-01 -7.08699703e-01 4.84100342e-01 9.68797505e-01 -5.77057242e-01 -6.90185189e-01 6.88287735e-01 -2.63893515e-01 -1.43238410e-01 -2.03127712e-01 5.09372294e-01 -4.02182013e-01 -1.74340278e-01 4.26395983e-01 3.26636493e-01 2.41385415e-01 -8.21878135e-01 1.11823939e-01 5.09782910e-01 2.01797843e-01 -4.53009546e-01 3.04794371e-01 1.01576090e-01 1.27475157e-01 -4.76038635e-01 -6.06153727e-01 -5.30818641e-01 -7.93756127e-01 4.32718754e-01 7.37097621e-01 -6.07418716e-01 -1.05260444e+00 4.19670306e-02 -1.07726574e+00 2.02838853e-01 2.65599668e-01 4.99640316e-01 -6.23246849e-01 7.10933506e-01 -8.06331456e-01 -4.73724931e-01 -4.64809746e-01 -1.43289459e+00 8.16063106e-01 5.12197241e-02 -7.04278797e-02 -6.50905967e-01 -9.22889337e-02 2.63110936e-01 1.33375913e-01 3.06838572e-01 1.45574367e+00 -9.52124715e-01 -4.69219208e-01 -4.98421118e-02 -2.38681272e-01 2.72258580e-01 6.46639287e-01 -1.97631493e-01 -8.07785809e-01 -3.66041481e-01 -2.32846454e-01 -6.22291267e-01 8.73344779e-01 2.78442144e-01 1.12505782e+00 -5.88640347e-02 -4.48906898e-01 5.12794971e-01 1.30278850e+00 3.98601681e-01 4.80187416e-01 1.20256186e-01 7.16233373e-01 8.08106303e-01 6.83603227e-01 1.74417645e-01 -3.60457902e-03 5.55623174e-01 3.99063826e-01 -1.87323600e-01 2.58385718e-01 -1.68433815e-01 9.55200121e-02 6.31832778e-01 -4.03257944e-02 -2.56246626e-01 -8.42712998e-01 3.26379567e-01 -2.15178156e+00 -6.53325438e-01 1.01300955e-01 2.01349640e+00 1.12622154e+00 -6.27836734e-02 -1.71526417e-01 7.36412629e-02 8.08033586e-01 -2.84632236e-01 -1.13081706e+00 -1.29430741e-02 -1.77862793e-01 1.82297856e-01 5.22429407e-01 2.36320496e-01 -1.19216371e+00 1.02965391e+00 6.60742807e+00 4.81919914e-01 -1.04957509e+00 -1.72406480e-01 5.13800502e-01 2.41619751e-01 -2.40206979e-02 1.26770232e-02 -6.73396409e-01 6.52517736e-01 9.02914345e-01 -8.70734006e-02 3.35193038e-01 9.46529388e-01 1.71800882e-01 2.30471820e-01 -1.19352281e+00 9.08494771e-01 -3.75900358e-01 -1.47801709e+00 -2.35179499e-01 1.49801403e-01 3.74368876e-01 6.58865124e-02 -3.48445982e-01 1.24983907e-01 3.46861660e-01 -1.18117380e+00 3.56062166e-02 3.02622437e-01 8.49150300e-01 -8.39354336e-01 8.54177475e-01 2.85781294e-01 -9.67872977e-01 7.54583254e-02 -6.90081596e-01 3.76925059e-02 -3.15858945e-02 6.22398973e-01 -1.02610421e+00 3.45266908e-01 4.25819933e-01 7.20962524e-01 -2.18291759e-01 8.17320228e-01 -1.92830205e-01 6.86104476e-01 -2.30837733e-01 -8.81452411e-02 1.84815049e-01 -4.83440220e-01 6.19696490e-02 9.18672383e-01 -3.35934572e-02 1.27054736e-01 5.73057652e-01 8.46491694e-01 -4.47933435e-01 4.27913487e-01 -4.84591097e-01 -7.52133608e-01 1.51567265e-01 1.28365755e+00 -6.90029204e-01 -2.32544616e-01 -6.71721101e-01 1.05558419e+00 5.79370975e-01 4.11571205e-01 -6.27343714e-01 -5.13108253e-01 1.05703211e+00 -1.43809155e-01 2.91871756e-01 -3.66578728e-01 1.06255770e-01 -1.18091083e+00 -1.19792916e-01 -8.29086900e-01 1.03992879e-01 -4.30765241e-01 -1.39393938e+00 3.09834301e-01 -5.95217168e-01 -1.07136452e+00 -1.12330593e-01 -1.02566588e+00 -1.03238963e-01 1.00671232e+00 -1.68208098e+00 -1.01861572e+00 1.47453949e-01 1.35001287e-01 4.61785734e-01 -2.85153478e-01 1.11970544e+00 4.47203457e-01 -6.45886004e-01 4.88418996e-01 5.87938249e-01 -1.84888076e-02 8.44066024e-01 -1.29992747e+00 4.05917078e-01 3.95726502e-01 -2.15785265e-01 8.98707747e-01 6.87421203e-01 -5.96432567e-01 -1.39669728e+00 -1.00537658e+00 9.72207367e-01 -2.34763950e-01 6.23946786e-01 -6.12250268e-01 -1.15432405e+00 4.10831064e-01 -2.08686113e-01 2.13101104e-01 1.20782185e+00 -7.80506954e-02 -2.98511684e-01 1.38898641e-01 -1.16234219e+00 4.62812811e-01 8.61642480e-01 -4.31087106e-01 -4.27206099e-01 5.87919176e-01 1.11091578e+00 9.35696080e-05 -1.00497091e+00 4.48168278e-01 6.65837109e-01 -4.59930182e-01 9.93292689e-01 -1.31446397e+00 3.38963509e-01 -6.52235568e-01 -7.98538104e-02 -1.02508378e+00 -6.34511173e-01 -4.26729262e-01 -1.60058096e-01 8.66155267e-01 5.73957860e-01 -6.60567403e-01 9.43223357e-01 5.23283124e-01 -6.52697980e-02 -1.03497338e+00 -4.67152685e-01 -5.06389439e-01 -2.33553257e-02 2.81387091e-01 7.52768397e-01 1.01321530e+00 -2.72736885e-02 5.43217778e-01 -5.25664032e-01 -8.50218311e-02 1.83229402e-01 5.34654915e-01 4.83899206e-01 -1.47468865e+00 -3.36984724e-01 3.72145362e-02 -5.04048944e-01 -1.08382988e+00 4.22888130e-01 -9.74191189e-01 -1.46684229e-01 -1.46554029e+00 2.96638578e-01 -2.25650996e-01 -9.50905323e-01 7.52781272e-01 -2.81779706e-01 1.19163692e-01 -3.36620748e-01 2.13427350e-01 -5.58439314e-01 6.91987634e-01 1.13312018e+00 -1.61520034e-01 -5.62528744e-02 -2.96824515e-01 -7.67367482e-01 5.90553105e-01 1.05247092e+00 -5.67079663e-01 -3.76313865e-01 4.13472056e-02 3.74756344e-02 -1.06075190e-01 -3.59395221e-02 -4.93122578e-01 -4.93554957e-02 -4.70069498e-01 5.42349398e-01 -6.06426597e-01 3.31157923e-01 -9.89129245e-01 5.13721593e-02 6.28636181e-01 -2.52573848e-01 3.39240171e-02 -6.42968938e-02 8.33893776e-01 -1.59859329e-01 -1.46523923e-01 7.61961102e-01 -2.32676983e-01 -6.37856781e-01 4.33394611e-01 -2.78061211e-01 -3.74348044e-01 1.07342327e+00 -2.03444168e-01 -1.47774905e-01 5.70904016e-02 -5.55892706e-01 2.59416431e-01 7.23597229e-01 2.91540593e-01 5.51913261e-01 -1.22915149e+00 -3.83331746e-01 1.96810961e-01 3.57781231e-01 -2.17740059e-01 -3.26202065e-02 6.37733459e-01 -8.15519691e-01 6.75383687e-01 -4.73503649e-01 -3.29267144e-01 -1.33961976e+00 9.41746593e-01 6.64631873e-02 -2.79273450e-01 -6.13071382e-01 4.40934747e-01 6.52741075e-01 -6.19441032e-01 5.77701218e-02 4.85124178e-02 -3.33838195e-01 -2.63079405e-01 4.68308926e-01 7.48495832e-02 7.61635378e-02 -5.58621287e-01 -3.87335718e-01 1.81247741e-01 -4.22869444e-01 6.31981552e-01 1.68187690e+00 1.08477190e-01 -3.95510972e-01 2.63425052e-01 1.43445945e+00 -3.53637785e-01 -1.35007668e+00 -3.67060363e-01 3.69039714e-01 -1.95066661e-01 -3.00851226e-01 -7.99904287e-01 -6.99880600e-01 8.65408719e-01 5.87625384e-01 -3.60925972e-01 8.80566597e-01 -1.47678658e-01 1.00928271e+00 8.63216996e-01 4.31229502e-01 -1.01728976e+00 -1.71473831e-01 4.87385362e-01 4.05467421e-01 -1.41065145e+00 -9.77960415e-03 -3.66876066e-01 -3.99182379e-01 1.10880113e+00 5.13644159e-01 1.33406758e-01 5.59488177e-01 1.78799301e-01 8.46503526e-02 -1.80106550e-01 -8.79722774e-01 -2.05479577e-01 1.57363653e-01 4.30693924e-01 9.69634414e-01 -5.37327752e-02 -8.88276100e-01 6.73882306e-01 2.65544116e-01 -6.26880750e-02 7.11511225e-02 1.08235574e+00 -5.23619056e-01 -1.58546042e+00 7.90214986e-02 5.33683717e-01 -9.26375687e-01 -1.89523518e-01 -7.67886281e-01 2.06571624e-01 9.22914594e-02 8.09904873e-01 -5.51983356e-01 -2.12375879e-01 4.27120179e-03 5.12368262e-01 2.02228874e-01 -6.84252083e-01 -1.93931982e-01 1.30236924e-01 -7.12759048e-02 -3.57017934e-01 -4.17115837e-01 -2.09033966e-01 -1.64701498e+00 -1.05378404e-01 -5.55537105e-01 3.41675937e-01 5.72409987e-01 7.65328705e-01 7.57196724e-01 3.95627677e-01 4.39540684e-01 -4.64604467e-01 -5.64223170e-01 -9.17828560e-01 -6.20015979e-01 7.86551833e-01 2.53269106e-01 -9.11948383e-01 1.07693657e-01 3.58664840e-01]
[4.74404239654541, 5.64155387878418]
0cee356a-2a4d-45bb-bed4-f846fb22daf5
efficient-zero-shot-event-extraction-with
2211.05156
null
https://arxiv.org/abs/2211.05156v2
https://arxiv.org/pdf/2211.05156v2.pdf
Efficient Zero-shot Event Extraction with Context-Definition Alignment
Event extraction (EE) is the task of identifying interested event mentions from text. Conventional efforts mainly focus on the supervised setting. However, these supervised models cannot generalize to event types out of the pre-defined ontology. To fill this gap, many efforts have been devoted to the zero-shot EE problem. This paper follows the trend of modeling event-type semantics but moves one step further. We argue that using the static embedding of the event type name might not be enough because a single word could be ambiguous, and we need a sentence to define the type semantics accurately. To model the definition semantics, we use two separate transformer models to project the contextualized event mentions and corresponding definitions into the same embedding space and then minimize their embedding distance via contrastive learning. On top of that, we also propose a warming phase to help the model learn the minor difference between similar definitions. We name our approach Zero-shot Event extraction with Definition (ZED). Experiments on the MAVEN dataset show that our model significantly outperforms all previous zero-shot EE methods with fast inference speed due to the disjoint design. Further experiments also show that ZED can be easily applied to the few-shot setting when the annotation is available and consistently outperforms baseline supervised methods.
['Dong Yu', 'Wenlin Yao', 'Hongming Zhang']
2022-11-09
null
null
null
null
['event-extraction']
['natural-language-processing']
[ 3.20729494e-01 3.84370118e-01 -4.68248576e-01 -5.32031894e-01 -8.28796327e-01 -5.68777561e-01 6.42670870e-01 5.86034477e-01 -6.06665969e-01 6.08281434e-01 4.77426231e-01 -2.32500091e-01 -2.16542378e-01 -1.07537067e+00 -6.75299048e-01 -4.71511781e-01 6.65433146e-03 4.08550322e-01 4.89525288e-01 -1.72089070e-01 -8.66323784e-02 8.31944216e-03 -1.38411486e+00 3.09106916e-01 4.94651049e-01 7.03633189e-01 1.34917960e-01 1.88596532e-01 -4.47946668e-01 9.20780957e-01 -4.35461938e-01 -5.90998113e-01 1.35425135e-01 -4.26694542e-01 -1.11927617e+00 -2.01450959e-01 2.92949341e-02 -4.37553786e-02 -2.27767363e-01 8.08998644e-01 4.68370467e-01 4.15063113e-01 5.69136322e-01 -1.37817121e+00 -2.50785798e-01 1.17525220e+00 -2.91405916e-01 3.66006613e-01 3.51957172e-01 -4.09335136e-01 1.54191303e+00 -1.10398483e+00 9.29934323e-01 1.05831385e+00 8.08576703e-01 4.15911585e-01 -1.22007823e+00 -5.17834842e-01 4.07976866e-01 5.33676565e-01 -1.34216774e+00 -2.58310556e-01 7.24542677e-01 -2.79506177e-01 1.34037197e+00 2.22632542e-01 5.13241172e-01 1.25000906e+00 -2.12141499e-01 8.39679897e-01 7.32457817e-01 -6.49804056e-01 3.80225986e-01 9.87493768e-02 7.13982821e-01 5.22094548e-01 2.44827598e-01 -1.24590978e-01 -5.61599135e-01 -2.53385067e-01 1.54165193e-01 2.28930768e-02 -2.09025234e-01 -3.48223507e-01 -1.25831521e+00 9.32551324e-01 6.45662099e-02 4.40270096e-01 -2.63512850e-01 4.62182797e-02 6.79345489e-01 1.95011005e-01 5.54437459e-01 6.67750061e-01 -7.93762028e-01 -3.02242428e-01 -8.88971746e-01 3.76394063e-01 1.03063631e+00 1.09382367e+00 7.45005965e-01 -4.78882432e-01 -2.20505297e-01 8.55798542e-01 1.01770118e-01 -9.59586650e-02 2.44307578e-01 -6.34949446e-01 4.49773669e-01 6.35075629e-01 1.32263973e-01 -6.97466493e-01 -4.68428940e-01 -1.92169622e-01 -3.41504604e-01 -3.03624570e-01 4.14570361e-01 -2.66020656e-01 -6.83135331e-01 1.88187230e+00 4.30422455e-01 6.95183814e-01 8.91621187e-02 6.18661523e-01 9.11055803e-01 5.93712747e-01 2.91471511e-01 -1.70329601e-01 1.76882076e+00 -8.36990178e-01 -9.86445308e-01 -5.26882887e-01 8.60672414e-01 -2.74928510e-01 1.00426447e+00 -8.71780440e-02 -7.99607217e-01 -6.93792989e-03 -1.19409466e+00 -2.43854001e-01 -8.36217940e-01 -1.67945236e-01 6.61292255e-01 4.52375501e-01 -3.13521951e-01 6.30029380e-01 -9.50546205e-01 -4.78693932e-01 3.14158916e-01 9.47591588e-02 -1.53487384e-01 6.24562427e-02 -1.88642347e+00 1.12297082e+00 7.42955327e-01 -2.20717967e-01 -4.46950316e-01 -1.00654757e+00 -1.25223756e+00 2.62876451e-01 9.82454777e-01 -6.26585186e-01 1.39224195e+00 -5.44576883e-01 -9.26438868e-01 7.60019779e-01 -5.10160923e-01 -7.34344900e-01 -4.93987016e-02 -2.76407033e-01 -4.20083910e-01 -1.03064189e-02 3.98281723e-01 1.37810484e-01 2.70014077e-01 -1.10794222e+00 -8.53897214e-01 -7.26206973e-02 4.83299047e-01 -3.59939821e-02 -4.84775603e-01 2.43561998e-01 -3.62826794e-01 -7.72154033e-01 -5.48303537e-02 -6.75359845e-01 -1.68538660e-01 -3.40143114e-01 -2.69731194e-01 -5.75052679e-01 5.23611724e-01 -4.24963534e-01 1.76187658e+00 -2.07106447e+00 1.54762968e-01 6.23285361e-02 2.28173122e-01 -8.44522119e-02 2.66761392e-01 6.47526920e-01 -2.55721509e-01 6.78891838e-02 -5.29760003e-01 -3.83916795e-01 3.89060050e-01 4.88839984e-01 -5.78569591e-01 1.79160386e-01 3.05505425e-01 9.04184878e-01 -1.39658511e+00 -6.34499431e-01 8.74663796e-03 2.07755372e-01 -4.53092903e-01 8.35975111e-02 -2.46527448e-01 -3.06705028e-01 -4.21675861e-01 3.87257963e-01 4.06067610e-01 -3.02145898e-01 4.71981615e-01 -4.06734586e-01 2.24213097e-02 8.18398058e-01 -1.43206525e+00 1.65054035e+00 -5.78749955e-01 4.61838841e-01 -3.83173645e-01 -1.24913299e+00 4.76667672e-01 5.48831165e-01 6.37012661e-01 -2.08175868e-01 1.71446800e-02 1.18068792e-01 -2.22358108e-01 -5.26334524e-01 4.45474267e-01 -4.73979175e-01 -4.36652809e-01 3.82825226e-01 3.25367272e-01 1.99611485e-01 4.44420695e-01 3.37312609e-01 1.28081119e+00 2.25614503e-01 6.60780191e-01 -1.00849584e-01 1.14195406e-01 7.10545853e-02 1.02482200e+00 8.76085162e-01 -2.34677233e-02 5.26535988e-01 6.17724419e-01 -3.44411969e-01 -6.69827044e-01 -1.20470226e+00 -2.89152503e-01 1.02989113e+00 3.41729611e-01 -1.22407806e+00 -4.59521562e-01 -1.10552180e+00 -2.09858850e-01 1.19810069e+00 -6.72536612e-01 -7.75446221e-02 -6.90175891e-01 -9.78657603e-01 5.25400698e-01 7.19818175e-01 2.99553961e-01 -8.53008091e-01 -7.32767165e-01 5.41344106e-01 -5.00700891e-01 -1.27889061e+00 -3.01755697e-01 5.32927871e-01 -3.20476443e-01 -1.13050830e+00 -1.53561562e-01 -7.73977757e-01 2.43328750e-01 -1.90617412e-01 1.36652172e+00 -2.82635629e-01 -2.61232965e-02 3.25325400e-01 -6.63795710e-01 -6.15600944e-01 -5.96530996e-02 1.21949136e-01 -1.14760473e-01 1.37587776e-02 1.02162504e+00 -5.81764162e-01 -2.29897246e-01 4.30325158e-02 -9.12839353e-01 7.44411349e-02 1.53259933e-01 8.65840137e-01 5.66970527e-01 2.18410403e-01 8.44396174e-01 -1.26428032e+00 3.47509503e-01 -8.69527757e-01 -2.12996528e-01 4.44082171e-01 -8.35529208e-01 1.79614127e-01 5.31247318e-01 -5.34530103e-01 -1.20479000e+00 6.88657314e-02 -5.53600714e-02 -1.14511847e-01 -6.72806278e-02 7.33110547e-01 -4.65065569e-01 7.00119257e-01 4.05967236e-01 -2.35466147e-03 -5.70396960e-01 -4.42235827e-01 4.64448363e-01 6.02958739e-01 3.20500553e-01 -5.80858946e-01 6.88158453e-01 4.59867209e-01 -2.21798509e-01 -6.10407174e-01 -1.35627139e+00 -6.85228050e-01 -6.20443225e-01 1.14819869e-01 8.70907426e-01 -8.31429005e-01 -5.53276204e-02 -7.75768459e-02 -1.26760113e+00 -2.26619646e-01 -5.74660540e-01 5.04905045e-01 -3.86379510e-01 3.96821469e-01 -6.26410782e-01 -6.91827595e-01 -1.33065149e-01 -7.80727684e-01 1.20124972e+00 -3.77770653e-03 -6.10252142e-01 -1.11613667e+00 1.82142630e-01 -2.07789421e-01 1.27898023e-01 3.16079527e-01 1.01687157e+00 -9.93286073e-01 -2.76452601e-01 -9.39905643e-02 -3.86906900e-02 -1.34167448e-01 2.36792579e-01 -4.86408025e-01 -8.34453523e-01 2.19451904e-01 -3.45654064e-03 -1.17673732e-01 9.91474986e-01 1.63519636e-01 9.43858325e-01 -3.02049756e-01 -6.27950311e-01 4.78741288e-01 1.40371180e+00 -1.43239126e-02 4.81205791e-01 5.43298542e-01 5.88697612e-01 5.21826804e-01 8.17124248e-01 5.18893063e-01 6.99483335e-01 8.55479181e-01 5.74121326e-02 1.30548123e-02 -4.03987765e-02 -4.74737436e-01 3.18678886e-01 4.44534272e-01 2.74655849e-01 -2.73924023e-01 -7.96557546e-01 9.91120875e-01 -1.92618442e+00 -1.32898426e+00 -8.39658156e-02 1.97065568e+00 1.23235476e+00 2.59748787e-01 5.48973344e-02 2.01974362e-01 5.82309783e-01 2.58766979e-01 -1.71599403e-01 -2.47444212e-01 -1.14947028e-01 4.93007421e-01 4.54229653e-01 4.87473011e-01 -1.30369914e+00 8.71227682e-01 5.61047220e+00 8.66778135e-01 -5.52486598e-01 4.13124651e-01 8.89042392e-02 -9.50564146e-02 -4.31030810e-01 5.15013874e-01 -1.13520133e+00 5.92943966e-01 9.92316782e-01 -4.55360889e-01 9.21750534e-03 5.87485194e-01 8.27838928e-02 7.81166628e-02 -1.50110245e+00 7.93524206e-01 7.87856206e-02 -1.29303086e+00 -1.79144531e-01 -1.77586183e-01 3.65259320e-01 -9.71943364e-02 -6.73002779e-01 6.61939263e-01 2.88949668e-01 -7.35645056e-01 6.19648576e-01 4.04026747e-01 5.23935914e-01 -6.17290556e-01 8.76739860e-01 3.01464170e-01 -1.57405365e+00 -5.53516299e-02 -2.77700812e-01 -1.11941472e-01 4.85544115e-01 7.26167500e-01 -7.74380744e-01 8.34275305e-01 5.77558160e-01 9.99247253e-01 -2.95670211e-01 8.34526598e-01 -4.84273970e-01 9.39832985e-01 -4.83886421e-01 -9.49166119e-02 9.86498743e-02 1.00507848e-01 6.75816834e-01 1.50067258e+00 1.04326822e-01 3.21603328e-01 2.46719420e-01 9.42917347e-01 -8.05780217e-02 5.78839965e-02 -6.09337986e-01 1.16360374e-01 7.18688250e-01 1.07604671e+00 -7.40902483e-01 -6.30637467e-01 -7.64610350e-01 9.14486468e-01 3.94467622e-01 1.97603852e-01 -1.09715915e+00 -7.64369130e-01 5.99736929e-01 -7.06540421e-02 5.15365541e-01 7.15476796e-02 -2.39592001e-01 -1.43655717e+00 1.37095064e-01 -4.81404454e-01 8.39622557e-01 -4.31867748e-01 -1.41008651e+00 3.19070905e-01 5.32210767e-01 -1.12184715e+00 -3.81094903e-01 -3.43519270e-01 -6.28655314e-01 4.37660545e-01 -1.46690536e+00 -1.03665578e+00 1.38601840e-01 2.33036876e-01 7.31359065e-01 3.44142646e-01 1.09534323e+00 5.31557560e-01 -6.05960548e-01 6.11347497e-01 -2.99619108e-01 3.88634950e-01 7.48788476e-01 -1.52775133e+00 2.64615595e-01 1.05902588e+00 4.25917596e-01 7.00613976e-01 8.90105367e-01 -7.09255993e-01 -1.21064007e+00 -1.24364924e+00 1.58822691e+00 -7.11289287e-01 8.29334199e-01 -6.93997383e-01 -9.42464232e-01 1.00760365e+00 1.77547455e-01 -8.14396422e-03 9.11784589e-01 5.27944922e-01 -5.31263113e-01 5.47265857e-02 -8.15503776e-01 6.08164907e-01 1.22215128e+00 -5.19181430e-01 -1.47018886e+00 4.49661076e-01 9.31603014e-01 -1.86083809e-01 -9.33080614e-01 4.02199984e-01 3.92463714e-01 -3.98357153e-01 9.12049294e-01 -7.47395039e-01 3.25831920e-01 -3.75094205e-01 -1.63463667e-01 -1.13945806e+00 -9.44617316e-02 -4.45022583e-01 -4.04652864e-01 1.57326627e+00 6.84454739e-01 -4.14065450e-01 3.21290016e-01 7.57230937e-01 2.60179136e-02 -7.66711116e-01 -9.07956064e-01 -9.23789263e-01 -9.24093202e-02 -7.80064106e-01 7.36514151e-01 1.21705604e+00 5.93218565e-01 7.68177271e-01 -3.36539894e-01 3.03322107e-01 5.26053190e-01 4.05243278e-01 3.45176846e-01 -1.16983473e+00 -4.62251335e-01 -1.64070204e-01 -2.72491097e-01 -9.19044614e-01 4.48866189e-01 -1.19660366e+00 1.45886466e-01 -1.72225535e+00 4.13259894e-01 -3.00678730e-01 -4.91245627e-01 9.55345690e-01 -4.72033888e-01 -1.21368617e-01 -5.25740758e-02 -2.32859731e-01 -7.37645864e-01 4.17651772e-01 4.22829390e-01 -1.62656203e-01 -1.43365070e-01 -1.86584115e-01 -7.12980390e-01 7.57145703e-01 5.97151518e-01 -7.67140508e-01 -4.20153052e-01 -3.10962170e-01 3.95101547e-01 -2.16847479e-01 3.69615287e-01 -5.29051661e-01 2.62172580e-01 -8.22610036e-02 -2.52110869e-01 -3.67532879e-01 2.19540223e-01 -9.54222381e-01 -1.71077847e-02 4.88603748e-02 -5.70634305e-01 -2.64221817e-01 2.31990181e-02 6.08269334e-01 -2.55553842e-01 -5.37579298e-01 3.46230775e-01 3.07709537e-02 -1.01336110e+00 1.78859845e-01 -3.04905474e-01 4.31135833e-01 9.00397301e-01 -1.08303837e-02 3.12349517e-02 -1.05181746e-01 -8.63942266e-01 3.08614552e-01 2.47430220e-01 4.86018777e-01 5.04638910e-01 -1.37683320e+00 -7.11618960e-01 -7.85747617e-02 3.96988988e-01 -8.89545213e-03 -1.00028373e-01 8.26151013e-01 2.28183195e-01 1.15404829e-01 4.27609473e-01 -1.67938963e-01 -1.12683606e+00 6.08831823e-01 2.77270764e-01 -5.52423298e-01 -1.03733683e+00 6.57184839e-01 2.08936006e-01 -4.31701839e-01 2.29265481e-01 -3.74041826e-01 -2.41674110e-01 3.30654085e-01 7.13664711e-01 2.27328911e-01 5.21765575e-02 -3.65306258e-01 -5.95885217e-01 2.73043782e-01 -5.97618148e-02 -1.50010422e-01 1.67168176e+00 -1.90923154e-01 3.50753590e-02 7.88278997e-01 1.24780393e+00 1.39624253e-01 -9.13148642e-01 -4.55017626e-01 6.58394635e-01 -1.77692875e-01 -2.80280728e-02 -6.37783825e-01 -7.30187476e-01 5.37012517e-01 1.29179731e-01 2.31749520e-01 1.09671175e+00 3.37682426e-01 1.00376391e+00 3.43413919e-01 3.17429334e-01 -1.18715096e+00 -4.10026044e-01 6.45685852e-01 4.76361901e-01 -1.02668583e+00 -6.68626130e-02 -6.51620686e-01 -5.33613443e-01 9.83567595e-01 3.98744851e-01 -1.47618473e-01 7.34348595e-01 5.91638207e-01 -3.59421313e-01 -3.77839833e-01 -8.77045870e-01 -5.55232942e-01 1.87455967e-01 2.90096790e-01 4.50423717e-01 -1.09006770e-01 -6.97238624e-01 1.23670375e+00 -1.01274699e-01 7.64682442e-02 4.67375904e-01 1.05550253e+00 -2.37432674e-01 -1.34833169e+00 9.49079320e-02 6.54008329e-01 -6.96191728e-01 -4.17338371e-01 -1.83727548e-01 8.31063509e-01 3.65662515e-01 1.07342923e+00 -5.30095398e-02 -1.94812268e-01 5.41178048e-01 6.45558953e-01 3.56109887e-01 -9.23440814e-01 -5.44408202e-01 -1.39690533e-01 4.89166617e-01 -5.54562271e-01 -4.94679868e-01 -7.87482083e-01 -1.51260388e+00 2.87877440e-01 -4.61934984e-01 3.13437313e-01 1.71181574e-01 1.28887737e+00 2.57017702e-01 5.98493397e-01 3.98323476e-01 -2.42641956e-01 -5.12952924e-01 -7.95629323e-01 -5.26219130e-01 6.74169123e-01 1.55295819e-01 -9.23337400e-01 -6.16089463e-01 1.44180328e-01]
[9.149552345275879, 9.104650497436523]
8d59c634-756f-4ce7-818d-ebc481e784de
analysis-of-chatgpt-on-source-code
2306.00597
null
https://arxiv.org/abs/2306.00597v2
https://arxiv.org/pdf/2306.00597v2.pdf
Analysis of ChatGPT on Source Code
This paper explores the use of Large Language Models (LLMs) and in particular ChatGPT in programming, source code analysis, and code generation. LLMs and ChatGPT are built using machine learning and artificial intelligence techniques, and they offer several benefits to developers and programmers. While these models can save time and provide highly accurate results, they are not yet advanced enough to replace human programmers entirely. The paper investigates the potential applications of LLMs and ChatGPT in various areas, such as code creation, code documentation, bug detection, refactoring, and more. The paper also suggests that the usage of LLMs and ChatGPT is expected to increase in the future as they offer unparalleled benefits to the programming community.
['Ahmed R. Sadik', 'Jibesh Patra', 'Frank Joublin', 'Antonello Ceravola']
2023-06-01
null
null
null
null
['code-generation']
['computer-code']
[-4.58040684e-01 3.76905620e-01 -2.67023593e-01 -2.06452414e-01 -4.93123472e-01 -2.97469497e-01 2.31742248e-01 6.37689650e-01 3.13691765e-01 3.24482560e-01 -1.07209496e-01 -6.63610816e-01 -4.74540628e-02 -6.00958288e-01 -3.60478818e-01 1.38772994e-01 -2.76001066e-01 2.76654750e-01 1.55113950e-01 -1.51730835e-01 8.71519566e-01 -6.37364388e-02 -1.51056242e+00 3.99722964e-01 1.18409419e+00 1.52497396e-01 3.19595128e-01 7.31227338e-01 -7.82497883e-01 1.37052453e+00 -7.73739636e-01 -4.87060964e-01 -1.89952284e-01 -3.17096651e-01 -9.05905783e-01 -2.26581827e-01 8.58174264e-02 2.06057318e-02 4.60714430e-01 1.11333394e+00 1.19329564e-01 -4.30186659e-01 1.97281558e-02 -1.44329524e+00 -6.51787996e-01 9.16356802e-01 -7.68437028e-01 -7.86272138e-02 7.58548558e-01 -1.18345199e-02 7.40464866e-01 -5.72337151e-01 5.43632388e-01 1.20806515e+00 1.02157271e+00 4.33750957e-01 -1.05635798e+00 -4.35870051e-01 -1.69503734e-01 1.36777675e-02 -1.09888589e+00 -1.33731924e-02 5.80035508e-01 -1.07765496e+00 1.63320684e+00 2.76474297e-01 6.44356191e-01 3.95929426e-01 6.76629364e-01 7.61200964e-01 5.09111345e-01 -1.20918310e+00 2.22173363e-01 5.94754994e-01 4.66580808e-01 1.13864613e+00 1.50705725e-01 -2.10487664e-01 -2.98972130e-01 -8.00800562e-01 6.19697392e-01 -1.77804187e-01 1.17858812e-01 -3.35062295e-02 -1.08894527e+00 9.98687625e-01 -1.57148585e-01 6.66459322e-01 -7.28433114e-03 3.53856415e-01 4.73338872e-01 4.73032773e-01 4.85253841e-01 8.96056533e-01 -4.85281825e-01 -9.96777952e-01 -8.53480041e-01 3.80145252e-01 1.18011093e+00 1.26281476e+00 8.00659716e-01 1.54123276e-01 4.84582067e-01 8.67994368e-01 6.23500764e-01 -1.84351623e-01 6.74401045e-01 -1.07410681e+00 4.70223755e-01 1.20479965e+00 7.69804642e-02 -1.15617001e+00 -2.64156014e-01 -4.42323908e-02 8.35813656e-02 4.55806881e-01 -7.41401687e-02 -1.59904078e-01 -2.95199931e-01 1.06616390e+00 -1.35331884e-01 -3.28339279e-01 -2.40863875e-01 1.07791703e-02 6.56491578e-01 7.55867541e-01 -1.36962272e-02 6.05643466e-02 7.61268377e-01 -1.19233668e+00 -5.10855556e-01 -5.56584835e-01 1.43549728e+00 -9.27400112e-01 9.20025647e-01 5.72903812e-01 -1.21412599e+00 -4.09127295e-01 -7.91797280e-01 7.27317035e-02 -8.02515745e-02 1.62388682e-01 9.93300498e-01 9.50159907e-01 -1.19443548e+00 6.13341391e-01 -1.10283160e+00 -4.09893394e-01 1.10801235e-02 1.78663552e-01 3.53336595e-02 -6.23044334e-02 -4.00383532e-01 9.09385979e-01 1.67570010e-01 -3.15861344e-01 -3.61269027e-01 -7.56389380e-01 -1.07932007e+00 -4.67582010e-02 2.82992274e-02 -3.10108781e-01 1.61227703e+00 -8.33845496e-01 -1.07164228e+00 6.56528652e-01 -1.76007807e-01 -1.89803571e-01 7.12103248e-02 -1.38565361e-01 -2.11754143e-01 -6.34262085e-01 2.50528097e-01 3.15684617e-01 3.37677956e-01 -7.51946390e-01 -6.35317802e-01 -3.26883458e-02 1.91999525e-01 -3.44254553e-01 -4.22294587e-01 7.62008965e-01 -2.65340030e-01 -3.34915936e-01 -2.58419245e-01 -7.19207048e-01 -4.63843137e-01 -2.33606204e-01 -5.69090769e-02 -3.99610907e-01 7.80328512e-01 -1.05384445e+00 1.65870440e+00 -2.02839875e+00 1.03431769e-01 9.02205035e-02 3.36976558e-01 2.32331932e-01 -2.39547297e-01 9.77447212e-01 6.00742847e-02 5.14643312e-01 8.58534351e-02 2.04463694e-02 6.45103455e-02 4.96868566e-02 2.10140705e-01 -5.73819168e-02 1.23113781e-01 8.12445998e-01 -8.30039263e-01 -4.64338899e-01 2.64495134e-01 -3.38187665e-02 -8.27453494e-01 1.06102183e-01 -6.60582781e-01 -1.57480240e-02 -4.26618457e-01 6.90651178e-01 2.36482486e-01 -2.46835172e-01 2.08554283e-01 8.98530066e-01 -5.25033057e-01 4.10339862e-01 -8.91829610e-01 1.52583218e+00 -7.46761858e-01 9.32463408e-01 -1.68289438e-01 -6.36315703e-01 1.32782733e+00 2.95302778e-01 -1.44755393e-02 -4.73831773e-01 -2.71522582e-01 2.01817960e-01 2.29034096e-01 -1.06913292e+00 4.00286168e-01 3.19833845e-01 -8.34490508e-02 9.87980425e-01 -8.58922973e-02 -2.09122539e-01 5.06148577e-01 2.73912996e-01 1.23364699e+00 1.91747591e-01 6.63718760e-01 -3.31597894e-01 2.85061657e-01 2.67736167e-01 4.12165135e-01 7.33092666e-01 2.69378573e-01 6.78510517e-02 9.15519416e-01 -5.02933681e-01 -9.81997252e-01 -3.57937485e-01 2.54499346e-01 1.12948000e+00 -3.43730688e-01 -1.06887996e+00 -9.94656563e-01 -6.22532427e-01 -2.25156397e-01 9.95725513e-01 -1.35461092e-01 -1.26889208e-02 -4.49252725e-01 -5.73662579e-01 5.47599018e-01 5.25796533e-01 9.07557458e-02 -1.16801882e+00 -7.46799648e-01 4.36794072e-01 -8.10258985e-02 -5.92033684e-01 -1.72767788e-01 -1.62845794e-02 -1.11382258e+00 -9.76807535e-01 -1.45057291e-01 -1.04253423e+00 6.17375970e-01 -1.68394819e-02 1.32566226e+00 9.33364272e-01 -6.38947427e-01 4.05879170e-01 -6.91128016e-01 -5.69485962e-01 -1.23614466e+00 1.40830338e-01 -5.64830542e-01 -8.53303611e-01 5.29932737e-01 -6.10560477e-01 3.23905379e-01 1.49586992e-02 -4.85521555e-01 1.62929788e-01 3.38385373e-01 7.66954303e-01 -3.82183731e-01 1.64651826e-01 5.18441796e-01 -1.06780481e+00 8.50792050e-01 -4.55667377e-01 -8.68337810e-01 3.78371269e-01 -9.07113254e-01 -1.21432096e-02 4.37390327e-01 -1.53515652e-01 -1.16687131e+00 -3.36026192e-01 -3.57832283e-01 3.81332308e-01 -2.58763582e-01 1.13392746e+00 1.62063673e-01 -6.21902645e-01 9.08553660e-01 -7.78305382e-02 3.25513668e-02 -7.07348347e-01 -8.02668855e-02 7.88681746e-01 -8.78237262e-02 -5.12833953e-01 6.21150613e-01 -4.65333164e-01 -6.31325006e-01 -8.61793101e-01 9.63372961e-02 -3.50016713e-01 -2.72611350e-01 -1.31906085e-02 2.68533379e-01 -5.54910302e-01 -2.17629060e-01 2.62794495e-01 -1.34216905e+00 -2.78914094e-01 9.32260901e-02 1.88875854e-01 -3.10288846e-01 7.06004202e-01 -7.84537435e-01 -9.04273450e-01 -2.04489231e-01 -1.30539429e+00 7.23703206e-01 3.36041957e-01 -8.25657487e-01 -1.40764666e+00 1.56112805e-01 4.00311202e-01 4.25873458e-01 1.52909011e-01 1.76164424e+00 -4.19632763e-01 -6.39235616e-01 -5.88903487e-01 5.45718260e-02 2.14948297e-01 4.31140028e-02 6.03342414e-01 -4.03824449e-01 -7.30287656e-02 -5.22620790e-02 -2.48294875e-01 -2.02370882e-01 1.29834071e-01 7.90003419e-01 -4.84022468e-01 -4.59566802e-01 1.22599818e-01 1.40690827e+00 7.91734815e-01 6.70256138e-01 5.95497668e-01 3.95081162e-01 5.39234400e-01 6.79837823e-01 5.56290865e-01 4.50479418e-01 4.46529597e-01 3.42526257e-01 2.03863382e-01 1.07576154e-01 -1.05832033e-02 4.27573442e-01 1.21123874e+00 -1.50616076e-02 1.83263436e-01 -1.72221947e+00 7.66018689e-01 -1.91428673e+00 -6.17644429e-01 -6.27994001e-01 1.89299881e+00 7.15956807e-01 1.80347741e-01 -3.95012461e-02 9.46719274e-02 4.31727916e-01 -4.07064557e-01 -3.88917662e-02 -9.95913982e-01 7.08621383e-01 1.91704839e-01 -3.44890177e-01 3.45204979e-01 -3.90331775e-01 6.56573296e-01 7.29243183e+00 5.69624245e-01 -9.07527387e-01 1.87394947e-01 1.22516222e-01 4.71500248e-01 -3.48888278e-01 4.94389862e-01 -6.54336154e-01 3.22007120e-01 1.14632308e+00 -6.52505994e-01 4.92485821e-01 1.55034113e+00 1.88110709e-01 -4.09550130e-01 -1.20783126e+00 6.24105453e-01 3.10983714e-02 -1.48452437e+00 -3.25003445e-01 -8.18262547e-02 7.96764314e-01 3.29486988e-02 -5.41130781e-01 4.26278979e-01 3.09969187e-01 -8.20209086e-01 6.88361824e-01 1.61818415e-01 1.74137838e-02 -7.66716003e-01 9.33410466e-01 8.19830060e-01 -8.77458096e-01 -3.25009495e-01 -3.67694438e-01 -8.02167058e-01 -1.72862783e-01 4.77471322e-01 -1.14668858e+00 4.68648016e-01 6.72185063e-01 7.36827612e-01 -1.09622359e+00 1.54583085e+00 -2.02333346e-01 6.79545522e-01 1.93421915e-01 -5.71577787e-01 -1.59778837e-02 -3.84217240e-02 2.79668063e-01 1.43656111e+00 3.40490729e-01 -5.53284407e-01 4.09406543e-01 1.12901068e+00 5.73707104e-01 8.83770548e-03 -5.32428682e-01 -6.41664922e-01 3.84058833e-01 1.02834082e+00 -6.61386967e-01 -2.64936894e-01 -7.69304514e-01 3.45186919e-01 2.51018673e-01 1.39310792e-01 -5.80319941e-01 -8.79716992e-01 4.12055105e-01 2.14840516e-01 -1.56494603e-01 -2.33296305e-01 -5.24374008e-01 -1.01031983e+00 2.29624629e-01 -1.22041690e+00 2.58878134e-02 -7.67450690e-01 -8.30158830e-01 5.89523315e-01 -6.21875108e-04 -9.40913916e-01 -7.81371057e-01 -4.72324669e-01 -9.49381888e-01 6.98108435e-01 -9.19992983e-01 -8.60500276e-01 -8.44208673e-02 -2.30364993e-01 7.01377571e-01 -4.00060385e-01 9.71594632e-01 1.96316272e-01 -4.06830400e-01 4.77430671e-01 1.18017949e-01 -4.79922332e-02 1.48095191e-01 -1.19377625e+00 8.76383066e-01 8.04528058e-01 1.43755928e-01 1.15813255e+00 7.47426569e-01 -7.02383399e-01 -1.31906152e+00 -9.45465088e-01 1.15721238e+00 -6.11270607e-01 6.86576307e-01 -3.86780530e-01 -9.35187101e-01 9.74702537e-01 2.88988382e-01 -6.22797549e-01 7.17234373e-01 4.41207618e-01 -3.27978998e-01 3.54759425e-01 -8.98669660e-01 4.28523749e-01 3.51727992e-01 -5.49542844e-01 -6.20002449e-01 5.41063726e-01 4.47013944e-01 -2.67544508e-01 -9.50374663e-01 -1.70938522e-01 3.45831901e-01 -1.27786112e+00 4.47463572e-01 -2.96668649e-01 7.65923083e-01 -2.48548724e-02 3.39074701e-01 -1.12578249e+00 -2.59964257e-01 -8.02620828e-01 1.79816306e-01 1.46659493e+00 6.84618533e-01 -6.60623193e-01 6.75111771e-01 9.17202771e-01 -3.27984720e-01 -3.17498595e-01 -6.05409861e-01 -7.90442526e-01 8.13699365e-02 -6.48150086e-01 5.40660083e-01 1.19492972e+00 9.93147492e-01 6.47092834e-02 -2.53591448e-01 -1.88794300e-01 2.68814743e-01 2.34188408e-01 8.28062773e-01 -1.40572298e+00 -6.98789299e-01 -3.52565557e-01 -6.00993037e-01 -5.08029163e-01 6.57610074e-02 -8.14590991e-01 8.12837183e-02 -1.61753392e+00 3.37169290e-01 -4.64970797e-01 5.16179025e-01 7.74893105e-01 -1.06398845e-02 -3.93978059e-01 4.38515842e-02 1.83091648e-02 -4.82134819e-01 -3.79814915e-02 4.60629880e-01 2.89112143e-03 -3.93876582e-01 2.68859774e-01 -7.07551062e-01 1.06494093e+00 7.50045180e-01 -7.84901857e-01 -2.77245641e-01 -6.25791907e-01 5.88031650e-01 3.62605363e-01 1.01063043e-01 -9.52868879e-01 1.60589039e-01 -1.96144328e-01 -1.56727448e-01 3.41007300e-02 -2.65169650e-01 -4.06594664e-01 3.51992607e-01 7.34675586e-01 -2.13000059e-01 3.79860580e-01 5.24726033e-01 2.85272729e-02 -2.83709824e-01 -1.23096597e+00 4.48398679e-01 -6.09436929e-01 -8.27577293e-01 -4.73782808e-01 -1.03391969e+00 -3.19927901e-01 1.09300005e+00 -2.76911855e-01 -3.54286224e-01 -5.80836311e-02 -2.26557463e-01 1.07673824e-01 7.61782706e-01 7.50969946e-01 5.53267837e-01 -8.20959330e-01 -3.05816263e-01 2.88448721e-01 3.51809919e-01 -4.85684574e-01 -9.07231420e-02 6.75860107e-01 -9.16702688e-01 5.01348495e-01 -2.74739444e-01 -5.63814461e-01 -1.70111954e+00 4.12655085e-01 -3.43012102e-02 -2.53538609e-01 -7.43783653e-01 7.46387720e-01 -1.13077275e-01 -7.13971317e-01 2.54802912e-01 -3.86155844e-01 -7.90390894e-02 -5.66616714e-01 8.51545870e-01 5.01719177e-01 1.99941635e-01 5.91727048e-02 -2.24617213e-01 4.27018374e-01 -3.26888114e-01 1.98792145e-01 1.43154228e+00 4.28685807e-02 -8.22517633e-01 7.17518985e-01 6.39722466e-01 -8.76638144e-02 -4.37946767e-01 1.82002291e-01 8.80337834e-01 -5.98034382e-01 -1.75180793e-01 -8.64265978e-01 -6.27525628e-01 1.02330935e+00 3.50253493e-01 4.31393236e-01 7.39518106e-01 1.02828093e-01 2.52063036e-01 4.75374043e-01 8.19236159e-01 -7.78770447e-01 1.62315831e-01 4.94392097e-01 7.10826635e-01 -9.51127291e-01 -1.54223517e-01 -3.95449668e-01 -3.34915578e-01 1.52116394e+00 9.51243460e-01 2.16313958e-01 1.65949151e-01 8.38671327e-01 5.13690189e-02 -1.73174992e-01 -1.12374842e+00 6.11208558e-01 -2.39120945e-01 8.39553833e-01 1.05890393e+00 -1.99303165e-01 -3.92417967e-01 4.66603488e-01 -1.86486989e-02 4.35692042e-01 1.30811942e+00 1.56849909e+00 -6.57344341e-01 -1.64615381e+00 -7.19526649e-01 5.41675806e-01 -2.98867583e-01 -3.09945822e-01 -4.59717870e-01 7.70963311e-01 1.13075882e-01 1.02980351e+00 -5.01674175e-01 -5.14213264e-01 -1.62053630e-02 3.46018165e-01 4.43731695e-01 -1.22528219e+00 -1.00253618e+00 1.14111071e-02 5.44432044e-01 -3.05337429e-01 4.06999104e-02 -5.89952588e-01 -1.07666075e+00 -4.25290942e-01 -6.83147609e-01 5.91092050e-01 7.56860554e-01 8.70690703e-01 4.69126701e-01 4.21520293e-01 4.90231179e-02 -5.17646730e-01 -3.13463181e-01 -9.55287695e-01 -3.99490893e-01 -3.99965078e-01 -8.02787915e-02 -2.90619135e-01 9.19475555e-02 2.64371276e-01]
[7.825216770172119, 7.645615100860596]
52095f56-60d3-4b09-aa3b-754c3c0e11e5
boundary-guided-context-aggregation-for
2110.14587
null
https://arxiv.org/abs/2110.14587v1
https://arxiv.org/pdf/2110.14587v1.pdf
Boundary Guided Context Aggregation for Semantic Segmentation
The recent studies on semantic segmentation are starting to notice the significance of the boundary information, where most approaches see boundaries as the supplement of semantic details. However, simply combing boundaries and the mainstream features cannot ensure a holistic improvement of semantics modeling. In contrast to the previous studies, we exploit boundary as a significant guidance for context aggregation to promote the overall semantic understanding of an image. To this end, we propose a Boundary guided Context Aggregation Network (BCANet), where a Multi-Scale Boundary extractor (MSB) borrowing the backbone features at multiple scales is specifically designed for accurate boundary detection. Based on which, a Boundary guided Context Aggregation module (BCA) improved from Non-local network is further proposed to capture long-range dependencies between the pixels in the boundary regions and the ones inside the objects. By aggregating the context information along the boundaries, the inner pixels of the same category achieve mutual gains and therefore the intra-class consistency is enhanced. We conduct extensive experiments on the Cityscapes and ADE20K databases, and comparable results are achieved with the state-of-the-art methods, clearly demonstrating the effectiveness of the proposed one.
['Di Huang', 'Hongyu Yang', 'Haoxiang Ma']
2021-10-27
null
null
null
null
['boundary-detection']
['computer-vision']
[ 2.46667102e-01 1.04171753e-01 -1.98977321e-01 -5.74886203e-01 -2.79178619e-01 -1.94409639e-01 4.71669018e-01 4.15499091e-01 -3.05034310e-01 4.59600836e-01 2.95833915e-01 8.98911729e-02 -2.27882415e-01 -9.31860030e-01 -4.52177376e-01 -9.18443441e-01 2.10457623e-01 5.55548295e-02 7.68757343e-01 -2.04637945e-01 3.28479260e-01 6.09866455e-02 -1.55539048e+00 4.18324560e-01 1.24875760e+00 1.29242241e+00 3.93229514e-01 -1.26466021e-01 -6.34438217e-01 3.71871203e-01 -2.42481306e-01 3.79574299e-02 1.28415436e-01 -3.65413785e-01 -7.55119085e-01 2.50792593e-01 3.93032104e-01 -2.35280227e-02 3.08476016e-02 1.24089801e+00 2.36173078e-01 1.56608269e-01 3.58412027e-01 -9.70383227e-01 -4.83525932e-01 5.72486997e-01 -7.76973069e-01 9.93769988e-02 4.19230126e-02 -1.60304278e-01 1.39116096e+00 -9.02224660e-01 5.49725652e-01 1.21327913e+00 4.65774685e-01 1.01800732e-01 -8.01932573e-01 -6.01094246e-01 9.46119785e-01 3.62701535e-01 -1.52413833e+00 1.40135139e-01 1.08116388e+00 -2.45759159e-01 4.44994211e-01 2.06889197e-01 7.37436056e-01 7.11163580e-01 -3.15949768e-01 1.01678109e+00 1.06037378e+00 -3.34100127e-01 2.34049976e-01 3.67416404e-02 4.56119657e-01 4.33865875e-01 3.56271744e-01 -2.68969834e-01 -3.95891905e-01 1.49990380e-01 6.53419912e-01 8.48535672e-02 -4.11173016e-01 -4.40176010e-01 -1.11237144e+00 6.14944577e-01 8.73854518e-01 5.67405522e-01 -2.61394769e-01 -1.39736041e-01 3.70364398e-01 -1.46738857e-01 7.03484476e-01 -6.24873340e-02 -3.82175416e-01 3.01502436e-01 -9.97557223e-01 1.92976519e-01 3.35531712e-01 9.39163923e-01 9.85127032e-01 -4.34165031e-01 -4.33835894e-01 9.68871295e-01 5.46852291e-01 1.53174609e-01 1.97851077e-01 -6.98396981e-01 5.63931525e-01 1.21581173e+00 4.18359553e-03 -1.25451458e+00 -3.98693413e-01 -6.81948960e-01 -7.27961898e-01 -9.85287726e-02 2.35250294e-01 1.22822337e-01 -1.04350877e+00 1.62314367e+00 7.86918104e-01 5.84192455e-01 -5.88871278e-02 1.04460311e+00 1.08487916e+00 4.20839190e-01 3.82282943e-01 4.56443913e-02 1.55008197e+00 -1.17706716e+00 -7.45075703e-01 -4.53821093e-01 5.99691629e-01 -6.39016807e-01 1.01269090e+00 1.57473966e-01 -4.84051645e-01 -6.31639540e-01 -1.08228636e+00 -5.91248572e-02 -6.30989194e-01 -8.56776610e-02 5.90900302e-01 3.58759135e-01 -8.35763216e-01 4.17119563e-01 -4.40772504e-01 -4.10205126e-01 6.13210618e-01 2.17960775e-02 -4.58923988e-02 -2.23282635e-01 -1.38551438e+00 3.73826295e-01 7.90109575e-01 4.50585067e-01 -3.17087770e-01 -5.41776896e-01 -8.57020080e-01 9.03765950e-03 7.92461991e-01 -5.40786326e-01 7.06115782e-01 -1.11696947e+00 -7.81795442e-01 7.32007384e-01 -1.93228900e-01 -1.75268322e-01 4.68844682e-01 -2.09637105e-01 -4.36348408e-01 3.58258516e-01 2.72671074e-01 9.49512661e-01 5.04883349e-01 -1.61252499e+00 -1.24490821e+00 -4.65774924e-01 1.20644942e-01 4.99649376e-01 -3.29969347e-01 -2.71601528e-01 -8.03618610e-01 -8.23108375e-01 6.25281513e-01 -6.88983202e-01 -9.52784941e-02 -6.32748455e-02 -3.17922205e-01 -4.40041214e-01 1.19928062e+00 -8.97964478e-01 1.62589276e+00 -2.27140236e+00 -1.72873512e-02 3.38245779e-01 1.84279561e-01 3.28420162e-01 4.37253080e-02 1.93999857e-02 8.11536238e-02 2.61964321e-01 -6.69072568e-01 -2.20002741e-01 -3.64756584e-02 3.53199124e-01 4.99582924e-02 4.38024253e-02 2.77913183e-01 8.85265410e-01 -9.65065956e-01 -7.36391962e-01 1.80725783e-01 4.01639462e-01 -4.37784612e-01 -1.56245381e-01 -2.43818864e-01 6.18934095e-01 -8.29759479e-01 6.45416498e-01 8.37836027e-01 -2.30349526e-01 1.60269365e-02 -1.87573969e-01 4.76764999e-02 1.17329508e-01 -1.40633416e+00 1.76288795e+00 -1.23449259e-01 2.21387520e-01 2.95071810e-01 -1.08843374e+00 1.16328859e+00 6.78141937e-02 3.73576671e-01 -9.50030029e-01 2.56601591e-02 2.90750831e-01 -2.01009229e-01 -3.42525721e-01 4.39723343e-01 1.96511194e-01 1.93236042e-02 -1.06114417e-01 -3.61141235e-01 1.68725893e-01 1.43083706e-01 1.90942436e-01 5.24598598e-01 3.59988362e-01 9.07897800e-02 -4.98819143e-01 8.00896347e-01 -7.42866248e-02 9.67646718e-01 5.24860144e-01 -4.60724860e-01 7.10175872e-01 4.09917414e-01 -2.81336695e-01 -7.20312238e-01 -6.83016241e-01 -1.59143656e-01 8.81337702e-01 8.92699778e-01 -3.03348333e-01 -1.04301560e+00 -9.25603807e-01 5.07387221e-02 4.28694308e-01 -7.35446453e-01 1.29678875e-01 -6.57191753e-01 -7.94906616e-01 1.03612818e-01 7.46429920e-01 1.36037791e+00 -9.60372329e-01 -3.45209032e-01 1.94520846e-01 -5.12220204e-01 -1.26444876e+00 -4.93143141e-01 -1.30145013e-01 -8.24453950e-01 -9.55136657e-01 -7.16922700e-01 -9.81138110e-01 5.08469939e-01 4.39844102e-01 8.33755791e-01 3.70698303e-01 -5.58350645e-02 -3.25546376e-02 -6.48058414e-01 -3.90008166e-02 1.42660499e-01 1.17872752e-01 -5.01994133e-01 4.14722234e-01 3.96863341e-01 -5.16617596e-01 -9.64998305e-01 4.45306450e-01 -8.55591059e-01 3.06708574e-01 5.33480942e-01 7.32982337e-01 9.18836296e-01 2.65994281e-01 6.67267382e-01 -9.88344669e-01 1.30365521e-01 -4.76906300e-01 -3.36499542e-01 3.96241844e-01 -7.50381947e-01 -2.20771194e-01 2.60453850e-01 -7.06178695e-02 -1.38820541e+00 -2.53118306e-01 5.26187301e-04 1.77981090e-02 -4.13732499e-01 3.72990787e-01 -6.39282823e-01 1.67484507e-01 2.01674402e-02 1.74099803e-01 -4.53376263e-01 -5.53130329e-01 4.43784207e-01 7.03712225e-01 2.99733251e-01 -5.55595815e-01 2.12492704e-01 8.06185424e-01 -1.34636208e-01 -6.87517226e-01 -1.03430879e+00 -8.50311041e-01 -6.35299504e-01 -1.89071253e-01 1.06482947e+00 -9.48623538e-01 -2.30601162e-01 6.44191146e-01 -7.87278771e-01 -1.61685050e-01 -1.28553703e-01 2.38226905e-01 -1.75277859e-01 5.78214526e-01 -4.03402925e-01 -6.02732956e-01 -1.65781066e-01 -1.01944244e+00 1.25101709e+00 6.89265907e-01 1.46125823e-01 -1.00717795e+00 -3.66231680e-01 5.14954031e-01 2.05844119e-01 3.01919311e-01 9.53724563e-01 -6.68082118e-01 -7.06340492e-01 1.91596791e-01 -8.18448603e-01 3.75333667e-01 3.33386958e-01 -2.31695041e-01 -1.09531391e+00 1.30978122e-01 -4.48620357e-02 2.68201023e-01 1.17485130e+00 3.98944855e-01 1.13126206e+00 2.01486982e-02 -5.65644145e-01 4.13762033e-01 1.56836581e+00 2.31007472e-01 5.60132325e-01 4.81966734e-01 9.95967686e-01 8.11502874e-01 9.39078391e-01 2.80467600e-01 6.50097787e-01 6.08319044e-01 5.46654761e-01 -3.79925996e-01 -3.40938121e-01 -1.29805669e-01 -2.12031845e-02 6.58103228e-01 1.74739193e-02 -1.13433860e-01 -8.90278935e-01 8.40952754e-01 -2.04467034e+00 -5.82448781e-01 -4.76319194e-01 1.93483615e+00 6.46420181e-01 4.00227189e-01 -2.09673978e-02 1.01571806e-01 1.18900776e+00 4.14505720e-01 -5.24845719e-01 -4.51790765e-02 -3.30320716e-01 -1.49976209e-01 2.52763629e-01 4.53971952e-01 -1.30961406e+00 1.06604397e+00 4.98764658e+00 1.25723124e+00 -7.13135183e-01 8.20804536e-02 8.67600501e-01 3.21139008e-01 -4.41826612e-01 3.04775566e-01 -9.13521886e-01 4.88268882e-01 4.54420075e-02 3.20508420e-01 8.85946974e-02 7.86729455e-01 3.33259314e-01 -6.11265779e-01 -6.98005021e-01 7.14298010e-01 -9.17949453e-02 -9.54362571e-01 1.56564683e-01 -3.30230128e-03 9.08866286e-01 -1.52790084e-01 -2.65996546e-01 8.32356736e-02 -3.11327241e-02 -6.27162457e-01 8.05094421e-01 5.08383811e-01 2.53169715e-01 -7.11646438e-01 9.78763521e-01 2.39145964e-01 -1.80612183e+00 -2.72485733e-01 -2.05993161e-01 1.47912413e-01 9.21238884e-02 9.27444518e-01 -2.42944181e-01 1.09882677e+00 8.54699373e-01 9.45418119e-01 -7.82848299e-01 1.00172973e+00 -4.44842190e-01 7.01840699e-01 -2.94398010e-01 2.45039567e-01 6.98001564e-01 -4.80525255e-01 4.15847033e-01 1.31642699e+00 -6.38368651e-02 1.46968395e-01 5.01028657e-01 9.19327974e-01 1.97737917e-01 4.43944782e-01 -1.17960587e-01 3.84204745e-01 4.41695392e-01 1.39163160e+00 -1.32171345e+00 -5.90868831e-01 -5.75464308e-01 7.52794683e-01 1.73980057e-01 4.93720412e-01 -8.24819207e-01 -4.52281326e-01 5.44853628e-01 -4.69555110e-02 5.84444463e-01 -1.04088783e-01 -7.45016456e-01 -9.18871880e-01 3.26819986e-01 -3.71985078e-01 5.55356681e-01 -6.77647591e-01 -1.18469131e+00 2.61224031e-01 7.21608996e-02 -1.04906654e+00 6.27489567e-01 -2.00601682e-01 -6.42597377e-01 8.75326753e-01 -2.03843141e+00 -1.30374968e+00 -7.11145103e-01 3.21264297e-01 5.48083663e-01 5.25570929e-01 2.38457561e-01 3.75937819e-01 -6.51748180e-01 2.97942758e-01 -1.45554587e-01 2.30586231e-01 4.44521576e-01 -1.15938175e+00 1.94391593e-01 1.01748776e+00 2.80200671e-02 4.45347190e-01 2.29910702e-01 -1.02241182e+00 -3.87932599e-01 -1.21407819e+00 7.76199639e-01 -4.48064990e-02 3.99917096e-01 -2.92349547e-01 -1.16551888e+00 2.63933420e-01 9.06066820e-02 7.54711479e-02 3.23718429e-01 2.45505609e-02 -2.68184662e-01 -2.24315062e-01 -8.25138688e-01 5.88017583e-01 1.42681909e+00 -3.21141154e-01 -6.74304664e-01 -1.78203285e-01 1.06399131e+00 -3.78598012e-02 -7.37565398e-01 7.51788914e-01 2.73476154e-01 -1.09433532e+00 8.31663609e-01 -1.33190406e-02 3.57081950e-01 -6.03681624e-01 -2.24385500e-01 -8.79761338e-01 -1.81824461e-01 -2.49572545e-02 2.34439164e-01 1.73832607e+00 1.59866035e-01 -6.83717072e-01 5.91007650e-01 2.82951236e-01 -3.39773655e-01 -8.49963248e-01 -8.54372978e-01 -7.00435936e-01 -1.67266373e-02 -4.86871362e-01 8.65285575e-01 9.29178596e-01 -2.06829101e-01 1.33681685e-01 1.00722134e-01 2.28639960e-01 4.61887360e-01 3.85799319e-01 3.26800019e-01 -1.32230783e+00 1.71828702e-01 -6.24476314e-01 -2.78622687e-01 -1.30025125e+00 -3.44473645e-02 -8.50476086e-01 9.13572386e-02 -1.91607952e+00 2.80766428e-01 -8.79771352e-01 -6.23417616e-01 4.27055776e-01 -7.48368084e-01 1.62182823e-01 1.24379963e-01 1.66281000e-01 -7.86499441e-01 7.34661043e-01 1.61981380e+00 -1.18122920e-01 -1.80620834e-01 -2.65341014e-01 -7.85113871e-01 9.35340285e-01 6.31323397e-01 -3.43306988e-01 -5.77582300e-01 -2.98662424e-01 1.19988367e-01 -4.73700732e-01 4.83316779e-01 -9.14044738e-01 1.65354073e-01 -1.53444588e-01 1.26998857e-01 -8.09173644e-01 -1.07426435e-01 -9.38913584e-01 -1.56622142e-01 7.11606592e-02 -1.71284467e-01 -5.64956248e-01 1.05299026e-01 8.14916193e-01 -4.76893514e-01 -1.13552645e-01 5.61111152e-01 -2.09504649e-01 -1.35162175e+00 3.13332230e-01 9.83937532e-02 3.58586460e-01 1.01112485e+00 -4.34039801e-01 -1.99936822e-01 7.03886822e-02 -8.62166464e-01 5.53330123e-01 4.51287508e-01 5.07061124e-01 3.61768812e-01 -1.29869938e+00 -4.33282703e-01 1.82938918e-01 3.54949445e-01 6.22552574e-01 6.33073211e-01 9.25559342e-01 -2.08351433e-01 2.74377882e-01 2.49083146e-01 -8.37180197e-01 -1.11875296e+00 4.04476613e-01 2.36155406e-01 -3.66375625e-01 -8.49644780e-01 7.99965560e-01 7.61121154e-01 -2.33895183e-01 1.47931367e-01 -6.51758254e-01 -6.64685488e-01 4.72541124e-01 3.43366295e-01 1.93946391e-01 3.63065600e-02 -7.20498145e-01 -5.07586539e-01 9.57078397e-01 6.91619292e-02 9.69922245e-02 1.09427893e+00 -6.70345902e-01 -1.66283458e-01 4.24445719e-01 7.70128608e-01 -1.77207917e-01 -1.47837222e+00 -4.66837734e-01 2.81579226e-01 -3.80560935e-01 1.21245965e-01 -7.02438414e-01 -1.30487633e+00 8.79508317e-01 5.50217450e-01 1.87883049e-01 1.42801619e+00 7.78187141e-02 9.00344849e-01 -1.27691805e-01 3.56340349e-01 -1.46271467e+00 -3.72765400e-02 2.91714817e-01 5.31195939e-01 -1.27077734e+00 -7.34484866e-02 -1.03994870e+00 -7.77609110e-01 7.54928052e-01 6.45217180e-01 -9.20076966e-02 7.85584390e-01 -1.06086828e-01 -7.16526853e-03 -2.43269771e-01 -1.91805080e-01 -7.36269236e-01 3.41404796e-01 4.30844456e-01 7.95729831e-03 1.90163597e-01 -5.37166536e-01 7.25818574e-01 3.73378485e-01 -9.18484405e-02 6.56544641e-02 8.69108319e-01 -7.24316955e-01 -1.02682519e+00 -3.52860004e-01 1.16046466e-01 -2.16680676e-01 -2.40086094e-01 -2.98557788e-01 8.37543964e-01 6.32131517e-01 1.07816458e+00 4.58237752e-02 -2.04485327e-01 1.53024971e-01 2.10659251e-01 2.02637035e-02 -6.85882807e-01 -4.68849927e-01 3.16687018e-01 -2.67035924e-02 -6.25366926e-01 -8.02135706e-01 -6.42123342e-01 -1.62669170e+00 2.35722944e-01 -6.51008189e-01 3.78436758e-03 4.85772371e-01 1.22645330e+00 4.01816130e-01 7.74365306e-01 5.42952120e-01 -3.92581224e-01 1.35044932e-01 -8.96382630e-01 -8.19182694e-01 6.00649834e-01 1.42976567e-01 -8.89812291e-01 -4.63903457e-01 -3.76184657e-02]
[9.539497375488281, -0.6345651149749756]
d16b47d8-a903-43c7-b8a5-c9189c1eeeb3
learning-to-recover-reasoning-chains-for
2004.02393
null
https://arxiv.org/abs/2004.02393v1
https://arxiv.org/pdf/2004.02393v1.pdf
Learning to Recover Reasoning Chains for Multi-Hop Question Answering via Cooperative Games
We propose the new problem of learning to recover reasoning chains from weakly supervised signals, i.e., the question-answer pairs. We propose a cooperative game approach to deal with this problem, in which how the evidence passages are selected and how the selected passages are connected are handled by two models that cooperate to select the most confident chains from a large set of candidates (from distant supervision). For evaluation, we created benchmarks based on two multi-hop QA datasets, HotpotQA and MedHop; and hand-labeled reasoning chains for the latter. The experimental results demonstrate the effectiveness of our proposed approach.
['Jun-Jie Huang', 'Xiaodan Zhu', 'Xiaoxiao Guo', 'Yufei Feng', 'Shiyu Chang', 'Wenhan Xiong', 'Murray Campbell', 'Mo Yu', 'Michael Greenspan']
2020-04-06
null
null
null
null
['multi-hop-question-answering']
['knowledge-base']
[-1.72166541e-01 6.33828282e-01 -4.12089944e-01 -5.06813467e-01 -1.55600035e+00 -7.27865815e-01 2.79263586e-01 2.68973261e-01 -3.55614722e-01 1.27783990e+00 1.92788601e-01 -2.65556693e-01 -4.22645062e-01 -6.83944285e-01 -9.46305692e-01 -4.23879534e-01 -1.07986115e-01 1.14907908e+00 1.04451334e+00 -3.22527796e-01 1.87345684e-01 -1.01740450e-01 -1.19588268e+00 7.68314123e-01 1.30946779e+00 9.32793021e-01 -2.31904283e-01 6.12252116e-01 -6.99835792e-02 1.77980304e+00 -5.06387770e-01 -1.10076427e+00 -4.30022515e-02 -7.62308776e-01 -1.62082374e+00 -2.41566941e-01 2.37956926e-01 -2.03241855e-01 -3.58605608e-02 1.11726332e+00 3.16770732e-01 4.59967665e-02 5.29138684e-01 -1.19547009e+00 -3.04395854e-01 1.28312612e+00 -4.20503974e-01 4.41613734e-01 8.39030385e-01 -3.19366157e-02 1.80949724e+00 -6.66391313e-01 1.14434886e+00 1.29878819e+00 4.37932462e-01 5.15998423e-01 -1.03555810e+00 -4.61121738e-01 1.16481945e-01 6.71940684e-01 -1.00818455e+00 -2.34706476e-01 7.16634333e-01 -1.54917046e-01 5.45081079e-01 -1.41870594e-04 3.52190644e-01 9.48563814e-01 -7.15190992e-02 1.17126524e+00 1.22882330e+00 -4.91371036e-01 5.74076593e-01 3.97803634e-01 7.99256802e-01 1.10145712e+00 -3.13136466e-02 -1.47136554e-01 -8.34087014e-01 -8.47749531e-01 2.43353263e-01 -6.23451710e-01 -3.68361861e-01 -4.00637656e-01 -1.06312561e+00 1.17666078e+00 3.68770897e-01 1.26113626e-03 -1.19827040e-01 -1.36940032e-01 1.71407700e-01 8.49371076e-01 5.15171979e-03 5.46875298e-01 -5.61904550e-01 9.41689126e-03 -5.94975233e-01 3.46669465e-01 1.27801824e+00 1.03925180e+00 5.43274403e-01 -9.59091425e-01 -5.34685433e-01 6.25975847e-01 5.95997572e-01 2.08296224e-01 -1.14883460e-01 -1.31245315e+00 1.05872512e+00 7.20477283e-01 4.58976954e-01 -6.20831907e-01 -1.47185981e-01 -1.53807979e-02 -1.62810609e-01 -2.95963705e-01 6.71551883e-01 -3.98507804e-01 -5.89332223e-01 1.68946111e+00 5.28843045e-01 2.60640234e-01 4.35461998e-01 1.06194282e+00 9.82249379e-01 5.54627180e-01 1.15270622e-01 -1.38776988e-01 1.36462641e+00 -1.51751733e+00 -5.19408107e-01 -1.49857640e-01 4.99601454e-01 -4.53952938e-01 8.35089564e-01 4.80699003e-01 -1.40613234e+00 -2.43864387e-01 -9.14204240e-01 4.65135798e-02 1.66307405e-01 -1.39017701e-01 4.39958811e-01 2.57510487e-02 -8.31803977e-01 5.08298337e-01 -3.65527421e-01 9.76058003e-03 2.99801856e-01 1.55380711e-01 -4.03852873e-02 -2.70692527e-01 -1.81950307e+00 8.43290508e-01 5.51351607e-01 9.65562067e-04 -1.14463830e+00 -1.22962527e-01 -6.72500730e-01 2.33227059e-01 7.45922863e-01 -6.22812331e-01 1.40948737e+00 -7.53176808e-01 -1.28079236e+00 8.21972072e-01 -4.15215753e-02 -6.84248865e-01 6.46268070e-01 -2.59466112e-01 -4.34322685e-01 6.95752323e-01 3.86623770e-01 4.55231011e-01 5.68373144e-01 -1.45133626e+00 -1.01512098e+00 -5.94099350e-02 4.17314589e-01 2.95565277e-01 2.58493632e-01 1.66465908e-01 -7.48455763e-01 -9.08198208e-02 8.79793800e-03 -7.55276442e-01 -3.51725727e-01 -2.25150093e-01 -6.97289705e-01 -6.59481645e-01 3.69857609e-01 -7.12893963e-01 7.59305656e-01 -1.69419050e+00 4.77977246e-01 5.05021751e-01 2.08670646e-01 -8.13291743e-02 -8.50430429e-02 4.51068938e-01 5.65921783e-01 -2.32618362e-01 -2.32919812e-01 -2.00050235e-01 -1.41241029e-02 4.48155403e-01 -4.82689977e-01 1.03135616e-01 1.62342131e-01 7.29613602e-01 -1.36818564e+00 -1.22554290e+00 -6.62968934e-01 -3.09822589e-01 -4.33318645e-01 5.66273928e-01 -8.62607539e-01 4.08517420e-01 -8.68701935e-01 7.91648090e-01 2.86005646e-01 -7.69163251e-01 4.37556475e-01 2.07360402e-01 7.22828805e-01 4.90285248e-01 -1.06262600e+00 1.63883686e+00 5.40102758e-02 1.50004879e-01 1.95794180e-01 -7.17856169e-01 9.25229728e-01 2.32335120e-01 -1.03681371e-01 -5.86462379e-01 3.68851833e-02 3.27080280e-01 7.99778327e-02 -7.93465912e-01 3.24919909e-01 -5.56536466e-02 -2.52795100e-01 5.55004835e-01 4.63139147e-01 -7.19639733e-02 4.27580297e-01 8.11169982e-01 1.25589728e+00 5.82924038e-02 5.83351888e-02 -3.87688982e-03 6.78247869e-01 5.43571949e-01 6.58651829e-01 1.04905224e+00 -4.93155688e-01 2.29822606e-01 9.70158637e-01 -7.32329562e-02 -3.58842015e-01 -1.15499485e+00 2.60141790e-01 1.18045902e+00 6.33675456e-01 -3.17227572e-01 -6.36462748e-01 -1.54576921e+00 -2.89114323e-02 7.34511256e-01 -6.02988720e-01 1.22182979e-03 -5.77040911e-01 -8.45463797e-02 6.24865711e-01 4.61138725e-01 5.17188907e-01 -1.32157898e+00 -2.78641075e-01 5.55687249e-02 -6.35931373e-01 -1.13045573e+00 -1.53159186e-01 2.90295392e-01 -6.43045068e-01 -1.57549357e+00 -1.39706790e-01 -1.18783295e+00 4.57202911e-01 -6.36826158e-02 1.63694477e+00 5.18982530e-01 4.91630673e-01 2.17951834e-01 -4.54167753e-01 3.61096971e-02 -6.01719439e-01 4.32934500e-02 -4.89422888e-01 -1.21547908e-01 3.92418057e-01 -1.39848709e-01 -4.75820720e-01 6.43596292e-01 -4.41630036e-01 -1.90649897e-01 6.00646555e-01 1.11814094e+00 7.16475070e-01 -3.18885744e-02 9.21929061e-01 -1.39868629e+00 9.65475798e-01 -9.55975294e-01 -4.92234230e-01 9.14273798e-01 -4.51989830e-01 2.59872109e-01 7.51469433e-01 -8.24772939e-02 -1.52900350e+00 -1.42233700e-01 5.60165988e-03 -4.67337631e-02 -8.83105397e-03 6.53786123e-01 -6.71901405e-02 1.92546904e-01 6.44945204e-01 -2.44031414e-01 -4.62977499e-01 -1.91509157e-01 6.45226955e-01 5.63186347e-01 4.68102187e-01 -1.06535280e+00 5.51644444e-01 3.57337147e-01 -5.00628352e-01 1.81244642e-01 -1.42009330e+00 -6.51048720e-01 -5.24765730e-01 -1.64683670e-01 7.22233057e-01 -8.52560341e-01 -6.20868742e-01 -2.87986606e-01 -1.25171387e+00 -3.73806864e-01 -4.48478073e-01 4.04330164e-01 -5.95526218e-01 3.20611030e-01 -1.19398272e+00 -7.58364975e-01 -3.40090334e-01 -9.86799538e-01 8.25640261e-01 4.11974460e-01 -2.62331098e-01 -8.73185575e-01 5.46619594e-01 1.02767754e+00 -2.75797814e-01 -1.59152560e-02 1.11745203e+00 -1.07662296e+00 -9.31029975e-01 -8.41499418e-02 -1.44286186e-01 2.25432888e-02 -4.33771133e-01 -1.32378176e-01 -9.02801275e-01 -1.36959879e-02 -1.41033083e-01 -1.24299979e+00 9.75565851e-01 -1.16918527e-01 5.88810205e-01 -1.72981083e-01 -6.36516571e-01 -1.49646446e-01 1.12177336e+00 1.25130042e-01 3.58669162e-01 2.55887628e-01 2.01583713e-01 8.46981525e-01 1.19724143e+00 -5.87814860e-02 4.87804204e-01 3.81352752e-01 3.76067966e-01 2.46772975e-01 1.12328336e-01 -7.08985090e-01 1.01107396e-01 9.77966189e-01 1.44496113e-01 -3.89246404e-01 -8.78200650e-01 7.41773009e-01 -2.32857156e+00 -8.13495874e-01 -2.16822445e-01 1.56701362e+00 1.23423994e+00 5.59120893e-01 1.83110461e-02 -9.04096738e-02 6.35828197e-01 -9.17141810e-02 -6.20088518e-01 -1.19071208e-01 -2.54870772e-01 1.57916725e-01 -9.14141461e-02 7.38149464e-01 -8.15899611e-01 8.15860450e-01 6.65838623e+00 7.90230095e-01 -8.12940001e-02 2.14077726e-01 8.00519168e-01 1.38565809e-01 -6.21865153e-01 4.46234614e-01 -7.51719654e-01 2.24322587e-01 8.76066864e-01 1.37260512e-01 1.49151459e-01 7.00796664e-01 -3.38612199e-01 -4.19120491e-01 -1.16112268e+00 2.23152205e-01 7.45116025e-02 -1.52069044e+00 -4.92390357e-02 -5.64949155e-01 1.01860154e+00 9.37806256e-03 -3.88480395e-01 5.41672766e-01 1.23991942e+00 -8.12650859e-01 7.79555976e-01 5.68673074e-01 2.09422782e-01 -6.68308735e-01 1.02246130e+00 8.18177938e-01 -8.44782710e-01 -3.69054675e-01 -3.88114244e-01 3.01808208e-01 3.58766496e-01 5.09137452e-01 -8.78084958e-01 8.70263875e-01 8.14949155e-01 4.57396388e-01 -5.51461995e-01 1.00115657e+00 -1.16632986e+00 1.27314556e+00 -9.50925751e-04 -4.43591088e-01 3.56282115e-01 -2.24657729e-01 3.22299212e-01 9.51743007e-01 -1.63613170e-01 3.25216800e-01 8.19784850e-02 9.98653829e-01 -5.58277845e-01 -1.24021158e-01 -2.19650432e-01 4.17732209e-01 6.49955451e-01 1.03775191e+00 -7.29580522e-01 -4.14269894e-01 -4.08790410e-01 5.84309995e-01 9.54417765e-01 3.58164907e-01 -9.73151445e-01 -2.36681104e-01 -1.48702919e-01 -3.79138023e-01 2.90472955e-01 4.72518772e-01 1.14954904e-01 -1.14924157e+00 8.89335424e-02 -1.07985771e+00 1.29620814e+00 -8.69763792e-01 -1.79309797e+00 8.10807645e-01 -1.48952097e-01 -9.66908813e-01 -1.50227726e-01 -1.82053715e-01 -8.07171881e-01 5.68566561e-01 -1.69357789e+00 -8.97726238e-01 -1.17422201e-01 8.19321156e-01 4.21262205e-01 -4.93638217e-02 5.57105958e-01 7.90415406e-02 -2.34896183e-01 3.05333525e-01 -7.73693770e-02 3.75663668e-01 7.31175065e-01 -1.48603845e+00 8.88242275e-02 5.93563199e-01 4.56503719e-01 3.93474162e-01 5.39023101e-01 -6.82936013e-01 -1.03668618e+00 -7.32942522e-01 8.81342411e-01 -7.39582002e-01 8.62899542e-01 -2.97393829e-01 -1.10148692e+00 8.35252941e-01 5.01620770e-01 -6.75549507e-02 6.99407756e-01 1.89854398e-01 -4.72817630e-01 1.15367331e-01 -1.35224378e+00 4.56906229e-01 9.68731403e-01 -4.48365241e-01 -1.24534476e+00 4.82851148e-01 1.03292012e+00 -4.36150998e-01 -7.37239361e-01 1.73389643e-01 1.69319883e-02 -1.01824546e+00 7.57314205e-01 -9.37437296e-01 6.78068638e-01 -4.44759578e-01 6.27894178e-02 -1.27658105e+00 -4.85834628e-02 -5.42854667e-01 -4.27231818e-01 1.14228308e+00 1.06726086e+00 -3.86051148e-01 1.02337945e+00 4.72305804e-01 1.66220322e-01 -8.60113919e-01 -1.05931830e+00 -4.02485549e-01 -1.15017705e-01 9.52909812e-02 4.26090986e-01 9.02700484e-01 4.57369030e-01 9.52278256e-01 -2.63361096e-01 3.51747006e-01 6.33711517e-01 7.55961537e-01 3.79801124e-01 -1.23671210e+00 -7.90961921e-01 4.83599342e-02 3.09944361e-01 -1.33519399e+00 3.97759080e-01 -7.77490318e-01 4.85401988e-01 -1.61331093e+00 4.90058064e-01 -6.32079720e-01 -3.22900474e-01 1.80056512e-01 -5.25769055e-01 -3.55446696e-01 -5.82681447e-02 3.86516154e-01 -1.46725905e+00 5.03544927e-01 1.31460369e+00 -3.52130592e-01 -2.05387287e-02 3.61526638e-01 -5.98228395e-01 7.48470664e-01 6.45065904e-01 -7.31374741e-01 -7.22381294e-01 -4.15767908e-01 5.34719348e-01 8.31555545e-01 1.46596849e-01 -5.93814969e-01 6.51441276e-01 -1.44023508e-01 -4.42753918e-02 -6.97405398e-01 3.17203581e-01 -6.32060409e-01 -4.68242586e-01 3.55112344e-01 -1.08911240e+00 -1.52760804e-01 -2.57805467e-01 1.16441345e+00 -5.49905777e-01 -5.78524411e-01 5.63474834e-01 -1.43850103e-01 -5.71568966e-01 -1.35089671e-02 -1.09723665e-01 6.29138887e-01 1.15205657e+00 4.58017290e-01 -8.33478749e-01 -7.04786718e-01 -9.09467161e-01 1.00364721e+00 4.19534780e-02 2.32280344e-02 7.54414678e-01 -1.14228117e+00 -8.91035318e-01 -5.07744968e-01 2.41605908e-01 1.91977680e-01 -9.21007022e-02 7.91617215e-01 -3.59137684e-01 2.05262154e-01 1.63737506e-01 -3.80015641e-01 -1.13988721e+00 6.04613543e-01 3.67723823e-01 -8.91771495e-01 -3.04971457e-01 1.24277401e+00 -5.27161002e-01 -5.54654300e-01 4.82097268e-01 -2.49827191e-01 -6.19706452e-01 -1.48947313e-01 3.29324603e-01 2.94041902e-01 -1.98078662e-01 -1.38017118e-01 -2.45971173e-01 1.09161422e-01 -1.42279223e-01 -2.30600357e-01 1.13669431e+00 -5.69058135e-02 -2.84299433e-01 3.34194481e-01 5.71913302e-01 2.90754676e-01 -1.17204452e+00 -7.03400016e-01 6.77805662e-01 -2.71349251e-01 -6.28369629e-01 -1.00095785e+00 -8.67268622e-01 5.91961265e-01 -6.19621202e-02 4.00596380e-01 6.81205690e-01 6.69519246e-01 7.12114930e-01 7.06642509e-01 6.36029899e-01 -1.15120161e+00 3.86678934e-01 3.96787286e-01 5.82177699e-01 -1.25813663e+00 -3.06449711e-01 -6.73911452e-01 -9.94804502e-01 8.47237706e-01 9.87108827e-01 -2.30026916e-02 4.01780307e-01 1.20860271e-01 7.46749789e-02 -7.32570469e-01 -1.32147467e+00 -1.61925718e-01 3.53883095e-02 4.60407257e-01 5.25020994e-02 -6.56233132e-02 -2.99336076e-01 8.15050364e-01 1.46754757e-01 -9.11405087e-02 3.46188366e-01 1.11000586e+00 -6.06028378e-01 -8.54119062e-01 -3.69207442e-01 4.29362714e-01 -1.23521715e-01 8.26740414e-02 -9.69610631e-01 5.48794329e-01 6.49804696e-02 1.43083227e+00 -4.17969763e-01 -3.19703192e-01 3.87145817e-01 4.53387536e-02 2.19880164e-01 -5.60189784e-01 -7.05910087e-01 -4.95904833e-01 7.76065350e-01 -4.13857400e-01 -6.59090102e-01 -2.85113901e-01 -1.42003334e+00 1.07847661e-01 -6.42354488e-01 1.21302640e+00 -2.77445674e-01 1.09454346e+00 9.83619094e-02 2.54344314e-01 5.68341672e-01 3.37251812e-01 -9.68673825e-01 -7.85667002e-01 -5.19277811e-01 5.75932145e-01 1.40613616e-01 -5.45263529e-01 -3.21777016e-01 -1.37475878e-01]
[10.975397109985352, 7.9300737380981445]
921d31f9-c69d-4125-82c2-20a24c447747
learning-long-term-dependencies-in
2006.04418
null
https://arxiv.org/abs/2006.04418v4
https://arxiv.org/pdf/2006.04418v4.pdf
Learning Long-Term Dependencies in Irregularly-Sampled Time Series
Recurrent neural networks (RNNs) with continuous-time hidden states are a natural fit for modeling irregularly-sampled time series. These models, however, face difficulties when the input data possess long-term dependencies. We prove that similar to standard RNNs, the underlying reason for this issue is the vanishing or exploding of the gradient during training. This phenomenon is expressed by the ordinary differential equation (ODE) representation of the hidden state, regardless of the ODE solver's choice. We provide a solution by designing a new algorithm based on the long short-term memory (LSTM) that separates its memory from its time-continuous state. This way, we encode a continuous-time dynamical flow within the RNN, allowing it to respond to inputs arriving at arbitrary time-lags while ensuring a constant error propagation through the memory path. We call these RNN models ODE-LSTMs. We experimentally show that ODE-LSTMs outperform advanced RNN-based counterparts on non-uniformly sampled data with long-term dependencies. All code and data is available at https://github.com/mlech26l/ode-lstms.
['Ramin Hasani', 'Mathias Lechner']
2020-06-08
null
http://proceedings.neurips.cc/paper/2020/hash/fa733611ef13bd333ebfbab7eed14b63-Abstract.html
http://proceedings.neurips.cc/paper/2020/file/fa733611ef13bd333ebfbab7eed14b63-Paper.pdf
neurips-2020-12
['sequential-image-classification']
['computer-vision']
[-1.95623890e-01 2.42088623e-02 -4.83318679e-02 7.51659786e-03 -3.10071141e-01 -4.06663835e-01 4.16760147e-01 -4.16606426e-01 -3.31446141e-01 6.61536098e-01 2.81076226e-02 -5.59427142e-01 -1.98291894e-02 -4.79801595e-01 -7.77900755e-01 -8.85007739e-01 -2.37961099e-01 2.37767547e-01 -1.44399345e-01 -3.13131690e-01 7.35387355e-02 5.87004662e-01 -1.06045842e+00 -1.94724072e-02 6.05157912e-01 9.30299520e-01 1.34374291e-01 9.97717321e-01 -2.63264440e-02 1.26400208e+00 -3.11890125e-01 1.30332321e-01 1.40235871e-01 -5.98870695e-01 -8.03226233e-01 -3.63746583e-01 -6.79754764e-02 -1.53194398e-01 -9.98255908e-01 9.06397700e-01 4.23603415e-01 3.99733692e-01 4.39902663e-01 -9.22718704e-01 -7.48020828e-01 3.89133781e-01 5.43473363e-02 4.16476220e-01 -4.87407446e-02 1.29017368e-01 7.90535927e-01 -9.80852902e-01 5.09908259e-01 9.27153409e-01 9.32536364e-01 8.63972366e-01 -1.17079282e+00 -3.47731233e-01 3.05163771e-01 1.19445257e-01 -1.23063362e+00 -5.51684976e-01 8.25900316e-01 -4.77205306e-01 1.30405676e+00 1.91576228e-01 7.17660487e-01 1.45792580e+00 5.03910959e-01 8.77643287e-01 7.58367181e-01 -2.80125141e-01 2.84561604e-01 -1.13660894e-01 3.44687849e-01 5.91739178e-01 -4.39734071e-01 3.08197349e-01 -4.77764934e-01 -1.10019542e-01 9.83579338e-01 3.37719202e-01 -4.66653019e-01 -1.05698025e-02 -9.74131346e-01 7.28619337e-01 5.85357308e-01 6.91942155e-01 -5.54075480e-01 3.47807825e-01 4.20934588e-01 7.71638453e-01 7.18011916e-01 3.95223677e-01 -6.76456332e-01 -3.34505558e-01 -9.01957929e-01 -6.67404458e-02 8.49575102e-01 6.30990803e-01 4.32739586e-01 4.94582862e-01 7.67893717e-02 6.52138829e-01 7.79750943e-02 3.72690499e-01 1.02493012e+00 -7.75346816e-01 2.35754535e-01 1.71883762e-01 7.92850032e-02 -9.13476110e-01 -5.57581961e-01 -8.90296221e-01 -1.34821391e+00 2.91201863e-02 3.48436564e-01 -3.64860088e-01 -9.74541664e-01 2.02690768e+00 -4.69671115e-02 5.23361504e-01 1.33400038e-01 8.21266949e-01 2.15923503e-01 1.15449679e+00 -3.80210042e-01 -5.69157600e-01 7.45532513e-01 -9.19377029e-01 -1.03320968e+00 -2.66335636e-01 8.17568481e-01 -2.71262109e-01 7.66058087e-01 2.05509156e-01 -1.24918091e+00 -4.57770258e-01 -7.72889495e-01 -1.61108702e-01 -4.92226362e-01 4.26775217e-02 9.71359313e-02 -1.27517089e-01 -1.26382136e+00 1.17601955e+00 -1.32124567e+00 -8.26882422e-02 -2.49804616e-01 3.66893470e-01 2.54228301e-02 6.29012883e-01 -1.56527913e+00 1.09571874e+00 -3.40157002e-03 9.48024988e-01 -7.59763062e-01 -6.77703202e-01 -6.80355310e-01 -2.15347901e-01 -9.50921327e-02 -5.18922389e-01 1.60316491e+00 -9.78156567e-01 -1.80485630e+00 4.65279132e-01 -6.29285634e-01 -7.56631017e-01 6.02492630e-01 -1.90195099e-01 -4.37836140e-01 -2.16865093e-01 -3.46369982e-01 9.59220901e-02 1.02026427e+00 -7.42361784e-01 2.34074537e-02 -2.09858492e-01 -4.23922122e-01 -4.26844433e-02 -2.32099265e-01 -3.06963593e-01 9.54995975e-02 -6.50366127e-01 3.54218841e-01 -1.16387475e+00 -4.69206035e-01 -1.92735910e-01 -4.96192217e-01 -1.57772023e-02 8.60883355e-01 -6.86194658e-01 1.58436596e+00 -2.04786372e+00 4.70139712e-01 -2.00195387e-02 3.48453373e-01 5.88464797e-01 -3.89984436e-02 6.23845339e-01 -2.20418349e-01 -1.97380763e-02 -2.74195433e-01 -6.09069943e-01 -1.02128066e-01 4.23858374e-01 -9.55141604e-01 5.86968064e-01 2.75963604e-01 1.19097221e+00 -9.41229403e-01 5.14814518e-02 4.75106351e-02 8.56104732e-01 -1.49929091e-01 2.62472391e-01 -2.05159619e-01 7.62023866e-01 -2.75440186e-01 7.05567822e-02 2.66935974e-01 -3.60626340e-01 -8.77972916e-02 1.37891695e-01 -5.99801719e-01 5.80231547e-01 -7.68338859e-01 1.50812745e+00 -8.15913856e-01 1.13506186e+00 1.35384630e-02 -1.06183696e+00 8.42163742e-01 6.59986734e-01 3.65611315e-01 -8.03869665e-01 1.71177402e-01 6.08799160e-01 -2.54311562e-02 -4.46912944e-01 1.91179231e-01 -2.73617208e-01 2.34337017e-01 4.68569338e-01 4.45888750e-02 4.32927996e-01 -4.76284921e-02 -3.81545499e-02 1.15195572e+00 -7.08115026e-02 -1.55049220e-01 -1.32448554e-01 2.99976647e-01 -4.38310176e-01 6.21314466e-01 6.80094242e-01 -3.18547711e-02 5.07589698e-01 5.11505902e-01 -5.50936401e-01 -1.10261059e+00 -7.72101760e-01 -1.87736690e-01 7.85879970e-01 -3.68198752e-01 -1.28998458e-01 -4.22616988e-01 2.88239811e-02 -2.84426391e-01 6.08124614e-01 -8.15178871e-01 -4.40889627e-01 -9.08275843e-01 -4.23623562e-01 6.61696851e-01 5.22412956e-01 2.97675073e-01 -1.20215666e+00 -8.29371095e-01 6.09825313e-01 -1.17379583e-01 -8.61082554e-01 -5.67764163e-01 6.19892418e-01 -1.18919671e+00 -5.22372127e-01 -1.13982046e+00 -7.47019529e-01 4.80127305e-01 -2.86556631e-01 9.28541660e-01 -1.30054519e-01 -5.63627370e-02 1.90776557e-01 -6.17399886e-02 5.05790710e-02 -4.24623042e-01 3.33069891e-01 2.09174722e-01 1.20563149e-01 -2.32517868e-02 -9.81728494e-01 -5.32826722e-01 1.71250284e-01 -8.03911865e-01 1.91502318e-01 3.45072895e-01 9.79008198e-01 5.32143950e-01 -2.94862002e-01 5.75744510e-01 -5.54468274e-01 6.28238201e-01 -6.45681679e-01 -6.17462456e-01 4.13515829e-02 -5.24275124e-01 4.84054208e-01 1.06161857e+00 -9.14108515e-01 -7.83569634e-01 -5.38884811e-02 -3.32167000e-01 -8.83291304e-01 2.98706800e-01 6.44107938e-01 4.69384700e-01 1.48998544e-01 3.49719912e-01 6.27770185e-01 1.87869277e-02 -6.70728624e-01 1.36781111e-01 2.89464861e-01 4.87905025e-01 -3.12705576e-01 4.17831421e-01 2.77317405e-01 -9.06031206e-02 -9.25678432e-01 -9.58676577e-01 -1.99316308e-01 -6.92128539e-01 -2.08601281e-01 6.43523991e-01 -5.80203831e-01 -6.97911024e-01 8.94995868e-01 -1.50962484e+00 -9.67991889e-01 -4.71221656e-01 5.11701107e-01 -5.97208500e-01 -2.90633500e-01 -1.33888531e+00 -1.16790414e+00 -3.16356987e-01 -6.48026288e-01 5.78361273e-01 3.74986269e-02 -2.27377623e-01 -1.62464154e+00 4.83489990e-01 -6.72921836e-01 7.80798554e-01 2.68548667e-01 6.99563444e-01 -4.87362772e-01 -1.95728227e-01 -2.55065203e-01 2.45093524e-01 4.57028449e-01 -1.04511775e-01 2.19491720e-01 -9.99439418e-01 -1.72828391e-01 6.39694571e-01 3.59863788e-02 1.10421419e+00 6.72180057e-01 8.45130861e-01 -7.24104404e-01 -2.30756968e-01 8.39683473e-01 1.28755796e+00 2.67783254e-01 4.34509873e-01 2.20145993e-02 8.30726504e-01 3.45131695e-01 -1.32218957e-01 2.77022034e-01 7.94042721e-02 2.92540401e-01 2.07249954e-01 -9.69264656e-02 2.07284525e-01 -4.92131382e-01 7.02564478e-01 1.58459651e+00 -2.82105189e-02 -1.03088409e-01 -9.90085900e-01 5.18437505e-01 -2.08476830e+00 -1.09927082e+00 -3.52985710e-01 2.15035439e+00 8.88279498e-01 2.44854271e-01 -5.97728975e-02 2.54341245e-01 6.55701578e-01 3.62492412e-01 -1.15145254e+00 -7.23862588e-01 -1.65008143e-01 1.66238785e-01 4.68110085e-01 7.73372471e-01 -7.95308113e-01 6.92728102e-01 6.45708847e+00 3.75941038e-01 -1.79114771e+00 2.35306844e-01 4.24547493e-01 -2.49003723e-01 -2.19561979e-02 -2.03100249e-01 -7.37196207e-01 5.72998047e-01 1.63194788e+00 -2.18575999e-01 5.02803266e-01 4.78251994e-01 5.33746600e-01 4.41251785e-01 -1.03186822e+00 7.66872942e-01 -4.24571395e-01 -1.28906512e+00 -3.10240149e-01 7.51790553e-02 6.53618217e-01 4.01097298e-01 4.06955808e-01 4.48594481e-01 5.50134629e-02 -1.03643417e+00 7.42544889e-01 9.20798421e-01 5.98649919e-01 -4.21426326e-01 2.88635194e-01 6.31715715e-01 -1.24672651e+00 -2.42602333e-01 -3.25345874e-01 -5.44584632e-01 3.56116027e-01 9.19150114e-01 -2.09759235e-01 7.90832937e-02 2.90040612e-01 1.15524185e+00 -1.42319441e-01 7.07162857e-01 -1.63015034e-02 8.15314949e-01 -4.16364908e-01 2.79125459e-02 4.85996008e-01 -3.31270128e-01 6.89664185e-01 1.15046799e+00 4.40066844e-01 -1.99701879e-02 -3.17792356e-01 1.15607095e+00 9.36037302e-02 -4.34878618e-01 -7.54214585e-01 -2.96904415e-01 3.04313391e-01 8.23359489e-01 -3.37486595e-01 -1.76108509e-01 -1.83099449e-01 1.17861557e+00 5.51842332e-01 8.69649470e-01 -8.57712626e-01 -4.30633873e-01 7.73074627e-01 -9.10333730e-03 2.00394824e-01 -6.17604792e-01 -6.21399395e-02 -1.32021475e+00 2.60410368e-01 -4.35389996e-01 1.26724228e-01 -7.22121418e-01 -1.11069787e+00 9.66018558e-01 -4.36272681e-01 -1.20288444e+00 -7.08444536e-01 -6.25477612e-01 -7.54344523e-01 9.47330117e-01 -1.50903249e+00 -7.08742261e-01 2.76327491e-01 5.82708180e-01 4.39199060e-01 2.77142346e-01 8.45668137e-01 3.57303709e-01 -1.01359558e+00 3.46368760e-01 4.92221862e-01 2.41767377e-01 1.52095348e-01 -1.15280676e+00 6.96001828e-01 7.69462764e-01 -3.36467735e-02 8.63254845e-01 8.51719439e-01 -3.70966226e-01 -1.50673282e+00 -1.17420471e+00 1.08934402e+00 -3.78252566e-01 1.09463787e+00 -4.53274310e-01 -1.32463646e+00 1.12877727e+00 -2.34960988e-02 3.00863624e-01 1.92766234e-01 -1.33962691e-01 -2.12452233e-01 1.92442402e-01 -5.53491712e-01 4.35818315e-01 1.01115322e+00 -9.98961926e-01 -4.37688231e-01 1.00168660e-01 7.55206943e-01 -5.52464962e-01 -6.16729617e-01 2.86609501e-01 5.04928470e-01 -9.67341065e-01 5.47861099e-01 -7.30300784e-01 3.21907371e-01 3.51985320e-02 1.32089227e-01 -1.57317913e+00 -2.42875397e-01 -1.16628444e+00 -8.58053029e-01 6.30314052e-01 4.95159268e-01 -1.15175903e+00 3.68136227e-01 5.95457196e-01 -1.94019407e-01 -1.13992810e+00 -9.73490894e-01 -1.12953115e+00 4.89606231e-01 -4.68483210e-01 3.58964913e-02 8.19650829e-01 2.24277675e-02 2.46787965e-01 -5.01255989e-01 1.65354669e-01 2.73161560e-01 -4.62359563e-02 5.76245636e-02 -1.06320095e+00 -2.67383873e-01 -6.58392668e-01 -1.20266072e-01 -1.51730180e+00 4.72301334e-01 -6.42744958e-01 2.14746326e-01 -1.31313443e+00 -4.05529827e-01 -3.13627124e-01 -5.06084919e-01 2.87384987e-01 2.17216551e-01 -1.59394652e-01 7.83323422e-02 3.89858693e-01 -2.52585977e-01 6.97173417e-01 1.12561297e+00 3.50364953e-01 -5.61616123e-01 3.19285482e-01 3.45451683e-02 6.71352744e-01 1.02440000e+00 -5.35448492e-01 -2.76661783e-01 -5.34991264e-01 4.67796326e-01 4.15357560e-01 5.68142414e-01 -9.81091440e-01 7.12909460e-01 2.24584639e-01 8.35079029e-02 -5.91301441e-01 6.01156831e-01 -4.60819274e-01 1.81623057e-01 8.08635831e-01 -7.53989041e-01 4.08756614e-01 1.35830894e-01 4.18447852e-01 -3.40652645e-01 -4.00842866e-03 6.94709599e-01 -1.14823431e-01 -2.19251856e-01 4.55005169e-01 -8.80796015e-01 1.65132508e-01 5.00834286e-01 4.27853577e-02 -6.73751608e-02 -6.70983553e-01 -1.02964604e+00 6.34839162e-02 4.05938961e-02 2.19941109e-01 4.60669279e-01 -1.36306739e+00 -4.26728576e-01 5.09707093e-01 -5.32678604e-01 -7.00430188e-04 2.96187013e-01 1.13155663e+00 -1.14760898e-01 6.79039598e-01 7.25143999e-02 -4.20575321e-01 -6.19695485e-01 4.45034117e-01 1.05117643e+00 -3.82169485e-01 -7.76712120e-01 8.66278648e-01 -9.55504104e-02 -4.64249820e-01 3.55567575e-01 -7.70607948e-01 2.06653208e-01 5.91002777e-02 5.43072104e-01 4.61230189e-01 7.81126916e-02 -5.63965917e-01 -1.64179057e-01 6.05711520e-01 1.44680753e-01 -3.24551314e-01 1.31939495e+00 -1.97692633e-01 -1.49593681e-01 1.51959360e+00 1.63706625e+00 -4.27906096e-01 -1.38774610e+00 -3.71320575e-01 8.11045915e-02 2.52862841e-01 1.03124410e-01 -2.67755896e-01 -1.24569750e+00 1.31590176e+00 3.65202010e-01 7.23540723e-01 8.70712817e-01 -3.85541558e-01 1.22116959e+00 4.73206788e-01 4.29214239e-02 -9.63598549e-01 -8.91694650e-02 1.29496753e+00 8.87039065e-01 -8.23829830e-01 -7.10530579e-01 4.76177961e-01 -2.81944513e-01 1.44340193e+00 2.51208276e-01 -5.06498754e-01 1.05174255e+00 4.42123652e-01 1.37145072e-01 -4.22150977e-02 -1.31715107e+00 -7.70869374e-04 1.29443228e-01 1.59280434e-01 5.18970013e-01 -7.32999370e-02 1.55274898e-01 4.02313054e-01 -1.66643858e-01 -4.12859395e-02 4.71235603e-01 8.38846803e-01 -2.16576293e-01 -7.13901818e-01 1.79417999e-04 1.74956024e-01 -3.65345806e-01 -7.16704503e-02 -1.59051076e-01 4.36309427e-01 -5.08027256e-01 6.81673348e-01 3.02100897e-01 -3.08738887e-01 1.70997828e-01 4.10944134e-01 2.48811781e-01 -3.40198696e-01 -5.23097038e-01 -1.11336252e-02 -3.63159657e-01 -6.69679284e-01 -1.81355290e-02 -5.50761700e-01 -1.51621509e+00 -4.79248703e-01 -1.33443356e-01 1.43741444e-01 6.40630543e-01 1.05983770e+00 5.42515695e-01 7.06785440e-01 7.32154787e-01 -1.06611836e+00 -8.17715168e-01 -8.17257285e-01 -4.71826881e-01 8.91932622e-02 1.34083557e+00 -2.29211539e-01 -7.87639201e-01 -1.37571126e-01]
[7.309007167816162, 3.370670795440674]
9a680f3a-700b-437d-a220-c48f579a7e44
a-step-towards-interpretable-multi-hop
null
null
https://aclanthology.org/2022.lrec-1.485
https://aclanthology.org/2022.lrec-1.485.pdf
A STEP towards Interpretable Multi-Hop Reasoning:Bridge Phrase Identification and Query Expansion
We propose an unsupervised method for the identification of bridge phrases in multi-hop question answering (QA). Our method constructs a graph of noun phrases from the question and the available context, and applies the Steiner tree algorithm to identify the minimal sub-graph that connects all question phrases. Nodes in the sub-graph that bridge loosely-connected or disjoint subsets of question phrases due to low-strength semantic relations are extracted as bridge phrases. The identified bridge phrases are then used to expand the query based on the initial question, helping in increasing the relevance of evidence that has little lexical overlap or semantic relation with the question. Through an evaluation on HotpotQA, a popular dataset for multi-hop QA, we show that our method yields: (a) improved evidence retrieval, (b) improved QA performance when using the retrieved sentences; and (c) effective and faithful explanations when answers are provided.
['Mihai Surdeanu', 'Fan Luo']
null
null
null
null
lrec-2022-6
['multi-hop-question-answering']
['knowledge-base']
[ 2.04261672e-02 7.00840354e-01 -2.83757865e-01 -2.51522452e-01 -1.46324742e+00 -8.13128889e-01 1.74561515e-01 9.38589692e-01 -2.03670129e-01 8.45805645e-01 6.19583786e-01 -5.03454983e-01 -6.99016690e-01 -9.29324806e-01 -5.65724134e-01 -2.63747901e-01 1.80207565e-01 9.03460741e-01 1.13640130e+00 -6.05935454e-01 4.68059093e-01 1.22691683e-01 -1.24172187e+00 3.38956416e-01 1.06532216e+00 6.28351450e-01 2.13681981e-01 5.26071310e-01 -6.65755630e-01 7.47985840e-01 -5.83706319e-01 -5.24521232e-01 -1.52270645e-01 -5.79955995e-01 -1.82666755e+00 -4.14702818e-02 3.00937831e-01 1.40410632e-01 -6.40169159e-02 8.54870141e-01 2.31211632e-01 1.54843509e-01 3.01555187e-01 -8.56673598e-01 -3.13819915e-01 8.46100092e-01 -1.69571325e-01 5.58325350e-01 1.04023814e+00 -4.58151728e-01 1.85203993e+00 -9.06610906e-01 9.65828896e-01 1.17871046e+00 3.26442271e-01 4.31936622e-01 -1.04729521e+00 -1.56939477e-01 -1.11390322e-01 4.86642569e-01 -1.21907234e+00 -1.92700863e-01 7.45388031e-01 2.57390648e-01 1.17677379e+00 4.73317653e-01 3.55864763e-01 5.83702624e-01 1.58896402e-01 3.37282836e-01 7.99474657e-01 -8.71045709e-01 3.01624328e-01 2.36341953e-01 8.29351246e-01 8.41931164e-01 2.11210232e-02 -6.04522347e-01 -6.71616077e-01 -8.27045739e-01 -9.88670290e-02 -7.92178154e-01 -3.91547740e-01 -6.23295978e-02 -7.32272744e-01 1.11561310e+00 5.63259840e-01 5.26577473e-01 -6.41000330e-01 -8.99975896e-02 -7.86025971e-02 3.38250220e-01 2.40036454e-02 7.83295333e-01 -5.51262975e-01 1.67480841e-01 -5.70686460e-01 3.29068780e-01 1.28762639e+00 9.23976481e-01 9.15993512e-01 -9.51511681e-01 -7.62557685e-02 8.41476381e-01 4.74948496e-01 3.18675697e-01 -9.07635840e-04 -1.25992906e+00 6.45665467e-01 1.01719713e+00 5.55790737e-02 -1.15781295e+00 -3.31122160e-01 -6.50474727e-02 7.89591298e-02 -8.68587792e-01 1.74427882e-01 2.27467492e-01 -4.49813157e-01 1.72366321e+00 6.85809076e-01 -4.82085586e-01 2.90398210e-01 6.58750355e-01 1.07774305e+00 6.82243884e-01 1.75426975e-01 -3.66674989e-01 1.96664226e+00 -8.50138366e-01 -7.51824141e-01 -3.62385690e-01 6.98179185e-01 -8.79617453e-01 1.00930250e+00 -2.50255615e-01 -1.16180480e+00 -7.36638084e-02 -6.70644462e-01 -3.36841613e-01 -2.38502920e-01 -5.73169112e-01 3.64469253e-02 2.83742845e-01 -1.22840321e+00 1.21268712e-01 -5.29609211e-02 -5.90754807e-01 -1.30421385e-01 3.91447783e-01 -1.56992897e-01 -3.56134117e-01 -1.67766798e+00 9.33284044e-01 4.71922129e-01 -4.51979160e-01 -4.35062379e-01 -4.66539413e-01 -8.14563036e-01 3.08671653e-01 9.09174740e-01 -1.05940115e+00 1.12900770e+00 -2.89312512e-01 -7.40606546e-01 8.01631689e-01 -6.48928821e-01 -4.04953182e-01 -4.61334825e-01 -4.27606590e-02 -3.50928932e-01 1.17082429e+00 6.82475448e-01 8.63815248e-01 6.55124784e-01 -1.44046140e+00 -6.86951756e-01 -3.14196944e-01 5.58239281e-01 4.40033048e-01 -1.48492500e-01 2.34292030e-01 -6.77309334e-01 -2.14194119e-01 6.37280762e-01 -7.05521107e-01 -3.06006372e-02 -4.71685469e-01 -5.81725359e-01 -8.42803121e-01 9.00017500e-01 -8.59373808e-01 1.44133091e+00 -1.70665288e+00 2.89813876e-01 5.32368839e-01 3.61619055e-01 -3.02837372e-01 -1.87501565e-01 8.86125445e-01 2.46538103e-01 3.43161345e-01 -4.43293899e-01 3.07758093e-01 -1.02022626e-01 7.42715061e-01 -6.08901143e-01 -3.12285274e-01 1.04925834e-01 9.54733670e-01 -1.06650841e+00 -1.22618437e+00 -5.30182898e-01 -1.97748661e-01 -5.54633677e-01 1.49922892e-01 -7.43209064e-01 3.33635956e-01 -8.62019002e-01 7.35085785e-01 9.92886275e-02 -5.13613343e-01 2.01008797e-01 -2.54992604e-01 4.41673160e-01 9.20946002e-01 -7.27118671e-01 1.42441785e+00 -6.63042217e-02 2.99483508e-01 -3.22414376e-02 -7.45198786e-01 9.12411869e-01 6.22412622e-01 3.80301833e-01 -7.05875695e-01 -9.63400975e-02 5.34092367e-01 -1.73900202e-01 -9.48568523e-01 6.42093539e-01 -5.49592357e-03 -2.66332895e-01 4.77079809e-01 1.44902781e-01 -4.35350358e-01 6.50728285e-01 9.24079120e-01 1.32690489e+00 -3.00730586e-01 3.02559942e-01 -4.42430347e-01 9.13326859e-01 5.24884403e-01 1.46388635e-01 7.16290534e-01 7.31402040e-02 1.47043467e-01 5.72762668e-01 7.32631981e-02 -7.87105501e-01 -1.03912890e+00 -1.19994588e-01 1.01409125e+00 4.70806986e-01 -7.33619571e-01 -8.08327675e-01 -1.04340696e+00 -1.29172653e-01 1.01566780e+00 -2.74386972e-01 -8.14316049e-02 -8.05958092e-01 -2.48766467e-02 4.11368817e-01 2.55699664e-01 4.24228907e-01 -1.16558444e+00 -4.06332374e-01 3.35999131e-01 -1.04529321e+00 -1.33059692e+00 -3.34296077e-01 5.52939214e-02 -9.77955163e-01 -1.38982129e+00 -2.73820937e-01 -1.07172012e+00 7.83838153e-01 3.76502782e-01 1.36048222e+00 7.05114782e-01 2.71925271e-01 9.34024036e-01 -5.23037910e-01 1.39456600e-01 -4.89416808e-01 9.57725719e-02 -3.81321937e-01 -5.71423173e-01 4.77949500e-01 -3.10857981e-01 -5.25623977e-01 5.14330864e-01 -9.06543851e-01 -5.87134957e-01 1.53115198e-01 7.13929057e-01 6.77950203e-01 -7.33788358e-03 9.94232655e-01 -6.79674029e-01 1.09316051e+00 -7.17714667e-01 -3.48291874e-01 6.86644018e-01 -7.19808996e-01 2.27611899e-01 2.13984072e-01 9.44510847e-02 -1.02000403e+00 -6.12922430e-01 -3.75485808e-01 3.13068151e-01 2.76871286e-02 8.98480296e-01 1.44487500e-01 -1.28147230e-01 8.29973519e-01 -7.13284612e-02 -4.55910772e-01 -3.39419991e-01 6.35050416e-01 5.60479283e-01 3.82762760e-01 -8.89753997e-01 6.35707319e-01 3.18896145e-01 5.41112535e-02 -7.18043208e-01 -1.11738944e+00 -9.18084264e-01 -5.02262175e-01 -1.34052828e-01 9.93529677e-01 -3.04240882e-01 -4.05290842e-01 -7.04695940e-01 -1.41186190e+00 3.60208124e-01 -5.61454058e-01 3.07743073e-01 -2.58064270e-01 7.70895660e-01 -6.26134217e-01 -7.54513323e-01 -4.11332786e-01 -8.58148158e-01 1.06117404e+00 4.01973844e-01 -6.42014623e-01 -9.52388763e-01 2.14849025e-01 1.10092032e+00 -6.01312667e-02 -3.48924577e-01 1.83864224e+00 -1.08489347e+00 -6.85897529e-01 -2.86710173e-01 -9.71976891e-02 -2.16314234e-02 1.32567650e-02 -3.97658467e-01 -5.51405251e-01 -2.52915006e-02 1.37158319e-01 -4.98211890e-01 5.55111945e-01 3.31135541e-02 4.79330570e-01 -4.52214271e-01 -4.74159122e-01 -4.93257850e-01 1.29067922e+00 1.67477086e-01 3.97257507e-01 4.23878878e-01 1.26409426e-01 1.15957999e+00 6.87617302e-01 -1.51977122e-01 6.63204074e-01 4.11924064e-01 3.03129613e-01 3.73510778e-01 -1.12987369e-01 -3.97323191e-01 6.13853708e-02 1.27024591e+00 3.40982676e-01 -3.90798450e-01 -9.98860657e-01 9.88046885e-01 -1.72669840e+00 -6.46338820e-01 -5.57916522e-01 1.73900437e+00 1.02203214e+00 1.97442695e-01 1.69190839e-02 1.38219729e-01 6.90275133e-01 -5.05265594e-02 -1.45442411e-01 -5.48702002e-01 -2.91483730e-01 6.17481470e-01 -8.11517015e-02 8.72745752e-01 -4.35731322e-01 1.10926521e+00 6.64676142e+00 7.52680421e-01 -6.20682798e-02 1.86889753e-01 1.80518940e-01 5.15218318e-01 -9.83890414e-01 5.99527657e-01 -7.52610505e-01 -1.12785697e-01 8.23871017e-01 -1.89921007e-01 -1.72483157e-02 3.48779112e-01 -1.21964522e-01 -5.51264703e-01 -7.64753222e-01 3.73275466e-02 8.38556737e-02 -1.42077923e+00 2.94985861e-01 -2.09188342e-01 5.58077514e-01 -2.30991468e-01 -4.69736099e-01 8.17942470e-02 7.96582550e-02 -5.02172470e-01 2.19695121e-01 1.89970151e-01 2.35394508e-01 -6.61501169e-01 8.88048470e-01 5.68277776e-01 -1.18780363e+00 -1.36743620e-01 -3.92264396e-01 4.34995949e-01 5.09244144e-01 3.50535899e-01 -8.15877855e-01 1.01880991e+00 7.53447831e-01 -1.18794687e-01 -5.17177701e-01 8.37983191e-01 -8.81690681e-01 9.24133301e-01 -4.38778549e-01 -3.95830005e-01 5.49592078e-01 -2.50082642e-01 9.73642230e-01 8.61560225e-01 7.09294900e-02 6.42770827e-01 2.99464054e-02 6.51574910e-01 -2.58310497e-01 5.56766093e-01 -4.20527756e-01 2.86821369e-03 7.12264776e-01 1.07075429e+00 -7.87762523e-01 -4.37135905e-01 -4.50119317e-01 5.90371668e-01 2.48181254e-01 4.14533108e-01 -3.71211231e-01 -3.88656378e-01 -1.78117886e-01 -1.30833983e-01 2.26904184e-01 2.59360056e-02 1.04360655e-01 -6.85512424e-01 3.60874206e-01 -1.06148696e+00 1.24245441e+00 -1.11374509e+00 -1.24111605e+00 7.40757406e-01 1.31702378e-01 -6.23755097e-01 -2.72464871e-01 -5.64557388e-02 -3.89107734e-01 8.32019210e-01 -1.46415901e+00 -8.10497046e-01 6.82995990e-02 7.01095343e-01 4.75246817e-01 2.50271291e-01 1.02112734e+00 3.67674455e-02 6.37265071e-02 3.86949070e-02 -5.11920750e-01 -2.38032132e-01 3.91983867e-01 -1.22562289e+00 -2.10679591e-01 5.83978117e-01 5.02606630e-01 7.89774179e-01 7.24732935e-01 -8.43121588e-01 -1.14554131e+00 -4.27636087e-01 1.87127924e+00 -5.40238023e-01 8.91900837e-01 1.59852490e-01 -1.37776804e+00 4.51557040e-01 6.48779213e-01 -6.34392440e-01 8.09005737e-01 7.25533813e-02 -4.07935798e-01 8.34066942e-02 -1.30816078e+00 5.65272093e-01 7.13510931e-01 -7.37795174e-01 -1.84427047e+00 7.76273131e-01 1.37514985e+00 -8.98093730e-02 -8.46786678e-01 3.72156203e-01 -5.29205687e-02 -5.36449671e-01 9.38246250e-01 -8.30288708e-01 4.27727222e-01 -2.16463253e-01 -3.39870900e-01 -7.86332250e-01 -1.38566568e-01 -6.56817973e-01 -1.27887130e-01 1.11265063e+00 9.84429121e-01 -4.69084591e-01 6.73677683e-01 5.60168386e-01 -1.13935314e-01 -8.74858856e-01 -1.30887151e+00 -4.57404971e-01 -3.97039764e-02 -1.91812530e-01 3.75512898e-01 7.61245787e-01 5.14074743e-01 8.82884383e-01 3.03050905e-01 3.87086093e-01 5.69673121e-01 4.83629018e-01 1.75695032e-01 -1.14498818e+00 -1.15647018e-01 5.85358590e-03 8.92923772e-02 -1.40355825e+00 2.62530208e-01 -9.90226448e-01 1.01735383e-01 -2.06031108e+00 2.50557363e-01 -2.87784636e-01 1.13458097e-01 3.08522463e-01 -3.98597687e-01 -1.05489321e-01 -2.31791675e-01 4.37665224e-01 -8.51985514e-01 3.62616807e-01 1.14254797e+00 -5.21954261e-02 -1.70899630e-02 -2.15598971e-01 -7.90181696e-01 6.36263371e-01 6.07885122e-01 -8.30833793e-01 -3.67833674e-01 -2.90041149e-01 6.44090652e-01 4.45104867e-01 2.27425471e-01 -3.93781245e-01 5.29385328e-01 6.16844669e-02 -4.44657862e-01 -7.67588556e-01 2.65451223e-01 -8.86857867e-01 -4.12330866e-01 5.56710303e-01 -6.48574829e-01 1.94906712e-01 8.09396505e-02 6.45387232e-01 -5.08978724e-01 -9.89792109e-01 2.87560135e-01 -1.11111522e-01 -5.25263488e-01 -1.42803892e-01 -3.19730014e-01 4.45515960e-01 5.49192309e-01 -6.81477264e-02 -5.60352564e-01 -6.92426145e-01 -8.03971767e-01 6.71691298e-01 -7.48435706e-02 2.41778716e-01 8.90327752e-01 -1.17128325e+00 -6.16819680e-01 -5.69187880e-01 3.71136159e-01 -9.85876173e-02 -2.73143239e-02 6.97661400e-01 -2.55765259e-01 8.06447089e-01 4.41033244e-01 -4.98398960e-01 -1.38814127e+00 4.35371071e-01 -6.13705022e-03 -5.30658364e-01 -6.37674928e-01 8.38512003e-01 -2.75816709e-01 -2.35605776e-01 6.45346791e-02 -7.67846107e-02 -6.13747180e-01 1.06245823e-01 8.77315700e-02 2.23618060e-01 1.24579832e-01 -7.55147338e-01 -4.33416039e-01 7.29707837e-01 1.12392485e-01 -5.82948446e-01 1.02677560e+00 -5.62369406e-01 -7.34767437e-01 2.46023521e-01 9.45656180e-01 4.31598544e-01 -1.78998664e-01 -7.35994399e-01 7.62601733e-01 -2.69958168e-01 -2.29874462e-01 -7.32807398e-01 -4.35648799e-01 3.21451426e-01 -5.89989759e-02 1.02511692e+00 1.07340860e+00 8.73792887e-01 1.21453154e+00 7.38984644e-01 4.46917802e-01 -9.09327388e-01 5.06912291e-01 5.64231515e-01 9.93379533e-01 -7.33079255e-01 -1.94227636e-01 -9.00905013e-01 -3.50282729e-01 1.12116611e+00 4.07917768e-01 3.18151206e-01 3.47034305e-01 -3.38220090e-01 -1.34172719e-02 -1.07309031e+00 -7.94065714e-01 -4.02131557e-01 5.01330972e-01 3.02359045e-01 -9.39069875e-03 -5.40804327e-01 -9.78715956e-01 5.36389530e-01 -3.01004976e-01 -5.14291525e-01 3.19513768e-01 1.08645272e+00 -1.02764392e+00 -1.15126109e+00 -4.79577839e-01 2.37696737e-01 -4.74696755e-01 -2.94119626e-01 -9.98967707e-01 7.83266187e-01 -1.54162183e-01 1.66476154e+00 -8.52644891e-02 -1.15184598e-02 4.82469410e-01 4.62877810e-01 4.40537393e-01 -7.93840885e-01 -7.51382172e-01 -2.34582908e-02 7.45050073e-01 -3.62010598e-01 -5.44218719e-01 -1.29871890e-01 -1.67071772e+00 9.91195366e-02 -7.85385728e-01 1.10648894e+00 4.28462625e-01 1.39126587e+00 4.32917088e-01 1.48820892e-01 5.53137422e-01 5.31281352e-01 -5.33576608e-01 -7.94726968e-01 -1.47784203e-01 4.90124822e-01 -1.53126325e-02 -3.29622984e-01 -5.00232577e-01 -1.82804003e-01]
[10.869186401367188, 7.953239440917969]
e5a98501-bda2-4a9e-b2f1-12639cf070b2
reusable-phrase-extraction-based-on-syntactic
null
null
https://aclanthology.org/2020.ccl-1.108
https://aclanthology.org/2020.ccl-1.108.pdf
Reusable Phrase Extraction Based on Syntactic Parsing
Academic Phrasebank is an important resource for academic writers. Student writers use the phrases of Academic Phrasebank organizing their research article to improve their writing ability. Due to the limited size of Academic Phrasebank, it can not meet all the academic writing needs. There are still a large number of academic phraseology in the authentic research article. In this paper, we proposed an academic phraseology extraction model based on constituency parsing and dependency parsing, which can automatically extract the academic phraseology similar to phrases of Academic Phrasebank from an unlabelled research article. We divided the proposed model into three main components including an academic phraseology corpus module, a sentence simplification module, and a syntactic parsing module. We created a corpus of academic phraseology of 2,129 words to help judge whether a word is neutral and general, and created two datasets under two scenarios to verify the feasibility of the proposed model.
['Christoph Zähner', 'Xiaojing Bai', 'Zan Hongying', 'Xuemin Duan']
null
null
null
null
ccl-2020-10
['constituency-parsing']
['natural-language-processing']
[-3.63636553e-01 3.51981103e-01 -4.35050488e-01 -7.07816407e-02 -5.65573931e-01 -7.16831744e-01 3.85589093e-01 2.75427848e-01 -3.63274425e-01 9.13999081e-01 5.36623538e-01 -7.46996582e-01 1.08570419e-01 -8.61140609e-01 -2.45389923e-01 -3.09958369e-01 9.29544866e-01 2.98019528e-01 1.70919269e-01 -4.07890290e-01 9.97440100e-01 5.44004023e-01 -9.72692370e-01 -3.61032449e-02 1.15638053e+00 8.66059810e-02 5.13493359e-01 4.69115049e-01 -9.60568547e-01 3.80388498e-01 -1.08721793e+00 -5.03723383e-01 -3.95523086e-02 -3.59460086e-01 -6.55641556e-01 1.88098603e-03 -1.21512987e-01 1.56383161e-02 -4.63764966e-02 1.38677430e+00 5.09217560e-01 1.04386374e-01 4.22211379e-01 -8.84462476e-01 -1.14806700e+00 1.11708248e+00 -4.69676346e-01 5.62379181e-01 4.32471931e-01 -2.78303057e-01 1.41275418e+00 -6.96706355e-01 6.53382123e-01 1.51656640e+00 3.99572015e-01 3.91989648e-01 -5.56401551e-01 -1.11810565e+00 3.99746388e-01 1.04859099e-01 -1.07747924e+00 -2.26203665e-01 7.71915913e-01 -3.13570619e-01 8.42788875e-01 1.20108560e-01 9.05043006e-01 1.09644604e+00 7.94776082e-01 5.54987729e-01 1.13893175e+00 -5.94903171e-01 -7.04766512e-02 4.68083560e-01 9.83827233e-01 2.30419472e-01 5.44798791e-01 -6.04456782e-01 -3.24072033e-01 -1.73980054e-02 6.21473014e-01 -2.36786976e-01 -2.50977367e-01 7.12415278e-01 -9.49256182e-01 8.70450556e-01 -4.36873704e-01 4.70375657e-01 -1.61224142e-01 -4.67309803e-01 3.86633515e-01 1.37288541e-01 1.96512997e-01 5.16436636e-01 -3.87277097e-01 -1.29551873e-01 -5.96867979e-01 5.92502534e-01 1.09206033e+00 1.46974325e+00 1.29332393e-01 -8.77772793e-02 -2.21823275e-01 6.67587936e-01 6.94075227e-01 5.90926409e-01 7.70459414e-01 -6.40315652e-01 5.40810168e-01 5.18940330e-01 2.15889998e-02 -8.03352237e-01 -1.21822208e-01 -5.78482091e-01 -4.01034653e-01 -6.99214995e-01 -1.87141672e-01 -2.71837592e-01 -4.96116221e-01 1.36924314e+00 1.02561563e-02 -1.61305308e-01 4.86001879e-01 5.88889539e-01 1.59642768e+00 1.18935454e+00 5.17432332e-01 -6.06597364e-01 1.92732370e+00 -1.05943012e+00 -1.38687015e+00 -3.02972376e-01 5.77676892e-01 -1.40269184e+00 1.05486774e+00 1.87659457e-01 -1.25148201e+00 -4.23767954e-01 -9.59162116e-01 -3.19317579e-01 -4.66069937e-01 -3.98878828e-02 3.17753464e-01 7.89364278e-01 -6.23534501e-01 2.14381590e-01 -3.77145350e-01 -2.82539576e-01 2.35238820e-01 2.84387022e-02 -6.31747842e-02 1.33930817e-01 -1.36838138e+00 8.54368091e-01 7.52971530e-01 -1.38934240e-01 6.89884499e-02 -7.36345708e-01 -7.27808952e-01 4.03537333e-01 1.12445511e-01 -4.66217637e-01 1.15889382e+00 -3.75014335e-01 -1.42976904e+00 7.81966567e-01 -3.15737069e-01 1.06751107e-01 -1.40375987e-01 -3.14661741e-01 -4.98361468e-01 -2.09777579e-01 5.35940051e-01 1.34408683e-01 4.31162834e-01 -8.12421739e-01 -9.97363031e-01 -3.29359502e-01 -1.09320305e-01 3.04415792e-01 -4.55072314e-01 6.94524527e-01 -5.91820121e-01 -1.17125463e+00 -2.11209375e-02 -7.97571599e-01 1.32657657e-03 -9.67135370e-01 -3.23256969e-01 -9.04895902e-01 7.43326545e-01 -9.88220811e-01 1.60140193e+00 -1.82589769e+00 9.70691666e-02 5.39469756e-02 1.31012555e-02 3.51037651e-01 9.78796463e-03 1.67905867e-01 -9.20415670e-02 7.40785241e-01 1.82947189e-01 -1.99046638e-02 1.34381890e-01 4.44431424e-01 -5.18375576e-01 -8.33586007e-02 1.55411422e-01 7.99943149e-01 -9.98365104e-01 -9.26033139e-01 -4.76574339e-02 4.32155915e-02 -3.29685569e-01 2.34123483e-01 -5.17278053e-02 2.20275030e-01 -1.10787630e+00 9.65709746e-01 7.59495676e-01 2.54578263e-01 -2.31922716e-02 1.49658144e-01 -6.05601251e-01 4.31279689e-01 -1.00609195e+00 1.47099948e+00 -3.76385480e-01 5.07913411e-01 -9.93433967e-03 -7.50125110e-01 1.22182906e+00 5.25821209e-01 7.74532184e-02 -5.40492117e-01 3.04091811e-01 1.77037582e-01 1.91430017e-01 -7.86279202e-01 7.60693491e-01 -3.31389189e-01 -3.45669568e-01 4.34307843e-01 3.84633690e-02 -3.50998938e-01 4.81850743e-01 1.25947967e-01 9.93877709e-01 2.54409224e-01 3.46214890e-01 -6.45084977e-01 1.08459520e+00 1.20097600e-01 1.13221705e+00 4.55052286e-01 -2.09809542e-01 2.73712695e-01 7.37781525e-01 -2.22081482e-01 -1.08429742e+00 -6.93900526e-01 -4.63110566e-01 9.30082619e-01 1.55268297e-01 -7.86415815e-01 -7.24144697e-01 -5.50409377e-01 -3.80776942e-01 1.22531462e+00 1.19015954e-01 2.49239743e-01 -9.42018449e-01 -9.13711607e-01 3.88883799e-01 2.58870453e-01 5.21618903e-01 -1.38450575e+00 -3.48567329e-02 3.82239610e-01 -2.02993616e-01 -1.32451952e+00 -3.88646215e-01 -8.25794786e-02 -5.63334882e-01 -9.67064977e-01 -3.30544651e-01 -1.14597452e+00 5.13964772e-01 4.54582758e-02 8.76450717e-01 1.26671726e-02 1.67774111e-01 -2.61398077e-01 -6.99529648e-01 -1.07538295e+00 -7.10887611e-01 4.98416424e-01 -1.44321606e-01 -9.64418232e-01 9.64807630e-01 -4.63407904e-01 -4.28541414e-02 -1.77224725e-01 -6.27637029e-01 -1.09553076e-01 8.73679280e-01 5.51878750e-01 3.92735690e-01 2.49403968e-01 8.31993759e-01 -1.05431652e+00 1.32916784e+00 -4.71651912e-01 -4.34091657e-01 2.32961342e-01 -6.07091248e-01 -9.88737494e-02 1.01694071e+00 -1.48197234e-01 -1.38023067e+00 -6.78550422e-01 -7.21659899e-01 6.89408109e-02 -1.12815410e-01 7.41271377e-01 -4.81562257e-01 1.77016482e-01 1.10820398e-01 1.19501933e-01 -3.06498021e-01 -5.62740326e-01 3.69713120e-02 1.09793031e+00 3.06581348e-01 -9.47351098e-01 8.52701783e-01 -5.54352462e-01 -6.38309941e-02 -8.84722650e-01 -1.17287374e+00 -4.72897917e-01 -6.20317638e-01 2.71395454e-03 8.55570018e-01 -7.82146156e-01 -7.80599475e-01 -9.80661884e-02 -1.66040838e+00 4.83293146e-01 -2.90691644e-01 6.28173053e-01 5.43808527e-02 6.59640193e-01 -8.37566376e-01 -4.26037788e-01 -7.53746092e-01 -1.26800406e+00 7.45543242e-01 7.95092702e-01 -2.64695257e-01 -9.89401877e-01 1.14386277e-02 5.84249854e-01 -1.01282589e-01 2.46947855e-02 1.26971674e+00 -9.41783190e-01 -2.66388245e-02 -1.01492442e-01 -1.18071295e-01 2.95077056e-01 1.03558011e-01 2.91041613e-01 -6.08080804e-01 4.15432192e-02 3.43402565e-01 1.36287764e-01 2.61346400e-01 6.42448962e-02 9.25910830e-01 -4.76464123e-01 -1.50462016e-01 3.32061768e-01 1.10713696e+00 5.19602060e-01 5.75294852e-01 9.82243717e-01 9.12244499e-01 6.11863613e-01 8.24949324e-01 1.66525468e-01 4.12193924e-01 1.63494289e-01 -2.94709504e-01 5.82258582e-01 1.45823285e-01 -6.61615729e-02 5.31743526e-01 1.75179386e+00 -1.41889423e-01 -3.47007871e-01 -1.03847027e+00 5.79888344e-01 -1.71978128e+00 -7.67908931e-01 -4.97753710e-01 1.43750274e+00 1.06829989e+00 6.11198545e-01 -4.15904582e-01 -9.71278623e-02 7.27028966e-01 7.57076219e-02 4.16066162e-02 -9.83043313e-01 -8.30458552e-02 5.60771346e-01 5.89147627e-01 4.63523477e-01 -7.49227405e-01 1.33782351e+00 6.20265293e+00 9.35219765e-01 -7.78623462e-01 -1.94436740e-02 1.28584117e-01 4.72687334e-01 -5.53617656e-01 3.13334584e-01 -1.62087178e+00 6.05312645e-01 1.12186241e+00 -8.04374635e-01 -3.55980098e-01 8.20369422e-01 5.68403900e-01 3.82082880e-01 -8.01814675e-01 6.48267508e-01 3.72147597e-02 -1.32177866e+00 5.00799537e-01 -2.26641502e-02 3.11559618e-01 -4.50215846e-01 -2.59132355e-01 5.29056489e-01 3.60275395e-02 -1.03855598e+00 5.14501452e-01 3.67205650e-01 -3.03384811e-02 -1.04160619e+00 1.22628713e+00 6.30722106e-01 -9.37786996e-01 1.97603449e-01 -6.46865427e-01 -3.31188589e-01 4.91560139e-02 3.82828951e-01 -5.31607687e-01 7.24126577e-01 5.45575738e-01 9.49487090e-01 -4.83370483e-01 6.39510691e-01 -7.57527947e-01 8.24326754e-01 -2.09578052e-02 -2.99391001e-01 3.84314328e-01 -6.82211399e-01 7.65257955e-01 1.40516651e+00 2.30729878e-01 6.42990649e-01 2.73078114e-01 9.03815925e-01 -2.73143649e-01 8.56720805e-01 -3.45009536e-01 -3.44361365e-01 9.28345084e-01 1.28259599e+00 -7.63881326e-01 -3.00586462e-01 -7.18358040e-01 3.22164804e-01 -9.07612145e-02 1.36431858e-01 -6.08008206e-01 -8.75774682e-01 2.20772177e-01 -7.71022439e-02 1.65917482e-02 -1.39184490e-01 -6.38640523e-01 -1.23948836e+00 1.55646816e-01 -1.00140584e+00 1.70405835e-01 -5.65385640e-01 -1.11201990e+00 2.82393664e-01 6.58890828e-02 -7.67044127e-01 1.78474277e-01 -6.87608004e-01 -1.08075154e+00 1.30514491e+00 -1.70206988e+00 -1.18480742e+00 -1.07429802e-01 -1.78525168e-02 1.12066281e+00 -3.51801544e-01 8.93735766e-01 1.80959970e-01 -1.20746779e+00 4.76490945e-01 -6.82671517e-02 2.89940417e-01 8.50767195e-01 -1.13626456e+00 3.13885421e-01 9.73571539e-01 -5.11383116e-01 1.18724751e+00 9.22442377e-01 -7.27501214e-01 -1.30811346e+00 -1.01004112e+00 1.42961311e+00 -4.21268523e-01 8.84465218e-01 -6.29005134e-02 -1.19388139e+00 9.51705575e-01 6.10324025e-01 -6.22669399e-01 1.01897430e+00 -2.65915282e-02 3.66189361e-01 2.03792974e-01 -9.58300889e-01 8.75665426e-01 6.41005218e-01 1.10851191e-01 -1.75642133e+00 5.91938019e-01 1.12624216e+00 -4.86310303e-01 -1.44901991e+00 8.08429271e-02 2.76384950e-01 -1.81482837e-01 8.39161396e-01 -5.44392884e-01 6.70049846e-01 -3.07914917e-04 3.43822390e-01 -1.03841686e+00 -5.28760076e-01 -7.73684621e-01 1.23262502e-01 1.77478540e+00 2.56408572e-01 -4.85005468e-01 7.44185627e-01 7.51925349e-01 -4.53647673e-01 -5.23268282e-01 -5.57874560e-01 -4.36981708e-01 5.63244760e-01 -2.04196703e-02 6.22341156e-01 1.09591913e+00 1.82921335e-01 1.22985089e+00 3.34855437e-01 6.07612692e-02 5.41007936e-01 1.37277201e-01 4.21063602e-01 -1.48058736e+00 2.38385394e-01 -7.02864349e-01 -1.28056511e-01 -8.59534681e-01 9.71870959e-01 -7.58165896e-01 -3.47671092e-01 -1.68891859e+00 1.39016345e-01 -5.00502706e-01 -8.87335166e-02 4.40953434e-01 -4.91754621e-01 -3.84930909e-01 -8.39627162e-02 4.80388880e-01 -7.00109079e-02 5.05442202e-01 1.38357747e+00 -4.63963822e-02 -6.02496505e-01 -1.78203225e-01 -1.12636268e+00 8.41601133e-01 9.24799144e-01 -6.39007986e-01 -3.12320679e-01 -1.56629741e-01 1.55571416e-01 6.09260052e-03 -3.69294405e-01 -2.57178158e-01 3.11105967e-01 -5.92736006e-01 1.49783909e-01 -6.64512992e-01 -3.90212536e-01 -5.42756200e-01 -3.62580627e-01 2.97035187e-01 -1.51463211e-01 4.09754336e-01 4.83521432e-01 9.38952267e-02 -3.01961154e-01 -1.02192616e+00 4.84594971e-01 -4.53838766e-01 -7.55191326e-01 -1.54872388e-01 -4.86967713e-01 3.01402628e-01 1.15663326e+00 -1.11153118e-01 -4.45128322e-01 4.05698895e-01 -4.58370060e-01 4.18370605e-01 2.54399002e-01 5.74076653e-01 6.06972039e-01 -1.04931259e+00 -8.18091869e-01 -6.05799295e-02 -8.58901218e-02 4.41016376e-01 -2.43041933e-01 4.71707642e-01 -1.01181555e+00 7.88698316e-01 -4.67099190e-01 7.19351545e-02 -1.39750183e+00 3.72421861e-01 -3.13726127e-01 -4.54774261e-01 -9.18668687e-01 5.62330306e-01 -1.01566814e-01 -3.90097171e-01 1.44968748e-01 -2.79835671e-01 -1.03827596e+00 3.98461595e-02 3.80265981e-01 1.79738522e-01 -2.47127376e-02 -8.46655846e-01 -1.58225626e-01 6.79338694e-01 -4.54794824e-01 3.76001187e-02 1.25883329e+00 -1.07154481e-01 -7.33781517e-01 2.00584114e-01 7.47960091e-01 4.62257892e-01 -1.44905850e-01 4.01633270e-02 2.30773017e-01 -3.36364865e-01 3.37539047e-01 -2.35447809e-01 -4.79008347e-01 5.89372694e-01 -1.24300547e-01 1.14878684e-01 7.88400412e-01 -3.76370907e-01 1.15207613e+00 6.43260479e-01 -1.30662113e-01 -1.55375051e+00 -4.33794379e-01 8.65708947e-01 6.57919586e-01 -1.00171900e+00 3.91575783e-01 -6.70839250e-01 -3.16900879e-01 1.46525919e+00 6.05010688e-01 -1.24440327e-01 6.69903398e-01 2.84410715e-01 -6.17781840e-02 -1.33668333e-01 -4.45886880e-01 3.62938195e-01 2.35668402e-02 3.41675222e-01 7.67550170e-01 -2.15605453e-01 -1.44790328e+00 1.27892232e+00 -7.86292076e-01 -1.40859857e-01 1.09065378e+00 1.23812962e+00 -7.94178188e-01 -1.54054201e+00 -7.08121181e-01 1.85833111e-01 -1.14017224e+00 -3.97481382e-01 -2.55623937e-01 8.28383029e-01 1.13656648e-01 9.23934579e-01 7.71473870e-02 5.77400764e-03 5.36759555e-01 2.55851418e-01 1.61078066e-01 -1.04899192e+00 -7.03086436e-01 2.66003340e-01 1.64280340e-01 1.27831638e-01 -5.45958042e-01 -5.69722593e-01 -1.41078842e+00 -5.96607506e-01 -3.60299647e-01 8.85234118e-01 6.10382617e-01 1.14839435e+00 -3.22224423e-02 8.40634286e-01 3.74633282e-01 -2.01503396e-01 -4.08023357e-01 -1.06040812e+00 -5.36614239e-01 -3.85369547e-03 -2.42356345e-01 -2.86085129e-01 -2.74416864e-01 2.41152078e-01]
[10.476665496826172, 10.022411346435547]
83170b79-0d06-41ae-a0d4-5897a2ce17e0
collaborative-unsupervised-visual
2108.06492
null
https://arxiv.org/abs/2108.06492v1
https://arxiv.org/pdf/2108.06492v1.pdf
Collaborative Unsupervised Visual Representation Learning from Decentralized Data
Unsupervised representation learning has achieved outstanding performances using centralized data available on the Internet. However, the increasing awareness of privacy protection limits sharing of decentralized unlabeled image data that grows explosively in multiple parties (e.g., mobile phones and cameras). As such, a natural problem is how to leverage these data to learn visual representations for downstream tasks while preserving data privacy. To address this problem, we propose a novel federated unsupervised learning framework, FedU. In this framework, each party trains models from unlabeled data independently using contrastive learning with an online network and a target network. Then, a central server aggregates trained models and updates clients' models with the aggregated model. It preserves data privacy as each party only has access to its raw data. Decentralized data among multiple parties are normally non-independent and identically distributed (non-IID), leading to performance degradation. To tackle this challenge, we propose two simple but effective methods: 1) We design the communication protocol to upload only the encoders of online networks for server aggregation and update them with the aggregated encoder; 2) We introduce a new module to dynamically decide how to update predictors based on the divergence caused by non-IID. The predictor is the other component of the online network. Extensive experiments and ablations demonstrate the effectiveness and significance of FedU. It outperforms training with only one party by over 5% and other methods by over 14% in linear and semi-supervised evaluation on non-IID data.
['Shuai Yi', 'Shuai Zhang', 'Yonggang Wen', 'Xin Gan', 'Weiming Zhuang']
2021-08-14
null
http://openaccess.thecvf.com//content/ICCV2021/html/Zhuang_Collaborative_Unsupervised_Visual_Representation_Learning_From_Decentralized_Data_ICCV_2021_paper.html
http://openaccess.thecvf.com//content/ICCV2021/papers/Zhuang_Collaborative_Unsupervised_Visual_Representation_Learning_From_Decentralized_Data_ICCV_2021_paper.pdf
iccv-2021-1
['federated-unsupervised-learning']
['methodology']
[ 8.98178890e-02 4.82456982e-01 -3.55515659e-01 -5.76297939e-01 -6.85186744e-01 -9.09583509e-01 3.75738829e-01 -2.93431878e-01 -5.90518594e-01 7.22147703e-01 3.31077017e-02 -1.75153300e-01 2.25028425e-01 -5.61363816e-01 -9.55493271e-01 -9.24510479e-01 -2.99155712e-04 4.31167692e-01 1.73338830e-01 4.58103478e-01 -3.76471996e-01 5.22899389e-01 -1.29377854e+00 3.28173369e-01 4.65675950e-01 1.41072345e+00 -4.73964168e-03 4.26128268e-01 5.56424223e-02 1.10225987e+00 -3.01228583e-01 -8.18984449e-01 9.93295372e-01 -2.68997490e-01 -6.95482671e-01 1.64703190e-01 2.32990667e-01 -9.64211822e-01 -5.64778149e-01 1.23152399e+00 3.87354523e-01 -1.24613598e-01 3.52279812e-01 -1.63590550e+00 -6.12113059e-01 7.36307025e-01 -5.04252732e-01 -3.48043054e-01 -3.37604970e-01 1.60550177e-01 8.43326867e-01 -3.81242216e-01 8.49282622e-01 8.47944260e-01 4.80185151e-01 8.34561944e-01 -1.31636631e+00 -9.39153314e-01 2.57690072e-01 6.22748174e-02 -1.14275217e+00 -6.71409726e-01 5.66426575e-01 -1.46275327e-01 4.71132874e-01 3.81163627e-01 1.32249013e-01 1.18277514e+00 -2.70390004e-01 7.59975910e-01 1.04936945e+00 -2.14283243e-01 4.10986423e-01 7.51894355e-01 -1.06986910e-02 5.15482068e-01 2.56507665e-01 5.80869280e-02 -6.45862222e-01 -5.27085245e-01 5.98450065e-01 4.03978795e-01 -1.53154746e-01 -7.20735371e-01 -6.94298685e-01 7.48024166e-01 3.75227988e-01 -1.38635188e-01 -3.11206013e-01 1.26648948e-01 5.22462726e-01 7.85280943e-01 4.69795674e-01 -1.04616791e-01 -6.55113816e-01 2.46962056e-01 -5.83831429e-01 -3.34796876e-01 1.11002100e+00 1.18503606e+00 1.14651000e+00 -2.58943021e-01 1.37099996e-01 5.34752071e-01 2.39546180e-01 4.22436208e-01 3.99610072e-01 -1.21541774e+00 6.14753246e-01 4.70815629e-01 -4.15666588e-03 -8.05321872e-01 1.62958652e-01 -1.42191529e-01 -1.04904890e+00 2.25462109e-01 3.73114944e-01 -6.04844987e-01 -4.46729332e-01 1.89881313e+00 4.84895110e-01 3.61218769e-03 2.53542334e-01 9.04499710e-01 2.91357368e-01 4.73631352e-01 -1.03637040e-01 -4.65075076e-01 9.91567373e-01 -1.04299819e+00 -5.04787743e-01 1.35214522e-01 6.48806036e-01 -2.40051568e-01 4.39890593e-01 1.95143834e-01 -9.32406306e-01 4.39087339e-02 -8.90991032e-01 -4.61851545e-02 -2.85985500e-01 6.85963929e-02 5.26124477e-01 6.74977899e-01 -1.13985968e+00 5.24245143e-01 -9.55382884e-01 -2.26798296e-01 8.92085075e-01 8.68674338e-01 -8.71928632e-01 -4.21092622e-02 -7.67108560e-01 2.79639065e-01 2.57693022e-01 -2.62624294e-01 -8.82343054e-01 -4.32380170e-01 -6.79443896e-01 2.90842324e-01 4.14156199e-01 -6.50668442e-01 1.27585530e+00 -1.61265135e+00 -1.57638752e+00 7.94547915e-01 -5.15910983e-02 -8.05480659e-01 9.20381665e-01 -3.15063521e-02 -1.50817722e-01 3.00953925e-01 -2.10893508e-02 4.45410669e-01 8.65183890e-01 -1.44210708e+00 -8.17442000e-01 -6.64497733e-01 6.56315088e-02 1.29866615e-01 -1.02211213e+00 -4.25241236e-03 -5.52479446e-01 -2.92349339e-01 -5.54281250e-02 -1.08884251e+00 -2.10613772e-01 6.12975836e-01 -5.39223969e-01 3.71081792e-02 1.39604914e+00 -5.39272547e-01 7.17121661e-01 -2.54237628e+00 -3.27236682e-01 4.43271697e-01 4.69400436e-01 3.32598567e-01 1.02824355e-02 2.42516279e-01 1.78265557e-01 4.38508429e-02 -1.00150034e-01 -9.09533262e-01 3.93360555e-02 4.34851289e-01 -5.00932217e-01 6.23409212e-01 -1.99532956e-01 6.41950667e-01 -7.31110036e-01 -6.05626047e-01 -7.20834583e-02 2.77441472e-01 -5.85292280e-01 4.12772506e-01 -2.18284708e-02 6.72901690e-01 -5.74754953e-01 5.28607965e-01 8.63368511e-01 -6.49945915e-01 7.53085911e-01 -3.38979401e-02 1.14645891e-01 -1.24335252e-01 -1.18681622e+00 1.40583861e+00 -1.74369544e-01 4.62715954e-01 5.78814089e-01 -1.05430424e+00 6.30016506e-01 5.74701130e-01 5.04025519e-01 -3.99694830e-01 1.46912456e-01 2.60853112e-01 -5.27553916e-01 -3.14475834e-01 -5.97551093e-02 1.56923145e-01 7.58727789e-02 8.48399580e-01 2.80290842e-01 7.11463273e-01 -4.34841305e-01 4.15923893e-01 1.35943222e+00 -3.62124681e-01 2.02871747e-02 1.94526091e-01 3.87410372e-01 -4.03289527e-01 9.28397000e-01 8.90023112e-01 -3.69320631e-01 5.82445383e-01 6.71010256e-01 -5.50472796e-01 -9.77822244e-01 -1.05374038e+00 2.39743978e-01 1.27786696e+00 1.63347349e-01 -3.38470876e-01 -7.44871497e-01 -1.26966119e+00 2.15240732e-01 3.93913984e-01 -4.49400216e-01 -5.98514043e-02 -1.89948127e-01 -2.38967240e-01 4.09895122e-01 3.17546725e-01 8.23089063e-01 -7.09793985e-01 -2.50452131e-01 -6.23202138e-02 2.66430322e-02 -1.10264754e+00 -5.41901469e-01 4.02638853e-01 -7.68751204e-01 -1.06415272e+00 -4.80829477e-01 -6.53481424e-01 1.07465518e+00 4.41453695e-01 5.95210075e-01 -3.05962890e-01 6.04620725e-02 5.97265422e-01 -9.06816497e-02 -4.02635604e-01 -4.45355922e-01 1.16098888e-01 2.38012671e-01 6.56184494e-01 1.87961966e-01 -8.83550406e-01 -6.58184826e-01 4.34470981e-01 -1.06602597e+00 -2.03164488e-01 5.53354800e-01 7.49948144e-01 6.06067717e-01 -1.16742730e-01 4.22002703e-01 -1.36791337e+00 3.16542625e-01 -6.80744350e-01 -9.34929311e-01 4.08167630e-01 -8.08520377e-01 1.72028821e-02 1.14100671e+00 -5.12120187e-01 -1.13175380e+00 5.97064435e-01 6.62343442e-01 -1.00309551e+00 3.55929928e-03 -3.91490459e-02 -4.49697882e-01 -1.53098732e-01 4.05859888e-01 1.34752497e-01 4.22940224e-01 -5.63820720e-01 5.44903874e-01 1.06256211e+00 5.12015224e-01 -2.87039042e-01 1.02814782e+00 8.71552050e-01 -1.38078064e-01 -4.22915220e-01 -4.14840549e-01 -3.36807370e-01 -3.84550631e-01 -1.95306055e-02 6.22816265e-01 -1.17479253e+00 -9.91037786e-01 5.76908708e-01 -1.10151100e+00 -1.67497978e-01 -5.72974443e-01 4.57923681e-01 -4.22100574e-01 4.55169588e-01 -6.46810770e-01 -9.42785561e-01 -5.15900016e-01 -9.38575864e-01 6.48737669e-01 2.26602331e-01 1.31792203e-01 -7.44787753e-01 -6.00136183e-02 5.62037408e-01 4.38757479e-01 -2.95769591e-02 5.81352651e-01 -1.11042964e+00 -9.83016133e-01 -3.26908827e-01 -3.56926620e-01 8.68845940e-01 1.58142477e-01 -3.94913226e-01 -1.30790102e+00 -5.70735812e-01 1.69534639e-01 -6.45621061e-01 6.53472006e-01 -1.58920825e-01 1.58207703e+00 -1.12754250e+00 -3.07839572e-01 7.99860477e-01 1.36657822e+00 -1.07756272e-01 3.15858185e-01 3.51219103e-02 5.58415830e-01 6.75014496e-01 1.85232647e-02 7.73293197e-01 1.91199139e-01 2.03444526e-01 7.04468966e-01 8.87062177e-02 2.30673268e-01 -5.37705421e-01 5.17462254e-01 7.09207058e-01 1.55264467e-01 -1.34343535e-01 -3.53367329e-01 3.52159053e-01 -2.20296812e+00 -8.96724045e-01 4.19561654e-01 2.45435047e+00 7.99135029e-01 -2.64071614e-01 -1.05485968e-01 -4.57878113e-01 8.10132504e-01 9.66934785e-02 -9.82608616e-01 -1.18266605e-01 -1.38185203e-01 -2.09556878e-01 1.13670480e+00 -6.50617527e-03 -1.09356713e+00 7.55054712e-01 5.26045799e+00 5.46520770e-01 -1.20382237e+00 4.63483900e-01 1.03443670e+00 -3.47570032e-01 -2.67557263e-01 1.97086737e-01 -6.35716021e-01 6.72439516e-01 9.31090713e-01 -1.99904829e-01 7.07692564e-01 1.27287745e+00 -4.00668830e-02 2.44302675e-01 -1.26141274e+00 1.06311870e+00 -9.02155638e-02 -1.46437919e+00 -1.07903570e-01 4.64999735e-01 9.75685954e-01 4.11069870e-01 1.30175859e-01 1.01732239e-01 7.28904009e-01 -5.74385405e-01 4.27001655e-01 4.82059211e-01 8.63179922e-01 -7.76059151e-01 5.72596252e-01 6.63481176e-01 -7.47887194e-01 -4.33302402e-01 -3.93582284e-01 2.24335968e-01 -1.57064915e-01 3.48431230e-01 -6.27296925e-01 4.27723289e-01 8.55028689e-01 5.12924731e-01 -2.54663497e-01 7.14431882e-01 -3.22690979e-02 6.94901347e-01 -6.64090991e-01 3.94053370e-01 8.73192474e-02 -2.88671523e-01 4.00641114e-01 7.02464700e-01 1.27420992e-01 6.08584844e-02 5.86563312e-02 5.84455073e-01 -8.66291285e-01 1.06203794e-01 -9.42245483e-01 1.50923714e-01 6.35294914e-01 1.25835586e+00 -1.82406768e-01 -2.98911750e-01 -6.67562366e-01 1.35134923e+00 6.98313773e-01 6.24349654e-01 -5.43859124e-01 -3.50910313e-02 7.67161608e-01 4.89126667e-02 3.72405946e-01 1.68168858e-01 8.29261094e-02 -1.30679643e+00 3.87159288e-01 -8.07896972e-01 6.95944011e-01 -3.69190246e-01 -1.81517136e+00 3.51272881e-01 -3.84103328e-01 -1.23162949e+00 -2.29927704e-01 -3.02467614e-01 -5.49836814e-01 5.22465825e-01 -1.42471874e+00 -1.12331665e+00 -4.53606769e-02 1.13139462e+00 -1.27010047e-01 -5.39323926e-01 9.50116277e-01 1.97009236e-01 -7.15098143e-01 1.09760416e+00 8.31898391e-01 5.62542021e-01 8.92744243e-01 -8.64672840e-01 -1.61273912e-01 7.36543298e-01 3.01269740e-01 4.81586397e-01 2.96609625e-02 -4.27902520e-01 -1.52043355e+00 -1.39992535e+00 7.02647924e-01 -9.37599689e-02 5.80309033e-01 -6.24646783e-01 -7.45600581e-01 1.14742243e+00 1.09144278e-01 7.56043613e-01 9.75774407e-01 -2.46697292e-01 -7.39891946e-01 -8.19696069e-01 -1.60256064e+00 4.20144707e-01 8.47472012e-01 -7.75802076e-01 1.40765026e-01 5.84477663e-01 8.27897727e-01 4.00396883e-02 -6.72695637e-01 2.16007400e-02 4.06508744e-01 -1.03824794e+00 4.74015713e-01 -8.59171748e-01 1.43537745e-02 -8.29364806e-02 -2.12596491e-01 -7.64574051e-01 4.18098047e-02 -1.32180059e+00 -4.82247412e-01 1.45328522e+00 3.26966882e-01 -1.23800409e+00 1.32404459e+00 1.29483724e+00 4.63332534e-01 -3.30761641e-01 -1.34065342e+00 -7.78979182e-01 -3.39456528e-01 -1.50958523e-01 4.22735870e-01 1.03350019e+00 -1.02757148e-01 8.85876864e-02 -6.23632014e-01 3.91180873e-01 9.30889368e-01 9.24607813e-02 1.06706727e+00 -9.83257532e-01 -4.44527239e-01 1.31096214e-01 -4.03469861e-01 -1.02964616e+00 4.04429793e-01 -9.53449070e-01 -2.60832787e-01 -8.33191037e-01 3.58016461e-01 -4.96569723e-01 -4.81516361e-01 7.93206513e-01 3.70970100e-01 -8.55867751e-03 3.60771000e-01 7.28168011e-01 -9.92384017e-01 4.83412743e-01 6.13029599e-01 -7.66309127e-02 -2.78398134e-02 3.13539624e-01 -8.29259932e-01 5.88793516e-01 7.82880962e-01 -7.40673304e-01 -5.91710150e-01 -5.23916602e-01 -1.52975813e-01 -9.38736945e-02 4.30076897e-01 -8.28299105e-01 6.89839900e-01 1.19676054e-01 3.27568769e-01 -2.26889849e-01 1.18402354e-01 -1.62667727e+00 3.10921609e-01 2.50180006e-01 -4.76496994e-01 -3.55483651e-01 -3.94244015e-01 1.03959370e+00 -1.08738430e-01 -2.25432925e-02 7.40197420e-01 -5.50105721e-02 -3.25639814e-01 6.48928940e-01 -1.05843961e-01 -1.81198418e-01 1.34977055e+00 -1.26599316e-02 -4.32608902e-01 -8.35408628e-01 -7.31242359e-01 5.17839551e-01 6.37011230e-01 8.82257819e-02 2.56722927e-01 -1.11788559e+00 -3.55174541e-01 3.87807518e-01 -6.69618457e-05 1.97415411e-01 2.27857620e-01 6.48072958e-01 -1.64360985e-01 4.99334596e-02 3.18873152e-02 -3.78868222e-01 -1.33991551e+00 6.93039119e-01 2.01820478e-01 -3.35416824e-01 -3.76121253e-01 7.19379842e-01 4.10995364e-01 -5.05663157e-01 5.91557741e-01 2.76070476e-01 3.27435672e-01 3.37619036e-02 5.24628699e-01 3.10429066e-01 -1.87313691e-01 -4.76809293e-01 -1.02419041e-01 -2.55824029e-01 -4.28649455e-01 -2.47065462e-02 1.55964804e+00 -3.04104120e-01 -2.95624346e-01 2.22295940e-01 1.84158981e+00 -1.40754566e-01 -1.75488138e+00 -6.68782890e-01 -3.50980431e-01 -4.88491356e-01 -1.75203919e-01 -5.13704717e-01 -1.55365908e+00 4.89088684e-01 6.65619433e-01 1.48930356e-01 1.16615176e+00 5.78617975e-02 8.39338183e-01 5.81014276e-01 6.22287273e-01 -1.17962778e+00 -1.87135488e-01 1.16818175e-01 2.77205259e-01 -1.31243289e+00 -9.99992341e-02 -2.62820184e-01 -7.34569192e-01 9.19546843e-01 6.06235743e-01 8.16028118e-02 8.28141451e-01 2.02858806e-01 1.95982978e-01 1.38322145e-01 -8.55120242e-01 2.72315621e-01 -3.23575109e-01 6.06927872e-01 -4.08191115e-01 -4.90483269e-02 1.72944769e-01 7.94670403e-01 1.73182502e-01 6.66033775e-02 1.57680288e-01 1.08745825e+00 4.07772399e-02 -1.25940716e+00 -1.35972336e-01 6.00271583e-01 -5.46293616e-01 2.70464689e-01 -5.30012906e-01 3.76826346e-01 6.97718700e-03 8.13247383e-01 1.03337787e-01 -4.06310558e-01 -2.03262605e-02 7.22655058e-02 -2.25096568e-01 -4.45250660e-01 -7.36463904e-01 -1.24375105e-01 -1.34812698e-01 -7.90373087e-01 -3.06928903e-01 -5.76594651e-01 -9.04463291e-01 -3.77998441e-01 -3.05688560e-01 2.31630862e-01 6.87328100e-01 5.76251447e-01 8.73840511e-01 -2.92800635e-01 1.39550304e+00 -3.97880465e-01 -1.19684267e+00 -4.57056791e-01 -8.27408969e-01 4.81138855e-01 3.22054088e-01 3.49198580e-02 -7.04276025e-01 2.42597878e-01]
[5.868849754333496, 6.444414138793945]
3e30cc61-6f12-4b1c-87c5-b080379d3aa4
sf-net-single-frame-supervision-for-temporal
2003.06845
null
https://arxiv.org/abs/2003.06845v6
https://arxiv.org/pdf/2003.06845v6.pdf
SF-Net: Single-Frame Supervision for Temporal Action Localization
In this paper, we study an intermediate form of supervision, i.e., single-frame supervision, for temporal action localization (TAL). To obtain the single-frame supervision, the annotators are asked to identify only a single frame within the temporal window of an action. This can significantly reduce the labor cost of obtaining full supervision which requires annotating the action boundary. Compared to the weak supervision that only annotates the video-level label, the single-frame supervision introduces extra temporal action signals while maintaining low annotation overhead. To make full use of such single-frame supervision, we propose a unified system called SF-Net. First, we propose to predict an actionness score for each video frame. Along with a typical category score, the actionness score can provide comprehensive information about the occurrence of a potential action and aid the temporal boundary refinement during inference. Second, we mine pseudo action and background frames based on the single-frame annotations. We identify pseudo action frames by adaptively expanding each annotated single frame to its nearby, contextual frames and we mine pseudo background frames from all the unannotated frames across multiple videos. Together with the ground-truth labeled frames, these pseudo-labeled frames are further used for training the classifier. In extensive experiments on THUMOS14, GTEA, and BEOID, SF-Net significantly improves upon state-of-the-art weakly-supervised methods in terms of both segment localization and single-frame localization. Notably, SF-Net achieves comparable results to its fully-supervised counterpart which requires much more resource intensive annotations. The code is available at https://github.com/Flowerfan/SF-Net.
['Gourab Kundu', 'Linchao Zhu', 'Fan Ma', 'Shengxin Zha', 'Matt Feiszli', 'Zheng Shou', 'Yi Yang']
2020-03-15
null
https://www.ecva.net/papers/eccv_2020/papers_ECCV/html/2314_ECCV_2020_paper.php
https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123490409.pdf
eccv-2020-8
['weakly-supervised-action-localization']
['computer-vision']
[ 4.93964434e-01 1.44263938e-01 -7.99440920e-01 -3.93658817e-01 -9.98507857e-01 -5.44859409e-01 3.02187920e-01 -6.83713332e-02 -4.44141448e-01 6.61510110e-01 1.05106615e-01 1.04382761e-01 3.93734425e-01 -3.71828347e-01 -8.13177645e-01 -8.29555094e-01 1.29480526e-01 7.12133199e-02 7.83996940e-01 2.56996453e-01 -1.40631825e-01 5.68402298e-02 -1.43127954e+00 5.24322867e-01 5.77821314e-01 1.11251640e+00 3.77217114e-01 5.78194261e-01 1.82349443e-01 1.16552353e+00 -4.27089185e-01 -3.88011634e-02 8.98253247e-02 -6.08319938e-01 -1.03641379e+00 5.39034367e-01 5.08234441e-01 -6.52630329e-01 -3.56060296e-01 9.31280077e-01 1.09120570e-01 3.23200375e-01 9.45633724e-02 -1.22164237e+00 -1.83583334e-01 6.43885553e-01 -7.43124187e-01 4.51913953e-01 5.98089755e-01 8.23145583e-02 1.16632926e+00 -9.65756774e-01 8.12366605e-01 9.95209277e-01 5.21408737e-01 6.43620014e-01 -9.76412237e-01 -5.41931808e-01 6.11577094e-01 3.49512607e-01 -1.24050677e+00 -6.00027442e-01 6.62369788e-01 -2.48070925e-01 5.92330694e-01 1.13605121e-02 5.85458517e-01 1.13003087e+00 -3.75608116e-01 1.35663307e+00 6.68665707e-01 -3.74664873e-01 4.22157459e-02 -4.97099698e-01 -7.08527640e-02 9.26765919e-01 -3.68730038e-01 -1.41248032e-01 -7.06745446e-01 1.28324062e-01 8.41251612e-01 1.78160548e-01 -4.44372118e-01 -8.34028572e-02 -1.53439581e+00 4.08691496e-01 2.40581691e-01 2.92102844e-01 -2.67667443e-01 3.55780810e-01 4.88448411e-01 -1.10203564e-01 6.44252181e-01 -1.43335924e-01 -6.76878870e-01 -4.37852472e-01 -9.86633003e-01 -5.97547740e-02 4.26615715e-01 1.00872791e+00 9.77457047e-01 -1.01076417e-01 -4.37033445e-01 6.89295352e-01 6.93090037e-02 2.40136266e-01 9.24948379e-02 -1.57805395e+00 5.05267322e-01 5.80458701e-01 3.16841930e-01 -5.26139498e-01 -1.16711378e-01 -8.62137303e-02 -5.15640080e-01 -1.59019917e-01 7.25435078e-01 -1.58667490e-01 -1.03799856e+00 1.65811265e+00 6.47522807e-01 8.51591408e-01 -1.63106814e-01 1.03701437e+00 6.23896658e-01 4.93410617e-01 1.08765207e-01 -3.62936437e-01 1.18039143e+00 -1.53876472e+00 -7.72019327e-01 -1.84154019e-01 8.21900070e-01 -6.37400389e-01 1.03981972e+00 2.05889761e-01 -9.53393042e-01 -7.46316075e-01 -7.10839391e-01 -5.41739389e-02 9.47483927e-02 6.05302691e-01 4.48413402e-01 5.78312017e-02 -7.65965462e-01 5.23880422e-01 -1.30628121e+00 -1.74685448e-01 5.49535215e-01 1.04571603e-01 -5.03749788e-01 -2.35585824e-01 -1.07018042e+00 4.12130296e-01 4.24294591e-01 1.03971526e-01 -1.26515031e+00 -5.79869628e-01 -1.20631683e+00 -1.42476186e-01 9.86956418e-01 -1.97358564e-01 1.59362590e+00 -1.31375146e+00 -1.42088532e+00 7.19853461e-01 -7.26909876e-01 -4.66722071e-01 4.82660145e-01 -3.76965970e-01 -1.73185840e-01 5.48790336e-01 4.51629996e-01 9.25267279e-01 8.26606631e-01 -1.04087889e+00 -1.08090758e+00 -1.89315360e-02 4.04455334e-01 1.80952460e-01 -5.03279753e-02 1.64045975e-01 -1.04391611e+00 -7.93329060e-01 9.41705853e-02 -1.02113688e+00 -6.41106442e-02 2.30509251e-01 -2.99716830e-01 -3.95131409e-01 1.11739874e+00 -5.54096103e-01 1.34689784e+00 -2.35468578e+00 9.18162391e-02 -2.42302746e-01 1.01921596e-01 1.72800064e-01 9.24421102e-03 -3.86915393e-02 -4.46240082e-02 -1.32894084e-01 -3.07248950e-01 -6.31962478e-01 -3.03128868e-01 5.10787785e-01 -3.21410112e-02 4.91868943e-01 2.64382720e-01 7.93700635e-01 -1.34654331e+00 -8.08565199e-01 3.18074495e-01 3.41891557e-01 -4.51851696e-01 1.59389883e-01 -4.55168426e-01 8.11733663e-01 -5.18518984e-01 1.01939428e+00 2.18657151e-01 -4.87722278e-01 1.51263624e-01 -2.37068132e-01 -3.66941988e-02 1.62760422e-01 -1.12981856e+00 1.99493372e+00 -2.14043319e-01 6.64810002e-01 5.24890833e-02 -1.01706707e+00 4.76980507e-01 5.03606856e-01 7.83035100e-01 -3.29397798e-01 -9.19728726e-03 6.36305958e-02 -3.27482820e-01 -5.30039549e-01 1.41477510e-01 1.69747919e-01 -3.31707555e-03 4.37312573e-01 2.78928638e-01 5.61118245e-01 5.25976777e-01 1.93887621e-01 1.27759027e+00 7.81552017e-01 3.11536044e-01 1.59091026e-01 7.18617618e-01 -1.24508984e-01 1.07960963e+00 4.29271549e-01 -6.73023999e-01 7.50176847e-01 4.73602831e-01 -4.44027275e-01 -7.48812020e-01 -8.04604232e-01 -9.90669522e-03 1.48037112e+00 4.42210197e-01 -7.32470453e-01 -9.74753082e-01 -1.13959420e+00 -2.10434839e-01 2.57580042e-01 -7.19924927e-01 1.73523769e-01 -8.21668923e-01 -2.35462144e-01 5.60780704e-01 9.16391909e-01 6.86505556e-01 -1.16428137e+00 -6.30327344e-01 1.35437191e-01 -6.97478771e-01 -1.47420597e+00 -1.00454772e+00 1.16448998e-01 -7.37530470e-01 -1.28264248e+00 -6.93254709e-01 -7.33516097e-01 7.01076031e-01 3.53749812e-01 9.63924646e-01 2.44018286e-01 6.34302944e-02 1.97717398e-01 -6.97257042e-01 6.85470924e-02 -1.20657682e-01 -1.31162763e-01 -1.16766304e-01 3.33103746e-01 3.87887567e-01 -2.73521900e-01 -6.42468154e-01 7.42765963e-01 -7.22201169e-01 2.35727981e-01 3.58942419e-01 8.06609929e-01 1.13749588e+00 -6.30009174e-02 4.51074302e-01 -8.08038712e-01 -3.13776940e-01 -3.45702589e-01 -4.05530363e-01 2.74966121e-01 -4.97187451e-02 -2.76019156e-01 3.97447556e-01 -5.56218684e-01 -1.18513131e+00 5.84706843e-01 1.16818845e-01 -9.40942407e-01 -3.52857441e-01 2.93906838e-01 -1.24907620e-01 1.66282609e-01 3.84523451e-01 1.18751593e-01 -2.92831481e-01 -4.96833891e-01 3.32291156e-01 4.46544319e-01 8.66802812e-01 -6.28073394e-01 4.59884554e-01 6.99197054e-01 -3.45503539e-01 -5.00311613e-01 -1.50781357e+00 -8.09613943e-01 -9.32785869e-01 -4.82614756e-01 1.19146037e+00 -1.07376099e+00 -4.72895086e-01 4.75293607e-01 -1.12417328e+00 -6.81839287e-01 -3.27185720e-01 4.44133580e-01 -6.66480899e-01 5.38957775e-01 -7.26041973e-01 -6.87890708e-01 -1.57774743e-02 -1.05205154e+00 1.55771077e+00 2.45101303e-01 -3.06684285e-01 -8.57731879e-01 -2.43219852e-01 6.74728096e-01 -3.69963199e-01 3.16284329e-01 2.14970425e-01 -3.18824440e-01 -6.48346782e-01 -7.30067343e-02 -5.88467382e-02 3.84990185e-01 3.47095817e-01 2.38087121e-02 -8.80604267e-01 -1.16436407e-01 -2.62647539e-01 -4.92205888e-01 9.19897735e-01 4.37773943e-01 1.33824539e+00 -5.05650863e-02 -3.96005601e-01 5.53165853e-01 9.30006921e-01 1.90117598e-01 3.77265662e-01 1.89433172e-01 8.75473738e-01 2.57760674e-01 1.34119070e+00 5.40429950e-01 2.67388552e-01 7.55787313e-01 2.55745828e-01 -1.83311641e-01 -1.84439942e-01 -3.37163687e-01 5.67956030e-01 3.24587107e-01 -4.25276726e-01 -1.99273035e-01 -7.35249877e-01 5.85345209e-01 -2.26046419e+00 -1.17303562e+00 9.49597955e-02 1.91988683e+00 1.01865375e+00 2.24181324e-01 3.78975183e-01 2.63736874e-01 1.02522385e+00 3.50176364e-01 -7.25207806e-01 3.82533967e-01 -6.33820891e-02 -5.37696071e-02 4.64412451e-01 4.96043712e-01 -1.51670206e+00 1.09495163e+00 5.71774292e+00 8.85531962e-01 -7.70373464e-01 3.57880890e-01 7.38333523e-01 -2.60803282e-01 1.66211456e-01 6.83840215e-02 -9.66846049e-01 5.71384430e-01 6.60381615e-01 2.30464131e-01 3.45307380e-01 8.81698310e-01 5.54826021e-01 -3.98165971e-01 -1.29840410e+00 7.61990368e-01 3.50544527e-02 -1.29491127e+00 -2.47204050e-01 -3.10306728e-01 8.43959689e-01 1.20751932e-01 -4.25093919e-01 1.24552786e-01 7.52214566e-02 -6.47315264e-01 9.44553256e-01 2.82095194e-01 1.04247928e+00 -5.18784106e-01 7.49041498e-01 4.28786695e-01 -1.68920755e+00 -9.82854590e-02 -4.51052189e-03 -1.11998662e-01 5.15845001e-01 4.77711618e-01 -2.94957966e-01 5.24444878e-01 9.20803487e-01 1.33575833e+00 -2.80374646e-01 6.98571026e-01 -5.91713667e-01 8.04839909e-01 -3.90351593e-01 5.58512092e-01 3.59244764e-01 4.80009541e-02 2.88858414e-01 1.11380696e+00 6.96466193e-02 3.45195651e-01 8.33647847e-01 5.40997207e-01 -9.06941891e-02 -3.32406312e-01 -1.25172287e-01 5.79579696e-02 5.40258110e-01 1.24673724e+00 -9.22882974e-01 -6.33894145e-01 -7.66345739e-01 1.20755672e+00 2.17719689e-01 4.96388435e-01 -1.26493466e+00 -5.12053929e-02 5.10230780e-01 1.40410572e-01 5.13736069e-01 -3.95777635e-04 2.08982266e-02 -1.28345406e+00 2.83691436e-01 -7.67602205e-01 6.93981409e-01 -8.40032399e-01 -8.65023971e-01 4.70597476e-01 1.41362578e-01 -1.49810457e+00 -1.28725886e-01 -1.98323950e-01 -5.70299506e-01 4.26663876e-01 -1.40971029e+00 -1.26256299e+00 -5.31023562e-01 7.96066821e-01 1.00954056e+00 2.32183009e-01 4.94520068e-01 4.34741020e-01 -8.18119884e-01 4.56678867e-01 -5.10183334e-01 4.78533924e-01 8.58916163e-01 -1.21958101e+00 1.18534066e-01 1.02909219e+00 2.95304209e-01 2.05029428e-01 2.90132195e-01 -7.50821590e-01 -7.80838549e-01 -1.27477062e+00 7.62688160e-01 -4.85035628e-01 7.35488594e-01 -1.11672752e-01 -8.60321939e-01 9.16474462e-01 -1.31553918e-01 6.48718715e-01 4.72041130e-01 -1.05762817e-01 -9.23935995e-02 -5.46534583e-02 -7.26218104e-01 3.20710629e-01 1.33993614e+00 -6.71405315e-01 -4.77413177e-01 3.49325955e-01 9.08872962e-01 -6.68348372e-01 -8.19839597e-01 4.68709171e-01 3.46587092e-01 -9.05863523e-01 8.23294759e-01 -2.95938939e-01 5.43250203e-01 -7.70160258e-01 -4.55223769e-02 -7.09942460e-01 -1.82968438e-01 -7.71635056e-01 -5.14627695e-01 1.30521774e+00 2.74074316e-01 -5.09355254e-02 1.02302372e+00 4.95856792e-01 -2.94051170e-01 -1.01411021e+00 -9.15224671e-01 -6.86867595e-01 -5.85098386e-01 -4.34657007e-01 1.88121393e-01 8.78677249e-01 3.29464115e-02 1.24103278e-01 -4.77212578e-01 1.93432838e-01 5.48160672e-01 1.25326067e-01 6.06750846e-01 -7.92882025e-01 -3.89957696e-01 7.17523322e-02 -2.77909309e-01 -1.48312247e+00 4.52421933e-01 -5.69914341e-01 4.03140575e-01 -1.20385003e+00 3.17511976e-01 -2.23054811e-01 -4.85032022e-01 1.01774478e+00 -4.38082248e-01 5.44100583e-01 9.57325995e-02 3.42253655e-01 -1.23768294e+00 3.33442211e-01 1.37700105e+00 2.34924033e-02 -1.61968291e-01 1.18287057e-02 -1.68348655e-01 1.08481431e+00 6.49402678e-01 -4.88347262e-01 -4.14566100e-01 -3.81602466e-01 -3.42500657e-01 2.92711407e-01 4.15015489e-01 -9.19637620e-01 2.12714121e-01 -3.00107270e-01 3.26461077e-01 -6.54698610e-01 2.56046057e-01 -7.02870369e-01 -5.81858233e-02 2.42475122e-01 -3.60871226e-01 -2.07982525e-01 -6.90216124e-02 8.40488195e-01 -5.40135264e-01 -2.00693890e-01 8.21382403e-01 -1.25716671e-01 -1.11746275e+00 5.09266138e-01 -2.58924067e-01 6.47090003e-02 1.34201443e+00 -3.30963761e-01 -1.70211345e-01 -3.06772351e-01 -1.09886372e+00 4.15113747e-01 5.52176893e-01 2.98535556e-01 4.28921282e-01 -1.30581760e+00 -4.07819152e-01 -4.34690155e-02 1.26304910e-01 3.82090926e-01 2.86638081e-01 1.14791787e+00 -2.69549549e-01 1.76150039e-01 1.42561361e-01 -8.96308243e-01 -1.51039028e+00 5.19929826e-01 2.71657348e-01 -1.27350166e-01 -7.83081412e-01 8.63444686e-01 3.81092846e-01 1.63046628e-01 4.59969372e-01 -5.32102406e-01 -9.91537794e-02 -1.45789133e-02 6.53156757e-01 3.54335546e-01 -3.44710886e-01 -8.39845538e-01 -4.74017590e-01 5.00893712e-01 1.36531174e-01 4.10590917e-02 1.01366341e+00 -2.08578587e-01 -6.75220368e-03 4.91207749e-01 1.06584561e+00 -1.43245474e-01 -2.04357791e+00 -4.32753444e-01 5.33123575e-02 -5.71596324e-01 -6.94990829e-02 -5.24221718e-01 -1.33661962e+00 5.93805969e-01 1.95305675e-01 -2.21367329e-01 1.41176081e+00 3.36734354e-01 8.79600525e-01 2.86173224e-01 3.61308187e-01 -1.14092290e+00 3.26528043e-01 4.29567367e-01 4.15382802e-01 -1.31632066e+00 -8.21282640e-02 -7.42165685e-01 -7.19833374e-01 9.13165808e-01 8.71177375e-01 5.00411093e-02 4.02633756e-01 4.05521721e-01 7.19471276e-02 8.32333118e-02 -6.92553282e-01 -3.97829473e-01 2.25382254e-01 3.75518978e-01 3.65913361e-01 -1.75957471e-01 7.82749522e-03 4.76987273e-01 4.35051203e-01 3.93279493e-01 2.68641651e-01 1.03427339e+00 -4.27971959e-01 -1.03663957e+00 -1.99384302e-01 1.83265746e-01 -6.73343956e-01 8.39886293e-02 -2.40286499e-01 5.89121163e-01 4.92963105e-01 1.02395272e+00 8.83750245e-02 -2.40697473e-01 9.23231393e-02 5.94416112e-02 3.01875770e-01 -6.94001257e-01 -2.43408024e-01 4.36065406e-01 2.84984499e-01 -9.95697677e-01 -1.10915101e+00 -8.85241628e-01 -1.62803888e+00 -7.04838336e-02 -4.78165567e-01 3.35281193e-02 -1.70427442e-01 1.11080992e+00 2.00417057e-01 4.81658667e-01 4.69241947e-01 -1.10153604e+00 -6.04254939e-03 -8.82289469e-01 -3.71925026e-01 6.23699784e-01 4.50174183e-01 -8.17511261e-01 -2.73182690e-01 7.77095497e-01]
[8.544309616088867, 0.5902863144874573]
69dc378e-fd8a-49a6-b817-e72859538a57
heterogeneous-molecular-graph-neural-networks
2009.1271
null
https://arxiv.org/abs/2009.12710v1
https://arxiv.org/pdf/2009.12710v1.pdf
Heterogeneous Molecular Graph Neural Networks for Predicting Molecule Properties
As they carry great potential for modeling complex interactions, graph neural network (GNN)-based methods have been widely used to predict quantum mechanical properties of molecules. Most of the existing methods treat molecules as molecular graphs in which atoms are modeled as nodes. They characterize each atom's chemical environment by modeling its pairwise interactions with other atoms in the molecule. Although these methods achieve a great success, limited amount of works explicitly take many-body interactions, i.e., interactions between three and more atoms, into consideration. In this paper, we introduce a novel graph representation of molecules, heterogeneous molecular graph (HMG) in which nodes and edges are of various types, to model many-body interactions. HMGs have the potential to carry complex geometric information. To leverage the rich information stored in HMGs for chemical prediction problems, we build heterogeneous molecular graph neural networks (HMGNN) on the basis of a neural message passing scheme. HMGNN incorporates global molecule representations and an attention mechanism into the prediction process. The predictions of HMGNN are invariant to translation and rotation of atom coordinates, and permutation of atom indices. Our model achieves state-of-the-art performance in 9 out of 12 tasks on the QM9 dataset.
['Zeren Shui', 'George Karypis']
2020-09-26
null
null
null
null
['formation-energy']
['miscellaneous']
[ 2.19002321e-01 1.78428426e-01 -6.24800026e-01 -2.88566202e-01 -6.68352144e-03 -2.88854271e-01 5.32535672e-01 6.69555247e-01 4.23960574e-02 8.27939928e-01 1.92517698e-01 -6.60684705e-01 4.48601842e-02 -1.23695612e+00 -1.08441460e+00 -8.56628835e-01 -4.05354291e-01 4.60675597e-01 9.77133363e-02 -4.14170653e-01 3.48971426e-01 7.41089046e-01 -8.15392852e-01 3.73750329e-01 8.19424570e-01 5.23858070e-01 7.30435327e-02 4.80412543e-01 -4.51895595e-02 1.25065410e+00 -2.17317894e-01 -5.03505766e-01 -1.46931976e-01 -5.09224415e-01 -9.40308928e-01 -5.43114305e-01 2.14510724e-01 2.00450629e-01 -7.38902450e-01 1.00250685e+00 6.82745278e-01 4.64959145e-01 8.38162422e-01 -9.19446826e-01 -9.44196463e-01 7.13730037e-01 -3.67281020e-01 -1.32252127e-01 5.05627453e-01 2.17133284e-01 1.29949546e+00 -8.15589786e-01 9.32951093e-01 1.42667210e+00 5.15615165e-01 3.71593326e-01 -1.35698330e+00 -6.42931461e-01 1.64273858e-01 5.76367974e-01 -1.31713736e+00 -1.22509331e-01 8.06300998e-01 -4.54738736e-01 1.72852159e+00 3.93993640e-03 6.76988482e-01 7.04855621e-01 1.03659856e+00 2.51241773e-01 4.22067106e-01 -1.03432007e-01 1.97927222e-01 -6.43886209e-01 3.34878355e-01 8.66111338e-01 2.81898707e-01 -7.39595518e-02 -7.23059475e-01 -4.81227070e-01 4.09570456e-01 3.05916995e-01 -2.14662686e-01 -5.11766374e-01 -1.06154478e+00 1.09767473e+00 1.20239806e+00 -1.04023345e-01 -5.16823411e-01 4.59978640e-01 3.51835519e-01 6.42780438e-02 3.96738470e-01 5.58451295e-01 -1.36442408e-01 3.82730514e-01 -1.48241311e-01 3.42829674e-01 9.48955715e-01 7.96521544e-01 8.78414512e-01 -2.65660793e-01 -2.01671887e-02 4.50150281e-01 5.29314101e-01 4.85183261e-02 1.58759028e-01 -2.84143478e-01 5.37668407e-01 8.68512213e-01 -2.28442371e-01 -1.34111130e+00 -7.33130097e-01 -2.59310216e-01 -1.28537369e+00 -1.99120522e-01 -2.13143021e-01 2.53218800e-01 -9.32520568e-01 1.57109952e+00 4.45655972e-01 2.08040001e-03 6.33256361e-02 4.86091822e-01 1.32710767e+00 9.98792052e-01 4.67229337e-01 -1.59513459e-01 8.83632779e-01 -1.21941614e+00 -6.95223570e-01 2.30506375e-01 1.13395214e+00 -3.95957053e-01 3.91051441e-01 1.88342631e-01 -1.09435046e+00 -5.78188419e-01 -1.09091806e+00 -3.39418739e-01 -7.70095587e-01 -5.47676623e-01 1.21424484e+00 2.91661620e-01 -9.04204309e-01 1.33482456e+00 -8.10690761e-01 1.61225080e-01 3.52060229e-01 9.27970350e-01 -5.79763114e-01 -1.88651577e-01 -1.52558434e+00 8.65800381e-01 5.74547887e-01 2.15937436e-01 -6.84477270e-01 -6.17465734e-01 -1.09828782e+00 1.06041618e-01 1.80888250e-01 -9.97116029e-01 7.83092916e-01 -5.32158315e-01 -1.35354578e+00 3.13803285e-01 -4.64392960e-01 -4.05674011e-01 -1.40710771e-01 3.26008320e-01 -4.49964672e-01 -2.82576066e-02 -2.77419090e-01 5.66877842e-01 2.17347190e-01 -8.24161410e-01 2.37066701e-01 -4.78804410e-01 1.09342538e-01 2.49586716e-01 9.45135131e-02 -1.06515549e-01 -2.36130431e-01 -3.20563942e-01 1.54458985e-01 -1.01682675e+00 -6.89640284e-01 -4.56746429e-01 -9.99928415e-01 -5.30908048e-01 2.90468544e-01 -3.02528530e-01 1.35980260e+00 -1.48118699e+00 6.75714791e-01 5.10418475e-01 7.39221156e-01 2.80771554e-01 -3.45659524e-01 1.14732730e+00 -4.51320738e-01 3.36734176e-01 -9.81164849e-05 -1.46773398e-01 1.04642719e-01 -2.37246919e-02 -2.73599084e-02 4.99797881e-01 4.46233526e-03 1.29772246e+00 -8.80638599e-01 -1.49024874e-01 1.54305205e-01 8.84757817e-01 -8.53149056e-01 1.37685969e-01 -6.65015101e-01 5.01588106e-01 -6.18309975e-01 5.09113371e-01 7.11590707e-01 -8.29077840e-01 6.51988924e-01 -4.71885324e-01 6.06056862e-02 6.87885463e-01 -6.38407350e-01 1.50920069e+00 2.47798726e-01 -2.34921109e-02 -4.84497130e-01 -8.96698058e-01 7.28199065e-01 2.28626698e-01 5.27174890e-01 -5.40196776e-01 -2.89512705e-02 -4.11934145e-02 6.61035955e-01 -1.32126600e-01 3.92943233e-01 -7.03005120e-02 4.70397919e-01 2.20518932e-01 -1.20101925e-02 -5.32325320e-02 2.41484523e-01 4.71852094e-01 1.07047164e+00 8.27215239e-02 7.10519373e-01 -7.84904230e-03 6.38687193e-01 -2.50236273e-01 2.88063139e-01 6.29727304e-01 2.65324295e-01 1.72639191e-01 5.50706089e-01 -9.71851826e-01 -8.43602836e-01 -5.55485129e-01 1.94455683e-02 1.24203825e+00 1.98470503e-01 -1.18943226e+00 -5.62628865e-01 -3.84614170e-01 1.57162949e-01 1.34294122e-01 -6.67494714e-01 -5.62781215e-01 -4.49392229e-01 -1.04554331e+00 3.72230820e-02 1.37048528e-01 1.24124125e-01 -1.31539214e+00 4.04879332e-01 6.59681380e-01 2.21607715e-01 -6.91102028e-01 -4.15597320e-01 3.41888875e-01 -7.82930911e-01 -1.26234090e+00 -1.81238189e-01 -5.41531622e-01 6.04541183e-01 4.68906343e-01 1.24127948e+00 3.59648794e-01 -2.73927003e-01 -3.10645789e-01 -1.20243534e-01 -2.53372788e-01 -3.72226179e-01 2.61654258e-01 3.24212350e-02 -1.72013447e-01 3.78774792e-01 -7.61286557e-01 -7.95543849e-01 5.65619580e-02 -7.54378617e-01 3.43400717e-01 4.17309046e-01 7.94791758e-01 8.43903482e-01 -2.06055239e-01 3.50728393e-01 -1.42760897e+00 5.37883937e-01 -7.08648622e-01 -2.45253041e-01 1.97630957e-01 -5.26695371e-01 2.48151183e-01 1.03491688e+00 8.78189877e-03 -4.03101563e-01 9.11071002e-02 -2.80326694e-01 -8.55996609e-02 1.10950358e-01 1.08969581e+00 -2.83835024e-01 -6.83965504e-01 4.52371150e-01 1.46927401e-01 -3.63339365e-01 -2.19700858e-01 4.70042378e-01 1.70063004e-01 -7.81517103e-02 -6.75459087e-01 3.47972989e-01 -9.22419690e-03 8.75066221e-01 -7.68929243e-01 -6.04270220e-01 -2.48685583e-01 -8.84400487e-01 2.41458014e-01 9.43000376e-01 -8.57988358e-01 -1.35484791e+00 3.27020735e-01 -1.45940375e+00 -2.09314257e-01 4.76732999e-01 3.94712567e-01 -3.23633671e-01 7.64045894e-01 -7.91865170e-01 -4.63474035e-01 -5.18842399e-01 -1.44220960e+00 8.40652764e-01 -5.77336065e-02 -1.63486600e-01 -1.16506994e+00 2.65958667e-01 3.07163745e-01 1.92834064e-01 5.03477275e-01 1.49475241e+00 -7.37998664e-01 -9.38880682e-01 -1.39526501e-01 -1.27683535e-01 -2.50996947e-01 2.19952509e-01 -5.91230616e-02 -5.96461892e-01 -4.50611681e-01 -7.63580322e-01 -3.01983356e-01 1.08024752e+00 4.82146442e-01 1.47310257e+00 -3.58866066e-01 -7.68101275e-01 8.53304327e-01 1.09444463e+00 4.60761935e-01 6.77278221e-01 -2.76983082e-02 1.56602573e+00 3.83736372e-01 -1.04477577e-01 2.79016584e-01 5.26102066e-01 7.56968200e-01 7.79082298e-01 -1.05384909e-01 1.61745980e-01 -4.85606223e-01 3.18609416e-01 1.07948649e+00 -6.94603086e-01 -7.35740602e-01 -8.72161388e-01 -3.87895912e-01 -1.92798400e+00 -1.01308239e+00 -4.43347275e-01 1.98605406e+00 8.38427067e-01 5.56673519e-02 -5.73009020e-03 -4.22078520e-01 6.51408911e-01 5.74562788e-01 -1.00050545e+00 -5.13323128e-01 -3.92737463e-02 3.68920863e-01 5.14906228e-01 6.91181242e-01 -1.05175114e+00 1.09087002e+00 6.21965408e+00 6.93272531e-01 -8.80243838e-01 -3.22454870e-01 7.84608424e-01 2.31165394e-01 -4.05186951e-01 1.09896332e-01 -8.03216398e-01 4.65941370e-01 1.14476430e+00 8.69190991e-02 6.00301147e-01 5.44599831e-01 1.93172231e-01 4.11011726e-01 -1.22808433e+00 8.86824012e-01 -7.16081038e-02 -1.97531581e+00 7.56331682e-01 2.52817750e-01 7.88868248e-01 2.96582490e-01 -5.98764569e-02 9.03493688e-02 3.95757914e-01 -1.76906872e+00 5.67710288e-02 5.93506873e-01 5.30127287e-01 -8.93221140e-01 5.14495909e-01 1.85706839e-01 -1.47909498e+00 4.17273939e-01 -8.01144838e-01 -3.00761133e-01 2.84263995e-02 3.45612437e-01 -7.08754420e-01 8.71490240e-01 -9.21811152e-04 1.25831354e+00 -3.89636815e-01 6.62780046e-01 -1.98609814e-01 4.28571761e-01 8.89811590e-02 -3.57219547e-01 3.24851811e-01 -8.21227312e-01 7.95536954e-03 1.03051770e+00 -5.50281592e-02 4.17657018e-01 4.09114867e-01 8.62925708e-01 -7.25966752e-01 4.23976094e-01 -8.83414447e-01 -4.42700028e-01 3.34643424e-01 1.04038465e+00 -3.79992217e-01 -2.06748143e-01 -6.40752852e-01 7.27240026e-01 7.55261540e-01 4.23017234e-01 -6.27751410e-01 -4.31580245e-01 7.77857482e-01 7.33230338e-02 2.93556391e-03 -3.73074591e-01 3.30987841e-01 -9.86429334e-01 -4.28829610e-01 -8.59737873e-01 2.29590788e-01 -7.32531607e-01 -1.26953101e+00 5.30242026e-01 -5.79659939e-01 -6.45457268e-01 1.27469152e-01 -1.07628226e+00 -7.23420143e-01 1.11475003e+00 -1.59163058e+00 -1.24208021e+00 9.85737219e-02 4.82735008e-01 1.38892671e-02 -8.01749751e-02 1.11738932e+00 1.44362729e-02 -8.18464935e-01 4.03253287e-01 2.59371430e-01 -7.64143020e-02 5.07345855e-01 -1.22460759e+00 9.80763376e-01 9.33394507e-02 4.18692417e-02 1.32554221e+00 4.17109013e-01 -1.00542748e+00 -2.04490161e+00 -1.31130624e+00 1.01113665e+00 -3.32485497e-01 6.35795057e-01 -4.82820630e-01 -9.94927406e-01 7.18129933e-01 1.00068003e-01 1.65700138e-01 1.06034505e+00 2.49849916e-01 -5.12160599e-01 4.42469031e-01 -6.03384912e-01 6.36958420e-01 1.39656937e+00 -8.81289840e-01 2.66903639e-02 1.02798927e+00 1.06328666e+00 -4.90520656e-01 -1.22917485e+00 4.09141809e-01 3.06946546e-01 -8.86607826e-01 1.37600815e+00 -1.36281216e+00 6.40910506e-01 -2.19454885e-01 -5.92440926e-03 -1.28874266e+00 -8.05946648e-01 -7.95123875e-01 -4.44370955e-01 2.89742947e-01 6.82275832e-01 -7.91372836e-01 8.74976158e-01 5.94229043e-01 -4.30395842e-01 -1.12207043e+00 -5.92798054e-01 -2.89142936e-01 1.52290717e-01 3.08458004e-02 8.32809031e-01 1.08297300e+00 3.82873535e-01 8.80363941e-01 -6.46707237e-01 4.83221784e-02 2.93594122e-01 3.19059312e-01 7.43009031e-01 -1.32597935e+00 -4.58846867e-01 -3.61703187e-01 -6.68199837e-01 -1.14435494e+00 4.80639368e-01 -1.55935907e+00 -5.42785287e-01 -1.63869309e+00 6.03774726e-01 2.45938506e-02 -5.05312264e-01 5.11661649e-01 -2.02559993e-01 5.14024980e-02 8.58135447e-02 2.80161768e-01 -9.63366687e-01 8.15134585e-01 1.65494204e+00 -4.78532612e-01 -1.57437593e-01 -2.83066958e-01 -5.33417344e-01 4.95384246e-01 6.61940455e-01 -2.92663217e-01 -2.92631209e-01 -1.42350659e-01 7.70817041e-01 1.63426384e-01 -5.20616025e-03 -8.08205068e-01 5.24168074e-01 -2.17898294e-01 5.45517623e-01 -5.71938336e-01 5.90416789e-01 -4.04741615e-01 4.64558452e-01 6.47635281e-01 -4.35999870e-01 2.08661988e-01 -1.60471439e-01 9.69168663e-01 -2.91482478e-01 1.92064017e-01 6.07437193e-01 -2.56000131e-01 -3.07278305e-01 1.18739069e+00 -1.16375513e-01 -5.22585392e-01 7.36907363e-01 4.07226058e-03 -4.77592498e-01 -2.35526487e-01 -8.23736012e-01 9.20089856e-02 4.85174030e-01 8.14974457e-02 6.42184138e-01 -1.36216903e+00 -2.56517828e-01 2.05631610e-02 1.33477241e-01 2.68674940e-02 2.12304756e-01 6.52893662e-01 -7.57772863e-01 7.71751344e-01 1.22542180e-01 -1.58476695e-01 -1.20882320e+00 8.33803773e-01 4.94988739e-01 -3.55731398e-01 -3.68239373e-01 8.01027477e-01 5.55419564e-01 -6.22142732e-01 4.81419824e-03 -3.00409615e-01 -2.58158624e-01 -2.40411103e-01 2.67330945e-01 -5.11992052e-02 6.77164197e-02 -8.97230029e-01 -3.08838487e-01 6.38310373e-01 -4.30855185e-01 8.51520717e-01 1.41198504e+00 4.97167200e-01 -6.58249795e-01 1.36919215e-01 1.36747944e+00 -1.91239700e-01 -7.52358675e-01 -2.94503301e-01 -1.18768990e-01 2.44005173e-02 -1.49897143e-01 -6.24518275e-01 -7.31018543e-01 1.07061350e+00 2.04476458e-03 6.51789783e-03 3.72449607e-01 -1.07693687e-01 8.23014855e-01 1.04068744e+00 4.18376356e-01 -5.11514008e-01 1.61939356e-02 8.62477243e-01 6.56936109e-01 -1.26890147e+00 2.78237611e-01 -6.82262123e-01 -3.03004682e-01 1.35852206e+00 4.30836409e-01 -1.76141839e-02 6.34938359e-01 -4.68269378e-01 -6.12334967e-01 -7.59630620e-01 -8.48381281e-01 1.73225433e-01 5.71853697e-01 3.72954875e-01 1.08224893e+00 2.83493280e-01 -1.59165815e-01 5.62587619e-01 -3.64744142e-02 -3.72516781e-01 1.80425659e-01 7.39947855e-01 -6.10053062e-01 -1.55898583e+00 1.84170246e-01 4.45521116e-01 -4.04847741e-01 -6.61076128e-01 -8.98315907e-01 5.52638471e-01 -3.97021370e-03 7.90215373e-01 -4.59841400e-01 -4.83801395e-01 1.21833645e-01 -1.04354136e-02 5.22983074e-01 -8.44038963e-01 -6.07045650e-01 -3.45824361e-01 9.93157178e-03 -7.37760723e-01 -4.40508932e-01 -5.13178147e-02 -1.51041877e+00 -7.47229815e-01 -2.50348419e-01 5.35067439e-01 3.40234667e-01 8.11565816e-01 7.90644944e-01 6.63935125e-01 5.80493927e-01 -1.11921620e+00 -2.53609866e-01 -8.56269181e-01 -7.28340030e-01 5.09933770e-01 2.15777323e-01 -5.15044510e-01 3.47787179e-02 -3.17151517e-01]
[5.153307914733887, 5.826826095581055]
4ed320c3-c78f-47f6-9a31-3f8eefda8c7c
learning-a-structured-latent-space-for
2203.1558
null
https://arxiv.org/abs/2203.15580v1
https://arxiv.org/pdf/2203.15580v1.pdf
Learning a Structured Latent Space for Unsupervised Point Cloud Completion
Unsupervised point cloud completion aims at estimating the corresponding complete point cloud of a partial point cloud in an unpaired manner. It is a crucial but challenging problem since there is no paired partial-complete supervision that can be exploited directly. In this work, we propose a novel framework, which learns a unified and structured latent space that encoding both partial and complete point clouds. Specifically, we map a series of related partial point clouds into multiple complete shape and occlusion code pairs and fuse the codes to obtain their representations in the unified latent space. To enforce the learning of such a structured latent space, the proposed method adopts a series of constraints including structured ranking regularization, latent code swapping constraint, and distribution supervision on the related partial point clouds. By establishing such a unified and structured latent space, better partial-complete geometry consistency and shape completion accuracy can be achieved. Extensive experiments show that our proposed method consistently outperforms state-of-the-art unsupervised methods on both synthetic ShapeNet and real-world KITTI, ScanNet, and Matterport3D datasets.
['Hongsheng Li', 'Xiaogang Wang', 'Qiang Wang', 'Chao Zhang', 'Kwan-Yee Lin', 'Yingjie Cai']
2022-03-29
null
http://openaccess.thecvf.com//content/CVPR2022/html/Cai_Learning_a_Structured_Latent_Space_for_Unsupervised_Point_Cloud_Completion_CVPR_2022_paper.html
http://openaccess.thecvf.com//content/CVPR2022/papers/Cai_Learning_a_Structured_Latent_Space_for_Unsupervised_Point_Cloud_Completion_CVPR_2022_paper.pdf
cvpr-2022-1
['point-cloud-completion']
['computer-vision']
[-7.17585683e-02 -9.43804309e-02 -3.06421131e-01 -6.04745567e-01 -1.03992438e+00 -5.80570579e-01 5.92898130e-01 -1.17123961e-01 1.92067966e-01 3.66375774e-01 6.05789758e-02 1.36476249e-01 -2.97813982e-01 -6.09749138e-01 -9.29244220e-01 -6.80886626e-01 3.51613581e-01 9.42010641e-01 1.31540954e-01 2.49215186e-01 2.11564884e-01 6.63926780e-01 -1.44823468e+00 9.08908695e-02 1.10605478e+00 6.49719298e-01 5.97797155e-01 -3.99286896e-02 -4.00655150e-01 2.71546096e-01 1.09833054e-01 -1.21148989e-01 4.40250456e-01 2.35164002e-01 -4.60998058e-01 5.44705749e-01 6.54010653e-01 -1.07751772e-01 -1.43478572e-01 1.10783970e+00 5.05908243e-02 -8.04208741e-02 7.75188029e-01 -1.25952196e+00 -6.64371014e-01 6.06415533e-02 -8.88718426e-01 -6.87503219e-01 2.27942780e-01 -1.19633459e-01 1.09271765e+00 -1.21077311e+00 7.44095743e-01 1.29000151e+00 5.10157228e-01 5.12055755e-02 -1.39937282e+00 -8.70988846e-01 1.44792333e-01 -3.49133730e-01 -1.66119385e+00 -2.98386902e-01 1.09043157e+00 -7.60125518e-01 5.89208603e-01 7.20062554e-02 4.76332515e-01 6.13931358e-01 -1.87630892e-01 6.06301725e-01 8.12865794e-01 -1.50086969e-01 2.69550353e-01 5.18560447e-02 -4.90514114e-02 6.60933554e-01 3.08879137e-01 -9.21624526e-02 -3.81040096e-01 -4.96921480e-01 1.00767207e+00 5.89499235e-01 -1.71308324e-01 -1.36241484e+00 -1.35061443e+00 7.60756850e-01 5.44818759e-01 1.62671849e-01 -2.91533083e-01 9.50405821e-02 -5.47120813e-03 -7.31484145e-02 5.22833228e-01 -6.39351085e-03 -4.17003810e-01 2.22103924e-01 -1.02738369e+00 2.71772593e-01 4.85044509e-01 1.51388407e+00 1.29628074e+00 -7.56973699e-02 1.08516105e-01 7.48554289e-01 7.80470073e-01 7.93934584e-01 -1.44704267e-01 -9.88777220e-01 9.35391068e-01 7.89682090e-01 1.58408672e-01 -1.17373037e+00 1.75546959e-01 -5.22018194e-01 -7.80139387e-01 2.39520386e-01 -7.83049613e-02 3.54458630e-01 -8.73162389e-01 1.56642509e+00 3.86730224e-01 6.18652582e-01 -2.49252319e-02 8.73061240e-01 4.42148089e-01 6.56049728e-01 -1.43460289e-01 -1.24588102e-01 1.00912535e+00 -9.04613554e-01 -5.13679922e-01 -1.16161138e-01 3.87790859e-01 -1.00873566e+00 8.79847884e-01 1.35067925e-01 -9.26922441e-01 -5.18345296e-01 -1.14691472e+00 -3.12328130e-01 1.62841246e-01 5.22823453e-01 7.41570592e-01 1.62716225e-01 -7.17702448e-01 4.69932348e-01 -1.05559301e+00 -1.33851573e-01 3.30500185e-01 3.19389224e-01 -6.73685789e-01 -5.10778725e-01 -4.32660580e-01 2.98364520e-01 1.71993077e-01 -1.18995987e-01 -8.63657355e-01 -6.78343952e-01 -9.86701846e-01 -6.08467720e-02 1.86326087e-01 -7.46450245e-01 7.27494538e-01 -3.22343588e-01 -1.11298430e+00 1.01237428e+00 -3.35028887e-01 9.36033055e-02 3.94223064e-01 -1.65350378e-01 -5.63754439e-02 5.70275076e-02 4.70590144e-01 6.22967601e-01 1.02398205e+00 -1.79003084e+00 -2.56244272e-01 -5.48822522e-01 -1.85061574e-01 3.47924471e-01 -1.33425236e-01 -3.71687204e-01 -8.06101382e-01 -5.53799272e-01 1.06051815e+00 -1.04958832e+00 -1.30016148e-01 2.91193783e-01 -3.05384874e-01 -2.02737778e-01 1.00160897e+00 -4.67507929e-01 7.11319268e-01 -2.34514570e+00 4.86898214e-01 4.52909827e-01 3.00305963e-01 -2.73876041e-01 -1.03133991e-01 3.10663700e-01 -2.96344817e-01 -1.27950609e-01 -5.89307010e-01 -9.92638052e-01 6.81451708e-02 6.64988339e-01 -5.81945896e-01 6.50994122e-01 4.78373468e-02 6.17828786e-01 -1.04492402e+00 -4.63601291e-01 3.28840792e-01 5.65936446e-01 -7.84504652e-01 2.01909781e-01 -2.39897847e-01 8.32347274e-01 -6.80104911e-01 8.39711070e-01 1.19156921e+00 -2.61427999e-01 -8.36474374e-02 -1.21413745e-01 -3.26560348e-01 -4.15360555e-02 -1.28496444e+00 2.49252224e+00 -2.64667660e-01 1.19672611e-01 1.18452363e-01 -8.09314311e-01 1.31391287e+00 2.36668140e-01 7.40663528e-01 -5.68128773e-04 -2.69127846e-01 5.03428459e-01 -6.22977674e-01 -2.58629382e-01 4.75132316e-01 -1.90837830e-01 4.06871503e-03 2.48350918e-01 8.40126574e-02 -5.31861365e-01 -2.93158710e-01 1.97392449e-01 7.41292775e-01 4.38158512e-01 -1.16101488e-01 -2.28848740e-01 6.55091345e-01 -1.59896284e-01 8.07932675e-01 3.69503707e-01 2.77737409e-01 1.05088747e+00 3.27076912e-01 -2.62710363e-01 -1.46539557e+00 -1.40646362e+00 -3.04118156e-01 3.11739177e-01 3.49859327e-01 -3.56117755e-01 -3.07059914e-01 -4.78414983e-01 2.07364514e-01 4.96588290e-01 -3.42977762e-01 1.33676052e-01 -3.95848155e-01 -6.72811717e-02 -9.83252749e-03 3.33816409e-01 3.64983380e-01 -5.19671977e-01 2.43388399e-01 3.12747322e-02 -2.32116669e-01 -1.05422020e+00 -4.71993446e-01 -2.34509230e-01 -1.22599924e+00 -9.29328203e-01 -6.46738827e-01 -9.67923105e-01 1.15117180e+00 7.59327352e-01 9.09086466e-01 1.37469396e-01 1.61715582e-01 2.56174684e-01 -3.13745350e-01 1.16899781e-01 -7.34537989e-02 8.45962949e-03 2.74504989e-01 2.80108690e-01 1.09238796e-01 -1.03176677e+00 -3.32796067e-01 3.78417850e-01 -9.35088873e-01 2.71019280e-01 5.80470502e-01 6.85372472e-01 1.04460621e+00 -3.92740332e-02 -1.55138180e-01 -5.91241479e-01 7.92781531e-04 -5.56927025e-01 -8.65103900e-01 3.16684693e-01 -3.12531829e-01 2.96991140e-01 2.46363431e-01 -1.54438084e-02 -9.63749707e-01 6.39456332e-01 4.36256453e-02 -1.28962517e+00 -5.05692773e-02 4.30654645e-01 -5.46060920e-01 -2.19504327e-01 2.73402721e-01 2.88874954e-01 -7.16826587e-04 -1.00131917e+00 4.60153699e-01 4.04760897e-01 6.68078899e-01 -1.02279246e+00 1.52139521e+00 9.67654705e-01 8.01507756e-02 -5.78959703e-01 -6.50571644e-01 -9.49946225e-01 -1.12176836e+00 2.03226537e-01 9.07130480e-01 -1.23859811e+00 -2.41752446e-01 2.66446412e-01 -1.32732570e+00 3.45453650e-01 -1.95626915e-01 5.41379035e-01 -6.76445127e-01 7.65656710e-01 -1.17093891e-01 -5.11126101e-01 1.38733521e-01 -1.26257849e+00 1.52456975e+00 -2.34842688e-01 1.89043343e-01 -6.55741692e-01 3.26313227e-01 3.14449489e-01 -1.86009616e-01 3.09324861e-01 6.35057271e-01 -1.54141605e-01 -1.21882057e+00 -2.08344340e-01 -3.03859919e-01 1.86720088e-01 3.18308651e-01 5.69436215e-02 -7.43117750e-01 -5.32855153e-01 1.99180469e-01 -1.74489334e-01 7.85540938e-01 1.43226348e-02 1.07337594e+00 3.32058556e-02 -4.65017796e-01 1.08785439e+00 1.55329609e+00 -3.37096065e-01 5.04849672e-01 -1.61441475e-01 1.08454788e+00 5.23608267e-01 7.76266336e-01 5.58255613e-01 4.25493121e-01 6.66439116e-01 5.66219211e-01 5.37538528e-02 2.62807369e-01 -7.53559351e-01 -1.04014035e-02 1.26666903e+00 -3.05764638e-02 3.12910765e-01 -1.02383506e+00 5.08682370e-01 -1.98600924e+00 -6.68191075e-01 -1.92641631e-01 2.38258576e+00 5.73718846e-01 -4.78143618e-02 -5.37465215e-01 -5.11898585e-02 9.07965302e-01 2.83170491e-01 -4.08677846e-01 3.91309261e-01 -2.05288474e-02 -1.04096040e-01 4.30492789e-01 6.15947783e-01 -1.08654714e+00 8.40570807e-01 5.11012936e+00 9.11486328e-01 -8.20268333e-01 2.66705662e-01 9.77167413e-02 2.56915987e-01 -8.86459768e-01 6.87425733e-01 -5.50723732e-01 3.00156683e-01 1.39639035e-01 5.60365394e-02 2.98004419e-01 1.10373712e+00 -5.45002520e-02 3.28709185e-01 -1.06528497e+00 1.39971304e+00 7.15975836e-02 -1.39681256e+00 2.41406217e-01 3.42043728e-01 1.13718796e+00 1.47829160e-01 6.96048373e-03 1.03802904e-01 2.35372201e-01 -5.60634732e-01 8.02184403e-01 8.13078821e-01 1.16551733e+00 -6.17656767e-01 3.45957577e-01 5.22564054e-01 -1.44669664e+00 3.00982744e-01 -7.65231311e-01 6.59946576e-02 2.56043077e-01 6.32488966e-01 -3.74910802e-01 9.65932548e-01 4.85372573e-01 1.14322424e+00 -3.51715237e-01 1.09554768e+00 -4.32201117e-01 2.38826185e-01 -4.56271827e-01 6.62196577e-01 1.38589039e-01 -7.02489376e-01 7.81234324e-01 5.27924657e-01 5.93784392e-01 2.87196547e-01 5.12014985e-01 1.17220163e+00 -2.68871849e-03 -4.88481708e-02 -7.14311123e-01 1.09960027e-01 6.67847872e-01 1.34764278e+00 -5.67606091e-01 -2.82609195e-01 -4.71750289e-01 8.22312832e-01 4.48588699e-01 3.42895687e-01 -6.44020617e-01 1.02471903e-01 7.07104027e-01 4.54294086e-02 2.92296708e-01 -7.18346179e-01 -5.93275964e-01 -1.55535138e+00 2.40325600e-01 -2.30240688e-01 8.09169188e-03 -9.38075721e-01 -1.38056469e+00 3.15106422e-01 1.02660678e-01 -2.01917124e+00 8.41357186e-02 -2.72977293e-01 -5.55337429e-01 9.03626919e-01 -1.53977787e+00 -1.45753658e+00 -6.06199265e-01 7.51710355e-01 2.82176703e-01 -2.91619867e-01 7.39187062e-01 3.69665235e-01 -1.60450578e-01 1.94246888e-01 4.10440505e-01 -6.05501570e-02 6.42347336e-01 -1.09865451e+00 3.17149401e-01 6.71384394e-01 2.20196262e-01 9.34452295e-01 4.56875563e-01 -9.40942824e-01 -1.35994768e+00 -1.13907897e+00 6.05966628e-01 -5.04034340e-01 3.64111304e-01 -5.41672051e-01 -1.11502552e+00 7.08169460e-01 -3.03928465e-01 2.48012453e-01 2.59969682e-01 -6.19514361e-02 -5.41270912e-01 -1.56656414e-01 -9.95732546e-01 2.54655540e-01 1.08529115e+00 -7.07104206e-01 -7.81426370e-01 6.27266288e-01 1.06061661e+00 -6.17012441e-01 -7.07707882e-01 5.61399043e-01 2.15417862e-01 -6.75893724e-01 1.13157547e+00 -1.55591995e-01 5.11350989e-01 -6.95455372e-01 -4.41402316e-01 -9.28035498e-01 -4.53125685e-01 -4.37919617e-01 1.41622037e-01 1.45465410e+00 -3.06588486e-02 -5.33002853e-01 1.01869631e+00 6.02402985e-01 -5.11963785e-01 -6.11181200e-01 -9.79165554e-01 -7.24536836e-01 1.68773439e-02 -3.83164406e-01 8.97463799e-01 1.19220841e+00 -4.84946489e-01 1.41818617e-02 -4.16711181e-01 6.01508141e-01 1.20666146e+00 4.56707060e-01 1.00940466e+00 -1.44387150e+00 -1.27316669e-01 1.87615380e-01 -5.57669222e-01 -1.29780257e+00 3.44191045e-01 -1.10631275e+00 -5.35331853e-02 -1.50049233e+00 2.06047118e-01 -9.40790534e-01 -1.91022128e-01 5.97609103e-01 2.38051429e-01 7.61659667e-02 1.11604393e-01 9.18132663e-01 -4.88009423e-01 1.03244388e+00 1.15014970e+00 -1.76519334e-01 -1.23038277e-01 -8.54656026e-02 -3.58966351e-01 7.36943841e-01 4.92334634e-01 -6.33355439e-01 -4.89747971e-01 -7.59480774e-01 1.51746437e-01 2.60406643e-01 4.12403286e-01 -1.16578233e+00 4.55719203e-01 -2.43117481e-01 1.37800977e-01 -1.21210134e+00 6.35255098e-01 -1.10751450e+00 4.75658536e-01 3.32890414e-02 5.15560694e-02 -2.17257261e-01 -9.10025463e-02 8.69225919e-01 -3.65525305e-01 -2.03612044e-01 6.01214468e-01 -5.12911826e-02 -4.94091094e-01 8.62385392e-01 5.73542297e-01 -2.82376885e-01 9.66501594e-01 -3.05491030e-01 1.73846468e-01 -1.24872550e-01 -6.58631563e-01 4.85902876e-01 9.87818480e-01 6.62457705e-01 9.31777358e-01 -1.79201472e+00 -7.13801920e-01 6.47506833e-01 5.08000970e-01 6.42914236e-01 3.83072257e-01 5.73774636e-01 -5.70059955e-01 4.70591336e-01 -2.41531372e-01 -1.22024071e+00 -1.03071880e+00 4.84980196e-01 -1.14411660e-01 -6.92498907e-02 -7.30017364e-01 6.69833362e-01 4.90717977e-01 -1.04071784e+00 8.04997422e-03 -2.65855104e-01 2.80915443e-02 -3.00447345e-01 2.04657018e-02 -5.87461703e-02 -1.33813843e-01 -1.00120163e+00 -2.35917881e-01 1.16443717e+00 6.31957725e-02 -8.74062255e-02 1.47924829e+00 -1.60532087e-01 -5.42433500e-01 3.67369592e-01 1.26811397e+00 2.88439244e-01 -1.61551809e+00 -5.57190359e-01 -1.89850613e-01 -1.06840181e+00 -1.37853742e-01 4.40677218e-02 -9.76462781e-01 8.71953011e-01 2.32341930e-01 -4.06695276e-01 7.13915169e-01 1.70711920e-01 6.48677886e-01 2.52246469e-01 8.54482770e-01 -6.35896742e-01 -2.02662777e-02 5.15154600e-01 1.13536453e+00 -1.06624556e+00 2.80156940e-01 -8.20303261e-01 -3.09250087e-01 8.48097444e-01 4.42897916e-01 -3.97773534e-01 7.18040168e-01 -2.32243329e-01 -3.43917668e-01 -4.26413029e-01 -5.43047249e-01 9.25534815e-02 3.86936873e-01 4.54545289e-01 -1.13740815e-02 6.04161397e-02 -4.32546176e-02 3.51959735e-01 -1.50643155e-01 -1.97543174e-01 9.20878425e-02 8.53295267e-01 -3.34949613e-01 -1.39210761e+00 -6.97952628e-01 2.16347501e-01 1.77397087e-01 1.18029155e-01 -1.50891887e-02 4.92944956e-01 2.54591227e-01 4.77479130e-01 7.00623840e-02 -3.62382948e-01 1.77775204e-01 -1.12905703e-01 2.02938423e-01 -9.83111143e-01 2.74676323e-01 3.47887158e-01 -5.27707696e-01 -5.01074731e-01 -5.10875404e-01 -8.54484499e-01 -1.35036695e+00 -5.95350452e-02 -4.14100766e-01 3.73876721e-01 7.73478389e-01 8.45560491e-01 4.15655017e-01 6.45329133e-02 7.67829299e-01 -1.24496388e+00 -4.96163636e-01 -6.34946167e-01 -7.24430323e-01 6.91870034e-01 2.44229198e-01 -1.08930779e+00 -5.18132389e-01 3.12204678e-02]
[8.250185012817383, -3.4549224376678467]
4136e42b-f712-48af-bdd7-067bc27be315
exploring-self-attention-for-crop-type
2210.13167
null
https://arxiv.org/abs/2210.13167v1
https://arxiv.org/pdf/2210.13167v1.pdf
Exploring Self-Attention for Crop-type Classification Explainability
Automated crop-type classification using Sentinel-2 satellite time series is essential to support agriculture monitoring. Recently, deep learning models based on transformer encoders became a promising approach for crop-type classification. Using explainable machine learning to reveal the inner workings of these models is an important step towards improving stakeholders' trust and efficient agriculture monitoring. In this paper, we introduce a novel explainability framework that aims to shed a light on the essential crop disambiguation patterns learned by a state-of-the-art transformer encoder model. More specifically, we process the attention weights of a trained transformer encoder to reveal the critical dates for crop disambiguation and use domain knowledge to uncover the phenological events that support the model performance. We also present a sensitivity analysis approach to understand better the attention capability for revealing crop-specific phenological events. We report compelling results showing that attention patterns strongly relate to key dates, and consequently, to the critical phenological events for crop-type classification. These findings might be relevant for improving stakeholder trust and optimizing agriculture monitoring processes. Additionally, our sensitivity analysis demonstrates the limitation of attention weights for identifying the important events in the crop phenology as we empirically show that the unveiled phenological events depend on the other crops in the data considered during training.
['Xiao Xiang Zhu', 'Dario Augusto Borges Oliveira', 'Ribana Roscher', 'Ivica Obadic']
2022-10-24
null
null
null
null
['type']
['speech']
[ 3.24556559e-01 3.47293854e-01 -4.58159357e-01 -4.10597742e-01 -1.45209104e-01 -8.21492910e-01 3.36549640e-01 8.97086203e-01 2.88875580e-01 4.39563364e-01 3.24035823e-01 -8.44672084e-01 -4.67857093e-01 -1.06872511e+00 -1.19011235e+00 -7.49749660e-01 -4.22801703e-01 -1.34422928e-01 -6.48403347e-01 -6.58554137e-01 -1.26865804e-01 5.40747404e-01 -1.77066123e+00 5.89275599e-01 8.81127179e-01 1.13155198e+00 6.76952600e-01 6.32979691e-01 3.38846892e-02 7.07437098e-01 -4.57575560e-01 -2.24815682e-01 4.70772386e-02 -1.41928747e-01 -4.54536885e-01 -8.75168741e-02 -4.21873294e-02 -3.30073267e-01 1.83846638e-01 1.07450986e+00 1.13263287e-01 -4.22030210e-01 4.78919089e-01 -1.22186208e+00 -1.08189499e+00 9.40331697e-01 -3.40873629e-01 1.83574721e-01 -6.25457540e-02 1.41368121e-01 1.24981511e+00 -4.35684323e-01 2.12065369e-01 9.21395421e-01 9.25671577e-01 -1.14854135e-01 -9.21432436e-01 -5.84245682e-01 5.33851206e-01 3.82197410e-01 -9.26840782e-01 -1.31948918e-01 5.88460624e-01 -5.80593288e-01 9.01383817e-01 3.99325848e-01 8.65785420e-01 9.18123960e-01 5.94197035e-01 7.05420315e-01 1.00350642e+00 -2.92484105e-01 -4.95256148e-02 2.83898264e-02 3.40057194e-01 5.10594249e-01 5.32075167e-01 5.37317514e-01 -4.90456253e-01 3.69241424e-02 6.03555977e-01 4.34260070e-01 -3.63387465e-01 -2.49596629e-02 -1.20683992e+00 9.83485639e-01 1.05994630e+00 2.98362374e-01 -1.11255896e+00 2.53650486e-01 3.28901082e-01 2.87412256e-01 8.36396396e-01 9.83907402e-01 -1.15086067e+00 4.16192710e-01 -7.10564494e-01 -6.53684288e-02 5.18165231e-01 8.46506178e-01 8.95134389e-01 3.33464518e-02 -2.29359180e-01 1.64594531e-01 2.80279040e-01 8.35168839e-01 -1.36322111e-01 -6.69886529e-01 4.37601171e-02 8.29180837e-01 3.67035240e-01 -1.18749702e+00 -5.14570534e-01 -6.50435209e-01 -8.09362590e-01 -7.14978576e-02 2.02970043e-01 -1.92885011e-01 -9.05414879e-01 1.63510430e+00 -1.30015910e-01 4.60712379e-03 2.57157147e-01 8.84473801e-01 5.98884344e-01 7.64948249e-01 3.46221149e-01 3.82625237e-02 1.83710790e+00 -2.64339864e-01 -1.01790321e+00 -2.24714532e-01 6.36486769e-01 -5.59899569e-01 8.83361220e-01 -8.44306052e-02 -4.98255014e-01 -5.32235682e-01 -1.08293760e+00 4.08212781e-01 -9.84960914e-01 3.55650991e-01 1.32567823e+00 2.06384063e-01 -5.65423965e-01 8.37618172e-01 -8.78963411e-01 -6.66159868e-01 5.21752775e-01 8.74827132e-02 -1.44802481e-01 2.19747081e-01 -1.53978777e+00 1.26433623e+00 5.08894384e-01 7.76549518e-01 -1.06864905e+00 -1.04698586e+00 -9.70375478e-01 5.59921384e-01 -3.94749343e-02 -2.22274557e-01 8.06274951e-01 -1.10929859e+00 -8.11229706e-01 9.89768326e-01 -1.16313480e-01 -8.58979285e-01 -2.69719779e-01 -2.81185120e-01 -5.53585589e-01 -7.39628077e-02 3.11056495e-01 7.01679111e-01 6.72307909e-01 -1.21451128e+00 -7.92355120e-01 -3.89903516e-01 1.78405300e-01 -7.74954036e-02 -3.80616814e-01 -1.45669490e-01 6.56514823e-01 -6.38540685e-01 9.74681079e-02 -6.84778094e-01 -1.03776917e-01 -8.37901141e-03 -1.33798346e-01 9.74184126e-02 5.04293740e-01 -8.55355680e-01 1.02597821e+00 -2.07880163e+00 -1.12281583e-01 -1.63190272e-02 1.55376211e-01 1.59215435e-01 -2.53609180e-01 4.91344690e-01 -3.46196145e-01 3.39208364e-01 -3.08897942e-01 4.78253573e-01 4.44666781e-02 3.54777306e-01 -6.66427433e-01 3.07038933e-01 1.10925949e+00 1.12758863e+00 -1.10771167e+00 2.35238090e-01 2.80034095e-01 5.09957671e-01 3.16811888e-03 3.66055727e-01 -4.68723625e-01 2.18334615e-01 -4.95460302e-01 1.07243574e+00 7.62553990e-01 -2.61157393e-01 2.75373340e-01 -5.91393590e-01 -4.21256840e-01 1.86589733e-01 -3.36403221e-01 1.16432416e+00 -3.85310918e-01 9.57696199e-01 -4.69638547e-03 -1.19020796e+00 9.85327721e-01 4.16526049e-01 3.98534626e-01 -5.87304592e-01 -2.42991298e-02 3.35564882e-01 1.11162543e-01 -5.92354357e-01 5.30442595e-01 2.33563650e-02 1.26789197e-01 1.70607001e-01 2.37389142e-03 4.08809409e-02 -3.44911486e-01 -4.48733211e-01 5.80497980e-01 3.46257389e-01 5.11805415e-01 -6.49545491e-01 -3.07063684e-02 3.81966382e-01 5.40479600e-01 5.19438148e-01 -3.16077679e-01 3.49337220e-01 6.93034112e-01 -1.05689681e+00 -8.20261002e-01 -3.64250541e-01 -3.93058509e-01 1.23263180e+00 -1.75667390e-01 -1.72199570e-02 -2.72647917e-01 -5.70803940e-01 4.68605250e-01 1.03552616e+00 -1.30958009e+00 -3.15298319e-01 -4.21640575e-02 -1.23075652e+00 6.63765252e-01 7.71556497e-01 3.58010679e-01 -1.26797819e+00 -1.25170112e+00 2.51580805e-01 -3.72357368e-01 -8.80446076e-01 2.99751490e-01 1.16407728e+00 -6.73163414e-01 -1.28187227e+00 -6.22144580e-01 -4.52843696e-01 5.84641397e-01 4.63700801e-01 1.24791563e+00 -2.19953875e-03 -1.09011102e-02 -3.41969691e-02 -6.66106939e-01 -1.11469817e+00 -3.56779158e-01 3.48782510e-01 -3.54714394e-01 -2.22212493e-01 8.58140171e-01 -2.05529854e-01 -5.49090266e-01 2.15880483e-01 -8.95398974e-01 6.15600124e-02 7.21195877e-01 8.83972883e-01 3.00379843e-01 1.27586097e-01 6.24498963e-01 -6.63686097e-01 1.21532969e-01 -1.01948607e+00 -5.60621023e-01 5.49048662e-01 -6.43208027e-01 3.67581606e-01 2.32203677e-01 -1.36670604e-01 -6.46556139e-01 9.75399762e-02 2.25902990e-01 -5.16156442e-02 -2.91821331e-01 1.24132919e+00 3.35466117e-02 2.40075454e-01 6.52367294e-01 9.00174677e-02 -1.06910430e-01 -1.93138704e-01 1.46139190e-01 6.34873152e-01 2.66073048e-01 -1.67109743e-01 4.57732618e-01 3.61923754e-01 -9.03919041e-02 -6.78991199e-01 -1.27562928e+00 -3.48570377e-01 -4.79563117e-01 -1.36312261e-01 8.57504487e-01 -1.19194388e+00 -7.53342867e-01 2.70056248e-01 -1.43797302e+00 -3.93870443e-01 -3.03269714e-01 5.13081849e-01 -2.25201368e-01 -2.89848089e-01 -1.62057027e-01 -8.26007485e-01 -4.19684231e-01 -8.79929364e-01 1.21912229e+00 1.61790863e-01 -2.34156668e-01 -9.97281969e-01 -2.84010142e-01 -1.40687034e-01 6.57674015e-01 5.97376466e-01 1.05955577e+00 -5.68631589e-01 -3.53206307e-01 -1.02868825e-01 -6.47850990e-01 -1.52206957e-01 5.48141062e-01 1.73396423e-01 -1.37083638e+00 1.62643269e-02 -2.81280186e-03 6.68827957e-03 1.00561965e+00 9.13465559e-01 8.81786227e-01 -2.81050116e-01 -2.40703642e-01 7.92249441e-01 1.22805226e+00 1.10605612e-01 4.31220680e-01 5.86900353e-01 4.84396905e-01 1.14944315e+00 1.01222169e+00 4.67079133e-01 5.10865986e-01 1.97139412e-01 1.12719274e+00 -6.69081390e-01 1.59284174e-01 -3.41655731e-01 4.69669908e-01 2.14792505e-01 -4.78945076e-02 -1.59071177e-01 -9.34313655e-01 9.37601507e-01 -1.91147768e+00 -9.95705545e-01 -2.82201767e-01 2.03948760e+00 6.01550519e-01 -2.33771116e-01 -4.04570490e-01 7.77840912e-02 6.13937736e-01 4.68535960e-01 -7.76031435e-01 -3.95212084e-01 -7.23150790e-01 1.58856928e-01 1.19356072e+00 2.69472390e-01 -1.24744916e+00 1.14700007e+00 6.26962852e+00 -4.76623848e-02 -1.30764544e+00 -1.63109615e-01 6.22689188e-01 4.10583526e-01 -5.11351168e-01 5.57426400e-02 -6.58419788e-01 2.03121156e-02 9.44615602e-01 1.67804919e-02 3.09706539e-01 5.96629560e-01 4.98810619e-01 -7.91642815e-02 -1.15135849e+00 1.89016432e-01 -3.21732342e-01 -1.31746352e+00 -2.85543595e-02 -6.59903586e-02 6.36644781e-01 1.10798426e-01 7.24021792e-02 1.73839107e-02 4.39546734e-01 -1.10225046e+00 5.59807777e-01 4.64079887e-01 7.32559979e-01 -5.07432818e-01 1.05039823e+00 -4.90631163e-02 -1.34935844e+00 -3.36688399e-01 -6.15120351e-01 -1.86950624e-01 -4.06822652e-01 7.10286796e-01 -9.28208888e-01 7.96118796e-01 9.57601666e-01 1.36583698e+00 -4.66805339e-01 4.78811860e-01 -3.43084574e-01 8.55314612e-01 -1.42247632e-01 8.05621222e-02 4.46650058e-01 1.84091657e-01 2.30832547e-01 1.01411772e+00 6.49651885e-01 1.02806792e-01 -4.33256418e-01 1.26153481e+00 2.84097135e-01 -2.52718955e-01 -9.65042412e-01 -5.61721563e-01 3.71081948e-01 1.04092050e+00 -6.15931451e-01 3.69526520e-02 -3.96830924e-02 7.44803607e-01 -9.49363038e-02 4.77744967e-01 -7.77453125e-01 -3.31188619e-01 1.02372956e+00 -2.72193197e-02 5.21476746e-01 -1.15786316e-02 -4.27206546e-01 -8.80253613e-01 -2.60838836e-01 -9.34034467e-01 1.05998874e-01 -1.18462217e+00 -1.00019109e+00 4.72376496e-01 -1.09238736e-01 -1.13048649e+00 4.82194647e-02 -7.35687613e-01 -6.45370901e-01 1.17243719e+00 -2.34086800e+00 -1.61947811e+00 -6.29590809e-01 5.61185069e-02 3.85607839e-01 1.17272241e-02 1.40135968e+00 -2.55029559e-01 -4.62201178e-01 1.52978465e-01 1.13817468e-01 -9.08507854e-02 5.28557777e-01 -1.04903483e+00 5.41587055e-01 8.87446821e-01 -1.07718468e-01 4.52184170e-01 8.07135165e-01 -6.55630350e-01 -1.26254141e+00 -1.43594050e+00 1.34385312e+00 -1.98851556e-01 8.98116946e-01 2.63925474e-02 -8.86402488e-01 8.08356404e-01 3.28579903e-01 -2.49460101e-01 7.86822200e-01 4.20762002e-01 -2.92427808e-01 -3.72667074e-01 -8.99115086e-01 2.04117686e-01 5.58603346e-01 -5.55551708e-01 -3.23324889e-01 5.48052251e-01 9.64281023e-01 -6.22117557e-02 -9.40344214e-01 5.76231837e-01 7.53136694e-01 -6.36598766e-01 6.45009935e-01 -8.91135871e-01 8.28029692e-01 -1.56817853e-01 -4.92060393e-01 -1.63087952e+00 -6.00581348e-01 -3.47659200e-01 -2.51309071e-02 9.00569916e-01 4.80758458e-01 -7.03244746e-01 4.33885604e-01 2.14604601e-01 -8.69675130e-02 -4.04888630e-01 -2.90363520e-01 -4.13706809e-01 -3.77096720e-02 -3.70312274e-01 1.25152194e+00 1.32194567e+00 -5.05073443e-02 -2.55264014e-01 -2.81053632e-01 9.65829670e-01 3.33668947e-01 3.55986774e-01 5.22394143e-02 -1.34727705e+00 2.11053386e-01 -4.25321519e-01 -2.83401042e-01 -4.84468609e-01 1.39486670e-01 -7.80130088e-01 -3.90768908e-02 -1.23228085e+00 -1.14917263e-01 -4.44685489e-01 -7.25999177e-01 9.59023774e-01 -5.78763664e-01 -2.43094414e-01 8.79035294e-02 4.65158895e-02 3.21622968e-01 4.85880941e-01 8.09093475e-01 -4.88184065e-01 -1.09142981e-01 2.74988919e-01 -1.09553146e+00 3.12581867e-01 9.94292855e-01 -6.12385154e-01 -1.04952484e-01 -8.82202089e-01 6.91428185e-01 1.64851341e-02 8.18463087e-01 -5.79614878e-01 -3.39122832e-01 -3.96436185e-01 2.59952039e-01 -7.66141236e-01 -2.83683270e-01 -1.11105740e+00 1.14218794e-01 7.77894855e-01 -7.38478780e-01 1.20186858e-01 6.57447815e-01 4.39578295e-01 -1.09112211e-01 -7.58969560e-02 1.06796540e-01 -5.32157086e-02 -6.96992517e-01 1.57126501e-01 -5.89272976e-01 -6.21244609e-01 8.54415357e-01 -3.98757160e-02 -2.63731629e-01 -2.21651331e-01 -7.07066238e-01 5.45664966e-01 2.14186504e-01 6.43428564e-01 2.54498303e-01 -1.23036802e+00 -1.15600801e+00 4.04430240e-01 4.92819607e-01 -3.28894556e-01 -1.38660550e-01 6.53643668e-01 -6.52702570e-01 7.68473387e-01 -5.99393189e-01 -6.21215284e-01 -8.97435129e-01 5.01055121e-01 4.41376090e-01 -3.13853830e-01 -1.25592455e-01 7.99173951e-01 2.03814700e-01 -3.68269354e-01 1.05886377e-01 -1.13360631e+00 -4.94021952e-01 5.32091856e-01 2.06646129e-01 -6.30427003e-02 1.18901424e-01 -4.33050066e-01 -4.27718133e-01 2.77213097e-01 3.36048543e-01 4.26118284e-01 1.76566362e+00 -9.93136615e-02 -2.48164870e-02 6.10017598e-01 7.68542051e-01 -6.96501136e-01 -1.43258142e+00 -1.25088051e-01 2.53789514e-01 -3.33449356e-02 3.06221783e-01 -1.08976686e+00 -1.31933153e+00 1.22590840e+00 9.57014859e-01 5.93666494e-01 1.31013548e+00 -3.79463762e-01 3.49465042e-01 4.84374404e-01 4.17403653e-02 -7.27108598e-01 -6.30130172e-01 3.74250799e-01 1.13186359e+00 -1.83423233e+00 -1.43212363e-01 -1.91497505e-01 -6.32518530e-01 1.20887005e+00 1.60023689e-01 1.87603697e-01 8.19079638e-01 1.80033192e-01 1.71570122e-01 -5.64557970e-01 -6.33205235e-01 -6.39048576e-01 2.00563341e-01 9.00809467e-01 7.76354492e-01 6.90361023e-01 2.68328607e-01 7.16447592e-01 -1.60047226e-02 3.01402453e-02 2.34061554e-01 6.16539598e-01 -3.74651730e-01 -6.77927494e-01 -4.69855964e-01 5.42847872e-01 -3.58902037e-01 -4.96520132e-01 -4.20972675e-01 4.83505577e-01 1.45835862e-01 1.00261831e+00 2.95363907e-02 -3.56220245e-01 3.38106602e-01 -4.79069464e-02 6.70906156e-02 -6.01449668e-01 -1.04799068e+00 -2.41486564e-01 -7.08166510e-02 -4.40014720e-01 -6.81158066e-01 -6.61448002e-01 -8.15579891e-01 -3.13966542e-01 -6.18055224e-01 3.00542731e-02 8.79658461e-01 8.39059293e-01 7.05649197e-01 9.02153969e-01 8.47881079e-01 -8.72940421e-01 -3.26909214e-01 -1.08691454e+00 -4.33766484e-01 9.55041647e-02 6.90625668e-01 -5.55984914e-01 -9.32995155e-02 1.48168653e-01]
[9.414546966552734, -1.5713423490524292]
d3b1407e-f43b-47c7-9894-a3553debc055
social-interactions-with-endogenous-group
2306.01544
null
https://arxiv.org/abs/2306.01544v1
https://arxiv.org/pdf/2306.01544v1.pdf
Social Interactions with Endogenous Group Formation
This paper explores the identification and estimation of social interaction models with endogenous group formation. We characterize group formation using a two-sided many-to-one matching model, where individuals select groups based on their preferences, while groups rank individuals according to their qualifications, accepting the most qualified until reaching capacities. The selection into groups leads to a bias in standard estimates of peer effects, which is difficult to correct for due to equilibrium effects. We employ the limiting approximation of a market as the market size grows large to simplify the selection bias. Assuming exchangeable unobservables, we can express the selection bias of an individual as a group-invariant nonparametric function of her preference and qualification indices. In addition to the selection correction, we show that the excluded variables in group formation can serve as instruments to tackle the reflection problem. We propose semiparametric distribution-free estimators that are root-n consistent and asymptotically normal.
['Xiaoting Sun', 'Shuyang Sheng']
2023-06-02
null
null
null
null
['selection-bias']
['natural-language-processing']
[-1.60487548e-01 3.34442943e-01 -6.99929595e-01 -4.30284590e-01 -2.48222470e-01 -4.38669473e-01 2.90763348e-01 1.15655199e-01 -5.51150739e-01 1.07449770e+00 2.99202293e-01 -4.52163935e-01 -5.50060272e-01 -1.03311551e+00 -5.43580115e-01 -5.65468431e-01 5.09482285e-04 7.64561415e-01 -2.64928728e-01 1.32906705e-01 2.19335467e-01 1.21638209e-01 -1.61745894e+00 -4.79433447e-01 1.57781887e+00 4.49017555e-01 9.89428833e-02 5.02104223e-01 1.23450965e-01 5.13900995e-01 -3.89766097e-01 -4.71772790e-01 7.02393591e-01 -3.89257967e-01 -4.51137364e-01 1.88607052e-01 2.07910970e-01 -4.93176907e-01 -1.24582060e-01 9.43778813e-01 5.28638005e-01 1.59275308e-01 1.18733895e+00 -1.58400905e+00 -9.14848268e-01 8.62849593e-01 -4.63243246e-01 -3.30615103e-01 5.78947246e-01 -3.50640714e-01 1.07270992e+00 -7.21658766e-01 5.71665347e-01 1.49575138e+00 4.16074395e-01 4.75305408e-01 -1.42476404e+00 -9.41106439e-01 1.20409109e-01 -4.33454305e-01 -1.23659706e+00 -3.02422822e-01 3.16247672e-01 -8.69935930e-01 7.81387761e-02 2.69459724e-01 7.53119946e-01 5.50666094e-01 3.30106497e-01 4.32446718e-01 1.21130478e+00 -5.84558606e-01 1.31156266e-01 1.90600619e-01 3.89399558e-01 2.87467718e-01 9.36118126e-01 2.50550598e-01 -4.49766189e-01 -6.13735974e-01 1.22960269e+00 2.98995733e-01 6.03968985e-02 -5.84984362e-01 -8.04752529e-01 1.14316952e+00 3.16479132e-02 6.93971217e-02 -7.21053600e-01 1.15412466e-01 -2.00119182e-01 9.22620595e-01 6.04109287e-01 3.74607325e-01 -4.18744147e-01 2.93659270e-01 -7.13603616e-01 7.00415552e-01 8.96738529e-01 8.09816897e-01 1.13628101e+00 -2.58859843e-01 -5.38377583e-01 6.28140271e-01 4.66522545e-01 9.17868376e-01 3.78914595e-01 -1.19395435e+00 3.40661913e-01 6.53348505e-01 5.97007275e-01 -6.14956737e-01 -3.28646630e-01 -5.15808523e-01 -6.64397359e-01 1.52305558e-01 9.79683876e-01 -6.59733593e-01 -4.02171016e-01 2.09325862e+00 3.45404118e-01 -2.03069732e-01 -1.56487644e-01 5.23041844e-01 2.90259600e-01 3.56090248e-01 7.44900182e-02 -6.48128211e-01 9.85669732e-01 -5.60483575e-01 -6.28980279e-01 2.44154587e-01 9.27904665e-01 -2.06533238e-01 6.45428002e-01 -2.63079088e-02 -1.30255997e+00 -3.07259858e-01 -3.62505801e-02 4.16896999e-01 1.32990211e-01 -3.24273705e-01 1.03752899e+00 9.50601518e-01 -1.32728422e+00 7.27085948e-01 -4.73255545e-01 -1.85385659e-01 8.10160786e-02 9.73129690e-01 -5.95091842e-02 2.82373071e-01 -1.12077045e+00 4.01996702e-01 -5.24417698e-01 -3.67635936e-01 -4.41673160e-01 -9.84009445e-01 -6.19691610e-01 3.04354519e-01 2.60059625e-01 -9.84178543e-01 1.21953857e+00 -1.55101633e+00 -1.30362916e+00 7.25143909e-01 -2.40527228e-01 -1.53826520e-01 7.84486592e-01 1.45582691e-01 -1.97015718e-01 -3.93964827e-01 8.02515566e-01 1.61825806e-01 5.36848664e-01 -1.08344209e+00 -8.82019401e-01 -6.39602065e-01 -1.85265746e-02 4.76695746e-01 -3.59950542e-01 3.59053761e-01 2.36719728e-01 -4.31203455e-01 1.35178464e-02 -9.49792564e-01 -4.06161457e-01 -5.05919456e-01 -8.54736716e-02 -4.19008255e-01 -2.05367014e-01 -5.62514007e-01 1.14055860e+00 -1.78063536e+00 6.66397363e-02 6.68318391e-01 2.22485736e-01 -6.21058285e-01 -2.09411457e-01 4.53057885e-01 9.09898356e-02 1.60401836e-01 2.10519433e-01 -3.94500703e-01 4.71673489e-01 -1.56727105e-01 6.28160387e-02 8.57296288e-01 -3.69785339e-01 8.33700538e-01 -6.96646333e-01 -3.83899957e-01 -3.50690633e-01 -2.10029125e-01 -7.75548875e-01 5.75940348e-02 6.38722599e-01 4.03261364e-01 -6.18069112e-01 5.27910352e-01 7.87528098e-01 -1.47053778e-01 4.20358658e-01 9.60103929e-01 -4.78425860e-01 3.38773727e-01 -1.32042825e+00 7.32090771e-01 -2.30110377e-01 1.87427044e-01 4.40409124e-01 -1.14705932e+00 6.53829694e-01 4.08143520e-01 4.62995440e-01 -4.10729885e-01 1.29037052e-01 5.79000533e-01 3.85426909e-01 -4.79435921e-03 3.52964461e-01 -3.89959663e-01 -1.99500680e-01 5.85423172e-01 1.49543539e-01 7.80985132e-02 2.21782461e-01 2.25169465e-01 7.59284794e-01 -2.17864722e-01 3.38007212e-01 -9.78253067e-01 1.38284937e-01 -6.38757646e-01 1.04064786e+00 1.31808448e+00 2.49602690e-01 1.91462100e-01 6.66108906e-01 2.42536798e-01 -8.81837428e-01 -1.09316838e+00 -1.69033796e-01 1.24140608e+00 1.12900153e-01 5.96747816e-01 -5.94451308e-01 -4.81248945e-01 8.84171426e-01 4.24719810e-01 -7.73870826e-01 1.33013263e-01 -1.45524129e-01 -6.87956333e-01 -7.25007281e-02 3.43700886e-01 6.77804872e-02 -8.61922622e-01 1.31742716e-01 7.48252422e-02 1.77529514e-01 9.08798277e-02 -5.06285310e-01 -2.91842908e-01 -8.26040030e-01 -8.31108868e-01 -1.17965627e+00 -7.51352787e-01 9.76237535e-01 2.22329065e-01 9.36927855e-01 4.40816730e-01 6.25502646e-01 5.72959125e-01 -9.19805765e-02 -7.58302927e-01 -4.33500111e-01 1.27258152e-01 6.76135361e-01 2.84556121e-01 5.82706153e-01 -4.91115808e-01 -7.09142208e-01 3.39093745e-01 -4.09453213e-01 -2.46256948e-01 2.86230236e-01 5.60279071e-01 -1.01475887e-01 -2.57025123e-01 1.10880196e+00 -9.52173591e-01 6.84526086e-01 -6.25090241e-01 -1.01199841e+00 2.62637377e-01 -8.76691222e-01 -2.75449365e-01 1.77623630e-01 -7.00300217e-01 -1.19817710e+00 -2.38005564e-01 6.05349660e-01 4.81300324e-01 3.75987887e-02 4.69239444e-01 -3.21098715e-01 -3.45603287e-01 2.22622037e-01 -3.29499364e-01 2.25408629e-01 -6.40895367e-01 -1.07045300e-01 8.83177519e-01 6.98316991e-02 -7.84484208e-01 1.16457474e+00 3.04403841e-01 1.64392650e-01 -6.22890770e-01 -3.41425568e-01 -4.81099695e-01 -5.09621084e-01 -1.61735132e-01 4.78963047e-01 -1.31829453e+00 -1.07887363e+00 5.89022577e-01 -6.58039093e-01 -5.80831051e-01 -6.53001070e-01 1.10173106e+00 -4.60361183e-01 1.91465452e-01 -8.50809991e-01 -1.43539715e+00 1.46362439e-01 -6.59554899e-01 6.07463479e-01 5.15561402e-01 -2.55682826e-01 -1.21022463e+00 -8.47114064e-03 3.34620774e-01 2.56483436e-01 -1.47552118e-01 4.31083232e-01 -7.08732903e-01 -7.69538581e-01 -2.21373782e-01 8.13273340e-02 -2.15003878e-01 6.29948229e-02 -7.33854175e-02 -3.06587547e-01 -5.70261657e-01 -1.48601964e-01 -4.46301326e-03 5.98179162e-01 1.14832544e+00 5.86582422e-01 -3.45501244e-01 -4.61946130e-01 2.66870290e-01 1.01909196e+00 2.66233176e-01 2.90238291e-01 1.99883550e-01 3.77312094e-01 1.02814353e+00 6.11908555e-01 7.48847485e-01 6.96227312e-01 3.18474054e-01 -4.83448319e-02 -2.05694482e-01 6.68565869e-01 -5.36826909e-01 3.11810523e-01 6.47756696e-01 -3.87695312e-01 -2.15497881e-01 -6.12185359e-01 8.15021932e-01 -1.94909203e+00 -9.86170352e-01 -5.56278169e-01 2.74534154e+00 8.15695763e-01 -2.61654764e-01 6.67835951e-01 -8.55045095e-02 8.83864522e-01 -6.38507605e-01 -2.02775761e-01 -1.06531531e-01 -3.39301914e-01 6.66940436e-02 1.05897844e+00 9.57899690e-01 -3.65762889e-01 5.83281159e-01 7.43897963e+00 4.64911908e-01 -4.40266043e-01 1.57860935e-01 5.81460774e-01 1.11513339e-01 -8.25512648e-01 6.14865065e-01 -9.92141664e-01 4.84321386e-01 6.39470339e-01 -8.31722379e-01 2.67896801e-01 2.32424751e-01 4.13480818e-01 -3.13448936e-01 -1.04992414e+00 3.86404485e-01 -3.54882985e-01 -6.37156606e-01 -4.51834798e-01 9.38706517e-01 1.38230467e+00 -6.07204497e-01 2.71340638e-01 2.91263819e-01 9.31792378e-01 -9.16590095e-01 7.46937215e-01 5.20310640e-01 6.73979104e-01 -7.95813143e-01 6.27423644e-01 6.29777610e-01 -6.84094191e-01 -7.17237949e-01 -5.18471122e-01 -1.15140903e+00 1.55921057e-01 7.21397698e-01 -3.08519691e-01 4.44706827e-01 2.06041619e-01 1.73081577e-01 -1.19404756e-01 1.16577077e+00 -6.21253951e-03 7.72658229e-01 -2.00259551e-01 1.69297174e-01 -2.72066683e-01 -9.34111774e-01 1.41024798e-01 3.72103989e-01 9.91531014e-01 3.48540545e-01 7.06916004e-02 8.50240767e-01 -3.99920970e-01 4.44758922e-01 -5.75281858e-01 9.17722136e-02 6.06330156e-01 7.71842778e-01 -5.26549935e-01 -3.41896445e-01 -7.92518675e-01 5.47194481e-01 -7.56414384e-02 5.82289040e-01 -3.03235531e-01 -2.53044426e-01 6.94108307e-01 6.55710757e-01 -1.65159672e-01 -1.03879571e-02 -1.36219919e-01 -1.40392768e+00 -2.78128058e-01 -6.90632403e-01 3.58564526e-01 -1.00620583e-01 -1.17108488e+00 -5.96016824e-01 1.29309133e-01 -6.74223363e-01 -3.81348670e-01 -1.50640249e-01 -4.58658278e-01 1.10145581e+00 -1.19441390e+00 -6.64040983e-01 1.94667235e-01 2.94196367e-01 -4.01768871e-02 -1.48340374e-01 3.83598715e-01 3.84944826e-01 -2.98997045e-01 6.53587580e-01 5.60766101e-01 -7.43777677e-02 7.28932559e-01 -1.37821925e+00 1.62993431e-01 5.27967274e-01 -3.37714523e-01 8.21389437e-01 6.10323966e-01 -1.10720754e+00 -1.00668037e+00 -7.02482045e-01 1.51962066e+00 -5.28844893e-01 6.77161276e-01 -5.88892043e-01 -6.19015574e-01 9.78524446e-01 -1.99621022e-02 -3.98372710e-01 7.30372190e-01 5.74372947e-01 3.18699569e-01 -1.34809211e-01 -1.21846747e+00 4.56622005e-01 1.28385782e+00 -4.06414211e-01 -2.36775219e-01 3.16604316e-01 5.11544883e-01 1.38223559e-01 -1.13627636e+00 9.47079509e-02 6.81833804e-01 -1.01002491e+00 5.82626939e-01 -5.27556062e-01 -6.43190369e-02 1.57928973e-01 2.42739320e-01 -1.32383752e+00 -8.60845327e-01 -8.37590814e-01 5.96048295e-01 1.59066522e+00 4.34855819e-01 -1.16930747e+00 9.09940481e-01 1.02057195e+00 5.32173693e-01 7.79390037e-02 -9.09610689e-01 -9.84325111e-01 6.49614334e-01 5.06888986e-01 1.03530037e+00 1.20565331e+00 2.71962464e-01 2.12302580e-01 -4.92407471e-01 -6.13668561e-02 9.36553955e-01 3.00145924e-01 8.98221374e-01 -2.18324018e+00 -4.39773947e-01 -8.97554979e-02 -8.34100023e-02 -1.05351222e+00 6.06402755e-01 -6.21165514e-01 -2.50102639e-01 -1.30880570e+00 6.94079101e-01 -6.95285201e-01 1.60769653e-02 -1.02944799e-01 -1.27259204e-02 -5.52550256e-02 1.03925698e-01 1.86106488e-02 -3.87269914e-01 2.59832114e-01 1.04736412e+00 1.88250601e-01 -7.16928422e-01 5.61084747e-01 -1.08695042e+00 6.17042422e-01 5.74764073e-01 -4.91433322e-01 -2.19365254e-01 -5.62583143e-03 5.08006990e-01 7.69803226e-01 1.85906291e-01 -3.23930442e-01 4.34176996e-02 -8.14032614e-01 1.74731851e-01 -2.04327092e-01 -1.19721033e-01 -6.17872655e-01 6.31695628e-01 4.56999391e-01 -5.11244357e-01 -1.52632147e-01 -6.90085530e-01 2.01243922e-01 1.14697024e-01 -2.98409075e-01 3.50841045e-01 1.08735964e-01 3.60131532e-01 5.64217031e-01 -8.09686184e-01 -1.35434493e-01 8.03941846e-01 -1.70676857e-01 2.88310815e-02 -1.04632330e+00 -5.93977273e-01 5.82102597e-01 1.12092364e+00 -5.53501816e-03 -2.58833095e-02 -1.55249226e+00 -1.05475903e+00 2.45788679e-01 -8.41051936e-02 -5.29677987e-01 -1.33309793e-02 7.21192837e-01 1.05979130e-01 2.59485245e-01 -1.49872959e-01 9.22337398e-02 -1.43247914e+00 3.61744910e-01 3.30027819e-01 -4.68617640e-02 2.27841094e-01 6.92176640e-01 8.29766452e-01 -5.72746098e-01 -1.30984234e-02 -1.51605867e-02 -2.77984291e-01 2.41847008e-01 2.78041184e-01 7.91381359e-01 -6.54646456e-01 -7.59456933e-01 -2.15477452e-01 5.19426107e-01 4.41224158e-01 -3.48430455e-01 9.97395933e-01 -6.78595126e-01 -3.90589058e-01 4.57831800e-01 6.81367218e-01 4.07827735e-01 -1.14549148e+00 -1.77202895e-01 1.32696358e-02 -7.14364111e-01 -6.11829281e-01 -1.75215334e-01 -7.96091139e-01 2.07424164e-01 2.06244424e-01 5.90642571e-01 6.89143181e-01 -1.71879724e-01 6.62177848e-03 -8.27201679e-02 4.71375734e-01 -1.27536023e+00 -4.61754680e-01 1.79297462e-01 4.40592855e-01 -1.13222396e+00 -8.66613351e-03 -6.41736329e-01 -1.62001580e-01 2.53475785e-01 3.18416893e-01 -3.09570193e-01 8.97409379e-01 -1.64497033e-01 -1.49824813e-01 3.25719476e-01 -6.91794455e-01 -1.87034279e-01 2.67128021e-01 8.91966462e-01 6.52782083e-01 6.11892998e-01 -1.33355057e+00 7.42461801e-01 -2.89990753e-01 -1.37896642e-01 9.21885014e-01 3.79730463e-01 -5.38174808e-01 -1.53644395e+00 -7.04741776e-01 9.61435676e-01 -9.48933601e-01 1.49784222e-01 -4.16703761e-01 7.57175386e-01 1.07358404e-01 1.08429086e+00 4.70014632e-01 4.09705460e-01 3.71509582e-01 4.17293087e-02 3.75130862e-01 -6.45297587e-01 -5.44224620e-01 2.33079836e-01 4.77745607e-02 4.13274392e-02 -6.20747745e-01 -1.24654925e+00 -8.49541485e-01 -7.29242861e-01 -8.28665316e-01 6.93865776e-01 1.30885318e-01 7.61573315e-01 4.92297374e-02 -1.02697030e-01 1.06228137e+00 -7.19313383e-01 -9.51261282e-01 -8.37029099e-01 -1.63872528e+00 4.46868151e-01 2.52240658e-01 -7.26404488e-01 -7.89478302e-01 -4.93998617e-01]
[7.910217761993408, 5.190557479858398]
31f99021-703d-45b6-8329-4108f8e9da94
put-at-semeval-2016-task-4-the-abc-of-twitter
null
null
https://aclanthology.org/S16-1018
https://aclanthology.org/S16-1018.pdf
PUT at SemEval-2016 Task 4: The ABC of Twitter Sentiment Analysis
null
['Mateusz Lango', 'Dariusz Brzezinski', 'Jerzy Stefanowski']
2016-06-01
null
null
null
semeval-2016-6
['twitter-sentiment-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]
[-7.284582614898682, 3.654465675354004]
b574064a-f504-451e-ab22-6f0c420e2e49
affective-manifolds-modeling-machine-s-mind
2208.13386
null
https://arxiv.org/abs/2208.13386v1
https://arxiv.org/pdf/2208.13386v1.pdf
Affective Manifolds: Modeling Machine's Mind to Like, Dislike, Enjoy, Suffer, Worry, Fear, and Feel Like A Human
After the development of different machine learning and manifold learning algorithms, it may be a good time to put them together to make a powerful mind for machine. In this work, we propose affective manifolds as components of a machine's mind. Every affective manifold models a characteristic group of mind and contains multiple states. We define the machine's mind as a set of affective manifolds. We use a learning model for mapping the input signals to the embedding space of affective manifold. Using this mapping, a machine or a robot takes an input signal and can react emotionally to it. We use deep metric learning, with Siamese network, and propose a loss function for affective manifold learning. We define margins between states based on the psychological and philosophical studies. Using triplets of instances, we train the network to minimize the variance of every state and have the desired distances between states. We show that affective manifolds can have various applications for machine-machine and human-machine interactions. Some simulations are also provided for verification of the proposed method. It is possible to have as many affective manifolds as required in machine's mind. More affective manifolds in the machine's mind can make it more realistic and effective. This paper opens the door; we invite the researchers from various fields of science to propose more affective manifolds to be inserted in machine's mind.
['Benyamin Ghojogh']
2022-08-29
null
null
null
null
['supervised-dimensionality-reduction']
['computer-vision']
[-2.82343537e-01 4.01490569e-01 3.38406749e-02 -6.19396687e-01 1.87849086e-02 -3.90018255e-01 4.47203636e-01 -3.42539176e-02 -8.22155252e-02 2.56952047e-01 -9.93073825e-03 1.60620481e-01 1.41233811e-02 -7.51809716e-01 -3.79922569e-01 -6.52573884e-01 -1.63840309e-01 2.88147926e-01 -3.55669320e-01 -4.30745602e-01 3.80261332e-01 3.77925158e-01 -1.22110891e+00 -1.18002139e-01 6.99257195e-01 9.11502540e-01 -2.92245355e-02 5.33427596e-01 -1.68455482e-01 7.02347338e-01 -4.09884483e-01 -3.74034762e-01 1.22213915e-01 -7.23122060e-01 -1.03069139e+00 2.44911134e-01 -1.79561451e-02 5.64229846e-01 -7.17621967e-02 1.38017559e+00 1.67065650e-01 3.26052010e-01 1.16987264e+00 -1.71798742e+00 -1.14303315e+00 2.46219918e-01 -5.14464438e-01 -3.19190294e-01 4.56537664e-01 -3.80981714e-02 9.11829650e-01 -8.53209794e-01 2.16598704e-01 1.46090758e+00 3.66458774e-01 7.96950638e-01 -1.08152449e+00 -5.73040307e-01 1.16633894e-02 4.43539321e-01 -1.26759672e+00 -2.16058880e-01 1.20995307e+00 -5.61539233e-01 5.78244150e-01 2.76265562e-01 9.14241076e-01 7.31824756e-01 6.08744621e-01 7.53796041e-01 1.05109346e+00 -5.62884092e-01 4.49044913e-01 4.38810915e-01 3.86776894e-01 9.58793938e-01 -2.74835408e-01 -3.69565159e-01 -2.43287265e-01 1.70241907e-01 5.16818345e-01 1.19555354e-01 3.35904211e-02 -5.14363468e-01 -1.40320826e+00 9.84033406e-01 4.81827915e-01 3.85962158e-01 -2.56207883e-01 8.22937563e-02 1.58709854e-01 5.59380114e-01 2.46047020e-01 7.39336848e-01 -7.48000965e-02 7.39544211e-03 -3.65557373e-01 -3.30283523e-01 7.67351747e-01 6.93118632e-01 1.00588930e+00 -1.10309377e-01 4.89883095e-01 6.91948831e-01 4.48738664e-01 5.08765757e-01 4.55314964e-01 -1.14016747e+00 -7.34886080e-02 9.80160177e-01 -4.93108714e-03 -1.46467042e+00 -7.21073151e-01 1.78111419e-01 -1.12365830e+00 2.55948395e-01 1.66841373e-02 -2.70251065e-01 -3.45040172e-01 1.86964524e+00 4.15284574e-01 2.66671598e-01 2.38186643e-01 1.19247305e+00 4.44844097e-01 7.97693789e-01 -2.74230272e-01 -2.89015949e-01 1.30123854e+00 -5.65892220e-01 -9.53437269e-01 -8.98509696e-02 7.27555990e-01 -6.35026813e-01 1.34636366e+00 3.78759950e-01 -8.89041722e-01 -5.43228328e-01 -1.28077304e+00 1.77516431e-01 -4.29567695e-01 -9.44649056e-03 7.61127591e-01 5.71167648e-01 -1.06696558e+00 6.73611701e-01 -8.00191045e-01 -7.53160477e-01 -1.00722715e-01 6.54102743e-01 -4.68228996e-01 6.31024003e-01 -1.53488517e+00 1.34975529e+00 4.40087914e-02 1.98309466e-01 -5.02766609e-01 3.32959853e-02 -9.74553525e-01 -3.27172011e-01 -2.91830063e-01 -8.94466639e-01 7.95119226e-01 -1.45013332e+00 -1.68730450e+00 1.08735430e+00 2.45193932e-02 -7.98518509e-02 -1.04694016e-01 -2.75599733e-02 -7.14076042e-01 1.14693463e-01 -1.14047974e-01 7.68397152e-01 1.04040277e+00 -1.02073336e+00 -2.12250486e-01 -6.47665560e-01 6.57307357e-02 2.91656733e-01 -6.37615979e-01 -1.42932475e-01 -9.86292511e-02 -1.07836574e-01 2.44388700e-01 -1.29517233e+00 -7.76833817e-02 -3.98563415e-01 -5.93624413e-01 -6.16498113e-01 8.51105213e-01 -6.27113879e-02 8.54494691e-01 -2.24885631e+00 6.35945618e-01 2.29087874e-01 2.39148796e-01 -8.24481621e-02 -1.73681468e-01 3.38477969e-01 -2.93358415e-01 -2.15130523e-02 -4.29492518e-02 -3.96396130e-01 2.34166622e-01 3.68993618e-02 -8.65681171e-02 9.99639213e-01 3.38166893e-01 7.08879054e-01 -7.42292643e-01 -5.93963623e-01 3.33503336e-01 2.98705548e-01 -3.75192076e-01 3.70624840e-01 9.08824205e-02 4.25473124e-01 -4.81536627e-01 2.28235722e-01 3.86164844e-01 -1.38477415e-01 -1.30628228e-01 -2.86116302e-01 3.04422557e-01 -2.15668961e-01 -1.11431277e+00 1.57253921e+00 -4.86462623e-01 6.53846860e-01 2.08542533e-02 -1.27695811e+00 1.35313475e+00 3.17088991e-01 4.96944845e-01 -3.36748391e-01 5.26937544e-01 -1.93912506e-01 1.35316923e-01 -6.95469379e-01 3.77243161e-01 -6.27998233e-01 -3.92683357e-01 7.36083746e-01 1.54237449e-01 -5.06053865e-01 -4.10665840e-01 1.74107447e-01 7.19110012e-01 -2.92301774e-01 8.82644057e-02 -6.01994097e-01 7.93096364e-01 -3.89302909e-01 3.54332477e-01 -2.07076028e-01 -5.21056950e-01 8.24125633e-02 5.49884915e-01 -4.01365787e-01 -8.27708721e-01 -1.28978205e+00 -4.80292216e-02 9.27218199e-01 5.00725627e-01 -4.14160974e-02 -9.56626296e-01 -4.23190653e-01 -1.97126433e-01 6.75443709e-01 -8.84131849e-01 -7.62484074e-01 -1.28948078e-01 -4.90382224e-01 2.05742508e-01 8.96888152e-02 3.22253168e-01 -1.19411218e+00 -3.53257984e-01 -3.41829717e-01 -2.07439110e-01 -7.77298808e-01 -5.58335781e-01 1.75285965e-01 -7.59554267e-01 -9.73876059e-01 -3.57911348e-01 -1.04686546e+00 6.92927599e-01 3.58761698e-02 1.00256646e+00 -2.73934305e-01 -4.92817909e-02 7.65804350e-01 -2.04879969e-01 -4.02115583e-01 -5.10900259e-01 -8.34772587e-02 8.03144693e-01 3.15359503e-01 6.81194723e-01 -6.20915711e-01 -6.17500722e-01 3.00569534e-01 -7.22892046e-01 -6.04529269e-02 1.81569815e-01 6.35013282e-01 1.81511238e-01 -7.02783167e-02 9.32968557e-01 -4.79574412e-01 9.23013210e-01 -5.80158710e-01 -8.30081105e-02 4.29646634e-02 -5.36132157e-01 3.34590793e-01 7.67681301e-01 -4.89764303e-01 -6.52067721e-01 6.27957731e-02 2.92838126e-01 -5.67298472e-01 -3.69994082e-02 1.65167928e-01 -3.13663691e-01 9.52042788e-02 6.75889850e-01 -5.15025817e-02 3.83978516e-01 3.24027948e-02 8.50995302e-01 9.93544281e-01 2.66072214e-01 -4.25273478e-01 5.27210832e-01 3.70183557e-01 1.08954363e-01 -9.97853398e-01 -7.67088175e-01 -2.26475209e-01 -6.89882755e-01 -5.86068988e-01 1.16497934e+00 -4.87503886e-01 -1.21464527e+00 2.95220554e-01 -1.01569891e+00 2.09825858e-02 -6.37701377e-02 5.99290073e-01 -9.65341210e-01 3.96477580e-01 -5.26944578e-01 -8.70038509e-01 -3.62204880e-01 -9.67448115e-01 7.47002363e-01 4.13749635e-01 -6.87216401e-01 -1.47799289e+00 3.46743196e-01 1.41274646e-01 -1.05482548e-01 1.88625306e-01 8.24804127e-01 -5.40679574e-01 -8.64967629e-02 -1.31430387e-01 3.00654203e-01 5.77194035e-01 4.77043658e-01 2.18904853e-01 -9.51423943e-01 -1.77875578e-01 5.80154777e-01 -4.59872574e-01 5.03427684e-01 1.03895009e-01 7.94877708e-01 -2.10044563e-01 -1.88795105e-01 4.58510160e-01 1.05221081e+00 5.40508687e-01 6.57596588e-01 2.77372360e-01 6.52298868e-01 9.99537170e-01 6.62741005e-01 2.33820200e-01 6.09341204e-01 2.13972762e-01 5.60110390e-01 -2.18830600e-01 6.94598079e-01 3.07732224e-02 6.99149489e-01 1.46881998e+00 2.56580234e-01 3.39248985e-01 -7.66981602e-01 3.91604394e-01 -1.74862289e+00 -9.64259624e-01 2.50204448e-02 2.01202512e+00 7.33398199e-01 -9.28586498e-02 8.71749148e-02 9.14827809e-02 8.95521700e-01 -1.36746749e-01 -6.62524104e-01 -1.07059848e+00 6.15658313e-02 -4.49097604e-02 -2.39953503e-01 6.01627290e-01 -1.21678853e+00 9.94952738e-01 6.18776321e+00 2.96538621e-01 -1.41893947e+00 -2.24337667e-01 7.91208327e-01 3.25590707e-02 -2.63480663e-01 -1.54453844e-01 -3.73705596e-01 3.36971253e-01 1.01035678e+00 -3.33410650e-01 5.12109280e-01 6.95323467e-01 2.63030916e-01 -9.68925655e-02 -1.45673394e+00 1.34886789e+00 7.84147754e-02 -7.48281658e-01 -2.58109152e-01 -2.48673216e-01 5.96336126e-01 -4.27002400e-01 4.66971159e-01 2.08350167e-01 2.45681882e-01 -1.32179070e+00 3.86622488e-01 9.16785598e-01 5.32471597e-01 -9.79557276e-01 5.37667871e-01 5.58711410e-01 -9.76611555e-01 1.79778576e-01 -5.03749490e-01 -4.26863164e-01 -3.42234075e-01 3.31014454e-01 -6.97683394e-01 1.88870415e-01 3.21300417e-01 8.41212928e-01 -4.03234750e-01 2.32452810e-01 1.26027584e-01 2.34386325e-01 -1.08036563e-01 -6.76872969e-01 1.76923603e-01 -8.86655509e-01 3.57815593e-01 9.35950994e-01 3.42104584e-01 2.22301826e-01 -1.34546772e-01 1.04800522e+00 5.39248250e-02 4.83069867e-01 -9.99224722e-01 -1.23496048e-01 2.62937307e-01 1.63960218e+00 -6.36310697e-01 -2.33051866e-01 -2.06420243e-01 1.32216728e+00 3.66126835e-01 8.74570608e-02 -9.24346626e-01 -5.87517440e-01 9.63646770e-01 -3.15471500e-01 -6.25916362e-01 -1.62055433e-01 -3.51256013e-01 -1.18893766e+00 -3.48079711e-01 -6.70148492e-01 7.05464929e-02 -8.00453901e-01 -1.45224977e+00 6.05855942e-01 -4.09942359e-01 -1.21642840e+00 -2.13110164e-01 -3.98873985e-01 -8.86703432e-01 6.66276276e-01 -5.42460263e-01 -7.88429856e-01 -8.12862888e-02 9.50848162e-01 2.02253666e-02 -3.89388025e-01 9.80969906e-01 -1.66390151e-01 -5.55715144e-01 3.97792011e-01 -1.27867669e-01 1.97690278e-01 8.02930772e-01 -1.48326957e+00 5.80166243e-02 3.51814538e-01 2.40477040e-01 9.05293703e-01 8.85653436e-01 -5.93336113e-02 -1.72552812e+00 -9.49411571e-01 5.82204640e-01 -6.09662235e-01 8.72417390e-01 -3.65055978e-01 -7.15269327e-01 7.44488597e-01 6.17255270e-01 -4.89353091e-01 9.28062320e-01 4.54550296e-01 2.10474655e-02 -3.10173362e-01 -1.17323840e+00 9.18616951e-01 5.40114999e-01 -7.51457572e-01 -9.83369410e-01 3.75122696e-01 6.88866973e-01 2.94658840e-01 -9.54532623e-01 9.30231586e-02 2.70574570e-01 -1.10885727e+00 6.15969539e-01 -6.38728738e-01 2.07949683e-01 -3.22123140e-01 -9.46954042e-02 -1.86432183e+00 -3.76271129e-01 -7.52845764e-01 1.76157206e-01 1.07452726e+00 5.61611503e-02 -7.93061495e-01 6.74031794e-01 6.69435203e-01 9.52247232e-02 -7.90784717e-01 -8.46980333e-01 -5.74080706e-01 4.77466375e-01 -3.35400820e-01 4.11349028e-01 1.31354034e+00 1.04207420e+00 1.04135275e+00 -2.35635892e-01 4.81913164e-02 5.75417876e-01 2.75019705e-01 8.02400053e-01 -1.24273729e+00 1.13966636e-01 -4.61641103e-01 -8.03456962e-01 -6.81794345e-01 7.08960831e-01 -1.03602612e+00 -1.71039388e-01 -1.31056285e+00 2.19843909e-01 -3.52416754e-01 -4.05359536e-01 9.26258191e-02 -2.36654542e-02 -4.00915779e-02 1.85350075e-01 7.89105445e-02 -6.55634880e-01 8.88794482e-01 1.54568815e+00 -1.57442808e-01 -2.36300498e-01 -7.96227083e-02 -6.32609248e-01 1.00943482e+00 1.14438713e+00 -1.97765585e-02 -4.76902157e-01 2.31378138e-01 3.74797523e-01 1.79361314e-01 6.79785535e-02 -1.01006186e+00 2.18400344e-01 -2.92342335e-01 3.59771460e-01 -2.46444449e-01 6.97033763e-01 -9.31202412e-01 1.60616845e-01 4.59403366e-01 -2.15347096e-01 1.86945111e-01 -2.22949803e-01 3.01055670e-01 -3.53697568e-01 -6.60579428e-02 1.25993741e+00 -5.63124605e-02 -4.79993075e-01 1.58865660e-01 -3.78450185e-01 3.58032063e-02 1.56260014e+00 -1.05643690e-01 -6.28569126e-02 -7.01809704e-01 -9.87501562e-01 3.67709100e-01 6.25570536e-01 5.16735196e-01 7.89480507e-01 -1.67493916e+00 -2.28213847e-01 2.23702416e-01 -3.67480936e-03 -5.86925805e-01 1.36232316e-01 9.19378459e-01 -4.11759675e-01 2.83822060e-01 -5.08701324e-01 -6.62290514e-01 -1.12384033e+00 8.85207772e-01 5.88905156e-01 4.61065859e-01 -1.52783975e-01 6.78338706e-01 2.97730476e-01 -6.95690155e-01 -1.05288578e-02 -2.26363808e-01 -4.10454810e-01 2.59531647e-01 1.51995480e-01 3.39255452e-01 -5.04419863e-01 -9.61408854e-01 -5.05248904e-01 8.09405327e-01 4.16431487e-01 -2.96170652e-01 9.75744724e-01 -2.32644081e-01 -4.88906413e-01 1.21304691e+00 1.67535079e+00 -1.84338823e-01 -6.73818886e-01 1.03100061e-01 -4.28461768e-02 2.18168199e-02 -2.76184916e-01 -1.40509382e-01 -9.08505917e-01 1.26688313e+00 8.69709492e-01 6.09646797e-01 1.18392968e+00 1.62898853e-01 6.86670065e-01 4.84182864e-01 3.80949616e-01 -1.46008337e+00 5.85795701e-01 5.98689437e-01 7.56737888e-01 -1.30396068e+00 -4.76558775e-01 2.83487737e-02 -9.98485327e-01 1.15246785e+00 6.08058333e-01 -5.04419923e-01 9.60350573e-01 -3.34360898e-02 4.43873644e-01 -3.04240942e-01 -8.44195545e-01 5.47287278e-02 2.63136387e-01 5.61281323e-01 5.24739206e-01 5.32091260e-01 3.33927199e-02 5.75046122e-01 -5.24132371e-01 -3.33846658e-01 6.19925141e-01 2.67546713e-01 -8.80423546e-01 -8.63339782e-01 -5.98315656e-01 1.72586471e-01 -2.23669738e-01 3.01193982e-01 -5.69839239e-01 5.22672534e-01 3.39008011e-02 1.27916396e+00 3.19979966e-01 -9.08301771e-01 1.58523366e-01 2.24307105e-01 5.73133469e-01 -5.35715520e-01 -1.75197363e-01 -2.33185604e-01 -3.59354585e-01 -5.11060476e-01 -5.36629498e-01 -5.94088972e-01 -1.70745099e+00 -5.82108140e-01 -1.55723706e-01 4.65508342e-01 5.08425772e-01 7.31785357e-01 2.51057148e-01 1.69635952e-01 1.09116006e+00 -5.41283190e-01 -4.26134735e-01 -1.09987342e+00 -9.26429331e-01 6.11680090e-01 1.75918296e-01 -8.11603546e-01 -5.14062405e-01 1.30282015e-01]
[13.006738662719727, 5.694408893585205]
1c2e44eb-7b63-4c91-b114-4cd4aeebb9cd
recurrent-pixel-embedding-for-instance
1712.08273
null
http://arxiv.org/abs/1712.08273v1
http://arxiv.org/pdf/1712.08273v1.pdf
Recurrent Pixel Embedding for Instance Grouping
We introduce a differentiable, end-to-end trainable framework for solving pixel-level grouping problems such as instance segmentation consisting of two novel components. First, we regress pixels into a hyper-spherical embedding space so that pixels from the same group have high cosine similarity while those from different groups have similarity below a specified margin. We analyze the choice of embedding dimension and margin, relating them to theoretical results on the problem of distributing points uniformly on the sphere. Second, to group instances, we utilize a variant of mean-shift clustering, implemented as a recurrent neural network parameterized by kernel bandwidth. This recurrent grouping module is differentiable, enjoys convergent dynamics and probabilistic interpretability. Backpropagating the group-weighted loss through this module allows learning to focus on only correcting embedding errors that won't be resolved during subsequent clustering. Our framework, while conceptually simple and theoretically abundant, is also practically effective and computationally efficient. We demonstrate substantial improvements over state-of-the-art instance segmentation for object proposal generation, as well as demonstrating the benefits of grouping loss for classification tasks such as boundary detection and semantic segmentation.
['Charless Fowlkes', 'Shu Kong']
2017-12-22
recurrent-pixel-embedding-for-instance-1
http://openaccess.thecvf.com/content_cvpr_2018/html/Kong_Recurrent_Pixel_Embedding_CVPR_2018_paper.html
http://openaccess.thecvf.com/content_cvpr_2018/papers/Kong_Recurrent_Pixel_Embedding_CVPR_2018_paper.pdf
cvpr-2018-6
['object-proposal-generation']
['computer-vision']
[ 4.19873625e-01 7.53404856e-01 -7.93487132e-02 -4.30489093e-01 -1.13359177e+00 -4.06359076e-01 3.07338417e-01 2.75819421e-01 -5.99257529e-01 3.79637510e-01 -1.38602003e-01 -1.62182540e-01 -2.34580651e-01 -6.49480462e-01 -9.41221774e-01 -9.49306726e-01 -1.31290555e-01 6.23240232e-01 4.15852547e-01 3.05198699e-01 5.26416898e-01 5.15233397e-01 -1.50368917e+00 -2.22642362e-01 1.09144580e+00 8.46851051e-01 2.11644292e-01 6.21296167e-01 5.73255681e-02 2.78984964e-01 -4.82308537e-01 -3.15286130e-01 2.74948537e-01 -3.61542702e-01 -7.35565722e-01 4.34880346e-01 5.57412088e-01 7.26088956e-02 -8.89569595e-02 9.61847544e-01 4.33699429e-01 5.48430085e-01 9.91321146e-01 -1.11082506e+00 -8.78514886e-01 5.90306461e-01 -6.56774163e-01 -1.00472383e-01 -2.46750459e-01 4.14176434e-02 1.45376658e+00 -7.92847931e-01 5.53650618e-01 1.04418373e+00 7.13091791e-01 7.09809303e-01 -1.61270487e+00 -2.58794904e-01 4.89821434e-01 -1.48340479e-01 -1.06043661e+00 -1.62046865e-01 6.50486469e-01 -5.68205535e-01 7.56202579e-01 2.02432930e-01 5.56956053e-01 4.46300238e-01 -1.22463018e-01 1.03624868e+00 4.99006540e-01 -4.99065250e-01 5.05931079e-01 1.76878065e-01 3.44646275e-01 8.54974747e-01 2.45382696e-01 -3.50911617e-01 -1.82445228e-01 2.52628773e-02 7.57897973e-01 -1.41761482e-01 -1.49328560e-01 -8.39151144e-01 -1.20608866e+00 1.11115682e+00 6.99967384e-01 -3.07458695e-02 -2.80187219e-01 5.08569062e-01 1.46510407e-01 -1.49603561e-01 7.26300001e-01 6.78516209e-01 -2.20122024e-01 2.08425745e-01 -9.99535799e-01 3.34195882e-01 7.24248052e-01 8.32124591e-01 8.32379043e-01 -2.27833837e-01 -1.32535785e-01 8.88914347e-01 5.83085418e-01 3.69553491e-02 3.66887331e-01 -1.35099185e+00 3.00623119e-01 5.13849437e-01 -4.01821546e-02 -7.80734241e-01 -4.46814567e-01 -2.69383192e-01 -4.78227645e-01 4.68565017e-01 5.49770176e-01 -1.45864516e-01 -1.00467527e+00 1.83216679e+00 4.27068919e-01 4.13118303e-01 -1.39628783e-01 9.52900708e-01 3.67532700e-01 5.29716671e-01 1.33575425e-01 1.32217631e-01 1.20375288e+00 -1.10774624e+00 -1.71826348e-01 -8.51196647e-02 6.68314278e-01 -5.55698097e-01 1.11623275e+00 9.17411745e-02 -1.37863779e+00 -2.53340662e-01 -1.03152966e+00 -3.52562428e-01 -2.04918981e-01 1.53455555e-01 7.46298254e-01 5.42110682e-01 -1.31615531e+00 1.07828844e+00 -1.14196384e+00 -1.92622349e-01 8.54957700e-01 4.51923102e-01 -3.27328891e-02 4.65105444e-01 -7.10456252e-01 5.41100085e-01 1.88002929e-01 -4.49191481e-02 -4.36078608e-01 -9.94315565e-01 -9.25143719e-01 2.43895575e-02 1.52701259e-01 -8.67791772e-01 1.15507030e+00 -9.92591918e-01 -1.57996595e+00 1.05274355e+00 -3.23581100e-01 -7.76824832e-01 6.83800876e-01 -2.05202475e-01 3.48568767e-01 2.83108860e-01 4.01524961e-01 1.08282578e+00 9.52750087e-01 -1.28134060e+00 -6.91035390e-01 -4.80382472e-01 -1.93091422e-01 4.82780218e-01 -1.89207286e-01 -2.60421932e-01 -5.29711664e-01 -6.95163190e-01 6.82974339e-01 -1.03007007e+00 -6.31562114e-01 5.08341372e-01 -5.34458399e-01 -5.06969035e-01 5.10360241e-01 -5.98328888e-01 8.01851034e-01 -2.06560016e+00 2.97745973e-01 3.97162467e-01 2.23091036e-01 -8.40580687e-02 8.81180242e-02 -4.51626405e-02 -4.54998612e-02 3.88124198e-01 -7.38188624e-01 -6.68950260e-01 2.10574776e-01 1.50031764e-02 -4.08179998e-01 7.61178672e-01 5.11001229e-01 9.25170541e-01 -8.45776618e-01 -3.60549957e-01 3.30146611e-01 3.28283072e-01 -8.53864849e-01 4.04657274e-02 -2.54662573e-01 1.48749799e-01 -5.04094839e-01 3.48539829e-01 4.63266939e-01 -4.87834364e-01 -3.36339980e-01 -2.08786968e-02 -3.50376479e-02 3.21106851e-01 -1.32858610e+00 1.72192693e+00 -2.86226332e-01 7.00589180e-01 -5.15769050e-02 -1.28691149e+00 6.52152836e-01 -1.15856394e-01 5.10791600e-01 -6.82629496e-02 -1.72514156e-01 3.41821648e-02 -2.57787287e-01 -2.62642473e-01 6.00892425e-01 7.32253194e-02 1.24921143e-01 6.42847657e-01 1.48357917e-02 -4.02094543e-01 3.49517278e-02 2.45040312e-01 8.04486930e-01 1.62380770e-01 -5.82936294e-02 -2.12416694e-01 1.49068579e-01 -3.55242938e-02 4.35472220e-01 8.05100262e-01 -1.49283871e-01 1.02098584e+00 7.08000898e-01 1.06899105e-01 -1.20378256e+00 -1.43919027e+00 -4.36791599e-01 1.05445969e+00 3.95441473e-01 1.38781890e-01 -1.02746713e+00 -6.02959216e-01 2.62541771e-01 7.61537015e-01 -6.76006973e-01 -1.01945885e-01 -6.31102979e-01 -7.61406839e-01 2.70107478e-01 6.02904737e-01 2.79903919e-01 -1.04156506e+00 -6.05529785e-01 3.31051648e-01 1.84747726e-01 -6.64786875e-01 -6.88156486e-01 2.73556530e-01 -1.06936467e+00 -1.05408883e+00 -9.68037486e-01 -1.09275484e+00 1.05742610e+00 2.83085465e-01 9.91134226e-01 -8.78924131e-02 -5.03656983e-01 6.15382731e-01 -6.79046735e-02 -4.97876741e-02 -6.93988428e-02 2.71013349e-01 -2.63335913e-01 -2.15230566e-02 2.45794356e-01 -3.13653350e-01 -8.01097810e-01 2.50034064e-01 -6.32241726e-01 -8.12773257e-02 3.26692551e-01 8.23554873e-01 9.15881395e-01 -2.01993927e-01 3.66722852e-01 -7.89318562e-01 4.87242937e-01 -3.30525994e-01 -6.58190787e-01 2.29012564e-01 -5.74235976e-01 2.35557452e-01 2.63442814e-01 -3.78228426e-01 -7.68144488e-01 1.12398848e-01 -2.79437732e-02 -1.89248055e-01 -7.45453760e-02 -5.24552055e-02 7.52550513e-02 -1.34983286e-01 5.17001092e-01 -6.08077794e-02 2.29275286e-01 -2.10592762e-01 7.64273465e-01 4.56014782e-01 5.48197508e-01 -6.61205888e-01 6.90497816e-01 7.98780918e-01 -1.55987844e-01 -9.27038789e-01 -7.36227930e-01 -7.17112958e-01 -7.09565043e-01 3.56408954e-02 1.20842087e+00 -7.00713158e-01 -7.71135569e-01 3.43739688e-01 -1.18129516e+00 -5.15767395e-01 -7.17270195e-01 3.81081820e-01 -9.94850039e-01 3.14180106e-01 -5.77425718e-01 -8.25070918e-01 -2.44250298e-01 -1.06928670e+00 1.33302689e+00 4.24573630e-01 -1.58643886e-01 -1.14104748e+00 -1.51926830e-01 3.27893168e-01 -9.88268927e-02 1.58135608e-01 8.72568727e-01 -7.40772784e-01 -8.75756383e-01 -3.75769921e-02 -2.98989803e-01 4.12715942e-01 7.66156465e-02 2.14524433e-01 -8.67164433e-01 -1.14227526e-01 -1.13855250e-01 -1.76259473e-01 1.28408408e+00 8.82925808e-01 1.43308771e+00 -9.64083970e-02 -5.34362853e-01 7.51986802e-01 1.14150143e+00 -1.64326608e-01 5.00519037e-01 3.01174521e-01 7.98103631e-01 8.81054044e-01 4.81020659e-01 1.49018601e-01 4.23065186e-01 5.47085464e-01 3.38117808e-01 -1.94528177e-01 -2.61716880e-02 -1.56613439e-01 -8.14442616e-03 5.11094630e-01 2.30425671e-01 -4.72424887e-02 -6.65780842e-01 7.20939457e-01 -2.02005982e+00 -9.34271753e-01 5.11075445e-02 2.34051752e+00 8.63569319e-01 3.02884728e-01 2.05201745e-01 -9.20063034e-02 9.43829596e-01 9.48780477e-02 -8.38437259e-01 -1.75801948e-01 -2.27049254e-02 2.00228602e-01 7.47459650e-01 7.32027292e-01 -1.39619827e+00 1.18670166e+00 6.04988480e+00 7.62391508e-01 -8.06447327e-01 -2.73260027e-02 9.66324687e-01 -2.01302581e-02 -5.64027488e-01 -4.34608050e-02 -8.94841135e-01 4.48887736e-01 4.41466451e-01 2.57965960e-02 4.05963510e-01 9.01787460e-01 2.44903773e-01 -1.19801991e-01 -1.22120452e+00 6.26936615e-01 1.29226297e-01 -1.50373137e+00 -8.49910304e-02 -3.80033068e-02 8.61766338e-01 5.44380508e-02 4.34300452e-01 3.20862159e-02 4.99281883e-01 -1.00877619e+00 7.11305976e-01 3.40032339e-01 3.86668116e-01 -6.73832715e-01 2.38697469e-01 3.29480357e-02 -9.23695982e-01 1.19936876e-01 -5.46272457e-01 3.84733766e-01 2.65087545e-01 7.26309180e-01 -8.94194007e-01 1.07966170e-01 5.03489673e-01 6.95358634e-01 -3.48557830e-01 1.22395337e+00 -3.18610489e-01 5.10512769e-01 -5.39898396e-01 -1.94083676e-02 3.74446303e-01 -6.82471097e-01 5.46664774e-01 1.10510421e+00 1.72450796e-01 -2.99981982e-02 1.11110397e-02 1.30670607e+00 -2.00799584e-01 -2.04700939e-02 -3.75413686e-01 1.63676620e-01 5.73370218e-01 1.12082815e+00 -1.11268711e+00 -3.00211668e-01 -1.31385922e-01 1.09186709e+00 4.14921999e-01 5.05547225e-01 -9.30946112e-01 -6.08349323e-01 9.09468412e-01 3.71039622e-02 6.19862378e-01 -1.92826554e-01 -8.93122256e-01 -8.43570650e-01 1.83723927e-01 -2.07387298e-01 2.04641506e-01 -4.88406688e-01 -1.32481968e+00 1.03235699e-01 -2.29256243e-01 -9.47168171e-01 -1.04498208e-01 -6.39507413e-01 -7.60762691e-01 8.04663777e-01 -1.47662282e+00 -9.48519588e-01 -1.31474668e-03 2.07321197e-01 7.27191269e-01 1.49982706e-01 5.12884438e-01 -6.59480989e-02 -6.10243976e-01 6.42277718e-01 1.87676743e-01 1.09556235e-01 5.13857782e-01 -1.79866743e+00 6.09995484e-01 7.04588890e-01 3.43334168e-01 6.71927989e-01 5.93695700e-01 -4.50081766e-01 -6.96779251e-01 -1.29968798e+00 6.49760306e-01 -4.25706387e-01 7.27264941e-01 -5.21566510e-01 -1.14142001e+00 7.48876214e-01 -2.59718388e-01 5.87286577e-02 4.83355671e-01 7.67133832e-02 -2.36671910e-01 1.54921144e-01 -1.19385421e+00 9.62393641e-01 1.01268208e+00 -3.71959269e-01 -6.98777676e-01 6.10639334e-01 9.18873727e-01 -4.00409281e-01 -6.94336236e-01 1.06122449e-01 2.74965286e-01 -7.71357775e-01 1.21924722e+00 -6.17853761e-01 4.22080547e-01 -2.63691276e-01 5.59629686e-02 -1.27513623e+00 -2.24108890e-01 -7.90442646e-01 1.24724574e-01 1.08749211e+00 6.53692424e-01 -7.77808249e-01 1.26558065e+00 9.19965863e-01 -2.94326633e-01 -1.03355110e+00 -1.01831579e+00 -5.69370508e-01 4.36203450e-01 -3.70658249e-01 1.29637003e-01 5.56818068e-01 -1.68357000e-01 -7.84145072e-02 3.12512189e-01 3.82703006e-01 9.91435945e-01 1.02727927e-01 5.39903283e-01 -1.20234692e+00 -3.37069750e-01 -7.48702943e-01 -5.34906983e-01 -1.46025050e+00 3.77477944e-01 -9.88122582e-01 2.45049611e-01 -1.65452588e+00 -5.56884706e-02 -8.19458961e-01 -2.07290351e-01 2.98838466e-01 -3.21166247e-01 2.08316371e-01 1.16789281e-01 2.77952492e-01 -5.28132021e-01 5.66435695e-01 1.00459766e+00 -1.12978727e-01 -5.37057281e-01 1.83879197e-01 -6.71894968e-01 8.69947851e-01 6.30240858e-01 -4.58197474e-01 -2.54770547e-01 -2.65667975e-01 -8.48007053e-02 -4.17581320e-01 4.70913351e-01 -7.95739293e-01 3.95848423e-01 1.33853555e-01 1.24210589e-01 -5.00572801e-01 3.04350436e-01 -5.39287090e-01 -3.93296242e-01 2.95667082e-01 -7.31723964e-01 -5.08531094e-01 7.71367252e-02 7.13180363e-01 1.33815497e-01 -5.55740416e-01 1.05703747e+00 7.10677430e-02 -6.12141967e-01 3.04989725e-01 -1.55176446e-01 3.69654745e-01 1.36640310e+00 -5.95445633e-01 -2.94605307e-02 -1.39224276e-01 -8.60246897e-01 3.39745283e-01 5.46632290e-01 1.59046561e-01 5.70132256e-01 -1.07578254e+00 -5.58753431e-01 1.48453295e-01 -1.09306082e-01 5.43808043e-01 5.95403425e-02 7.60413587e-01 -5.72712123e-01 1.33137718e-01 3.33091050e-01 -9.10008729e-01 -1.06342924e+00 1.93093657e-01 4.46053684e-01 2.15355307e-02 -9.92669404e-01 1.31987345e+00 3.92614901e-01 -6.72815144e-01 4.05064076e-01 -4.33196396e-01 1.40155315e-01 1.45818532e-01 1.77360311e-01 4.01205182e-01 -1.57355696e-01 -1.63520321e-01 -2.02859059e-01 8.85703623e-01 -1.90268904e-01 -3.44932824e-01 1.31615400e+00 -9.59662572e-02 2.91995481e-02 5.92473209e-01 1.29738760e+00 -3.31433833e-01 -1.82354796e+00 -2.66079064e-02 1.32952556e-01 -2.54163116e-01 -1.59372509e-01 -3.91175330e-01 -9.05278981e-01 8.02365839e-01 4.25713807e-01 3.14531326e-01 5.85295677e-01 4.35959637e-01 6.87404633e-01 2.30134368e-01 3.24480832e-02 -1.47015083e+00 2.49118015e-01 2.91892290e-01 6.13716185e-01 -1.26582670e+00 -3.04987747e-02 -4.83618259e-01 -6.64530993e-01 1.03612840e+00 3.47367287e-01 -7.30683208e-01 6.40658915e-01 -9.82010663e-02 -9.06850025e-02 -1.46084398e-01 -3.72834086e-01 -1.77280992e-01 3.83574665e-01 6.48413718e-01 2.97191232e-01 -4.87211719e-02 -9.85555798e-02 1.37630701e-01 -2.04421103e-01 -4.90662247e-01 3.95879835e-01 4.49968576e-01 -7.46199727e-01 -6.10131800e-01 -1.45419046e-01 6.68914676e-01 -2.93382764e-01 -5.31892292e-02 -7.34789297e-02 8.13140213e-01 -1.58578366e-01 6.51693463e-01 6.62616432e-01 2.72339821e-01 1.43775806e-01 -4.72283624e-02 4.00368214e-01 -8.31220388e-01 -1.95479512e-01 -7.34162459e-04 -3.36998373e-01 -5.77481389e-01 -2.22579241e-01 -9.54764605e-01 -1.70302415e+00 1.63278311e-01 -4.50411081e-01 3.85817885e-02 8.57793987e-01 7.89089978e-01 2.60289073e-01 6.12812877e-01 6.60243154e-01 -1.14374161e+00 -9.40995693e-01 -5.01369655e-01 -6.18854344e-01 4.69146222e-01 3.81125331e-01 -4.58228618e-01 -7.81998992e-01 8.87099653e-02]
[9.539765357971191, 0.4792215824127197]
f1135f64-82d3-4ea9-9871-76e77cee7f88
learn-to-resolve-conversational-dependency-a
2106.11575
null
https://arxiv.org/abs/2106.11575v1
https://arxiv.org/pdf/2106.11575v1.pdf
Learn to Resolve Conversational Dependency: A Consistency Training Framework for Conversational Question Answering
One of the main challenges in conversational question answering (CQA) is to resolve the conversational dependency, such as anaphora and ellipsis. However, existing approaches do not explicitly train QA models on how to resolve the dependency, and thus these models are limited in understanding human dialogues. In this paper, we propose a novel framework, ExCorD (Explicit guidance on how to resolve Conversational Dependency) to enhance the abilities of QA models in comprehending conversational context. ExCorD first generates self-contained questions that can be understood without the conversation history, then trains a QA model with the pairs of original and self-contained questions using a consistency-based regularizer. In our experiments, we demonstrate that ExCorD significantly improves the QA models' performance by up to 1.2 F1 on QuAC, and 5.2 F1 on CANARD, while addressing the limitations of the existing approaches.
['Jaewoo Kang', 'Jungsoo Park', 'Hyunjae Kim', 'Gangwoo Kim']
2021-06-22
null
https://aclanthology.org/2021.acl-long.478
https://aclanthology.org/2021.acl-long.478.pdf
acl-2021-5
['question-rewriting']
['natural-language-processing']
[-1.06928078e-02 5.50534070e-01 1.70971602e-01 -6.13981903e-01 -9.65893209e-01 -7.80282855e-01 5.95776916e-01 8.90491158e-02 -2.37239569e-01 1.04360878e+00 6.82207406e-01 -4.98027653e-01 6.29367074e-04 -7.72554159e-01 -2.53428191e-01 -1.19209029e-01 3.29251498e-01 8.63953471e-01 3.58473241e-01 -1.00992322e+00 1.68598697e-01 -8.73935819e-02 -1.20061040e+00 7.22999871e-01 1.39871573e+00 3.28319430e-01 1.80219516e-01 8.31813514e-01 -7.26361752e-01 1.23984456e+00 -1.12594712e+00 -6.73062742e-01 -4.30388153e-01 -1.01578259e+00 -1.86192858e+00 -1.77603766e-01 1.21988162e-01 -4.79452878e-01 -2.24005029e-01 8.34521472e-01 3.08644056e-01 7.90209398e-02 3.77924442e-01 -1.11929131e+00 -3.48421484e-01 8.22053790e-01 1.61493450e-01 3.83988529e-01 1.04600620e+00 7.22594485e-02 1.31155646e+00 -2.30458483e-01 6.30276918e-01 1.71389055e+00 4.56684351e-01 1.09064496e+00 -8.28382254e-01 -2.29612812e-01 4.01125215e-02 5.20857096e-01 -6.95168436e-01 -5.53502500e-01 5.67610264e-01 -4.92449254e-02 1.22197723e+00 5.45327127e-01 2.58643448e-01 8.52617085e-01 1.07662603e-01 7.84577429e-01 1.11816967e+00 -6.26506031e-01 8.93801376e-02 -3.64252441e-02 6.81285977e-01 3.83398354e-01 -2.84448653e-01 -4.43664521e-01 -4.74421293e-01 -4.01741445e-01 3.36219162e-01 -7.73948967e-01 -3.29087228e-01 3.67963046e-01 -6.96334004e-01 1.20737433e+00 1.66728452e-01 3.42271507e-01 -2.16171473e-01 -3.97772700e-01 3.92892033e-01 6.16546690e-01 -1.25025283e-03 9.50863898e-01 -4.32884961e-01 -5.18413901e-01 -5.61269596e-02 5.91024160e-01 1.41942465e+00 7.20461309e-01 4.91186261e-01 -5.41696250e-01 -2.47188017e-01 9.90879357e-01 1.62016362e-01 4.31378961e-01 4.19304103e-01 -1.49889624e+00 7.36261666e-01 7.48840094e-01 4.69009221e-01 -8.41292262e-01 -5.33296585e-01 2.33822301e-01 -1.55651748e-01 -4.64980423e-01 8.55585039e-01 -4.40100729e-01 -2.28895634e-01 1.92654705e+00 5.20801127e-01 -1.85814887e-01 8.35832119e-01 8.70893002e-01 1.30715430e+00 7.90471673e-01 1.77998513e-01 -4.21309024e-01 1.87752855e+00 -1.17637980e+00 -1.30071235e+00 -4.72246617e-01 8.98463726e-01 -8.53577971e-01 1.25034142e+00 3.05861011e-02 -1.28536355e+00 -1.98732466e-01 -7.47703612e-01 -3.83973688e-01 2.70480305e-01 -4.15041775e-01 5.43069661e-01 4.25538719e-01 -7.79159307e-01 2.82830745e-01 -6.10474348e-01 -3.75318289e-01 -2.48788878e-01 8.90556425e-02 -1.85862407e-01 -2.58950979e-01 -1.91428125e+00 1.35121357e+00 1.73797563e-01 -4.76227514e-02 -4.74277973e-01 -3.46118301e-01 -1.11595440e+00 1.99611545e-01 5.54807067e-01 -6.95736051e-01 2.03008914e+00 -6.16582990e-01 -1.85943103e+00 6.94228411e-01 -6.26664400e-01 -6.00604594e-01 2.42402509e-01 -5.12245953e-01 -4.21512753e-01 3.77690732e-01 3.78584743e-01 5.38211823e-01 1.64078444e-01 -1.04060435e+00 -7.45785773e-01 -2.83135504e-01 9.65320170e-01 7.76089907e-01 5.59609793e-02 3.06467116e-01 -2.79992282e-01 8.65900367e-02 5.60277179e-02 -8.79810393e-01 -4.99985553e-02 -5.95086753e-01 -2.99748451e-01 -8.33448648e-01 7.41781116e-01 -8.46416652e-01 1.38538206e+00 -1.76406467e+00 2.66050816e-01 -4.86864954e-01 -5.34961596e-02 4.46825176e-01 -3.09067667e-01 7.94404089e-01 2.97252715e-01 -5.18324673e-02 -4.05033082e-01 -1.85795754e-01 1.80392005e-02 6.90944552e-01 -4.63303924e-01 -2.59429038e-01 4.87748265e-01 8.25072110e-01 -9.85103488e-01 -4.91306871e-01 -5.25877476e-02 9.66298506e-02 -5.79278469e-01 9.08714831e-01 -7.03515291e-01 6.16500437e-01 -5.14685333e-01 6.08778633e-02 5.53382099e-01 -1.92579567e-01 7.53273368e-01 1.50791302e-01 5.25073558e-02 1.36911869e+00 -9.32113409e-01 1.61961293e+00 -5.83916306e-01 4.59922254e-01 3.64602923e-01 -6.75148487e-01 7.40195453e-01 5.56824744e-01 -1.97359428e-01 -8.29090178e-01 -5.80900302e-03 7.97365457e-02 3.32512438e-01 -8.74460757e-01 5.18790543e-01 -3.29423308e-01 -2.84564734e-01 8.89805615e-01 -1.16513059e-01 -4.21001196e-01 3.72338772e-01 5.39469063e-01 1.05642366e+00 -2.74375379e-01 3.95366192e-01 -9.55346152e-02 9.80228424e-01 4.71630067e-01 5.47212899e-01 5.72707117e-01 -3.55432570e-01 3.12506437e-01 9.24424767e-01 -3.82984519e-01 -4.29117322e-01 -8.02550256e-01 -1.11050270e-01 1.22322619e+00 2.62642294e-01 -7.43802726e-01 -1.12936068e+00 -8.05340588e-01 -3.73147011e-01 1.22097659e+00 -1.93771988e-01 5.95175214e-02 -1.14363050e+00 -5.14482439e-01 7.45172858e-01 2.29058534e-01 7.58758128e-01 -1.50877714e+00 -5.76517642e-01 4.44568276e-01 -1.16976249e+00 -1.27643228e+00 -2.07129031e-01 -2.97243237e-01 -6.40944123e-01 -1.47534144e+00 -3.72050852e-02 -7.62462080e-01 1.57899499e-01 2.86091566e-01 1.33871555e+00 4.61969703e-01 2.46371746e-01 2.53402352e-01 -5.05849063e-01 -1.98957995e-01 -9.38676417e-01 1.56826749e-01 -3.48151624e-01 -4.63220924e-01 5.45960605e-01 -3.42596650e-01 -3.60264659e-01 5.28567135e-01 -6.90607965e-01 6.57941997e-02 -1.23615809e-01 9.28858399e-01 -2.70147622e-01 -2.72521019e-01 9.77182925e-01 -1.30413783e+00 1.25359750e+00 -3.48568469e-01 -1.84989512e-01 3.14062566e-01 -2.81007439e-01 2.22438395e-01 6.68837070e-01 -3.95455211e-02 -1.59024775e+00 -4.55535799e-01 -5.70186973e-01 5.85815489e-01 -2.56766617e-01 6.12119377e-01 -2.91910499e-01 3.00483733e-01 7.68494487e-01 -7.03464374e-02 1.45291194e-01 -3.66215050e-01 5.02354860e-01 9.05716121e-01 6.68790281e-01 -1.06309569e+00 3.05954427e-01 4.38630804e-02 -5.78499913e-01 -6.71043754e-01 -1.36808336e+00 -5.87692440e-01 -4.73971158e-01 8.23572651e-02 8.92907381e-01 -7.55258262e-01 -8.81879628e-01 2.95587331e-01 -1.64812839e+00 -2.63900131e-01 -5.81340902e-02 2.52048790e-01 -5.30706644e-01 8.03350687e-01 -1.16252792e+00 -8.24020207e-01 -6.01079106e-01 -9.48421061e-01 5.19388258e-01 5.48705518e-01 -7.65604794e-01 -9.49845910e-01 1.96848199e-01 1.07133818e+00 5.28611720e-01 -1.55636609e-01 1.31781662e+00 -9.20529723e-01 -9.62696820e-02 2.76554614e-01 -7.94832557e-02 1.53373361e-01 3.02649558e-01 -4.50435907e-01 -8.96337092e-01 -5.88996634e-02 4.11580682e-01 -6.62563920e-01 9.02776793e-02 -1.06158301e-01 2.74536997e-01 -6.40689671e-01 -5.33706835e-03 -2.31154382e-01 6.62128866e-01 2.96833634e-01 8.10256243e-01 2.65958667e-01 1.16277613e-01 1.05688405e+00 8.42933476e-01 9.37342346e-02 1.08647633e+00 7.60553300e-01 2.96630114e-01 4.78513718e-01 -1.41246900e-01 -1.27551973e-01 2.80413002e-01 1.12659323e+00 2.84771740e-01 -1.76524788e-01 -8.90397549e-01 7.76870310e-01 -1.93193817e+00 -7.91275382e-01 -5.35261393e-01 1.58658099e+00 1.47066391e+00 2.48013344e-02 -9.53238755e-02 -7.14226365e-02 8.05063426e-01 1.50041565e-01 -3.33618373e-01 -7.28543520e-01 1.48456125e-02 2.36195698e-01 -6.13085985e-01 1.19448984e+00 -8.37146759e-01 1.23127925e+00 6.50750637e+00 1.77106246e-01 -6.82058692e-01 3.74604687e-02 1.41159728e-01 5.67858398e-01 -3.00410241e-01 3.31143439e-01 -7.51833916e-01 8.18784460e-02 1.19386220e+00 -2.77993113e-01 3.07827711e-01 5.40726900e-01 -2.92531326e-02 -1.71037421e-01 -1.17215478e+00 3.82996708e-01 1.62323624e-01 -1.15042496e+00 1.33431733e-01 -4.39967394e-01 3.45501840e-01 -5.05924523e-01 -6.77264512e-01 6.72097862e-01 4.36355889e-01 -1.03278720e+00 4.42059971e-02 -5.67412563e-02 2.14607120e-01 -6.44913435e-01 1.09579039e+00 7.23609328e-01 -6.67305768e-01 -2.94035096e-02 -3.57443511e-01 -5.79524040e-01 5.30962467e-01 -1.39838353e-01 -1.15913379e+00 6.97108507e-01 4.22854781e-01 6.85542300e-02 -2.20727414e-01 5.78207254e-01 -8.28464568e-01 8.16283166e-01 -1.71823427e-01 -1.75526679e-01 2.44796708e-01 -1.89220488e-01 5.95941007e-01 1.09311891e+00 -2.22477555e-01 8.48984182e-01 8.07534382e-02 5.54177880e-01 4.70369793e-02 1.56571463e-01 -3.77428974e-03 1.91438213e-01 8.00023258e-01 1.09051943e+00 8.49735644e-03 -3.70930701e-01 -3.03019583e-01 9.09533143e-01 5.67592382e-01 7.42311478e-02 -4.68390048e-01 -4.26909745e-01 5.66251695e-01 -6.24906681e-02 -7.46641159e-02 -7.92350695e-02 6.59171939e-02 -1.08338034e+00 5.08388095e-02 -1.40625560e+00 9.22055781e-01 -9.16428208e-01 -1.37945294e+00 7.95397639e-01 3.00866496e-02 -5.76178610e-01 -7.56266892e-01 -3.62267703e-01 -1.00124443e+00 1.00340605e+00 -1.60397851e+00 -8.91302407e-01 -2.10002318e-01 5.24580240e-01 7.60370135e-01 2.96857268e-01 1.40038192e+00 7.45463073e-02 -3.39790046e-01 3.00512195e-01 -5.64709306e-01 3.06914300e-01 9.01887119e-01 -1.38476765e+00 4.82066929e-01 6.07648432e-01 -1.39101863e-01 9.38081324e-01 9.34419036e-01 -3.98761690e-01 -1.22031331e+00 -5.09816408e-01 1.51487124e+00 -5.59655130e-01 6.50472462e-01 5.80621138e-02 -1.54272389e+00 7.03951776e-01 7.75506735e-01 -5.43948948e-01 7.99929380e-01 4.54749018e-01 -4.58915472e-01 2.86637276e-01 -1.19987643e+00 5.77416539e-01 7.29444206e-01 -7.18922198e-01 -1.73860180e+00 5.54920733e-01 1.07347822e+00 -7.72913992e-01 -8.27128053e-01 2.73746818e-01 -4.90829162e-02 -7.94003367e-01 6.50787830e-01 -1.27777135e+00 4.48851585e-01 -2.04701051e-01 -3.06716692e-02 -1.25566888e+00 1.24737352e-01 -8.72700930e-01 -9.73661765e-02 1.25804532e+00 4.00161415e-01 -6.22475266e-01 4.87427920e-01 9.85307515e-01 -2.89860070e-01 -2.08600476e-01 -1.07064605e+00 -2.02426732e-01 3.59731555e-01 -1.08506702e-01 5.74174881e-01 1.19437969e+00 7.59424150e-01 1.13045835e+00 -1.99276730e-01 1.83454916e-01 1.23266838e-01 4.21846598e-01 8.28529835e-01 -1.22138798e+00 -3.15064996e-01 -3.01209483e-02 3.38990331e-01 -1.46722412e+00 3.98188949e-01 -4.28724289e-01 2.44717017e-01 -1.76518524e+00 -1.42239472e-02 -3.91736418e-01 4.07829821e-01 3.73738557e-01 -6.13213658e-01 -3.81759882e-01 3.43734175e-01 2.22083837e-01 -7.65890777e-01 5.93839109e-01 1.41814363e+00 1.71223581e-02 -3.11414272e-01 1.21854834e-01 -8.56654465e-01 6.75535440e-01 8.82909179e-01 -3.94202650e-01 -3.47472608e-01 -5.76754451e-01 -4.29932624e-02 7.30776906e-01 -8.00886229e-02 -5.58771431e-01 4.84604716e-01 -3.29780251e-01 -5.70317268e-01 -3.00947249e-01 4.63684499e-01 -1.60267964e-01 -4.21231776e-01 4.73119527e-01 -5.70124030e-01 2.56230384e-01 3.94224584e-01 2.33600020e-01 -4.78111804e-01 -6.47163749e-01 4.13855970e-01 -2.59345114e-01 -4.43320245e-01 -5.56595683e-01 -6.51345193e-01 6.51389003e-01 6.45903051e-01 4.89697725e-01 -8.43431056e-01 -9.57576454e-01 -7.73140252e-01 8.42245877e-01 -9.39465873e-03 4.71911073e-01 3.97902071e-01 -9.12341654e-01 -9.51965272e-01 -2.59583682e-01 -8.27348605e-02 2.80620217e-01 3.93552780e-01 3.49109381e-01 -6.49762571e-01 7.62427270e-01 -2.57968783e-01 -4.55596149e-01 -1.46429229e+00 5.76487146e-02 5.05605519e-01 -6.14582300e-01 -4.44295675e-01 5.86648464e-01 1.05647981e-01 -1.04284286e+00 9.84463245e-02 -8.66416842e-02 -7.89828062e-01 -1.23804606e-01 8.64569604e-01 -1.30978813e-02 1.35474682e-01 -6.59925580e-01 -3.85909796e-01 2.01184914e-01 -3.59333634e-01 -2.35331357e-01 1.00389326e+00 -4.15696681e-01 -4.33359355e-01 2.54615813e-01 6.39396608e-01 1.40947759e-01 -8.63218307e-01 -4.51994151e-01 2.77268231e-01 -2.30181679e-01 -6.74783885e-01 -1.11376643e+00 -1.34874091e-01 9.23702478e-01 -1.79437995e-01 5.60356617e-01 6.97907388e-01 1.62236989e-01 1.14675975e+00 9.07837272e-01 2.34217674e-01 -8.14629793e-01 1.96783155e-01 1.16880596e+00 1.08741188e+00 -1.01990581e+00 -2.54734367e-01 -9.41721678e-01 -9.73616242e-01 1.03023612e+00 1.07987475e+00 2.57028580e-01 -8.06501061e-02 2.17719339e-02 7.28223085e-01 -3.99203479e-01 -1.11409020e+00 -8.67306069e-02 -7.93775544e-02 4.62261975e-01 6.14754438e-01 5.41522168e-03 -7.89915919e-01 8.23304057e-01 -5.74215531e-01 -4.50019479e-01 8.76704216e-01 9.96194363e-01 -5.10268807e-01 -1.40244377e+00 -4.22044814e-01 -7.39025101e-02 -4.84702229e-01 -1.05604842e-01 -7.82257676e-01 8.02811027e-01 -6.10065460e-01 1.65600753e+00 4.63910997e-02 -2.68615671e-02 6.04674697e-01 3.58314067e-01 2.10013390e-01 -8.37341964e-01 -9.44695115e-01 -4.63471532e-01 1.01839983e+00 -3.75074327e-01 -5.33695996e-01 -4.92797762e-01 -1.65755451e+00 -3.09888303e-01 -4.62476045e-01 9.96473491e-01 2.14967266e-01 1.42926812e+00 3.48206311e-01 2.02849224e-01 4.08591449e-01 1.22477867e-01 -9.64356124e-01 -1.20414031e+00 1.26565427e-01 7.33772337e-01 1.35922283e-01 -3.71441931e-01 -3.65691572e-01 -1.15098953e-01]
[12.262101173400879, 8.020942687988281]
a12534ad-64a3-4716-9aae-38c6f67a782c
finite-time-analysis-of-minimax-q-learning
2306.057
null
https://arxiv.org/abs/2306.05700v2
https://arxiv.org/pdf/2306.05700v2.pdf
Finite-Time Analysis of Minimax Q-Learning for Two-Player Zero-Sum Markov Games: Switching System Approach
The objective of this paper is to investigate the finite-time analysis of a Q-learning algorithm applied to two-player zero-sum Markov games. Specifically, we establish a finite-time analysis of both the minimax Q-learning algorithm and the corresponding value iteration method. To enhance the analysis of both value iteration and Q-learning, we employ the switching system model of minimax Q-learning and the associated value iteration. This approach provides further insights into minimax Q-learning and facilitates a more straightforward and insightful convergence analysis. We anticipate that the introduction of these additional insights has the potential to uncover novel connections and foster collaboration between concepts in the fields of control theory and reinforcement learning communities.
['Donghwan Lee']
2023-06-09
null
null
null
null
['q-learning']
['methodology']
[ 7.79071897e-02 3.21547508e-01 -3.88741195e-01 2.17238635e-01 -7.10426569e-01 -7.46666014e-01 2.99069464e-01 3.90366822e-01 -5.90759397e-01 1.00392163e+00 -2.94383407e-01 -7.15537667e-01 -8.80047441e-01 -5.73536754e-01 -3.70067626e-01 -6.74120188e-01 -5.87840974e-01 1.81343526e-01 -3.35797928e-02 -3.74408334e-01 4.26435202e-01 4.62766774e-02 -1.00877607e+00 -4.23145860e-01 7.55625486e-01 8.98215473e-01 6.76406082e-03 8.55922759e-01 2.53655165e-01 8.65431011e-01 -4.17151332e-01 -2.27499604e-01 3.70420039e-01 -6.34690821e-01 -8.58632028e-01 1.06798120e-01 -1.32443830e-01 -2.13288411e-01 -3.68595533e-02 1.12615812e+00 5.22692263e-01 6.54446602e-01 6.23093486e-01 -1.55147564e+00 -1.99194357e-01 3.66908342e-01 -5.53247273e-01 3.96736145e-01 2.02641100e-01 5.26851058e-01 1.25752366e+00 -1.52144700e-01 3.88095140e-01 1.27771997e+00 5.77848434e-01 6.93174422e-01 -1.26574302e+00 -5.53609371e-01 6.78192228e-02 -5.37647046e-02 -8.72363389e-01 -3.83709259e-02 4.83230263e-01 -4.39869076e-01 7.42104709e-01 -1.60810113e-01 1.07782030e+00 6.95056140e-01 4.58203763e-01 9.04579937e-01 1.27575421e+00 -5.15493691e-01 7.48134971e-01 -1.47006914e-01 -1.20836392e-01 8.23443770e-01 1.60717145e-01 8.52408290e-01 -3.90477151e-01 -2.64100730e-01 9.36945736e-01 6.33977726e-02 1.62176564e-01 -7.53976822e-01 -8.23706388e-01 1.27714300e+00 4.14697267e-02 -5.94275855e-02 -5.97511053e-01 6.85222030e-01 2.47602269e-01 6.11458361e-01 4.04104292e-01 8.15029800e-01 -3.51757258e-01 -5.76036692e-01 -7.17848539e-01 6.20025337e-01 1.02712369e+00 5.72718322e-01 7.33002305e-01 3.52608301e-02 -1.46750242e-01 1.89608485e-01 5.96180975e-01 4.08930391e-01 -1.49875745e-01 -1.76984191e+00 4.03909385e-01 6.68966100e-02 5.27799785e-01 -5.37062705e-01 -3.61059517e-01 -4.42517072e-01 -3.96006942e-01 5.86621463e-01 6.69483960e-01 -7.80124664e-01 -3.41097027e-01 1.96287775e+00 2.37418562e-01 -3.51125561e-02 2.76764870e-01 8.22420239e-01 -1.85522109e-01 6.09917819e-01 4.45955023e-02 -8.01698685e-01 8.06174934e-01 -6.02445900e-01 -6.76731944e-01 1.26814768e-01 7.85441995e-01 -4.21494961e-01 8.45772743e-01 3.48923683e-01 -1.36530054e+00 -7.50328526e-02 -8.01786482e-01 7.52106309e-01 1.99670359e-01 -8.12176168e-01 6.09040380e-01 7.95782030e-01 -1.01923156e+00 9.00107265e-01 -9.30705130e-01 -4.90138322e-01 5.66866636e-01 4.12529588e-01 3.84233177e-01 2.11495608e-01 -1.25751412e+00 8.14730465e-01 1.57562196e-01 -1.42836794e-01 -1.11071050e+00 -7.07485020e-01 -4.88188952e-01 1.22977607e-01 1.16207862e+00 -6.78292096e-01 1.77711236e+00 -1.20949674e+00 -1.93471086e+00 2.29655713e-01 3.10004115e-01 -3.52586836e-01 6.96311712e-01 1.17262803e-01 3.57926428e-01 2.51971871e-01 2.32778609e-01 8.80557373e-02 6.74452782e-01 -8.65098476e-01 -6.60643160e-01 -2.84836560e-01 3.23858708e-01 4.42387402e-01 -6.11249357e-02 -2.11348623e-01 4.61023033e-01 -4.66111183e-01 -6.52586818e-01 -9.41033185e-01 -9.08158720e-01 -7.24333525e-02 -1.62247773e-02 -2.95847386e-01 8.29923227e-02 -1.82388835e-02 9.94291902e-01 -1.86042356e+00 1.96839407e-01 4.82895821e-01 2.24796891e-01 -4.48283106e-02 -3.91848564e-01 9.93216932e-01 1.66604444e-01 2.30044618e-01 1.39343487e-02 1.37184501e-01 2.15856835e-01 2.86144525e-01 -1.55739143e-01 5.94829917e-01 1.19423382e-01 1.08847642e+00 -1.14750206e+00 -3.29371870e-01 -1.03093684e-02 -2.25736916e-01 -8.76996458e-01 2.76890188e-01 -2.45373681e-01 2.95206338e-01 -7.97409236e-01 1.15381412e-01 2.45591149e-01 -3.20748210e-01 5.78710377e-01 5.89384913e-01 -3.69504780e-01 -1.68158039e-01 -1.34550107e+00 1.22234261e+00 -2.59450465e-01 4.75541443e-01 2.57980436e-01 -1.38276863e+00 3.18886071e-01 3.47652495e-01 9.04415071e-01 -7.60566294e-01 2.02419683e-01 -1.89788509e-02 2.68054809e-02 -4.46868479e-01 4.18950468e-01 -7.93240190e-01 -1.56324670e-01 9.33289230e-01 -2.87283882e-02 -1.98610336e-01 2.85988182e-01 3.30952078e-01 9.03671682e-01 -1.07980810e-01 5.72865725e-01 -6.80206060e-01 1.40274659e-01 1.62776280e-02 3.03548843e-01 1.16867447e+00 -7.82256722e-01 -3.51893395e-01 8.65389407e-01 -4.93834652e-02 -8.63650143e-01 -1.29122424e+00 2.25890085e-01 1.30020988e+00 1.95542186e-01 -4.50600088e-01 -5.13385475e-01 -5.76116562e-01 3.46718132e-01 3.12428474e-01 -6.29826427e-01 -3.27265561e-01 -2.25266784e-01 -5.39806545e-01 1.32615402e-01 3.80690724e-01 1.65774927e-01 -8.94120753e-01 -9.19722199e-01 4.07073170e-01 1.08717501e-01 -3.08905780e-01 -7.06593394e-01 2.67086536e-01 -1.05396354e+00 -1.21528041e+00 -6.05417788e-01 -5.13570607e-01 3.02865982e-01 -5.64884953e-02 8.96369815e-01 -2.10328311e-01 4.03490998e-02 1.11277604e+00 -2.34351546e-01 -3.78227741e-01 -6.12378597e-01 5.48766479e-02 1.44467577e-01 -8.50195512e-02 -1.20631658e-01 -2.29533181e-01 -7.32662797e-01 5.13027348e-02 -8.96670043e-01 -2.90969163e-01 3.54308963e-01 8.12269628e-01 3.30846906e-01 1.14928328e-01 8.89460206e-01 -5.52827954e-01 1.29548645e+00 -5.01745224e-01 -1.15073359e+00 2.07236677e-01 -1.05219913e+00 4.65907931e-01 6.68304324e-01 -5.00960767e-01 -9.11098301e-01 -9.65847000e-02 2.33604759e-01 -1.10303620e-02 4.34643239e-01 6.65615022e-01 3.10428411e-01 -1.05105996e-01 3.61751407e-01 1.52175665e-01 5.61795950e-01 -1.26755372e-01 6.36491060e-01 4.06263053e-01 -1.18815236e-01 -6.97423935e-01 6.06306732e-01 2.26453587e-01 3.42361122e-01 -6.80444658e-01 -5.50354421e-01 -3.73879671e-01 -1.24841690e-01 -5.14223993e-01 7.39978373e-01 -6.37672722e-01 -1.66091144e+00 2.29664713e-01 -7.23587573e-01 -8.39237988e-01 -7.67723322e-01 4.16105419e-01 -1.31512356e+00 3.75504762e-01 -6.20834827e-01 -1.49842775e+00 2.02588424e-01 -7.42255867e-01 4.46561188e-01 5.94645143e-01 4.61140499e-02 -1.35250306e+00 7.60892987e-01 1.79473951e-01 2.43139192e-01 -6.55963123e-02 5.98508716e-01 -3.20152342e-01 -6.89259291e-01 2.14924470e-01 2.48781532e-01 1.03931479e-01 -2.17024356e-01 -7.27017075e-02 -2.60711670e-01 -7.47957349e-01 -1.05623767e-01 -6.01909459e-01 7.03114569e-01 7.81453729e-01 5.68835855e-01 -5.10954738e-01 6.52146265e-02 1.63712755e-01 1.81222701e+00 4.47848439e-01 9.37849507e-02 2.88602799e-01 8.13182220e-02 7.28598356e-01 7.08658516e-01 9.53413844e-01 2.93276757e-01 3.97343129e-01 5.01045704e-01 1.77862361e-01 8.43591630e-01 -3.16167653e-01 4.68750119e-01 5.31218171e-01 -2.24452689e-01 -3.62893054e-03 -7.88990557e-01 3.58336717e-01 -2.17353296e+00 -1.32325041e+00 3.21514130e-01 2.30947518e+00 9.66277122e-01 -9.26316679e-02 7.40921438e-01 -1.03468306e-01 6.35475457e-01 3.57601941e-02 -7.20503449e-01 -7.99240351e-01 2.38456190e-01 5.57859302e-01 6.22260153e-01 7.83971846e-01 -8.87197375e-01 8.25834692e-01 7.94349432e+00 8.38774323e-01 -6.30239844e-01 9.43032354e-02 4.63179588e-01 -4.31684792e-01 -2.08458140e-01 1.60027519e-01 -3.08570653e-01 3.14023346e-01 1.07701910e+00 -7.07024753e-01 1.01626539e+00 6.67318702e-01 8.58361065e-01 -3.86679232e-01 -8.62703025e-01 5.66932440e-01 -6.52265131e-01 -1.30649626e+00 -4.80742872e-01 5.29342353e-01 1.00632370e+00 -2.02137128e-01 1.23768874e-01 3.95768821e-01 8.64193857e-01 -9.92464960e-01 5.50432146e-01 4.27818298e-01 6.51256740e-01 -1.18393564e+00 3.34857523e-01 4.00541335e-01 -9.79468465e-01 -6.53911829e-01 -1.44279208e-02 -8.96220088e-01 1.00106999e-01 1.19826131e-01 -2.18969256e-01 2.13113084e-01 1.86320588e-01 6.49976790e-01 3.38936821e-02 9.68687236e-01 1.32466316e-01 6.96781516e-01 -1.01715602e-01 -5.78579783e-01 5.36281288e-01 -5.67604065e-01 5.01758575e-01 6.91799700e-01 -7.74911568e-02 7.62046874e-02 5.10420501e-01 1.01780140e+00 1.34556577e-01 2.16963008e-01 -3.83295834e-01 -5.12309253e-01 1.32782280e-01 9.25653577e-01 -8.04462254e-01 -6.17802665e-02 -4.65937346e-01 4.65383828e-01 2.86646962e-01 6.00067914e-01 -6.67900681e-01 -5.29965699e-01 1.04736066e+00 1.27094677e-02 3.88556719e-01 -2.67960787e-01 -6.11443445e-02 -9.52828228e-01 -3.48701864e-01 -7.41304636e-01 4.00670439e-01 -1.92641437e-01 -1.38444412e+00 -3.95722121e-01 1.78853888e-02 -7.71464109e-01 -5.79831183e-01 -4.68554795e-01 -5.62181413e-01 5.66140592e-01 -1.41840076e+00 -3.12912941e-01 5.76754451e-01 5.62260211e-01 1.28067806e-01 -2.41419543e-02 2.89675564e-01 -1.61296040e-01 -6.62705004e-01 3.12165856e-01 9.32219088e-01 -9.71622020e-02 1.80055022e-01 -1.42106032e+00 -2.08166987e-01 6.26396000e-01 -1.02669127e-01 2.60356963e-01 7.54667759e-01 -3.10700476e-01 -1.57085431e+00 -8.47739875e-01 1.66724175e-01 -4.03581589e-01 1.05504322e+00 3.48638818e-02 -3.58259648e-01 4.03002262e-01 1.89523354e-01 -3.21112603e-01 6.20742917e-01 8.88477415e-02 2.56913245e-01 -1.53077736e-01 -8.11072648e-01 5.85406542e-01 6.79175436e-01 -6.34445190e-01 -5.30177534e-01 1.21644437e-01 5.47905803e-01 2.32779860e-01 -7.59333134e-01 -2.93148071e-01 5.45117259e-01 -6.08415782e-01 6.88670814e-01 -1.11309624e+00 4.52524871e-01 2.02689350e-01 4.43461584e-03 -1.45008397e+00 -3.53620738e-01 -1.46796763e+00 -1.00569941e-01 7.06290364e-01 1.57364249e-01 -7.31704354e-01 9.58202839e-01 3.93423766e-01 3.83765906e-01 -8.39946926e-01 -1.19184995e+00 -1.09095228e+00 8.37646067e-01 -4.31073189e-01 -1.74760818e-01 6.43051922e-01 7.69888461e-01 2.69678771e-01 -3.36245954e-01 -3.11291605e-01 1.01477802e+00 7.45962933e-02 3.33074898e-01 -1.00223279e+00 -6.04983270e-01 -6.24349773e-01 -5.35928039e-03 -1.09258103e+00 1.94363073e-01 -7.00975537e-01 1.99391857e-01 -1.20801044e+00 4.93634939e-01 -1.60473347e-01 -5.26936471e-01 1.00615770e-01 -1.66822836e-01 2.40982175e-02 7.17790365e-01 6.34415308e-03 -1.08954453e+00 8.80114079e-01 1.49308562e+00 1.29452527e-01 -5.58802545e-01 4.72676128e-01 -7.84569085e-01 2.02790454e-01 8.78296256e-01 -7.22770214e-01 -6.27105832e-01 1.64186552e-01 6.68140531e-01 6.35307074e-01 4.73199934e-01 -3.77329826e-01 1.66667849e-01 -1.00070453e+00 -3.82865399e-01 -3.80016491e-02 -2.01550558e-01 -3.98843557e-01 -4.72386211e-01 1.09192002e+00 -6.97986960e-01 -2.86060665e-03 -9.13539007e-02 9.55211163e-01 1.21871032e-01 -3.67245555e-01 8.98358822e-01 -3.97244066e-01 -3.42459440e-01 3.33354086e-01 -1.25527024e+00 8.24826479e-01 1.16801381e+00 -1.53817922e-01 1.53774351e-01 -7.93910682e-01 -7.27443218e-01 5.81238449e-01 1.41838908e-01 -1.63243726e-01 4.39042866e-01 -1.00793374e+00 -3.90498042e-01 -3.03410273e-02 -3.66350085e-01 -6.80952430e-01 5.93039021e-02 8.86176646e-01 -8.34885761e-02 3.73251379e-01 -4.58028883e-01 -2.30708271e-01 -8.54015172e-01 4.69844341e-01 8.01059246e-01 -5.77004135e-01 4.14693803e-02 2.86930144e-01 3.75398770e-02 -2.29008421e-01 5.32834083e-02 -2.35354006e-01 4.71964367e-02 9.90624577e-02 2.25812033e-01 7.91866958e-01 -7.75674462e-01 -7.19629228e-02 -2.02363759e-01 3.54739219e-01 3.21220815e-01 -8.23248804e-01 1.23687482e+00 -4.32676673e-01 2.05309957e-01 4.57796782e-01 8.83918881e-01 -4.10331041e-01 -1.80669785e+00 -7.53028095e-02 1.91398457e-01 -9.87624153e-02 -3.88476253e-02 -4.59055305e-01 -9.04247582e-01 9.01521921e-01 6.29432023e-01 5.04781067e-01 9.10910726e-01 -9.40720215e-02 3.78964663e-01 5.06699920e-01 2.91944474e-01 -1.56420135e+00 6.38204336e-01 6.29177451e-01 2.26196617e-01 -1.13872659e+00 -1.76445648e-01 2.77167588e-01 -5.60420096e-01 1.06371391e+00 3.55616897e-01 -3.36971760e-01 9.34371531e-01 2.44232282e-01 -1.31865010e-01 1.95634980e-02 -1.19169557e+00 -5.80814004e-01 -2.53970891e-01 7.40416944e-01 1.58269346e-01 6.35761991e-02 -5.41397691e-01 5.40436447e-01 1.92920148e-01 3.19323301e-01 8.37700009e-01 1.08822095e+00 -5.09181917e-01 -1.13362443e+00 -1.33814856e-01 2.56096572e-01 -5.65540493e-01 -6.24242313e-02 -2.29086429e-01 6.74411476e-01 -5.26958227e-01 1.16596079e+00 7.37264678e-02 4.47597988e-02 1.24005415e-01 -1.77976780e-03 7.81407654e-01 -4.51326072e-01 -6.36791289e-01 5.24991676e-02 -3.72361451e-01 -6.96258485e-01 -6.24369860e-01 -7.18230724e-01 -1.06420505e+00 -6.31207883e-01 -4.68568385e-01 4.56674606e-01 2.05716714e-01 1.16838014e+00 2.12845057e-01 2.62443602e-01 9.09658074e-01 -4.08731878e-01 -1.47312713e+00 -4.39209610e-01 -9.60644782e-01 1.84199452e-01 4.29661810e-01 -4.79814768e-01 -2.85412908e-01 -2.73998141e-01]
[4.187510967254639, 2.617335796356201]
33c95b59-7ba7-4fff-9fe7-d1cd884c2e56
automated-identification-of-tree-species-by
2210.0929
null
https://arxiv.org/abs/2210.09290v1
https://arxiv.org/pdf/2210.09290v1.pdf
Automated Identification of Tree Species by Bark Texture Classification Using Convolutional Neural Networks
Identification of tree species plays a key role in forestry related tasks like forest conservation, disease diagnosis and plant production. There had been a debate regarding the part of the tree to be used for differentiation, whether it should be leaves, fruits, flowers or bark. Studies have proven that bark is of utmost importance as it will be present despite seasonal variations and provides a characteristic identity to a tree by variations in the structure. In this paper, a deep learning based approach is presented by leveraging the method of computer vision to classify 50 tree species, on the basis of bark texture using the BarkVN-50 dataset. This is the maximum number of trees being considered for bark classification till now. A convolutional neural network(CNN), ResNet101 has been implemented using transfer-learning based technique of fine tuning to maximise the model performance. The model produced an overall accuracy of >94% during the evaluation. The performance validation has been done using K-Fold Cross Validation and by testing on unseen data collected from the Internet, this proved the model's generalization capability for real-world uses.
['Sahil Faizal']
2022-10-03
null
null
null
null
['texture-classification']
['computer-vision']
[ 2.45744780e-01 4.19955775e-02 -8.99339914e-02 -1.70539305e-01 1.56651095e-01 -6.22292161e-01 6.75734997e-01 3.67481202e-01 -2.26442188e-01 7.13444352e-01 2.15131771e-02 -5.65454602e-01 -5.74995100e-01 -8.83163095e-01 9.60063860e-02 -6.55467868e-01 -4.92860466e-01 4.29588169e-01 4.38782096e-01 -2.04462763e-02 4.01563376e-01 1.09612083e+00 -1.72705305e+00 1.40951067e-01 5.34965277e-01 1.12614238e+00 5.95709622e-01 9.59652603e-01 -2.27534994e-01 7.53574908e-01 -7.98371613e-01 -8.23945254e-02 3.36467773e-01 7.09439665e-02 -1.19143963e+00 6.81447014e-02 4.61466372e-01 -4.86948550e-01 4.19133931e-01 5.96641839e-01 3.46450061e-01 -2.43066117e-01 8.00560951e-01 -7.94681728e-01 -2.75624663e-01 4.47962552e-01 -2.02074677e-01 3.51692885e-01 -1.16405278e-01 2.41202004e-02 8.66889000e-01 -3.33824605e-01 4.24726129e-01 1.03826213e+00 5.21860540e-01 1.18339971e-01 -1.31451797e+00 -5.31132996e-01 -3.11304539e-01 4.68131840e-01 -1.13516510e+00 3.55082490e-02 6.50112391e-01 -6.15902841e-01 5.58778226e-01 3.92924666e-01 7.83741832e-01 6.42804861e-01 3.39883476e-01 2.74378985e-01 1.50974417e+00 -4.93538201e-01 4.05081481e-01 6.23689890e-02 4.19800356e-02 3.08887661e-01 1.80165425e-01 1.77246779e-01 1.24595642e-01 3.20355803e-01 7.48754144e-01 -4.06404108e-01 -4.41305898e-02 -1.52894661e-01 -6.96359754e-01 9.05705988e-01 9.20059025e-01 8.42675686e-01 -6.24385774e-01 -1.62632868e-01 7.25174963e-01 1.69738084e-01 1.93657368e-01 4.13137376e-01 -6.78093791e-01 6.45968318e-02 -9.02032137e-01 4.01682453e-04 1.00612724e+00 1.53606445e-01 2.44240493e-01 1.68910660e-02 1.31591424e-01 9.44930136e-01 2.14209810e-01 1.91948310e-01 6.00039542e-01 -7.32962191e-01 -3.75250071e-01 9.16545630e-01 -4.37297016e-01 -8.56807351e-01 -4.08776939e-01 -6.27467096e-01 -9.93851721e-01 7.36421347e-01 3.78223807e-01 2.59572029e-01 -1.12025309e+00 1.40200031e+00 5.03064513e-01 -8.76141116e-02 -4.64844592e-02 6.39230251e-01 1.01246834e+00 4.58423704e-01 4.19458151e-01 6.82472289e-02 1.30341673e+00 -5.08426547e-01 -2.38984436e-01 1.56052381e-01 2.65935123e-01 -8.84033263e-01 4.09580797e-01 7.17742980e-01 -2.25013867e-01 -7.18446732e-01 -1.30030453e+00 4.73769575e-01 -9.61023033e-01 4.29382175e-01 8.58290195e-01 9.72977579e-01 -8.38340998e-01 9.12729919e-01 -3.80139947e-01 -9.19361532e-01 5.12024343e-01 4.31065112e-01 -6.15272224e-01 2.26829574e-01 -6.31487131e-01 9.92183328e-01 1.04042125e+00 4.82304633e-01 -8.65795970e-01 -2.02638656e-01 -3.34325805e-02 4.92347479e-02 4.41196635e-02 -1.74037710e-01 1.16295612e+00 -1.19091201e+00 -1.43390155e+00 1.10284197e+00 4.15737867e-01 -8.47148061e-01 3.87509108e-01 -1.43860066e-02 -5.85083365e-01 1.95748419e-01 1.74302369e-01 6.65362477e-01 8.93693626e-01 -1.18460429e+00 -8.11351240e-01 -4.68638003e-01 -3.44671286e-03 -3.75336260e-01 -9.39494148e-02 -1.35766342e-01 5.66666186e-01 -5.20488381e-01 1.65278196e-01 -1.00389004e+00 1.90265223e-01 -5.71772717e-02 -1.73651591e-01 -1.26492539e-02 1.11353350e+00 -1.01682293e+00 7.73335159e-01 -1.77610648e+00 -2.93278277e-01 1.87317759e-01 1.75361723e-01 7.16921508e-01 1.34670332e-01 6.07680917e-01 -2.04603896e-01 2.44529754e-01 -2.51925826e-01 6.55240417e-01 -5.16107619e-01 5.44630468e-01 -1.09868288e-01 1.86388806e-01 2.55872518e-01 3.90655309e-01 -5.13798356e-01 -5.00003874e-01 5.00231802e-01 4.22070563e-01 3.37623090e-01 1.40089214e-01 6.43233359e-02 3.93829942e-01 -3.55574101e-01 8.25314760e-01 8.98870707e-01 2.93547839e-01 8.48046541e-02 -2.43881136e-01 -2.04671219e-01 -1.78538978e-01 -9.86666024e-01 9.06660259e-01 -5.96351564e-01 8.38011324e-01 3.89101595e-01 -9.19896007e-01 1.26389861e+00 2.57842928e-01 3.38543981e-01 -3.31089616e-01 6.53747544e-02 4.45768028e-01 6.23296797e-01 -2.96967000e-01 1.25861436e-01 6.64729029e-02 7.27130473e-01 -1.37037650e-01 1.03063725e-01 -1.97497427e-01 2.97294021e-01 -2.42024451e-01 9.77893531e-01 9.56457630e-02 4.28188831e-01 -5.60217679e-01 8.20204675e-01 7.74493739e-02 6.26784042e-02 2.38327324e-01 -3.95747244e-01 2.24702582e-01 2.98705608e-01 -4.67691869e-01 -9.39612806e-01 -8.94424975e-01 -8.26539218e-01 1.03062797e+00 -5.54384470e-01 8.14906955e-02 -6.76490188e-01 -4.51227307e-01 4.25233953e-02 6.22582316e-01 -7.84267008e-01 -3.30740176e-02 -1.86192617e-01 -5.09263754e-01 3.88247728e-01 4.12858188e-01 9.71126854e-01 -1.50636256e+00 -1.20822477e+00 3.75780076e-01 4.92623389e-01 -7.94688344e-01 5.74567676e-01 8.64456654e-01 -1.13541806e+00 -1.30310607e+00 -6.63637400e-01 -5.90450883e-01 6.28813803e-02 3.78935901e-03 9.80011463e-01 -1.01073772e-01 -7.56555617e-01 5.53792715e-02 -6.87300205e-01 -4.98930216e-01 -5.84146917e-01 4.92188245e-01 -4.95465219e-01 -2.11574644e-01 2.50648052e-01 -9.77631867e-01 -5.54972649e-01 1.51042581e-01 -7.68172622e-01 -2.39119351e-01 9.92437363e-01 8.08733225e-01 1.15412451e-01 4.24160510e-01 7.89016247e-01 -8.86229277e-01 6.08009458e-01 -1.55814067e-01 -4.45880532e-01 1.66947350e-01 -5.93820810e-01 -6.02493770e-02 6.67258978e-01 -1.92482993e-01 -6.35011673e-01 1.53400466e-01 -1.04201578e-01 1.15141660e-01 -6.64318025e-01 4.62380856e-01 -1.30100295e-01 -2.95043468e-01 7.31542408e-01 -4.32482697e-02 2.86449194e-02 -6.27256930e-01 1.83162570e-01 9.33731556e-01 7.09242880e-01 -2.90295213e-01 5.04616380e-01 1.20827436e-01 4.43369180e-01 -1.22874510e+00 -6.57927096e-01 -6.86450243e-01 -1.15294194e+00 -4.81526464e-01 9.87651467e-01 -2.79361695e-01 -9.03909385e-01 5.79360545e-01 -9.14848685e-01 -1.71988651e-01 -8.91655535e-02 2.36181200e-01 -2.35489503e-01 2.02898189e-01 -7.51282498e-02 -1.17525268e+00 -7.54596889e-01 -5.53432643e-01 8.39821517e-01 2.81719387e-01 -7.53465369e-02 -8.40561390e-01 3.54178622e-02 2.64253855e-01 5.39202929e-01 5.94673097e-01 1.08194804e+00 -6.64040804e-01 -2.13969842e-01 -5.82825720e-01 -3.76436234e-01 5.76750457e-01 4.26860571e-01 5.61380148e-01 -1.29111123e+00 6.01492226e-02 -1.42185465e-01 -1.92520782e-01 8.43551278e-01 3.89239699e-01 1.09609890e+00 -1.94725189e-02 -5.21290042e-02 2.66348869e-01 1.66409683e+00 4.98590320e-01 6.25970364e-01 7.38522649e-01 2.90705204e-01 7.41030574e-01 4.35865581e-01 3.55640262e-01 -4.17376965e-01 2.29960412e-01 1.22645438e+00 7.04601333e-02 -1.77560419e-01 4.42804815e-03 -1.41944647e-01 2.30324969e-01 -3.00976276e-01 5.08867912e-02 -9.86012757e-01 4.04736280e-01 -1.21875489e+00 -9.88830626e-01 -4.29497153e-01 2.45394921e+00 2.63550252e-01 4.92862672e-01 3.39202732e-01 8.37904990e-01 7.30224609e-01 -4.36629057e-02 -3.31892163e-01 -9.56469774e-01 -1.34749949e-01 7.36759186e-01 9.05403197e-01 2.55413294e-01 -1.45202363e+00 9.23885286e-01 5.99569845e+00 6.21133089e-01 -1.62020707e+00 -2.88381368e-01 6.15670860e-01 5.89714766e-01 6.46693766e-01 1.37766674e-01 -4.79027897e-01 2.21497223e-01 1.03726363e+00 3.57379735e-01 3.26348156e-01 9.35587168e-01 2.51470506e-01 -6.31847441e-01 -5.83266199e-01 4.85648721e-01 -3.20708007e-01 -9.50642884e-01 -1.72013655e-01 2.07594454e-01 1.20238394e-01 1.00400727e-02 -4.44056392e-01 1.81037366e-01 2.04799235e-01 -1.37637734e+00 4.28331733e-01 1.42945975e-01 3.93805623e-01 -6.50607467e-01 9.31088984e-01 4.47392941e-01 -1.41486752e+00 -3.98687571e-01 -3.88613015e-01 -5.96408285e-02 -2.65209675e-01 4.06021863e-01 -1.36516666e+00 7.03316271e-01 9.32697356e-01 3.68815899e-01 -1.26916087e+00 1.21586537e+00 -7.68241808e-02 9.14650619e-01 -4.62090880e-01 -2.25773960e-01 3.00856084e-01 -3.41364741e-01 2.10179761e-01 1.12246478e+00 1.52150184e-01 -6.00426197e-01 1.21893838e-01 6.92583025e-01 5.24167836e-01 4.82377470e-01 -5.15379012e-01 -1.93966046e-01 3.25524777e-01 1.62148869e+00 -1.07201815e+00 -5.24967127e-02 2.93983757e-01 7.88485289e-01 -1.14597483e-02 -1.89094096e-01 -2.71311015e-01 -2.26096734e-01 -9.83494371e-02 3.27030778e-01 5.32932222e-01 -1.78063095e-01 -1.31215796e-01 -8.45415071e-02 -9.98839438e-02 -4.86698925e-01 2.70202488e-01 -7.54629672e-01 -1.14877558e+00 6.67136490e-01 -4.01024893e-03 -8.65217805e-01 -1.07892431e-01 -1.18960214e+00 -7.37483323e-01 1.12848127e+00 -1.01460516e+00 -1.67811108e+00 -6.52123749e-01 -1.80676021e-02 4.14009124e-01 -2.77904451e-01 1.16224110e+00 -8.89898613e-02 -3.66608322e-01 -2.52354205e-01 2.65329659e-01 -1.98372584e-02 2.95194387e-01 -1.69180179e+00 -1.72867507e-01 4.03746277e-01 -5.11410646e-03 -4.84081022e-02 5.78423262e-01 -5.40475905e-01 -8.19586039e-01 -1.09740424e+00 6.18526101e-01 1.21476784e-01 6.29523158e-01 1.26336858e-01 -8.03938210e-01 1.94902513e-02 3.88990164e-01 -2.26433173e-01 7.17113733e-01 9.60112661e-02 -1.50732622e-01 -4.14366126e-01 -1.27375019e+00 1.18662871e-01 4.07612234e-01 -1.67363584e-01 -3.38928133e-01 2.37051040e-01 -2.44924247e-01 2.43250102e-01 -1.14698946e+00 4.65498835e-01 8.68843675e-01 -1.17693818e+00 7.99280941e-01 -4.39875543e-01 5.95842600e-01 -2.97117263e-01 -2.33838558e-01 -1.01431966e+00 -5.00735164e-01 -8.94622058e-02 1.78918108e-01 1.66972673e+00 1.75179049e-01 -5.60243309e-01 9.28526342e-01 -7.88298920e-02 1.74501985e-01 -5.04324436e-01 -9.51741517e-01 -9.60634828e-01 1.81985155e-01 -7.38622323e-02 3.31477553e-01 5.63604712e-01 -8.90549242e-01 6.55769169e-01 5.86981177e-02 -1.21081769e-01 2.96753883e-01 1.74418211e-01 8.90368283e-01 -2.07761621e+00 -2.04809666e-01 -7.41218090e-01 -1.12326026e+00 -2.46663913e-01 7.70251080e-02 -9.13719296e-01 -2.76902378e-01 -1.61350167e+00 -3.84440348e-02 -4.08310950e-01 9.44754202e-03 3.55692685e-01 2.27852598e-01 -5.95451146e-02 1.32936195e-01 2.17947870e-01 4.89384532e-01 1.54181093e-01 1.07745564e+00 -2.34148264e-01 -7.04363734e-02 3.96329433e-01 -2.52375811e-01 3.96557868e-01 1.11183619e+00 -1.34088635e-01 -3.02172899e-01 2.00027972e-01 -3.11552197e-01 -2.25328833e-01 6.49257421e-01 -1.34394169e+00 -3.78014326e-01 -1.48258219e-02 5.34850240e-01 -7.30526567e-01 3.42985600e-01 -1.35818362e+00 5.68690896e-01 8.67466033e-01 -1.83866814e-01 -1.85654953e-01 2.10715666e-01 3.72657537e-01 -9.14316997e-02 -8.54782701e-01 8.47222924e-01 -1.93047106e-01 -8.52575362e-01 -7.35955834e-02 -3.15300077e-01 -5.50972402e-01 1.03172874e+00 -7.39137590e-01 -1.87038332e-01 -2.10640967e-01 -8.55335474e-01 -1.84467018e-01 1.17918923e-01 2.35766664e-01 -1.39097432e-02 -1.03157663e+00 -8.21641684e-01 -2.14375332e-01 2.50890970e-01 -1.25652462e-01 -5.95506951e-02 6.56857848e-01 -1.30568624e+00 3.97589236e-01 -8.69726419e-01 -8.42609942e-01 -1.70443106e+00 5.80144703e-01 3.73655409e-01 -2.05958217e-01 -6.45209968e-01 4.51471597e-01 -3.57064575e-01 -2.72581369e-01 1.75851330e-01 -2.30605245e-01 -9.31742668e-01 1.94794327e-01 1.33286742e-02 3.91354173e-01 3.33637923e-01 -7.96317518e-01 -2.07459301e-01 4.54426408e-01 7.30945915e-02 4.36185896e-02 1.45284224e+00 5.01206160e-01 -1.95794597e-01 6.64134085e-01 9.52207804e-01 -2.32753515e-01 -8.67345691e-01 4.20819312e-01 4.18387353e-01 -6.00912929e-01 4.06913310e-01 -1.22518659e+00 -1.06758821e+00 9.27145123e-01 1.42960322e+00 8.88437271e-01 1.20771837e+00 -4.07704562e-01 1.51357383e-01 4.02925193e-01 1.68698840e-02 -1.12757683e+00 -5.80222249e-01 2.40167081e-01 1.21517158e+00 -1.33665657e+00 2.93791518e-02 -5.04273951e-01 -1.96110517e-01 1.58196771e+00 3.57884884e-01 -3.08388263e-01 9.01055872e-01 2.75562294e-02 1.99519238e-03 -9.90924332e-03 -5.41838169e-01 -6.07659638e-01 7.04057962e-02 1.12590575e+00 9.28448439e-01 6.30178690e-01 -7.52383411e-01 5.07980175e-02 -4.90013719e-01 1.17983431e-01 1.18123315e-01 9.77317274e-01 -7.12326944e-01 -1.12284184e+00 -4.84224409e-01 8.52596343e-01 -4.44077283e-01 1.29289269e-01 -1.17282283e+00 1.09508371e+00 3.33344460e-01 9.47575510e-01 -3.14428508e-01 -3.56970519e-01 4.11244743e-02 3.79519835e-02 6.39146924e-01 -4.02873605e-01 -1.06106675e+00 -2.54686713e-01 2.64755100e-01 1.82853281e-01 -5.51834106e-01 -4.21191871e-01 -5.84697545e-01 -4.56870228e-01 -6.16081715e-01 1.81680918e-02 1.16891849e+00 7.46667206e-01 -1.15873761e-01 7.11031079e-01 7.51079857e-01 -6.90246224e-01 -6.22150421e-01 -1.45245445e+00 -7.90621996e-01 1.28393680e-01 -1.50358409e-01 -7.23579288e-01 -5.89873455e-02 1.21194229e-01]
[9.343017578125, -1.42855703830719]
60fbe48e-abfe-48ca-99ee-54f1ebad6f8c
indoor-scene-generation-from-a-collection-of
2108.09022
null
https://arxiv.org/abs/2108.09022v1
https://arxiv.org/pdf/2108.09022v1.pdf
Indoor Scene Generation from a Collection of Semantic-Segmented Depth Images
We present a method for creating 3D indoor scenes with a generative model learned from a collection of semantic-segmented depth images captured from different unknown scenes. Given a room with a specified size, our method automatically generates 3D objects in a room from a randomly sampled latent code. Different from existing methods that represent an indoor scene with the type, location, and other properties of objects in the room and learn the scene layout from a collection of complete 3D indoor scenes, our method models each indoor scene as a 3D semantic scene volume and learns a volumetric generative adversarial network (GAN) from a collection of 2.5D partial observations of 3D scenes. To this end, we apply a differentiable projection layer to project the generated 3D semantic scene volumes into semantic-segmented depth images and design a new multiple-view discriminator for learning the complete 3D scene volume from 2.5D semantic-segmented depth images. Compared to existing methods, our method not only efficiently reduces the workload of modeling and acquiring 3D scenes for training, but also produces better object shapes and their detailed layouts in the scene. We evaluate our method with different indoor scene datasets and demonstrate the advantages of our method. We also extend our method for generating 3D indoor scenes from semantic-segmented depth images inferred from RGB images of real scenes.
['Xin Tong', 'Bin Zhou', 'Yu-Xiao Guo', 'Ming-Jia Yang']
2021-08-20
null
http://openaccess.thecvf.com//content/ICCV2021/html/Yang_Indoor_Scene_Generation_From_a_Collection_of_Semantic-Segmented_Depth_Images_ICCV_2021_paper.html
http://openaccess.thecvf.com//content/ICCV2021/papers/Yang_Indoor_Scene_Generation_From_a_Collection_of_Semantic-Segmented_Depth_Images_ICCV_2021_paper.pdf
iccv-2021-1
['scene-generation']
['computer-vision']
[ 4.87246156e-01 2.29307935e-01 4.88366872e-01 -6.26359165e-01 -6.76908016e-01 -9.48798835e-01 5.03284037e-01 -3.61991853e-01 9.18842852e-03 4.85142559e-01 1.87092602e-01 -2.06310824e-01 4.36866850e-01 -1.29653347e+00 -1.30567801e+00 -5.23689270e-01 3.88089508e-01 7.79290199e-01 1.96512025e-02 2.58798420e-01 1.00653321e-02 6.86486065e-01 -1.44074452e+00 1.49774075e-01 7.70341754e-01 1.13939881e+00 6.05498075e-01 1.08639848e+00 -5.12006879e-01 7.42253482e-01 -7.50778258e-01 4.31958511e-02 7.74512887e-01 -5.32862127e-01 -6.83849990e-01 8.04653227e-01 7.24668026e-01 -6.38164818e-01 -4.60686505e-01 7.71492839e-01 2.13372156e-01 2.46793315e-01 8.37067902e-01 -1.14481497e+00 -7.85277843e-01 -8.09475183e-02 -1.82754263e-01 -3.69887918e-01 6.20449841e-01 1.49861455e-01 3.47281128e-01 -7.88408399e-01 5.08260667e-01 1.38370728e+00 4.42041963e-01 7.42551744e-01 -1.46200824e+00 -3.94497812e-01 3.99695337e-01 -6.08913422e-01 -1.25066733e+00 -7.89056569e-02 1.14415228e+00 -6.24998271e-01 7.63519049e-01 1.87130988e-01 1.00033605e+00 1.19401324e+00 7.70664811e-02 7.28878260e-01 1.29838085e+00 -5.52078336e-02 6.33437157e-01 1.49300277e-01 -4.92753565e-01 8.39541793e-01 -5.36059551e-02 -1.46470696e-01 7.05584437e-02 -2.73696240e-02 1.29531276e+00 6.44780278e-01 -8.20481181e-02 -9.51728344e-01 -1.16455579e+00 7.11635172e-01 7.27081418e-01 -2.17200905e-01 -2.23167509e-01 3.83271575e-01 -9.94853452e-02 -2.40947202e-01 5.54400146e-01 2.23337457e-01 -2.84496456e-01 1.32592410e-01 -7.70346403e-01 3.87117893e-01 6.43279374e-01 1.45245862e+00 1.02715802e+00 1.32330269e-01 -6.42813742e-02 4.28048074e-01 3.49433988e-01 1.08323109e+00 1.19535752e-01 -1.36888552e+00 5.79941571e-01 8.02193403e-01 1.17868982e-01 -5.63432336e-01 5.21951495e-03 1.62193060e-01 -8.02498043e-01 1.61818787e-01 2.90657580e-01 1.24642730e-01 -1.50416982e+00 1.44201910e+00 4.90414351e-01 1.05797827e-01 2.27434635e-02 7.38195240e-01 9.17908549e-01 7.98371971e-01 -3.14554214e-01 4.47194546e-01 7.94516146e-01 -9.59784687e-01 -2.70920157e-01 -5.74763298e-01 5.04687950e-02 -3.67124289e-01 1.26040566e+00 2.40644142e-02 -1.13657129e+00 -8.09900105e-01 -7.31603086e-01 -3.52891773e-01 -4.55116510e-01 -1.77574992e-01 6.22502804e-01 7.08987713e-01 -9.99414921e-01 -3.31054814e-03 -9.03480470e-01 -1.40591398e-01 6.99363589e-01 1.39601260e-01 -2.24075064e-01 -4.20051783e-01 -4.03769612e-01 2.94309169e-01 2.98464477e-01 -2.00248346e-01 -1.74863029e+00 -7.56981671e-01 -1.33096707e+00 -2.75644481e-01 1.25321791e-01 -1.16085231e+00 1.07375658e+00 -6.03918970e-01 -1.26767004e+00 1.15714359e+00 -2.39533052e-01 -2.42305249e-02 3.32074225e-01 -1.08263828e-01 3.39007705e-01 1.76833756e-02 2.73748577e-01 7.48545647e-01 7.27170348e-01 -2.13926792e+00 -1.72181740e-01 -5.97037375e-01 4.64411765e-01 3.81800711e-01 3.86670917e-01 -7.81242371e-01 -4.80527848e-01 -4.41474587e-01 5.46482861e-01 -9.12389278e-01 -6.36672378e-01 9.18845981e-02 -7.97550499e-01 4.95928645e-01 7.68799841e-01 -4.44518983e-01 2.13982314e-01 -2.13866186e+00 3.44991654e-01 1.54886216e-01 2.42018595e-01 -6.49246335e-01 8.27328861e-02 -2.54072528e-02 4.46347922e-01 6.56955987e-02 -7.19618082e-01 -1.04878473e+00 9.28659439e-02 6.02338433e-01 -5.45384109e-01 2.86363214e-01 -1.82554662e-01 1.07498574e+00 -1.09509158e+00 -3.09348524e-01 7.11884916e-01 8.04687977e-01 -8.14310253e-01 7.91691601e-01 -6.12082958e-01 9.97940481e-01 -6.09070957e-01 6.63277507e-01 7.63903141e-01 -2.87118256e-01 -1.00434177e-01 3.15478086e-01 3.76968265e-01 2.92118073e-01 -9.14156377e-01 2.36960053e+00 -8.23152781e-01 2.59895891e-01 -1.26029551e-01 -7.09760666e-01 1.18906319e+00 -7.72365462e-03 5.42737365e-01 -4.48122740e-01 5.35437502e-02 -1.40408546e-01 -1.03061950e+00 -3.61018389e-01 5.74800551e-01 -1.79534629e-01 -5.77049017e-01 4.47764248e-01 9.92380679e-02 -1.51117122e+00 -4.69395667e-01 2.98594743e-01 9.71288919e-01 5.19586861e-01 -7.69555569e-02 1.98804557e-01 1.16483070e-01 -1.22152738e-01 1.44973218e-01 7.34904826e-01 1.18695050e-01 9.88844872e-01 1.03277504e-01 -5.32043934e-01 -1.46205556e+00 -1.94875193e+00 1.77185293e-02 3.41834247e-01 3.31959456e-01 -3.21189240e-02 -9.01498854e-01 -8.36681962e-01 2.27953084e-02 7.22235143e-01 -8.37801456e-01 1.21838704e-01 -3.40717375e-01 -3.03130835e-01 8.05286989e-02 6.18071377e-01 7.45904386e-01 -8.54646504e-01 -7.85269201e-01 -1.59725413e-01 -3.45358104e-01 -1.51258647e+00 -5.05430758e-01 2.35462815e-01 -1.04500329e+00 -1.01142108e+00 -5.77421606e-01 -8.61953735e-01 1.31298351e+00 4.72153842e-01 1.44277263e+00 -4.19867933e-01 -4.32300568e-01 1.07834828e+00 -1.10253781e-01 -4.45372045e-01 -5.13044357e-01 -2.59085447e-01 -5.33833504e-02 -1.35615245e-01 2.13734414e-02 -7.71702528e-01 -6.01939797e-01 4.64401096e-02 -1.00156629e+00 5.34507275e-01 2.07189340e-02 4.72269148e-01 1.19408357e+00 3.08724046e-01 -2.79564053e-01 -1.04667914e+00 5.23350481e-03 -4.01025027e-01 -8.03724408e-01 -1.48855343e-01 -1.13129973e-01 -3.02334838e-02 8.57785821e-01 -2.52779424e-01 -1.10878265e+00 4.87976879e-01 1.30509198e-01 -7.37979531e-01 -7.40571797e-01 -4.27052736e-01 -5.63766778e-01 3.00890714e-01 4.68248576e-01 6.22382343e-01 -3.76797050e-01 -3.10105741e-01 6.15561962e-01 1.72371581e-01 6.17862642e-01 -9.19052780e-01 1.33508027e+00 9.55223083e-01 -2.49864273e-02 -4.59200531e-01 -1.16345358e+00 -2.80423611e-01 -1.06066167e+00 -7.53357857e-02 1.30555761e+00 -1.22679365e+00 -4.33565289e-01 4.58538741e-01 -1.22094905e+00 -1.00789988e+00 -7.64432490e-01 1.94701463e-01 -9.74900603e-01 -2.15419069e-01 -3.79954696e-01 -8.37891877e-01 2.22314999e-01 -1.19239914e+00 1.73490036e+00 4.60819138e-04 1.10213459e-01 -1.25012064e+00 8.64279643e-02 6.97243333e-01 1.41520411e-01 8.70182216e-01 8.25565159e-01 2.16319904e-01 -1.36928713e+00 1.52583897e-01 -5.42324372e-02 5.47660470e-01 5.58511376e-01 -4.63438690e-01 -1.31505060e+00 -6.42165840e-02 2.92961359e-01 -3.46910298e-01 4.46723998e-01 6.03106141e-01 1.86111581e+00 -4.55312818e-01 -2.17232972e-01 1.26011014e+00 1.67912626e+00 3.89999181e-01 5.37754655e-01 4.35590409e-02 1.20332682e+00 5.25976837e-01 1.58385009e-01 4.10082817e-01 6.72274947e-01 3.21640402e-01 8.15472007e-01 -2.37157851e-01 -1.23431385e-01 -1.01383257e+00 2.24627614e-01 5.96788347e-01 3.89941372e-02 -2.14806005e-01 -7.72287190e-01 5.43698430e-01 -1.13471663e+00 -8.36243868e-01 3.49906683e-01 2.01713443e+00 6.22777820e-01 1.04588933e-01 -3.94193903e-02 9.89058893e-03 1.62919328e-01 3.40029269e-01 -9.41195369e-01 -3.36479783e-01 -6.23100884e-02 2.45607987e-01 5.61520159e-01 6.58790588e-01 -8.86325121e-01 6.99249685e-01 6.08641768e+00 1.12339497e-01 -7.32013285e-01 1.67753808e-02 8.85905623e-01 -3.16792935e-01 -9.31066036e-01 -1.95225880e-01 -7.34482110e-01 3.63441139e-01 4.71249253e-01 2.00991690e-01 8.70779455e-01 1.22679341e+00 1.02468822e-02 -1.56443212e-02 -1.31758118e+00 1.21992588e+00 3.68965000e-01 -1.30010784e+00 2.45976478e-01 3.08143705e-01 1.36713970e+00 -6.95790797e-02 1.38635248e-01 1.01311773e-01 7.51399457e-01 -1.28170013e+00 9.97475505e-01 5.12978852e-01 9.75556016e-01 -4.87316072e-01 8.38669762e-02 5.01747847e-01 -1.20736957e+00 7.52606764e-02 -2.86317527e-01 -6.08823262e-03 1.69301555e-01 6.07079387e-01 -1.00154948e+00 3.12660933e-01 8.94423246e-01 6.15090072e-01 -4.41975534e-01 5.79709709e-01 -5.12228489e-01 2.28674695e-01 -2.85328060e-01 2.48783380e-01 1.15796819e-01 -3.27607185e-01 2.43257508e-01 7.49702692e-01 4.68188524e-01 4.43286300e-02 3.97866756e-01 1.66055906e+00 -3.70556414e-01 -3.76320690e-01 -1.31048596e+00 3.51251870e-01 5.24517953e-01 9.12683725e-01 -8.88031781e-01 -4.07729626e-01 -2.02848062e-01 1.35483193e+00 6.52629212e-02 5.83382905e-01 -9.67740774e-01 -5.59603348e-02 4.77761358e-01 4.05273795e-01 2.20061734e-01 -7.27782190e-01 -5.78397393e-01 -1.11252713e+00 4.72901091e-02 -3.78718048e-01 -2.58567154e-01 -1.34430194e+00 -1.10883880e+00 5.67569554e-01 1.50378704e-01 -1.10625708e+00 -5.99960871e-02 -5.51432848e-01 -2.86191136e-01 7.61336744e-01 -1.26607239e+00 -1.21032572e+00 -1.00249481e+00 8.81506205e-01 7.78390169e-01 2.85923749e-01 1.06407344e+00 -1.18348919e-01 -1.84531845e-02 4.93577197e-02 -3.26870568e-02 2.45676026e-01 7.60982782e-02 -1.67609096e+00 7.63952613e-01 6.48586035e-01 2.04922140e-01 3.64309430e-01 2.26914525e-01 -4.27043408e-01 -1.37826073e+00 -1.61697519e+00 2.57649630e-01 -1.21651506e+00 -9.60540548e-02 -1.19751394e+00 -3.95527750e-01 1.09728456e+00 -1.31581485e-01 2.07646519e-01 5.24142623e-01 -4.39466298e-01 -3.25742900e-01 -3.45010613e-03 -1.49129009e+00 4.55033123e-01 1.58638096e+00 -7.17533767e-01 -5.09810627e-01 3.58829498e-01 1.24740565e+00 -1.05840015e+00 -8.03244352e-01 2.38595709e-01 1.42572492e-01 -1.03726089e+00 1.40538311e+00 -4.84754611e-03 7.96532989e-01 -4.53163654e-01 -6.49418175e-01 -1.43925023e+00 6.99054673e-02 -1.10713243e-01 -1.10464513e-01 7.69799650e-01 -1.92578156e-02 -2.91290909e-01 1.05163646e+00 7.81831503e-01 -4.26434606e-01 -6.14360869e-01 -6.84959173e-01 -6.07414007e-01 -1.93516091e-02 -5.81076145e-01 1.04466295e+00 6.22453094e-01 -1.12489986e+00 1.67852223e-01 5.49758272e-03 3.24850202e-01 1.04631078e+00 6.29272461e-01 1.42818153e+00 -8.81979346e-01 -2.69422382e-01 5.38673326e-02 -3.86948198e-01 -1.58177197e+00 3.07206541e-01 -8.84200454e-01 2.50119716e-01 -1.99326372e+00 3.20356190e-01 -7.75314987e-01 2.12484404e-01 2.59482592e-01 1.03979968e-01 2.75650322e-01 2.18034126e-02 -1.48254335e-01 -4.52951312e-01 8.13281894e-01 1.84156024e+00 -3.82821143e-01 -3.83232325e-01 -5.76109476e-02 -4.41205829e-01 6.73882306e-01 5.64887106e-01 -2.74475664e-01 -9.16999817e-01 -7.19679594e-01 -9.41362530e-02 4.58705947e-02 8.81563008e-01 -1.03417790e+00 -3.46357316e-01 -4.87829417e-01 9.15999234e-01 -8.32111955e-01 8.72067750e-01 -1.08515024e+00 2.46196613e-01 1.56049162e-01 -7.68419504e-02 -1.60296127e-01 2.24453866e-01 5.27258456e-01 1.91590086e-01 3.91119123e-01 5.04780829e-01 -7.85585880e-01 -5.15324056e-01 7.00897515e-01 -3.41922529e-02 1.40840396e-01 1.09516013e+00 -6.27047420e-01 2.79641966e-03 -3.18516433e-01 -6.06853783e-01 -6.91934377e-02 1.18130624e+00 5.09873271e-01 1.09484315e+00 -1.44888043e+00 -1.99746862e-01 7.80753970e-01 7.12019503e-02 1.30506647e+00 4.04837161e-01 -3.13807905e-01 -6.45355821e-01 4.25843626e-01 -8.49212930e-02 -9.64380026e-01 -7.35948920e-01 7.59800732e-01 5.71460962e-01 6.32869229e-02 -7.32315898e-01 9.44617033e-01 1.01810944e+00 -1.10830879e+00 1.02126770e-01 -8.77131224e-01 5.32022417e-01 -8.00795794e-01 -8.91752483e-04 -7.11095110e-02 -3.43722224e-01 -5.12518942e-01 -6.67752251e-02 7.22124696e-01 7.66425550e-01 -6.60821721e-02 1.41831291e+00 -2.32312262e-01 -4.58566919e-02 9.27109838e-01 1.29400134e+00 2.91805804e-01 -1.86647594e+00 -3.70967574e-02 -8.93392503e-01 -8.79052401e-01 -2.43954778e-01 -5.77404618e-01 -1.11852193e+00 8.11632276e-01 4.04877543e-01 -1.31578997e-01 1.10966396e+00 3.68252903e-01 8.69054675e-01 9.80744287e-02 9.14675474e-01 -6.41216695e-01 6.13276541e-01 4.14078832e-01 7.65913367e-01 -1.19228160e+00 -2.05659389e-01 -3.82731974e-01 -3.13269079e-01 6.84339404e-01 7.84819663e-01 -2.71023601e-01 6.00404561e-01 4.11187559e-01 -1.39233157e-01 -2.45147645e-01 -1.73282698e-01 1.55617222e-01 1.68172330e-01 8.89041841e-01 -4.33485350e-03 3.83021951e-01 1.00067806e+00 2.81105489e-01 -6.85099483e-01 -2.94958115e-01 4.80144769e-01 7.67917275e-01 -1.61341697e-01 -8.90906036e-01 -7.52273321e-01 2.30297118e-01 1.34704739e-01 6.74187914e-02 -1.47860765e-01 4.52478260e-01 4.36834008e-01 5.96598208e-01 3.66688251e-01 -4.54176664e-01 4.67303544e-01 -2.38675326e-02 9.67373252e-01 -1.12096345e+00 9.00886580e-02 -2.86412746e-01 -6.04722023e-01 -7.22925186e-01 -4.72949386e-01 -5.21484673e-01 -1.32015514e+00 -1.87600672e-01 3.87536496e-01 -2.00322390e-01 9.46773350e-01 5.51166594e-01 2.07242325e-01 8.18009853e-01 9.95710969e-01 -1.30006981e+00 1.88936546e-01 -5.09643674e-01 -6.57542467e-01 7.06557572e-01 5.53589165e-01 -6.63142264e-01 -4.43502516e-01 5.06447077e-01]
[9.07532024383545, -3.217766523361206]
21bda1c6-8d1d-421f-b10d-a13cd729a768
kuaipedia-a-large-scale-multi-modal-short
2211.00732
null
https://arxiv.org/abs/2211.00732v2
https://arxiv.org/pdf/2211.00732v2.pdf
Kuaipedia: a Large-scale Multi-modal Short-video Encyclopedia
Online encyclopedias, such as Wikipedia, have been well-developed and researched in the last two decades. One can find any attributes or other information of a wiki item on a wiki page edited by a community of volunteers. However, the traditional text, images and tables can hardly express some aspects of an wiki item. For example, when we talk about ``Shiba Inu'', one may care more about ``How to feed it'' or ``How to train it not to protect its food''. Currently, short-video platforms have become a hallmark in the online world. Whether you're on TikTok, Instagram, Kuaishou, or YouTube Shorts, short-video apps have changed how we consume and create content today. Except for producing short videos for entertainment, we can find more and more authors sharing insightful knowledge widely across all walks of life. These short videos, which we call knowledge videos, can easily express any aspects (e.g. hair or how-to-feed) consumers want to know about an item (e.g. Shiba Inu), and they can be systematically analyzed and organized like an online encyclopedia. In this paper, we propose Kuaipedia, a large-scale multi-modal encyclopedia consisting of items, aspects, and short videos lined to them, which was extracted from billions of videos of Kuaishou (Kwai), a well-known short-video platform in China. We first collected items from multiple sources and mined user-centered aspects from millions of users' queries to build an item-aspect tree. Then we propose a new task called ``multi-modal item-aspect linking'' as an expansion of ``entity linking'' to link short videos into item-aspect pairs and build the whole short-video encyclopedia. Intrinsic evaluations show that our encyclopedia is of large scale and highly accurate. We also conduct sufficient extrinsic experiments to show how Kuaipedia can help fundamental applications such as entity typing and entity linking.
['Bing Qin', 'Zhongyuan Wang', 'Yangqiu Song', 'Ming Liu', 'Ruiji Fu', 'Zepeng Zhai', 'Yuzhou Zhang', 'Haojie Pan']
2022-10-28
null
null
null
null
['entity-typing']
['natural-language-processing']
[-5.30278981e-01 -1.98746607e-01 -6.48068190e-01 -5.38478121e-02 -8.28829765e-01 -9.19476390e-01 3.32786977e-01 2.31698781e-01 -3.10077101e-01 7.82779157e-01 6.17741883e-01 1.38187066e-01 -5.32260314e-02 -1.02823818e+00 -1.03765202e+00 -1.90243602e-01 1.78742297e-02 6.89450651e-02 1.14618629e-01 -4.86605436e-01 4.24676389e-03 -4.45390850e-01 -1.59941661e+00 4.74925369e-01 7.72809267e-01 9.85316932e-01 3.80931437e-01 3.96398455e-01 -4.60797787e-01 9.04442787e-01 -6.54720485e-01 -1.06252873e+00 -1.81932807e-01 -3.88295323e-01 -7.63675511e-01 -9.63957049e-03 4.79035407e-01 -3.96486908e-01 -5.28381944e-01 1.29134703e+00 2.47418359e-01 -2.35869586e-01 4.15129244e-01 -1.53737342e+00 -1.18608880e+00 1.13264298e+00 -4.87237632e-01 -1.55813739e-01 9.85665858e-01 5.45020727e-03 1.21524668e+00 -6.84884906e-01 1.28717172e+00 7.98002839e-01 8.28592837e-01 -1.18938107e-02 -4.36861455e-01 -7.75549948e-01 1.05296724e-01 5.14715075e-01 -1.33960462e+00 1.66163612e-02 4.36178684e-01 -5.38895130e-01 4.72643048e-01 4.44932818e-01 1.14835298e+00 1.11240363e+00 -6.79913163e-02 9.25759792e-01 6.42067671e-01 -2.84967404e-02 -4.62775230e-01 3.69453549e-01 -1.74135081e-02 7.37375319e-01 6.33423507e-01 -5.95138311e-01 -8.26524258e-01 1.36307552e-01 7.21005857e-01 1.91064328e-01 -5.59385896e-01 -1.58434898e-01 -1.66425586e+00 6.23560548e-01 3.02977115e-01 2.03222632e-01 -2.67427266e-01 7.22576752e-02 7.71216869e-01 3.31708044e-01 3.56823027e-01 3.94375771e-01 -4.54588711e-01 -4.63337123e-01 -2.86606401e-01 2.46084109e-01 1.12123656e+00 1.78287840e+00 9.98907804e-01 -4.33009624e-01 1.46364018e-01 1.04072142e+00 4.50182520e-02 6.23758137e-01 3.06398809e-01 -9.78594780e-01 7.89242923e-01 8.85807037e-01 1.84327260e-01 -1.34132195e+00 -1.10693388e-01 -7.41166621e-02 -6.89782441e-01 -7.11327195e-01 3.76136720e-01 -4.93868321e-01 -3.21694851e-01 1.42159522e+00 3.86038899e-01 -6.97987601e-02 -2.68009305e-01 1.02222419e+00 1.47571731e+00 7.69455492e-01 4.34916094e-02 -2.48763725e-01 1.95925713e+00 -9.98474240e-01 -1.19782495e+00 2.89243996e-01 7.23877788e-01 -9.84810233e-01 1.23635054e+00 3.71973485e-01 -9.79488969e-01 -4.66552466e-01 -9.27150249e-01 -2.88044959e-01 -9.12471771e-01 3.97700928e-02 6.18449688e-01 1.33230790e-01 -3.96085501e-01 4.42460448e-01 -1.87843010e-01 -6.67811155e-01 4.24248993e-01 -2.29863554e-01 -9.77242053e-01 -3.79649550e-01 -1.62256277e+00 5.58588624e-01 4.98746991e-01 -3.36468846e-01 -4.66917753e-01 -8.40470672e-01 -1.09919369e+00 -2.31327668e-01 1.08727372e+00 -6.42875910e-01 1.08223033e+00 -9.92347479e-01 -9.18752015e-01 8.93920958e-01 1.94253698e-01 9.74831581e-02 2.26621345e-01 -6.99951887e-01 -9.65987146e-01 2.80878365e-01 3.98726553e-01 3.66632670e-01 4.75708932e-01 -8.07149827e-01 -1.09244192e+00 -9.36590433e-02 7.86171556e-01 2.21226543e-01 -7.68329024e-01 2.87282944e-01 -1.18860209e+00 -1.00025153e+00 -5.20964921e-01 -7.03231871e-01 5.10594308e-01 1.46122605e-01 -6.28180683e-01 5.51027991e-03 7.87304401e-01 -1.05926907e+00 1.93736768e+00 -2.07199311e+00 1.87321037e-01 -1.35780975e-01 3.39504898e-01 5.57009988e-02 -2.56150514e-02 7.96796560e-01 2.37757891e-01 3.90581399e-01 -2.13131085e-02 4.60705072e-01 9.98738408e-02 2.11799726e-01 2.48106003e-01 2.20909104e-01 -4.23487127e-01 9.55831170e-01 -1.36077559e+00 -6.56593919e-01 -3.58352542e-01 4.03437495e-01 -4.86547142e-01 1.41613558e-01 -2.89214849e-01 -1.38125524e-01 -4.16882277e-01 8.13597500e-01 2.70319402e-01 -4.61851358e-01 1.64991528e-01 -8.04781079e-01 -3.29526901e-01 2.34252706e-01 -1.12609172e+00 1.95310175e+00 -4.90418553e-01 7.29722798e-01 3.13211083e-02 -6.57578170e-01 4.65014994e-01 3.72638673e-01 5.01524866e-01 -4.31262136e-01 2.30071247e-02 1.47894487e-01 -5.02038717e-01 -9.91160452e-01 7.32254982e-01 4.86721545e-01 -6.43756688e-01 4.68240738e-01 2.47949153e-01 1.79630905e-01 6.58614099e-01 7.48625398e-01 9.20138121e-01 4.75265443e-01 6.65305912e-01 -1.28079340e-01 3.47597599e-01 4.77193326e-01 4.26813036e-01 3.22013795e-01 1.50834709e-01 5.61158061e-01 6.13277614e-01 -3.72924954e-01 -1.17618299e+00 -7.40390003e-01 -6.90662488e-02 1.16878867e+00 5.06137073e-01 -1.17941809e+00 -7.58023083e-01 -7.13158488e-01 6.40195087e-02 2.95421660e-01 -5.74190915e-01 2.04570621e-01 -4.53634828e-01 -3.16943228e-01 2.56663978e-01 3.98094088e-01 9.49679792e-01 -9.25752878e-01 -6.30258396e-02 4.09385741e-01 -8.65115941e-01 -1.08437335e+00 -1.24378443e+00 -3.71529043e-01 -4.46714126e-02 -1.47836709e+00 -8.85941803e-01 -9.40899312e-01 4.63686705e-01 6.00659609e-01 1.44470620e+00 1.49532676e-01 -1.83855742e-01 5.86343706e-01 -8.76244426e-01 -3.48928601e-01 1.07910465e-02 6.91326782e-02 1.65829122e-01 -1.01318076e-01 5.48306882e-01 -3.98561120e-01 -5.28713942e-01 8.55554879e-01 -9.88704562e-01 1.72004879e-01 4.48876083e-01 5.82950830e-01 7.13306606e-01 1.33971795e-01 6.18499458e-01 -1.51886129e+00 3.33910793e-01 -1.23775434e+00 -7.87085891e-02 4.23490137e-01 -1.73315361e-01 -3.92840773e-01 7.15550900e-01 -6.24467969e-01 -7.10121036e-01 -5.07889450e-01 -4.93813977e-02 -2.25445747e-01 9.16574746e-02 1.12614071e+00 -5.77164292e-01 2.13430852e-01 2.41677061e-01 1.07246011e-01 -2.69864142e-01 -5.97727537e-01 5.87290645e-01 8.62248123e-01 7.55332947e-01 -4.94064510e-01 8.65941942e-01 1.70976982e-01 -5.50450325e-01 -7.03031361e-01 -1.06988263e+00 -8.43220890e-01 -3.52647513e-01 -6.64368153e-01 9.07661736e-01 -1.24709141e+00 -9.18736517e-01 5.85529387e-01 -9.24273908e-01 1.01141207e-01 -1.19227812e-01 3.14051002e-01 -3.28536510e-01 3.14345002e-01 -6.78901494e-01 1.40425563e-02 -7.34589174e-02 -5.52782476e-01 5.41088820e-01 3.24183404e-01 -1.13718666e-01 -9.25649226e-01 -7.46849701e-02 4.41795766e-01 3.57831448e-01 1.40759826e-01 6.94934785e-01 -5.64851522e-01 -5.91291428e-01 -2.35292003e-01 -5.84339380e-01 1.94321752e-01 3.48408967e-01 9.99243036e-02 -3.10564756e-01 7.35374913e-02 -5.37710428e-01 -2.51638651e-01 3.98602009e-01 -1.52937338e-01 1.21377313e+00 -1.11387587e+00 -2.67889410e-01 5.88416100e-01 1.55412042e+00 5.41927814e-02 8.13912511e-01 7.69635141e-01 1.06471646e+00 4.98098165e-01 1.01923263e+00 4.49399889e-01 9.25484419e-01 8.58294129e-01 3.55969578e-01 3.93660277e-01 -1.55246601e-01 -7.71666348e-01 4.97052819e-01 1.68409467e+00 -4.60227460e-01 -2.32274726e-01 -5.19896686e-01 7.98731983e-01 -1.85976171e+00 -1.26350212e+00 -2.54522681e-01 2.05112720e+00 9.23505068e-01 -1.75811142e-01 4.73230511e-01 -3.42031181e-01 8.50982785e-01 6.52116686e-02 -4.01756525e-01 1.50262892e-01 -2.48868674e-01 -5.65503657e-01 5.18746197e-01 -9.98478830e-02 -1.01974487e+00 4.63056833e-01 4.69021320e+00 1.20376241e+00 -5.85034072e-01 3.62435549e-01 1.23303451e-01 -1.95052996e-02 -5.58386803e-01 -1.48792103e-01 -9.57610309e-01 9.53347266e-01 7.26531267e-01 -4.90928292e-01 4.04971689e-01 1.01113403e+00 -1.56902075e-01 -5.79134189e-02 -8.22723866e-01 1.19411457e+00 3.95145327e-01 -1.70474184e+00 2.25123540e-01 -1.40900880e-01 7.90738344e-01 -3.01784486e-01 -2.29790211e-01 5.07943332e-01 1.59168422e-01 -5.67584515e-01 9.08136189e-01 4.85107690e-01 1.26636863e+00 -7.49098897e-01 7.75269270e-01 2.23697007e-01 -1.75259423e+00 2.55290508e-01 -4.15112138e-01 3.27129751e-01 3.12742740e-01 5.35824120e-01 1.43563867e-01 9.60017145e-01 1.06447637e+00 1.36911213e+00 -3.66556704e-01 1.35997212e+00 -3.68153363e-01 3.52643669e-01 -1.12471536e-01 -3.31612080e-01 7.91974217e-02 -4.28268403e-01 5.24342000e-01 1.37145233e+00 7.36768961e-01 4.51530010e-01 -1.17394840e-02 2.86048383e-01 -7.85699904e-01 3.96822274e-01 -9.21825111e-01 -3.65249246e-01 6.82926476e-01 1.43189681e+00 -2.68649369e-01 -4.81601447e-01 -1.03065062e+00 9.04248118e-01 3.50021809e-01 1.78661019e-01 -1.03872633e+00 -8.37797046e-01 8.28971624e-01 5.24487078e-01 2.84138352e-01 3.55306938e-02 4.61057961e-01 -1.64652812e+00 3.29026356e-02 -9.53264892e-01 5.83036244e-01 -1.12673604e+00 -1.48279285e+00 5.39425850e-01 8.44809487e-02 -1.77359807e+00 1.52503595e-01 -3.71899903e-01 -3.74491096e-01 2.42162213e-01 -1.35551083e+00 -1.23525453e+00 -5.81486583e-01 6.55156970e-01 5.84335685e-01 -3.78597826e-02 3.93516183e-01 9.67164934e-01 -5.07391274e-01 5.99316418e-01 1.37833714e-01 3.82263303e-01 1.04197824e+00 -1.01981330e+00 -7.28331553e-03 5.71065843e-01 2.22554103e-01 6.39834642e-01 5.40384531e-01 -8.64954710e-01 -1.74937284e+00 -1.29641604e+00 9.52747703e-01 -6.39848650e-01 1.19643092e+00 -2.30465353e-01 -9.19474483e-01 1.05510426e+00 4.63352591e-01 -2.74736971e-01 8.92599285e-01 1.06453873e-01 -4.37578529e-01 -1.20036721e-01 -7.30406165e-01 3.65491062e-01 1.50841641e+00 -6.15051746e-01 -5.78853667e-01 6.09978914e-01 9.97830510e-01 -4.18246448e-01 -1.27846086e+00 1.74794104e-02 8.84101808e-01 -7.06402123e-01 9.92974162e-01 -7.12810099e-01 8.11609983e-01 -3.24929655e-01 -1.57593742e-01 -1.48148263e+00 -2.73914099e-01 -7.84090102e-01 -2.72863984e-01 1.74834824e+00 4.49953139e-01 -3.28723848e-01 5.44948041e-01 2.36218080e-01 -3.41433078e-01 -5.78827262e-01 -2.88396895e-01 -7.98657238e-01 -4.77592319e-01 -5.25882542e-01 8.38039219e-01 1.16075122e+00 3.55188936e-01 2.43635535e-01 -7.74480581e-01 -1.05579384e-01 4.23258722e-01 6.64533749e-02 9.26146328e-01 -9.73647892e-01 -2.57105887e-01 -1.52798221e-01 -3.47270072e-01 -1.11812532e+00 -2.92604923e-01 -8.04427803e-01 -3.00479174e-01 -1.59662735e+00 7.29439735e-01 -3.92999262e-01 1.05329499e-01 5.03119230e-01 -2.24903062e-01 4.46708709e-01 2.05320850e-01 4.73364472e-01 -1.07666111e+00 3.69365700e-02 1.52863634e+00 -2.44221389e-01 1.17567718e-01 -4.51057523e-01 -9.99633908e-01 7.51800895e-01 3.09545398e-01 -3.15632969e-01 -1.18601628e-01 -4.18398410e-01 1.19914567e+00 8.21064562e-02 1.00515902e-01 -7.73825347e-01 2.49514982e-01 -2.23238289e-01 -6.89995885e-02 -4.88258302e-01 9.95684043e-02 -9.01209176e-01 7.66328216e-01 6.65020719e-02 -8.66112113e-02 1.57706216e-01 -1.31750509e-01 6.03319049e-01 -6.14369154e-01 -2.11772382e-01 2.29121208e-01 -2.97163874e-01 -1.14640164e+00 5.57541430e-01 -2.23011106e-01 3.57215911e-01 1.10367787e+00 -5.31813800e-02 -9.67275381e-01 -7.52621770e-01 -6.61933720e-01 4.67815042e-01 6.10976219e-01 4.36490506e-01 3.96208405e-01 -1.76138377e+00 -6.36066020e-01 -3.27642351e-01 6.77233160e-01 -2.59303749e-01 5.65109372e-01 8.67619097e-01 -7.68985569e-01 3.25933635e-01 -4.50256079e-01 -3.23945545e-02 -1.13010573e+00 6.93634450e-01 -3.82617176e-01 -1.24554314e-01 -6.97410703e-01 7.78409719e-01 2.72304505e-01 -2.70610482e-01 1.42577365e-01 -1.53913289e-01 -5.45008540e-01 7.02462196e-01 8.36573482e-01 4.11584377e-01 -3.31931025e-01 -8.06230485e-01 -1.33966759e-01 7.74020910e-01 6.02205247e-02 5.00323534e-01 1.42580521e+00 -6.46016598e-01 -2.02726483e-01 5.95416188e-01 1.56599092e+00 1.14181399e-01 -7.62441218e-01 -1.58728570e-01 -4.87145483e-01 -6.02281928e-01 -3.98661017e-01 -8.29021394e-01 -1.34375501e+00 4.74947423e-01 -1.39307022e-01 4.50207531e-01 1.05003011e+00 2.27949336e-01 1.42703140e+00 3.81545901e-01 8.39049816e-01 -1.15778768e+00 -4.08699028e-02 1.97542205e-01 9.85048532e-01 -1.22806585e+00 1.12255782e-01 -5.85058510e-01 -8.17728102e-01 1.07277751e+00 5.58077753e-01 1.75531149e-01 8.07396591e-01 -7.52992630e-02 -5.74479885e-02 -3.33690912e-01 -6.33748114e-01 -2.38019973e-01 3.40211779e-01 6.81189239e-01 1.02006026e-01 1.73148036e-01 -4.33617324e-01 1.13064528e+00 -2.34822765e-01 1.17705949e-01 8.63310993e-01 7.06794798e-01 -5.55912852e-01 -9.56299305e-01 -1.35457024e-01 8.35423708e-01 -5.66106975e-01 -1.68738719e-02 8.80372822e-02 1.14795935e+00 3.00738245e-01 4.72056568e-01 -1.29102960e-01 -5.20921350e-01 3.47642154e-01 -3.07887733e-01 6.73478916e-02 -5.81739128e-01 -6.89001739e-01 -1.68213323e-01 6.43825293e-01 -7.08089888e-01 -7.09542096e-01 -5.91389835e-01 -8.58219564e-01 -7.54832149e-01 -2.66448200e-01 4.37647671e-01 7.68904626e-01 6.83015406e-01 1.85528144e-01 2.57404417e-01 3.89989465e-01 -5.82877517e-01 4.67460722e-01 -6.37836874e-01 -8.01639616e-01 7.26470649e-01 -2.11079881e-01 -7.70083010e-01 -1.39143273e-01 4.99796450e-01]
[10.124886512756348, 0.8491062521934509]
609a1c76-ae33-4792-b2f2-ad50b2066262
a-parallel-english-serbian-bulgarian
null
null
https://aclanthology.org/2022.clib-1.17
https://aclanthology.org/2022.clib-1.17.pdf
A Parallel English - Serbian - Bulgarian - Macedonian Lexicon of Named Entities
This paper describes the creation of a parallel multilingual lexicon of named entities from English to three South Slavic languages: Serbian, Bulgarian and Macedonian, with Wikipedia as a source. The basics of the proposed methodology are well known. This methodology provides a cheap opportunity to build multilingual lexicons, without having expertise in target languages. Wikipedia’s database dump can be freely downloaded in SQL and XML formats. The method presented here has been used to build a Python application that extracts the English – Serbian – Bulgarian – Macedonian parallel titles from Wikipedia and classifies them using the English Wikipedia category system. The extracted named entity sets have been classified into five classes: PERSON, ORGANIZATION, LOCATION, PRODUCT, and MISC (miscellaneous). It has been achieved using Wikipedia metadata. The quality of classification has been checked manually on 1,000 randomly chosen named entities. The following are the results obtained: 97% for precision and 90% for recall.
['Aleksandar Petrovski']
null
null
null
null
clib-2022-9
['miscellaneous']
['miscellaneous']
[-6.96953237e-01 1.40656427e-01 -1.33166447e-01 -2.12478593e-01 -5.26824653e-01 -8.93415332e-01 8.62352550e-01 5.88119149e-01 -1.02634752e+00 1.45506394e+00 2.09374189e-01 -2.58284599e-01 -5.46899438e-02 -1.02087522e+00 -3.51425260e-01 -1.28300935e-01 2.00612262e-01 8.01350713e-01 2.94942170e-01 -4.08251673e-01 4.16720212e-01 5.76365530e-01 -1.53847289e+00 -1.19921476e-01 1.11425936e+00 3.68911654e-01 2.34982207e-01 4.16283667e-01 -5.06102562e-01 8.40684116e-01 -5.89330792e-01 -8.46895576e-01 2.08600730e-01 -6.88515278e-03 -1.02417779e+00 -2.85619497e-01 1.35470301e-01 1.81610271e-01 5.24828210e-02 1.04674840e+00 3.88160825e-01 -1.26249984e-01 9.14801836e-01 -8.44295621e-01 -3.78641248e-01 6.46704316e-01 -5.38546182e-02 -1.49205342e-01 5.77163279e-01 -2.99651474e-01 7.70679712e-01 -1.02794158e+00 1.31857979e+00 8.77760410e-01 7.64488339e-01 -2.20331158e-02 -7.39546657e-01 -6.56603277e-01 -5.44934034e-01 3.34079340e-02 -1.98856747e+00 -2.28109688e-01 -1.07217535e-01 -8.25334430e-01 1.09252405e+00 5.73640801e-02 4.55384493e-01 2.75610059e-01 2.46670917e-01 1.34712130e-01 1.30271864e+00 -8.76120031e-01 5.25778867e-02 1.17392945e+00 3.24566782e-01 5.38330615e-01 9.88685489e-01 -3.42912942e-01 -4.42507207e-01 -1.53897822e-01 4.34313536e-01 -5.02774298e-01 1.61320511e-02 -4.37051922e-01 -1.09399390e+00 6.30516589e-01 -8.59532878e-02 8.50519121e-01 -5.16555429e-01 -5.16701400e-01 6.67396128e-01 3.00486416e-01 2.93775350e-01 5.16838789e-01 -8.24391127e-01 -2.05449715e-01 -9.34728324e-01 3.18523139e-01 1.25808799e+00 1.62376153e+00 8.79105747e-01 -3.08266073e-01 5.70628226e-01 1.00709283e+00 5.65199018e-01 7.39640236e-01 5.41798651e-01 -2.22593412e-01 6.26997411e-01 1.03090036e+00 6.02738142e-01 -1.01696312e+00 -4.52942222e-01 2.98204962e-02 -2.17323735e-01 3.01834885e-02 4.67614353e-01 -4.38929170e-01 -8.32353234e-01 1.21739495e+00 4.76547986e-01 -6.85896158e-01 6.14739120e-01 4.50982749e-01 1.27031934e+00 5.55389583e-01 4.89049762e-01 -2.33858496e-01 1.55273974e+00 -4.93413836e-01 -8.89446080e-01 3.42108816e-01 4.82029796e-01 -1.30538118e+00 4.39412862e-01 1.64517432e-01 -9.91010606e-01 -4.65788096e-01 -8.40862095e-01 -5.09619713e-03 -1.26555049e+00 5.47831655e-01 4.79116052e-01 1.01902223e+00 -8.96605372e-01 2.90781170e-01 -5.95961392e-01 -9.69458878e-01 -2.88308173e-01 3.08242828e-01 -8.31691146e-01 3.60351413e-01 -1.48907578e+00 1.34316659e+00 1.10052204e+00 -2.61244386e-01 -3.03864479e-01 -2.55129486e-01 -1.07191479e+00 -2.82298446e-01 1.55833095e-01 -3.41134369e-01 7.98738778e-01 -6.99631333e-01 -1.05360210e+00 1.40508044e+00 1.63970366e-01 -3.10814083e-01 6.12010598e-01 2.31148582e-02 -1.02350736e+00 5.48322313e-02 8.61444652e-01 2.04619229e-01 -7.70297721e-02 -9.45445001e-01 -1.22950041e+00 -3.63152504e-01 3.57566364e-02 2.46244594e-01 -2.52192706e-01 4.54249710e-01 -6.21812582e-01 -7.35878646e-01 -1.51241589e-02 -6.97376549e-01 1.23687126e-01 -8.23382616e-01 -1.40779331e-01 -1.38099238e-01 6.79732934e-02 -1.12580454e+00 1.54230249e+00 -1.63106072e+00 -1.80840611e-01 3.79997998e-01 -1.46328971e-01 2.91922867e-01 5.85392535e-01 1.06519604e+00 9.94714275e-02 5.45873679e-02 2.09558569e-02 4.06315744e-01 1.12109706e-01 -1.27559071e-02 2.26887777e-01 2.63286442e-01 -1.65341347e-01 3.50355625e-01 -9.00555313e-01 -9.55012739e-01 2.23638341e-01 3.15479457e-01 -5.69898970e-02 -2.13425830e-01 2.46002689e-01 1.74822025e-02 -3.36630315e-01 6.41765416e-01 7.76406109e-01 3.72164875e-01 5.94963789e-01 -1.77229881e-01 -7.41735160e-01 4.20727372e-01 -1.56209278e+00 1.32730377e+00 -7.34682262e-01 4.68207538e-01 2.63085701e-02 -3.60710561e-01 1.08694494e+00 6.19145095e-01 3.43814254e-01 -5.27908862e-01 2.09297985e-01 8.68908942e-01 -4.36355412e-01 -7.83590555e-01 1.11729646e+00 -1.58997819e-01 -4.44042742e-01 2.83117294e-01 6.23280764e-01 -3.93700004e-02 1.03642178e+00 2.50419736e-01 3.43746543e-01 3.77101719e-01 1.11594307e+00 -5.61430633e-01 9.61919844e-01 6.52236938e-01 5.26918590e-01 4.21120405e-01 -8.32024366e-02 6.79553626e-03 3.02105278e-01 -2.91427314e-01 -1.42742622e+00 -9.62775767e-01 -4.44663882e-01 8.64318252e-01 -9.59148109e-02 -4.71860021e-01 -7.23286510e-01 -4.97727633e-01 -1.46709397e-01 6.42944157e-01 -1.70118004e-01 4.90784287e-01 -3.72317940e-01 -4.36602652e-01 7.50609159e-01 -9.25004333e-02 5.46440005e-01 -1.26509798e+00 -4.24877673e-01 3.62185687e-01 -2.03288794e-01 -1.00938141e+00 9.53036696e-02 -7.32185394e-02 -4.37469333e-01 -1.19963884e+00 -9.03886914e-01 -1.16199470e+00 3.80573899e-01 -3.81099463e-01 1.17023230e+00 -2.75703073e-01 -2.15052560e-01 4.70704257e-01 -4.56725985e-01 -6.60351515e-01 -6.63850248e-01 5.33083856e-01 1.85099527e-01 -4.70460683e-01 8.61979961e-01 -1.28995568e-01 1.47421127e-02 2.24685863e-01 -7.97120035e-01 -3.70402843e-01 5.03054261e-01 4.60217118e-01 2.39681661e-01 4.50599901e-02 7.98244178e-01 -1.39202690e+00 3.95142347e-01 -6.60824418e-01 -8.04035306e-01 4.38607275e-01 -5.67929864e-01 -1.02330849e-01 5.38172781e-01 2.23497450e-01 -1.27683723e+00 3.46477032e-01 -2.84448296e-01 5.87506890e-01 -4.79353130e-01 8.43907833e-01 -3.62080216e-01 -8.83377343e-02 5.49003065e-01 2.23690465e-01 -3.69656503e-01 -4.55517143e-01 3.37459803e-01 1.16990471e+00 4.11266297e-01 -5.16914129e-01 5.28519332e-01 1.31923914e-01 -5.48865736e-01 -1.09330630e+00 -1.07740581e-01 -1.02514923e+00 -1.17239654e+00 -3.51199836e-01 9.09592330e-01 -1.20220661e+00 -3.86826605e-01 5.81015170e-01 -9.82865930e-01 2.78099924e-01 -4.91221473e-02 7.68421650e-01 -3.68824959e-01 2.54455477e-01 -5.66795230e-01 -8.44474435e-01 -4.50995147e-01 -6.02973700e-01 4.83295321e-01 5.33156276e-01 -2.97499180e-01 -1.24234664e+00 3.00740182e-01 2.63248116e-01 6.67137951e-02 1.73240364e-01 7.35432267e-01 -1.03963184e+00 -9.16729793e-02 -3.22902560e-01 -1.97551936e-01 2.14719906e-01 1.63954452e-01 6.27270341e-02 -4.40753996e-01 -1.81887280e-02 -3.22981060e-01 -2.18013190e-02 3.30211669e-01 -1.45380408e-01 -2.65367627e-01 -2.25629687e-01 -4.14784998e-01 -3.49953696e-02 1.97800195e+00 3.35055262e-01 6.29892528e-01 9.70003903e-01 3.72690558e-01 5.31955898e-01 1.07764268e+00 3.06908816e-01 6.24464571e-01 4.35054958e-01 -3.20754856e-01 1.12455830e-01 3.26009765e-02 -1.99468955e-02 2.64126748e-01 1.03100491e+00 -3.20206434e-01 1.51780784e-01 -1.40661752e+00 1.07700288e+00 -1.47260773e+00 -9.58789945e-01 -2.93897510e-01 2.47215414e+00 1.05146742e+00 9.60021168e-02 1.95604101e-01 -1.75047830e-01 9.21586454e-01 -3.87441188e-01 2.71405905e-01 -4.20868367e-01 -1.27713516e-01 4.86505896e-01 1.00399590e+00 5.34777164e-01 -1.33816648e+00 1.12413895e+00 5.45229292e+00 6.49532139e-01 -9.22808111e-01 1.45998523e-01 -2.16898501e-01 4.47827935e-01 2.35261209e-02 4.66537662e-02 -1.26360846e+00 3.83903474e-01 1.14633143e+00 -7.31899559e-01 -4.37997952e-02 8.66823018e-01 1.87627494e-01 -3.87250334e-01 -3.68733048e-01 7.42677748e-01 1.78869471e-01 -1.26866388e+00 -7.70957544e-02 -5.21547869e-02 8.50883901e-01 2.00877666e-01 -6.88608766e-01 4.96692598e-01 3.47252846e-01 -5.60650766e-01 9.82122421e-01 6.96428299e-01 9.25408721e-01 -9.51080620e-01 1.16634607e+00 2.07901508e-01 -1.31951964e+00 2.67642647e-01 -4.91193265e-01 2.05382958e-01 6.04212172e-02 3.45755219e-01 -1.18789840e+00 1.04989457e+00 5.18583119e-01 3.82065386e-01 -6.18029296e-01 1.25101507e+00 -9.64237303e-02 2.34908491e-01 -2.81591326e-01 -3.15911591e-01 2.85556763e-01 -4.83539045e-01 4.05539840e-01 1.63320124e+00 3.71254295e-01 -3.05543393e-01 -6.58438578e-02 2.36917943e-01 -3.56106795e-02 1.17737544e+00 -7.94884503e-01 -7.60712847e-02 6.86763525e-01 1.20506406e+00 -1.05375493e+00 -5.31709790e-01 -5.01366138e-01 5.55892169e-01 1.79350656e-02 5.65451682e-02 -6.03247106e-01 -1.43876398e+00 1.04431681e-01 2.53409296e-01 2.70909041e-01 -6.04375042e-02 -4.71618176e-02 -1.20189834e+00 -1.00958653e-01 -1.00250661e+00 6.07886314e-01 -4.10971582e-01 -8.80219579e-01 6.62838817e-01 8.55838209e-02 -1.11637664e+00 -3.58791173e-01 -8.15421641e-01 1.36064038e-01 1.00347257e+00 -9.66594994e-01 -1.13542557e+00 7.85841942e-02 2.60257810e-01 1.49772510e-01 -6.50817811e-01 9.65431392e-01 9.48733032e-01 -2.36430556e-01 2.79093504e-01 3.85196507e-01 5.89000762e-01 8.84518504e-01 -1.34891200e+00 9.31034461e-02 6.98992372e-01 -5.64193055e-02 1.03287649e+00 6.81166351e-01 -9.68692482e-01 -6.31774306e-01 -9.73593116e-01 1.99594939e+00 -3.26611072e-01 1.00050271e+00 -2.07376495e-01 -4.47420716e-01 6.83801591e-01 5.53973019e-01 -6.34560347e-01 8.43219340e-01 1.07962199e-01 -1.00069664e-01 7.78656080e-02 -1.33595669e+00 3.41165990e-01 3.35858941e-01 -4.87783492e-01 -7.96108603e-01 5.12328744e-01 6.29197583e-02 -3.70665640e-01 -1.46900070e+00 -1.53457090e-01 6.35798395e-01 -7.77469039e-01 6.87488258e-01 -5.52576423e-01 1.75205991e-02 -5.97567737e-01 7.71702453e-02 -1.00231004e+00 2.98635632e-01 -2.79483885e-01 7.64934838e-01 1.76540661e+00 9.90402877e-01 -9.31294084e-01 4.28879172e-01 4.85534668e-01 1.66155428e-01 4.24000062e-02 -8.44274819e-01 -6.33998752e-01 1.57974541e-01 -1.51386276e-01 3.78028601e-01 1.25702035e+00 4.09042537e-01 3.74873608e-01 -1.19971089e-01 1.02989219e-01 2.45467007e-01 -2.36140698e-01 8.09266329e-01 -1.43373013e+00 3.68819445e-01 -1.31344110e-01 -7.52631426e-01 2.57246830e-02 6.39656410e-02 -8.79118681e-01 -1.82591438e-01 -1.89241648e+00 7.61111453e-02 -7.07075119e-01 6.28162995e-02 3.07413906e-01 2.74149060e-01 2.00197384e-01 1.32862315e-01 2.31506273e-01 -4.29397255e-01 -1.71992451e-01 5.08268714e-01 2.26571143e-01 -2.21119493e-01 -3.58762965e-02 -4.50391710e-01 6.91878974e-01 6.28203571e-01 -6.77419603e-01 1.71908081e-01 1.56040657e-02 6.86746359e-01 -2.46652484e-01 -1.93343133e-01 -1.15603411e+00 2.89425075e-01 -7.38932416e-02 8.72341022e-02 -6.57296419e-01 -1.44098938e-01 -9.05462205e-01 5.22274315e-01 4.81147796e-01 2.45413184e-02 1.08897053e-01 1.56428263e-01 6.07546084e-02 -5.04057467e-01 -8.73316288e-01 7.69315004e-01 -4.07070190e-01 -1.06621742e+00 -3.08743030e-01 -5.40077090e-01 1.41034618e-01 1.36233568e+00 -8.59285891e-02 5.07036224e-02 -3.63660902e-02 -9.36264873e-01 1.46174178e-01 6.93646967e-01 1.85483813e-01 -1.34321183e-01 -1.21291673e+00 -6.10711813e-01 -1.58908293e-01 4.72134978e-01 -7.72363424e-01 7.54277334e-02 5.47489941e-01 -1.35965824e+00 9.85495269e-01 -7.08500445e-01 2.59038452e-02 -1.21931136e+00 4.05767947e-01 1.12492926e-01 -4.46404606e-01 -6.92171007e-02 2.27421910e-01 -6.30274057e-01 -1.19543386e+00 -8.85968357e-02 8.82419348e-02 -9.66825008e-01 6.19486094e-01 1.83138460e-01 5.75258076e-01 2.95492679e-01 -1.29543495e+00 -5.42639434e-01 5.98728716e-01 3.76474112e-02 -3.68373424e-01 1.27715087e+00 -3.60083073e-01 -3.99470389e-01 6.48834467e-01 8.93247545e-01 7.65899897e-01 6.49837255e-02 6.49058372e-02 4.78385359e-01 -6.16362542e-02 -4.66301501e-01 -7.87148178e-01 -4.73990381e-01 2.52603501e-01 4.76217151e-01 3.02699953e-01 7.89697230e-01 -2.76723742e-01 3.53473932e-01 3.13460857e-01 8.44561398e-01 -1.60214460e+00 -1.22014439e+00 7.65539765e-01 4.35071081e-01 -9.52243149e-01 1.42386869e-01 -4.78852004e-01 -8.01042974e-01 1.36459196e+00 1.84048176e-01 7.05827028e-02 9.14222658e-01 -2.30421498e-02 5.03641367e-01 -1.95813879e-01 -2.66413867e-01 -6.52799487e-01 8.28203410e-02 5.48025846e-01 9.92833793e-01 1.72632739e-01 -1.57793534e+00 4.99201834e-01 -4.28052932e-01 2.88824439e-01 9.20928478e-01 1.10642612e+00 -3.73105437e-01 -1.18608761e+00 -5.14497936e-01 4.16730523e-01 -1.00179923e+00 -1.71160623e-01 -3.03679705e-01 1.56541133e+00 4.25611526e-01 8.02555382e-01 -1.20356254e-01 4.62456457e-02 4.41249877e-01 1.94498003e-01 7.00186715e-02 -8.95792902e-01 -1.01929009e+00 -8.62358361e-02 6.40977979e-01 8.05425569e-02 -7.61751235e-01 -7.04297721e-01 -1.35791707e+00 -2.91022390e-01 -3.61110419e-01 8.61630738e-01 1.04693663e+00 8.00819933e-01 6.92035034e-02 -1.53567836e-01 1.55699790e-01 -2.84761518e-01 5.14256693e-02 -1.05730116e+00 -6.92718685e-01 3.89455378e-01 -3.98517549e-01 -7.08270490e-01 -8.28399584e-02 3.33430320e-01]
[9.690922737121582, 9.623844146728516]
a85560ad-52b3-4882-ad01-0113da9fa92b
progressively-normalized-self-attention
2105.08468
null
https://arxiv.org/abs/2105.08468v2
https://arxiv.org/pdf/2105.08468v2.pdf
Progressively Normalized Self-Attention Network for Video Polyp Segmentation
Existing video polyp segmentation (VPS) models typically employ convolutional neural networks (CNNs) to extract features. However, due to their limited receptive fields, CNNs can not fully exploit the global temporal and spatial information in successive video frames, resulting in false-positive segmentation results. In this paper, we propose the novel PNS-Net (Progressively Normalized Self-attention Network), which can efficiently learn representations from polyp videos with real-time speed (~140fps) on a single RTX 2080 GPU and no post-processing. Our PNS-Net is based solely on a basic normalized self-attention block, equipping with recurrence and CNNs entirely. Experiments on challenging VPS datasets demonstrate that the proposed PNS-Net achieves state-of-the-art performance. We also conduct extensive experiments to study the effectiveness of the channel split, soft-attention, and progressive learning strategy. We find that our PNS-Net works well under different settings, making it a promising solution to the VPS task.
['Ling Shao', 'Debesh Jha', 'Huazhu Fu', 'Geng Chen', 'Deng-Ping Fan', 'Yu-Cheng Chou', 'Ge-Peng Ji']
2021-05-18
null
null
null
null
['video-polyp-segmentation']
['computer-vision']
[ 2.90291607e-01 -1.31540522e-01 -4.42875504e-01 -1.78765103e-01 -4.14934129e-01 -3.07826966e-01 1.69730872e-01 -1.05927229e-01 -4.58511621e-01 3.14844728e-01 4.22488153e-02 -3.35009545e-01 5.55530250e-01 -7.44491875e-01 -1.03849137e+00 -4.25028622e-01 -9.33975056e-02 -3.71802568e-01 9.17973399e-01 3.98475565e-02 1.75935030e-01 1.65237278e-01 -1.26966310e+00 5.28861880e-01 8.65508735e-01 1.16246700e+00 3.70517194e-01 8.38034511e-01 1.83194295e-01 1.00744808e+00 -3.82378995e-01 -1.80834338e-01 4.66471404e-01 -1.37070939e-01 -6.89099133e-01 3.34594697e-02 1.98813379e-01 -6.79757714e-01 -8.08631599e-01 8.39302599e-01 4.33239579e-01 -6.11497350e-02 2.34392032e-01 -9.63071704e-01 -5.00133097e-01 4.78028506e-01 -8.46132994e-01 6.67369127e-01 1.89556524e-01 3.84147048e-01 7.64750600e-01 -6.41061544e-01 4.47772503e-01 9.60465074e-01 8.16610992e-01 3.79634053e-01 -8.92968237e-01 -5.92997193e-01 3.72142315e-01 1.96552351e-01 -1.14766800e+00 -1.07642889e-01 4.05925602e-01 -1.09492831e-01 1.08449447e+00 5.32717295e-02 1.14411747e+00 1.06123543e+00 2.62260735e-01 1.29428971e+00 6.62756920e-01 -9.66181457e-02 8.11737776e-02 -3.40111941e-01 1.98846489e-01 8.87986839e-01 1.56741932e-01 6.32392690e-02 -5.52162051e-01 1.29446447e-01 1.29528987e+00 1.00705951e-01 -5.39767981e-01 -4.31854613e-02 -1.24112117e+00 6.06815159e-01 8.07492197e-01 2.88946837e-01 -4.05136257e-01 5.38592219e-01 6.92067027e-01 1.27398089e-01 3.68587315e-01 1.23442888e-01 -6.41079485e-01 -3.27615261e-01 -1.10629058e+00 9.72480979e-03 5.74618161e-01 1.20931029e+00 5.14138699e-01 3.89451534e-02 -3.43973011e-01 4.64354515e-01 5.60706249e-03 8.67047012e-02 7.12509394e-01 -1.11191773e+00 3.74551088e-01 4.40424174e-01 -1.01897724e-01 -9.83255148e-01 -3.88581425e-01 -5.93737662e-01 -8.71879637e-01 -2.16584638e-01 1.33010641e-01 -3.16389352e-01 -1.18968153e+00 1.49225569e+00 1.34581253e-01 6.64093316e-01 -2.15368569e-02 1.04349637e+00 9.96660054e-01 7.99030602e-01 1.10353634e-01 -3.72567773e-02 1.14470983e+00 -1.73897970e+00 -3.81238818e-01 -3.79907221e-01 5.73544264e-01 -4.98361140e-01 9.69847739e-01 2.09417537e-01 -1.17275274e+00 -8.38366091e-01 -1.16026449e+00 -2.43942514e-01 3.35235670e-02 1.32768244e-01 8.71280611e-01 4.44602281e-01 -1.17945433e+00 8.41019690e-01 -1.30338323e+00 -1.64498180e-01 8.09476435e-01 6.54987156e-01 4.54939008e-02 -1.12909824e-01 -8.78074884e-01 -6.43105619e-03 4.34992135e-01 2.58767992e-01 -9.70583141e-01 -7.60403872e-01 -8.27845991e-01 3.17419350e-01 4.78980958e-01 -8.04463863e-01 1.24682593e+00 -1.25802791e+00 -1.56288254e+00 4.70288664e-01 -2.38725662e-01 -7.57508934e-01 4.66652721e-01 -2.81498164e-01 -4.06614318e-02 5.05466580e-01 1.11468233e-01 1.22228205e+00 1.06917417e+00 -7.69326568e-01 -7.65822351e-01 4.96077314e-02 2.19239637e-01 2.03647181e-01 -4.09697831e-01 -5.79191037e-02 -9.82619941e-01 -8.68772089e-01 2.31176034e-01 -9.95302737e-01 -5.52236736e-01 8.68459791e-02 -3.35559070e-01 -3.02087963e-02 8.92833292e-01 -5.33521950e-01 1.14989531e+00 -2.34946656e+00 -2.80871354e-02 -1.02175541e-01 3.80574316e-01 6.94171011e-01 -1.67243779e-01 -1.85619488e-01 5.39482646e-02 6.11510016e-02 -3.26002717e-01 -1.98388234e-01 -4.88698840e-01 2.60142744e-01 -4.26779799e-02 3.59196812e-01 4.59201783e-01 1.14558017e+00 -1.08185053e+00 -6.30557954e-01 2.64879763e-01 5.42527437e-01 -8.49514604e-01 1.39953762e-01 -2.26135105e-01 5.11282563e-01 -4.61261153e-01 8.59392583e-01 6.23775423e-01 -5.85577726e-01 8.19396600e-02 -2.83436686e-01 -1.87114209e-01 1.75741576e-02 -5.01964092e-01 1.94830930e+00 -3.11010897e-01 1.00423551e+00 -1.46595299e-01 -9.53763127e-01 4.19988871e-01 4.52146530e-02 4.28163797e-01 -7.78001547e-01 3.98441851e-01 3.45179558e-01 2.62474641e-02 -4.63586509e-01 6.09719038e-01 4.36018735e-01 2.70760059e-01 7.10420534e-02 1.42545432e-01 4.42927331e-01 3.41055959e-01 1.09754056e-01 1.32021487e+00 2.56544322e-01 1.07531697e-01 -2.59267032e-01 4.42170024e-01 -9.94397402e-02 7.65377223e-01 7.53342986e-01 -5.85337698e-01 8.10780406e-01 6.55578911e-01 -5.63529730e-01 -8.69940758e-01 -8.15328240e-01 1.99524965e-02 9.68792558e-01 5.36883891e-01 -5.59365511e-01 -7.37313271e-01 -7.14590251e-01 -4.38846290e-01 -4.80495766e-03 -5.46224058e-01 -5.08524999e-02 -8.49885762e-01 -7.42996812e-01 3.95887613e-01 1.05960798e+00 9.33617651e-01 -1.16574514e+00 -1.16614640e+00 5.47600746e-01 -2.45215103e-01 -1.45077276e+00 -5.57618797e-01 -5.96225597e-02 -1.10697091e+00 -1.16761363e+00 -1.20835876e+00 -1.09525979e+00 6.50022686e-01 6.09605193e-01 9.01218235e-01 1.38116181e-01 -2.68224180e-01 1.50233731e-01 -5.30588627e-01 1.64282277e-01 7.26532266e-02 3.71871978e-01 -4.26705569e-01 -8.50092918e-02 8.70265439e-02 -5.69849789e-01 -1.17267621e+00 2.79137880e-01 -8.54084253e-01 3.35745484e-01 7.31301665e-01 8.56012702e-01 7.45593429e-01 -8.24048743e-02 2.60939926e-01 -6.83383048e-01 2.10670769e-01 -4.00107324e-01 -4.81005639e-01 3.20980810e-02 -9.06614885e-02 -2.87995070e-01 7.65788198e-01 -5.55346072e-01 -7.39859879e-01 1.94421262e-01 -2.84023464e-01 -7.56143212e-01 2.07711905e-01 4.08438534e-01 3.13916057e-01 -4.84890193e-01 2.57755727e-01 3.38560909e-01 -2.31399536e-01 -1.75824508e-01 -3.40530612e-02 5.28384387e-01 7.27992475e-01 -2.56915092e-01 2.41871566e-01 6.74044967e-01 -2.74732888e-01 -9.25028026e-01 -7.23899603e-01 -4.92051035e-01 -3.60191286e-01 -1.58821404e-01 9.43825305e-01 -1.25401819e+00 -7.53351092e-01 6.71213865e-01 -9.68386531e-01 -6.28671288e-01 -1.31254911e-01 4.74808902e-01 -5.40111303e-01 6.92901492e-01 -1.06821656e+00 -3.82741779e-01 -5.58271587e-01 -1.40615988e+00 1.02700818e+00 5.71413100e-01 7.42349923e-02 -7.15984285e-01 -3.75093877e-01 2.14782059e-01 3.34137470e-01 2.82749653e-01 4.87906903e-01 -1.54109403e-01 -9.71695065e-01 4.20870706e-02 -7.98046410e-01 3.61789167e-01 -8.31990615e-02 2.16351561e-02 -8.13967466e-01 -4.32285339e-01 7.41348509e-03 -3.28724355e-01 1.35513628e+00 7.89475203e-01 1.70134747e+00 -1.66071460e-01 -2.99236387e-01 1.10220647e+00 1.46859944e+00 2.36167416e-01 8.21219027e-01 2.32687965e-01 8.01826239e-01 -2.12948807e-02 6.28586829e-01 4.00921702e-01 3.27142835e-01 3.43700945e-01 6.27068818e-01 -4.35640961e-01 -1.88499182e-01 -4.04185019e-02 3.32340777e-01 8.53977740e-01 -1.96316957e-01 -2.28010178e-01 -6.22802854e-01 7.04107463e-01 -1.96139288e+00 -4.77907777e-01 -7.89065063e-02 1.88429606e+00 4.11857247e-01 2.91648269e-01 1.09428518e-01 2.56126355e-02 7.51980007e-01 4.33176637e-01 -6.68460429e-01 -2.51925647e-01 7.18876300e-03 3.53437781e-01 8.28377843e-01 1.11447722e-01 -1.41568995e+00 1.09937954e+00 6.12068748e+00 8.80479217e-01 -1.32165837e+00 1.36324927e-01 8.66002738e-01 -1.99012421e-02 1.97938353e-01 -1.33205920e-01 -5.78597009e-01 6.03989422e-01 6.75328016e-01 1.07535578e-01 3.38055342e-01 8.68302405e-01 -3.34754661e-02 -1.15768395e-01 -7.82511115e-01 1.00000548e+00 -4.10234891e-02 -1.44541216e+00 -1.06863298e-01 -3.62948142e-02 9.81601119e-01 3.95183176e-01 7.12563619e-02 2.50610858e-01 -4.94056791e-02 -7.00468957e-01 6.85131609e-01 -9.58397239e-03 8.30264270e-01 -7.32012451e-01 8.11367810e-01 -4.67658713e-02 -1.44187391e+00 -2.55574495e-01 -3.89186531e-01 -2.48442981e-02 2.04762429e-01 3.10810566e-01 -4.38023597e-01 4.46708560e-01 9.08213198e-01 1.24059129e+00 -5.18005669e-01 1.23229206e+00 -2.97101699e-02 6.00543499e-01 -4.43185210e-01 -1.38558792e-02 7.74195671e-01 2.23830506e-01 3.64230216e-01 1.37443411e+00 2.90256679e-01 2.41297871e-01 1.97972402e-01 6.70314014e-01 -2.55435526e-01 -1.16367325e-01 -3.69895622e-02 -3.09286565e-01 -6.42300621e-02 1.01511836e+00 -1.23959279e+00 -5.86305678e-01 -6.39842451e-01 1.15744436e+00 2.20171466e-01 2.95907587e-01 -1.13273740e+00 -3.58738273e-01 6.11923516e-01 1.10925891e-01 8.59478056e-01 -2.07186982e-01 -2.07845703e-01 -1.34573364e+00 5.46930134e-02 -6.72570527e-01 1.27197072e-01 -5.99545062e-01 -8.94198596e-01 7.12017655e-01 -4.57906991e-01 -1.32791722e+00 1.47629738e-01 -6.70640290e-01 -5.13705671e-01 2.02701464e-01 -1.85326135e+00 -9.55649912e-01 -5.20050585e-01 6.63546026e-01 9.22632694e-01 1.59416378e-01 3.66584957e-01 3.94799292e-01 -6.51904583e-01 6.10951602e-01 -1.05574660e-01 3.33177984e-01 3.51774454e-01 -8.42230678e-01 7.53525078e-01 9.66138422e-01 -1.45689622e-01 2.76897639e-01 2.77919054e-01 -6.39461219e-01 -1.18784809e+00 -1.44617224e+00 2.82063663e-01 3.98388684e-01 5.28560638e-01 -8.34523812e-02 -8.18297625e-01 6.76811278e-01 1.92375869e-01 5.49396694e-01 3.28824639e-01 -3.60619813e-01 -1.80448398e-01 3.70431133e-02 -8.47787499e-01 6.61836743e-01 1.27026308e+00 -1.56689048e-01 -1.64684728e-01 4.46497977e-01 1.20902300e+00 -9.49836433e-01 -5.89590192e-01 6.27028406e-01 5.53819537e-01 -1.25170457e+00 1.07034564e+00 -2.81571776e-01 8.97620916e-01 -1.35640845e-01 8.77924263e-02 -8.40281487e-01 -3.42569202e-01 -6.91855192e-01 -3.35184723e-01 5.83025634e-01 2.15213448e-01 -5.47450483e-01 9.50293124e-01 1.47770286e-01 -4.58957672e-01 -1.17647731e+00 -8.34408700e-01 -6.56763077e-01 -3.82850170e-01 -5.77359498e-01 3.51822615e-01 6.15082145e-01 -1.96903124e-01 3.28039303e-02 -3.15759689e-01 2.44465604e-01 4.00544316e-01 9.81381908e-02 3.92525077e-01 -5.65503240e-01 -5.73183775e-01 -3.27652186e-01 -6.74050093e-01 -1.71787250e+00 -6.70999065e-02 -3.31883818e-01 4.09127995e-02 -1.35306072e+00 2.79344618e-01 -2.94418514e-01 -3.71176809e-01 5.18549025e-01 -2.49458387e-01 6.58915043e-01 3.35289806e-01 1.81374028e-01 -9.35552955e-01 4.82492924e-01 1.45864809e+00 -1.18361518e-01 -4.92420644e-01 -9.29311663e-02 -5.16545594e-01 9.71161246e-01 6.60146058e-01 -3.04442048e-01 -3.31867099e-01 -7.35721290e-01 -1.03881493e-01 1.22238919e-01 4.28162247e-01 -1.38664556e+00 3.53454530e-01 2.22509518e-01 4.63440746e-01 -7.26143658e-01 3.48769516e-01 -5.80924511e-01 -3.63915831e-01 8.60393584e-01 -3.22388373e-02 -9.39998552e-02 3.67194682e-01 7.60561526e-01 -3.49183947e-01 -6.10301234e-02 8.10830414e-01 -3.47142041e-01 -9.50268328e-01 5.98633766e-01 -3.93177927e-01 1.20649502e-01 1.03939772e+00 -2.62472123e-01 -1.47113770e-01 -1.36147484e-01 -6.62558138e-01 3.50567818e-01 4.46474880e-01 2.60057569e-01 7.33312488e-01 -1.13599277e+00 -3.60544652e-01 3.85687917e-01 -2.19312519e-01 3.39322984e-01 5.54678261e-01 8.43847096e-01 -1.09531224e+00 6.03344381e-01 -2.07427964e-01 -9.78113651e-01 -1.13820338e+00 5.97281754e-01 1.46596134e-01 -2.98324645e-01 -1.02527952e+00 1.08895385e+00 4.96869683e-01 1.40078068e-01 2.79304713e-01 -7.09820390e-01 -5.81499003e-02 -3.38164121e-01 5.89762568e-01 2.54011512e-01 -1.28137201e-01 -3.67258340e-01 -2.25627735e-01 5.51302314e-01 -3.23465556e-01 2.16526464e-01 1.21137846e+00 1.58528592e-02 1.97052747e-01 -1.12654649e-01 1.31507087e+00 -5.16390145e-01 -1.77943432e+00 -4.76603471e-02 -5.90498328e-01 -6.94769740e-01 1.81050852e-01 -2.76387423e-01 -1.65621483e+00 8.53148699e-01 4.95281994e-01 -1.57792791e-04 1.44257832e+00 -2.57249355e-01 1.41232014e+00 1.07925341e-01 2.75633991e-01 -8.88038576e-01 2.54959106e-01 4.62557107e-01 4.59714115e-01 -1.13977075e+00 -1.70873344e-01 -8.34697008e-01 -4.73098755e-01 1.14693737e+00 8.21529567e-01 -5.73645830e-01 5.60945928e-01 2.56307751e-01 -2.58958995e-01 7.67827183e-02 -6.16067231e-01 -1.50721103e-01 -1.48115098e-01 4.39857781e-01 2.47168764e-01 -1.95413649e-01 -2.78816313e-01 6.05871141e-01 7.36863986e-02 3.73312354e-01 6.21175051e-01 1.07134402e+00 -2.10441455e-01 -7.13250458e-01 1.13702804e-01 5.28361022e-01 -5.36261559e-01 -2.20171556e-01 7.53354877e-02 6.96848273e-01 2.47084409e-01 6.93499446e-01 3.41828912e-01 -4.66177732e-01 1.25096729e-02 -6.87637269e-01 5.41959465e-01 -4.03349727e-01 -7.00435519e-01 3.57330501e-01 -2.91773647e-01 -9.66544032e-01 -6.22092247e-01 -5.41220665e-01 -1.28227234e+00 -2.40807399e-01 -2.23631784e-01 -3.48142207e-01 4.00235355e-01 8.04272771e-01 5.04010797e-01 1.03857923e+00 5.93503952e-01 -1.12428463e+00 -1.07679911e-01 -6.33667111e-01 -2.29259521e-01 1.41020948e-02 3.52530986e-01 -3.09803337e-01 -2.29167581e-01 -4.40398464e-03]
[9.299015045166016, -0.04466262087225914]
35a5d822-80dc-40f7-8cc9-2876d156da13
fashion-cut-unsupervised-domain-adaptation
2305.0558
null
https://arxiv.org/abs/2305.05580v1
https://arxiv.org/pdf/2305.05580v1.pdf
Fashion CUT: Unsupervised domain adaptation for visual pattern classification in clothes using synthetic data and pseudo-labels
Accurate product information is critical for e-commerce stores to allow customers to browse, filter, and search for products. Product data quality is affected by missing or incorrect information resulting in poor customer experience. While machine learning can be used to correct inaccurate or missing information, achieving high performance on fashion image classification tasks requires large amounts of annotated data, but it is expensive to generate due to labeling costs. One solution can be to generate synthetic data which requires no manual labeling. However, training a model with a dataset of solely synthetic images can lead to poor generalization when performing inference on real-world data because of the domain shift. We introduce a new unsupervised domain adaptation technique that converts images from the synthetic domain into the real-world domain. Our approach combines a generative neural network and a classifier that are jointly trained to produce realistic images while preserving the synthetic label information. We found that using real-world pseudo-labels during training helps the classifier to generalize in the real-world domain, reducing the synthetic bias. We successfully train a visual pattern classification model in the fashion domain without real-world annotations. Experiments show that our method outperforms other unsupervised domain adaptation algorithms.
["Noel E. O'Connor", 'Philip Kelly', 'Martina Naughton', 'Alex Martinelli', 'Enric Moreu']
2023-05-09
null
null
null
null
['unsupervised-domain-adaptation']
['methodology']
[ 6.42674387e-01 4.32455540e-03 -2.81672716e-01 -1.00084245e+00 -8.13538790e-01 -8.39486003e-01 2.04564691e-01 1.11664392e-01 -2.83331871e-01 6.79349363e-01 -1.41632065e-01 -9.46434513e-02 3.77846360e-01 -8.85755897e-01 -1.01697135e+00 -3.80644858e-01 5.96630156e-01 8.36809993e-01 8.79229158e-02 -1.00875005e-01 1.27271581e-02 1.99718997e-01 -1.64162612e+00 6.64452910e-01 8.88528109e-01 9.99885380e-01 2.67807722e-01 4.74198163e-01 -1.77108437e-01 5.78881621e-01 -7.95539975e-01 -5.47685444e-01 6.08981967e-01 -5.37186384e-01 -5.55764318e-01 8.77987027e-01 5.40934384e-01 -3.08219790e-01 2.66065210e-01 1.28760374e+00 7.07748532e-02 1.30930513e-01 6.84427738e-01 -1.49231458e+00 -1.29642594e+00 2.02373043e-01 -3.77259225e-01 -3.30246985e-01 3.50160897e-01 3.44916850e-01 5.93576014e-01 -7.20376909e-01 1.08144212e+00 1.08785796e+00 6.96489513e-01 4.65103686e-01 -1.94151652e+00 -7.35061586e-01 7.79979974e-02 -1.21996505e-02 -1.24616873e+00 -3.60087961e-01 1.08150518e+00 -5.08574009e-01 5.51541269e-01 -3.62981223e-02 7.10752308e-01 1.40038574e+00 -9.45814420e-03 6.86588705e-01 1.39625597e+00 -6.49191439e-01 5.22712469e-01 8.73505294e-01 -1.66959152e-01 3.76332313e-01 2.82553911e-01 8.90662223e-02 -3.27956766e-01 2.39442527e-01 8.03355098e-01 -2.27119159e-02 1.95624419e-02 -8.55220795e-01 -8.82608354e-01 9.67904985e-01 5.03438771e-01 9.40616766e-04 -3.49202454e-01 -4.03105825e-01 1.74426854e-01 7.88557410e-01 6.04493976e-01 9.20453787e-01 -4.89708275e-01 1.78873524e-01 -9.17362034e-01 2.42101341e-01 7.32152879e-01 1.11113346e+00 7.46806681e-01 5.10855615e-02 2.21180350e-01 9.88863468e-01 2.28444442e-01 4.71198767e-01 4.87592757e-01 -1.06729496e+00 1.65332928e-01 6.49890184e-01 4.44656491e-01 -1.00885487e+00 -4.44479845e-02 -4.99750048e-01 -5.37785530e-01 4.44558382e-01 9.03373659e-01 2.32047737e-01 -1.28098154e+00 1.59538746e+00 1.56155676e-01 -3.95077318e-01 7.44010359e-02 1.00803995e+00 4.58019763e-01 5.41232586e-01 2.99403220e-01 -1.79476887e-02 1.00729632e+00 -8.89295518e-01 -5.73548436e-01 -5.28756082e-01 5.40963352e-01 -8.95941377e-01 1.54606080e+00 5.17411113e-01 -8.43191862e-01 -9.21103656e-01 -1.27025938e+00 -1.87995099e-02 -6.61644340e-01 7.49624372e-02 4.51002985e-01 8.19276035e-01 -6.58267558e-01 4.07520503e-01 -4.24886435e-01 -5.90838313e-01 4.90731835e-01 1.38132319e-01 -4.92709726e-01 -4.21170741e-01 -9.57642674e-01 8.00956368e-01 5.89740336e-01 -2.10277706e-01 -5.56734622e-01 -6.93391263e-01 -1.12389588e+00 -2.43157953e-01 2.79814690e-01 -4.36928809e-01 1.45227790e+00 -1.86379397e+00 -1.23270786e+00 9.96624529e-01 9.54963118e-02 -3.42174858e-01 4.03312474e-01 2.16401458e-01 -6.77884459e-01 -3.58884074e-02 2.87233353e-01 1.04791725e+00 1.10152960e+00 -1.75007164e+00 -5.48767030e-01 -3.56683016e-01 -2.36239076e-01 2.94338702e-03 -7.78231472e-02 -5.12706876e-01 -1.82563111e-01 -7.16732681e-01 2.60677725e-01 -1.10400152e+00 -1.05300501e-01 1.19978480e-01 2.34234165e-02 2.09465697e-01 7.86589026e-01 -7.00387120e-01 5.58389843e-01 -2.20407295e+00 -4.34280157e-01 3.25036287e-01 -1.75669014e-01 1.37166128e-01 -2.62309581e-01 5.89195602e-02 -1.54558599e-01 -1.10449165e-01 -1.14527874e-01 -8.47960413e-02 5.02927229e-03 3.65122736e-01 -2.53474593e-01 5.74818440e-02 2.87025005e-01 9.48151886e-01 -1.06674469e+00 -3.96002173e-01 2.50555545e-01 2.82542199e-01 -6.29980206e-01 7.90611580e-02 -5.94223440e-01 6.18536174e-01 -2.56571546e-02 5.29001534e-01 7.79240966e-01 -4.32690263e-01 3.46488178e-01 -3.47072035e-01 5.12930810e-01 1.80651527e-02 -1.20481420e+00 1.74322033e+00 -6.02036715e-01 4.94654298e-01 -2.06916600e-01 -9.47294354e-01 1.12466121e+00 -1.23220354e-01 1.86305210e-01 -1.12186456e+00 6.34365603e-02 1.98334441e-01 -8.75874907e-02 -4.70809549e-01 5.11265993e-01 -4.57856625e-01 -2.95766413e-01 4.58025694e-01 8.78380239e-02 -3.93226355e-01 2.16245517e-01 -1.70022119e-02 6.22220933e-01 4.30967599e-01 -4.37550955e-02 1.26050368e-01 -5.11825755e-02 6.93468273e-01 6.20126665e-01 5.48909009e-01 -1.96181744e-01 5.88392079e-01 1.28672525e-01 -7.14825571e-01 -1.49650490e+00 -1.35248411e+00 6.46035711e-04 1.10503948e+00 1.88780293e-01 9.04491395e-02 -6.42516375e-01 -9.44760561e-01 8.03010911e-02 1.19589221e+00 -5.09280086e-01 -4.39471334e-01 -2.22485021e-01 -5.51562786e-01 2.27322727e-02 6.55237377e-01 6.78329706e-01 -1.03872871e+00 -4.64465231e-01 3.58130693e-01 -1.70907885e-01 -1.10198486e+00 -4.96092618e-01 1.52336255e-01 -7.14331985e-01 -8.68479252e-01 -4.32539165e-01 -1.16758060e+00 1.04802608e+00 1.50215790e-01 1.43071723e+00 -4.11153883e-01 -3.30903798e-01 2.69501895e-01 -3.34738344e-01 -4.28801030e-01 -9.75263536e-01 -1.82920426e-01 -1.88646585e-01 1.09141596e-01 9.09935117e-01 -2.67156094e-01 -5.31108439e-01 6.60727978e-01 -9.84797359e-01 2.88272202e-01 5.93950450e-01 1.16314840e+00 8.34914923e-01 3.23675036e-01 7.23927796e-01 -1.26797390e+00 6.25689387e-01 -1.79824829e-01 -4.44953054e-01 2.41652057e-01 -8.32714200e-01 2.58569658e-01 5.72733760e-01 -8.20377350e-01 -1.48742843e+00 5.81878901e-01 1.81839585e-01 -3.11869562e-01 -4.94971037e-01 2.19287574e-01 -2.87268847e-01 1.70325175e-01 1.06403232e+00 2.84214579e-02 1.20072380e-01 -4.91779983e-01 4.83757764e-01 7.60457933e-01 5.58471143e-01 -3.19559306e-01 5.94918847e-01 2.98394382e-01 -3.54799718e-01 -5.77360630e-01 -9.55468297e-01 -2.18650252e-01 -8.11636031e-01 -1.73142850e-01 6.36146486e-01 -8.23509276e-01 -7.24891648e-02 2.68118620e-01 -6.52029514e-01 -4.06360805e-01 -7.08414674e-01 4.08687204e-01 -5.32715619e-01 -1.09510645e-02 -3.87519956e-01 -4.24010247e-01 1.75460652e-01 -1.00572240e+00 9.86908197e-01 9.80523974e-02 -5.52672923e-01 -7.95870006e-01 -1.49420843e-01 4.94455546e-01 3.06386173e-01 2.88132638e-01 9.65477645e-01 -5.79604924e-01 -4.53359842e-01 -4.05434221e-01 -2.97252655e-01 6.07008219e-01 3.18414688e-01 -3.40989351e-01 -8.77888083e-01 -2.34802708e-01 -2.15506628e-02 -5.81388116e-01 5.56841254e-01 2.16370001e-01 1.12807763e+00 -2.81051427e-01 -2.29064971e-01 1.59517646e-01 1.39490366e+00 4.67375249e-01 5.40291011e-01 2.64778703e-01 4.50506330e-01 7.85127044e-01 9.79866147e-01 -4.61393036e-02 2.37326488e-01 7.01126099e-01 1.66364200e-02 -2.90775865e-01 -3.65108281e-01 -8.19613934e-01 -4.42164689e-02 2.41273269e-01 4.72882032e-01 -1.64586991e-01 -5.39860487e-01 6.83871329e-01 -1.64123499e+00 -8.87523890e-01 2.47719303e-01 2.33478498e+00 9.97485995e-01 4.42151994e-01 2.65501142e-01 1.24144204e-01 6.63620174e-01 -4.62154835e-01 -7.76760697e-01 -5.27142942e-01 1.39005892e-02 1.76685661e-01 6.37221396e-01 3.43827605e-01 -1.12762976e+00 8.50762963e-01 6.44316244e+00 4.48060453e-01 -1.01736009e+00 1.27607778e-01 8.39580297e-01 -2.94779222e-02 -3.47596765e-01 -2.09513202e-01 -3.93339038e-01 4.93998438e-01 8.58478487e-01 1.20907001e-01 3.78628045e-01 1.28548777e+00 2.76906248e-02 -1.15163654e-01 -1.27441013e+00 1.23953032e+00 3.16875905e-01 -1.01360607e+00 7.38753006e-02 -6.14262372e-02 9.49771762e-01 -6.20737135e-01 3.19812268e-01 2.92999268e-01 6.60929561e-01 -9.18264389e-01 6.97949171e-01 2.48146057e-01 8.94135535e-01 -7.37779200e-01 5.51232457e-01 2.70940185e-01 -7.79865503e-01 1.74397845e-02 -3.37288618e-01 -9.32601374e-03 3.49918082e-02 4.60717171e-01 -1.31685567e+00 -6.49412647e-02 6.06698394e-01 4.28463519e-01 -8.23723018e-01 7.56544292e-01 -7.98804760e-02 2.98236251e-01 -1.91581860e-01 9.09045264e-02 -1.38583750e-01 -1.69908360e-01 -3.19991484e-02 9.33393061e-01 2.79588073e-01 -3.43457758e-01 3.63293707e-01 9.21138287e-01 -8.79796594e-02 -9.61830765e-02 -8.03913236e-01 -1.14285648e-01 2.67120510e-01 9.79369044e-01 -9.14981604e-01 -3.37548643e-01 -3.29076052e-01 1.34092855e+00 4.04697806e-02 5.29714286e-01 -6.50537133e-01 -2.89638460e-01 3.30847293e-01 4.85592127e-01 2.87530869e-01 -9.88672525e-02 -6.00584090e-01 -1.02018118e+00 2.00440697e-02 -1.02172005e+00 2.93744594e-01 -1.00850344e+00 -1.55389297e+00 5.19112825e-01 -1.38539514e-02 -1.38893914e+00 -5.83411396e-01 -7.63952494e-01 1.71874478e-01 6.45597100e-01 -1.08608353e+00 -1.26155162e+00 -4.05541599e-01 3.40820551e-01 8.03903520e-01 -1.17312670e-01 8.20043504e-01 3.43818665e-01 1.53491618e-02 7.18283355e-01 -8.36821273e-03 1.54774532e-01 1.15560794e+00 -1.30197477e+00 4.03155595e-01 5.08825481e-01 3.45233500e-01 3.63393903e-01 8.94239247e-01 -7.15295315e-01 -9.97603953e-01 -1.25781512e+00 7.38506734e-01 -4.89517212e-01 2.80346304e-01 -4.75691974e-01 -8.75895381e-01 8.66503000e-01 4.38474938e-02 -1.06977761e-01 8.89490604e-01 -4.36883382e-02 -7.42743254e-01 -3.58081132e-01 -1.60459447e+00 5.52139401e-01 8.89593720e-01 -4.63419825e-01 -5.45304537e-01 3.53563815e-01 4.72676903e-01 -2.63644665e-01 -9.16948080e-01 1.59918487e-01 6.58765197e-01 -6.65310979e-01 8.48884463e-01 -6.21852994e-01 3.72054517e-01 -4.67262357e-01 -1.15718886e-01 -1.66320145e+00 -3.10907155e-01 -3.15606967e-02 3.63413781e-01 1.12643182e+00 7.77147710e-01 -4.37639534e-01 1.06697273e+00 1.05504215e+00 2.44044647e-01 -9.47006717e-02 -3.44933093e-01 -8.55385900e-01 -1.14610717e-01 -4.50976998e-01 7.76300907e-01 1.00368226e+00 -2.31072530e-01 5.20886064e-01 -2.80821621e-01 -1.36216834e-01 6.64847016e-01 3.83535117e-01 8.90656710e-01 -1.34397852e+00 -3.90269846e-01 5.37304208e-02 -4.43314523e-01 -9.13395941e-01 -2.77880970e-02 -8.68136346e-01 1.50895804e-01 -1.37563443e+00 4.42023240e-02 -6.11653745e-01 -1.88298784e-02 3.94338727e-01 2.94861764e-01 8.50571156e-01 9.21215042e-02 1.30334705e-01 -4.57865119e-01 9.39131901e-02 1.42867768e+00 -3.88747633e-01 -4.09455538e-01 -8.33430439e-02 -7.97750831e-01 7.36869693e-01 8.74638081e-01 -4.70400453e-01 -7.10055709e-01 -3.48265141e-01 2.38604099e-01 -4.40194905e-01 5.01221597e-01 -8.56419981e-01 -1.03483252e-01 -8.63884389e-02 1.17662823e+00 -3.90114695e-01 3.90710384e-01 -1.15200663e+00 4.41929728e-01 2.62059897e-01 -4.91721183e-01 -1.25084758e-01 1.24919914e-01 6.48276091e-01 -2.15631649e-01 -2.34935984e-01 9.50107396e-01 -3.95681351e-01 -1.05449450e+00 -2.78806001e-01 -1.03223369e-01 -1.10420391e-01 1.20778275e+00 -4.73705947e-01 -1.62974223e-01 -5.23070276e-01 -1.01314151e+00 -1.13016136e-01 1.01007962e+00 7.48287976e-01 4.33941573e-01 -1.53546250e+00 -3.98504972e-01 7.01724887e-01 4.71614003e-01 -2.46568158e-01 8.65702704e-02 3.88605334e-02 -6.00502431e-01 1.47820994e-01 -4.85269099e-01 -7.43520737e-01 -1.00753140e+00 1.08876002e+00 7.43211806e-02 -1.65981978e-01 -3.91543925e-01 5.95386267e-01 1.41117305e-01 -4.84206736e-01 7.71853700e-02 -3.11381161e-01 2.59517014e-01 -5.74513748e-02 3.70901495e-01 -9.32821631e-02 2.13787556e-01 -5.21895528e-01 9.42180976e-02 3.12407196e-01 -3.02213639e-01 -2.51823545e-01 1.02311540e+00 -2.24258870e-01 5.47899425e-01 4.30604070e-01 1.17522132e+00 -3.83047968e-01 -1.44823456e+00 -2.77416050e-01 -7.48680234e-02 -9.01748538e-01 -8.15467834e-02 -1.19170868e+00 -9.12481308e-01 6.35601580e-01 9.43075120e-01 2.74106443e-01 1.39277911e+00 -6.06524609e-02 7.05484986e-01 2.43022159e-01 6.73923254e-01 -1.45513260e+00 3.32990110e-01 -2.88881510e-01 7.10876882e-01 -1.68720412e+00 -1.93660855e-01 -6.24772668e-01 -1.22829223e+00 7.14141607e-01 8.18133056e-01 -3.92654017e-02 4.95257735e-01 8.54179561e-02 3.29684764e-01 4.83911783e-02 -3.34983438e-01 4.65477034e-02 2.08169714e-01 1.04300714e+00 1.83307856e-01 8.79056603e-02 9.98091027e-02 4.06774849e-01 -3.20239305e-01 3.49982142e-01 2.78433859e-01 1.00299728e+00 -1.63572088e-01 -1.41465712e+00 -4.74723876e-01 5.15393734e-01 -1.29990622e-01 2.57399142e-01 -4.15868372e-01 7.81139970e-01 4.22082782e-01 1.10034490e+00 4.20892507e-01 -2.52371490e-01 4.58057404e-01 4.19411212e-01 7.01270401e-01 -6.63258612e-01 -1.72866061e-01 4.12050746e-02 1.31599069e-01 -3.42593819e-01 -4.85759974e-01 -5.08858085e-01 -7.68947184e-01 -1.41644562e-02 -3.23143154e-02 3.57187651e-02 6.75847650e-01 5.85177958e-01 4.60782558e-01 3.68930876e-01 5.21632493e-01 -6.23826087e-01 -4.17610705e-01 -8.16830158e-01 -6.96984708e-01 1.12371373e+00 7.60824010e-02 -7.13706911e-01 -4.00044909e-03 7.83077002e-01]
[9.907181739807129, 1.595145583152771]
9830d336-1a53-4307-8364-1509277b2154
relationformer-a-unified-framework-for-image
2203.10202
null
https://arxiv.org/abs/2203.10202v1
https://arxiv.org/pdf/2203.10202v1.pdf
Relationformer: A Unified Framework for Image-to-Graph Generation
A comprehensive representation of an image requires understanding objects and their mutual relationship, especially in image-to-graph generation, e.g., road network extraction, blood-vessel network extraction, or scene graph generation. Traditionally, image-to-graph generation is addressed with a two-stage approach consisting of object detection followed by a separate relation prediction, which prevents simultaneous object-relation interaction. This work proposes a unified one-stage transformer-based framework, namely Relationformer, that jointly predicts objects and their relations. We leverage direct set-based object prediction and incorporate the interaction among the objects to learn an object-relation representation jointly. In addition to existing [obj]-tokens, we propose a novel learnable token, namely [rln]-token. Together with [obj]-tokens, [rln]-token exploits local and global semantic reasoning in an image through a series of mutual associations. In combination with the pair-wise [obj]-token, the [rln]-token contributes to a computationally efficient relation prediction. We achieve state-of-the-art performance on multiple, diverse and multi-domain datasets that demonstrate our approach's effectiveness and generalizability.
['Bjoern Menze', 'Volker Tresp', 'Georgios Kaissis', 'Sahand Sharifzadeh', 'Jiazhen Pan', 'Hongwei Li', 'Ivan Ezhov', 'Johannes Paetzold', 'Bastian Wittmann', 'Rajat Koner', 'Suprosanna Shit']
2022-03-19
null
null
null
null
['scene-graph-generation']
['computer-vision']
[ 3.12324643e-01 4.29128975e-01 -2.64210701e-01 -5.08445442e-01 -5.70232987e-01 -3.75341356e-01 8.34531307e-01 3.83497477e-01 1.27996162e-01 3.67710739e-01 1.21575579e-01 -3.15284692e-02 -3.17810863e-01 -1.16441572e+00 -7.63819098e-01 -3.30217391e-01 -5.36969863e-02 4.06014323e-01 6.39183640e-01 -1.55201983e-02 -3.63269658e-03 6.38720989e-01 -1.52949440e+00 3.86909962e-01 7.50297129e-01 1.38612223e+00 3.10626060e-01 4.86960560e-01 -2.64622062e-01 1.39979124e+00 -1.63404018e-01 -4.60372329e-01 1.57999367e-01 -1.17095821e-01 -1.03078496e+00 4.01640028e-01 3.38330239e-01 -3.01739722e-01 -4.96533483e-01 7.97385812e-01 8.10553282e-02 1.60350621e-01 8.48323584e-01 -1.34164917e+00 -4.10466164e-01 5.09000719e-01 -6.90601230e-01 1.23576885e-02 1.02923967e-01 2.44772509e-01 1.43856049e+00 -1.01089323e+00 8.16515744e-01 1.27377617e+00 2.37453923e-01 -8.27360898e-02 -1.22399104e+00 -5.65641820e-01 3.75565588e-01 4.76399243e-01 -1.56310749e+00 -2.63541788e-01 1.03641784e+00 -6.18284225e-01 9.67418015e-01 4.86158840e-02 7.70075560e-01 4.26340312e-01 -9.62040201e-02 9.63407815e-01 7.70949006e-01 -2.64488906e-01 -1.39935732e-01 7.45263174e-02 -1.00105003e-01 1.11312783e+00 1.16462320e-01 -1.55460060e-01 -5.31654894e-01 6.50802329e-02 1.11813748e+00 -8.96997005e-03 7.14926645e-02 -3.94401491e-01 -1.33656991e+00 5.77592492e-01 8.52470100e-01 -3.89612131e-02 -5.22335231e-01 1.69845924e-01 1.22505061e-01 -3.25230539e-01 4.00883794e-01 3.32877129e-01 -3.51675242e-01 4.55041021e-01 -5.91922283e-01 1.81739688e-01 6.25190377e-01 1.22289133e+00 1.12794983e+00 -2.59667963e-01 -5.87515771e-01 7.97870398e-01 4.01072204e-01 5.49313985e-02 -2.06202254e-01 -8.17957044e-01 5.46312034e-01 1.17258990e+00 -8.58975053e-02 -1.28895390e+00 -3.75140667e-01 -5.68881392e-01 -8.24199319e-01 -5.11717866e-04 1.87819511e-01 1.75624773e-01 -8.00130785e-01 1.49618292e+00 7.88681567e-01 7.41888106e-01 -1.03103191e-01 7.91323185e-01 1.21498370e+00 3.69727314e-01 2.71332860e-01 1.38177380e-01 1.62098515e+00 -1.43683457e+00 -3.10728550e-01 -1.51557907e-01 5.56020141e-01 -7.49868751e-01 6.33102238e-01 1.09619781e-01 -8.92994046e-01 -5.70808172e-01 -6.38630569e-01 -2.90646791e-01 -3.21916580e-01 1.86399281e-01 8.63532543e-01 -1.19777255e-01 -7.12014675e-01 4.07044381e-01 -4.43871677e-01 -2.28974596e-01 9.32288408e-01 2.53511250e-01 -5.45563638e-01 -1.94195956e-01 -8.68608415e-01 8.43053758e-01 4.08581167e-01 2.26307884e-01 -8.94353211e-01 -6.53642058e-01 -9.64719534e-01 1.30634010e-01 7.83042848e-01 -1.04608440e+00 9.22084451e-01 -5.74848056e-01 -1.11186504e+00 9.23407614e-01 -3.38217914e-01 -4.01536733e-01 3.13355565e-01 -9.61369053e-02 -1.66990444e-01 4.59773391e-01 1.89222932e-01 8.50272119e-01 6.51663780e-01 -1.45733809e+00 -8.03682446e-01 -2.45018587e-01 2.27713630e-01 3.82736534e-01 -2.25050747e-01 6.23850934e-02 -9.37463820e-01 -6.57607257e-01 2.15949506e-01 -5.07420957e-01 -4.33190942e-01 3.56039405e-01 -9.12136018e-01 -5.09389520e-01 7.80787826e-01 -4.95778680e-01 1.05928075e+00 -1.90365636e+00 3.71845774e-02 4.69180048e-01 8.63560259e-01 1.48479387e-01 -4.02828991e-01 3.30086142e-01 -2.02400256e-02 -9.36388448e-02 7.23615289e-02 -4.13977385e-01 -1.67575255e-01 2.53918052e-01 -5.14083616e-02 2.15196088e-01 7.84330666e-01 1.20673943e+00 -1.07967997e+00 -1.00628996e+00 5.46732843e-01 4.42197561e-01 -4.30317819e-01 2.86602855e-01 -5.21605968e-01 4.48460460e-01 -9.25308704e-01 8.75772238e-01 4.13937718e-01 -6.42608106e-01 4.43351194e-02 -7.52309620e-01 1.06446639e-01 3.39737684e-01 -1.01712298e+00 1.55820727e+00 -3.33987653e-01 4.26368743e-01 -2.01544896e-01 -1.03932667e+00 1.20744395e+00 1.21872857e-01 7.55917192e-01 -6.21872485e-01 8.01415071e-02 -3.00714672e-02 -4.06229608e-02 -4.40528393e-01 3.88040572e-01 1.37482479e-01 2.90401131e-01 1.84978530e-01 6.10992052e-02 -3.17177474e-02 2.41691872e-01 5.92814207e-01 1.23436534e+00 3.20759594e-01 5.35311997e-01 -1.35036603e-01 6.52036011e-01 2.77341790e-02 4.42066103e-01 5.86742342e-01 1.55102843e-02 4.76905495e-01 7.47570038e-01 -2.94357032e-01 -7.79512048e-01 -1.05095911e+00 7.98058659e-02 7.33009934e-01 5.61212361e-01 -6.79899871e-01 -4.79698092e-01 -8.63255501e-01 1.80901829e-02 4.76954341e-01 -6.22417092e-01 -5.45545332e-02 -5.17060459e-01 -4.35743451e-01 3.78593653e-01 6.43666983e-01 4.63623166e-01 -1.06053591e+00 -3.09872240e-01 3.68182778e-01 -2.29167268e-01 -1.56225491e+00 -3.74166727e-01 -5.30458661e-03 -7.47158587e-01 -1.30901468e+00 -1.46347895e-01 -6.98368967e-01 9.25950646e-01 3.23851824e-01 1.36132276e+00 3.13129574e-01 -5.89209199e-01 2.28236452e-01 -3.58864516e-01 -1.96215943e-01 3.84336412e-02 -8.40708315e-02 -4.57174450e-01 2.80249745e-01 6.86857924e-02 -6.31354213e-01 -8.20061862e-01 4.35697198e-01 -6.55682087e-01 5.28950751e-01 9.29832220e-01 7.08235562e-01 1.06451273e+00 2.84122169e-01 3.28675330e-01 -1.01553810e+00 5.53001426e-02 -5.82740426e-01 -5.03929973e-01 4.40403193e-01 -3.82162422e-01 4.01750654e-02 4.72488463e-01 -3.78905177e-01 -1.15533006e+00 4.21359599e-01 2.30767563e-01 -6.18293405e-01 -1.68660969e-01 4.05533582e-01 -4.32960302e-01 2.96328776e-02 4.20176417e-01 1.73566073e-01 -1.95818752e-01 -2.90436327e-01 6.21438682e-01 1.48886755e-01 7.13308096e-01 -6.63371861e-01 9.32468891e-01 5.19065440e-01 3.08557153e-01 -5.07958949e-01 -1.21256733e+00 -6.10450268e-01 -7.68509388e-01 -4.02993321e-01 9.06222641e-01 -1.08629131e+00 -7.85943925e-01 3.33574027e-01 -1.29843271e+00 -3.32246572e-01 -4.93655950e-01 3.54899704e-01 -4.50845242e-01 1.17374629e-01 -4.98737484e-01 -5.87668002e-01 -1.46504804e-01 -9.94534850e-01 1.35467494e+00 2.01323897e-01 2.63856761e-02 -9.82835174e-01 -5.38776159e-01 6.00161552e-01 -2.01363973e-02 3.99496794e-01 8.91883075e-01 -4.69107270e-01 -1.33723629e+00 -5.53071909e-02 -1.01279485e+00 1.54069006e-01 3.27461451e-01 5.78531586e-02 -7.53979325e-01 3.09235573e-01 -6.01542294e-01 -1.46162897e-01 7.35788286e-01 2.89002836e-01 1.34982252e+00 -5.80970943e-01 -6.19487941e-01 4.81371224e-01 1.31273329e+00 -1.71984985e-01 6.37179911e-01 3.71533968e-02 1.18351424e+00 7.94577181e-01 8.30490291e-01 5.52895606e-01 9.62902367e-01 7.35971808e-01 6.13227367e-01 -4.41878289e-01 -5.67439377e-01 -3.21800202e-01 -2.12638099e-02 4.27878171e-01 -1.15720578e-01 -3.24231178e-01 -8.33274722e-01 5.07384479e-01 -2.06539464e+00 -7.64777660e-01 -5.02199888e-01 1.76978981e+00 7.67065287e-01 1.49795428e-01 4.09070328e-02 -2.57636160e-01 6.43328488e-01 1.18540004e-01 -4.66951549e-01 2.91696131e-01 -1.43666252e-01 9.24619511e-02 3.31342965e-01 4.81929392e-01 -1.04161656e+00 1.39843142e+00 5.44682741e+00 1.07859683e+00 -8.08484614e-01 7.83245824e-03 7.27084100e-01 2.95751750e-01 -4.00421500e-01 2.77345121e-01 -8.34843159e-01 2.26382446e-02 6.78233281e-02 2.19715703e-02 1.98668242e-01 7.91178107e-01 6.02443814e-02 -2.70024180e-01 -1.30131221e+00 8.92452300e-01 -2.45727878e-02 -1.79275167e+00 3.29566121e-01 1.70759559e-01 6.39786959e-01 -2.30876461e-01 -3.71710241e-01 5.69478422e-02 3.89888674e-01 -9.64024067e-01 8.29564095e-01 7.98578620e-01 7.50657976e-01 -4.08151031e-01 5.25816619e-01 2.40417391e-01 -1.89423788e+00 1.58378005e-01 -1.55986231e-02 4.99621034e-02 3.06793034e-01 1.05965853e+00 -1.12985337e+00 9.50164437e-01 4.65498447e-01 1.09444153e+00 -5.93410969e-01 8.88976395e-01 -4.13786978e-01 5.19330025e-01 -4.30461049e-01 2.28386924e-01 1.09347485e-01 -1.05351672e-01 4.74315256e-01 1.19324732e+00 -2.85262000e-02 3.78694624e-01 4.98099804e-01 1.13742089e+00 -2.05795556e-01 9.53768343e-02 -3.15606564e-01 1.10057354e-01 5.87844789e-01 1.74554408e+00 -1.05649948e+00 -5.42079866e-01 -4.67528552e-01 7.92613328e-01 5.98215103e-01 2.79921204e-01 -7.50900924e-01 -2.45188072e-01 5.48427105e-01 3.36065859e-01 4.70986485e-01 -3.66188101e-02 -3.67465913e-01 -8.72995913e-01 2.32933670e-01 -3.13415915e-01 3.96169424e-01 -9.07956719e-01 -1.46755445e+00 6.44782364e-01 1.56442240e-01 -1.24597597e+00 1.52590815e-02 -3.46737891e-01 -5.39824009e-01 8.49121332e-01 -1.81075907e+00 -1.85040772e+00 -6.44598067e-01 7.74699330e-01 3.76675099e-01 1.90847740e-02 6.51439428e-01 3.50713521e-01 -6.78588569e-01 2.99219072e-01 -7.18181491e-01 3.90699029e-01 3.76803249e-01 -1.21648335e+00 1.88155025e-01 7.57302225e-01 2.39033177e-01 3.84153515e-01 2.62398362e-01 -7.63395190e-01 -1.01651430e+00 -1.70936716e+00 8.42882812e-01 -4.53479856e-01 8.10047567e-01 -3.60829145e-01 -6.21301353e-01 7.78691530e-01 -2.23005325e-01 6.23736680e-01 2.94093251e-01 5.13890572e-02 -5.34570873e-01 -3.14308733e-01 -9.05287623e-01 6.43715680e-01 1.37807274e+00 -4.64855820e-01 -2.22811908e-01 5.64118922e-01 7.06911504e-01 -5.07301390e-01 -1.01342094e+00 5.05963326e-01 4.51323122e-01 -7.68579960e-01 1.20329404e+00 -4.19357121e-01 7.26670265e-01 -5.91229558e-01 -8.00419450e-02 -7.23047435e-01 -4.90335226e-01 -4.35825020e-01 -4.99893874e-01 1.36319482e+00 3.85961086e-01 -4.33280736e-01 8.49207759e-01 4.31088805e-01 -1.03529766e-01 -1.22374606e+00 -5.11986077e-01 -6.41825199e-01 -6.06530070e-01 -5.34368873e-01 6.66442096e-01 8.70269299e-01 -3.12288105e-01 6.75891042e-01 -3.16773444e-01 4.16030467e-01 8.19982946e-01 2.63027757e-01 9.91491854e-01 -1.26316738e+00 -3.33973706e-01 -4.35220987e-01 -6.71839476e-01 -1.26214790e+00 1.54929802e-01 -9.19402242e-01 2.88879918e-03 -2.03530240e+00 3.93667459e-01 -9.86967146e-01 -3.27061474e-01 8.09494853e-01 -3.45560193e-01 2.47249216e-01 1.03470549e-01 1.81370720e-01 -9.21585619e-01 5.34995317e-01 1.62511492e+00 -2.17991963e-01 -1.21323369e-01 -1.70782685e-01 -7.57694066e-01 8.49232554e-01 4.06378359e-01 -4.92364079e-01 -5.31183004e-01 -2.61926025e-01 -1.36417197e-02 1.21575236e-01 8.05025816e-01 -6.90029085e-01 2.73754746e-01 -3.40274513e-01 3.40267539e-01 -7.08948851e-01 5.35087645e-01 -6.50075614e-01 6.74329698e-02 6.63886294e-02 -1.99992269e-01 -3.54667634e-01 -1.59872174e-01 6.53901696e-01 -3.41993958e-01 2.26030827e-01 4.54067469e-01 -2.30596662e-01 -8.97178590e-01 6.81272805e-01 2.54904717e-01 -3.49937901e-02 1.28455424e+00 -3.01433355e-01 -3.72821331e-01 -1.63499594e-01 -7.10174918e-01 5.27739704e-01 -1.17906053e-02 4.28334266e-01 7.61600852e-01 -1.21444440e+00 -8.16488743e-01 -8.51603821e-02 4.70390707e-01 7.21207500e-01 4.09179717e-01 1.03380632e+00 -2.11177930e-01 9.81104821e-02 6.46339506e-02 -7.89218426e-01 -1.33907640e+00 3.02681208e-01 2.15645179e-01 -6.05882943e-01 -9.13115323e-01 1.15138960e+00 7.07526147e-01 -2.33347550e-01 -4.76819724e-02 -3.78357619e-01 -2.89270371e-01 -4.57321778e-02 3.03245872e-01 1.44250318e-01 -1.39684767e-01 -7.60189533e-01 -4.45074588e-01 5.07125735e-01 -1.83018148e-01 1.09470516e-01 1.19359493e+00 -2.73936596e-02 -3.88613164e-01 1.32699311e-01 9.56726193e-01 -2.05912367e-01 -1.31689370e+00 -6.94589496e-01 -7.78963715e-02 -5.88422418e-01 7.01265931e-02 -7.14115202e-01 -1.28524852e+00 5.95295846e-01 -2.78399661e-02 -1.36377722e-01 1.18031287e+00 6.40044391e-01 7.12980926e-01 2.35165507e-01 3.08047891e-01 -7.88356841e-01 4.93601412e-01 2.26739526e-01 8.30959439e-01 -1.25442314e+00 3.20443720e-01 -1.37401259e+00 -7.23186493e-01 8.53157878e-01 9.21428144e-01 3.26336361e-02 7.49565601e-01 1.53421596e-01 -4.12884980e-01 -4.88867790e-01 -7.66319156e-01 -5.32893240e-01 6.85420215e-01 6.13697052e-01 1.26924992e-01 2.30145559e-01 2.19639152e-01 4.29902375e-01 3.66618223e-02 8.66233781e-02 1.14590034e-01 7.06619322e-01 -2.61221439e-01 -1.13132143e+00 -8.07328522e-02 7.94699073e-01 1.79198384e-02 -2.39891738e-01 -2.05888644e-01 6.09136641e-01 6.17864132e-01 8.46296787e-01 2.28425637e-01 -5.39706469e-01 2.97143787e-01 -4.51275885e-01 4.67645854e-01 -9.08634543e-01 -4.24150527e-01 1.26046121e-01 2.84711808e-01 -8.07407439e-01 -6.12154663e-01 -6.05035424e-01 -1.64966393e+00 7.28711933e-02 -4.32906538e-01 -2.92414904e-01 2.69476742e-01 1.18601501e+00 4.80962068e-01 8.81166518e-01 5.32904088e-01 -7.01193869e-01 1.00632012e-01 -7.45612085e-01 -5.96756041e-01 4.66476172e-01 7.10417107e-02 -8.76912534e-01 2.10040420e-01 2.54142255e-01]
[10.303046226501465, 1.6414434909820557]
1df59c62-7091-4a73-9299-8b8dcba89f32
meetingbank-a-benchmark-dataset-for-meeting
2305.17529
null
https://arxiv.org/abs/2305.17529v1
https://arxiv.org/pdf/2305.17529v1.pdf
MeetingBank: A Benchmark Dataset for Meeting Summarization
As the number of recorded meetings increases, it becomes increasingly important to utilize summarization technology to create useful summaries of these recordings. However, there is a crucial lack of annotated meeting corpora for developing this technology, as it can be hard to collect meetings, especially when the topics discussed are confidential. Furthermore, meeting summaries written by experienced writers are scarce, making it hard for abstractive summarizers to produce sensible output without a reliable reference. This lack of annotated corpora has hindered the development of meeting summarization technology. In this paper, we present MeetingBank, a new benchmark dataset of city council meetings over the past decade. MeetingBank is unique among other meeting corpora due to its divide-and-conquer approach, which involves dividing professionally written meeting minutes into shorter passages and aligning them with specific segments of the meeting. This breaks down the process of summarizing a lengthy meeting into smaller, more manageable tasks. The dataset provides a new testbed of various meeting summarization systems and also allows the public to gain insight into how council decisions are made. We make the collection, including meeting video links, transcripts, reference summaries, agenda, and other metadata, publicly available to facilitate the development of better meeting summarization techniques. Our dataset can be accessed at: https://meetingbank.github.io
['Fei Liu', 'Hassan Foroosh', 'Franck Dernoncourt', 'Hanieh Deilamsalehy', 'Tim Ganter', 'Yebowen Hu']
2023-05-27
null
null
null
null
['meeting-summarization']
['natural-language-processing']
[ 2.70077407e-01 2.94498205e-01 -1.17597193e-01 -3.63735229e-01 -1.64367652e+00 -9.80118155e-01 5.26749134e-01 7.86520600e-01 -2.43518919e-01 1.05133235e+00 1.14554214e+00 -1.56204745e-01 1.03038400e-01 -2.79785663e-01 -2.39808902e-01 -2.59165823e-01 3.55175972e-01 4.06523466e-01 -4.50271778e-02 -1.80810347e-01 5.46189725e-01 2.21453067e-02 -1.21141839e+00 6.32003069e-01 1.05913413e+00 9.65712517e-02 3.47204059e-01 9.71642733e-01 -1.19702563e-01 4.24729675e-01 -1.44939935e+00 -3.19951802e-01 -1.89098224e-01 -8.10434639e-01 -9.81182814e-01 1.81693479e-01 5.36817849e-01 -2.63001829e-01 -1.74788877e-01 6.52417541e-01 6.43353581e-01 3.60864669e-01 5.04895329e-01 -8.38302612e-01 -2.47944936e-01 9.55176413e-01 -4.77634013e-01 4.56556112e-01 9.31290388e-01 -1.62196547e-01 1.23369873e+00 -3.67870241e-01 7.51435459e-01 8.53569388e-01 3.94068033e-01 4.84580427e-01 -8.80299866e-01 -5.05593896e-01 1.62596405e-01 -1.83869287e-01 -1.00617671e+00 -8.86228979e-01 5.06544411e-01 -4.40803915e-01 9.88547862e-01 7.61772096e-01 6.45527065e-01 1.03177297e+00 7.74707347e-02 7.04332352e-01 4.08105373e-01 -5.69418490e-01 -2.98260432e-02 1.46924734e-01 3.43642324e-01 2.47767851e-01 6.35654271e-01 -9.66953814e-01 -7.93890834e-01 -3.79807949e-01 3.30614924e-01 -2.42794380e-01 -7.44929254e-01 3.04916263e-01 -1.35032415e+00 4.26670730e-01 -1.45823345e-01 6.07582629e-01 -2.84197301e-01 -1.68986708e-01 6.36182487e-01 1.79201275e-01 6.21342599e-01 7.48409092e-01 -2.88218670e-02 -5.90989113e-01 -1.40659952e+00 5.54883361e-01 1.23958600e+00 8.50175679e-01 3.25319111e-01 -2.31130481e-01 -2.77959794e-01 8.61571491e-01 -7.28390887e-02 1.69607833e-01 3.09703827e-01 -1.16217089e+00 9.97107267e-01 5.26377082e-01 4.36767578e-01 -1.17219353e+00 -2.17094138e-01 -1.97289661e-01 -7.26885796e-01 -4.85844254e-01 3.97546530e-01 -3.74854535e-01 -3.39720815e-01 1.32983387e+00 -1.30633041e-01 -2.88510382e-01 8.53164345e-02 4.88249063e-01 1.42766857e+00 9.72265482e-01 -3.00494075e-01 -7.44661808e-01 1.41079831e+00 -8.85133862e-01 -1.07993102e+00 -1.12615697e-01 7.53922999e-01 -1.20970321e+00 8.60979021e-01 3.19154382e-01 -1.62529457e+00 -2.74767429e-01 -1.06003654e+00 -4.70069721e-02 6.69512749e-02 4.91835102e-02 2.13941529e-01 2.90129006e-01 -1.12778509e+00 4.93631005e-01 -8.92893493e-01 -7.98573911e-01 1.66676134e-01 1.30643919e-01 -3.26349825e-01 -2.79236492e-02 -6.90650344e-01 7.12446868e-01 1.04204021e-01 2.21312176e-02 -8.14530626e-02 -6.22700155e-01 -1.10712910e+00 1.21888043e-02 2.03758553e-01 -6.15828037e-01 1.85860527e+00 -5.36350906e-01 -1.18671751e+00 7.61612952e-01 -5.81661940e-01 -7.60809854e-02 4.21235293e-01 -2.50906646e-01 -3.04393768e-01 1.80673659e-01 4.20595944e-01 3.02666962e-01 2.02714518e-01 -9.60275412e-01 -7.64415562e-01 -7.43535534e-02 -3.50839235e-02 5.35221040e-01 -2.96387285e-01 2.60950446e-01 -6.03633702e-01 -4.93482709e-01 -7.08614737e-02 -6.18650973e-01 -1.36991307e-01 -9.73705232e-01 -6.16042256e-01 -3.42776775e-01 4.90929276e-01 -1.18735111e+00 1.81884694e+00 -2.10272932e+00 1.10785097e-01 -2.26556346e-01 2.87684023e-01 1.57280892e-01 3.23702884e-03 1.11881554e+00 2.61242688e-01 4.82931048e-01 -2.26713903e-02 -5.12015760e-01 -6.18860908e-02 -1.82607085e-01 -2.08761692e-01 3.43098760e-01 -5.03030606e-02 5.79062879e-01 -9.89252329e-01 -6.32754743e-01 -1.23564497e-01 9.49998274e-02 -3.94415706e-01 1.51388496e-01 -2.67335046e-02 7.05681205e-01 -6.77112579e-01 3.91140133e-01 3.16240072e-01 -2.15122849e-01 2.01041162e-01 1.15810938e-01 -5.73673666e-01 1.02704203e+00 -1.05421841e+00 1.89591515e+00 -2.11626843e-01 1.13666224e+00 1.06029414e-01 -8.41622710e-01 9.78272736e-01 7.08480775e-01 3.93375665e-01 -1.79651335e-01 -3.16771716e-02 2.16700867e-01 -2.56704450e-01 -5.21800697e-01 1.11013293e+00 1.36794806e-01 -6.47313714e-01 7.23389745e-01 -3.41947526e-01 -6.92399859e-01 9.62148726e-01 6.51501596e-01 1.39087307e+00 -3.58444303e-01 5.18095553e-01 2.72003021e-02 3.00637186e-01 5.31632185e-01 6.44230008e-01 6.75020933e-01 -1.08280055e-01 8.90781939e-01 4.43579823e-01 -1.62664458e-01 -1.03201866e+00 -6.72324300e-01 8.73694345e-02 6.30669713e-01 -2.42272541e-01 -8.24362278e-01 -8.39127958e-01 -1.68345153e-01 -3.48408818e-01 7.11782992e-01 -9.99106020e-02 2.91115671e-01 -8.64398241e-01 -3.19668591e-01 4.78053510e-01 2.50064820e-01 4.36114579e-01 -1.01196396e+00 -4.25550073e-01 4.33525532e-01 -1.07266414e+00 -9.80652750e-01 -9.33780313e-01 -1.20045185e-01 -6.71474516e-01 -9.04826343e-01 -8.17428112e-01 -9.47280884e-01 4.72545385e-01 4.24275160e-01 1.25237417e+00 1.46185279e-01 -1.25795901e-01 4.35662627e-01 -4.04354334e-01 -7.12364137e-01 -7.12188244e-01 6.10635281e-01 -1.06423371e-01 -5.99590182e-01 3.19077164e-01 -4.30614859e-01 -3.39706808e-01 -2.60425806e-02 -6.84971273e-01 2.20734537e-01 2.84309894e-01 4.14940953e-01 2.14579418e-01 -1.94405034e-01 9.21198785e-01 -9.07038093e-01 1.12082851e+00 -2.36316934e-01 -1.52472883e-01 1.90270394e-01 1.44145042e-01 -4.70667541e-01 2.96475530e-01 8.28759074e-02 -1.03676105e+00 -4.69488770e-01 1.04412496e-01 5.38745463e-01 5.96891216e-04 8.00832272e-01 4.26937826e-02 7.21609592e-01 5.70944130e-01 -8.74175057e-02 -2.91370839e-01 -4.45420802e-01 -9.56949070e-02 1.14674401e+00 6.92873359e-01 -2.46091858e-01 5.75589478e-01 3.36384475e-02 -8.02151322e-01 -1.27181482e+00 -1.15987396e+00 -1.09849787e+00 -4.74031627e-01 -3.31114352e-01 8.03839803e-01 -1.04441297e+00 -4.72562194e-01 -5.03713125e-03 -1.26546419e+00 -3.34491581e-01 -3.14091861e-01 5.95804989e-01 -3.20451677e-01 4.74231333e-01 -6.77341938e-01 -6.10027134e-01 -5.00278234e-01 -7.09358811e-01 8.27877939e-01 5.79523742e-01 -1.30453992e+00 -8.62543106e-01 2.72169888e-01 8.61939311e-01 2.17837002e-02 5.34517109e-01 5.70938110e-01 -7.48429358e-01 -2.90833771e-01 -4.58048970e-01 1.38788149e-01 1.90895453e-01 3.45488280e-01 4.67802256e-01 -6.16881311e-01 -2.83680856e-01 -2.40751281e-01 -2.28669316e-01 7.22044826e-01 6.69324160e-01 6.86302245e-01 -5.85016072e-01 -3.97361368e-01 1.13773890e-01 8.44092369e-01 1.97280142e-02 4.53009993e-01 4.81253892e-01 2.94832706e-01 6.01492941e-01 6.46250427e-01 5.95825970e-01 6.46987140e-01 4.73820150e-01 -2.83603162e-01 1.86983645e-01 2.05765087e-02 -7.07583502e-02 4.57697064e-01 1.88701749e+00 -2.33030334e-01 -5.57079792e-01 -9.99583542e-01 9.35704648e-01 -1.89998651e+00 -1.26766002e+00 -2.94076055e-01 2.11006284e+00 1.03402495e+00 7.37282634e-02 1.72094271e-01 1.38526708e-01 7.99698770e-01 4.33411390e-01 -6.47592843e-02 -7.28896201e-01 3.08379624e-02 -4.21144590e-02 1.03562146e-01 4.75984812e-01 -8.60117555e-01 4.12009150e-01 5.99769974e+00 3.86482805e-01 -7.29706526e-01 -5.43676242e-02 4.16281164e-01 -5.13909578e-01 -3.39977384e-01 -2.35626310e-01 -9.50940013e-01 3.09538901e-01 1.19867742e+00 -8.61035407e-01 -5.37661090e-02 3.81440073e-01 8.54001820e-01 -3.82114172e-01 -1.28891408e+00 8.11841011e-01 1.54704481e-01 -1.71766710e+00 -1.35524899e-01 2.14171335e-02 9.91382778e-01 -1.53836712e-01 -3.53175193e-01 2.90005744e-01 4.64363933e-01 -9.01882708e-01 5.39641976e-01 1.91344470e-01 6.59582615e-01 -7.01362789e-01 8.28515530e-01 4.77534711e-01 -1.12096071e+00 2.63961554e-01 -3.79594564e-01 -4.11582261e-01 4.39522892e-01 5.46158969e-01 -1.03935611e+00 7.99160302e-01 5.67475259e-01 8.04619670e-01 -2.21247256e-01 1.28297985e+00 -6.29373044e-02 6.90680146e-01 -1.65108874e-01 -3.85484844e-02 2.50258327e-01 -1.71630204e-01 6.95352674e-01 1.48104954e+00 5.40400445e-01 3.34203720e-01 6.49086475e-01 1.99713498e-01 -4.31043088e-01 2.37948254e-01 -6.05913997e-01 -3.55712950e-01 8.67416501e-01 1.24072719e+00 -7.33528793e-01 -4.08279657e-01 -2.95406371e-01 7.33044982e-01 9.85540673e-02 3.56558889e-01 -4.40745562e-01 -6.40357912e-01 3.79150093e-01 9.24083292e-02 1.04630599e-02 -3.37963521e-01 -3.60200882e-01 -1.09365952e+00 2.28127182e-01 -1.24678004e+00 5.02729714e-01 -4.87295508e-01 -6.47913396e-01 5.28851509e-01 6.04951531e-02 -1.08038485e+00 -4.08993214e-01 4.56049174e-01 -9.64086771e-01 7.94980764e-01 -1.17755198e+00 -6.94641829e-01 -6.32318735e-01 -9.78931412e-02 1.16835880e+00 1.84773460e-01 7.86984921e-01 2.46840343e-01 -7.98106253e-01 4.37173039e-01 1.36384517e-01 1.11440614e-01 1.16440463e+00 -1.11138046e+00 3.19914639e-01 5.79044580e-01 8.27694908e-02 5.14944136e-01 1.07075858e+00 -6.92724764e-01 -9.38994586e-01 -9.25083101e-01 1.42746687e+00 -6.60873234e-01 5.93580782e-01 -7.46852756e-02 -9.34305429e-01 8.62029493e-01 6.98814571e-01 -9.97826576e-01 1.03402579e+00 1.29340202e-01 3.23350906e-01 3.23559158e-02 -5.98825037e-01 6.70445383e-01 7.74993837e-01 -2.41306484e-01 -1.01549256e+00 5.19913971e-01 6.62508070e-01 -5.58066905e-01 -8.33806455e-01 -1.82139426e-01 2.81908602e-01 -5.76625943e-01 2.91279256e-01 -8.38932991e-02 5.12563109e-01 -7.84940496e-02 2.61061251e-01 -1.51916778e+00 -7.80730397e-02 -1.17545033e+00 3.84077936e-01 1.85237885e+00 7.08292663e-01 -2.89131552e-01 6.59345925e-01 6.28428757e-01 -6.31929755e-01 -3.17132920e-01 -5.83917916e-01 -4.76326525e-01 -1.62501335e-01 -5.22467755e-02 4.03378010e-01 9.78168547e-01 6.97021306e-01 5.57176292e-01 -4.94237393e-02 -1.03789762e-01 2.60485470e-01 8.63514096e-02 1.02754593e+00 -1.38720703e+00 -7.26568103e-02 -4.92319942e-01 3.46718803e-02 -1.00988245e+00 -1.08149752e-01 -7.36473143e-01 2.35057384e-01 -2.42273641e+00 3.75149190e-01 -1.58046648e-01 2.09074974e-01 2.41671115e-01 -2.02841043e-01 1.52072296e-01 1.70275927e-01 5.53990066e-01 -8.55841756e-01 1.00278184e-01 1.27983749e+00 -1.60522237e-01 -7.77907729e-01 2.02709213e-01 -1.08751547e+00 8.21674347e-01 9.26707208e-01 -3.57104838e-01 -1.87093124e-01 -5.43446600e-01 1.63953632e-01 3.11007559e-01 -3.16811174e-01 -1.03925967e+00 3.75116765e-01 -1.83016330e-01 1.08507328e-01 -1.00060856e+00 2.11924985e-01 -1.28973937e-02 1.72169149e-01 1.61104262e-01 -5.24761796e-01 2.27420315e-01 3.27593863e-01 3.43529642e-01 -5.54275393e-01 -2.06867129e-01 2.54788250e-01 -3.53999555e-01 -1.26234323e-01 -1.82418317e-01 -6.69515729e-01 4.75908697e-01 8.12658787e-01 -5.06906390e-01 -4.31542635e-01 -8.86645973e-01 -5.70684910e-01 4.93317097e-01 7.08843708e-01 9.88669470e-02 1.80730626e-01 -8.84374678e-01 -1.25914705e+00 -4.67015207e-01 4.48949933e-02 5.36866426e-01 3.23763847e-01 9.55320954e-01 -6.77854598e-01 6.42316341e-01 3.26743722e-02 -4.03997421e-01 -1.64310741e+00 -3.33793640e-01 -2.91327298e-01 -5.63925445e-01 -1.01113200e+00 6.56827867e-01 1.73570439e-01 -7.21169040e-02 4.59958464e-01 -4.42897946e-01 -5.36151588e-01 5.56765318e-01 1.05260003e+00 3.75828981e-01 1.45065621e-01 -3.90983403e-01 -2.55852342e-01 1.64279073e-01 -2.99049079e-01 -2.29795009e-01 1.61570036e+00 -4.70535964e-01 -9.84685346e-02 9.19057727e-01 8.57311606e-01 4.30106640e-01 -8.86619747e-01 1.94835197e-02 1.78148508e-01 -2.65286893e-01 -2.34747574e-01 -5.03207326e-01 -2.42587447e-01 4.40631360e-01 -6.60955071e-01 6.75683260e-01 8.47944558e-01 7.67721608e-02 9.94625568e-01 3.42208117e-01 -7.12674484e-02 -1.13543022e+00 1.36908680e-01 5.85487545e-01 1.14859951e+00 -1.06360304e+00 3.66695940e-01 -3.78656656e-01 -5.91426849e-01 9.60426927e-01 2.47404113e-01 1.77261606e-01 1.98066458e-02 1.42220706e-01 2.19018802e-01 -2.04913706e-01 -9.35006142e-01 3.12709391e-01 1.60150886e-01 5.05607426e-01 9.02322769e-01 1.20648675e-01 -6.34427547e-01 6.25402272e-01 -6.69277370e-01 -1.38420686e-01 1.37879503e+00 9.78790045e-01 -6.85385168e-01 -1.06322169e+00 -4.91806597e-01 6.23596609e-01 -8.23961735e-01 -3.66388401e-03 -6.46081626e-01 9.25330281e-01 -6.03462398e-01 1.32919085e+00 2.68444359e-01 2.75032818e-01 5.11439025e-01 7.50985667e-02 2.12873071e-01 -1.35844529e+00 -8.59928012e-01 1.54557362e-01 8.63039970e-01 -5.90802543e-02 -4.71723139e-01 -9.54501629e-01 -1.12683702e+00 -8.25863838e-01 -2.55656570e-01 7.70343781e-01 6.41977549e-01 8.04351926e-01 3.36867422e-01 8.16497922e-01 2.79456556e-01 -9.43711698e-01 -4.37955886e-01 -1.15278149e+00 -4.16975796e-01 2.01762587e-01 3.77666265e-01 -4.67021158e-03 -2.74784029e-01 4.86690193e-01]
[12.616031646728516, 9.416236877441406]
058e78ac-e6e2-4e27-bbf3-d075d36ccd2c
revisiting-the-stability-of-stochastic
null
null
https://openreview.net/forum?id=oQyb8NrFzu
https://openreview.net/pdf?id=oQyb8NrFzu
Revisiting the Stability of Stochastic Gradient Descent: A Tightness Analysis
The technique of algorithmic stability has been used to capture the generalization power of several learning models, especially those trained with stochastic gradient descent (SGD). This paper investigates the tightness of the algorithmic stability bounds for SGD given by~\cite{hardt2016train}. We show that the analysis of~\cite{hardt2016train} is tight for convex objective functions, but loose for non-convex objective functions. In the non-convex case we provide a tighter upper bound on the stability (and hence generalization error), and provide evidence that it is asymptotically tight up to a constant factor. However, deep neural networks trained with SGD exhibit much better stability and generalization in practice than what is suggested by these (tight) bounds, namely, linear or exponential degradation with time for SGD with constant step size. We aim towards characterizing deep learning loss functions with good generalization guarantees, despite being trained using SGD with constant step size. We propose the Hessian Contractive (HC) condition, which specifies the contractivity of regions containing local minima in the neural network loss landscape. We provide empirical evidence that this condition holds for several loss functions, and provide theoretical evidence that the known tight SGD stability bounds for convex and non-convex loss functions can be circumvented by HC loss functions, thus partially explaining the generalization of deep neural networks.
['Mayank Goswami', 'Chao Chen', 'Vamsi Pritham Pingali', 'Wenjia Zhang', 'Samuel Bald', 'Yikai Zhang']
2021-01-01
null
null
null
null
['exponential-degradation']
['time-series']
[-1.62489891e-01 3.41379166e-01 -4.68218029e-02 -4.42425638e-01 -7.53523588e-01 -5.80511868e-01 3.65779437e-02 8.12521353e-02 -7.51684189e-01 8.60210717e-01 -1.92085326e-01 -4.56524581e-01 -4.93009478e-01 -3.49916786e-01 -1.10203016e+00 -1.09482884e+00 -3.73208910e-01 1.83258951e-01 1.02134138e-01 -2.73198664e-01 -1.78749725e-01 6.06456935e-01 -1.31878507e+00 -3.34781915e-01 8.55031788e-01 1.25136793e+00 -1.02502108e-01 5.54069400e-01 2.50876933e-01 3.05636704e-01 -2.86433637e-01 -6.25080526e-01 5.84380090e-01 -2.05262721e-01 -8.24877620e-01 -3.09966803e-01 7.98504949e-01 -1.37364149e-01 -3.15696090e-01 1.45279384e+00 5.84692121e-01 3.34196121e-01 6.66405678e-01 -1.34907031e+00 -4.52999771e-01 5.90893090e-01 -3.20783168e-01 2.95160472e-01 -2.77550071e-01 -1.47420913e-01 1.17869031e+00 -7.96507955e-01 4.02805388e-01 1.11448753e+00 1.23507273e+00 8.98463845e-01 -1.41781533e+00 -4.84375894e-01 3.46787095e-01 -1.69344738e-01 -1.33925462e+00 -2.95700103e-01 4.11938876e-01 -4.44080055e-01 8.67351174e-01 3.48982543e-01 2.64660060e-01 8.70517015e-01 3.11186850e-01 9.31213796e-01 7.18443632e-01 -2.88198650e-01 2.86044180e-01 4.14369792e-01 4.53109145e-01 9.59451199e-01 4.33076829e-01 6.67305067e-02 -2.61554092e-01 -7.81298354e-02 6.33190870e-01 -2.45454147e-01 -5.28067410e-01 -5.35484850e-01 -5.47926545e-01 1.01735330e+00 6.16733611e-01 3.12743038e-01 -1.07909314e-01 3.07237774e-01 6.85649514e-01 7.20139265e-01 6.85661972e-01 4.04119313e-01 -6.23180330e-01 -1.69575602e-01 -8.64251971e-01 3.25364918e-01 1.01010132e+00 9.35660541e-01 5.45482934e-01 2.08202839e-01 5.53145856e-02 8.14945161e-01 1.36967748e-01 3.53823662e-01 4.58480000e-01 -9.91720676e-01 5.98226130e-01 3.12255591e-01 -4.68270667e-03 -7.39662468e-01 -7.42841423e-01 -9.52603996e-01 -8.97334993e-01 5.12923062e-01 7.60893226e-01 -3.88453275e-01 -4.13970381e-01 2.43997049e+00 1.04125559e-01 -3.15273702e-01 5.82672693e-02 8.12172592e-01 2.53173560e-01 4.29740757e-01 -1.75785586e-01 -4.19227362e-01 5.42336166e-01 -6.52767003e-01 -3.74089360e-01 -4.60709296e-02 1.12020516e+00 -9.17566940e-02 1.50295210e+00 4.09885108e-01 -1.25047302e+00 -1.24139816e-01 -1.22367632e+00 -1.65506959e-01 -6.94508255e-02 -5.98323792e-02 2.87979394e-01 7.59200633e-01 -1.31715930e+00 1.08996689e+00 -9.90748227e-01 -1.62776798e-01 4.22061414e-01 6.10915780e-01 -2.89189816e-01 1.89151987e-01 -1.01579726e+00 9.02833700e-01 3.55311573e-01 3.00433636e-01 -7.41509497e-01 -1.09987271e+00 -6.39300227e-01 1.23124383e-01 -2.19879989e-02 -4.52882081e-01 1.12784922e+00 -9.26218629e-01 -1.40745914e+00 1.03902972e+00 8.65989029e-02 -9.22248960e-01 1.09084797e+00 -4.46467817e-01 1.70082346e-01 -1.52043268e-01 -4.07402664e-01 3.52074236e-01 6.53977334e-01 -1.07218575e+00 -3.02947342e-01 -5.97729027e-01 2.13870615e-01 2.14413598e-01 -8.09606969e-01 -1.13729246e-01 1.60174772e-01 -4.06392843e-01 -2.36521028e-02 -9.63311374e-01 -1.06613472e-01 4.02561963e-01 -2.90080905e-01 -4.47047561e-01 6.60890460e-01 -6.04669094e-01 1.38021159e+00 -2.09340668e+00 2.24320620e-01 2.78811634e-01 2.78378844e-01 3.67077529e-01 -1.14795253e-01 1.32301852e-01 -3.46225984e-02 1.71349615e-01 -6.75592482e-01 -4.51349676e-01 5.07737815e-01 2.10563302e-01 -3.20649147e-01 8.66890192e-01 3.04124877e-02 7.24286735e-01 -6.46621644e-01 -5.98893240e-02 -2.96892643e-01 4.44988668e-01 -6.55912340e-01 -2.56686479e-01 4.51390743e-02 -2.16427147e-01 -2.46428177e-01 1.17410384e-01 7.94188678e-01 -1.39935538e-01 -2.49383062e-01 1.99676659e-02 3.89208947e-03 4.68959101e-02 -1.13660312e+00 1.35067952e+00 -4.81712341e-01 9.45714414e-01 6.23882294e-01 -1.48996210e+00 7.67565548e-01 4.97058444e-02 4.81378198e-01 -2.43038565e-01 6.82262108e-02 5.44012189e-01 -1.91058010e-01 -3.10301989e-01 -7.01923221e-02 -6.69775426e-01 2.09863991e-01 2.25164890e-01 4.97476719e-02 8.57765228e-02 -3.76758352e-02 -9.03620124e-02 1.06016588e+00 -1.84079275e-01 -3.25344622e-01 -9.64028180e-01 6.08674467e-01 -4.16696131e-01 4.29145992e-01 6.69937253e-01 -5.00232399e-01 5.09356856e-01 8.53732586e-01 -2.60562807e-01 -1.06701076e+00 -1.12850285e+00 -5.74772477e-01 1.33160603e+00 7.58701144e-03 -5.89068718e-02 -9.91801381e-01 -6.80920601e-01 1.70012727e-01 6.33179426e-01 -8.54037106e-01 -6.36128008e-01 -5.54536402e-01 -1.14707875e+00 8.12259793e-01 5.52162290e-01 5.39039612e-01 -4.39583600e-01 -4.75896418e-01 7.84100071e-02 3.84259105e-01 -7.71012127e-01 -6.82685435e-01 5.91135859e-01 -1.14567077e+00 -8.73947024e-01 -8.44888568e-01 -1.01945114e+00 6.00984752e-01 -2.76090086e-01 1.00252986e+00 4.36072759e-02 -5.27642667e-03 5.68830788e-01 2.14679301e-01 -3.77283454e-01 -2.67426103e-01 3.49376082e-01 4.76504117e-01 -1.36153042e-01 -9.53864083e-02 -5.42403758e-01 -5.96570134e-01 3.37401181e-01 -1.03280556e+00 -3.58472854e-01 2.82281667e-01 8.11126709e-01 5.41282833e-01 1.78901538e-01 5.85504174e-01 -5.86676955e-01 9.02157962e-01 -3.76468092e-01 -7.68708408e-01 2.37714827e-01 -9.09009635e-01 4.56219584e-01 8.34905863e-01 -4.70495969e-01 -9.08793271e-01 -2.08086446e-02 -3.00296903e-01 -5.52813351e-01 4.41225320e-01 4.51108426e-01 -2.23143632e-03 -5.07136166e-01 9.36634123e-01 7.45296627e-02 1.08710408e-01 -6.49079502e-01 1.24156661e-01 1.79467037e-01 4.35577422e-01 -8.51008117e-01 7.89949059e-01 4.60363477e-01 2.89070636e-01 -7.30700850e-01 -1.06293344e+00 -2.20585853e-01 -3.50180537e-01 -2.06911676e-02 5.54001272e-01 -5.22344589e-01 -9.01397347e-01 3.37513357e-01 -7.63469398e-01 -6.14680350e-01 -5.26695609e-01 5.53641260e-01 -7.73720562e-01 5.48381396e-02 -8.92182171e-01 -9.78032768e-01 -4.41559434e-01 -7.93080509e-01 6.53503001e-01 -3.64268273e-02 2.16186911e-01 -1.57502437e+00 4.56287861e-02 -9.39369202e-02 3.85600269e-01 4.91680354e-01 9.65713263e-01 -8.34104657e-01 8.58583227e-02 -2.01207891e-01 -1.62544012e-01 9.72222388e-01 -3.15994918e-01 -7.09018856e-02 -7.68296421e-01 -7.18894422e-01 5.24995327e-01 -3.99274081e-01 1.05024016e+00 7.24267244e-01 1.06100011e+00 -6.97314262e-01 1.46181034e-02 9.73202050e-01 1.69749892e+00 -1.74929291e-01 1.54610276e-01 4.39832598e-01 5.49112320e-01 6.26585305e-01 8.47200155e-02 3.18325937e-01 -9.03332606e-02 4.18210864e-01 5.08934975e-01 5.19239530e-02 1.69244587e-01 3.55671234e-02 6.41813278e-01 7.65342653e-01 -2.26399899e-02 9.60739776e-02 -8.06011260e-01 4.06156391e-01 -1.87023842e+00 -7.60750353e-01 -1.46431938e-01 2.44568276e+00 1.17007828e+00 4.35139090e-01 2.50364661e-01 2.74185240e-01 5.96107244e-01 -1.91882834e-01 -1.00481355e+00 -8.84820223e-01 -4.29914117e-01 1.05645522e-01 8.71813357e-01 9.16452348e-01 -1.02802801e+00 5.47931671e-01 6.92088509e+00 9.51774478e-01 -1.21534061e+00 1.04538955e-01 6.71629190e-01 -4.26341087e-01 -2.75922656e-01 -5.88652492e-01 -9.07730997e-01 2.73486704e-01 9.37133133e-01 -5.05703270e-01 4.44873303e-01 1.04050744e+00 1.81261927e-01 4.36103016e-01 -1.31764913e+00 9.26774442e-01 -3.03325504e-01 -1.21071291e+00 -3.80028933e-01 2.48735920e-01 1.07868946e+00 3.03585231e-01 3.60126108e-01 3.81697744e-01 1.97430566e-01 -9.80382144e-01 7.03002930e-01 1.70339912e-01 5.39667189e-01 -9.93813455e-01 7.26439357e-01 4.59910899e-01 -8.36355567e-01 -2.65409172e-01 -5.89373469e-01 5.55796400e-02 -2.35742271e-01 7.90169358e-01 -4.44512635e-01 1.56822130e-01 6.32987261e-01 6.11942768e-01 -4.39221114e-01 1.18526030e+00 1.32861644e-01 6.48360670e-01 -7.04200983e-01 -1.02160677e-01 5.11863351e-01 -4.07212943e-01 7.15293109e-01 1.34841430e+00 2.01439396e-01 -2.58208692e-01 -2.10758373e-01 8.31061959e-01 -4.34253126e-01 1.17682874e-01 -4.34046596e-01 2.33073637e-01 3.54734957e-02 7.32282758e-01 -3.53527129e-01 8.48689973e-02 -1.84043631e-01 8.46312463e-01 7.06189334e-01 5.69316983e-01 -8.76145840e-01 -6.50301635e-01 8.37000549e-01 1.27667904e-01 2.44332910e-01 -2.06833839e-01 -5.73173165e-01 -1.03708410e+00 7.02321887e-01 -4.82431263e-01 3.94082010e-01 -1.56966552e-01 -1.37398410e+00 4.92600709e-01 5.20173572e-02 -8.14236164e-01 2.90505718e-02 -1.07626343e+00 -5.04882693e-01 6.63091719e-01 -1.18047559e+00 -5.18580794e-01 2.40733996e-01 6.85606718e-01 9.45849717e-02 9.38864946e-02 3.30581933e-01 4.31817770e-01 -8.20268989e-01 1.16881979e+00 8.39703977e-01 -1.16245903e-01 1.96427435e-01 -1.61559355e+00 -1.32528454e-01 7.23094344e-01 -1.07982785e-01 5.47136486e-01 1.07235587e+00 1.66092329e-02 -1.30230916e+00 -1.04639184e+00 6.36644185e-01 -3.33978623e-01 8.48083913e-01 -4.32118565e-01 -1.24877405e+00 5.66387236e-01 -3.87046099e-01 1.67651191e-01 3.66512567e-01 -1.81044284e-02 -4.01118636e-01 -4.78732616e-01 -1.39300537e+00 4.26559538e-01 1.28482437e+00 -5.74758112e-01 -4.10362512e-01 5.45485616e-01 8.10033917e-01 -2.04278171e-01 -1.00036395e+00 4.93352264e-01 4.87691343e-01 -9.14079845e-01 8.50513518e-01 -9.06318069e-01 1.14095651e-01 9.52272862e-02 -2.69852459e-01 -1.26069474e+00 2.06162445e-02 -9.60308135e-01 -3.67642671e-01 8.29921782e-01 6.36150718e-01 -9.18992102e-01 9.40761268e-01 1.02600610e+00 -4.13471431e-01 -1.24751401e+00 -1.52854323e+00 -1.45421565e+00 9.47059870e-01 -4.31010485e-01 -1.02225132e-01 7.21017182e-01 6.37400448e-02 -3.69079970e-02 7.65779838e-02 -4.27637994e-03 6.28114581e-01 -4.04976547e-01 2.15589032e-02 -1.19572115e+00 -1.85914829e-01 -1.00852966e+00 -5.29585183e-01 -1.05132747e+00 4.63875264e-01 -9.60237384e-01 6.06831685e-02 -1.15091586e+00 -3.60464789e-02 -4.68921840e-01 -6.46109343e-01 3.74821901e-01 6.07658885e-02 1.48960892e-02 4.46668118e-02 1.73606575e-01 -4.66208190e-01 7.15488911e-01 1.15198767e+00 -3.02711576e-02 -4.15966094e-01 3.32566172e-01 -6.23472810e-01 8.28789890e-01 7.59209514e-01 -3.07193577e-01 -5.27927697e-01 -5.82750499e-01 5.29893339e-01 -4.99481589e-01 2.77148217e-01 -1.13993526e+00 1.68870330e-01 2.03666508e-01 -2.66885739e-02 -2.66091861e-02 4.75720577e-02 -6.99508309e-01 -3.07937443e-01 8.10114324e-01 -8.54569912e-01 -1.07366024e-02 3.29862773e-01 4.71544951e-01 1.82295293e-02 -4.80388641e-01 1.16079676e+00 4.07312214e-01 -1.04964245e-02 5.28170049e-01 2.08729673e-02 6.11364305e-01 7.83678412e-01 -6.27587438e-02 -6.56554922e-02 -3.74948829e-01 -8.54308844e-01 2.21130654e-01 1.97477624e-01 1.13158941e-01 4.61474597e-01 -1.42311490e+00 -7.61839688e-01 -1.70015858e-03 -2.78715283e-01 2.02330705e-02 -7.70289078e-02 1.14405906e+00 -5.72050631e-01 3.45152736e-01 1.53355196e-01 -4.79944795e-01 -9.75666404e-01 5.09224355e-01 8.83213222e-01 -3.12531233e-01 -6.09640777e-01 1.33005881e+00 3.93514872e-01 -2.28718221e-01 7.81909883e-01 -7.30922699e-01 4.19848293e-01 -1.17240988e-01 2.89502591e-01 6.35710537e-01 3.29003304e-01 -3.17001551e-01 -4.17487502e-01 5.70651114e-01 -1.25362743e-02 -1.01682588e-01 1.42805588e+00 -2.56976560e-02 5.17287999e-02 4.72724319e-01 2.10456252e+00 -3.68118852e-01 -1.84615076e+00 -3.10181081e-01 1.78274050e-01 -1.14026591e-02 3.29129919e-02 -5.56214511e-01 -1.19365466e+00 9.95862186e-01 8.82632852e-01 4.95980412e-01 1.08640254e+00 -2.68481579e-02 7.05282807e-01 7.59498179e-01 8.64534155e-02 -1.32010114e+00 -2.58201867e-01 6.00654840e-01 1.25160992e+00 -9.76646662e-01 -2.41418481e-01 7.67231137e-02 -2.44379327e-01 1.17015862e+00 4.43442136e-01 -3.90965462e-01 8.91091406e-01 1.96951851e-01 -3.52599055e-01 1.46470323e-01 -5.89962304e-01 1.50191933e-01 3.71931762e-01 4.03299481e-01 2.03454748e-01 -8.85478556e-02 -3.68404001e-01 5.96657395e-01 -4.91883546e-01 -2.98410743e-01 1.03213072e-01 5.66183329e-01 -5.72670877e-01 -5.87044656e-01 7.60446489e-02 3.03878993e-01 -5.55800200e-01 -6.14445508e-02 -3.11794400e-01 8.15066218e-01 -1.29992470e-01 5.23560524e-01 -1.62419707e-01 -3.32094908e-01 3.91679496e-01 8.18542317e-02 6.34288907e-01 -5.42325601e-02 -4.58889991e-01 -4.76302266e-01 -1.04170740e-01 -5.65861583e-01 -8.48208591e-02 -5.69369555e-01 -1.33776331e+00 -4.49500442e-01 -2.21186921e-01 1.78687841e-01 5.98382950e-01 9.82246757e-01 5.90786189e-02 1.57989413e-01 6.63898945e-01 -5.78201532e-01 -1.34179986e+00 -6.87877715e-01 -6.17547512e-01 5.12625992e-01 8.37186992e-01 -4.58353609e-01 -1.06852615e+00 -1.48012996e-01]
[7.753108978271484, 3.826568841934204]
68662ba5-ab87-4c07-9f23-df7dbf90086b
considering-nested-tree-structure-in-sentence
null
null
https://aclanthology.org/2021.emnlp-main.330
https://aclanthology.org/2021.emnlp-main.330.pdf
Considering Nested Tree Structure in Sentence Extractive Summarization with Pre-trained Transformer
Sentence extractive summarization shortens a document by selecting sentences for a summary while preserving its important contents. However, constructing a coherent and informative summary is difficult using a pre-trained BERT-based encoder since it is not explicitly trained for representing the information of sentences in a document. We propose a nested tree-based extractive summarization model on RoBERTa (NeRoBERTa), where nested tree structures consist of syntactic and discourse trees in a given document. Experimental results on the CNN/DailyMail dataset showed that NeRoBERTa outperforms baseline models in ROUGE. Human evaluation results also showed that NeRoBERTa achieves significantly better scores than the baselines in terms of coherence and yields comparable scores to the state-of-the-art models.
['Manabu Okumura', 'Hidetaka Kamigaito', 'Naoki Kobayashi', 'Jingun Kwon']
null
null
null
null
emnlp-2021-11
['extractive-document-summarization']
['natural-language-processing']
[ 1.87396929e-01 5.75554073e-01 -2.13448852e-01 -4.29009467e-01 -1.12663364e+00 -5.74605048e-01 5.75330913e-01 3.59246105e-01 -3.50051850e-01 9.12335455e-01 1.48796880e+00 2.60743778e-02 1.60323814e-01 -6.32601619e-01 -6.64602935e-01 -3.16932321e-01 -2.79159267e-02 2.83260643e-01 3.24207582e-02 -3.43439728e-01 5.59628367e-01 8.12594146e-02 -1.03874946e+00 8.79637182e-01 1.02432525e+00 4.45416749e-01 2.24217996e-01 1.18527138e+00 -2.25113139e-01 1.03517604e+00 -1.32000756e+00 -6.10346735e-01 -1.51021779e-01 -8.60052645e-01 -1.21316648e+00 4.48142402e-02 8.84489596e-01 -6.63218975e-01 -4.80374515e-01 7.07304835e-01 5.78893304e-01 1.41403720e-01 6.33609354e-01 -4.70494956e-01 -6.73273385e-01 1.30258250e+00 -4.64404911e-01 3.81316781e-01 4.88708407e-01 -2.84169227e-01 1.55726290e+00 -7.21874297e-01 9.61946547e-01 1.31148732e+00 5.70245445e-01 5.89921772e-01 -8.98598790e-01 -2.17036799e-01 1.15468994e-01 1.84605084e-02 -7.07457781e-01 -5.85160911e-01 5.63996255e-01 7.85343647e-02 1.61950982e+00 6.36589706e-01 6.54509842e-01 1.11674666e+00 7.63821363e-01 1.23890829e+00 1.66511640e-01 -2.18573928e-01 9.49270055e-02 -5.19265294e-01 7.77488172e-01 6.43646061e-01 5.21142960e-01 -7.47897029e-01 -7.78566182e-01 5.21074459e-02 7.54167959e-02 -3.81887615e-01 -2.21598729e-01 4.01470810e-01 -1.30627263e+00 8.05816531e-01 3.72791529e-01 3.83501291e-01 -6.67628288e-01 1.81043938e-01 9.87953484e-01 1.82513803e-01 7.60757029e-01 9.89420533e-01 -2.20225960e-01 -1.42031938e-01 -1.42455673e+00 5.62574148e-01 9.76440310e-01 1.13517213e+00 2.80363023e-01 3.63685280e-01 -9.48182642e-01 6.36410594e-01 -3.45234752e-01 1.99975714e-01 5.66564918e-01 -1.15660381e+00 9.82949197e-01 6.17391288e-01 -7.51168951e-02 -8.30980599e-01 -2.93408483e-01 -5.60648501e-01 -1.19697273e+00 -3.88243586e-01 -3.49569649e-01 -4.37328100e-01 -8.11379552e-01 1.29118156e+00 -3.63832057e-01 -4.06269282e-01 4.90270287e-01 4.58850026e-01 1.58659697e+00 1.19835281e+00 -3.23293686e-01 -6.07204020e-01 1.13963771e+00 -1.53936696e+00 -1.14103591e+00 -3.94951224e-01 5.63310385e-01 -5.47763824e-01 7.91781425e-01 3.21134895e-01 -1.52837121e+00 -4.94625092e-01 -1.40409195e+00 -6.05051041e-01 1.66236237e-01 3.47706199e-01 5.58719516e-01 1.52565435e-01 -1.27839327e+00 7.23022521e-01 -6.80608630e-01 -4.18038219e-01 5.61085761e-01 -1.19054504e-02 -5.45271397e-01 1.96331948e-01 -8.80139232e-01 9.55906034e-01 7.52261996e-01 -7.03777596e-02 -8.53556931e-01 -3.94300669e-01 -1.07071304e+00 6.36104703e-01 2.85410434e-01 -1.10040808e+00 1.74287748e+00 -4.94118422e-01 -1.43317974e+00 4.37747121e-01 -5.48971593e-01 -1.06840265e+00 1.48626834e-01 -7.90662885e-01 3.06283068e-02 5.02546608e-01 3.72084618e-01 7.24178731e-01 5.68328440e-01 -9.25886095e-01 -6.16815746e-01 -5.25746085e-02 1.04063503e-01 3.97363156e-01 -2.56829113e-01 1.97897732e-01 -2.50211835e-01 -7.39448726e-01 -1.97326913e-01 -4.32967931e-01 -2.03286037e-01 -1.05874205e+00 -1.10958004e+00 -5.75868309e-01 8.05077314e-01 -1.08984303e+00 1.73986912e+00 -1.54127002e+00 3.38155180e-01 -6.34103835e-01 3.68833333e-01 3.61118466e-01 -3.16512525e-01 1.02685654e+00 9.90556255e-02 4.06345904e-01 -3.76816392e-01 -6.08341217e-01 -1.02747150e-01 -5.09728640e-02 -6.37329698e-01 -6.43276349e-02 1.77512169e-01 1.12784171e+00 -1.02040565e+00 -7.80055463e-01 -1.55573040e-01 -6.37619849e-03 -5.55044293e-01 3.19505215e-01 -4.19222683e-01 1.66445877e-02 -4.51721281e-01 2.87133515e-01 3.18259716e-01 -6.30221292e-02 1.82224780e-01 -8.34811106e-02 -1.62701353e-01 9.70575035e-01 -2.88179696e-01 1.99014437e+00 -3.26922387e-01 1.11741328e+00 -3.53784531e-01 -7.92042673e-01 8.82764161e-01 4.96691018e-01 1.48270622e-01 -4.48728681e-01 1.05804540e-01 5.35501502e-02 -8.42348784e-02 -4.54540998e-01 1.52626204e+00 2.25241452e-01 -4.71640050e-01 5.31253874e-01 2.88254291e-01 -5.69638252e-01 8.41838479e-01 8.14405143e-01 1.50625122e+00 -9.72523987e-02 6.74123764e-01 -2.67882407e-01 3.42549562e-01 2.06529666e-02 5.80911934e-01 9.48649228e-01 1.06640533e-01 1.07319427e+00 7.65488863e-01 -2.69732088e-01 -1.11647010e+00 -8.11167061e-01 3.45760196e-01 8.04605961e-01 -2.59311438e-01 -1.21388674e+00 -1.09339154e+00 -9.98883903e-01 -4.49660420e-01 1.38286674e+00 -4.95478928e-01 -3.65239847e-03 -1.03236961e+00 -3.91196549e-01 6.48899138e-01 5.94264567e-01 7.95933485e-01 -1.34177446e+00 -7.55148828e-01 2.58204222e-01 -7.86564350e-01 -1.03067255e+00 -8.88874173e-01 -5.37663437e-02 -1.09080887e+00 -8.07240069e-01 -6.12086654e-01 -7.96564162e-01 4.15481120e-01 3.71164382e-01 1.39286840e+00 -1.86819918e-02 2.66435832e-01 -2.87077427e-02 -5.29944003e-01 -6.13673031e-01 -7.03528345e-01 7.93841422e-01 -4.02886152e-01 -5.39901853e-01 1.11917347e-01 -3.93243581e-01 -5.13273954e-01 -7.25165308e-01 -8.73571217e-01 3.64662081e-01 6.45219743e-01 8.00873220e-01 2.06827119e-01 1.13556832e-02 1.07335794e+00 -9.66755450e-01 1.28886747e+00 2.05920590e-03 2.47031614e-01 2.95192629e-01 -2.40536079e-01 6.64625168e-02 8.24034095e-01 1.95702195e-01 -1.38032508e+00 -3.77438545e-01 -2.02591166e-01 3.24102372e-01 1.66402712e-01 7.93458819e-01 5.06023057e-02 1.15731299e+00 7.01180995e-01 3.41201156e-01 -4.01983321e-01 -3.66048008e-01 4.74744618e-01 8.41229320e-01 9.15930986e-01 -8.95525292e-02 3.15674514e-01 2.49330446e-01 -4.02650267e-01 -1.17530465e+00 -1.37748754e+00 -4.04610634e-01 -7.26543427e-01 -2.00107936e-02 8.02573085e-01 -9.26610947e-01 -2.60154396e-01 2.43286997e-01 -1.84971440e+00 9.36599448e-02 -6.45205319e-01 2.05828860e-01 -5.71148336e-01 6.75114810e-01 -7.57875860e-01 -3.61103594e-01 -1.39935088e+00 -7.68737197e-01 1.18672216e+00 4.29917514e-01 -9.71449196e-01 -7.58718491e-01 1.16863966e-01 5.01950324e-01 9.63969380e-02 3.70227993e-01 8.58974516e-01 -7.67759204e-01 -4.05192494e-01 -3.99670213e-01 2.38757972e-02 5.72595894e-01 2.25251809e-01 1.77887112e-01 -6.44102693e-01 -3.45906049e-01 1.50503600e-02 -4.13918406e-01 1.56266820e+00 6.03083789e-01 9.24695492e-01 -9.66170549e-01 -5.77919744e-02 1.18007138e-01 8.70368779e-01 -1.55238686e-02 8.16223025e-01 2.55269319e-01 5.34627318e-01 5.05948722e-01 4.95094895e-01 4.06854600e-01 4.86186832e-01 1.14793703e-01 2.02171102e-01 1.79261088e-01 -5.39807916e-01 -4.12358731e-01 6.78293288e-01 1.46820104e+00 1.08242020e-01 -8.03312659e-01 -4.15080398e-01 9.61648405e-01 -2.08098006e+00 -1.40898979e+00 -3.07246923e-01 1.59604251e+00 1.30010688e+00 2.48896986e-01 -1.92489699e-01 -1.19495563e-01 5.82096338e-01 9.03079152e-01 -3.45517009e-01 -9.85668421e-01 -3.70989501e-01 2.04098970e-01 7.75190517e-02 5.06915748e-01 -1.04270518e+00 1.10194767e+00 6.77251387e+00 7.14505792e-01 -6.91915154e-01 -1.57046348e-01 4.33615655e-01 -3.78773481e-01 -4.14193928e-01 -1.95167046e-02 -7.57712722e-01 1.55379921e-01 1.08042109e+00 -6.56139076e-01 -1.76596388e-01 6.98745847e-01 3.72553945e-01 -2.89439857e-01 -1.25824380e+00 3.16157281e-01 4.57624644e-01 -1.99753010e+00 5.25848567e-01 -3.37855279e-01 1.08492827e+00 3.93055119e-02 -4.69183236e-01 4.15024936e-01 4.33002919e-01 -9.09530938e-01 7.56343067e-01 5.18451452e-01 4.21644151e-01 -8.46803963e-01 9.84645307e-01 5.15097857e-01 -6.56499445e-01 2.82409877e-01 -5.18267035e-01 -1.04717053e-01 3.02578032e-01 4.92752910e-01 -1.09986508e+00 9.55933034e-01 4.77116972e-01 1.06283343e+00 -7.43629575e-01 7.62163579e-01 -5.82799733e-01 9.03904498e-01 1.58362746e-01 -3.32655519e-01 4.59254265e-01 7.70868501e-03 9.97902811e-01 1.84364974e+00 2.65893131e-01 4.49302047e-02 -1.04518980e-01 6.06755793e-01 -6.78606331e-01 1.12195173e-02 -5.69944441e-01 -1.32156119e-01 4.20284510e-01 1.21675551e+00 -5.73803365e-01 -7.58591712e-01 -9.47220400e-02 1.05500042e+00 3.43156487e-01 2.13277221e-01 -4.35943246e-01 -8.44611645e-01 1.15146585e-01 -2.01185331e-01 5.28157592e-01 -1.47089183e-01 -5.80434144e-01 -1.36162186e+00 1.85832858e-01 -9.28682983e-01 2.73969620e-01 -8.88469160e-01 -7.85721719e-01 8.60690117e-01 4.26754765e-02 -9.49674428e-01 -5.68229616e-01 2.72288442e-01 -1.14208531e+00 6.11605048e-01 -1.19304156e+00 -1.19315016e+00 -1.43946692e-01 -1.74321339e-01 1.40583551e+00 -2.65843779e-01 9.26464736e-01 -5.15229464e-01 -6.56641006e-01 2.27736503e-01 1.50989443e-01 1.59758002e-01 6.30397916e-01 -1.64871871e+00 1.08884478e+00 1.35693157e+00 -9.48965829e-03 8.25100899e-01 9.63773072e-01 -8.64555836e-01 -1.10528624e+00 -1.14268792e+00 1.53577054e+00 -3.15117836e-01 2.09983647e-01 -2.05445319e-01 -8.10955882e-01 8.75819445e-01 1.44363427e+00 -9.66979802e-01 5.89217007e-01 1.85020879e-01 -1.64677531e-01 -4.88685183e-02 -5.71581483e-01 7.71694899e-01 8.44106197e-01 -2.79032618e-01 -1.41020823e+00 3.54267269e-01 1.29542947e+00 -4.18768734e-01 -5.88539600e-01 2.85697818e-01 3.19797277e-01 -9.98247445e-01 5.47693253e-01 -7.74396300e-01 1.30360746e+00 7.15109333e-02 1.56363577e-01 -1.67093050e+00 -2.98517436e-01 -8.41986418e-01 -4.14244831e-01 1.43742907e+00 3.44079167e-01 -8.20382982e-02 6.36071682e-01 1.57816991e-01 -8.15244198e-01 -6.08614624e-01 -7.49269903e-01 -5.49931824e-01 2.09648922e-01 4.60921898e-02 4.45765495e-01 4.78538424e-01 3.12376231e-01 1.28602672e+00 -2.88300395e-01 -3.02003205e-01 3.03342521e-01 4.10387099e-01 9.60048318e-01 -8.99544179e-01 1.95098206e-01 -6.43527150e-01 1.59568131e-01 -1.14564717e+00 4.74209875e-01 -1.02837932e+00 2.66256601e-01 -2.53824520e+00 5.85488141e-01 7.19768763e-01 2.67554879e-01 3.42656732e-01 -3.31418961e-01 -2.82178521e-01 3.29601675e-01 1.09152302e-01 -1.06547964e+00 9.70321476e-01 1.34666586e+00 -5.34035683e-01 -1.85386389e-01 -9.66460258e-02 -1.22076011e+00 6.61241710e-01 8.07285547e-01 -4.07547027e-01 -3.46340567e-01 -5.21198690e-01 7.99836814e-02 4.44388509e-01 -1.52888834e-01 -8.33414555e-01 4.02926087e-01 1.94618121e-01 1.37393281e-01 -1.47903597e+00 9.60320234e-02 1.25185415e-01 -4.52011198e-01 3.57313991e-01 -9.37635124e-01 1.41187340e-01 8.96033049e-02 3.08452815e-01 -4.66196120e-01 -6.41038954e-01 3.92023295e-01 -3.01853657e-01 -3.23229849e-01 -1.63567439e-01 -6.88994348e-01 2.43840471e-01 4.21389550e-01 -1.85503177e-02 -7.67129600e-01 -8.03950906e-01 -1.58410773e-01 3.80649418e-01 9.19570848e-02 4.81485933e-01 8.67125273e-01 -9.76649702e-01 -1.43078411e+00 -4.11312759e-01 -1.28797188e-01 4.59020674e-01 2.38461673e-01 2.73784786e-01 -7.58991122e-01 8.89984310e-01 -1.01912513e-01 -1.58907056e-01 -1.51255333e+00 1.45475954e-01 -1.24667950e-01 -8.27536285e-01 -9.89361048e-01 5.59957147e-01 1.36047617e-01 -2.00098976e-01 2.13167667e-01 -7.43298590e-01 -4.41253871e-01 3.32641512e-01 8.00752223e-01 5.08812129e-01 2.51659304e-01 -3.49521339e-01 -6.52755573e-02 2.20789686e-02 -7.00024605e-01 -1.03777550e-01 1.62601793e+00 -1.94586545e-01 -4.29433018e-01 2.92655289e-01 1.14237738e+00 1.99735895e-01 -9.72873747e-01 -5.70425428e-02 2.15934828e-01 4.18671705e-02 -4.87892479e-02 -8.25082839e-01 -5.17685354e-01 7.72118747e-01 -6.77540064e-01 3.62930745e-01 9.42680359e-01 -9.29755643e-02 1.23058152e+00 1.03517663e+00 -1.86747238e-01 -1.26392984e+00 4.20232296e-01 1.00940204e+00 1.63276494e+00 -8.95476699e-01 5.24413049e-01 -1.48239493e-01 -9.81685102e-01 1.37874961e+00 4.59545493e-01 -3.58440846e-01 -1.06833346e-01 -5.73124830e-03 -2.74668097e-01 -2.42571443e-01 -1.20158017e+00 7.80344903e-02 3.37292463e-01 3.05673927e-01 7.32840180e-01 5.38067520e-03 -6.84476852e-01 8.85366976e-01 -9.01152670e-01 -1.84006900e-01 1.23618364e+00 8.51776123e-01 -7.11893260e-01 -6.77089572e-01 6.39856607e-02 6.24736369e-01 -7.04308271e-01 -3.00001085e-01 -8.08062494e-01 6.64521039e-01 -6.34248972e-01 1.21888256e+00 1.03210613e-01 -2.49800935e-01 5.28346002e-01 3.39300074e-02 3.73932451e-01 -1.13820207e+00 -1.11169755e+00 1.31036356e-01 8.30285251e-01 -2.42916450e-01 -3.51541877e-01 -7.78430521e-01 -1.48463321e+00 -4.34570223e-01 -1.87368661e-01 3.82926941e-01 3.65207970e-01 7.52018809e-01 4.89139408e-01 1.08830464e+00 5.82958102e-01 -8.17590475e-01 -6.30436957e-01 -1.48073018e+00 -2.02487543e-01 6.46133646e-02 5.25779843e-01 4.66642588e-01 2.71940548e-02 1.70586005e-01]
[12.525839805603027, 9.494762420654297]
985ffd40-4b1c-4756-8dfd-da54c0778b86
transferring-a-semantic-representation-for
1706.03725
null
http://arxiv.org/abs/1706.03725v1
http://arxiv.org/pdf/1706.03725v1.pdf
Transferring a Semantic Representation for Person Re-Identification and Search
Learning semantic attributes for person re-identification and description-based person search has gained increasing interest due to attributes' great potential as a pose and view-invariant representation. However, existing attribute-centric approaches have thus far underperformed state-of-the-art conventional approaches. This is due to their non-scalable need for extensive domain (camera) specific annotation. In this paper we present a new semantic attribute learning approach for person re-identification and search. Our model is trained on existing fashion photography datasets -- either weakly or strongly labelled. It can then be transferred and adapted to provide a powerful semantic description of surveillance person detections, without requiring any surveillance domain supervision. The resulting representation is useful for both unsupervised and supervised person re-identification, achieving state-of-the-art and near state-of-the-art performance respectively. Furthermore, as a semantic representation it allows description-based person search to be integrated within the same framework.
['Zhiyuan Shi', 'Timothy M. Hospedales', 'Tao Xiang']
2017-06-12
transferring-a-semantic-representation-for-1
http://openaccess.thecvf.com/content_cvpr_2015/html/Shi_Transferring_a_Semantic_2015_CVPR_paper.html
http://openaccess.thecvf.com/content_cvpr_2015/papers/Shi_Transferring_a_Semantic_2015_CVPR_paper.pdf
cvpr-2015-6
['person-search']
['computer-vision']
[ 2.29692429e-01 -1.72602192e-01 -1.79012462e-01 -7.09833086e-01 -8.42189789e-01 -6.11654639e-01 1.11546087e+00 4.40104127e-01 -7.18363285e-01 7.06629932e-01 4.35909003e-01 4.49424893e-01 -2.12632284e-01 -5.24125338e-01 -2.65888900e-01 -4.77208763e-01 3.58338118e-01 1.18023264e+00 1.98235348e-01 -1.03270583e-01 1.18758604e-02 5.34985483e-01 -2.08926415e+00 1.99468762e-01 4.68074471e-01 7.94814408e-01 -1.10642865e-01 5.87728202e-01 2.01569498e-01 2.99759626e-01 -3.87015313e-01 -8.44986498e-01 1.66105554e-01 -1.81394100e-01 -1.12725568e+00 3.56849998e-01 7.96458602e-01 -2.35138774e-01 -3.81581515e-01 1.01183212e+00 7.17945755e-01 2.17819139e-01 6.98524773e-01 -1.28218246e+00 -5.63012481e-01 -9.85546783e-02 -4.01956402e-02 5.51076271e-02 9.06181931e-01 -7.88628086e-02 6.76355302e-01 -6.20425642e-01 6.78373337e-01 1.35295129e+00 9.09427285e-01 1.00098562e+00 -1.48827732e+00 -4.30362910e-01 1.57483995e-01 5.05099893e-01 -1.68204582e+00 -6.54949963e-01 6.07438564e-01 -3.92076313e-01 1.02314734e+00 4.60179359e-01 7.87766159e-01 1.37130880e+00 -7.89045930e-01 7.71862626e-01 1.12990415e+00 -5.01460195e-01 1.26603730e-02 6.11253738e-01 1.41575769e-01 6.95622087e-01 3.96783054e-01 1.42755851e-01 -5.73343456e-01 -1.97092891e-01 5.81169069e-01 2.41631970e-01 2.18584910e-02 -7.73474991e-01 -1.16610515e+00 6.42408848e-01 4.79141712e-01 2.19878525e-01 -2.67890602e-01 1.91447899e-01 6.27860188e-01 2.15995997e-01 5.28911710e-01 3.11325163e-01 -2.44868591e-01 -9.54148918e-02 -9.84167397e-01 6.43390238e-01 7.39246845e-01 1.05521917e+00 8.00569832e-01 -4.02217060e-01 -2.44351387e-01 9.42392528e-01 1.55350298e-01 5.93044400e-01 3.49552184e-01 -7.19101787e-01 8.62831101e-02 9.34165120e-01 3.23965997e-01 -6.13337517e-01 -4.72576439e-01 -5.26535332e-01 -6.74924552e-01 5.86755481e-03 4.04444307e-01 4.68979537e-01 -1.07274640e+00 1.74771416e+00 4.14549381e-01 2.60243565e-01 1.96342513e-01 8.42651188e-01 8.73390555e-01 1.76929440e-02 5.88904262e-01 2.19183743e-01 1.92349350e+00 -8.30709696e-01 -4.10619497e-01 -2.69665003e-01 5.12870848e-01 -5.78903377e-01 6.61551833e-01 -7.68723488e-02 -8.05300832e-01 -5.89580595e-01 -6.58321679e-01 5.91458753e-02 -8.46721292e-01 1.21801592e-01 4.87581372e-01 1.04678273e+00 -1.16316366e+00 2.33050659e-01 -4.43032265e-01 -1.25427353e+00 5.71294010e-01 5.75247467e-01 -9.83325362e-01 -2.78543234e-01 -9.77654099e-01 9.22775507e-01 4.64811265e-01 -4.30714250e-01 -8.31572235e-01 -7.02517688e-01 -8.34049940e-01 -2.18172103e-01 3.46343040e-01 -1.25258005e+00 1.14162695e+00 -7.01236606e-01 -9.96501267e-01 1.71253729e+00 -3.75687122e-01 -4.49545145e-01 6.06840372e-01 -1.14326447e-01 -6.91591084e-01 2.50926435e-01 5.47502398e-01 6.67137980e-01 8.05879891e-01 -1.50506008e+00 -8.98228884e-01 -8.31519723e-01 1.51374742e-01 3.34337980e-01 -4.92199093e-01 3.88657451e-01 -8.47893655e-01 -7.58960128e-01 -2.08725974e-01 -1.14539754e+00 -5.54693714e-02 1.30154900e-02 -1.96757823e-01 -4.87121493e-01 6.63796008e-01 -7.15347826e-01 8.32808137e-01 -1.84937167e+00 1.81663394e-01 4.96762432e-02 1.10031059e-02 4.93355811e-01 6.40162528e-02 4.78848666e-01 1.05531188e-02 -9.98911932e-02 -1.79593153e-02 -8.70855689e-01 1.06194846e-01 2.95413490e-02 2.48656213e-01 6.17671609e-01 -1.24785960e-01 9.47868705e-01 -9.48124647e-01 -5.22347391e-01 6.31987989e-01 5.71232975e-01 -1.86183065e-01 3.77849400e-01 1.53325215e-01 5.22136569e-01 -3.87162894e-01 9.29511130e-01 3.73551399e-01 -3.00215408e-02 3.36498059e-02 -3.92974854e-01 9.39562693e-02 -3.04456711e-01 -1.19234931e+00 1.87415385e+00 -2.61205286e-01 3.25819790e-01 -7.63998255e-02 -1.05836022e+00 8.19098592e-01 3.18821937e-01 5.20319343e-01 -6.97479010e-01 -1.09568454e-01 2.12186471e-01 -8.90292585e-01 -2.25225270e-01 5.42031229e-01 -2.33440530e-02 -4.09376025e-01 3.01006168e-01 1.88980266e-01 4.57982063e-01 1.57820985e-01 7.08519146e-02 6.58530891e-01 2.00781256e-01 6.14290416e-01 -3.48007292e-01 1.13650346e+00 2.47261241e-01 2.27030560e-01 7.68848360e-01 -3.71973753e-01 5.36140800e-01 -4.09664840e-01 -7.72961557e-01 -1.17870080e+00 -1.08182204e+00 -9.12344456e-02 1.37714708e+00 3.55933040e-01 -4.31073785e-01 -9.45609868e-01 -7.97049403e-01 1.57045007e-01 4.82951224e-01 -5.95035553e-01 -1.95616424e-01 -5.29334068e-01 -6.38327003e-01 7.42327511e-01 6.92977965e-01 8.87805402e-01 -7.64795065e-01 -4.16064441e-01 1.55390441e-01 -2.93523490e-01 -1.31663704e+00 -3.27576160e-01 -2.74778873e-01 -4.15549576e-01 -1.23947036e+00 -1.14351034e+00 -9.04314041e-01 7.07040191e-01 3.04904908e-01 1.13291121e+00 8.40204507e-02 -4.55850720e-01 1.28341985e+00 -3.86150450e-01 -4.12871361e-01 -4.00733739e-01 1.76762134e-01 4.74597484e-01 1.81252226e-01 8.92997384e-01 -1.66914478e-01 -7.81890094e-01 5.23673356e-01 -5.32727063e-01 -4.55933779e-01 2.55239159e-01 6.64604485e-01 4.87047195e-01 -1.75869435e-01 5.18815994e-01 -7.48890817e-01 4.75057095e-01 2.44698450e-02 -3.25811356e-01 6.24146998e-01 -7.82374322e-01 -1.25832632e-01 2.28207305e-01 -2.58203834e-01 -1.17171478e+00 3.59046727e-01 4.50443253e-02 -1.25004545e-01 -8.93490672e-01 -1.87653527e-01 -3.31941038e-01 -3.99988025e-01 4.95378315e-01 3.49765003e-01 -1.25705555e-01 -6.87763393e-01 4.27334815e-01 8.70972395e-01 1.05648649e+00 -3.40345144e-01 9.12323773e-01 7.62544215e-01 -6.96830126e-03 -6.78830206e-01 -7.88838029e-01 -1.23054719e+00 -1.03181839e+00 -2.88070142e-01 1.07314670e+00 -1.35938692e+00 -8.43909383e-01 6.66159630e-01 -9.86753106e-01 1.59091681e-01 -2.72208065e-01 1.83750600e-01 -5.88988602e-01 4.66481954e-01 -2.01865137e-01 -8.72507274e-01 -3.31513077e-01 -8.10205519e-01 1.37312889e+00 3.34060133e-01 -4.17822331e-01 -1.04222775e+00 -1.11043714e-02 7.74203122e-01 2.43583992e-01 1.76594540e-01 4.41451877e-01 -9.61772144e-01 -2.97690928e-01 -8.38222086e-01 -3.95403206e-01 2.86630895e-02 1.14928067e-01 -9.95995700e-01 -1.27548981e+00 -6.98911369e-01 -5.78683078e-01 -3.08479637e-01 9.47968364e-01 -3.49441469e-02 7.26613462e-01 -8.95599574e-02 -8.24044943e-01 6.16406560e-01 1.43430924e+00 -3.14541131e-01 3.35449636e-01 8.50105464e-01 6.78021848e-01 7.65624046e-01 5.29545784e-01 4.29211646e-01 7.86247671e-01 1.28851008e+00 2.42685556e-01 -2.59313971e-01 -5.29105961e-01 -3.54432404e-01 -1.75929442e-02 -1.71799093e-01 -4.89638746e-01 -1.92623749e-01 -9.68942642e-01 6.87950730e-01 -2.19198179e+00 -1.21118164e+00 1.29343048e-01 2.41393971e+00 3.27152371e-01 -4.47329551e-01 6.60134077e-01 8.44289735e-02 9.75932658e-01 -8.54764134e-02 -3.64950866e-01 3.16234156e-02 -1.71125516e-01 -2.94739366e-01 6.89153790e-01 2.30449945e-01 -1.61876249e+00 1.08000243e+00 6.04311132e+00 7.03916132e-01 -2.78088927e-01 3.22665125e-01 2.26434678e-01 2.80448765e-01 9.87156630e-02 -2.29422197e-01 -1.13965094e+00 2.60981739e-01 6.26986086e-01 -7.96843320e-02 3.72677624e-01 9.11875904e-01 -2.58306175e-01 8.07676464e-03 -1.37905681e+00 1.71270251e+00 5.64615846e-01 -1.19789290e+00 2.55252063e-01 8.27804729e-02 5.40696263e-01 -4.89951581e-01 -8.47655460e-02 1.48358941e-01 2.06141159e-01 -9.70570028e-01 6.50289774e-01 5.75629711e-01 1.18097055e+00 -7.44911909e-01 9.97047067e-01 1.83548212e-01 -1.59704828e+00 -2.87062317e-01 -2.66751021e-01 1.86281174e-01 3.89325976e-01 -1.65362760e-01 -7.30328202e-01 7.75264144e-01 9.79309738e-01 6.93763554e-01 -8.71454418e-01 9.64318037e-01 7.75613114e-02 5.45352139e-02 -2.27193132e-01 2.60434389e-01 8.52354616e-02 1.40231103e-01 5.42969823e-01 1.51693916e+00 9.07048285e-02 -6.51257411e-02 6.10521615e-01 4.12507683e-01 4.13184352e-02 -1.21717773e-01 -5.22198975e-01 3.25918138e-01 5.34030795e-01 9.90558326e-01 -7.14002073e-01 -5.50848246e-01 -4.15237814e-01 1.58948898e+00 3.16677123e-01 2.40505323e-01 -4.79732186e-01 9.09767374e-02 9.30954218e-01 2.87046760e-01 1.58034459e-01 1.58776462e-01 1.79914922e-01 -1.05186784e+00 -1.15956977e-01 -7.20098197e-01 9.57806408e-01 -4.23020571e-01 -1.58239281e+00 5.85666418e-01 4.00356978e-01 -9.47684586e-01 -6.85226023e-01 -4.82161880e-01 1.80895329e-02 8.70162129e-01 -1.52172887e+00 -2.08383179e+00 -7.30022252e-01 9.58589375e-01 5.80072582e-01 -7.98477590e-01 1.46087658e+00 4.61542755e-01 -2.01659456e-01 7.62250602e-01 -1.05234064e-01 1.65320873e-01 9.39020693e-01 -1.24706864e+00 3.76515239e-01 7.28662789e-01 7.66464397e-02 5.18845201e-01 7.09806800e-01 -6.98946297e-01 -1.03609669e+00 -1.28024352e+00 1.10733438e+00 -9.72184658e-01 1.86042711e-01 -4.48909461e-01 -5.29035866e-01 5.76834977e-01 -1.44090369e-01 2.51940060e-02 8.72135282e-01 2.34964088e-01 -6.16641462e-01 -2.86376774e-01 -1.37327147e+00 2.55215466e-01 1.54725909e+00 -9.17340815e-01 -5.72012842e-01 4.40803051e-01 2.48446926e-01 6.88854977e-02 -7.15451896e-01 2.60031611e-01 5.71459353e-01 -9.57536757e-01 1.73984718e+00 -6.02455020e-01 -5.03158331e-01 -2.90238947e-01 -2.57102072e-01 -1.02738714e+00 -6.84417486e-01 -3.11745793e-01 4.76834550e-02 1.58831131e+00 -1.50100276e-01 -5.64176559e-01 9.45189357e-01 7.78784633e-01 2.69461870e-01 -8.72032046e-02 -1.09346080e+00 -1.10394633e+00 -5.02932072e-01 -1.75693080e-01 7.69443870e-01 7.19047487e-01 -3.39272439e-01 3.66863191e-01 -5.88543475e-01 3.74689966e-01 1.19008374e+00 -5.40040247e-02 8.53487849e-01 -1.73480690e+00 1.25748292e-01 -3.45841169e-01 -1.05923712e+00 -6.70057535e-01 4.26168203e-01 -1.01957488e+00 -2.91481465e-01 -1.67411566e+00 7.83592939e-01 -5.25217772e-01 -2.09209129e-01 4.38183904e-01 -1.60578907e-01 7.24046588e-01 2.99220145e-01 5.25062025e-01 -1.04487693e+00 4.05507118e-01 3.25618654e-01 -2.41983578e-01 1.29979059e-01 1.12369724e-01 -6.58681989e-01 6.46529973e-01 7.32493579e-01 -3.07566494e-01 -2.83859402e-01 -3.54526043e-01 -1.62765130e-01 -5.09596884e-01 1.09408259e+00 -1.31232989e+00 3.47984046e-01 3.33420664e-01 6.87883735e-01 -3.80856454e-01 7.20783412e-01 -1.06712806e+00 3.94295365e-01 4.47087765e-01 -2.24297136e-01 3.99795026e-02 -5.01292497e-02 1.02512109e+00 -1.14495791e-01 -2.58144319e-01 6.98805571e-01 -5.45850396e-01 -1.31871223e+00 4.65645552e-01 -9.55074728e-02 -1.38282537e-01 1.16005361e+00 -7.34023631e-01 -8.71045068e-02 -3.68942529e-01 -7.83311248e-01 -4.66490444e-03 8.97797108e-01 5.72727084e-01 4.30105627e-01 -1.33875108e+00 -8.30968142e-01 1.65867373e-01 8.03208411e-01 -5.10576248e-01 3.08097482e-01 1.55033767e-01 -1.31541997e-01 5.70350170e-01 -2.71864861e-01 -6.60823882e-01 -1.79195535e+00 7.63912082e-01 2.37347797e-01 -2.02235803e-01 -5.52538037e-01 6.90153778e-01 7.77408406e-02 -5.58732271e-01 2.68797755e-01 7.91824579e-01 -4.25521195e-01 1.01038076e-01 6.91318989e-01 6.09980881e-01 -1.03475206e-01 -1.22905648e+00 -8.88313353e-01 8.52398694e-01 -6.49303421e-02 -1.12773553e-02 1.17830575e+00 -4.83449548e-01 1.43975005e-01 -4.09009457e-02 1.01752543e+00 -3.76779854e-01 -9.56662714e-01 -4.02701676e-01 2.71821678e-01 -5.37013292e-01 -2.37574294e-01 -9.59507227e-01 -4.52853233e-01 4.67112273e-01 1.03600156e+00 -1.07617732e-02 9.05580342e-01 4.98919904e-01 5.31045914e-01 5.21622479e-01 7.18473911e-01 -1.33956647e+00 1.78971887e-02 1.07698746e-01 6.16459548e-01 -1.55408812e+00 3.16411376e-01 -5.99231064e-01 -5.82980216e-01 8.01407337e-01 4.24096167e-01 3.11186105e-01 3.66587996e-01 -2.59010911e-01 1.95257720e-02 -2.06684232e-01 -7.03252777e-02 -7.52336681e-01 5.23633718e-01 1.37187254e+00 9.31111202e-02 1.59851208e-01 1.77760899e-01 4.48608637e-01 1.04651719e-01 -1.43766016e-01 -1.32647470e-01 6.48376286e-01 -2.99415022e-01 -1.38170052e+00 -3.92682105e-01 2.24597797e-01 -2.24314287e-01 5.91886379e-02 -5.34934878e-01 8.20949614e-01 2.85657048e-01 9.82371569e-01 -1.24981478e-02 -1.87169120e-01 6.08560205e-01 1.70032680e-01 5.87576091e-01 -5.30867457e-01 -6.91690445e-01 -3.47381681e-01 4.56234753e-01 -6.21822298e-01 -9.52198982e-01 -8.81612837e-01 -6.72988653e-01 -3.35395455e-01 -6.60449564e-02 -4.85943966e-02 4.01212245e-01 8.39703679e-01 3.44195426e-01 -1.46906066e-03 1.65867537e-01 -7.81300843e-01 -5.14964640e-01 -6.93549454e-01 -3.80848199e-01 1.16862452e+00 8.04169402e-02 -7.92129755e-01 1.96462736e-01 3.66813779e-01]
[14.654324531555176, 0.9955406188964844]
9d3833db-c17c-472d-9f86-10aaf0671473
efficient-and-accurate-multi-scale
2102.12135
null
https://arxiv.org/abs/2102.12135v1
https://arxiv.org/pdf/2102.12135v1.pdf
Efficient and Accurate Multi-scale Topological Network for Single Image Dehazing
Single image dehazing is a challenging ill-posed problem that has drawn significant attention in the last few years. Recently, convolutional neural networks have achieved great success in image dehazing. However, it is still difficult for these increasingly complex models to recover accurate details from the hazy image. In this paper, we pay attention to the feature extraction and utilization of the input image itself. To achieve this, we propose a Multi-scale Topological Network (MSTN) to fully explore the features at different scales. Meanwhile, we design a Multi-scale Feature Fusion Module (MFFM) and an Adaptive Feature Selection Module (AFSM) to achieve the selection and fusion of features at different scales, so as to achieve progressive image dehazing. This topological network provides a large number of search paths that enable the network to extract abundant image features as well as strong fault tolerance and robustness. In addition, ASFM and MFFM can adaptively select important features and ignore interference information when fusing different scale representations. Extensive experiments are conducted to demonstrate the superiority of our method compared with state-of-the-art methods.
['Guixu Zhang', 'Aiwen Jiang', 'Faming Fang', 'Juncheng Li', 'Qiaosi Yi']
2021-02-24
null
null
null
null
['image-dehazing']
['computer-vision']
[ 2.22595453e-01 -5.89103103e-01 1.99539393e-01 -2.77248900e-02 -2.35305995e-01 4.14605699e-02 1.57003522e-01 -1.67320400e-01 -1.12657584e-01 4.42289650e-01 4.16449346e-02 1.56069323e-01 -5.65832436e-01 -9.97837961e-01 -2.87702173e-01 -9.81672585e-01 -1.13421589e-01 -4.63477522e-01 6.67791605e-01 -3.28143895e-01 5.91369033e-01 6.93638444e-01 -1.79936290e+00 6.54490650e-05 1.18857002e+00 1.22856069e+00 5.18685043e-01 4.25381690e-01 1.55191927e-03 7.54782915e-01 -8.16619217e-01 1.14468768e-01 2.12265998e-01 -2.19629884e-01 -4.47272182e-01 3.96079421e-01 2.46477097e-01 -3.49855632e-01 -7.34211624e-01 1.32849908e+00 5.81123292e-01 3.34520727e-01 1.94724932e-01 -1.22179258e+00 -9.18860078e-01 7.75407851e-02 -8.10585499e-01 5.80614269e-01 -4.98519577e-02 2.22897813e-01 4.25987571e-01 -1.01490688e+00 2.22762764e-01 1.21116626e+00 2.83428282e-01 1.03520021e-01 -6.92449272e-01 -8.70524704e-01 2.15113282e-01 7.02146471e-01 -1.80551207e+00 -4.21205550e-01 1.05060816e+00 -2.13744324e-02 5.80823064e-01 2.63427526e-01 5.93996286e-01 2.92091429e-01 4.69913423e-01 7.65891135e-01 9.92916226e-01 -3.06918591e-01 1.12571545e-01 -5.30689470e-02 -1.19928651e-01 8.63202274e-01 3.96159261e-01 -2.29138527e-02 -7.28384078e-01 3.65497358e-02 1.03316653e+00 6.04554355e-01 -7.00609028e-01 -5.12280837e-02 -1.25990009e+00 6.85440958e-01 8.87452006e-01 5.42953968e-01 -5.73548734e-01 1.10916406e-01 6.98008686e-02 3.68898511e-01 4.14946169e-01 4.37349290e-01 -8.16789418e-02 3.81700367e-01 -9.47718799e-01 -4.39374894e-02 1.90278411e-01 6.93773925e-01 1.17282891e+00 2.76324660e-01 -1.66259363e-01 8.59354258e-01 1.24249756e-02 3.16390187e-01 4.64398891e-01 -9.37353730e-01 2.46372193e-01 8.34950686e-01 -1.69222519e-01 -1.63860619e+00 -1.44567639e-01 -5.06531358e-01 -1.41532457e+00 4.16445106e-01 -3.38438660e-01 1.02853514e-01 -1.07905912e+00 1.31101680e+00 3.46251279e-01 4.73109841e-01 9.44827795e-02 9.53524411e-01 5.89431584e-01 1.06698918e+00 -3.70548666e-01 -3.40245336e-01 1.28981900e+00 -1.03528452e+00 -7.84590721e-01 -2.58805305e-01 8.64731744e-02 -8.59255493e-01 7.20425725e-01 4.61781204e-01 -9.96576607e-01 -7.66005456e-01 -1.40259790e+00 2.98364833e-02 -4.48438555e-01 -3.65512967e-02 5.30942202e-01 1.71827391e-01 -1.03839076e+00 5.88994503e-01 -6.15282595e-01 -6.99412674e-02 6.51487648e-01 5.26779056e-01 -3.08389723e-01 -5.74621260e-01 -1.26630878e+00 6.60133660e-01 3.87017399e-01 3.39965135e-01 -7.74901688e-01 -4.47175801e-01 -7.71377921e-01 1.60886765e-01 6.04820967e-01 -5.83780348e-01 5.99052250e-01 -7.75520802e-01 -1.15165949e+00 1.49376661e-01 -9.98584479e-02 -2.92189885e-02 4.66365628e-02 -9.92678255e-02 -6.67590022e-01 5.68650484e-01 9.99533236e-02 5.76285839e-01 1.28065968e+00 -1.11066675e+00 -9.33086157e-01 -3.34059507e-01 8.75984430e-02 3.56153965e-01 -7.96439111e-01 1.07844442e-01 -4.59040314e-01 -1.02957189e+00 3.68637919e-01 -4.38534439e-01 -3.67451757e-01 3.10097069e-01 -3.00948560e-01 -4.27554511e-02 1.12889707e+00 -6.29814506e-01 1.45589590e+00 -2.40863585e+00 4.09898192e-01 1.52875066e-01 5.67208469e-01 4.61879820e-01 -1.80658326e-01 3.85246277e-01 1.83764264e-01 1.28031820e-01 -4.15459216e-01 -5.41359633e-02 -4.78579968e-01 1.82218164e-01 -2.16518521e-01 5.22904515e-01 4.67572480e-01 6.10429168e-01 -5.64135551e-01 -7.05112815e-01 3.59039307e-01 6.69423878e-01 -4.50832337e-01 2.85064638e-01 1.91082075e-01 2.45632157e-01 -7.81309962e-01 9.55482602e-01 1.02849305e+00 -3.29306543e-01 -5.04538596e-01 -4.19542432e-01 -3.32120210e-01 -3.76576036e-01 -1.21954453e+00 1.57629681e+00 -3.38529259e-01 3.84225696e-01 1.58431739e-01 -8.18382680e-01 8.72352123e-01 6.01250902e-02 2.93353945e-01 -6.17055476e-01 1.91265911e-01 2.03221545e-01 -1.68636084e-01 -6.26975000e-01 5.72949708e-01 -3.89075391e-02 1.05911911e-01 1.62897259e-01 -1.24598563e-01 -4.42799143e-02 6.86913133e-02 1.36296690e-01 1.01447082e+00 -4.28642929e-01 2.22096488e-01 -2.80934721e-01 9.82912481e-01 -1.40361086e-01 8.60540748e-01 3.25012267e-01 -1.90859467e-01 6.53777540e-01 -1.53132416e-02 -5.99870324e-01 -6.61819994e-01 -6.86876476e-01 4.05542068e-02 6.29823804e-01 7.65051126e-01 -2.42123902e-01 -5.84697962e-01 -5.51390290e-01 -4.94575202e-02 2.64583230e-01 -5.91390371e-01 -7.47243285e-01 -6.37984335e-01 -6.82385266e-01 1.53165385e-01 1.21695518e-01 1.19330263e+00 -9.19636905e-01 -5.53981245e-01 2.23794013e-01 -1.77190989e-01 -7.73435056e-01 -6.11750245e-01 -7.66957402e-02 -6.59678340e-01 -1.03215694e+00 -7.30132520e-01 -1.03918183e+00 8.86026442e-01 1.14757359e+00 5.08009672e-01 7.08147585e-01 -4.31711674e-01 -1.11299559e-01 -5.61448872e-01 -2.26672545e-01 -4.93673645e-02 8.58988799e-03 5.17379306e-03 3.30538213e-01 5.04330508e-02 -8.97060692e-01 -9.38772678e-01 4.46379364e-01 -1.59386396e+00 7.17003495e-02 1.01023149e+00 9.24082637e-01 5.37539959e-01 9.29324269e-01 5.64223647e-01 -4.30800110e-01 4.31186438e-01 -4.89358276e-01 -5.29388368e-01 3.15975785e-01 -6.02756917e-01 -5.72080538e-02 7.86396265e-01 -5.11187375e-01 -9.41894710e-01 -3.51898998e-01 8.79163146e-02 -7.64646232e-01 6.20336197e-02 4.83165532e-01 -3.56648415e-01 -6.90526128e-01 2.72383153e-01 6.74914062e-01 1.37167066e-01 -5.46033919e-01 6.99617192e-02 7.03670800e-01 5.28005064e-01 -1.92748219e-01 1.24055970e+00 4.93961275e-01 1.06069505e-01 -7.27691174e-01 -7.63381839e-01 -2.34627992e-01 -3.87535155e-01 -1.41307935e-01 6.81771457e-01 -9.60082114e-01 -4.16485339e-01 7.40814030e-01 -9.16800141e-01 1.99427068e-01 -9.56893619e-03 1.76520586e-01 1.27237020e-02 5.93217015e-01 -5.46238065e-01 -5.59399247e-01 -4.79134560e-01 -1.16259348e+00 8.39302778e-01 6.70443654e-01 5.99405825e-01 -6.63156569e-01 -5.00451148e-01 6.54042587e-02 8.24758768e-01 3.48473340e-01 7.69130349e-01 -2.63225771e-02 -9.94547844e-01 -7.29732960e-02 -5.65049112e-01 4.27834213e-01 5.33246040e-01 -6.62169755e-02 -6.11801326e-01 -5.73248625e-01 3.13894689e-01 -6.14937805e-02 1.12406123e+00 1.98959813e-01 1.21365738e+00 -3.80844980e-01 -3.56868029e-01 8.72278988e-01 1.49451160e+00 1.95401639e-01 7.29875386e-01 5.76502502e-01 7.83999622e-01 4.88510579e-01 7.96401799e-01 4.18175995e-01 2.42036164e-01 3.25487584e-01 7.19238281e-01 -2.75690258e-01 -1.69800043e-01 9.04113799e-03 9.28738862e-02 8.63977611e-01 1.85797498e-01 -8.70212987e-02 -5.65773904e-01 5.10872602e-01 -1.75894916e+00 -6.67305052e-01 3.48779052e-01 1.95403028e+00 5.24771929e-01 6.87543973e-02 -4.60568994e-01 2.47473985e-01 1.07093298e+00 4.50056225e-01 -6.23982608e-01 2.14749485e-01 -2.65519202e-01 -6.32459372e-02 2.85548747e-01 9.62480679e-02 -1.16289365e+00 8.16852272e-01 5.62144279e+00 1.27052188e+00 -1.22539175e+00 -3.18365358e-02 6.23491108e-01 -5.77703342e-02 -1.61216900e-01 1.87128112e-02 -4.30646926e-01 6.89598143e-01 2.97500342e-01 -2.73935348e-01 6.77183747e-01 5.74222505e-01 -7.08397776e-02 -1.43184140e-01 -2.76899666e-01 1.04103386e+00 2.28577003e-01 -1.39372730e+00 2.60043234e-01 7.89547116e-02 9.66135800e-01 -3.42286021e-01 1.30957678e-01 -1.47947390e-02 4.36636619e-02 -8.44098806e-01 4.52078700e-01 5.66971123e-01 8.21528018e-01 -1.15217555e+00 7.87994504e-01 2.48232722e-01 -1.50006890e+00 -4.54199940e-01 -7.12251365e-01 1.34837076e-01 -1.35185150e-02 9.28463936e-01 -1.54195214e-03 9.78555679e-01 8.49849403e-01 8.65256906e-01 -7.33377755e-01 1.23402333e+00 -1.84759542e-01 1.62255377e-01 -1.82043537e-01 2.97495216e-01 2.63689041e-01 -3.67680900e-02 4.46473986e-01 7.63372719e-01 6.43476903e-01 3.20950538e-01 3.51741076e-01 5.72429597e-01 -4.22686618e-03 -4.48606722e-02 -3.42121303e-01 7.55575001e-02 7.46319413e-01 1.33322382e+00 -6.33723259e-01 -2.80205995e-01 -4.21402246e-01 1.09673369e+00 3.15209717e-01 2.86827773e-01 -5.80688596e-01 -9.79283810e-01 9.17615294e-01 -1.71017572e-01 3.67150426e-01 -2.31209278e-01 1.74017474e-01 -1.20863485e+00 -1.00728516e-02 -8.52652073e-01 3.93294752e-01 -7.66502380e-01 -1.20973754e+00 7.87881315e-01 -2.63007134e-01 -1.49788439e+00 3.92710149e-01 -2.34931588e-01 -6.60346270e-01 9.66946304e-01 -2.08160949e+00 -1.05065203e+00 -7.95492053e-01 8.63573194e-01 6.87076807e-01 -1.20647855e-01 3.13547015e-01 3.85693580e-01 -8.39206398e-01 2.85515368e-01 -1.72261223e-02 -1.40689269e-01 7.00635970e-01 -7.23529518e-01 1.57170057e-01 1.30379975e+00 -3.42535377e-01 6.31854832e-01 5.12975216e-01 -5.93061030e-01 -1.52104402e+00 -1.37242544e+00 2.07864612e-01 2.54066527e-01 3.73328149e-01 2.02492669e-01 -1.15674913e+00 1.51050299e-01 1.87865883e-01 3.89331609e-01 3.06651860e-01 -6.79030657e-01 -2.92841554e-01 -4.71536964e-01 -1.15553021e+00 5.10977864e-01 9.27507818e-01 -3.14292580e-01 -3.48851770e-01 1.58514678e-01 1.13056803e+00 -1.15349464e-01 -7.13723123e-01 6.11609936e-01 1.26555666e-01 -1.04536092e+00 1.00363874e+00 6.35237386e-03 3.06593150e-01 -8.80640566e-01 -2.33366147e-01 -1.28866208e+00 -7.25048125e-01 -6.83400512e-01 -1.89071283e-01 1.19675434e+00 -3.11633408e-01 -8.03819776e-01 4.32105660e-01 -3.71870995e-02 -2.28219509e-01 -9.68346775e-01 -9.48578238e-01 -6.52458847e-01 -4.47109759e-01 1.47600159e-01 9.38349187e-01 9.35646415e-01 -4.53555077e-01 -9.29327961e-03 -5.80794096e-01 6.48583889e-01 7.46458471e-01 3.25121075e-01 4.40990031e-01 -1.24236333e+00 1.58263907e-01 -4.17946190e-01 -6.49734855e-01 -8.59402120e-01 -1.35487750e-01 -4.44998205e-01 7.90410340e-02 -1.59571147e+00 2.63527989e-01 -3.54025513e-01 -7.28166223e-01 4.91607994e-01 -5.62037349e-01 3.82386684e-01 7.22132996e-02 6.08797014e-01 -6.30038440e-01 8.72612178e-01 1.52508914e+00 -1.13017932e-01 2.59843990e-02 -2.41126284e-01 -1.00860977e+00 5.15250027e-01 8.18670750e-01 -4.52012956e-01 -3.65198582e-01 -5.53071916e-01 -7.64664784e-02 1.02460280e-03 2.95608461e-01 -1.29800880e+00 5.47679245e-01 -3.42393547e-01 6.98416352e-01 -4.90766943e-01 3.37373048e-01 -8.77728999e-01 4.68699709e-02 4.91574347e-01 9.68512446e-02 1.50998652e-01 7.94457793e-02 7.33453035e-01 -6.49794519e-01 6.01311103e-02 7.63658166e-01 -2.23296851e-01 -1.02474856e+00 6.24297678e-01 -2.46816590e-01 -4.64744389e-01 1.04317832e+00 -3.58051121e-01 -5.39627492e-01 -1.52471364e-01 -2.14233920e-01 5.53600490e-01 7.00801134e-01 4.91574943e-01 1.25695860e+00 -1.36519539e+00 -6.96088135e-01 4.53826040e-01 8.92654657e-02 2.81121284e-01 7.66754568e-01 7.24789500e-01 -6.56984329e-01 -6.21640086e-02 -4.34502661e-01 -2.83172399e-01 -1.23611772e+00 9.18789625e-01 1.35735631e-01 -1.54490456e-01 -6.78783000e-01 7.79766798e-01 3.23484510e-01 2.61621475e-02 -1.22921519e-01 2.16479331e-01 -3.10140818e-01 -1.57888800e-01 9.87286747e-01 3.74306172e-01 -1.62582397e-01 -6.04828000e-01 -2.75671244e-01 7.91438758e-01 -2.89087117e-01 3.19852948e-01 1.44060814e+00 -5.75283468e-01 -6.07635200e-01 -6.65571690e-02 1.14536440e+00 -6.60359561e-02 -1.39073217e+00 -5.86449862e-01 -3.45881224e-01 -1.01162291e+00 5.33100724e-01 -4.25852418e-01 -1.48800647e+00 8.60720634e-01 6.68087721e-01 2.63429582e-01 1.80714536e+00 -4.04873580e-01 1.11560988e+00 3.01287770e-01 4.53524560e-01 -8.27695191e-01 4.37771827e-01 1.61803588e-01 8.17063332e-01 -9.18541789e-01 2.55364209e-01 -6.41093135e-01 -3.89194727e-01 1.13201463e+00 7.88068771e-01 -3.12062651e-01 5.75502932e-01 4.69850115e-02 -7.77451620e-02 -2.36525685e-01 -6.25065625e-01 -1.99727386e-01 2.66905785e-01 4.63166833e-01 -2.76371062e-01 -3.72703195e-01 5.78756854e-02 3.27036440e-01 2.50066668e-01 -9.00272951e-02 5.44291377e-01 1.28763425e+00 -1.00110137e+00 -8.28362703e-01 -5.55739701e-01 3.80847335e-01 -3.83685023e-01 -1.13086917e-01 4.29440141e-02 5.38753271e-01 2.47150540e-01 1.12144339e+00 -2.42982566e-01 -8.60229492e-01 1.07575864e-01 -3.84629846e-01 1.54926568e-01 -4.56969708e-01 -3.51009071e-01 4.34244610e-02 -4.80038404e-01 -5.25536954e-01 -2.96346813e-01 -2.11038843e-01 -1.12112057e+00 -3.22111458e-01 -6.57935083e-01 3.32708597e-01 3.48342985e-01 8.58729720e-01 6.36890590e-01 9.30924654e-01 1.19932544e+00 -1.05975103e+00 -2.68003553e-01 -7.84622073e-01 -8.11333179e-01 1.42306357e-03 6.42513692e-01 -7.29323685e-01 -4.55658555e-01 -2.04177991e-01]
[10.904335021972656, -2.9747817516326904]
bc634f2b-3d58-483b-a72d-a85c6cea8ae2
multimodal-brain-age-estimation-using
2307.04639
null
https://arxiv.org/abs/2307.04639v1
https://arxiv.org/pdf/2307.04639v1.pdf
Multimodal brain age estimation using interpretable adaptive population-graph learning
Brain age estimation is clinically important as it can provide valuable information in the context of neurodegenerative diseases such as Alzheimer's. Population graphs, which include multimodal imaging information of the subjects along with the relationships among the population, have been used in literature along with Graph Convolutional Networks (GCNs) and have proved beneficial for a variety of medical imaging tasks. A population graph is usually static and constructed manually using non-imaging information. However, graph construction is not a trivial task and might significantly affect the performance of the GCN, which is inherently very sensitive to the graph structure. In this work, we propose a framework that learns a population graph structure optimized for the downstream task. An attention mechanism assigns weights to a set of imaging and non-imaging features (phenotypes), which are then used for edge extraction. The resulting graph is used to train the GCN. The entire pipeline can be trained end-to-end. Additionally, by visualizing the attention weights that were the most important for the graph construction, we increase the interpretability of the graph. We use the UK Biobank, which provides a large variety of neuroimaging and non-imaging phenotypes, to evaluate our method on brain age regression and classification. The proposed method outperforms competing static graph approaches and other state-of-the-art adaptive methods. We further show that the assigned attention scores indicate that there are both imaging and non-imaging phenotypes that are informative for brain age estimation and are in agreement with the relevant literature.
['Daniel Rueckert', 'Alexander Hammers', 'Rolandos Alexandros Potamias', 'Vasileios Baltatzis', 'Kyriaki-Margarita Bintsi']
2023-07-10
null
null
null
null
['age-estimation', 'graph-learning', 'graph-construction', 'age-estimation']
['computer-vision', 'graphs', 'graphs', 'miscellaneous']
[ 2.19921917e-01 4.25773352e-01 1.72896951e-01 -5.39307594e-01 -1.91952884e-01 -1.96209788e-01 4.81833577e-01 5.37758946e-01 -6.70916975e-01 7.37661481e-01 3.03661704e-01 -1.62662551e-01 -2.05875516e-01 -8.54800344e-01 -5.45950174e-01 -8.48572671e-01 -6.23329043e-01 7.08179474e-01 1.23087116e-01 5.62479272e-02 1.17281757e-01 4.63400692e-01 -1.26647496e+00 -1.07502453e-01 1.08250976e+00 1.08044779e+00 2.13608727e-01 5.67566454e-01 9.48200673e-02 3.52799237e-01 -4.82118070e-01 -7.12871134e-01 -1.49534701e-03 -1.39568433e-01 -5.62914193e-01 -2.79904082e-02 6.35273099e-01 -1.94120720e-01 -1.71564564e-01 1.13525951e+00 8.01870227e-01 -8.90455097e-02 7.82049298e-01 -1.18533456e+00 -6.03837311e-01 7.83212304e-01 -5.20048022e-01 4.91286784e-01 -9.55855995e-02 2.28248239e-01 1.15345013e+00 -3.74752700e-01 7.35021710e-01 1.17989576e+00 7.20658958e-01 5.52906871e-01 -1.01625347e+00 -4.51022446e-01 2.58351505e-01 5.85576296e-01 -8.73165488e-01 -1.40904114e-01 7.75624752e-01 -6.78506374e-01 6.80694878e-01 1.01745315e-01 1.08985114e+00 1.28127384e+00 4.21982884e-01 2.89407074e-01 9.55671549e-01 -2.17957869e-01 2.16841936e-01 -5.17879307e-01 4.22558933e-01 1.02988255e+00 3.63009155e-01 -1.10406756e-01 -3.05632055e-01 -1.65452793e-01 6.70384228e-01 -1.45758823e-01 -3.90169352e-01 -2.96102524e-01 -1.32418239e+00 8.25573266e-01 9.06676233e-01 2.33679265e-01 -6.52661860e-01 2.14173555e-01 4.32361603e-01 -1.01711145e-02 7.11907327e-01 4.74819124e-01 -4.16691512e-01 2.03028902e-01 -6.79512739e-01 1.48533955e-01 4.30295914e-01 3.31701577e-01 4.88488793e-01 -9.43360478e-02 -4.25494075e-01 8.62764478e-01 3.06357861e-01 2.09154651e-01 4.33397382e-01 -7.14843869e-01 4.18701679e-01 9.24701989e-01 -5.59682846e-01 -6.74689353e-01 -1.04547668e+00 -7.04281867e-01 -1.05371034e+00 2.02808902e-01 7.54418433e-01 -2.08165586e-01 -1.14368975e+00 2.04478765e+00 3.58273625e-01 -1.50073995e-03 -3.87210041e-01 8.45837593e-01 8.73277724e-01 -7.38606900e-02 4.97707009e-01 1.44427530e-02 1.75148749e+00 -8.07934761e-01 -5.32772601e-01 -4.01413023e-01 6.82149947e-01 -9.63057727e-02 9.00850713e-01 1.01652123e-01 -9.29042041e-01 -1.55459628e-01 -8.15517426e-01 -1.88147679e-01 -4.66090381e-01 1.25430763e-01 8.81324589e-01 4.94974405e-01 -1.32363045e+00 7.97195256e-01 -1.00453305e+00 -5.07379651e-01 9.33743298e-01 4.88719404e-01 -4.98061866e-01 -1.52204171e-01 -1.20704103e+00 9.40057576e-01 4.26743507e-01 2.50169933e-01 -5.74558794e-01 -8.26939285e-01 -7.85229385e-01 2.20507056e-01 1.75978392e-01 -1.36898196e+00 5.29141366e-01 -7.36731589e-01 -8.45901847e-01 8.91241252e-01 1.24442354e-01 -6.09257340e-01 6.41810834e-01 -3.20334025e-02 -1.64212003e-01 4.37270343e-01 7.09205717e-02 8.64549696e-01 8.82966816e-01 -6.01575375e-01 -1.46584228e-01 -9.77545977e-01 2.46126112e-03 8.89748186e-02 -4.33422983e-01 -7.32740201e-03 -3.00745845e-01 -6.29805207e-01 -1.49059929e-02 -9.07629490e-01 -2.55541950e-01 3.03852588e-01 -4.80247736e-01 -2.18279555e-01 4.69054461e-01 -1.22801983e+00 1.12912285e+00 -1.74912250e+00 4.68555361e-01 2.39678174e-01 8.90120983e-01 -1.41908988e-01 -4.73964587e-02 -7.55613595e-02 -3.29397917e-01 2.06949562e-01 -4.57066983e-01 -3.35438550e-01 -2.13806376e-01 -1.60984173e-01 5.91807485e-01 5.24178505e-01 3.45057040e-01 1.12200844e+00 -9.53371108e-01 -5.54347813e-01 -7.63824303e-03 6.34774387e-01 -5.12234151e-01 1.05384670e-01 -1.19309671e-01 6.66632175e-01 -2.76002765e-01 4.45064485e-01 3.28054994e-01 -2.91446686e-01 2.05791995e-01 -4.97251064e-01 3.72961879e-01 -8.96491036e-02 -3.93797606e-01 1.48076522e+00 -7.84393772e-02 4.01575565e-01 -9.82957184e-02 -1.10205984e+00 7.06535101e-01 7.52091333e-02 6.10513151e-01 -4.15250182e-01 3.23065609e-01 -4.14378271e-02 6.86430752e-01 -5.84681153e-01 -2.07358286e-01 1.62469521e-01 2.15451449e-01 3.42363894e-01 2.21133739e-01 2.88152844e-01 5.32524884e-01 3.08795631e-01 1.48754621e+00 -3.48297730e-02 3.44759464e-01 -4.13881660e-01 5.43001235e-01 -4.28912520e-01 5.58308542e-01 3.78471285e-01 -2.65867621e-01 5.34650028e-01 1.06026840e+00 -3.23931634e-01 -1.14533186e+00 -9.90841687e-01 -4.37137112e-02 8.40881705e-01 -4.90232855e-01 -5.26596606e-01 -9.95688558e-01 -8.68163645e-01 -9.50037166e-02 4.50413316e-01 -1.07266831e+00 -3.89809102e-01 -3.68482709e-01 -1.18026829e+00 2.87142992e-01 7.00123250e-01 3.43488187e-01 -9.70256984e-01 -4.29951042e-01 1.39951870e-01 -8.38327110e-02 -1.20620298e+00 -4.76558626e-01 -1.01896510e-01 -1.27366495e+00 -1.42753625e+00 -7.89597929e-01 -4.18369472e-01 1.13772333e+00 -3.53171617e-01 1.20813024e+00 3.73719603e-01 -4.73301411e-01 3.64689201e-01 -2.29374513e-01 -4.24805492e-01 -3.42401952e-01 4.72913861e-01 -1.32430732e-01 1.19238414e-01 2.42592599e-02 -9.25659060e-01 -8.75911117e-01 -1.06772698e-01 -6.32264674e-01 1.73676714e-01 7.98616886e-01 7.77616084e-01 4.04746652e-01 -2.72327989e-01 6.81081593e-01 -1.14914525e+00 6.15675926e-01 -4.60779458e-01 -3.96125823e-01 2.75141776e-01 -6.40404642e-01 3.55242729e-01 3.93650323e-01 -4.75818038e-01 -6.50372148e-01 -1.96636114e-02 -1.49655968e-01 -1.49934828e-01 -1.27406344e-02 6.52754128e-01 -3.87648106e-01 -2.94542406e-02 4.98266011e-01 -3.85236770e-01 3.27464461e-01 -3.92210126e-01 3.00154597e-01 1.90607980e-01 3.29404682e-01 -4.36040312e-01 3.72262388e-01 2.23397553e-01 5.08047342e-01 -7.33623564e-01 -6.30145848e-01 -9.02001858e-02 -8.12925994e-01 -4.42933619e-01 1.04750204e+00 -4.60130274e-01 -6.20450556e-01 6.79957449e-01 -9.03791010e-01 -4.68731433e-01 6.00637496e-02 4.29246873e-01 -2.86239475e-01 4.51331407e-01 -5.44715643e-01 -5.55167496e-01 -8.71603549e-01 -1.27685356e+00 1.10460854e+00 1.07648306e-01 -2.03253105e-01 -1.32138121e+00 -3.50686729e-01 3.36769283e-01 3.96325618e-01 6.08408093e-01 1.61097169e+00 -6.27847910e-01 -2.21341386e-01 -7.74837956e-02 -4.30109382e-01 3.50635499e-02 1.86867230e-02 -1.00628652e-01 -7.56284058e-01 -1.88286081e-01 -4.10884172e-01 8.67078528e-02 1.00312495e+00 6.99019969e-01 1.34951115e+00 -1.83978081e-01 -2.96191752e-01 4.52579051e-01 1.11171687e+00 -2.26030082e-01 8.59376729e-01 1.81501076e-01 1.09424603e+00 7.74602354e-01 -4.80847619e-02 2.83381850e-01 6.04090154e-01 5.99217713e-01 6.63250983e-01 -2.80271202e-01 -2.19868377e-01 3.03357422e-01 2.76812077e-01 5.43623924e-01 -5.35871208e-01 -5.40009812e-02 -9.44625795e-01 3.04502100e-01 -1.71626961e+00 -6.07390165e-01 -5.15039861e-01 2.26601768e+00 7.44971037e-01 2.12555081e-01 3.41733247e-01 -1.10776432e-01 8.49955976e-01 1.10930251e-02 -6.89850569e-01 -6.86012506e-02 -5.92654422e-02 3.76259208e-01 4.56368804e-01 1.80197954e-01 -9.47104573e-01 5.91173232e-01 5.44535780e+00 1.31030023e-01 -1.06336713e+00 1.64898619e-01 8.28984261e-01 -4.01336374e-03 -1.16001807e-01 -1.48431957e-01 -3.08595568e-01 4.72301871e-01 1.05974460e+00 -1.32157803e-01 4.83140022e-01 5.73731005e-01 2.19284028e-01 -1.34965509e-01 -1.27327132e+00 7.73821115e-01 -6.23948581e-04 -9.11174059e-01 -6.63778484e-02 1.76632553e-01 1.94351837e-01 1.17667414e-01 -2.04988211e-01 8.45569074e-02 -6.82080630e-03 -8.24648321e-01 5.05223930e-01 8.30067694e-01 7.41748273e-01 -5.30055940e-01 8.80863428e-01 -9.25066099e-02 -1.09970224e+00 -1.39969051e-01 -1.66511521e-01 3.75247419e-01 2.97879040e-01 1.04976106e+00 -8.16207767e-01 4.67215657e-01 7.19486296e-01 7.27038085e-01 -1.38613582e+00 1.23632371e+00 -4.84464526e-01 4.80973363e-01 -7.95093402e-02 2.54151314e-01 -7.88163543e-02 -3.82079571e-01 4.87342954e-01 7.76289999e-01 4.23497528e-01 -1.90857500e-01 -1.38399705e-01 1.02111161e+00 -2.57288367e-01 2.75152951e-01 -3.52612913e-01 -2.63297111e-01 7.90609140e-03 1.61316538e+00 -1.00855219e+00 -3.01006466e-01 -3.72649580e-01 6.97075188e-01 7.23465860e-01 1.64906710e-01 -6.83072090e-01 -1.63632348e-01 5.95769286e-01 2.77206689e-01 2.04555914e-02 -2.38638401e-01 2.91579938e-03 -1.02862513e+00 -3.23221199e-02 -5.90840399e-01 5.47025979e-01 -9.31883872e-01 -1.56983352e+00 5.63669741e-01 4.26051542e-02 -5.81743300e-01 -1.13639690e-01 -6.76671803e-01 -7.40892112e-01 8.02586555e-01 -1.27217841e+00 -1.29701889e+00 -6.40074551e-01 3.25302690e-01 1.74603630e-02 7.82924611e-03 6.96762800e-01 4.05956089e-01 -8.23260665e-01 3.70076984e-01 -4.30022776e-01 2.69406825e-01 6.18503511e-01 -1.59378254e+00 3.96192014e-01 8.20295811e-01 -1.29276350e-01 6.17737949e-01 4.26327616e-01 -9.66990888e-01 -1.01782763e+00 -1.18682873e+00 7.58311808e-01 -3.06586325e-01 8.85132134e-01 -4.55315828e-01 -9.99393940e-01 5.78841627e-01 -9.15552080e-02 8.79832953e-02 5.18788695e-01 2.12023139e-01 -2.32530728e-01 -1.39382377e-01 -9.57522810e-01 6.12391114e-01 1.44825506e+00 -1.79990590e-01 -2.06546307e-01 5.03609955e-01 6.02327287e-01 -2.44974047e-01 -1.22767234e+00 2.77816325e-01 5.53644717e-01 -8.48347425e-01 1.01530588e+00 -6.85431659e-01 3.93226653e-01 1.83546752e-01 4.61179793e-01 -1.55823898e+00 -4.28774416e-01 -1.16352603e-01 -2.86679387e-01 1.17228222e+00 4.50660467e-01 -8.62867296e-01 5.95430493e-01 7.54019022e-01 -2.78037220e-01 -8.91769946e-01 -7.46399641e-01 -4.29324001e-01 -3.36730108e-02 -2.31370345e-01 5.54057896e-01 6.38124824e-01 -3.95241231e-01 6.28742754e-01 3.26635577e-02 1.68723121e-01 8.79602134e-01 -3.69187504e-01 2.05610603e-01 -1.86166370e+00 -5.20592891e-02 -7.44241714e-01 -8.52948844e-01 2.82573383e-02 4.71043557e-01 -1.20503795e+00 -4.45476532e-01 -1.88433158e+00 2.66829103e-01 -3.24530095e-01 -3.60707909e-01 5.97567379e-01 -5.86288393e-01 5.36385737e-02 -1.13482833e-01 -2.00859949e-01 -2.20639378e-01 3.87289405e-01 1.31100655e+00 -3.10112447e-01 -2.62663327e-02 -8.40330198e-02 -6.64768994e-01 5.20830870e-01 1.01227236e+00 -4.05563146e-01 -4.52290207e-01 -2.59046882e-01 2.01290548e-01 -2.65295058e-01 5.20512402e-01 -9.73425210e-01 -1.08772613e-01 3.59582096e-01 5.66384017e-01 -7.41519257e-02 1.68957040e-01 -7.08297491e-01 1.26852110e-01 5.92907190e-01 -1.22979477e-01 2.81807840e-01 -1.92236498e-01 4.89020616e-01 3.14012498e-01 -2.64057338e-01 6.55252099e-01 -2.01886922e-01 -2.94568896e-01 8.31359982e-01 3.16109695e-02 1.93113148e-01 7.52538383e-01 3.32791395e-02 -4.47961301e-01 -3.30433607e-01 -1.00221121e+00 1.64016590e-01 3.79608542e-01 1.96460664e-01 3.70066911e-01 -1.10727680e+00 -7.64389396e-01 -6.98380247e-02 1.03319302e-01 -1.86628312e-01 3.33196104e-01 1.39299047e+00 -4.01908010e-01 -1.34737149e-01 -5.96157730e-01 -6.10587239e-01 -1.35791576e+00 4.86923128e-01 4.22744125e-01 -5.74990273e-01 -7.71045506e-01 5.61193228e-01 2.22578168e-01 -4.11552712e-02 9.84934792e-02 -5.62472165e-01 -6.42672241e-01 4.38513517e-01 3.71387303e-01 3.31795216e-01 1.95219308e-01 -6.06228650e-01 -3.09380680e-01 4.49800164e-01 -5.87802492e-02 9.00547206e-02 1.82271266e+00 -6.63839839e-03 -5.16611993e-01 2.37643793e-01 1.01266670e+00 -3.77331525e-01 -1.14617372e+00 1.79117862e-02 2.08976164e-01 -4.83571663e-02 1.74659595e-01 -6.36549354e-01 -1.67799115e+00 9.51823473e-01 7.49952435e-01 9.05639976e-02 1.07947218e+00 1.32143095e-01 5.83626330e-01 -1.42565630e-02 3.70359421e-01 -7.35622048e-01 1.26795722e-02 1.38266578e-01 9.98546302e-01 -1.10679197e+00 2.57870823e-01 -5.25259972e-01 -4.08367544e-01 1.30920434e+00 6.59311891e-01 1.00767642e-01 4.49407458e-01 -8.52827877e-02 -2.78094441e-01 -4.93015707e-01 -3.90159756e-01 -3.32827330e-01 6.07590318e-01 8.69278848e-01 3.85865659e-01 2.11640131e-02 -6.22016072e-01 7.43153512e-01 -3.16622943e-01 -1.61997244e-01 2.94938743e-01 4.76118863e-01 -1.01955466e-01 -1.21251702e+00 -7.75510445e-02 1.12694275e+00 -5.08086801e-01 -2.04555109e-01 -4.85469133e-01 5.85062623e-01 1.45710260e-01 4.37345207e-01 -3.14679556e-02 -4.89240363e-02 1.77292630e-01 1.93136305e-01 7.81300128e-01 -5.76341510e-01 -6.22346222e-01 -2.83283055e-01 4.54941362e-01 -5.70538998e-01 -3.92635435e-01 -6.05807841e-01 -1.25928450e+00 -2.50773821e-02 -2.40771338e-01 -2.16607392e-01 6.79431975e-01 9.22046065e-01 4.04229522e-01 1.03209662e+00 -3.29427160e-02 -9.85301137e-01 2.16588378e-02 -1.14189994e+00 -4.45837110e-01 6.33376539e-01 -9.08540487e-02 -1.03742719e+00 -1.73058406e-01 1.02375455e-01]
[12.382786750793457, 3.360534191131592]
a77654fe-b5dd-4f6d-acbe-972a7d03c2c3
2305-14562
2305.14562
null
https://arxiv.org/abs/2305.14562v1
https://arxiv.org/pdf/2305.14562v1.pdf
GiPH: Generalizable Placement Learning for Adaptive Heterogeneous Computing
Careful placement of a computational application within a target device cluster is critical for achieving low application completion time. The problem is challenging due to its NP-hardness and combinatorial nature. In recent years, learning-based approaches have been proposed to learn a placement policy that can be applied to unseen applications, motivated by the problem of placing a neural network across cloud servers. These approaches, however, generally assume the device cluster is fixed, which is not the case in mobile or edge computing settings, where heterogeneous devices move in and out of range for a particular application. We propose a new learning approach called GiPH, which learns policies that generalize to dynamic device clusters via 1) a novel graph representation gpNet that efficiently encodes the information needed for choosing a good placement, and 2) a scalable graph neural network (GNN) that learns a summary of the gpNet information. GiPH turns the placement problem into that of finding a sequence of placement improvements, learning a policy for selecting this sequence that scales to problems of arbitrary size. We evaluate GiPH with a wide range of task graphs and device clusters and show that our learned policy rapidly find good placements for new problem instances. GiPH finds placements with up to 30.5% lower completion times, searching up to 3X faster than other search-based placement policies.
['Carlee Joe-Wong', 'Bob Iannucci', 'Yanqi Zhou', 'James Laudon', 'Aviral Shrivastava', 'Harshul Singh', 'Edward Andert', 'Chaoran Zhang', 'Yi Hu']
2023-05-23
null
null
null
null
['edge-computing']
['time-series']
[ 1.40721068e-01 1.61220171e-02 -8.24718297e-01 2.02172603e-02 -7.55818963e-01 -8.09644580e-01 -3.58147979e-01 2.98801184e-01 3.56116109e-02 5.62477410e-01 -2.24255562e-01 -8.67012441e-01 -5.51970422e-01 -7.07496166e-01 -1.15185654e+00 -6.92552030e-01 -3.41462284e-01 9.45372820e-01 3.07960212e-01 7.84270316e-02 2.63805211e-01 5.40627718e-01 -9.77966785e-01 1.98894188e-01 5.42631805e-01 8.34418952e-01 5.91422617e-01 8.90265882e-01 1.81573823e-01 4.59760696e-01 -6.73356414e-01 7.92954396e-03 5.45135856e-01 2.57018488e-02 -8.84080768e-01 2.73404479e-01 -5.81828579e-02 -4.51529883e-02 -4.32201236e-01 8.07723820e-01 7.24177182e-01 -3.73577029e-02 2.63359189e-01 -1.68312490e+00 -6.38537765e-01 9.39594626e-01 -9.11182165e-01 5.01003206e-01 1.34807631e-01 8.76093581e-02 1.00598562e+00 1.51461795e-01 4.85049039e-01 6.86110973e-01 6.20852351e-01 3.71269703e-01 -1.02164662e+00 -5.08766711e-01 4.75629866e-01 1.83686361e-01 -1.35529935e+00 -1.69436201e-01 6.83425128e-01 -2.25107044e-01 1.25974202e+00 2.56892025e-01 4.20526743e-01 8.85650992e-01 3.07525635e-01 6.29342914e-01 4.08837855e-01 -4.12154466e-01 6.67270303e-01 -2.39186972e-01 -1.91276848e-01 3.37098032e-01 4.43695873e-01 -3.28830689e-01 -3.18104059e-01 -4.75627303e-01 6.93505585e-01 2.43581265e-01 -1.80073515e-01 -6.44119740e-01 -7.63026834e-01 3.84303570e-01 7.57063389e-01 2.69773424e-01 -6.38670981e-01 6.51797593e-01 4.40072656e-01 1.59957886e-01 2.69689918e-01 7.40360260e-01 -8.70034933e-01 -2.65193850e-01 -7.24638164e-01 5.28748445e-02 1.12557995e+00 1.29161310e+00 6.96217477e-01 -1.88769951e-01 -1.60560504e-01 4.91143942e-01 -1.68253407e-01 4.85139787e-01 7.04911053e-02 -8.09003711e-01 7.47155130e-01 5.77048659e-01 4.64492030e-02 -1.11823034e+00 -5.18686533e-01 -4.88246500e-01 -7.34359622e-01 -2.73686230e-01 -6.54769093e-02 -6.26293898e-01 -6.51692927e-01 1.86115646e+00 2.46852726e-01 6.72976017e-01 -3.79624516e-01 7.15798974e-01 -2.97758095e-02 8.76176953e-01 -3.35214846e-02 -1.21175855e-01 8.70388746e-01 -1.13050163e+00 -2.17548922e-01 -5.08308947e-01 7.15658784e-01 -3.96456778e-01 8.42961252e-01 3.12682003e-01 -1.01626992e+00 -2.07501367e-01 -1.06334901e+00 4.69680399e-01 -3.07814062e-01 -6.43285513e-02 7.58539081e-01 7.14702010e-01 -1.77050233e+00 5.55474460e-01 -9.69845653e-01 -7.20268786e-01 5.54414213e-01 1.17869258e+00 2.16954306e-01 -4.62292403e-01 -3.45378339e-01 4.10477042e-01 3.59959364e-01 -2.48585060e-01 -8.62594485e-01 -7.87194133e-01 -4.77694452e-01 5.12598932e-01 8.24198008e-01 -7.91997015e-01 1.15015900e+00 -7.57043600e-01 -1.19412005e+00 3.29519778e-01 -1.57531220e-02 -3.24426651e-01 -1.12943873e-01 2.03824058e-01 -2.41898790e-01 -1.30590945e-01 1.62491962e-01 4.19386894e-01 7.62870610e-01 -1.30470765e+00 -7.95700192e-01 -3.28159422e-01 5.48291326e-01 3.88710588e-01 -6.50300980e-01 -2.02491492e-01 -8.90405774e-01 -6.37754127e-02 -1.94216803e-01 -1.24401224e+00 -6.80596828e-01 -3.70190859e-01 -5.47262907e-01 -2.67581463e-01 8.98875475e-01 -3.88559669e-01 1.39979684e+00 -1.93581808e+00 2.91243374e-01 5.35028756e-01 5.96176088e-01 9.66208652e-02 -2.77440697e-01 5.26985466e-01 2.92854816e-01 2.51209408e-01 3.11356068e-01 -1.63316563e-01 -3.34059484e-02 4.47592705e-01 -4.05700281e-02 2.04259515e-01 -1.65401042e-01 1.12294161e+00 -1.00466824e+00 -8.58935434e-03 8.58346894e-02 7.64326975e-02 -8.95352662e-01 -3.84220667e-02 -3.73160690e-01 1.41359597e-01 -7.33947039e-01 6.58737302e-01 6.36374652e-01 -1.20466757e+00 8.21681917e-01 1.12231001e-01 4.83640403e-01 7.88512006e-02 -8.88255298e-01 1.54100573e+00 -6.79444432e-01 4.33540076e-01 1.76733315e-01 -1.02588463e+00 5.89103699e-01 3.38175595e-02 1.04937375e+00 -3.98277968e-01 2.04037562e-01 1.19882219e-01 -1.46962583e-01 -4.82042819e-01 3.13925892e-01 4.06352907e-01 -2.47206673e-01 5.53139806e-01 -4.46375340e-01 2.74025172e-01 -6.75817057e-02 1.49694800e-01 2.19748449e+00 -4.50120449e-01 -2.40875836e-02 -2.89714813e-01 -8.72307271e-02 8.25850219e-02 5.60145557e-01 9.80237603e-01 2.27797627e-02 3.19963902e-01 7.20613182e-01 -5.27264237e-01 -1.01350749e+00 -9.33688402e-01 5.11820436e-01 1.33818722e+00 4.49587703e-01 -5.16037703e-01 -9.13500428e-01 -7.53605425e-01 1.20836556e-01 1.54687807e-01 -3.10916990e-01 -2.38077492e-01 -7.27798223e-01 -7.19574749e-01 -1.27715304e-01 7.60151207e-01 3.73950489e-02 -1.05384362e+00 -5.45181394e-01 3.01126242e-01 2.70059314e-02 -1.42664397e+00 -9.18975949e-01 6.56359732e-01 -8.92883778e-01 -1.27227926e+00 -2.91810006e-01 -1.23721159e+00 1.02583849e+00 4.48426425e-01 1.35000038e+00 4.71893549e-01 -2.30096802e-01 4.43268448e-01 -2.84645021e-01 -6.07417971e-02 -1.64608181e-01 9.85772252e-01 2.33949572e-02 -3.30719054e-01 2.55467325e-01 -8.11116040e-01 -4.70091730e-01 2.40089849e-01 -7.65925705e-01 -2.35569686e-01 6.15783811e-01 4.76215094e-01 7.63678849e-01 7.34578490e-01 5.76103985e-01 -1.26646149e+00 8.83439541e-01 -9.00009930e-01 -7.33885229e-01 4.30678755e-01 -5.12832046e-01 -1.30966425e-01 9.17276621e-01 -6.28035069e-01 -7.35641122e-02 2.96778589e-01 3.60814705e-02 -9.46353376e-01 1.83478780e-02 2.02692762e-01 -3.31800640e-01 -4.28815067e-01 7.01319635e-01 -2.68900752e-01 -4.37716305e-01 -7.12812021e-02 1.20497666e-01 5.77291310e-01 3.02933872e-01 -8.98208380e-01 6.62139654e-01 1.24087431e-01 1.15307011e-01 -4.21704352e-01 -5.46625555e-01 -6.35701120e-01 -4.30186838e-01 7.64934868e-02 5.74858427e-01 -6.83840930e-01 -1.13221121e+00 -6.63150698e-02 -8.78948212e-01 -1.00484729e+00 -2.35994145e-01 -3.05223465e-01 -5.54928541e-01 3.03848516e-02 -6.00893974e-01 -4.85003948e-01 -3.34121883e-01 -1.41979480e+00 1.28789401e+00 2.86126852e-01 5.84492423e-02 -1.18774498e+00 -2.68668264e-01 5.77752143e-02 6.50232852e-01 1.82953537e-01 1.42023075e+00 -4.55451578e-01 -1.08152902e+00 -7.00735524e-02 -1.35405287e-01 -8.79460201e-02 3.02378476e-01 -3.05321842e-01 -4.76515383e-01 -8.67117226e-01 -3.87531310e-01 7.04441145e-02 2.10727587e-01 7.74550021e-01 1.74357820e+00 -5.90132236e-01 -1.06860399e+00 7.94412017e-01 1.68235993e+00 5.00378847e-01 5.22485971e-01 2.24731252e-01 8.38105381e-01 -1.24201842e-01 1.58844814e-01 4.01861697e-01 4.07500356e-01 7.65997171e-01 7.60512233e-01 -2.01035917e-01 6.95424080e-02 -2.97735989e-01 -1.67888850e-02 6.85106874e-01 3.80208611e-01 -8.90032768e-01 -1.04121208e+00 6.91130698e-01 -2.15025187e+00 -2.16144368e-01 6.24308765e-01 2.12077904e+00 3.53336722e-01 4.32982028e-01 3.59799653e-01 1.25513449e-02 8.88132513e-01 -4.36353050e-02 -1.13797593e+00 -3.15752298e-01 3.61444116e-01 3.10124010e-01 9.85491514e-01 1.67166695e-01 -8.95804286e-01 1.00775301e+00 6.22219515e+00 7.96620190e-01 -1.22585773e+00 2.00373024e-01 8.73470008e-01 6.57581687e-02 -1.54171050e-01 -9.32270288e-02 -6.61604345e-01 6.08444154e-01 1.06939840e+00 -3.01067442e-01 1.08879995e+00 1.03419387e+00 -6.45419508e-02 2.15884805e-01 -1.24272287e+00 1.37917340e+00 -7.46470168e-02 -1.56767499e+00 -4.27558273e-01 3.02059442e-01 1.11237097e+00 2.55498767e-01 2.77887791e-01 2.43569449e-01 6.04418278e-01 -7.96321094e-01 7.30127469e-02 -2.14018762e-01 7.33124316e-01 -8.67415130e-01 5.26915371e-01 3.70939046e-01 -1.21416593e+00 -5.95415294e-01 -5.40745139e-01 1.67518556e-02 7.24806488e-02 5.27362585e-01 -1.28310657e+00 3.05278391e-01 8.85686636e-01 6.84809506e-01 -4.28457737e-01 1.41602886e+00 4.65069041e-02 5.46255648e-01 -5.17644882e-01 -1.23497352e-01 1.23140238e-01 2.59223878e-01 2.92737156e-01 6.89563751e-01 4.85679448e-01 -6.64981827e-02 6.81282043e-01 5.19005358e-01 -6.30080163e-01 -1.02229699e-01 -6.75867915e-01 -1.68868735e-01 8.92510951e-01 1.25660551e+00 -1.27340484e+00 9.11834463e-02 -4.87640239e-02 9.62223649e-01 4.62519765e-01 5.03924847e-01 -8.11789691e-01 -4.28792415e-03 8.90561342e-01 4.35274065e-01 4.61195707e-01 -2.73565024e-01 -4.61127311e-01 -5.73035896e-01 2.48171955e-01 -7.60335565e-01 3.69089991e-01 -5.95723212e-01 -1.25150871e+00 7.44948983e-01 -1.71561360e-01 -8.36722732e-01 -1.21698454e-01 -6.05947495e-01 -6.78586602e-01 5.19002676e-01 -1.27987111e+00 -6.25355422e-01 -2.94146091e-01 3.56675178e-01 4.65959311e-01 1.66965686e-02 5.05324960e-01 4.51028526e-01 -7.31921554e-01 7.64405131e-01 3.54056247e-02 -4.13759798e-01 4.34321135e-01 -1.18963277e+00 8.26055348e-01 7.14233100e-01 1.40599474e-01 2.80488312e-01 3.70397866e-01 -6.12712860e-01 -1.86999285e+00 -1.26089132e+00 5.62773168e-01 -4.97009844e-01 5.76944351e-01 -5.22985995e-01 -5.67934871e-01 7.46385932e-01 5.86758777e-02 1.90165818e-01 4.39795315e-01 4.27025229e-01 3.05089168e-02 -3.34081590e-01 -1.09394169e+00 6.35867655e-01 1.28263521e+00 -2.81739771e-01 5.69052756e-01 1.03959370e+00 1.16383815e+00 -8.70639205e-01 -6.04114711e-01 3.80029917e-01 -1.22516617e-01 -3.18061411e-01 8.73226464e-01 -7.99492776e-01 5.05965427e-02 -2.04014465e-01 -1.76336274e-01 -1.23747039e+00 -6.00446165e-01 -9.34133232e-01 -4.32042390e-01 5.99424005e-01 6.49028003e-01 -3.32316160e-01 1.37629712e+00 4.89166886e-01 -1.85878918e-01 -1.29584312e+00 -6.18896544e-01 -7.26857543e-01 -2.79112130e-01 -1.92562848e-01 1.02047276e+00 7.47931302e-01 -8.85880813e-02 5.27677715e-01 -3.26176643e-01 7.37857819e-01 2.55805135e-01 3.92564051e-02 5.93207002e-01 -1.00526559e+00 -8.31691027e-01 -3.11747909e-01 -3.99963856e-01 -1.39468181e+00 2.99354136e-01 -9.72818494e-01 -2.35244930e-02 -1.77338624e+00 2.19071969e-01 -1.17707634e+00 -4.89544749e-01 8.04997861e-01 -2.48136241e-02 -2.36112297e-01 2.15905905e-01 8.75189379e-02 -1.23353314e+00 -2.06139371e-01 9.95034575e-01 -1.43432602e-01 -6.75177753e-01 4.73624051e-01 -1.09447372e+00 2.10461006e-01 1.06535184e+00 -4.90765631e-01 -8.49471748e-01 -9.04392421e-01 6.21010244e-01 2.81573296e-01 -1.02767214e-01 -1.20192099e+00 7.25420296e-01 -1.89199716e-01 7.06101432e-02 -2.65518785e-01 6.27600923e-02 -1.10607123e+00 2.72055835e-01 3.25311095e-01 -2.17757344e-01 7.54969418e-01 4.56584960e-01 8.20459962e-01 5.50143600e-01 1.78976372e-01 3.10383022e-01 2.83766627e-01 -7.34952092e-01 8.59701395e-01 3.93526405e-02 1.59194142e-01 1.20473218e+00 -1.77648868e-02 -3.17275167e-01 -3.50861460e-01 -7.10564494e-01 5.39000571e-01 6.76032782e-01 6.32077932e-01 3.92518371e-01 -1.05867410e+00 -1.63066179e-01 -1.17706545e-02 -1.05435081e-01 1.08400121e-01 2.20942557e-01 5.96352279e-01 -6.52306795e-01 2.70057052e-01 -1.25019148e-01 -7.36070395e-01 -9.21894312e-01 1.11204565e+00 2.69349843e-01 -6.96628869e-01 -6.05822921e-01 7.25181043e-01 9.10630748e-02 -1.68509841e-01 4.12339658e-01 -2.47687325e-01 3.98573995e-01 -8.36417437e-01 2.10041255e-01 2.48340696e-01 3.28643382e-01 -4.19095345e-02 -4.05046672e-01 3.12036186e-01 -1.81470916e-01 5.23188412e-01 1.33359063e+00 -1.08787827e-01 -2.14546695e-01 -1.51702985e-01 1.16201437e+00 -3.17923635e-01 -1.26211905e+00 -2.73177594e-01 1.72960475e-01 -4.36367452e-01 1.13204280e-02 -6.74913585e-01 -1.55002892e+00 4.05661851e-01 4.66117263e-01 4.00571585e-01 1.60146999e+00 2.04915240e-01 9.21763897e-01 4.39720094e-01 9.91982162e-01 -8.58255148e-01 1.64973646e-01 2.17274487e-01 1.49266675e-01 -8.57227325e-01 -2.61412889e-01 -5.90443909e-01 -1.97548941e-01 7.74319828e-01 8.67143452e-01 -7.88727328e-02 8.79027009e-01 8.24278176e-01 -4.25466478e-01 -2.60798812e-01 -9.01695549e-01 1.44138366e-01 -1.55293375e-01 8.44074368e-01 -2.78755784e-01 4.66545165e-01 3.48465145e-01 3.72209877e-01 -2.15479564e-02 -5.17766587e-02 3.13853323e-01 1.11132658e+00 -2.73594916e-01 -1.39401639e+00 1.13157369e-01 9.02427852e-01 -2.59589851e-01 2.49142684e-02 -6.86577633e-02 3.54685605e-01 1.28850937e-01 1.01710010e+00 6.17477596e-02 -7.68419325e-01 2.72815466e-01 -3.56119275e-01 3.84559274e-01 -9.30483162e-01 -6.98799014e-01 -8.93308371e-02 -4.01840687e-01 -7.10928500e-01 1.65381998e-01 -1.57232329e-01 -1.06051183e+00 -5.32334745e-01 -2.40451291e-01 7.65639246e-02 5.94745755e-01 7.08293796e-01 9.57731307e-01 1.06134093e+00 8.41712415e-01 -9.15506244e-01 -3.72389287e-01 -1.38339221e-01 -6.43985271e-01 -1.51393861e-01 2.23111406e-01 -4.35233265e-01 -1.70649718e-02 -3.81402463e-01]
[5.471680641174316, 3.0657191276550293]
acbad825-6c4c-4c38-8b7d-743c7ce6ebd8
socrates-a-stereo-camera-trap-for-monitoring
2209.0907
null
https://arxiv.org/abs/2209.09070v2
https://arxiv.org/pdf/2209.09070v2.pdf
SOCRATES: A Stereo Camera Trap for Monitoring of Biodiversity
The development and application of modern technology is an essential basis for the efficient monitoring of species in natural habitats and landscapes to trace the development of ecosystems, species communities, and populations, and to analyze reasons of changes. For estimating animal abundance using methods such as camera trap distance sampling, spatial information of natural habitats in terms of 3D (three-dimensional) measurements is crucial. Additionally, 3D information improves the accuracy of animal detection using camera trapping. This study presents a novel approach to 3D camera trapping featuring highly optimized hardware and software. This approach employs stereo vision to infer 3D information of natural habitats and is designated as StereO CameRA Trap for monitoring of biodivErSity (SOCRATES). A comprehensive evaluation of SOCRATES shows not only a $3.23\%$ improvement in animal detection (bounding box $\text{mAP}_{75}$) but also its superior applicability for estimating animal abundance using camera trap distance sampling. The software and documentation of SOCRATES is provided at https://github.com/timmh/socrates
['Hjalmar S. Kühl', 'Volker Steinhage', 'Timm Haucke']
2022-09-19
null
null
null
null
['stereo-matching-1']
['computer-vision']
[-7.14408979e-03 -6.52109087e-01 -9.36087593e-02 -1.12582989e-01 5.61353005e-02 -6.59777164e-01 3.40982169e-01 5.09828389e-01 -1.01611924e+00 5.50790608e-01 3.01452484e-02 -4.80126649e-01 1.97863922e-01 -9.52754200e-01 -4.33266252e-01 -4.47177768e-01 -4.83769119e-01 2.64148980e-01 4.15304452e-01 -5.41047491e-02 3.53782594e-01 9.49367583e-01 -1.56757092e+00 -4.06588286e-01 2.22135395e-01 4.84609246e-01 5.91257036e-01 6.87149823e-01 -6.82373866e-02 -8.49098340e-02 -2.60852635e-01 -2.21045122e-01 2.96899348e-01 -4.65017445e-02 -1.93824366e-01 -3.76047581e-01 1.24972299e-01 -9.44360733e-01 1.02958597e-01 8.99878263e-01 2.82293588e-01 -2.37057641e-01 6.51890993e-01 -6.99615598e-01 3.27550359e-02 3.68954062e-01 -8.29259336e-01 6.07526958e-01 1.76214382e-01 3.42883468e-01 6.12652361e-01 -5.53423703e-01 3.21395516e-01 9.91895258e-01 7.32654989e-01 5.21028675e-02 -1.57560539e+00 -9.50712025e-01 -2.78362036e-01 4.53776717e-02 -1.70933557e+00 -3.94892156e-01 3.28529388e-01 -6.97154582e-01 7.73268461e-01 2.81137496e-01 1.12877333e+00 5.47867358e-01 1.69067770e-01 3.04953367e-01 9.86616910e-01 -2.36521751e-01 1.87512606e-01 1.34968743e-01 -9.50212032e-02 5.91302097e-01 9.05310929e-01 3.51926297e-01 -4.81992334e-01 -4.05571222e-01 1.12477314e+00 5.84418952e-01 2.61234306e-03 -3.08868647e-01 -1.06353104e+00 8.79692137e-01 6.41800463e-01 1.75359219e-01 -3.73327702e-01 1.07242674e-01 3.11595231e-01 -1.81840554e-01 5.02568364e-01 2.65301287e-01 3.90699580e-02 -2.99458325e-01 -1.06368208e+00 8.72712880e-02 4.37607586e-01 5.28547168e-01 7.95459330e-01 -7.94560686e-02 4.18823719e-01 8.17858458e-01 4.29244518e-01 1.32445288e+00 1.15760252e-01 -9.94521558e-01 1.94299012e-01 9.73245144e-01 3.06821913e-01 -1.30633104e+00 -3.98062587e-01 -3.18959624e-01 -6.59620702e-01 2.28500128e-01 1.15587190e-01 2.22630292e-01 -2.63963223e-01 1.26585853e+00 5.34993291e-01 -2.33965471e-01 -4.53643680e-01 6.20279968e-01 5.22475064e-01 7.03462005e-01 1.42483696e-01 -4.16544676e-02 1.13923407e+00 -5.86251020e-02 -1.23714343e-01 -4.45567310e-01 3.12553048e-01 -3.46956372e-01 6.41617596e-01 -8.49298611e-02 -6.10546052e-01 -7.79698193e-02 -8.99259269e-01 2.82543749e-01 -6.81392848e-01 1.49156690e-01 5.53411663e-01 5.61225176e-01 -1.02938330e+00 2.94052482e-01 -1.09809589e+00 -7.01964080e-01 4.19385910e-01 2.75303185e-01 -5.54240644e-01 2.83604950e-01 -4.77502584e-01 8.54992270e-01 -5.76407798e-02 2.15793386e-01 -8.34213912e-01 -3.98209482e-01 -8.59535933e-01 1.15241082e-02 -1.24912463e-01 -1.62984859e-02 9.56939280e-01 -2.67843753e-01 -8.52159142e-01 1.26452470e+00 -7.86559433e-02 -4.76667851e-01 3.49193513e-01 2.97136698e-02 4.88489680e-02 2.81274348e-01 6.27980053e-01 8.68167639e-01 4.80396748e-01 -8.50651085e-01 -9.00989115e-01 -8.92881215e-01 -5.83125167e-02 1.28891185e-01 -3.45382094e-01 -9.14560165e-03 2.55325466e-01 -3.30832005e-01 1.14863932e-01 -7.61733472e-01 -2.99177080e-01 5.47523081e-01 5.47392488e-01 4.32264239e-01 4.48994339e-01 -4.20613438e-01 1.21338332e+00 -1.95907414e+00 -1.01452783e-01 3.46773397e-03 1.07220158e-01 3.87711346e-01 1.86742023e-01 6.57455504e-01 6.11758113e-01 1.13303490e-01 -5.61198056e-01 -3.00130894e-04 -2.55819529e-01 1.50591493e-01 1.53814796e-02 8.46379697e-01 -2.00958461e-01 2.24055305e-01 -1.04327750e+00 -5.74727535e-01 8.46220553e-01 5.99647582e-01 -4.84520435e-01 -3.18505801e-02 4.33996886e-01 2.18746409e-01 -3.76713365e-01 9.87360120e-01 6.75749958e-01 2.58993655e-01 1.60456613e-01 3.70895445e-01 -1.02260876e+00 -9.80772972e-02 -9.43829775e-01 1.12327230e+00 -4.05274510e-01 7.05332577e-01 3.76519471e-01 -5.80441177e-01 1.00962555e+00 -1.32263273e-01 4.81730729e-01 -3.02245110e-01 1.70340985e-01 3.46229076e-01 -4.04717654e-01 -2.29512006e-01 5.77453077e-01 -1.06129192e-01 -5.76554202e-02 1.65010437e-01 -1.49937496e-01 -3.42127204e-01 2.23148435e-01 -7.66639858e-02 7.73163140e-01 -2.16479581e-02 7.67165363e-01 -8.44356656e-01 3.86464357e-01 3.88791114e-01 1.26556650e-01 5.61021924e-01 -5.25791287e-01 -7.71040097e-02 2.00204253e-01 -6.82873309e-01 -9.38163459e-01 -1.05657876e+00 -6.97318316e-01 8.49228323e-01 3.48759055e-01 -2.70493925e-01 -4.27052110e-01 1.08303837e-01 4.07211661e-01 4.15062904e-01 -5.37522972e-01 2.49615498e-02 -2.25524887e-01 -9.30100799e-01 5.24717927e-01 4.52740520e-01 6.18620098e-01 -8.09832454e-01 -1.31471741e+00 4.79076244e-02 -4.17280868e-02 -5.91458559e-01 -9.31924060e-02 2.37011448e-01 -1.17259645e+00 -9.94511306e-01 -5.48333406e-01 -2.63793766e-01 6.80121601e-01 1.00859308e+00 6.38546526e-01 7.35269040e-02 -6.01728380e-01 2.43814811e-01 -3.95790130e-01 -3.15347016e-01 -4.05326784e-02 1.51757682e-02 3.25291157e-01 -4.26185787e-01 8.99556041e-01 -9.24352646e-01 -8.10795605e-01 5.41846752e-01 -7.73496151e-01 -9.86593068e-02 6.25837326e-01 3.52497339e-01 6.58508599e-01 -2.50948668e-01 3.44904512e-02 -3.21510434e-01 2.75848862e-02 -5.28777122e-01 -1.25086284e+00 -1.67893350e-01 -4.39709127e-02 -1.58865258e-01 4.29574311e-01 1.66425165e-02 -2.98131585e-01 4.00048822e-01 5.03282696e-02 7.49569461e-02 -4.01358694e-01 2.53776282e-01 3.89100075e-01 -1.54378965e-01 8.23581159e-01 1.55157149e-01 2.38079697e-01 -6.28422201e-01 -3.81544501e-01 9.25421059e-01 4.75452363e-01 -1.81630179e-01 6.45560324e-01 9.56982732e-01 2.43782878e-01 -1.38811338e+00 -1.81587189e-01 -1.02296436e+00 -8.57853591e-01 -5.51329494e-01 6.41694605e-01 -1.12984884e+00 -1.05043888e+00 3.75657767e-01 -7.09265590e-01 -3.21308851e-01 1.51498824e-01 7.37507224e-01 -1.29648075e-01 1.59820631e-01 -4.21500765e-02 -1.04301095e+00 -4.03587878e-01 -1.09107769e+00 1.24508941e+00 2.29899019e-01 -2.53799427e-02 -8.03234696e-01 3.77654642e-01 2.41168037e-01 5.32981932e-01 5.08719742e-01 9.74509418e-02 2.56319374e-01 -4.29955631e-01 -6.31628990e-01 -3.23138207e-01 -5.57366535e-02 2.10024774e-01 2.34647840e-01 -5.69071352e-01 -2.46781409e-01 -1.52245119e-01 2.52583921e-01 7.61258841e-01 5.43701947e-01 6.42354071e-01 -2.16255888e-01 -5.29015720e-01 5.67610919e-01 1.72896707e+00 1.27382830e-01 2.57845312e-01 4.42402840e-01 1.48310676e-01 5.35473406e-01 4.18768018e-01 7.84869730e-01 2.52061844e-01 7.58358479e-01 7.67739356e-01 2.28776321e-01 2.77954698e-01 -4.20146495e-01 5.30575037e-01 1.71368465e-01 -2.04947725e-01 2.90705115e-01 -1.22292817e+00 8.08057785e-01 -1.36349833e+00 -1.17751408e+00 -4.79699463e-01 2.63473225e+00 2.69028395e-01 -2.86329061e-01 4.61706400e-01 -1.00506142e-01 9.45732355e-01 -1.95603743e-02 -3.12268704e-01 -3.65071267e-01 1.78238645e-01 4.88948561e-02 1.29887915e+00 7.90090501e-01 -9.53300714e-01 7.94364274e-01 6.37649155e+00 4.40312088e-01 -1.09962344e+00 -2.87276417e-01 -1.13599570e-02 -3.07282329e-01 1.51203871e-01 3.11455466e-02 -1.18405950e+00 3.95470828e-01 6.54064000e-01 -1.57026380e-01 5.24336576e-01 8.21925461e-01 6.77776992e-01 -8.30346107e-01 -4.33780491e-01 8.81603658e-01 -2.74717826e-02 -9.49766517e-01 -5.42279519e-02 5.00269532e-01 1.24507688e-01 2.73333907e-01 -2.57174134e-01 -1.34216174e-01 3.07650506e-01 -4.86697644e-01 6.39531314e-01 2.63651051e-02 8.63384366e-01 -5.65444708e-01 8.00403059e-01 4.25615609e-01 -1.48204732e+00 8.50676186e-03 -6.47154629e-01 -5.93060076e-01 1.82984751e-02 1.85780928e-01 -7.84197390e-01 -1.35892749e-01 1.12792504e+00 8.10212433e-01 -8.21007967e-01 1.20989418e+00 -5.25468364e-02 3.86075675e-01 -8.90844762e-01 -3.64203125e-01 3.10298771e-01 -4.67235267e-01 5.82712114e-01 1.17092133e+00 4.78556216e-01 1.54834732e-01 -1.61902875e-01 7.16436684e-01 1.81287616e-01 1.03040077e-01 -8.51122022e-01 -8.01464766e-02 6.95580065e-01 1.19273806e+00 -9.88780618e-01 -4.09719646e-02 -1.85574099e-01 4.35099214e-01 -1.95110552e-02 -3.86750668e-01 -6.08880877e-01 -2.52939135e-01 5.51435173e-01 5.95362842e-01 1.99238777e-01 -5.95353365e-01 -1.09288171e-01 -9.46540058e-01 -1.81980312e-01 -8.61938149e-02 3.08935434e-01 -4.47752863e-01 -5.45230508e-01 2.40764946e-01 6.41064107e-01 -1.46418715e+00 5.62203154e-02 -6.46241486e-01 -2.35789150e-01 5.89872062e-01 -1.19009745e+00 -9.35461164e-01 -4.19150561e-01 1.98467940e-01 2.90677100e-01 1.11122787e-01 7.60884166e-01 1.06458671e-01 -2.98587769e-01 -7.42417499e-02 5.32071710e-01 -2.19793409e-01 1.01389721e-01 -9.39862072e-01 1.01672500e-01 6.84195101e-01 -1.55443206e-01 3.37838799e-01 6.30250096e-01 -8.34634423e-01 -1.27906358e+00 -8.37528050e-01 6.85667455e-01 -2.89438009e-01 8.45569015e-01 -4.65350986e-01 -2.90944546e-01 4.55675244e-01 -3.99006158e-01 -3.41762722e-01 8.68689120e-01 -1.77431807e-01 9.14722979e-02 -6.30689979e-01 -1.40338206e+00 5.37695944e-01 9.01684284e-01 -2.36799642e-01 -1.82245165e-01 4.01532613e-02 -1.11388832e-01 6.20014071e-02 -7.46164143e-01 1.74784958e-01 1.14387035e+00 -9.90911424e-01 9.48647320e-01 2.31956825e-01 3.21038932e-01 -5.10957479e-01 -4.89931852e-01 -7.72094727e-01 -2.66505659e-01 2.68286824e-01 7.51758158e-01 7.70624518e-01 1.76721334e-01 -5.37010670e-01 5.32159328e-01 1.87144637e-01 1.65275782e-01 1.72479913e-01 -9.26340103e-01 -8.16267788e-01 -1.27888486e-01 9.81831849e-02 2.52159536e-01 2.85629928e-01 4.96626832e-02 -1.80084884e-01 -2.30460122e-01 4.39061105e-01 9.41441476e-01 -1.82710569e-02 8.91887903e-01 -1.30478168e+00 1.19222857e-01 -4.16728705e-01 -8.58091652e-01 -4.73381042e-01 -2.57366210e-01 -5.09619176e-01 -1.36637703e-01 -1.29970598e+00 3.99851292e-01 -6.07956983e-02 4.07595366e-01 4.82223094e-01 2.24841014e-01 5.92579305e-01 -8.92894343e-02 5.10572374e-01 -2.05899209e-01 3.38123679e-01 5.29019833e-01 5.49966916e-02 -2.86119431e-01 3.92658077e-02 -3.88988674e-01 5.00659406e-01 9.72609401e-01 -6.68042421e-01 1.21885858e-01 -6.24274433e-01 6.07478730e-02 2.32864451e-02 9.70096886e-01 -1.28685760e+00 4.76529710e-02 -3.52102935e-01 2.04319015e-01 -7.95053601e-01 4.25216943e-01 -1.27488768e+00 5.60610473e-01 1.12446332e+00 1.01412818e-01 5.95566351e-03 5.25437772e-01 5.33721745e-01 1.04791112e-01 -4.42649871e-01 9.84061301e-01 -5.83908021e-01 -5.20695567e-01 6.45283759e-02 -9.82531607e-01 -5.55809617e-01 8.83074045e-01 -4.59596038e-01 -2.54663914e-01 1.25229703e-02 -1.01481616e-01 1.43708229e-01 1.00266504e+00 -1.13658477e-02 3.17351609e-01 -8.13417077e-01 -6.50367081e-01 1.87329903e-01 2.53773123e-01 -1.58469215e-01 2.21564859e-01 6.77522898e-01 -1.61272752e+00 4.33408409e-01 -7.35299706e-01 -7.29308248e-01 -1.34485006e+00 1.89919591e-01 2.53850758e-01 2.66203165e-01 -2.29470626e-01 5.55083513e-01 -3.32726240e-02 -1.45581976e-01 -1.87173933e-02 -3.31076905e-02 -1.29471064e-01 4.44385875e-03 8.00297141e-01 5.12702167e-01 -3.04777384e-01 -9.19217885e-01 -6.59316897e-01 5.79841971e-01 3.10061723e-01 -8.30380321e-02 1.66079307e+00 -4.31049705e-01 -2.98529983e-01 4.24784213e-01 8.83512199e-01 -1.56141788e-01 -1.13378119e+00 1.08688593e-01 -9.09513161e-02 -9.46339488e-01 3.46307755e-01 -2.79475391e-01 -8.41949046e-01 9.76727068e-01 8.13407958e-01 2.28756741e-01 6.78753078e-01 -1.10514462e-01 1.18813559e-01 2.97068626e-01 7.88793981e-01 -7.53545761e-01 -6.33581519e-01 -1.16156355e-01 8.32068443e-01 -1.26974809e+00 4.16887850e-01 1.85643554e-01 -2.11412966e-01 9.15060878e-01 2.90285528e-01 -1.28737569e-01 5.64514518e-01 1.80653304e-01 -3.31868798e-01 -1.48284748e-01 -8.04589391e-02 -4.43569988e-01 -5.88146865e-01 7.68541098e-01 2.89472014e-01 4.71127152e-01 -5.23306251e-01 -4.69850861e-02 -2.10877672e-01 -1.55526444e-01 5.53301871e-01 9.93671894e-01 -9.07340467e-01 -7.30263054e-01 -6.52844787e-01 4.43649024e-01 -2.42560431e-01 -8.53030905e-02 -6.80803418e-01 6.61759198e-01 -8.09701383e-02 8.88777196e-01 1.05070353e-01 -1.98570997e-01 2.04086244e-01 -4.06597197e-01 1.61968917e-01 -3.92893046e-01 -5.82189977e-01 -1.38954625e-01 -1.99376151e-01 -1.93465978e-01 -7.32166231e-01 -9.75452006e-01 -5.08155346e-01 -9.13467705e-01 -2.73647755e-01 3.05981636e-01 1.21437550e+00 2.01776549e-01 1.19611286e-01 -5.05047679e-01 7.05364168e-01 -1.12570143e+00 7.71896616e-02 -7.55121469e-01 -8.96307349e-01 -3.25415134e-01 8.67154300e-02 -7.36350775e-01 -6.69418573e-01 -9.05002877e-02]
[8.612133979797363, -1.1187125444412231]
d7422b6c-4faf-4fa4-b5f5-0e76d904beb0
word-embedding-for-response-to-text-1
1908.01969
null
https://arxiv.org/abs/1908.01969v1
https://arxiv.org/pdf/1908.01969v1.pdf
Word Embedding for Response-To-Text Assessment of Evidence
Manually grading the Response to Text Assessment (RTA) is labor intensive. Therefore, an automatic method is being developed for scoring analytical writing when the RTA is administered in large numbers of classrooms. Our long-term goal is to also use this scoring method to provide formative feedback to students and teachers about students' writing quality. As a first step towards this goal, interpretable features for automatically scoring the evidence rubric of the RTA have been developed. In this paper, we present a simple but promising method for improving evidence scoring by employing the word embedding model. We evaluate our method on corpora of responses written by upper elementary students.
['Haoran Zhang', 'Diane Litman']
2019-08-06
word-embedding-for-response-to-text
https://aclanthology.org/P17-3013
https://aclanthology.org/P17-3013.pdf
acl-2017-7
['automated-essay-scoring']
['natural-language-processing']
[ 1.03861451e-01 1.95951208e-01 -1.07690096e-01 -5.85579574e-01 -9.09723580e-01 -6.24853253e-01 5.12720406e-01 7.54725695e-01 -4.07438576e-01 5.74527442e-01 4.59107667e-01 -6.29024506e-01 -3.30923527e-01 -7.06581593e-01 -2.02833802e-01 -7.14736581e-02 6.93541765e-01 2.34307379e-01 4.55280930e-01 -2.57834822e-01 1.03263330e+00 5.11548936e-01 -1.51719153e+00 2.97273129e-01 1.15997910e+00 2.76243091e-01 2.08176672e-01 1.04500246e+00 -5.16353667e-01 1.19826484e+00 -8.85826468e-01 -8.08947384e-01 -3.97053361e-01 -6.38228655e-01 -1.12160861e+00 -1.77103896e-02 8.27834189e-01 -5.45121491e-01 1.28250942e-01 1.03957522e+00 4.24250513e-01 2.95132875e-01 8.73359621e-01 -8.88357520e-01 -7.64877439e-01 7.86016583e-01 -2.31363043e-01 4.15980965e-01 7.31725931e-01 -2.60935128e-01 1.16258347e+00 -9.70157623e-01 3.98881584e-01 8.63638103e-01 4.09562409e-01 6.76864684e-01 -1.00769699e+00 -5.43842554e-01 -1.25602916e-01 4.71826106e-01 -8.41827691e-01 -3.27981025e-01 8.61687303e-01 -8.04748118e-01 7.74264336e-01 2.67861396e-01 8.58867526e-01 6.86445534e-01 1.53146729e-01 7.37760723e-01 1.55050063e+00 -1.05626464e+00 2.18839601e-01 7.11445332e-01 8.09015036e-01 8.18623900e-01 4.08046722e-01 -5.70000708e-01 -8.55933726e-01 -6.09155092e-03 1.98461920e-01 -4.31501061e-01 4.94868159e-02 -3.06892823e-02 -5.32674909e-01 9.41738784e-01 -3.61898541e-01 5.89419544e-01 -1.94108099e-01 -2.80992966e-02 3.40300530e-01 5.60719669e-01 5.51357329e-01 7.14320302e-01 -3.73075634e-01 -8.93658936e-01 -9.31573510e-01 2.89568841e-01 8.86635840e-01 5.32933831e-01 1.72201917e-01 1.07227704e-02 -1.97402790e-01 1.17621398e+00 6.76886618e-01 2.60451943e-01 8.30798268e-01 -9.18383241e-01 4.69275385e-01 9.49569702e-01 6.60097450e-02 -9.07435000e-01 1.04618803e-01 5.50234206e-02 1.34748012e-01 6.40842676e-01 5.13728440e-01 2.92937532e-02 -6.32989585e-01 1.31846845e+00 4.25619632e-01 -4.69594687e-01 3.41574699e-02 5.07708013e-01 9.95161891e-01 5.62644184e-01 6.09848015e-02 -1.69723272e-01 1.42245245e+00 -7.86346376e-01 -1.09197092e+00 3.56373668e-01 9.82798457e-01 -1.23420775e+00 1.46981001e+00 8.72682631e-01 -1.51694334e+00 -5.77717602e-01 -1.15335381e+00 -2.17825681e-01 -5.28623879e-01 5.47240973e-02 1.70002744e-01 1.19802654e+00 -8.64114344e-01 5.52165210e-01 -5.58026075e-01 -1.84822336e-01 2.16936529e-01 2.15020448e-01 -2.97737002e-01 2.18568865e-04 -7.97861934e-01 1.19096434e+00 -1.87445581e-01 -3.21804017e-01 -2.88076401e-01 -4.89668280e-01 -8.20564389e-01 2.26293743e-01 4.49783029e-03 -3.36966179e-02 1.74138868e+00 -3.63564938e-01 -2.02110934e+00 9.34942424e-01 -4.35090959e-02 2.91670322e-01 1.35067716e-01 -2.68603384e-01 -1.92112908e-01 2.31083974e-01 1.15180135e-01 2.12065987e-02 4.03121322e-01 -7.84472644e-01 -4.81231153e-01 -3.19817692e-01 -7.27153644e-02 2.08331332e-01 -1.13078153e+00 6.97929680e-01 1.04038253e-01 -6.69098258e-01 -3.28409043e-03 -6.33977711e-01 1.84267640e-01 -2.92298704e-01 3.78182679e-01 -1.16049170e+00 5.23602605e-01 -8.02094698e-01 1.74866164e+00 -1.74045336e+00 -1.77929714e-01 3.70172352e-01 3.01769465e-01 6.50863469e-01 -7.53336251e-02 6.56722605e-01 5.03934845e-02 2.74655700e-01 1.80620715e-01 -2.45006215e-02 2.87389338e-01 7.85887316e-02 -1.96113557e-01 -1.27150625e-01 1.02781825e-01 5.26973426e-01 -1.13278699e+00 -9.53192949e-01 1.77738860e-01 1.18295193e-01 -4.15797710e-01 6.04789078e-01 1.52834043e-01 -3.58789623e-01 -5.87013900e-01 4.28289592e-01 9.33395028e-02 -9.38304737e-02 6.46538064e-02 6.14009500e-01 -4.82028574e-01 1.02508402e+00 -1.19264650e+00 1.15597975e+00 -5.72918713e-01 9.59577143e-01 -4.36467618e-01 -7.22588122e-01 1.04778636e+00 3.82234573e-01 1.95743769e-01 -5.29256105e-01 1.37201846e-01 2.76984543e-01 2.03709051e-01 -9.88711536e-01 8.44300747e-01 -1.02136366e-01 1.42202929e-01 1.05898094e+00 2.57358819e-01 -6.65978134e-01 7.19130635e-01 3.97626847e-01 1.22794771e+00 1.68572456e-01 4.64250445e-01 -1.89922497e-01 6.17196739e-01 -6.45866320e-02 1.76352989e-02 5.05527675e-01 -3.01601171e-01 2.79103607e-01 3.37369770e-01 -3.12574089e-01 -7.41093278e-01 -6.35463715e-01 -8.69603083e-02 1.30843866e+00 -7.48910546e-01 -6.10885561e-01 -8.68707299e-01 -7.94677138e-01 -3.02691638e-01 8.12142015e-01 -3.64080459e-01 -3.30020813e-03 -5.55507779e-01 -1.55471563e-01 2.15877131e-01 6.91883743e-01 -9.22230631e-02 -1.04454076e+00 -7.33783066e-01 5.24249494e-01 -3.14253196e-02 -6.99720502e-01 -2.52861381e-01 2.69459963e-01 -7.67354488e-01 -9.22973394e-01 -5.57452142e-01 -7.70702362e-01 6.75940812e-01 3.71498704e-01 9.09845948e-01 6.32698655e-01 4.85275798e-02 7.45491624e-01 -6.14007056e-01 -7.91647077e-01 -6.86419427e-01 -1.10878527e-01 -5.78973219e-02 -6.51377618e-01 8.42260480e-01 -3.71234357e-01 -1.31969750e-01 -3.36192138e-02 -9.32348967e-01 -1.02747619e-01 4.19657588e-01 5.49674571e-01 6.00976646e-02 -2.21733481e-01 6.02222919e-01 -8.54476631e-01 1.40657818e+00 -2.76926272e-02 -7.49336123e-01 3.50059152e-01 -1.13845992e+00 9.89463627e-02 6.56348884e-01 -5.59100151e-01 -9.60257173e-01 -4.34471726e-01 -4.72923249e-01 1.60428450e-01 -1.72507793e-01 7.56463349e-01 2.85295516e-01 -4.26763088e-01 7.85648346e-01 -5.64386286e-02 -1.14219971e-01 -3.45898598e-01 -7.48592243e-02 1.00923407e+00 2.59114623e-01 -9.71234262e-01 7.60983229e-01 -4.64896142e-01 1.10829594e-02 -9.43566144e-01 -1.18256664e+00 -6.45701170e-01 -4.75219101e-01 -7.76776969e-01 5.70501626e-01 -5.12929082e-01 -6.73567116e-01 5.23200892e-02 -1.30837595e+00 -4.84493405e-01 -2.26507097e-01 9.38582361e-01 -2.87618637e-01 4.47988331e-01 -6.03032231e-01 -8.32559705e-01 -2.00208098e-01 -9.97313738e-01 5.51945806e-01 4.19320047e-01 -7.32651472e-01 -1.04163921e+00 7.35495627e-01 1.05424631e+00 9.82204229e-02 -2.64235348e-01 9.53987241e-01 -8.88870001e-01 -1.17379330e-01 -2.56618798e-01 1.04192823e-01 7.38943577e-01 -1.56793356e-01 6.04230225e-01 -8.78504932e-01 2.75294363e-01 8.99516121e-02 -7.63815701e-01 3.13743442e-01 -9.35332999e-02 1.12248504e+00 -4.49074894e-01 4.88451064e-01 -2.37416685e-01 1.11713135e+00 2.15518519e-01 2.98986048e-01 4.97397691e-01 4.02946293e-01 9.36089873e-01 7.63606608e-01 2.08162248e-01 3.26747328e-01 5.82151592e-01 -2.15132445e-01 6.25325561e-01 -3.80822003e-01 -2.31743902e-01 7.91083872e-01 1.87915742e+00 -1.69328272e-01 -1.96344420e-01 -1.02273870e+00 7.43065357e-01 -1.41570282e+00 -1.14924872e+00 -8.21025491e-01 1.87443662e+00 1.24055254e+00 2.20346615e-01 1.18403725e-01 6.58137202e-01 5.69180176e-02 -1.60730034e-01 3.21573675e-01 -1.33405733e+00 5.93552232e-01 7.36606598e-01 -1.38905168e-01 8.27804923e-01 -2.96102554e-01 6.15473807e-01 6.54325676e+00 6.50310516e-01 -7.41158485e-01 -1.53477028e-01 1.81723371e-01 3.76677841e-01 -5.19679368e-01 -1.22262143e-01 -9.61605489e-01 3.90561372e-01 1.18960810e+00 -2.27645233e-01 -1.59060396e-02 6.93262815e-01 3.47651362e-01 -2.18392611e-01 -8.30667377e-01 2.77216613e-01 2.63632119e-01 -9.76808965e-01 -1.15711227e-01 9.42139849e-02 8.31060648e-01 -5.87543428e-01 -3.42031457e-02 3.16608936e-01 6.12588882e-01 -6.47997439e-01 5.57905614e-01 2.25235298e-01 4.68214363e-01 -6.11316383e-01 8.17836165e-01 1.84508324e-01 -6.84171259e-01 7.65363500e-02 -3.53864849e-01 -6.44777596e-01 -5.18760502e-01 2.46673673e-01 -1.37495291e+00 -1.97786037e-02 4.43982631e-01 4.15727273e-02 -9.19311762e-01 9.67022955e-01 -8.02469492e-01 1.15459955e+00 -2.65145767e-02 -9.46753085e-01 2.09231228e-01 -2.74249464e-01 1.16496004e-01 1.16445410e+00 4.27640527e-01 4.17780429e-01 -2.13039085e-01 4.87171888e-01 -1.35329187e-01 6.59044802e-01 -6.50473297e-01 -3.19433421e-01 4.44296002e-01 1.42034912e+00 -7.96401143e-01 -4.33530033e-01 -4.77839828e-01 4.32884932e-01 6.07964814e-01 -1.72063038e-01 -3.38503569e-01 -6.24167621e-01 2.18525931e-01 4.38116379e-02 2.33706273e-03 -1.50225252e-01 -6.37587011e-01 -7.64179945e-01 -5.07183559e-03 -1.00929761e+00 3.13058853e-01 -1.05639899e+00 -1.02968383e+00 -7.68368132e-03 1.40513763e-01 -9.87043321e-01 -3.57536614e-01 -9.22755599e-01 -8.91139030e-01 9.24098074e-01 -1.27890313e+00 -5.90993404e-01 -3.09818685e-01 1.80368006e-01 7.39199579e-01 -2.42316782e-01 8.55406523e-01 1.51861534e-01 -4.34852719e-01 6.68108165e-01 -8.46857727e-02 -2.98473313e-02 8.38203907e-01 -1.81382978e+00 -2.26475149e-01 9.66069818e-01 5.29489994e-01 7.12665081e-01 9.35704708e-01 -5.19660115e-01 -1.12247074e+00 -3.88747007e-01 1.74943411e+00 -6.31642878e-01 1.12177193e+00 1.10928612e-02 -1.20363700e+00 1.60431951e-01 6.09550893e-01 -6.28047943e-01 1.33713019e+00 2.47598171e-01 -2.87819564e-01 3.50544788e-02 -8.22966933e-01 6.09686255e-01 2.65027374e-01 -6.05356216e-01 -1.57528949e+00 3.21738750e-01 2.16458842e-01 -2.17556491e-01 -1.17772853e+00 -2.37894785e-02 5.75378597e-01 -5.60829222e-01 5.19220173e-01 -5.92615604e-01 1.10811138e+00 -8.37505311e-02 2.18484104e-01 -1.28838050e+00 -1.55820981e-01 -5.28960705e-01 -4.37195785e-02 1.45477974e+00 3.00618172e-01 -2.15297565e-01 8.21452379e-01 7.74177372e-01 -3.85450393e-01 -7.87180722e-01 -5.66476583e-01 -5.39644122e-01 2.28497863e-01 -3.89621884e-01 2.97038853e-01 1.17095661e+00 5.37798345e-01 4.92862493e-01 1.73518613e-01 -2.01209188e-01 3.56403321e-01 -1.16861559e-01 8.70986402e-01 -1.44303119e+00 4.70390916e-02 -7.27174401e-01 -2.33624056e-01 -5.99063873e-01 2.16520593e-01 -7.41410613e-01 3.20271440e-02 -1.49061275e+00 3.58294129e-01 4.20993418e-02 -2.77760506e-01 4.83922958e-01 -4.60002959e-01 2.31221557e-01 8.76572728e-02 -2.58411467e-01 -4.94852901e-01 1.89176515e-01 1.19183016e+00 1.02326475e-01 -7.61732310e-02 -4.70449775e-02 -7.58584023e-01 7.91065097e-01 1.09463680e+00 -7.72740483e-01 -5.11262059e-01 -2.61437148e-01 6.41792476e-01 -8.78395215e-02 -4.90998067e-02 -8.77337515e-01 4.36777562e-01 -6.30194366e-01 1.26847744e-01 -6.71188653e-01 -7.61600435e-02 -6.20504975e-01 -6.38018012e-01 8.91251937e-02 -7.19857156e-01 4.80648220e-01 7.28438050e-02 -1.01238437e-01 -3.93425703e-01 -1.35118484e+00 5.50802767e-01 1.61151186e-01 -1.56029776e-01 -4.83321518e-01 -8.91462862e-01 1.49006229e-02 6.90807760e-01 -4.32512879e-01 -3.62392813e-01 -2.70208329e-01 -1.21290140e-01 7.64577910e-02 2.01949239e-01 2.39395976e-01 9.86694455e-01 -1.30216694e+00 -5.85032821e-01 9.79222450e-03 1.73309118e-01 -3.40716958e-01 -2.22498387e-01 7.20588863e-01 -7.18389094e-01 6.04247212e-01 -3.94066453e-01 4.71729971e-02 -2.06457663e+00 6.88439310e-02 -3.11327189e-01 -5.57744205e-01 -2.21366301e-01 6.62157595e-01 -5.85897624e-01 -3.19825381e-01 2.03974545e-01 -3.85235816e-01 -8.92294765e-01 2.02662796e-01 1.00426733e+00 7.20617771e-01 3.26530457e-01 -3.67962807e-01 4.31533217e-01 3.67838323e-01 -2.01254651e-01 -4.41393822e-01 1.57988024e+00 1.48085535e-01 -7.97848627e-02 7.85926163e-01 9.97666240e-01 6.84411645e-01 -3.27899516e-01 1.54563874e-01 3.78089964e-01 -6.03454590e-01 9.30845067e-02 -9.00079608e-01 -2.47291908e-01 1.08924508e+00 3.28464270e-01 6.33403540e-01 8.08335781e-01 -3.67594212e-01 3.98531944e-01 6.23427212e-01 -8.08711722e-02 -1.65593123e+00 4.54141289e-01 5.09012520e-01 6.31824195e-01 -1.26542437e+00 3.35006297e-01 -1.10401519e-01 -3.19458276e-01 1.65433693e+00 6.12721682e-01 3.94860417e-01 9.59518105e-02 -1.72667652e-02 1.31651863e-01 -4.34042126e-01 -8.78294706e-01 7.93136731e-02 5.73415458e-01 3.26015800e-01 1.22759402e+00 5.94703034e-02 -1.16487849e+00 3.93740058e-01 -3.65052521e-01 -2.59421282e-02 1.29478705e+00 1.16289353e+00 -8.33484471e-01 -1.47310936e+00 -6.28395736e-01 5.63703358e-01 -9.33246195e-01 -1.55912936e-02 -9.64000285e-01 6.22544646e-01 -3.85383010e-01 1.24128103e+00 -5.96850276e-01 -1.80804715e-01 3.48144054e-01 6.63020849e-01 7.34858572e-01 -8.69411230e-01 -9.42108095e-01 -3.31985056e-01 2.65404344e-01 -1.06333442e-01 -6.26802087e-01 -7.08876491e-01 -9.17849243e-01 -3.46868664e-01 -5.99511266e-01 3.85637492e-01 9.22321022e-01 8.95992994e-01 -3.49525332e-01 4.67555314e-01 5.85320175e-01 -1.25333533e-01 -8.83830965e-01 -1.14835238e+00 -1.75423980e-01 2.82461822e-01 -9.27202180e-02 -4.62328017e-01 -3.28617722e-01 5.73159643e-02]
[11.285449981689453, 9.256958961486816]
409cced2-12f6-41cf-875c-f8c33061d234
camil-context-aware-multiple-instance
2305.05314
null
https://arxiv.org/abs/2305.05314v1
https://arxiv.org/pdf/2305.05314v1.pdf
CAMIL: Context-Aware Multiple Instance Learning for Whole Slide Image Classification
Cancer diagnoses typically involve human pathologists examining whole slide images (WSIs) of tissue section biopsies to identify tumor cells and their subtypes. However, artificial intelligence (AI)-based models, particularly weakly supervised approaches, have recently emerged as viable alternatives. Weakly supervised approaches often use image subsections or tiles as input, with the overall classification of the WSI based on attention scores assigned to each tile. However, this method overlooks the potential for false positives/negatives because tumors can be heterogeneous, with cancer and normal cells growing in patterns larger than a single tile. Such errors at the tile level could lead to misclassification at the tumor level. To address this limitation, we developed a novel deep learning pooling operator called CHARM (Contrastive Histopathology Attention Resolved Models). CHARM leverages the dependencies among single tiles within a WSI and imposes contextual constraints as prior knowledge to multiple instance learning models. We tested CHARM on the subtyping of non-small cell lung cancer (NSLC) and lymph node (LN) metastasis, and the results demonstrated its superiority over other state-of-the-art weakly supervised classification algorithms. Furthermore, CHARM facilitates interpretability by visualizing regions of attention.
['Chris Bakal', 'Mat De Vries', 'Chen Jin', 'Avi Arampatzis', 'Olga Fourkioti']
2023-05-09
null
null
null
null
['whole-slide-images', 'multiple-instance-learning']
['computer-vision', 'methodology']
[ 6.11347079e-01 5.17280340e-01 -6.30091548e-01 -8.30020234e-02 -1.27624309e+00 -5.04185677e-01 5.26578546e-01 6.86732888e-01 -2.41435423e-01 7.22443402e-01 2.56052405e-01 -5.06882727e-01 -6.32674322e-02 -6.92347646e-01 -5.67679048e-01 -1.34939706e+00 7.98632279e-02 5.71408451e-01 2.41793916e-01 3.40313703e-01 9.96632800e-02 5.13028145e-01 -1.19545114e+00 1.07480025e+00 7.30324388e-01 9.00949061e-01 1.42512694e-01 7.80107856e-01 -4.06925350e-01 1.18871272e+00 -4.75376070e-01 -1.13690145e-01 -1.52518958e-01 -1.01056777e-01 -9.18793440e-01 2.90078353e-02 4.18882996e-01 1.97114557e-01 6.66329861e-02 1.03718126e+00 1.56891912e-01 -5.72760165e-01 7.89482355e-01 -1.22647309e+00 -3.49722594e-01 4.02357697e-01 -7.37354934e-01 1.68837532e-01 -2.87426829e-01 3.97261322e-01 1.20353043e+00 -8.64341676e-01 7.67836273e-01 7.40265727e-01 8.04369390e-01 5.13583302e-01 -1.38670623e+00 -5.80995381e-01 8.04032832e-02 2.67958231e-02 -1.35808647e+00 -1.37503698e-01 2.76027620e-01 -4.91704524e-01 1.05808139e+00 7.20803082e-01 6.80136383e-01 8.90574396e-01 5.73466361e-01 8.71731520e-01 1.17138326e+00 -2.65108883e-01 2.78601676e-01 1.95682034e-01 3.15085143e-01 8.15390110e-01 4.56161529e-01 -3.12672764e-01 -2.78583795e-01 -2.90983975e-01 5.21345556e-01 4.93665367e-01 -1.33507594e-01 -3.76485847e-02 -1.53417635e+00 6.16440773e-01 8.04230869e-01 3.86541843e-01 -7.25827441e-02 7.03801513e-02 4.21844840e-01 -6.72334880e-02 6.24481976e-01 4.91151929e-01 -1.52526051e-01 3.25401485e-01 -8.64033759e-01 -1.91947281e-01 3.71889889e-01 3.84120435e-01 5.39734125e-01 -7.03738093e-01 -5.55467367e-01 5.80721855e-01 1.67112947e-01 -8.13528597e-02 5.40657461e-01 -3.46215278e-01 9.82626155e-02 1.25883567e+00 -1.77664548e-01 -6.33041441e-01 -6.85855210e-01 -7.61539161e-01 -1.23626900e+00 3.44885111e-01 4.80079919e-01 3.47754896e-01 -1.30454254e+00 1.38698268e+00 2.01978803e-01 2.54306823e-01 -1.01907343e-01 7.02046514e-01 8.51570666e-01 2.57367998e-01 4.70388710e-01 1.31108716e-01 1.60150599e+00 -8.44360232e-01 -6.50454521e-01 -1.19436875e-01 1.21836650e+00 -3.91303360e-01 9.66548085e-01 6.69789016e-02 -8.59430850e-01 -3.56347896e-02 -7.37398267e-01 -1.42208576e-01 -5.62037587e-01 1.48286223e-01 5.93523204e-01 2.17127666e-01 -1.16876864e+00 1.82419404e-01 -1.09926879e+00 -5.25253952e-01 1.10630655e+00 5.01171350e-01 -5.08413315e-01 1.67165287e-02 -6.22604072e-01 7.85778582e-01 3.74510661e-02 2.07208753e-01 -7.22223759e-01 -1.27403581e+00 -5.93370378e-01 1.99797317e-01 1.28635496e-01 -5.77308476e-01 9.73249495e-01 -1.05210090e+00 -8.86798501e-01 1.43068647e+00 -5.68579376e-01 -3.21115077e-01 1.91431686e-01 6.15865350e-01 -1.67953014e-01 1.36004150e-01 8.85384008e-02 7.21262693e-01 3.18097562e-01 -1.13481081e+00 -8.07686865e-01 -4.79033291e-01 -2.71663368e-01 3.70828286e-02 -2.66160637e-01 -3.21365297e-01 -2.39427015e-01 -5.58410645e-01 1.05656609e-01 -9.86910105e-01 -3.89561355e-01 5.21523356e-01 -6.77205443e-01 -1.89835444e-01 7.91395545e-01 -3.06387067e-01 1.11442029e+00 -2.23338962e+00 -1.93458814e-02 3.99664223e-01 5.94551384e-01 6.81630075e-02 -6.40186146e-02 -1.28837124e-01 -1.48336142e-01 5.66323459e-01 -2.22531140e-01 -2.61113137e-01 -1.88618213e-01 2.49313638e-01 1.89581156e-01 5.84485471e-01 6.77569926e-01 1.17787349e+00 -8.75527084e-01 -8.31622601e-01 -3.03839333e-02 1.68413326e-01 -4.85845953e-01 2.71544158e-02 -2.30265647e-01 1.74405932e-01 -1.56465381e-01 1.04675865e+00 3.70449990e-01 -9.85100925e-01 3.02070707e-01 -2.54451722e-01 1.51926756e-01 1.07734054e-01 -5.13878942e-01 1.16575611e+00 -2.06285641e-01 6.73646271e-01 2.28606761e-01 -9.15448368e-01 3.81565541e-01 3.65364790e-01 5.91753185e-01 -3.43329668e-01 -1.19653970e-01 1.66895926e-01 2.99822837e-01 -6.24156952e-01 9.97056253e-03 -3.53247643e-01 1.70963451e-01 4.04055923e-01 -2.18707174e-01 1.06304690e-01 -9.75224525e-02 1.27631381e-01 1.64636779e+00 -5.40441453e-01 5.61743021e-01 -5.16375959e-01 2.19912305e-01 4.30800825e-01 6.64482772e-01 7.64764905e-01 -4.24332857e-01 8.27015936e-01 8.91596496e-01 -6.07223213e-01 -8.70215535e-01 -1.16446602e+00 -4.95425224e-01 9.00122285e-01 -4.70660329e-02 -4.81009930e-02 -4.02605057e-01 -9.20522034e-01 1.68199554e-01 1.67200699e-01 -1.25774372e+00 -3.74410003e-02 -3.16523314e-01 -1.23256779e+00 5.06954610e-01 6.45897508e-01 1.99208766e-01 -9.24913347e-01 -3.43051165e-01 1.30444199e-01 -1.54106364e-01 -7.80941486e-01 -1.49065554e-01 7.95186758e-01 -7.09835231e-01 -1.45998502e+00 -6.51326776e-01 -1.05006373e+00 1.38624215e+00 2.00162977e-01 1.18343484e+00 4.78488058e-01 -9.49129224e-01 -5.04458770e-02 -1.02982551e-01 -6.59608841e-01 -4.97047901e-01 2.74756044e-01 -3.95973861e-01 1.44468084e-01 6.65220559e-01 5.92269674e-02 -6.69274807e-01 2.82265246e-01 -1.04256570e+00 4.20330733e-01 9.33966219e-01 1.14898956e+00 1.03985643e+00 -3.58810648e-02 3.94082457e-01 -1.42765737e+00 1.59025311e-01 -7.33189344e-01 -6.43159524e-02 4.10902053e-01 -1.70470849e-01 -9.96588245e-02 5.43439388e-01 -4.21771526e-01 -8.54387820e-01 1.77448317e-02 1.13587990e-01 3.50829139e-02 -2.98752785e-01 5.68569481e-01 6.67088339e-03 -4.85062785e-02 7.30022371e-01 -6.00492433e-02 2.23172903e-01 1.81935534e-01 -5.34023046e-01 7.18442559e-01 3.99244040e-01 -1.70413956e-01 4.70547467e-01 9.08069193e-01 1.88838854e-01 -7.04531193e-01 -9.91416752e-01 -5.01544952e-01 -4.43807483e-01 -2.11573660e-01 1.03748453e+00 -7.76696503e-01 -8.12828243e-01 2.19431669e-01 -7.80533552e-01 -7.17523098e-01 -3.12506258e-01 3.94608408e-01 -9.99259129e-02 -2.25485206e-01 -9.28286731e-01 -4.12562609e-01 -1.40862554e-01 -1.11817718e+00 1.37397075e+00 5.64427674e-02 -7.17965305e-01 -1.05797684e+00 1.46760158e-02 3.45423013e-01 4.55691487e-01 5.26482284e-01 1.58304453e+00 -7.25227296e-01 -7.07592070e-01 -5.28628588e-01 -4.49933976e-01 -2.30792657e-01 2.68076003e-01 3.59462082e-01 -1.14898157e+00 -2.35274479e-01 -4.82779354e-01 -2.90470958e-01 9.13766384e-01 4.95742977e-01 1.58086956e+00 -3.91587228e-01 -9.44040596e-01 6.97610557e-01 1.42672920e+00 -5.58543801e-02 5.41669011e-01 4.34065461e-01 5.93209386e-01 5.62686443e-01 1.63625240e-01 -2.51919427e-03 1.23162150e-01 2.78739959e-01 6.09729052e-01 -8.27842832e-01 -1.00622796e-01 1.18104219e-01 -6.78364933e-02 2.74347126e-01 4.98167314e-02 -3.40480715e-01 -1.18488944e+00 5.91084540e-01 -1.63848746e+00 -8.86800647e-01 -2.10854962e-01 1.90500486e+00 8.72511744e-01 1.92247525e-01 -4.56148863e-01 1.53684676e-01 6.90972328e-01 1.33585520e-02 -7.11929739e-01 -1.84353232e-01 -1.53556123e-01 9.60769206e-02 3.81695896e-01 2.50002801e-01 -9.55588996e-01 4.36099887e-01 6.26739311e+00 8.11643362e-01 -1.14420247e+00 2.33586747e-02 1.36627817e+00 -4.08671319e-01 -4.31592524e-01 -3.45779181e-01 -6.22101665e-01 3.19665611e-01 5.01857817e-01 2.31775001e-01 -1.50604278e-01 4.48274791e-01 9.87962037e-02 -2.61852831e-01 -1.34255290e+00 6.58039629e-01 -1.59237370e-01 -1.67906535e+00 1.84816331e-01 3.92820925e-01 9.00264025e-01 7.93608353e-02 1.79384649e-01 1.70547992e-01 4.15269017e-01 -1.33215129e+00 6.77434579e-02 4.97655988e-01 9.09261644e-01 -4.60378557e-01 1.33838892e+00 1.36297435e-01 -9.13652718e-01 -1.69607345e-02 -2.35978276e-01 2.39672780e-01 -5.62922716e-01 6.27280831e-01 -1.29048336e+00 -2.50858963e-02 6.51410758e-01 5.33113718e-01 -8.06167662e-01 7.80814230e-01 2.21364141e-01 7.32393801e-01 -1.53082207e-01 -1.45957857e-01 4.31502610e-01 5.61907172e-01 1.28418744e-01 1.42742109e+00 2.11942315e-01 1.42458215e-01 3.48005481e-02 7.68027723e-01 -7.38982707e-02 -1.89291373e-01 -1.95309043e-01 -1.17781190e-02 1.46234706e-01 1.56145585e+00 -1.12798870e+00 -4.07933176e-01 -5.43149531e-01 5.16651094e-01 4.89791185e-01 4.50904280e-01 -6.17792666e-01 -6.09136233e-03 7.81827807e-01 4.64934975e-01 -1.02637112e-01 5.17103434e-01 -8.66563141e-01 -8.65415454e-01 -2.31480762e-01 -6.52822793e-01 7.03891814e-01 -5.31377316e-01 -1.45833659e+00 3.98649484e-01 -5.79294562e-01 -1.36960948e+00 3.26224118e-01 -7.13551700e-01 -8.57640028e-01 6.04946256e-01 -1.63293326e+00 -1.17123783e+00 -6.11332715e-01 2.09240705e-01 3.94071162e-01 7.73488879e-02 1.08698916e+00 -2.48275414e-01 -7.68150270e-01 7.16250896e-01 -2.73143351e-02 1.97715327e-01 6.77353978e-01 -1.52385461e+00 -1.47266001e-01 2.56318063e-01 -4.36852217e-01 4.69806880e-01 3.28146040e-01 -4.65831876e-01 -1.06363642e+00 -1.45517409e+00 8.27210248e-01 -3.36651206e-01 6.83721066e-01 -4.07179087e-01 -1.16070175e+00 6.17462635e-01 -1.04478002e-02 6.69324338e-01 1.38252389e+00 -3.31979394e-02 -1.82376564e-01 3.73434685e-02 -1.25778043e+00 8.27040970e-01 7.19479561e-01 -5.06567776e-01 1.34533364e-02 4.89450395e-01 2.75345504e-01 -2.92391747e-01 -9.79049325e-01 5.67971408e-01 5.05040526e-01 -7.78555810e-01 7.49420226e-01 -7.05417991e-01 6.65913105e-01 -4.24430490e-01 2.25530148e-01 -1.09185898e+00 -8.11267138e-01 1.02078952e-01 1.67797580e-01 8.32820475e-01 8.98325503e-01 -6.08448625e-01 1.24072433e+00 7.93552458e-01 -3.07747751e-01 -1.09957480e+00 -9.57423687e-01 -2.95050293e-01 5.00388779e-02 -6.82982430e-02 4.90565479e-01 9.73018527e-01 5.34964144e-01 -1.91709667e-01 3.48018885e-01 3.27729106e-01 4.30225819e-01 -3.57122975e-03 4.56360102e-01 -1.22429609e+00 -1.80904791e-01 -7.63133347e-01 -7.09892094e-01 -1.75980330e-01 7.91702569e-02 -1.31862414e+00 8.75399411e-02 -1.54227722e+00 7.92606950e-01 -4.25382435e-01 -6.42763734e-01 9.08791363e-01 -5.11211932e-01 6.62952304e-01 -2.50158101e-01 4.80452508e-01 -6.80316150e-01 -1.44930169e-01 1.24062622e+00 -6.46136165e-01 1.61782220e-01 -1.12020038e-01 -8.54227304e-01 7.94902802e-01 7.41657734e-01 -4.95379865e-01 7.96969384e-02 -2.76225775e-01 1.58839285e-01 -3.34057435e-02 6.14536464e-01 -8.94689739e-01 4.60049033e-01 -1.93976343e-01 9.23815489e-01 -5.86383879e-01 1.11243166e-02 -8.34776580e-01 1.99634790e-01 7.14446247e-01 -7.87190855e-01 -2.84395248e-01 2.16427505e-01 6.06704473e-01 -3.32257271e-01 8.55070502e-02 8.21763754e-01 -3.27919841e-01 -1.86166495e-01 4.03644949e-01 -7.39040673e-01 -3.72482926e-01 1.13885844e+00 -4.98581141e-01 -7.45816469e-01 1.73664108e-01 -6.62119389e-01 2.77493894e-01 5.59628308e-01 -4.25099246e-02 3.53738517e-01 -1.05816948e+00 -7.86938429e-01 2.81516820e-01 7.48026788e-01 4.16770846e-01 4.02761519e-01 1.17878103e+00 -6.41690433e-01 4.52269912e-01 1.24693066e-01 -9.36780810e-01 -1.38119161e+00 3.63754153e-01 5.12598395e-01 -8.24711442e-01 -4.02066946e-01 9.67214465e-01 8.16673100e-01 -3.87031019e-01 3.65590423e-01 -5.62517464e-01 1.45041002e-02 -2.01109588e-01 5.79504073e-01 1.62845422e-02 2.74522811e-01 -2.23525628e-01 -4.44967240e-01 1.47165745e-01 -5.71061075e-01 3.76770377e-01 1.14699364e+00 3.49431485e-01 -3.62323225e-01 4.64718163e-01 1.31544971e+00 -2.92540193e-01 -1.21265399e+00 -3.08034360e-01 1.02383316e-01 -6.65427968e-02 6.95722476e-02 -9.56590414e-01 -1.02549922e+00 6.57891214e-01 4.77945387e-01 1.37480304e-01 9.19777930e-01 2.01623142e-01 4.25245464e-01 9.40230116e-02 -5.42249084e-02 -7.83142447e-01 -1.19119115e-01 1.17472917e-01 4.51083690e-01 -1.49423683e+00 -7.12796375e-02 -4.44941252e-01 -4.06547695e-01 1.26276731e+00 7.28186846e-01 4.70926501e-02 5.29634953e-01 8.90066922e-01 1.99172065e-01 -3.40400100e-01 -1.24640822e+00 -4.65846844e-02 9.77317616e-02 4.51792657e-01 7.38039970e-01 3.26639205e-01 3.81006077e-02 5.99158585e-01 2.04836234e-01 2.19854601e-02 2.61417031e-01 9.86472070e-01 -4.34082657e-01 -5.97639382e-01 -4.50072289e-01 1.12666214e+00 -6.43380225e-01 -2.03085572e-01 -6.49288476e-01 7.98496842e-01 9.66418311e-02 5.66289306e-01 5.47067761e-01 -1.64174080e-01 -1.06594726e-01 -5.82701564e-02 7.92294890e-02 -7.69004047e-01 -9.64329839e-01 6.02797680e-02 -2.87523896e-01 -2.62821853e-01 -2.27430016e-01 -4.71730173e-01 -1.35860336e+00 2.29273625e-02 -1.49113685e-01 3.50447521e-02 2.50308722e-01 8.53441775e-01 3.62314284e-01 8.83596063e-01 3.97339433e-01 -5.87410748e-01 -9.13943746e-04 -8.73371601e-01 -6.33203030e-01 4.68484044e-01 6.12975061e-01 -2.44535580e-01 -5.41733205e-01 1.77482918e-01]
[15.103022575378418, -2.890406608581543]
73226b40-a692-4bc2-a19d-647af110a2a8
temporal-consistency-loss-for-high-resolution
2104.09259
null
https://arxiv.org/abs/2104.09259v1
https://arxiv.org/pdf/2104.09259v1.pdf
Temporal Consistency Loss for High Resolution Textured and Clothed 3DHuman Reconstruction from Monocular Video
We present a novel method to learn temporally consistent 3D reconstruction of clothed people from a monocular video. Recent methods for 3D human reconstruction from monocular video using volumetric, implicit or parametric human shape models, produce per frame reconstructions giving temporally inconsistent output and limited performance when applied to video. In this paper, we introduce an approach to learn temporally consistent features for textured reconstruction of clothed 3D human sequences from monocular video by proposing two advances: a novel temporal consistency loss function; and hybrid representation learning for implicit 3D reconstruction from 2D images and coarse 3D geometry. The proposed advances improve the temporal consistency and accuracy of both the 3D reconstruction and texture prediction from a monocular video. Comprehensive comparative performance evaluation on images of people demonstrates that the proposed method significantly outperforms the state-of-the-art learning-based single image 3D human shape estimation approaches achieving significant improvement of reconstruction accuracy, completeness, quality and temporal consistency.
['Adrian Hilton', 'Armin Mustafa', 'Akin Caliskan']
2021-04-19
null
null
null
null
['3d-human-reconstruction']
['computer-vision']
[-1.23801611e-01 -5.17948925e-01 -1.23593271e-01 -1.71139762e-01 -3.19521517e-01 -2.23535225e-01 3.52015018e-01 -6.15603149e-01 -1.07240617e-01 8.02353203e-01 1.50047183e-01 4.80756730e-01 1.58279851e-01 -3.37262303e-01 -8.21768939e-01 -5.06844282e-01 -1.71035767e-01 6.60559237e-01 2.75125474e-01 2.23831460e-01 -7.39221945e-02 6.60183728e-01 -1.46158791e+00 2.72447407e-01 3.01449257e-03 1.29211485e+00 1.51610419e-01 7.91541576e-01 1.78609595e-01 8.99123192e-01 2.10614260e-02 -4.11265865e-02 4.77145344e-01 -3.67124766e-01 -6.46580279e-01 6.96032584e-01 8.65396202e-01 -8.10282409e-01 -8.32500994e-01 3.75716835e-01 6.18417680e-01 1.31863594e-01 7.08649695e-01 -8.14187825e-01 -5.62829971e-01 -4.26403552e-01 -5.05710602e-01 -8.54284391e-02 1.16444862e+00 1.83078513e-01 2.32874095e-01 -1.11300457e+00 1.14483321e+00 1.51066792e+00 1.29947579e+00 6.99840367e-01 -1.31822836e+00 -1.73995242e-01 2.68444475e-02 2.82835811e-01 -1.38748205e+00 -4.44764465e-01 7.55403757e-01 -5.36183119e-01 9.65585887e-01 2.38868162e-01 1.33685648e+00 1.14383948e+00 4.39713627e-01 9.59546268e-01 1.31756556e+00 -3.47159237e-01 -7.58306161e-02 -1.74193144e-01 -4.79154497e-01 1.30608606e+00 1.48574179e-02 7.50073612e-01 -7.87796557e-01 -2.41439819e-01 1.59862697e+00 1.27019331e-01 -2.23801196e-01 -8.90069962e-01 -1.25385332e+00 2.28365377e-01 -4.13878039e-02 -2.00133789e-02 -5.68966150e-01 4.81336355e-01 4.52365726e-01 4.42751080e-01 6.97572947e-01 -3.94760996e-01 -4.47785020e-01 -6.59706518e-02 -9.90489960e-01 6.67249024e-01 6.28945589e-01 1.33472478e+00 4.08945352e-01 4.20468420e-01 -7.18521699e-02 5.18802524e-01 3.40049654e-01 8.53354692e-01 9.15045962e-02 -1.28191781e+00 -1.21065713e-01 3.57079297e-01 4.96755123e-01 -9.80350554e-01 -4.70393509e-01 3.29869576e-02 -8.32124174e-01 2.52054036e-01 4.27572638e-01 2.17975929e-01 -7.49209404e-01 1.18348205e+00 6.46643460e-01 2.93117374e-01 -3.36584806e-01 1.48506546e+00 1.22578418e+00 5.30015707e-01 -2.33893439e-01 -3.72308254e-01 8.14910889e-01 -9.23685849e-01 -7.64096558e-01 2.65371084e-01 -7.58689120e-02 -1.02795112e+00 5.01848280e-01 4.33959484e-01 -1.73284245e+00 -9.29249585e-01 -5.40396988e-01 -1.11184858e-01 2.81419665e-01 2.36567900e-01 3.90337199e-01 3.69013041e-01 -1.00131667e+00 8.56555283e-01 -7.19336689e-01 -4.38351482e-01 4.20128219e-02 3.07998329e-01 -6.43350840e-01 -3.89039546e-01 -6.41619027e-01 9.74562347e-01 -1.60348609e-01 -3.28857861e-02 -1.11224449e+00 -7.17979252e-01 -1.19093895e+00 -6.90294027e-01 2.48585716e-01 -1.31028855e+00 1.01362097e+00 -8.18792403e-01 -1.57826424e+00 1.44428921e+00 -2.32948378e-01 -3.30352604e-01 9.94850576e-01 -4.23949450e-01 2.85168644e-02 4.66964334e-01 -2.72538692e-01 7.26580322e-01 1.30774856e+00 -1.45229685e+00 -2.14382172e-01 -4.01000530e-01 -4.21492428e-01 4.72282320e-01 7.00966418e-01 -2.92957842e-01 -4.48606640e-01 -7.35006630e-01 5.28826296e-01 -1.08603036e+00 -1.53057188e-01 9.05684948e-01 1.99529365e-01 1.73909348e-02 8.49386752e-01 -1.21292889e+00 4.86361593e-01 -1.68341362e+00 6.39343977e-01 -1.50292858e-01 -6.96997531e-03 -4.30118628e-02 1.17248058e-01 5.78109920e-02 3.94748598e-01 -5.76406240e-01 3.45528089e-02 -7.88371086e-01 -1.61282137e-01 4.32590336e-01 9.30390134e-02 1.10454667e+00 -2.16722444e-01 1.01965094e+00 -8.63228798e-01 -7.45911419e-01 7.84537494e-01 7.89253235e-01 -3.22480500e-01 6.75994813e-01 -1.75482556e-02 9.05363262e-01 -1.83044195e-01 1.01342356e+00 7.14681327e-01 -1.37648627e-01 2.23262534e-01 -5.52519202e-01 -5.45368902e-02 -2.63685167e-01 -1.14749610e+00 2.19679904e+00 -1.98363185e-01 3.19908261e-01 3.11692446e-01 -8.22657049e-01 9.93362427e-01 7.24732101e-01 1.10553467e+00 -5.96315861e-01 -1.98946029e-04 8.01025704e-02 -1.03274524e+00 -6.95792735e-01 4.50180143e-01 -3.46347660e-01 1.38491645e-01 2.42273882e-01 1.77849934e-01 -5.20959377e-01 -5.72910964e-01 -3.08411360e-01 6.52979970e-01 1.17571676e+00 3.34525794e-01 -2.15455517e-01 6.16561055e-01 -5.15130199e-02 5.10075510e-01 3.89853477e-01 -3.63533288e-01 9.60596144e-01 -3.07256877e-01 -1.32873023e+00 -1.75764024e+00 -1.48001194e+00 3.77206318e-02 3.29313904e-01 3.84545654e-01 -7.32055381e-02 -3.17037463e-01 -3.17022890e-01 3.74927431e-01 -1.83719441e-01 -5.83554983e-01 3.02876830e-01 -1.01559389e+00 -7.64046097e-03 1.09496854e-01 4.76142794e-01 4.95114893e-01 -8.98016691e-01 -8.66965950e-01 1.09271042e-01 -4.40611213e-01 -1.45967519e+00 -7.34895051e-01 -4.29437488e-01 -1.19194925e+00 -1.02352750e+00 -1.23902166e+00 -9.04011011e-01 5.62955916e-01 3.54940057e-01 1.32792795e+00 1.13070101e-01 -6.99993074e-01 1.29171240e+00 -1.06239699e-01 3.63293380e-01 -2.05194727e-01 -8.14777792e-01 5.45050561e-01 -5.00623621e-02 -1.28470287e-01 -5.42570531e-01 -7.24361897e-01 6.58167362e-01 -3.38702679e-01 2.98405886e-01 9.80262086e-02 1.05215502e+00 9.54501092e-01 -2.87261963e-01 5.29370941e-02 -1.49616718e-01 -1.86193153e-01 -8.04048404e-02 -5.85176826e-01 1.22661233e-01 -2.64647335e-01 -1.70383707e-01 2.81586319e-01 -6.31144464e-01 -1.16203868e+00 5.80083668e-01 1.14496849e-01 -1.17639387e+00 -1.20406605e-01 -9.67820883e-02 3.94326776e-01 -4.67082411e-01 5.41657150e-01 7.69996941e-01 5.02131402e-01 -6.18850231e-01 1.59742653e-01 9.60985050e-02 7.21656024e-01 -7.68792331e-01 5.86074114e-01 9.10754025e-01 2.27126732e-01 -9.70160186e-01 -5.10001719e-01 -6.81973636e-01 -1.31364453e+00 -6.94526851e-01 9.24711943e-01 -1.31889415e+00 -8.05667698e-01 6.37091756e-01 -1.17868257e+00 -3.64089727e-01 -3.33878428e-01 6.73674881e-01 -1.46321285e+00 8.54605198e-01 -8.91731441e-01 -1.01743221e+00 -5.77840388e-01 -7.94123650e-01 1.49390280e+00 -3.26934576e-01 -3.63101572e-01 -1.10829115e+00 6.86160997e-02 5.20087719e-01 1.30859196e-01 7.19479263e-01 4.32524294e-01 6.57243907e-01 -8.15598547e-01 7.86049291e-02 1.53820561e-02 2.95008093e-01 1.00525789e-01 -5.29434085e-01 -7.40849197e-01 -6.79300129e-01 1.19429849e-01 -4.85004902e-01 4.59079862e-01 1.03041458e+00 7.82145798e-01 -3.03929299e-01 -1.65909350e-01 7.92740047e-01 1.42159331e+00 -1.61632285e-01 4.70319659e-01 -1.56601623e-01 7.16851234e-01 6.86339378e-01 7.64537632e-01 8.55783761e-01 2.19067052e-01 1.09557211e+00 1.18488364e-01 1.15674056e-01 -6.60562992e-01 -2.79016346e-01 4.26802963e-01 8.55409145e-01 -6.88829005e-01 3.60071331e-01 -3.12307984e-01 4.65799212e-01 -2.04452729e+00 -1.12790155e+00 1.16546355e-01 2.24518490e+00 5.04100859e-01 -3.88232797e-01 5.04256546e-01 1.49403259e-01 6.00610077e-01 -4.51802649e-03 -4.78175610e-01 -8.76890961e-03 -3.27809304e-01 -7.65564591e-02 2.92608708e-01 5.57055533e-01 -9.73312855e-01 8.42795849e-01 7.28743553e+00 4.48796988e-01 -7.86540747e-01 2.64456004e-01 2.79021531e-01 -2.61201799e-01 -2.10045531e-01 -3.10657829e-01 -5.60938179e-01 4.67694215e-02 2.80895501e-01 1.79189518e-01 4.90932673e-01 5.85493505e-01 3.88402194e-01 -8.46975446e-02 -1.18679559e+00 1.88578069e+00 4.49862152e-01 -1.52628195e+00 7.26890564e-03 -6.13024682e-02 1.05041111e+00 -5.01019001e-01 -2.30650622e-02 -2.42236301e-01 -1.82395920e-01 -1.15851736e+00 1.27089584e+00 1.08848178e+00 1.14544868e+00 -5.05030215e-01 3.35896522e-01 2.63787419e-01 -1.52190924e+00 1.87761620e-01 -6.39606893e-01 -2.14264646e-01 5.15858829e-01 3.58842701e-01 -4.42999810e-01 5.75092733e-01 1.00270927e+00 1.14992285e+00 -2.83953343e-02 8.99607718e-01 3.11157405e-01 -2.66133044e-02 -1.18039586e-01 2.20405772e-01 5.54151498e-02 -1.74971640e-01 8.03492606e-01 1.10106921e+00 3.21200699e-01 5.21130860e-01 6.23896897e-01 8.51034403e-01 3.20037097e-01 -4.98865955e-02 -1.05503619e+00 4.53293800e-01 -2.07403582e-02 8.27392638e-01 -2.51189709e-01 -3.92254591e-01 -3.59951586e-01 1.34708643e+00 1.24668382e-01 2.07410946e-01 -7.16741860e-01 8.83436620e-01 2.93063074e-01 5.76618612e-01 4.15330023e-01 -7.51822948e-01 -2.04232648e-01 -1.29280293e+00 1.74051985e-01 -6.16474926e-01 3.09808552e-01 -1.18548024e+00 -1.40458775e+00 4.37048942e-01 1.70082524e-01 -1.38034773e+00 -3.38413984e-01 -5.17624676e-01 1.79272637e-01 5.90256512e-01 -1.13498592e+00 -1.50545752e+00 -5.43353379e-01 9.68345165e-01 8.87198150e-01 -3.34217519e-01 7.62503207e-01 1.35589242e-01 4.69270974e-01 3.06377023e-01 -1.70353547e-01 -2.92338371e-01 6.32494390e-01 -8.93611729e-01 3.28306913e-01 2.86673754e-01 -8.56034160e-02 -1.23439394e-01 5.83548069e-01 -9.80209112e-01 -1.95210660e+00 -9.50483501e-01 7.21101701e-01 -7.70938218e-01 -2.54187435e-01 -2.23049670e-02 -5.91006696e-01 6.92725360e-01 -4.78363447e-02 4.26911861e-01 1.27674341e-01 -3.94239277e-01 -1.01384453e-01 4.51829545e-02 -1.64502943e+00 3.30305904e-01 1.51667154e+00 -4.16092306e-01 -5.70465863e-01 2.97204643e-01 2.96469480e-01 -8.17258656e-01 -1.23794651e+00 4.40085143e-01 1.20521235e+00 -9.81316268e-01 1.56399381e+00 -3.49163622e-01 2.13603571e-01 -1.64040774e-01 -4.89053100e-01 -7.07797945e-01 -5.65762222e-01 -8.22718918e-01 -6.42216146e-01 2.76453882e-01 -5.95059812e-01 1.63334191e-01 1.00339484e+00 3.43620926e-01 -5.50733097e-02 -7.86350965e-01 -1.19527280e+00 -1.03989875e+00 -2.48589277e-01 -1.08205020e-01 6.04122169e-02 5.96915662e-01 -3.65996599e-01 -2.79493123e-01 -1.43548298e+00 -1.60522476e-01 1.35058486e+00 3.32609922e-01 8.10233533e-01 -1.02490282e+00 -1.10590875e-01 2.26590574e-01 -6.54279470e-01 -1.42877972e+00 2.41821244e-01 -5.41433871e-01 -1.31210089e-01 -1.27205217e+00 3.41894597e-01 -3.57813358e-01 4.31091577e-01 1.70120243e-02 4.65796381e-01 5.99915922e-01 1.09505057e-01 2.88643360e-01 -5.61720669e-01 6.84286594e-01 1.68800592e+00 -5.36426157e-02 2.97915787e-02 -4.05995175e-02 5.77911139e-01 7.25124896e-01 2.00907782e-01 -1.63826063e-01 -2.39887550e-01 -4.27169889e-01 -1.12551771e-01 7.22510338e-01 1.01195621e+00 -9.43615496e-01 3.12693007e-02 -2.25796938e-01 1.09105396e+00 -1.01966333e+00 9.35629666e-01 -9.70707238e-01 8.72101724e-01 7.08266258e-01 2.52927560e-02 3.99441481e-01 1.01169668e-01 8.10011208e-01 2.10333645e-01 8.00894126e-02 1.19070745e+00 -7.77338982e-01 -9.76033211e-01 6.41100347e-01 -2.73937017e-01 -2.55625278e-01 8.13923359e-01 -6.19090259e-01 4.96690899e-01 -5.23894787e-01 -1.27943099e+00 -4.21582818e-01 8.93702209e-01 4.30317044e-01 1.50775850e+00 -1.84733260e+00 -8.76662254e-01 3.99808645e-01 -4.39807892e-01 -2.89622277e-01 5.34942091e-01 8.28435957e-01 -9.20092404e-01 5.41112542e-01 -7.02843010e-01 -1.20738328e+00 -1.68496764e+00 5.92610419e-01 6.35494888e-01 -5.03270626e-02 -1.29432452e+00 4.32550311e-01 6.37450442e-02 -6.83598518e-01 2.28710011e-01 -1.83863476e-01 3.60927016e-01 -5.98421216e-01 1.78704619e-01 7.70561039e-01 -2.57999092e-01 -1.06961334e+00 -3.09718966e-01 1.28939724e+00 5.12596369e-01 -1.14643626e-01 1.19989884e+00 -6.36587143e-01 1.69088200e-01 5.13024449e-01 1.06060505e+00 -4.33895946e-01 -1.86303329e+00 -5.14046192e-01 -6.25125229e-01 -9.68907177e-01 -1.88030064e-01 -5.81338823e-01 -9.27942455e-01 5.34922421e-01 8.39528561e-01 -5.86465776e-01 8.80531847e-01 -1.52447326e-02 1.09677148e+00 2.82169413e-02 9.58046019e-01 -9.12172973e-01 4.92078543e-01 3.81914228e-01 1.10022974e+00 -1.26197577e+00 5.20818532e-01 -4.86809731e-01 -4.46004957e-01 1.20276821e+00 4.39732790e-01 -4.40080851e-01 8.41600239e-01 -8.64570215e-03 -1.72301814e-01 -1.93330720e-01 -7.14191854e-01 1.93017438e-01 8.06670964e-01 8.65028381e-01 2.81822950e-01 -9.09845252e-03 -2.05141902e-01 3.60699072e-02 2.53236562e-01 1.28174290e-01 2.49121264e-01 8.93865526e-01 -2.03123868e-01 -7.19748080e-01 -6.42412484e-01 -8.66063908e-02 -1.44703552e-01 3.65381241e-01 -5.86413853e-02 6.77329063e-01 8.02588165e-02 5.33232212e-01 -1.23226993e-01 -1.19142830e-01 3.36632371e-01 -1.30859211e-01 1.70692670e+00 -1.79258928e-01 -3.62250209e-01 3.59579206e-01 1.62072510e-01 -8.45285416e-01 -8.88510227e-01 -8.35589588e-01 -8.23173702e-01 -6.80287659e-01 2.31500342e-01 -2.28670076e-01 4.25873280e-01 8.22312176e-01 8.89284834e-02 -1.58507735e-01 4.51967806e-01 -1.56600952e+00 -4.04242516e-01 -4.48330164e-01 -7.22023487e-01 6.53668106e-01 4.71824646e-01 -1.06446922e+00 1.21539822e-02 6.04252160e-01]
[7.207547664642334, -1.2036736011505127]
1a176d9b-fd92-41e5-8bb6-ac33196836d2
linguistic-generalization-and
1904.00157
null
https://arxiv.org/abs/1904.00157v3
https://arxiv.org/pdf/1904.00157v3.pdf
Linguistic generalization and compositionality in modern artificial neural networks
In the last decade, deep artificial neural networks have achieved astounding performance in many natural language processing tasks. Given the high productivity of language, these models must possess effective generalization abilities. It is widely assumed that humans handle linguistic productivity by means of algebraic compositional rules: Are deep networks similarly compositional? After reviewing the main innovations characterizing current deep language processing networks, I discuss a set of studies suggesting that deep networks are capable of subtle grammar-dependent generalizations, but also that they do not rely on systematic compositional rules. I argue that the intriguing behaviour of these devices (still awaiting a full understanding) should be of interest to linguists and cognitive scientists, as it offers a new perspective on possible computational strategies to deal with linguistic productivity beyond rule-based compositionality, and it might lead to new insights into the less systematic generalization patterns that also appear in natural language.
['Marco Baroni']
2019-03-30
null
null
null
null
['systematic-generalization']
['reasoning']
[ 2.74993092e-01 3.09825629e-01 -1.30947396e-01 -3.33001554e-01 4.60373640e-01 -8.31006289e-01 8.64112258e-01 2.01393262e-01 -4.74352419e-01 3.39467257e-01 5.08033156e-01 -1.02204382e+00 -4.15773004e-01 -1.03561342e+00 -4.35497314e-01 -2.84392267e-01 -1.90069377e-01 4.78664607e-01 8.49530250e-02 -8.66777599e-01 3.64915073e-01 6.95496321e-01 -1.62508702e+00 4.94369328e-01 9.41139996e-01 6.64096355e-01 2.34572753e-01 5.23259640e-01 -4.33280170e-01 1.15791738e+00 -3.39535594e-01 -7.56476223e-01 4.00776416e-02 -4.20791060e-01 -9.36188638e-01 -1.63805857e-01 4.36379582e-01 -1.44062802e-01 -1.88347250e-01 1.17806482e+00 3.61111239e-02 2.07668304e-01 6.83021307e-01 -7.73327708e-01 -1.31782746e+00 1.09912884e+00 -3.99583802e-02 5.27370036e-01 1.15292169e-01 5.13899863e-01 1.35721207e+00 -4.16845560e-01 4.98137563e-01 1.64165103e+00 8.78264248e-01 5.80592573e-01 -1.27278209e+00 -5.66585481e-01 4.56595272e-01 -3.45264748e-02 -8.00845265e-01 -2.90322065e-01 5.91138363e-01 -4.65378672e-01 1.38465321e+00 -6.54620230e-02 7.42001951e-01 1.09063900e+00 4.88056749e-01 5.82001805e-01 1.06929469e+00 -5.48809290e-01 -8.81022289e-02 -2.59564012e-01 1.80398881e-01 1.11719179e+00 6.65708363e-01 1.30072579e-01 -5.49349010e-01 -9.94665734e-03 8.81576180e-01 -1.64512753e-01 -7.24888593e-02 6.94895089e-02 -1.26121998e+00 1.02036381e+00 4.88845557e-01 9.85891342e-01 -1.79658204e-01 3.81717324e-01 7.76908815e-01 5.53525031e-01 8.73003006e-02 1.08466876e+00 -6.62917852e-01 1.21778004e-01 -7.11503863e-01 5.88815808e-01 7.21199989e-01 7.98366606e-01 4.67427820e-01 3.66444468e-01 4.27392185e-01 7.89986730e-01 8.05756226e-02 3.73719126e-01 7.89448977e-01 -1.13588476e+00 1.54163674e-01 8.42479646e-01 -3.21681768e-01 -1.21232331e+00 -6.80569768e-01 -4.02438998e-01 -9.67717469e-01 -3.25441249e-02 8.29541028e-01 6.82133064e-02 -4.07866478e-01 2.15650320e+00 -4.48608309e-01 -7.10672140e-01 -1.37018993e-01 5.78720450e-01 4.86639649e-01 4.71504182e-01 5.21555781e-01 -1.46852550e-03 1.45519805e+00 -3.82377625e-01 -4.73448753e-01 -4.19918180e-01 8.49616766e-01 -4.51459110e-01 1.54382777e+00 5.47660649e-01 -1.32068288e+00 -5.67163169e-01 -1.04636621e+00 -5.21149397e-01 -5.79797626e-01 -2.83398420e-01 1.38683510e+00 7.62502909e-01 -1.19265223e+00 6.61785722e-01 -6.02482080e-01 -6.90967977e-01 2.48562008e-01 3.04195613e-01 -5.61391748e-02 2.40530059e-01 -1.31733525e+00 1.07158935e+00 5.28108239e-01 9.01182294e-02 -5.65652788e-01 -5.57107329e-01 -8.60250831e-01 3.09751451e-01 5.16693175e-01 -1.07167339e+00 1.45860803e+00 -1.41279411e+00 -1.33971930e+00 1.13989627e+00 -1.26297012e-01 -5.72322905e-01 1.01790242e-01 5.21751940e-02 -3.96824896e-01 -5.76110929e-02 -1.08483016e-01 4.17621583e-01 4.82396543e-01 -1.01258373e+00 -5.28822660e-01 -4.24132586e-01 3.55038792e-01 -2.57731024e-02 -3.12923521e-01 2.86402285e-01 5.58975756e-01 -8.40053678e-01 1.17019124e-01 -7.93673277e-01 5.07707931e-02 -1.34260412e-02 -2.99286786e-02 -7.11558700e-01 -4.17344868e-02 -3.46941620e-01 1.12484753e+00 -1.76515150e+00 7.84584656e-02 -5.46879172e-02 4.82192993e-01 1.83658868e-01 -2.14985594e-01 4.80327904e-01 -4.59903516e-02 4.90420431e-01 -4.33238894e-01 5.41918427e-02 3.69668573e-01 2.09422797e-01 -7.63584912e-01 2.02365726e-01 3.71711820e-01 1.36379695e+00 -1.08553410e+00 -7.44660944e-02 -1.53422192e-01 2.84936819e-02 -6.49898827e-01 -2.03431055e-01 -5.63323498e-01 -9.44287553e-02 -4.84408066e-02 4.16328639e-01 1.99675128e-01 -1.66050702e-01 6.93871439e-01 1.90457448e-01 -2.60368139e-01 7.70203531e-01 -5.63164413e-01 1.59190214e+00 -4.25542027e-01 8.18111718e-01 8.21083412e-03 -1.20328808e+00 6.33843720e-01 9.04355124e-02 -2.75055587e-01 -7.76634455e-01 4.82176691e-01 3.65312010e-01 1.09129894e+00 -3.42592835e-01 5.10759473e-01 -6.98305726e-01 -2.88398206e-01 7.69820035e-01 8.15953016e-02 -4.94902521e-01 4.52703446e-01 8.16471949e-02 9.52394664e-01 -6.39772788e-02 4.13836509e-01 -8.47303033e-01 6.16673768e-01 2.04316769e-02 3.67155969e-01 9.52629149e-01 -1.62981346e-01 -1.16761155e-01 5.89179158e-01 -9.92948413e-01 -1.33032942e+00 -1.08541107e+00 5.18193617e-02 1.76939237e+00 -4.55945849e-01 -1.66986197e-01 -5.92174590e-01 -6.69059306e-02 3.68183963e-02 8.05561900e-01 -6.41362786e-01 -5.13124526e-01 -8.19076777e-01 -5.90926111e-01 1.04586804e+00 6.72550142e-01 4.71220791e-01 -1.39648616e+00 -7.35830009e-01 1.56203330e-01 1.51622757e-01 -9.56164658e-01 -7.87431523e-02 3.91416699e-01 -1.13745511e+00 -7.41931856e-01 -4.10476387e-01 -9.63601649e-01 4.24456328e-01 1.95618063e-01 1.32483125e+00 6.32853270e-01 5.57752922e-02 1.22636706e-01 -2.17304248e-02 -6.97427630e-01 -8.54922950e-01 3.25976610e-01 4.54130888e-01 -7.00501442e-01 6.97700500e-01 -9.34006691e-01 -2.18272105e-01 -9.48397294e-02 -1.08760226e+00 -1.76263288e-01 3.87094617e-01 6.15167797e-01 -3.28105301e-01 2.08605826e-01 8.14800560e-01 -8.56065333e-01 1.37905502e+00 -4.29738104e-01 -4.20353472e-01 1.12647854e-01 -5.07998526e-01 3.17576349e-01 9.47064698e-01 -4.17158097e-01 -1.14320219e+00 -6.17049575e-01 -4.40694876e-02 2.65836775e-01 -2.43572876e-01 6.12031579e-01 1.44353544e-03 1.68616340e-01 9.02622998e-01 2.99078017e-01 6.85437173e-02 -2.02183157e-01 5.66535950e-01 2.38693714e-01 5.55653989e-01 -1.17696130e+00 5.44626534e-01 3.47620219e-01 1.31708384e-01 -1.08915901e+00 -7.06185639e-01 1.55126750e-01 -4.95997012e-01 1.85866505e-01 8.04605305e-01 -6.30018055e-01 -9.45231974e-01 3.62756699e-01 -1.40370846e+00 -6.25008225e-01 -1.80520266e-01 2.94868257e-02 -6.85373425e-01 4.77344304e-01 -1.10377991e+00 -6.68125451e-01 -2.19117478e-01 -1.02212441e+00 4.99152482e-01 5.38294241e-02 -9.60751414e-01 -1.39397943e+00 -2.37518311e-01 4.86598499e-02 6.33196235e-01 -2.62777209e-01 1.87840116e+00 -7.33319759e-01 -3.29814434e-01 1.58574924e-01 -2.57246733e-01 3.21002245e-01 2.83888616e-02 1.03133038e-01 -8.19524825e-01 1.95745334e-01 4.68816385e-02 -4.31777686e-01 8.85142207e-01 3.05110067e-01 1.20525718e+00 -3.10537636e-01 3.86462174e-02 3.55928212e-01 1.13210547e+00 2.35094219e-01 2.95364916e-01 1.98497415e-01 3.67490739e-01 9.10141528e-01 -3.99357229e-01 -1.28911883e-01 5.09581447e-01 1.10083349e-01 7.04354495e-02 3.27831596e-01 -8.06338042e-02 -3.11148643e-01 1.82533175e-01 1.02635908e+00 -4.00940478e-01 -2.24268049e-01 -1.16318834e+00 5.75756848e-01 -1.65555930e+00 -1.22160959e+00 -2.48703733e-03 1.95279562e+00 8.79872799e-01 4.63843256e-01 3.23587030e-01 2.86986083e-01 4.52903032e-01 9.27767456e-02 -4.15159732e-01 -1.30768311e+00 -5.52516699e-01 3.98877710e-01 9.30891559e-02 4.46355760e-01 -5.14222324e-01 1.27691913e+00 7.75544691e+00 2.24311084e-01 -1.20311391e+00 -1.82951707e-02 4.13618863e-01 3.03690255e-01 -5.85352659e-01 -3.18241566e-01 -4.87981200e-01 1.18917972e-01 1.04918838e+00 -2.92697340e-01 7.09257543e-01 5.89370787e-01 7.89712146e-02 2.29406878e-02 -1.42227995e+00 4.89701241e-01 -9.33845118e-02 -1.38818276e+00 4.84348506e-01 8.83203074e-02 6.44039333e-01 -5.67834936e-02 1.91477984e-01 4.59116131e-01 7.84928858e-01 -1.29105389e+00 8.64149570e-01 2.68734634e-01 4.50066149e-01 -7.27449775e-01 2.95911342e-01 5.88151157e-01 -8.65102589e-01 -4.54352319e-01 -6.21232629e-01 -9.80962932e-01 -4.74412851e-02 4.36213434e-01 -2.90273815e-01 -9.49281082e-02 4.20573533e-01 2.95376211e-01 -4.16294217e-01 3.85705054e-01 -4.68474865e-01 5.10363102e-01 -1.14737667e-01 -3.34085137e-01 4.43677664e-01 -3.36583965e-02 2.32384473e-01 1.33444715e+00 5.40931746e-02 9.93827730e-02 -2.92781085e-01 1.40439689e+00 -2.20480204e-01 3.30470838e-02 -9.54041302e-01 -5.69607019e-01 4.32865858e-01 7.45418847e-01 -1.03526604e+00 -4.58947361e-01 -5.66345990e-01 7.67565668e-01 6.14615023e-01 2.64710426e-01 -3.96855235e-01 -2.77408898e-01 7.29742110e-01 2.47229010e-01 -1.78266287e-01 -7.26424813e-01 -7.76941061e-01 -1.14690924e+00 -2.50473142e-01 -1.17775881e+00 1.07160456e-01 -7.05948949e-01 -1.24996030e+00 4.04610425e-01 -1.67183783e-02 -1.68489397e-01 -2.75719136e-01 -1.16312242e+00 -5.93537331e-01 7.75112033e-01 -1.02661383e+00 -9.30261791e-01 1.83425814e-01 2.62781858e-01 6.38700008e-01 -1.68043748e-01 8.73506069e-01 -9.30863395e-02 -2.10974768e-01 4.73542124e-01 -2.65239149e-01 1.27086326e-01 2.74978071e-01 -1.17987704e+00 7.07975686e-01 7.53357470e-01 2.30847061e-01 1.48648715e+00 7.35290825e-01 -3.52222592e-01 -1.20535052e+00 -6.31751001e-01 1.29836392e+00 -6.94695234e-01 9.94556129e-01 -5.00131249e-01 -1.03213453e+00 9.34221387e-01 5.49852312e-01 -5.66377640e-01 5.46801746e-01 4.46344763e-01 -7.84859598e-01 1.33046672e-01 -8.15090477e-01 9.61135685e-01 1.53736472e+00 -7.70137906e-01 -1.24923146e+00 1.23878896e-01 9.46055293e-01 3.12106818e-01 -4.42739338e-01 1.66208655e-01 7.73304045e-01 -9.96111751e-01 7.09753335e-01 -9.96232986e-01 9.12496030e-01 -2.48907898e-02 -3.49657908e-02 -1.29313076e+00 -8.52935970e-01 -4.42329198e-01 2.40900517e-01 8.12639594e-01 4.77459580e-01 -9.02422667e-01 2.96108246e-01 6.77122533e-01 -3.36467147e-01 -5.57275414e-01 -5.66355407e-01 -9.07024324e-01 9.57309067e-01 -6.17886662e-01 7.51530886e-01 1.17962980e+00 4.97777820e-01 5.76220214e-01 1.98996618e-01 -3.11255544e-01 2.53651679e-01 -2.16521509e-02 3.69041175e-01 -1.64268386e+00 -1.71546698e-01 -1.40861213e+00 -2.68499345e-01 -8.78204286e-01 6.62309468e-01 -1.27794826e+00 -3.05145979e-01 -1.25289404e+00 1.31775081e-01 -5.78630157e-02 -3.15908454e-02 4.07696903e-01 5.42611852e-02 1.17398933e-01 1.92387551e-01 1.49419606e-02 -2.71267295e-01 1.60803571e-01 1.22944760e+00 -9.33520868e-02 2.17902940e-02 -5.31784117e-01 -1.38812065e+00 1.10154760e+00 9.29522097e-01 -1.09923832e-01 -5.02678692e-01 -9.12876368e-01 9.64692414e-01 -6.48417413e-01 1.06807448e-01 -7.82612264e-01 3.32331518e-03 -4.11537081e-01 6.80557266e-02 2.00041085e-01 -2.32153907e-01 -6.46107376e-01 -3.20715129e-01 7.69441247e-01 -6.15054429e-01 5.18130898e-01 4.19757545e-01 5.44144101e-02 1.44943476e-01 -1.50933012e-01 8.23129475e-01 -7.82678545e-01 -7.29799271e-01 -1.79698408e-01 -1.20899522e+00 9.26863477e-02 5.25425613e-01 -9.01914313e-02 -4.43611145e-01 -2.44462699e-01 -5.09750187e-01 -1.55718535e-01 5.89252234e-01 6.09903157e-01 1.19027413e-01 -1.10376430e+00 -3.17378223e-01 -5.33301905e-02 1.47240281e-01 -2.08804935e-01 -9.26497430e-02 4.29540604e-01 -7.56455898e-01 8.41135859e-01 -2.38779441e-01 -2.86831800e-02 -5.38232684e-01 9.12007809e-01 6.39907241e-01 1.38173224e-02 -5.43879271e-01 1.03862154e+00 5.50335705e-01 -5.99086642e-01 -5.35602458e-02 -1.04671907e+00 2.18943283e-01 -6.93428963e-02 6.09026670e-01 2.08242565e-01 -1.02692112e-01 -4.12160248e-01 -2.72075355e-01 3.33615839e-01 -4.34952974e-02 1.18004680e-01 1.19128180e+00 -9.99977579e-04 -6.76854730e-01 8.20773005e-01 5.75065136e-01 -4.54365276e-02 -3.85535240e-01 -3.74336153e-01 3.57394725e-01 1.57020286e-01 -3.39069545e-01 -8.49337161e-01 -6.58111155e-01 1.25493503e+00 -1.92523256e-01 5.06554186e-01 9.29052174e-01 7.53335431e-02 5.78797162e-01 7.33802021e-01 3.82544935e-01 -1.15976620e+00 2.55004726e-02 1.05143559e+00 1.06969368e+00 -8.99469674e-01 -9.90319774e-02 -1.35410786e-01 -1.64364219e-01 1.37717414e+00 4.74039376e-01 -4.30778205e-01 3.86453211e-01 2.87365139e-01 -2.56699443e-01 -2.53953367e-01 -1.04988086e+00 -2.36950845e-01 -7.61005357e-02 6.20026469e-01 1.12938130e+00 2.47484401e-01 -7.38613367e-01 7.04703033e-01 -7.36084878e-01 -3.88675816e-02 5.07654846e-01 5.03730416e-01 -7.02047050e-01 -1.06678259e+00 -2.94114113e-01 3.91142398e-01 -3.98024797e-01 -4.08151746e-01 -7.08584785e-01 8.06482673e-01 4.34744567e-01 7.80275881e-01 1.82842612e-01 -1.54943660e-01 1.65881798e-01 5.68646193e-01 8.38981867e-01 -8.49459052e-01 -7.86121964e-01 -3.89006734e-01 9.98260453e-02 -2.68699527e-01 -2.29904443e-01 -6.74220622e-01 -1.30401850e+00 -8.21775436e-01 1.73268095e-01 -7.99870491e-02 2.91579455e-01 1.06505251e+00 6.97862878e-02 4.50737000e-01 -2.49148965e-01 -6.00264907e-01 -6.00507379e-01 -1.00685072e+00 -5.66004932e-01 3.77444774e-01 1.85041279e-01 -4.26393837e-01 -3.56116593e-01 4.29614857e-02]
[10.4725980758667, 8.884968757629395]
14199ad6-b984-416b-93ea-8a5210e0e2fe
full-pulse-tomographic-reconstruction-with
1802.02242
null
http://arxiv.org/abs/1802.02242v1
http://arxiv.org/pdf/1802.02242v1.pdf
Full-pulse Tomographic Reconstruction with Deep Neural Networks
Plasma tomography consists in reconstructing the 2D radiation profile in a poloidal cross-section of a fusion device, based on line-integrated measurements along several lines of sight. The reconstruction process is computationally intensive and, in practice, only a few reconstructions are usually computed per pulse. In this work, we trained a deep neural network based on a large collection of sample tomograms that have been produced at JET over several years. Once trained, the network is able to reproduce those results with high accuracy. More importantly, it can compute all the tomographic reconstructions for a given pulse in just a few seconds. This makes it possible to visualize several phenomena -- such as plasma heating, disruptions and impurity transport -- over the course of a discharge.
['Horácio Fernandes', 'Pedro J. Carvalho', 'Diogo R. Ferreira']
2018-02-02
null
null
null
null
['tomographic-reconstructions']
['medical']
[-3.20303351e-01 -2.45252639e-01 4.62009817e-01 -3.18461329e-01 -5.12896419e-01 -3.68282527e-01 8.04903209e-01 1.83290839e-01 -2.11576641e-01 1.02193034e+00 9.43538472e-02 -5.00473022e-01 -2.93058027e-02 -1.01093733e+00 -4.65384990e-01 -7.86069691e-01 -2.38493398e-01 1.20668411e+00 -2.91297466e-01 -1.13501400e-01 1.93591326e-01 1.37454665e+00 -9.52453911e-01 3.25553752e-02 4.52331245e-01 1.20214820e+00 -7.56304488e-02 3.98996532e-01 9.42088068e-02 9.13558662e-01 -5.16006231e-01 -5.55473305e-02 -1.42083004e-01 -6.36121392e-01 -7.00640321e-01 1.35728210e-01 -2.60155290e-01 -4.23278004e-01 -6.20900929e-01 8.95907938e-01 3.11557204e-01 3.04675579e-01 8.58785927e-01 -3.59942257e-01 9.73334908e-02 3.92474532e-01 -4.52369183e-01 5.66084921e-01 7.32709020e-02 2.65485436e-01 3.11419398e-01 -8.05894375e-01 6.72000706e-01 4.96170342e-01 6.11982942e-01 1.14953220e-01 -1.17594731e+00 -6.61264881e-02 -1.07633913e+00 3.64904813e-02 -9.33304608e-01 -4.49650913e-01 7.73255110e-01 -5.98635912e-01 1.28085673e+00 1.01413786e-01 1.02783298e+00 8.65518272e-01 6.83035910e-01 3.71943086e-01 1.04872096e+00 -1.56697020e-01 4.07414794e-01 -2.33180616e-02 -3.50089222e-02 5.52547216e-01 6.25256896e-02 6.98841810e-01 -2.44640633e-01 -7.75918961e-02 7.77906954e-01 -3.98264945e-01 -7.28857756e-01 -1.71549246e-01 -7.69372642e-01 8.73462081e-01 5.08171320e-01 7.51676440e-01 -7.48174846e-01 -2.86104113e-01 7.66130030e-01 1.35995895e-01 7.31546581e-01 9.24806297e-01 -2.79992491e-01 -4.11066055e-01 -1.42137718e+00 4.11853701e-01 8.42164576e-01 2.51327664e-01 6.98314726e-01 5.47056973e-01 1.14361890e-01 3.09000790e-01 -2.74874866e-01 5.26277006e-01 6.10497653e-01 -5.07106662e-01 -2.48418450e-01 8.17717314e-02 4.15048867e-01 -5.49808323e-01 -8.68734837e-01 -3.35764527e-01 -1.45651674e+00 4.95178282e-01 1.88373685e-01 -6.98186100e-01 -9.00802791e-01 1.14923918e+00 2.83392340e-01 1.65044382e-01 2.23068058e-01 1.07918751e+00 7.13035583e-01 9.40040410e-01 -6.25034332e-01 -7.77913272e-01 1.04179144e+00 -8.40520144e-01 -8.07674527e-01 -9.76851881e-02 4.36824709e-01 -4.81694371e-01 3.20498854e-01 4.89719987e-01 -1.52784383e+00 -5.15893757e-01 -1.23966885e+00 3.43295157e-01 -1.76892370e-01 -1.80403039e-01 6.38403952e-01 7.41711399e-03 -1.08420539e+00 1.41183770e+00 -9.12224591e-01 -6.19998090e-02 3.81436288e-01 -6.49949759e-02 -4.50536728e-01 2.80486256e-01 -9.87513900e-01 1.02081621e+00 3.41860563e-01 1.59926012e-01 -1.19618928e+00 -8.38435888e-01 -6.07216597e-01 1.61390260e-01 -2.45281547e-01 -6.13017380e-01 1.87312317e+00 -2.03313828e-01 -1.73355985e+00 4.50318277e-01 -3.18309844e-01 -7.92860866e-01 5.15693367e-01 8.39544088e-02 -5.90284526e-01 4.16179270e-01 -7.05075338e-02 -1.03514217e-01 7.26058364e-01 -1.21321642e+00 -1.83040872e-01 -1.84927434e-01 -5.00331581e-01 3.72537524e-02 4.86453623e-01 9.20317098e-02 -5.75296171e-02 1.89706609e-02 2.37873554e-01 -3.34313095e-01 -4.48298216e-01 -5.70271134e-01 -5.09128392e-01 3.08381289e-01 9.19771671e-01 -6.69996917e-01 2.59013087e-01 -1.84769201e+00 3.02264869e-01 -9.77377687e-03 3.45552266e-02 2.88852125e-01 6.13271773e-01 6.92369580e-01 -4.34103966e-01 -1.04854450e-01 -7.95728922e-01 -5.91306269e-01 -4.72540706e-01 -2.79123727e-02 -8.32729638e-01 8.09797466e-01 -2.83810526e-01 9.53752935e-01 -5.95156848e-01 2.38090366e-01 4.96359527e-01 5.08475244e-01 3.00987810e-02 5.46915054e-01 -4.42384511e-01 9.86539721e-01 -2.68333495e-01 2.12147132e-01 1.13607204e+00 -1.83854967e-01 -1.33464918e-01 -7.38952607e-02 -5.62951624e-01 2.91686565e-01 -1.32256985e-01 1.69562852e+00 -6.49924517e-01 9.49022174e-01 2.66689360e-01 -1.28415334e+00 8.44227016e-01 8.16185117e-01 6.63056254e-01 -1.08268690e+00 6.20318711e-01 3.11248720e-01 -1.48241088e-01 -5.21218836e-01 5.87843835e-01 -1.13301909e+00 1.40416592e-01 9.17626321e-01 9.89815220e-02 -8.31197083e-01 1.52590320e-01 -1.05896533e-01 1.09091938e+00 -5.89615047e-01 9.52632278e-02 -3.48318279e-01 1.12161398e-01 2.51130015e-01 5.31665459e-02 4.97105449e-01 2.36130431e-01 6.74367011e-01 4.81153846e-01 -9.59846318e-01 -1.49708736e+00 -9.06355798e-01 -7.15700865e-01 -3.64348650e-01 -3.71996202e-02 -1.70528293e-01 -7.68667996e-01 -1.83773234e-01 -1.67934418e-01 8.90643299e-01 -5.71729958e-01 1.52750546e-02 -5.98869741e-01 -1.07784343e+00 5.54387927e-01 4.45263922e-01 6.32794857e-01 -1.83451843e+00 -1.01185584e+00 2.11262599e-01 1.83189232e-02 -1.06373811e+00 3.44106078e-01 4.48856652e-01 -9.00863051e-01 -1.05755126e+00 -4.28476483e-01 -1.86293304e-01 4.62991327e-01 -2.07221359e-01 1.25174701e+00 -1.57982204e-02 -3.98858011e-01 -3.68003964e-01 -1.87602028e-01 -1.70131311e-01 -5.67910016e-01 -5.26718438e-01 3.46127361e-01 -2.22003654e-01 5.72499335e-02 -8.81320477e-01 -1.96866572e-01 -3.62255543e-01 -6.03697956e-01 7.08995983e-02 1.84154466e-01 9.04530466e-01 4.69449699e-01 7.65729725e-01 1.77358225e-01 -6.91633523e-01 8.12536836e-01 -4.43553835e-01 -1.01911771e+00 -2.21002623e-01 -2.41164908e-01 -2.24412069e-01 1.29656756e+00 1.72114268e-01 -1.37280142e+00 -5.91937542e-01 -5.12967348e-01 -5.78609943e-01 -5.50586879e-01 7.67616868e-01 3.69978517e-01 -2.09451076e-02 8.28604877e-01 4.76752430e-01 -6.25888854e-02 -4.62150514e-01 6.39947206e-02 2.62038529e-01 8.71166825e-01 -1.63036168e-01 6.73048198e-01 7.17365980e-01 2.00740755e-01 -1.16495574e+00 -8.34653080e-01 -2.68621236e-01 -6.51183486e-01 -2.49397174e-01 9.14155960e-01 -6.50927901e-01 -8.51354897e-01 6.80016756e-01 -1.24178958e+00 -6.54715002e-01 -6.66862369e-01 7.41568804e-01 -6.58103287e-01 2.21980602e-01 -9.65998411e-01 -6.51717842e-01 -6.64081573e-01 -1.11419761e+00 6.85829759e-01 5.84205449e-01 -1.39330998e-02 -1.15538549e+00 4.32962716e-01 -2.63676256e-01 6.42521620e-01 4.59774911e-01 8.64080906e-01 -1.54076749e-02 -3.06155860e-01 -1.43175974e-01 -2.09650233e-01 4.65136133e-02 6.11757338e-02 -1.21701576e-01 -1.11004972e+00 -5.86993992e-01 1.05666447e+00 -5.32358527e-01 8.77886236e-01 9.22126949e-01 1.24408865e+00 -7.57372901e-02 -5.57145596e-01 1.18027914e+00 1.46410978e+00 4.58210498e-01 4.19815451e-01 -1.06854036e-01 2.22846031e-01 1.02112196e-01 2.66864836e-01 5.89657605e-01 -6.00725114e-01 4.53290135e-01 3.09106320e-01 -3.33016604e-01 2.94708848e-01 1.29476145e-01 -1.76687017e-01 6.26080394e-01 -1.03714220e-01 -3.37433845e-01 -1.00744331e+00 3.50768596e-01 -1.39637578e+00 -1.03349066e+00 -3.26753229e-01 1.92775691e+00 3.36618632e-01 1.08274169e-01 -3.35590661e-01 1.34420618e-01 1.29987538e-01 3.06188703e-01 -6.94691718e-01 -7.85338640e-01 -1.48470923e-01 6.71621203e-01 3.59883517e-01 7.12844014e-01 -6.60886288e-01 5.99985719e-01 7.36986017e+00 2.22225115e-01 -1.74469817e+00 1.97602093e-01 4.68467802e-01 -5.96893206e-02 -2.92763352e-01 -1.29225612e-01 -5.96874952e-02 2.07079783e-01 1.32523608e+00 -4.87069368e-01 7.42127597e-01 5.06591320e-01 2.91537642e-01 -6.95636570e-01 -7.41519749e-01 8.71619046e-01 -7.25368410e-02 -1.97590923e+00 -2.85292298e-01 -1.08240955e-01 7.01614439e-01 6.93047881e-01 -3.35186511e-01 6.13219626e-02 6.79810420e-02 -1.27535248e+00 6.40095890e-01 7.01487064e-01 1.04349923e+00 -9.23877299e-01 7.86007345e-01 8.96080196e-01 -5.91369629e-01 2.22525477e-01 -4.70603436e-01 -3.33809964e-02 1.15753329e+00 1.15912998e+00 -1.15953469e+00 1.08390391e+00 7.14894772e-01 5.78185141e-01 4.57982533e-02 1.01476371e+00 -2.55593002e-01 5.38564265e-01 -4.43521649e-01 3.81962001e-01 3.83465677e-01 -7.96251833e-01 4.87981558e-01 9.33204055e-01 7.45636702e-01 3.41802120e-01 -2.85984546e-01 1.07237303e+00 -6.86078593e-02 -5.28519690e-01 -1.05530930e+00 -6.05751351e-02 -1.20745912e-01 1.63001752e+00 -7.19534874e-01 -3.42912525e-01 1.82026908e-01 7.14155734e-01 4.48928922e-01 3.64541978e-01 -6.16254807e-01 -7.54044205e-02 4.79745477e-01 2.66172647e-01 1.08548664e-01 -3.49010199e-01 -3.51525545e-01 -8.87200415e-01 -2.18280897e-01 -2.21992686e-01 -1.46150857e-01 -1.22165942e+00 -1.12923396e+00 1.25772190e+00 3.22232544e-02 -5.22997439e-01 -6.36012673e-01 -6.11366630e-01 -1.31444645e+00 1.24772608e+00 -1.36528826e+00 -3.70063275e-01 -2.79982775e-01 2.73373038e-01 2.11703449e-01 -2.40888491e-01 1.18665111e+00 4.10977416e-02 -5.01776457e-01 -4.03794527e-01 3.34497362e-01 -1.83179498e-01 -2.18726113e-01 -1.27882218e+00 7.73833096e-01 7.73315609e-01 -2.47956607e-02 1.70379385e-01 1.09064472e+00 -5.88346660e-01 -1.09664094e+00 -7.81090319e-01 5.70209980e-01 2.77398829e-03 4.39234942e-01 -1.41800985e-01 -1.21130979e+00 7.27449834e-01 7.98937857e-01 3.20935696e-01 2.57519484e-01 -9.89289433e-02 5.71066082e-01 2.91669130e-01 -1.17279088e+00 1.96933746e-01 4.59643304e-01 -7.69504666e-01 -3.91326964e-01 7.33880401e-01 1.07829332e-01 -8.65281165e-01 -6.90803349e-01 5.92139781e-01 -1.69050053e-01 -1.54286253e+00 4.65996087e-01 -4.76604074e-01 4.77097750e-01 -1.70004189e-01 4.37231779e-01 -1.91261971e+00 -7.60561451e-02 -6.09934568e-01 -1.13881536e-01 5.05548358e-01 2.92013735e-01 -4.60631132e-01 8.27305973e-01 -4.72580343e-02 -5.15618443e-01 -9.69191611e-01 -1.16603374e+00 -3.23651820e-01 2.69458711e-01 -4.00284588e-01 6.80456579e-01 8.36035490e-01 2.94436693e-01 2.04468474e-01 -3.16061020e-01 5.24860084e-01 4.20307547e-01 6.39797628e-01 3.40315670e-01 -1.23779368e+00 -2.48792216e-01 -2.22644210e-02 1.36326641e-01 -8.01446915e-01 4.41384763e-01 -9.50976551e-01 2.52482802e-01 -1.53731120e+00 -8.34018216e-02 -5.16397655e-01 3.31797659e-01 1.43422633e-01 7.25536525e-01 1.20584652e-01 -3.45227480e-01 2.42638990e-01 2.27435783e-01 8.28314722e-01 1.26159871e+00 8.75508934e-02 7.61636794e-02 3.03412136e-02 7.41531923e-02 7.75876403e-01 1.18735921e+00 -2.87436724e-01 -2.09802598e-01 -3.66824657e-01 -5.58428019e-02 9.56081569e-01 2.73169875e-01 -1.28888142e+00 3.06306422e-01 1.88969880e-01 6.87581837e-01 -8.51765871e-01 8.34256709e-01 -3.93401086e-01 3.81300539e-01 2.04634070e-01 2.74006724e-01 1.53746411e-01 6.52056694e-01 1.16851076e-01 -7.28420258e-01 -5.07936180e-01 9.77688134e-01 -4.42069501e-01 -8.78415257e-02 5.18609405e-01 -6.86628819e-01 -2.09852681e-01 9.32335496e-01 4.71829981e-01 -1.65690616e-01 -2.20211729e-01 -6.46004498e-01 -2.07068250e-01 6.24276638e-01 -2.32376605e-01 5.83268881e-01 -9.93574858e-01 -4.10824984e-01 6.54765904e-01 -5.41728199e-01 3.23565900e-01 6.70532048e-01 8.09340000e-01 -1.09572303e+00 6.11083686e-01 -2.58438915e-01 -6.81138396e-01 -6.51298344e-01 5.80664933e-01 1.19619012e+00 -4.41566676e-01 -1.41421092e+00 6.76448047e-01 -2.34948043e-02 -3.37747484e-01 -6.37785375e-01 -2.56314486e-01 -2.78277248e-01 -2.14766353e-01 7.00296402e-01 -2.27247089e-01 6.51580513e-01 -6.99983120e-01 2.68641114e-01 3.59200835e-01 1.55663237e-01 -3.22261721e-01 1.47678339e+00 4.12305325e-01 -3.86300087e-01 4.87084627e-01 1.04600179e+00 -3.41129750e-01 -1.27124226e+00 1.06005140e-01 -5.77715814e-01 -2.99331576e-01 5.65520346e-01 -7.23240197e-01 -1.31868315e+00 1.11434555e+00 9.51742753e-02 6.39942229e-01 1.04738081e+00 -1.09097756e-01 7.61546075e-01 3.46285582e-01 3.98217440e-01 -5.52825093e-01 -4.87020582e-01 7.26214588e-01 8.52441788e-01 -7.84658372e-01 -1.17161676e-01 1.19871594e-01 -3.74806702e-01 1.36712146e+00 2.04784021e-01 -5.61128668e-02 9.71738338e-01 7.34806538e-01 1.40666679e-01 -8.68367255e-01 -7.81463146e-01 5.33186555e-01 -2.30267227e-01 9.11968276e-02 1.75349265e-01 1.88671634e-01 1.81537732e-01 -5.04981168e-02 -6.69731379e-01 -4.69032526e-02 8.93689632e-01 6.17101789e-01 -3.37012321e-01 -6.95767224e-01 -2.07214579e-01 5.45146406e-01 -2.81466544e-01 1.04158513e-01 2.92273462e-01 5.32455623e-01 -1.32058248e-01 7.09346831e-01 4.90646362e-01 6.11500479e-02 1.97402477e-01 1.06428877e-01 4.29879248e-01 -2.84968764e-01 -3.71774763e-01 -2.12151304e-01 1.55757755e-01 -4.28196102e-01 -1.73393846e-01 -7.08220243e-01 -1.50088465e+00 -6.61021054e-01 -2.80719530e-02 8.21369648e-01 6.73663318e-01 1.26017094e+00 -2.19615906e-01 8.13408136e-01 7.05127895e-01 -1.36812186e+00 -2.26465344e-01 -1.24874413e+00 -1.51850116e+00 1.38097957e-01 5.75869918e-01 -5.16263604e-01 -5.38931370e-01 -5.20515919e-01]
[6.518337249755859, 3.458902597427368]
a6e2366d-7025-45b3-bf6f-c860833e97b6
monte-carlo-inference-for-semiparametric
2306.05498
null
https://arxiv.org/abs/2306.05498v1
https://arxiv.org/pdf/2306.05498v1.pdf
Monte Carlo inference for semiparametric Bayesian regression
Data transformations are essential for broad applicability of parametric regression models. However, for Bayesian analysis, joint inference of the transformation and model parameters typically involves restrictive parametric transformations or nonparametric representations that are computationally inefficient and cumbersome for implementation and theoretical analysis, which limits their usability in practice. This paper introduces a simple, general, and efficient strategy for joint posterior inference of an unknown transformation and all regression model parameters. The proposed approach directly targets the posterior distribution of the transformation by linking it with the marginal distributions of the independent and dependent variables, and then deploys a Bayesian nonparametric model via the Bayesian bootstrap. Crucially, this approach delivers (1) joint posterior consistency under general conditions, including multiple model misspecifications, and (2) efficient Monte Carlo (not Markov chain Monte Carlo) inference for the transformation and all parameters for important special cases. These tools apply across a variety of data domains, including real-valued, integer-valued, compactly-supported, and positive data. Simulation studies and an empirical application demonstrate the effectiveness and efficiency of this strategy for semiparametric Bayesian analysis with linear models, quantile regression, and Gaussian processes.
['Bohan Wu', 'Daniel R. Kowal']
2023-06-08
null
null
null
null
['gaussian-processes']
['methodology']
[ 5.48197515e-02 -4.00825232e-01 -2.42581442e-01 -3.67679179e-01 -8.48602474e-01 -5.77515543e-01 4.69677210e-01 -6.06965926e-03 -2.51686275e-01 1.28810287e+00 -3.97848040e-01 -5.47384560e-01 -5.92735112e-01 -8.27704132e-01 -6.12457335e-01 -9.27623868e-01 -2.05045529e-02 6.48912966e-01 4.37777257e-03 4.21689749e-01 2.71005094e-01 6.25723600e-01 -1.21770251e+00 -8.47533226e-01 1.09071851e+00 8.94164681e-01 -1.43287510e-01 5.55995643e-01 2.77590811e-01 1.95068076e-01 -4.37022746e-01 -5.17281950e-01 2.08097789e-02 -6.97854087e-02 -2.39263661e-02 -1.55859113e-01 -1.63595945e-01 -4.45765883e-01 1.83827072e-01 1.19612598e+00 3.40632588e-01 2.93722242e-01 1.42259979e+00 -1.34846818e+00 -4.28343505e-01 3.49053800e-01 -1.07742333e+00 3.67443077e-02 2.82604080e-02 -1.49919719e-01 7.09287405e-01 -9.08368111e-01 -1.95960000e-01 1.33436072e+00 8.94654691e-01 -2.53923059e-01 -1.46864080e+00 -8.63449574e-01 -1.95356414e-01 -2.82403022e-01 -1.74667478e+00 -3.62273842e-01 3.38057339e-01 -7.60541737e-01 4.09904420e-01 1.46804526e-02 3.61523062e-01 9.39375937e-01 5.53598821e-01 1.83744490e-01 1.23472750e+00 -4.04873371e-01 5.10411203e-01 1.19636670e-01 4.13375527e-01 2.08995923e-01 9.35748935e-01 3.55450094e-01 -1.69668540e-01 -8.35783720e-01 1.12246895e+00 4.10259515e-02 -1.41080993e-03 -3.93734992e-01 -7.32747734e-01 1.09025455e+00 -6.70917392e-01 -1.85246423e-01 -6.30089760e-01 1.77235648e-01 2.25899562e-01 -8.53805169e-02 5.90244591e-01 -2.90960699e-01 -3.64436030e-01 -3.49809557e-01 -1.01456726e+00 2.64041394e-01 9.43578959e-01 1.05015767e+00 5.19111931e-01 1.57410741e-01 -3.65734130e-01 7.27549791e-01 6.47992373e-01 1.28950953e+00 1.09468319e-03 -7.52496779e-01 3.78466845e-01 -1.94714725e-01 8.43403816e-01 -7.82307267e-01 -1.70917362e-01 -2.82238036e-01 -1.01213312e+00 -1.10605821e-01 6.59823775e-01 -2.68066198e-01 -7.62928307e-01 1.70588732e+00 5.12769520e-01 1.84945881e-01 -1.38755262e-01 4.57972586e-01 5.31385168e-02 8.14759016e-01 2.30827227e-01 -7.14878142e-01 1.19491804e+00 -8.06092620e-02 -9.68848944e-01 5.23977391e-02 1.06620535e-01 -7.62880087e-01 1.03320670e+00 4.14104164e-01 -1.15961921e+00 -2.83652753e-01 -6.65506124e-01 3.37754637e-01 1.57885879e-01 1.56608611e-01 4.79911894e-01 1.00481117e+00 -5.41686118e-01 2.74505705e-01 -1.12680387e+00 -9.56470966e-02 1.93507731e-01 3.46516073e-01 3.63407545e-02 -1.71830490e-01 -1.06740069e+00 6.83607340e-01 2.38979697e-01 3.28478307e-01 -5.58463335e-01 -8.46020281e-01 -6.25884295e-01 3.85157853e-01 1.53794080e-01 -5.98320901e-01 1.19404364e+00 -3.36931676e-01 -1.61860991e+00 1.96629390e-01 -3.57325494e-01 -1.83611691e-01 6.16010487e-01 -1.98109880e-01 -3.27867419e-01 -1.11542821e-01 1.92385063e-01 -3.08017403e-01 1.15548563e+00 -8.18782210e-01 -5.55754721e-01 -3.79735917e-01 -5.69488227e-01 -1.02241658e-01 -9.41642299e-02 1.54085636e-01 -2.90993899e-01 -6.80267453e-01 1.16725191e-01 -6.66988134e-01 -1.04624182e-01 -2.87426502e-01 -3.60305786e-01 -5.14985211e-02 2.59303749e-01 -9.75305617e-01 1.16758692e+00 -2.42432189e+00 -3.82540554e-01 9.11190093e-01 -4.33228552e-01 -3.17417502e-01 3.13220471e-01 4.43519533e-01 2.25338936e-02 -2.77820956e-02 -5.27524769e-01 -5.98212183e-02 2.50045270e-01 1.06343895e-01 -5.97716093e-01 9.94384229e-01 -1.40328994e-02 4.93354201e-01 -5.75955033e-01 -5.48275828e-01 2.14274228e-01 3.66751790e-01 -3.89151901e-01 1.27484128e-01 3.39682102e-01 3.10540736e-01 -5.00860155e-01 6.05333090e-01 1.03613222e+00 -1.96686357e-01 2.34200090e-01 5.40788248e-02 -1.86774403e-01 -2.15011105e-01 -1.58911133e+00 4.64291215e-01 -2.86619872e-01 2.01869890e-01 1.43206254e-01 -1.25254846e+00 1.07173991e+00 1.88791558e-01 2.83599526e-01 -1.34808406e-01 1.02600619e-01 2.58647710e-01 -2.35753983e-01 -3.29370379e-01 2.07894504e-01 -6.11275971e-01 -2.74560541e-01 4.59767461e-01 4.30672895e-03 -2.52987921e-01 1.71667024e-01 -3.06536943e-01 5.90847611e-01 2.32461497e-01 9.92828488e-01 -3.21827412e-01 2.23082438e-01 -4.11246270e-01 6.34209573e-01 1.05845702e+00 6.32802099e-02 2.23562121e-01 7.58771062e-01 2.93594688e-01 -9.22714412e-01 -1.54970121e+00 -5.91903627e-01 8.89016986e-01 -5.29450141e-02 2.62366772e-01 -3.45628649e-01 -4.12714817e-02 3.58585685e-01 1.09583652e+00 -5.73148072e-01 -1.81429669e-01 -2.48729542e-01 -1.18706143e+00 2.66579241e-01 6.23600602e-01 7.59486482e-02 -3.96139741e-01 -1.98028490e-01 3.56292665e-01 -7.66816586e-02 -8.82509470e-01 -2.18876109e-01 1.89712018e-01 -1.06754911e+00 -1.03167939e+00 -7.59261370e-01 -1.19779274e-01 5.90694726e-01 -4.68110479e-02 5.85598946e-01 -6.00233316e-01 9.85817090e-02 5.04836440e-01 1.42211676e-01 -5.03862441e-01 -3.36161882e-01 -4.00889724e-01 7.96016082e-02 1.35055348e-01 9.70141515e-02 -5.15953958e-01 -2.20752046e-01 6.19416237e-01 -7.02316880e-01 -6.36802316e-01 5.79525709e-01 9.79402542e-01 6.07042730e-01 3.00361931e-01 8.07145953e-01 -6.22012377e-01 8.13100398e-01 -6.64418519e-01 -1.32285309e+00 6.17598057e-01 -4.20756847e-01 -1.24694817e-01 4.29870069e-01 -6.43551707e-01 -1.54876626e+00 -2.91680753e-01 4.14206952e-01 -2.27676049e-01 -2.70956326e-02 9.24793005e-01 -1.71240717e-01 3.52477223e-01 4.14733469e-01 1.04053378e-01 1.62291918e-02 -5.06908774e-01 5.97866960e-02 6.75320268e-01 7.79613674e-01 -1.00256073e+00 8.36360037e-01 5.62238455e-01 4.79376644e-01 -7.55622089e-01 -6.82364285e-01 -3.14365476e-01 -5.72452962e-01 1.37554020e-01 4.50809300e-01 -8.52902651e-01 -6.84115529e-01 5.10894358e-01 -9.25501227e-01 -1.30083859e-01 -2.17312157e-01 1.16493583e+00 -8.06964576e-01 4.95861620e-01 -1.79989219e-01 -1.43382621e+00 2.63728723e-02 -8.05824876e-01 7.21127570e-01 1.53374627e-01 -1.84925169e-01 -1.27443886e+00 1.52107954e-01 -2.31597554e-02 1.73425481e-01 1.32841930e-01 1.14066529e+00 -7.13679433e-01 -3.96020710e-02 -8.26940179e-01 -3.77950400e-01 3.68703097e-01 1.74527615e-01 6.60841107e-01 -5.39179146e-01 -3.21963429e-01 3.67031954e-02 8.73518586e-02 3.32688272e-01 1.18989754e+00 1.03612602e+00 -2.67223716e-01 -3.58620048e-01 3.01233709e-01 1.02493703e+00 2.63570935e-01 5.01440108e-01 -1.77606359e-01 2.02434487e-03 5.13585389e-01 1.03346360e+00 1.13157058e+00 1.65463790e-01 3.98568779e-01 1.09996125e-01 2.26680979e-01 7.56512284e-01 -1.21422119e-01 4.31891084e-01 4.91269261e-01 -1.06774338e-01 -1.92445666e-01 -7.87637472e-01 3.47892255e-01 -1.93664420e+00 -1.02399325e+00 -2.92828828e-01 2.92692494e+00 9.64706600e-01 -1.15890354e-01 1.97385982e-01 -9.74977463e-02 1.30951989e+00 -5.33631921e-01 -6.42414987e-01 -2.81500190e-01 1.42431587e-01 3.93826336e-01 7.23914444e-01 4.20276523e-01 -9.96701062e-01 4.01006579e-01 7.25749540e+00 1.05391073e+00 -5.79391479e-01 1.44893125e-01 4.76978660e-01 1.73931807e-01 -2.16254309e-01 2.69543022e-01 -9.90199566e-01 5.97452700e-01 1.06440818e+00 -4.68832582e-01 1.74821973e-01 7.13014662e-01 6.38936341e-01 -4.71572667e-01 -8.72571886e-01 7.17737138e-01 -5.35598099e-01 -6.29608691e-01 -3.17197621e-01 3.13998014e-01 8.07557523e-01 -5.13782203e-01 1.55202180e-01 3.72750312e-01 5.07563412e-01 -1.00514102e+00 7.33320594e-01 1.01174319e+00 8.74388993e-01 -1.02863955e+00 1.03394413e+00 3.73041958e-01 -7.36790955e-01 -2.03526884e-01 -5.91871738e-01 -4.84528467e-02 3.60318482e-01 1.05380917e+00 -6.61632776e-01 4.32920545e-01 5.03245056e-01 3.38653564e-01 -1.46133766e-01 1.40076888e+00 -2.20567301e-01 1.01525116e+00 -7.20320404e-01 4.53584529e-02 -2.39398777e-01 -7.02798665e-01 2.75779456e-01 1.05968726e+00 9.43308711e-01 -3.53745632e-02 -1.59839988e-01 8.82555604e-01 5.44983208e-01 1.71537519e-01 -3.44458103e-01 -9.25307870e-02 1.01324463e+00 8.50736976e-01 -6.68508053e-01 -3.50079924e-01 -5.24406731e-01 5.77425174e-02 -6.28318489e-02 9.20290053e-01 -1.20490742e+00 -4.47033823e-01 2.60600626e-01 -5.71130402e-03 5.77535689e-01 -2.41379023e-01 -5.43482065e-01 -8.24755847e-01 -1.11510493e-01 -5.13493598e-01 5.25908828e-01 -4.75316018e-01 -1.55150402e+00 -4.73923057e-01 8.31977189e-01 -9.86267209e-01 -7.55093277e-01 -4.40661103e-01 -6.49661958e-01 1.20889354e+00 -1.09062338e+00 -8.39929879e-01 1.17308207e-01 6.55804157e-01 -2.46072978e-01 -2.66790893e-02 5.59651017e-01 -1.13776654e-01 -7.63864934e-01 4.81192827e-01 8.22011471e-01 -2.92162716e-01 7.17648566e-01 -9.64112937e-01 -1.12651601e-01 7.47584939e-01 -5.56076944e-01 7.77826250e-01 9.94577706e-01 -8.71942282e-01 -9.38662291e-01 -8.43428016e-01 4.82551485e-01 1.52112752e-01 1.09171212e+00 -1.63615286e-01 -7.98841953e-01 6.93189561e-01 -4.67980742e-01 -7.45736212e-02 8.81374478e-01 2.95893252e-01 -8.52231905e-02 -4.64940891e-02 -1.32271469e+00 5.92263103e-01 1.98549882e-01 -1.02897339e-01 -3.43893200e-01 1.74606025e-01 1.28984764e-01 6.10815436e-02 -1.20612705e+00 6.23446584e-01 8.83863807e-01 -8.07763636e-01 9.34562504e-01 -5.28724492e-01 -9.04753432e-02 -1.35718286e-01 -4.50758040e-01 -9.81894970e-01 -1.57237634e-01 -7.37219751e-01 -3.18029314e-01 1.41840518e+00 1.91479042e-01 -1.08914351e+00 1.10593148e-01 7.64376879e-01 1.13093369e-01 -3.32650125e-01 -1.37637877e+00 -1.10706377e+00 2.62964576e-01 -6.01943910e-01 5.64282000e-01 6.10923707e-01 -3.54590833e-01 -1.75262347e-01 -4.51544404e-01 4.78847980e-01 9.86794114e-01 1.66664258e-01 9.03763533e-01 -1.49008596e+00 -3.29723030e-01 -2.66472220e-01 -1.96422264e-01 -9.38259363e-01 3.28338027e-01 -1.80561692e-01 1.56308800e-01 -1.08364272e+00 4.78760988e-01 -3.23924869e-01 1.33449305e-02 2.47887313e-01 -1.83277056e-01 -1.91736460e-01 -5.51616609e-01 2.22745225e-01 7.16548786e-02 7.47462749e-01 9.47949767e-01 3.33560020e-01 -1.44785136e-01 7.55198956e-01 -5.64390123e-01 8.84420395e-01 6.79517567e-01 -6.75705433e-01 -4.72658992e-01 4.50896680e-01 2.41615459e-01 4.62366372e-01 6.99984491e-01 -3.74820232e-01 -1.28186718e-01 -7.84766912e-01 3.96046251e-01 -8.54554653e-01 3.05508643e-01 -7.77121842e-01 3.72022361e-01 2.82695651e-01 4.43206839e-02 -3.11550230e-01 8.77913758e-02 9.22237039e-01 -1.16442144e-01 -6.24405384e-01 9.72356319e-01 5.02196848e-01 1.33375287e-01 7.21843913e-02 -5.12796223e-01 4.54107523e-02 9.56079304e-01 -2.84092784e-01 -1.11541152e-01 -7.75476396e-01 -9.07927990e-01 -2.68437620e-02 1.12344958e-01 -2.38320231e-01 4.04465795e-01 -1.23421419e+00 -6.91781998e-01 2.16068253e-01 -2.67348379e-01 -1.77261069e-01 2.60839403e-01 1.31095695e+00 -5.46668135e-02 3.74122530e-01 1.41284227e-01 -6.78589404e-01 -8.66803408e-01 2.86912352e-01 3.65683697e-02 -2.39255920e-01 -9.74566787e-02 2.48509675e-01 4.08039838e-01 -3.16179454e-01 -3.86284143e-02 -5.51531076e-01 1.18643522e-01 8.39809552e-02 2.36638471e-01 9.26394463e-01 -2.62537718e-01 -2.78473079e-01 -1.82036474e-01 4.44220394e-01 2.58780003e-01 -4.09163445e-01 1.04114652e+00 -4.62058008e-01 -2.47810364e-01 9.46284890e-01 8.00379872e-01 -7.87100382e-03 -1.52482557e+00 -1.87557071e-01 -3.52416001e-02 -5.07359564e-01 -9.69857350e-02 -4.12674099e-01 -3.85226846e-01 8.13129961e-01 1.39411122e-01 2.47572616e-01 7.88821578e-01 -2.42096618e-01 -8.68730396e-02 1.91295043e-01 1.80978924e-01 -9.32052791e-01 -2.91442454e-01 3.36826354e-01 7.66345322e-01 -8.50370288e-01 2.72880882e-01 -5.39949417e-01 -3.12786669e-01 1.08482647e+00 1.70182884e-01 5.63644804e-02 1.05243647e+00 2.75949121e-01 -5.46528041e-01 2.42287114e-01 -4.70499516e-01 3.09989482e-01 3.77809823e-01 7.71018684e-01 2.68858582e-01 2.36851543e-01 -4.28071201e-01 1.08380818e+00 -8.59064981e-02 1.98001731e-02 6.00139201e-01 7.82663822e-01 -4.11806494e-01 -6.12594068e-01 -8.21573615e-01 5.97425401e-01 -5.87916791e-01 -7.13720545e-02 4.28963840e-01 1.19706190e+00 -7.32672989e-01 9.45883870e-01 3.01013082e-01 5.96333086e-01 1.61584422e-01 1.39391765e-01 4.28903461e-01 -3.28834116e-01 4.35851097e-01 6.54781044e-01 -6.31840229e-02 -5.61362728e-02 -1.68107286e-01 -1.20618594e+00 -8.53108943e-01 -4.55907345e-01 -7.15375900e-01 1.92926779e-01 6.49671376e-01 1.07338214e+00 -1.77094378e-02 4.78170626e-02 7.84529388e-01 -6.06544077e-01 -1.48751330e+00 -1.16912019e+00 -1.16308534e+00 -2.52102345e-01 -8.63975007e-03 -1.14515972e+00 -6.64886832e-01 4.99925436e-03]
[7.069799423217773, 4.2025017738342285]
2b8686be-cbf8-4ea8-893c-889ee3ac4f90
lost-in-the-middle-how-language-models-use
2307.03172
null
https://arxiv.org/abs/2307.03172v1
https://arxiv.org/pdf/2307.03172v1.pdf
Lost in the Middle: How Language Models Use Long Contexts
While recent language models have the ability to take long contexts as input, relatively little is known about how well the language models use longer context. We analyze language model performance on two tasks that require identifying relevant information within their input contexts: multi-document question answering and key-value retrieval. We find that performance is often highest when relevant information occurs at the beginning or end of the input context, and significantly degrades when models must access relevant information in the middle of long contexts. Furthermore, performance substantially decreases as the input context grows longer, even for explicitly long-context models. Our analysis provides a better understanding of how language models use their input context and provides new evaluation protocols for future long-context models.
['Percy Liang', 'Fabio Petroni', 'Michele Bevilacqua', 'Ashwin Paranjape', 'John Hewitt', 'Kevin Lin', 'Nelson F. Liu']
2023-07-06
null
null
null
null
['retrieval', 'question-answering']
['methodology', 'natural-language-processing']
[ 1.63125515e-01 -3.75388980e-01 -4.37745243e-01 -3.26024860e-01 -1.15931571e+00 -1.06489182e+00 8.19763184e-01 9.16123748e-01 -9.50038612e-01 4.52718705e-01 5.94776869e-01 -8.02533925e-01 -2.28551894e-01 -5.72346210e-01 -4.57147300e-01 1.80426450e-03 3.82748805e-02 4.39948708e-01 6.25135243e-01 -3.04048657e-01 3.95779669e-01 1.49333149e-01 -1.28725302e+00 7.65059292e-01 5.44670939e-01 4.88582671e-01 3.90035063e-01 1.03975368e+00 -8.51996601e-01 8.09205472e-01 -7.81795502e-01 -5.37404343e-02 -1.19575649e-01 -6.97315410e-02 -1.39914155e+00 -7.64406502e-01 7.47682750e-01 -2.80212671e-01 -1.91042498e-02 6.37424052e-01 2.73074657e-01 3.07209074e-01 4.56392735e-01 -7.35212564e-01 -6.76359177e-01 7.06203759e-01 1.37338877e-01 9.01857078e-01 7.45883644e-01 2.12345570e-01 1.42098784e+00 -9.57163632e-01 7.23464906e-01 1.16443026e+00 3.25182438e-01 2.90144414e-01 -1.09878755e+00 -1.53741300e-01 6.32132828e-01 1.66350290e-01 -1.15951812e+00 -2.99896538e-01 2.78256983e-01 -3.85269940e-01 1.63256085e+00 4.51272517e-01 1.63685650e-01 7.02011824e-01 2.83484936e-01 6.80903673e-01 8.30304742e-01 -6.11522019e-01 8.34992602e-02 -1.18659690e-01 8.88378739e-01 1.63769260e-01 1.52046263e-01 -7.88536966e-02 -5.26737928e-01 -6.54575169e-01 7.43287206e-02 -4.33539338e-02 3.28383222e-02 4.65975493e-01 -1.08585954e+00 6.22472525e-01 1.74823865e-01 8.91050994e-01 -2.63855785e-01 3.12969804e-01 2.80146062e-01 6.09285712e-01 2.10599914e-01 9.03258622e-01 -5.24665534e-01 -3.56646299e-01 -8.68924677e-01 5.83422363e-01 1.03454292e+00 8.96556795e-01 7.64501691e-01 -5.61998606e-01 -5.88145077e-01 8.42034996e-01 1.33975655e-01 5.63709557e-01 3.38507295e-01 -1.03874087e+00 6.15835786e-01 7.13881195e-01 4.25646842e-01 -6.68059111e-01 -2.76744783e-01 -2.95716673e-01 5.64600453e-02 -5.09054422e-01 8.46742213e-01 -4.21454981e-02 -6.14243865e-01 1.89262474e+00 5.13776690e-02 -1.90464795e-01 2.17425957e-01 6.01538241e-01 6.34743929e-01 9.36805308e-01 6.37506485e-01 -1.73143551e-01 1.57317746e+00 -7.94549525e-01 -6.24827862e-01 -8.63053203e-01 9.96600211e-01 -1.20005989e+00 1.69740868e+00 -1.40174538e-01 -1.06062472e+00 -4.58566576e-01 -5.89980304e-01 -6.36632264e-01 -5.80808520e-01 -4.64483023e-01 4.19665575e-01 1.96683943e-01 -1.22266996e+00 1.60778955e-01 -3.34210724e-01 -5.31193912e-01 -3.73591214e-01 5.10746315e-02 -2.66510304e-02 -4.01700675e-01 -1.43901241e+00 9.82272625e-01 2.16569185e-01 -1.23283774e-01 -4.74726439e-01 -6.68981433e-01 -6.43045127e-01 1.02053680e-01 4.84684736e-01 -6.75656497e-01 1.78431571e+00 -6.62689090e-01 -6.69075131e-01 7.02262223e-01 -6.02317274e-01 -1.93840802e-01 -1.30078986e-01 -5.53114951e-01 -3.85781020e-01 1.29128203e-01 -9.81475115e-02 3.99820060e-01 3.64861608e-01 -9.96140838e-01 -4.86966729e-01 -1.90996915e-01 6.56287193e-01 3.83928448e-01 -2.95346051e-01 5.58826506e-01 -6.92549467e-01 -5.50731480e-01 -1.06279887e-01 -7.16667891e-01 -1.37255460e-01 -2.10599020e-01 1.99270546e-02 -4.81364816e-01 6.45015240e-01 -5.70760369e-01 1.95051682e+00 -1.81300592e+00 -2.45322376e-01 1.78866938e-01 1.93475168e-02 2.21651435e-01 -4.16458994e-01 1.02098596e+00 1.35929495e-01 5.13504148e-01 -8.34741741e-02 4.09267060e-02 2.43231412e-02 3.61019194e-01 -6.79731786e-01 -4.87493932e-01 -9.99063700e-02 1.12684608e+00 -9.92999673e-01 -5.41147709e-01 -2.68153638e-01 1.68564625e-03 -7.22087502e-01 2.34275877e-01 -6.54921293e-01 -2.30575755e-01 -5.53204536e-01 4.51399148e-01 -2.50262506e-02 -4.94249314e-01 4.23077904e-02 9.50063094e-02 2.09244876e-03 7.59696126e-01 -8.99561226e-01 1.38018024e+00 -6.32431328e-01 5.52806258e-01 2.13305697e-01 -2.26069435e-01 3.99798512e-01 2.76870698e-01 3.65279503e-02 -9.81084883e-01 -3.66069883e-01 2.54434764e-01 1.40218779e-01 -8.47841084e-01 8.93440902e-01 -9.96960774e-02 -3.02705914e-01 9.02258635e-01 -5.66934586e-01 -1.89965963e-01 5.26578546e-01 6.72333121e-01 1.19225359e+00 -4.58151549e-01 1.83229059e-01 -3.02996814e-01 6.26084864e-01 -4.01267856e-02 6.84038037e-03 1.27490044e+00 -1.25235900e-01 2.00549006e-01 5.53873539e-01 -2.69310147e-01 -8.88241529e-01 -9.16131973e-01 -1.65167497e-03 1.97442174e+00 1.29590660e-01 -9.26421463e-01 -5.66929996e-01 -8.91213298e-01 2.18591690e-01 9.04274106e-01 -4.02628154e-01 -2.03710094e-01 -7.55064666e-01 -2.38652721e-01 5.65204799e-01 7.63048887e-01 3.18487436e-02 -1.12659216e+00 -2.92087317e-01 3.16834241e-01 -3.73400897e-01 -1.08400881e+00 -8.97650063e-01 1.03664167e-01 -9.16869462e-01 -9.61144269e-01 -3.57754767e-01 -7.19603658e-01 3.92611176e-01 1.04716770e-01 1.50966144e+00 6.67196453e-01 1.93259984e-01 8.81297529e-01 -2.93263048e-01 -1.12786576e-01 -3.70387286e-01 2.96870738e-01 -4.18567806e-01 -6.94212317e-01 7.02660382e-01 -1.85516849e-01 -6.02703989e-01 4.38827991e-01 -1.22013724e+00 -2.96552479e-01 3.91434938e-01 7.52144694e-01 4.49416459e-01 -6.61859989e-01 7.13262200e-01 -1.00169158e+00 1.17690778e+00 -7.46304989e-01 -2.71512568e-01 7.13764369e-01 -6.50856674e-01 5.74078619e-01 6.17226064e-01 -4.12481457e-01 -8.80176187e-01 -6.66618645e-01 -4.26104546e-01 2.32334137e-01 -1.59864157e-01 9.70257640e-01 2.46308178e-01 4.22481209e-01 9.05418813e-01 -7.84181058e-02 -3.60999495e-01 -6.05004549e-01 3.53169143e-01 5.81755936e-01 2.43087873e-01 -1.20203233e+00 3.99206430e-01 -4.35628854e-02 -5.10120690e-01 -7.44946182e-01 -7.63615787e-01 -8.24573696e-01 -1.60662904e-01 4.57603484e-02 6.97331488e-01 -5.54651976e-01 -3.66996378e-01 1.02294169e-01 -1.08404481e+00 -5.88186622e-01 -1.97760627e-01 2.35508233e-01 7.79147893e-02 5.86231470e-01 -7.24092662e-01 -7.23932147e-01 -4.61252987e-01 -1.06983566e+00 9.67609048e-01 2.52332222e-02 -7.91011751e-01 -1.12993670e+00 -4.05886620e-02 2.11170942e-01 6.74273074e-01 -6.04607046e-01 1.63019598e+00 -9.19165492e-01 -5.90647936e-01 -2.19035074e-01 -1.52722280e-02 5.81609085e-02 -7.79564083e-02 -2.24831372e-01 -7.63719022e-01 -2.25590393e-01 -4.10790324e-01 -4.52909112e-01 8.50823283e-01 1.17133399e-02 1.09011543e+00 -4.60325211e-01 -2.97736943e-01 -1.10062547e-02 1.21135497e+00 2.13469788e-01 1.99358448e-01 5.74055687e-03 2.45267466e-01 6.17078185e-01 5.79351187e-01 -1.84062898e-01 5.31491041e-01 5.59307992e-01 -1.40356332e-01 3.99546593e-01 1.84792001e-02 -2.74502248e-01 2.83207536e-01 8.45896125e-01 5.34376800e-01 -3.86817724e-01 -1.31937969e+00 7.87335753e-01 -1.57887888e+00 -7.96584666e-01 1.15537681e-01 2.23348761e+00 1.06488800e+00 4.14955556e-01 -1.19427010e-01 -3.19078356e-01 2.66068906e-01 3.91680479e-01 -6.26836896e-01 -6.47590518e-01 1.08325751e-02 1.20283097e-01 2.70298496e-02 1.17301297e+00 -6.60053432e-01 1.04412222e+00 8.29616642e+00 5.79819381e-01 -9.61371005e-01 -3.04823592e-02 4.43797916e-01 -2.83377290e-01 -9.26750720e-01 3.33512843e-01 -9.52082694e-01 3.14064890e-01 1.16313636e+00 -2.31812820e-01 2.48612180e-01 6.79406285e-01 6.88137561e-02 -6.99795902e-01 -1.39686310e+00 6.47105217e-01 -7.26764724e-02 -1.12801194e+00 3.39798421e-01 -1.22683093e-01 3.56878757e-01 -9.49754473e-03 -1.22640260e-01 5.53364754e-01 1.30730346e-01 -8.88938129e-01 3.42320025e-01 5.86982429e-01 6.23789907e-01 -4.78986979e-01 3.70361894e-01 6.72345698e-01 -1.12636590e+00 -3.25746298e-01 -2.56388456e-01 -3.03246766e-01 1.50123924e-01 2.90740907e-01 -8.88355136e-01 -1.05518224e-02 2.96623737e-01 8.45529959e-02 -1.03389835e+00 7.85098910e-01 -1.59721166e-01 9.82822895e-01 -6.04186535e-01 -4.35207188e-01 2.05908880e-01 2.45294809e-01 5.98960221e-01 1.42985725e+00 5.22883162e-02 3.08629483e-01 3.51494491e-01 5.64125180e-01 -6.86802194e-02 2.42705002e-01 -5.89412212e-01 -2.56528378e-01 7.78382778e-01 7.69454360e-01 -3.63907903e-01 -6.71733558e-01 -5.63306034e-01 4.01217669e-01 4.30749267e-01 8.30818534e-01 -2.59400636e-01 -4.13633168e-01 6.02647245e-01 3.47546041e-01 -2.86641300e-01 -4.79972720e-01 -1.87105551e-01 -1.14396942e+00 3.64062786e-01 -8.75469744e-01 9.92343307e-01 -9.67246056e-01 -1.07224905e+00 4.04782832e-01 5.29510617e-01 -3.75459909e-01 -6.76597953e-01 -6.11037493e-01 -5.85894406e-01 1.11117685e+00 -1.48133254e+00 -7.82115340e-01 6.27243593e-02 3.76092315e-01 5.82909048e-01 1.99821189e-01 9.77162957e-01 2.46631518e-01 -1.34457871e-01 7.02498436e-01 1.50450403e-02 7.56896194e-03 9.15430725e-01 -1.11063135e+00 4.17615473e-01 9.16635215e-01 2.53178060e-01 1.39027429e+00 5.30582190e-01 -7.02596903e-01 -1.38942003e+00 -5.44570923e-01 1.58139181e+00 -9.65698004e-01 5.90269089e-01 -2.43199527e-01 -1.42725730e+00 6.80237651e-01 4.33502138e-01 -1.36547670e-01 9.82148290e-01 5.27366698e-01 -6.46215737e-01 -1.73635617e-01 -7.08256066e-01 7.31718779e-01 8.56747508e-01 -1.26737916e+00 -1.17602861e+00 1.00317582e-01 9.92668808e-01 -1.20426178e-01 -7.22114325e-01 2.19767123e-01 6.13017499e-01 -3.85785997e-01 9.81099546e-01 -8.79380345e-01 1.92816824e-01 -2.26012155e-01 -4.38864440e-01 -8.05316627e-01 -9.86108780e-02 -4.38340455e-01 -4.23892975e-01 9.10542130e-01 8.63990605e-01 -4.85922694e-01 3.07409555e-01 1.06523168e+00 1.07522751e-03 -8.72849762e-01 -5.15312195e-01 -4.23052818e-01 5.18889129e-01 -7.43961334e-01 6.38361275e-01 7.37194240e-01 2.00570524e-01 5.36352158e-01 4.28091645e-01 7.52038285e-02 6.31215516e-03 3.74641418e-01 3.08044434e-01 -8.91406298e-01 -2.56234258e-01 -6.27298355e-01 3.47347289e-01 -1.46827328e+00 9.02035683e-02 -9.37032223e-01 -4.11451869e-02 -1.67839241e+00 1.27042234e-01 -6.40437484e-01 -3.78988713e-01 6.11906230e-01 -9.23873901e-01 -1.61264077e-01 1.47311136e-01 2.60906965e-01 -8.33502889e-01 -4.29842025e-02 9.92959678e-01 1.12767190e-01 -6.81264773e-02 -1.36784062e-01 -8.47255528e-01 5.21107495e-01 4.70306009e-01 -3.51264894e-01 -7.04369426e-01 -9.14000392e-01 8.63520205e-01 6.26622885e-02 2.22418994e-01 -8.11661184e-01 4.87983704e-01 -6.13016605e-01 2.16586113e-01 -7.21782386e-01 1.87252522e-01 -6.39835119e-01 -3.41596782e-01 1.23163059e-01 -1.00826204e+00 7.12210715e-01 4.69590962e-01 3.54599684e-01 -2.10503101e-01 -5.50336361e-01 3.92297626e-01 -1.85524270e-01 -8.78677249e-01 1.35038048e-02 -7.28280485e-01 8.16173732e-01 5.16069710e-01 8.87109526e-03 -3.48361582e-01 -4.54526484e-01 -8.25455904e-01 5.45492172e-01 4.50509578e-01 6.48214221e-01 4.92794096e-01 -1.04609084e+00 -3.30480993e-01 2.60481667e-02 4.51310694e-01 -1.11256510e-01 2.43569724e-02 2.95300037e-01 -3.15094143e-01 6.11566901e-01 6.83350921e-01 -5.36615491e-01 -1.24109089e+00 5.90283275e-01 3.21363360e-01 -5.88846982e-01 -1.33394450e-01 8.30652356e-01 -2.56102923e-02 -3.52026403e-01 3.91219854e-01 -8.50305378e-01 -1.55518755e-01 1.53411895e-01 7.50072718e-01 -1.17011636e-01 1.69090912e-01 -3.94730300e-01 -3.74845415e-01 5.80001771e-01 -3.98871273e-01 -4.61034328e-01 6.64370775e-01 -2.42257074e-01 -1.39181077e-01 5.85521281e-01 1.26720035e+00 1.38407990e-01 -6.61926031e-01 -6.06010139e-01 4.97245371e-01 -3.21298182e-01 -1.30096838e-01 -1.08110082e+00 -1.10167027e-01 9.78206575e-01 1.52243108e-01 3.31878275e-01 9.21542287e-01 1.70659751e-01 8.30268443e-01 1.10844934e+00 2.28940859e-01 -1.18229032e+00 2.25685865e-01 1.25905228e+00 9.86431420e-01 -8.11372221e-01 -1.61745265e-01 6.47213757e-02 -3.63425195e-01 7.80770183e-01 8.10647428e-01 3.74230713e-01 5.76642632e-01 2.65080035e-01 2.66338468e-01 -2.26902649e-01 -1.26202285e+00 -1.25200510e-01 5.00637949e-01 7.99319595e-02 7.73643076e-01 -3.19473922e-01 -4.45409238e-01 5.82316995e-01 -2.27549136e-01 -3.19369316e-01 1.91287160e-01 1.34400523e+00 -6.68894112e-01 -1.36727095e+00 -3.59530717e-01 5.58030367e-01 -8.03820729e-01 -6.42383575e-01 -6.94886684e-01 5.30824244e-01 -2.84861296e-01 1.15669656e+00 1.87166147e-02 2.11335495e-02 4.51370984e-01 1.04942453e+00 2.07762763e-01 -1.00065005e+00 -1.12183678e+00 1.67235022e-03 2.74619371e-01 -6.04241610e-01 5.46295159e-02 -4.35316443e-01 -1.29914618e+00 -1.31642535e-01 -2.71157980e-01 5.88072956e-01 3.74087900e-01 1.09903383e+00 5.83388090e-01 9.67407506e-03 1.27202094e-01 8.06032270e-02 -6.24314666e-01 -9.80519891e-01 5.97159304e-02 5.76957703e-01 7.06470132e-01 -6.01283051e-02 -2.83491135e-01 -6.39462546e-02]
[11.242130279541016, 8.06555461883545]
f792d43e-cbce-4135-9eae-9669cc7bcd42
rpnet-a-deep-learning-approach-for-robust-r
2004.08103
null
https://arxiv.org/abs/2004.08103v1
https://arxiv.org/pdf/2004.08103v1.pdf
RPnet: A Deep Learning approach for robust R Peak detection in noisy ECG
Automatic detection of R-peaks in an Electrocardiogram signal is crucial in a multitude of applications including Heart Rate Variability (HRV) analysis and Cardio Vascular Disease(CVD) diagnosis. Although there have been numerous approaches that have successfully addressed the problem, there has been a notable dip in the performance of these existing detectors on ECG episodes that contain noise and HRV Irregulates. On the other hand, Deep Learning(DL) based methods have shown to be adept at modelling data that contain noise. In image to image translation, Unet is the fundamental block in many of the networks. In this work, a novel application of the Unet combined with Inception and Residual blocks is proposed to perform the extraction of R-peaks from an ECG. Furthermore, the problem formulation also robustly deals with issues of variability and sparsity of ECG R-peaks. The proposed network was trained on a database containing ECG episodes that have CVD and was tested against three traditional ECG detectors on a validation set. The model achieved an F1 score of 0.9837, which is a substantial improvement over the other beat detectors. Furthermore, the model was also evaluated on three other databases. The proposed network achieved high F1 scores across all datasets which established its generalizing capacity. Additionally, a thorough analysis of the model's performance in the presence of different levels of noise was carried out.
['Vignesh R', 'Jayaraj Joseph', 'Sricharan Vijayarangan', 'Preejith SP', 'Mohansankar Sivaprakasam', 'Balamurali Murugesan']
2020-04-17
null
null
null
null
['heart-rate-variability']
['medical']
[ 3.07281822e-01 -1.59692004e-01 8.26628581e-02 -1.91662878e-01 -6.76856518e-01 -1.88390553e-01 8.67278054e-02 1.58730686e-01 -2.10056484e-01 6.52565479e-01 -4.59281132e-02 -5.54301776e-02 -3.09964389e-01 -4.56484586e-01 -6.31884187e-02 -6.34031951e-01 -4.97213632e-01 -6.91506490e-02 -2.18131170e-01 -1.83372498e-01 -3.84183452e-02 5.68797410e-01 -1.17057633e+00 1.68870091e-01 5.27212918e-01 1.16932094e+00 -2.10990131e-01 8.52709293e-01 4.88320082e-01 5.23050249e-01 -9.07919586e-01 -7.36599639e-02 3.32807899e-01 -7.54077256e-01 -2.58651674e-01 -1.10378072e-01 2.69132882e-01 -1.23055369e-01 -2.12356210e-01 6.86308265e-01 1.28542268e+00 -2.05063865e-01 4.60177302e-01 -8.29307497e-01 -1.73601702e-01 3.02574813e-01 -5.86068034e-01 8.10600758e-01 1.95473462e-01 1.13554664e-01 5.98650098e-01 -8.40229869e-01 3.88506055e-01 7.97530949e-01 1.30798340e+00 1.02023527e-01 -1.39425862e+00 -6.17692709e-01 -6.36178970e-01 3.29719819e-02 -1.57153475e+00 -2.52724409e-01 9.09008324e-01 -3.47739130e-01 8.76767397e-01 2.84394860e-01 7.80453920e-01 8.91837120e-01 4.64010745e-01 2.68551141e-01 1.14528370e+00 -4.98156101e-01 1.73717700e-02 8.52740258e-02 1.77094191e-01 3.01698327e-01 3.09708178e-01 3.10499400e-01 -3.88977826e-01 -1.52396277e-01 9.00338888e-01 -1.42869860e-01 -2.90081382e-01 1.23914180e-03 -1.06373894e+00 6.76093876e-01 2.61405408e-01 5.08270264e-01 -5.22188842e-01 -1.15060918e-01 7.69836247e-01 4.20052230e-01 2.84518421e-01 5.31560600e-01 -4.46115553e-01 -1.29735649e-01 -1.21729898e+00 1.90192461e-01 7.89852321e-01 3.71210545e-01 3.14275846e-02 4.97891635e-01 -3.88816655e-01 8.01768959e-01 9.81521159e-02 4.14193809e-01 4.98913825e-01 -7.33638823e-01 2.73756772e-01 4.30234164e-01 -2.72940278e-01 -1.47604370e+00 -8.71456683e-01 -9.82462645e-01 -1.48011029e+00 4.51425575e-02 2.49307185e-01 -3.87482852e-01 -9.23000395e-01 1.59206593e+00 7.34689459e-02 1.51501432e-01 6.91532120e-02 1.03118086e+00 1.12189913e+00 4.07233149e-01 1.16646588e-02 -5.17767668e-01 1.34462190e+00 -9.89989787e-02 -9.63918090e-01 -7.59890825e-02 2.02383086e-01 -6.56685889e-01 4.67837453e-01 4.91051733e-01 -8.88276577e-01 -9.93485093e-01 -1.15690708e+00 2.82290518e-01 6.37180507e-02 2.73306996e-01 2.48217583e-01 9.11099017e-01 -9.72394049e-01 7.73531258e-01 -5.56013346e-01 -4.76974100e-01 4.48509783e-01 3.96206617e-01 -1.44868791e-01 2.31947914e-01 -1.38976765e+00 8.93658459e-01 3.40058714e-01 4.34475124e-01 -6.17005587e-01 -4.78802115e-01 -6.20308757e-01 -9.86661296e-03 3.78484093e-02 -5.74305475e-01 6.70587838e-01 -9.25435543e-01 -1.01969397e+00 8.98679614e-01 1.23399235e-01 -8.67688656e-01 6.91798925e-01 -1.14573263e-01 -8.39874506e-01 2.23528489e-01 -8.36230218e-02 8.17184597e-02 9.93240714e-01 -7.06506848e-01 -3.72178137e-01 -3.88688505e-01 -4.06194597e-01 -6.64549619e-02 5.11847027e-02 -3.71158868e-02 -1.13552988e-01 -8.95851195e-01 3.27040851e-01 -8.83301854e-01 -3.03609818e-02 -3.41417730e-01 -3.48106176e-01 1.41980454e-01 4.85292226e-01 -8.10204208e-01 1.47033525e+00 -2.29715633e+00 -1.36620477e-01 4.80316877e-01 3.38437825e-01 5.90431750e-01 1.08858228e-01 5.38240492e-01 -3.51951480e-01 1.63360924e-01 -2.73462296e-01 1.76477447e-01 -2.73051769e-01 1.58296108e-01 -4.93268780e-02 6.55584455e-01 3.30438048e-01 7.69993722e-01 -3.82233948e-01 -3.96002561e-01 1.89139366e-01 7.49897003e-01 -1.83522061e-01 5.05310856e-02 4.52984571e-01 7.68019676e-01 -5.11648618e-02 4.50207293e-01 4.94425297e-01 -1.13126114e-01 3.13637733e-01 -4.64705497e-01 4.19397242e-02 -1.32027775e-01 -1.30707777e+00 1.37865293e+00 1.21901639e-01 7.10328698e-01 -1.56751081e-01 -1.15610504e+00 1.22146416e+00 8.54144394e-01 8.30976129e-01 -8.03897858e-01 1.89270213e-01 2.87015200e-01 5.57884395e-01 -7.62573242e-01 -1.28773838e-01 -3.57363254e-01 1.58672959e-01 1.63549140e-01 2.20916439e-02 2.04163641e-01 1.03857808e-01 -2.42115304e-01 1.21753001e+00 -2.43941974e-02 6.52212501e-01 -3.17269236e-01 7.27366984e-01 -1.69231012e-01 8.04235816e-01 1.02938187e+00 -4.72251803e-01 9.50780571e-01 4.59245384e-01 -8.06576312e-01 -8.08652461e-01 -8.38583708e-01 -5.66598177e-01 3.18683535e-01 -1.67112246e-01 -3.89481515e-01 -4.74075496e-01 -2.21321121e-01 -9.12065059e-02 1.30912825e-01 -5.35688281e-01 -2.00068429e-01 -7.51223564e-01 -1.13818192e+00 1.05300689e+00 5.71620047e-01 6.07538521e-01 -1.32059383e+00 -1.09861898e+00 4.90361631e-01 -2.56666720e-01 -1.08940089e+00 6.62573949e-02 3.89890224e-01 -1.10271192e+00 -1.29650044e+00 -7.55852401e-01 -6.74956739e-01 1.51696950e-01 -2.59829760e-01 1.27814877e+00 6.88965321e-02 -9.40310061e-01 2.82457441e-01 -1.84535459e-01 -7.29417026e-01 -3.84307206e-01 6.66591106e-03 1.23598367e-01 -3.76034975e-02 4.18023795e-01 -6.76574409e-01 -8.36551309e-01 9.83392149e-02 -7.17403173e-01 -4.56906855e-01 5.85476875e-01 9.42291617e-01 5.91036499e-01 2.57343352e-01 1.19310009e+00 -7.20649481e-01 9.15660143e-01 -2.70032763e-01 -2.67653912e-01 -3.17378968e-01 -7.95632482e-01 -4.09560055e-01 5.24772584e-01 -2.67203242e-01 -4.56891090e-01 -1.40290447e-02 -3.74941498e-01 -2.69530505e-01 -1.73962608e-01 5.53367317e-01 2.20109150e-01 1.52884930e-01 8.38523865e-01 1.01162121e-01 2.46256918e-01 -2.94811934e-01 -2.70217121e-01 5.09672165e-01 5.95852733e-01 -1.81891605e-01 5.92281103e-01 2.06674978e-01 3.37402701e-01 -1.12937450e+00 -7.12810874e-01 -4.98306394e-01 -6.95300043e-01 -9.51971784e-02 9.28330600e-01 -9.32319641e-01 -5.39739668e-01 5.16439140e-01 -8.48458827e-01 1.86940074e-01 -3.35020691e-01 5.70783615e-01 -3.36469591e-01 3.56293052e-01 -5.62497199e-01 -1.05946684e+00 -8.09734344e-01 -7.80486941e-01 5.02858818e-01 1.60628095e-01 -5.03141463e-01 -8.25864136e-01 1.99744150e-01 8.75765234e-02 5.80893755e-01 9.53505337e-01 8.66034448e-01 -8.38730097e-01 5.99208772e-02 -4.23919499e-01 -1.01504512e-02 6.76326275e-01 2.49965236e-01 -3.08231801e-01 -1.18928790e+00 -4.41542804e-01 4.34108734e-01 1.35810366e-02 6.21558547e-01 7.57208407e-01 6.28110051e-01 4.56384495e-02 -3.26752067e-02 5.30031621e-01 1.45293903e+00 3.74524176e-01 9.20404315e-01 2.53029615e-01 4.26211745e-01 3.67422670e-01 2.76038557e-01 4.65542614e-01 -7.25653395e-02 5.44677377e-01 2.95229286e-01 -6.97499275e-01 -1.37635708e-01 3.85680765e-01 1.34012088e-01 7.18753517e-01 -3.79593670e-01 1.64317638e-01 -9.26815331e-01 3.96194637e-01 -1.58159935e+00 -1.12916648e+00 -4.97468978e-01 2.13664889e+00 5.87867320e-01 2.91335821e-01 4.12249625e-01 7.06079423e-01 6.21408761e-01 -4.24026931e-03 -6.22348249e-01 -5.30818701e-01 -2.81887025e-01 6.55727684e-01 2.37354651e-01 -1.20653421e-01 -1.36711740e+00 5.76947965e-02 6.89668798e+00 2.37556528e-02 -1.24599063e+00 -1.22203536e-01 6.93412602e-01 1.07298858e-01 6.75251961e-01 -4.84717041e-01 -4.18337673e-01 2.78984815e-01 1.09148109e+00 -9.59121361e-02 8.14119913e-03 4.81522202e-01 6.17969573e-01 -3.08010504e-02 -1.01392984e+00 1.30368161e+00 1.09255232e-01 -1.00157726e+00 -5.30122697e-01 -5.82439899e-02 4.25557017e-01 -7.01899007e-02 -9.53397825e-02 4.16679800e-01 -7.94088662e-01 -1.32362401e+00 2.23897561e-01 5.52629650e-01 8.92150521e-01 -9.19744730e-01 1.23920298e+00 3.31332773e-01 -1.08708060e+00 -2.32883275e-01 -2.22010508e-01 -2.41563320e-01 -6.24067262e-02 8.16144049e-01 -8.83249283e-01 6.90440953e-01 7.62821674e-01 7.14733899e-01 -5.44102967e-01 1.10492051e+00 1.08463921e-01 7.48597145e-01 -1.60481676e-01 5.29277146e-01 -2.67416209e-01 -6.23751879e-02 7.11000204e-01 1.47993600e+00 2.57313132e-01 1.08194001e-01 1.74454346e-01 7.71260738e-01 1.24208622e-01 1.68367982e-01 -7.27896333e-01 2.06286266e-01 1.13502160e-01 1.21356237e+00 -6.69962227e-01 -2.89531708e-01 -3.25239539e-01 4.97128487e-01 -3.64798695e-01 2.62178332e-01 -9.82718706e-01 -4.09753203e-01 3.86592716e-01 4.61710513e-01 1.85261562e-01 1.73170432e-01 -4.64419037e-01 -6.88203216e-01 1.65519938e-01 -1.22033048e+00 4.60613579e-01 -3.54795605e-01 -1.28029823e+00 7.01416671e-01 -1.95931301e-01 -1.43695617e+00 -2.06100449e-01 -2.19233796e-01 -5.82336485e-01 1.12460041e+00 -1.33685637e+00 -6.20032132e-01 -4.24907476e-01 4.96348143e-01 3.37112665e-01 -1.51388615e-01 1.08626449e+00 7.27874219e-01 -5.52711666e-01 5.10932922e-01 -2.82422543e-01 4.07519907e-01 6.78633690e-01 -1.13311696e+00 1.55699119e-01 1.02191377e+00 1.14814803e-01 6.83425665e-01 7.01586604e-01 -5.80074251e-01 -1.11982763e+00 -1.17122900e+00 6.46462977e-01 -5.05592339e-02 5.68746477e-02 3.94771323e-02 -9.62775588e-01 3.18469703e-01 1.11105755e-01 1.53984576e-01 7.07792878e-01 -1.51204944e-01 1.37430564e-01 -3.38825941e-01 -1.11745548e+00 5.92719726e-02 5.53762197e-01 -2.62749851e-01 -7.12972045e-01 4.13715636e-04 -1.92282632e-01 -4.71808046e-01 -1.14906967e+00 7.55155683e-01 8.30342948e-01 -1.13775611e+00 1.10315931e+00 -3.68288934e-01 1.46930039e-01 -3.21020395e-01 6.05483353e-02 -1.04080117e+00 -3.94535244e-01 -6.70095444e-01 -1.97483703e-01 1.21568012e+00 1.67126760e-01 -5.68469644e-01 4.32142049e-01 2.80708045e-01 -1.93902757e-02 -6.60401702e-01 -8.77616346e-01 -7.35530615e-01 -3.02205652e-01 -3.97832155e-01 -6.32213056e-02 8.74164045e-01 -2.07612082e-01 5.45864940e-01 -6.49070978e-01 2.38923822e-02 6.87436640e-01 -8.42766464e-02 5.57555914e-01 -1.64209914e+00 -2.90602028e-01 -1.14024781e-01 -7.13960648e-01 -2.01812416e-01 -4.68498796e-01 -1.00825381e+00 -9.35294256e-02 -1.55389261e+00 7.24973530e-02 -2.91357815e-01 -7.07869411e-01 2.50195771e-01 -1.83485672e-01 6.19410872e-01 1.07136585e-01 2.64638364e-01 3.54331173e-02 -1.14949875e-01 1.00652766e+00 1.16027016e-02 -6.15263700e-01 2.63062358e-01 -5.46805501e-01 7.07365870e-01 1.10811794e+00 -6.69848144e-01 -3.98525983e-01 2.95041680e-01 2.67448537e-02 3.68821353e-01 4.77916509e-01 -1.40096927e+00 -1.32556096e-01 5.88451803e-01 9.63159859e-01 -6.81246519e-01 2.85088390e-01 -7.23435700e-01 5.91571987e-01 7.49402344e-01 -2.42911354e-01 3.69795471e-01 2.95106739e-01 4.12360787e-01 -3.24375153e-01 1.61479220e-01 1.05195761e+00 -1.75854892e-01 -3.22332829e-01 7.52496393e-03 -3.29816759e-01 3.68344992e-01 9.06316936e-01 -4.68887061e-01 1.29619762e-01 -1.29010558e-01 -1.03591144e+00 -1.57331467e-01 -1.49784133e-01 1.99977949e-01 7.19916582e-01 -1.14292252e+00 -9.59734201e-01 3.79964590e-01 -3.62643078e-02 -2.33455479e-01 2.72989899e-01 1.46964693e+00 -5.64360261e-01 3.61324489e-01 -3.93196672e-01 -1.00570333e+00 -1.47155344e+00 3.14938307e-01 6.71527803e-01 -3.56467754e-01 -1.27727473e+00 3.64332557e-01 -1.87344640e-01 1.35022193e-01 3.70794326e-01 -3.49148273e-01 -4.75266010e-01 2.92975098e-01 4.54592526e-01 4.74975705e-01 2.62828380e-01 -7.13664830e-01 -5.55389941e-01 5.73869944e-01 2.19532803e-01 2.36732781e-01 1.31334829e+00 3.28453514e-03 2.41665784e-02 6.48428440e-01 9.56447661e-01 -2.52523690e-01 -7.27076173e-01 -3.15433070e-02 1.98352218e-01 -1.03201486e-01 -2.12920588e-02 -9.41218615e-01 -1.28261030e+00 9.37843382e-01 1.37097168e+00 4.36346859e-01 1.35339558e+00 -6.59149706e-01 6.21125996e-01 2.58590654e-02 2.30011325e-02 -1.00866461e+00 -1.24583123e-02 2.74061739e-01 6.70842111e-01 -1.13359916e+00 2.63584852e-01 -8.28949362e-03 -6.12113297e-01 1.38400018e+00 9.43003520e-02 -3.16841125e-01 7.34413326e-01 2.70372242e-01 3.80266845e-01 -4.03271079e-01 -2.98562944e-01 -1.72088817e-02 3.13356906e-01 7.45880544e-01 8.07699382e-01 -9.25729275e-02 -6.79057360e-01 5.08826494e-01 7.49032199e-02 4.41044450e-01 5.43618381e-01 8.78918171e-01 -1.36679068e-01 -7.42072463e-01 -4.85846221e-01 6.05873168e-01 -1.18816042e+00 -2.38428917e-02 -1.38794512e-01 9.63787556e-01 2.19729543e-01 1.05916321e+00 -3.84415835e-01 -1.89927503e-01 5.78518212e-01 3.28233808e-01 2.26275370e-01 -5.15865862e-01 -1.16484690e+00 3.97824138e-01 4.58181538e-02 -4.20575798e-01 -5.21280468e-01 -4.81502265e-01 -1.15229154e+00 1.91942424e-01 -1.99523896e-01 -1.33801892e-01 4.74611253e-01 6.52159214e-01 3.14047664e-01 1.10073316e+00 5.17742634e-01 -4.77448016e-01 -5.37886620e-01 -1.02708638e+00 -8.97476256e-01 5.73642194e-01 5.51411569e-01 -3.39637220e-01 -3.24927062e-01 2.31766924e-01]
[14.29918098449707, 3.2790777683258057]
a57ab889-011a-4d3d-b075-8a7a496da8df
eggs-eigen-gap-guided-search-making-subspace
2107.12183
null
https://arxiv.org/abs/2107.12183v4
https://arxiv.org/pdf/2107.12183v4.pdf
A Simple Approach to Automated Spectral Clustering
The performance of spectral clustering heavily relies on the quality of affinity matrix. A variety of affinity-matrix-construction (AMC) methods have been proposed but they have hyperparameters to determine beforehand, which requires strong experience and leads to difficulty in real applications, especially when the inter-cluster similarity is high and/or the dataset is large. In addition, we often need to choose different AMC methods for different datasets, which still depends on experience. To solve these two challenging problems, in this paper, we present a simple yet effective method for automated spectral clustering. First, we propose to find the most reliable affinity matrix via grid search or Bayesian optimization among a set of candidates given by different AMC methods with different hyperparameters, where the reliability is quantified by the \textit{relative-eigen-gap} of graph Laplacian introduced in this paper. Second, we propose a fast and accurate AMC method based on least squares representation and thresholding and prove its effectiveness theoretically. Finally, we provide a large-scale extension for the automated spectral clustering method, of which the time complexity is linear with the number of data points. Extensive experiments of natural image clustering show that our method is more versatile, accurate, and efficient than baseline methods.
['Mingbo Zhao', 'Zhao Zhang', 'Haijun Zhang', 'Yiheng Tu', 'Jicong Fan']
2021-07-23
null
null
null
null
['image-clustering']
['computer-vision']
[ 2.47948915e-02 -6.70263767e-01 4.26113196e-02 -2.31402367e-02 -8.83202374e-01 -5.23741722e-01 3.88392173e-02 8.41920450e-02 -3.82345647e-01 4.10139710e-01 -2.15557039e-01 -2.24011093e-02 -4.86024320e-01 -4.82306123e-01 -3.53336245e-01 -1.16273570e+00 -6.57886788e-02 5.63612878e-01 5.84021151e-01 -6.93411231e-02 3.59589785e-01 1.84806362e-01 -1.30714953e+00 -1.57884240e-01 1.29814816e+00 8.10248375e-01 4.19034094e-01 3.10829192e-01 -2.02503115e-01 8.06842670e-02 -3.40253592e-01 -1.83657706e-01 2.48088300e-01 -6.58724785e-01 -6.60814404e-01 2.76186913e-01 -2.02221312e-02 4.44255948e-01 1.81985646e-01 1.47343969e+00 5.52090406e-01 6.29590079e-02 9.22286749e-01 -1.38450634e+00 -3.54115754e-01 5.21090925e-01 -1.12035787e+00 -3.62601578e-02 3.94789726e-02 -1.98742524e-01 8.15505505e-01 -1.07848370e+00 3.54613960e-01 1.04850066e+00 7.69533217e-01 2.02210292e-01 -1.66232383e+00 -7.60192752e-01 9.84174460e-02 5.02888203e-01 -1.95955491e+00 -2.38678947e-01 9.85975981e-01 -5.79479575e-01 1.01009645e-01 3.00780267e-01 6.74520314e-01 6.79954112e-01 -4.47810471e-01 5.70102453e-01 1.14539206e+00 -4.99139458e-01 3.90741259e-01 1.15522221e-01 9.77785215e-02 8.86358678e-01 3.05296838e-01 -4.78062004e-01 -1.53441504e-01 -5.03074944e-01 6.30298853e-01 4.26857211e-02 -4.92492199e-01 -7.01719642e-01 -1.30593753e+00 8.49367201e-01 2.98085600e-01 5.09830058e-01 -1.59129664e-01 -7.60431141e-02 8.86297151e-02 5.41195879e-03 1.25922576e-01 3.12263280e-01 -8.48784745e-02 9.34992582e-02 -9.36456442e-01 -2.14457482e-01 5.83082736e-01 7.60492325e-01 1.10001457e+00 -3.61517042e-01 2.97935963e-01 1.17256784e+00 2.94422686e-01 6.28247738e-01 5.63554227e-01 -7.74095237e-01 2.67669857e-01 7.67995059e-01 1.98168936e-03 -1.74706328e+00 -4.41075265e-01 -1.64447963e-01 -1.35494578e+00 -8.28944743e-02 4.53132510e-01 6.51770011e-02 -5.19607544e-01 1.56449318e+00 5.34078360e-01 3.79516065e-01 -2.49034941e-01 1.12511277e+00 3.99043888e-01 6.81007683e-01 -2.79100716e-01 -7.45482206e-01 1.10921824e+00 -7.20065355e-01 -7.36063004e-01 4.25657518e-02 4.67201173e-01 -9.25633430e-01 1.20169675e+00 4.95378017e-01 -7.31041074e-01 -3.97078305e-01 -9.64434385e-01 4.52366799e-01 -2.32711241e-01 3.88047487e-01 4.96174037e-01 6.00816309e-01 -9.67054009e-01 3.27979296e-01 -7.63243616e-01 -4.40917075e-01 -3.94187719e-02 4.34475899e-01 -2.48024747e-01 1.58802152e-01 -9.62520242e-01 3.82952005e-01 6.00042880e-01 1.44196540e-01 -1.92318425e-01 -1.49074107e-01 -3.37692738e-01 -8.08985829e-02 6.04152203e-01 -2.29093045e-01 5.49894929e-01 -9.09640133e-01 -1.37317705e+00 4.84281957e-01 -9.68979374e-02 -1.59781761e-02 2.79158473e-01 1.87390149e-01 -2.84864813e-01 3.44536066e-01 2.61393934e-01 3.94772589e-01 8.62466276e-01 -1.44586182e+00 -3.83661002e-01 -4.62344140e-01 -4.84047025e-01 1.81657642e-01 -4.81761098e-01 -1.56744331e-01 -1.05839860e+00 -6.37603521e-01 6.94978416e-01 -1.14013374e+00 -4.44377422e-01 -1.61022291e-01 -3.58604848e-01 -2.96922624e-01 7.63008118e-01 -4.41671759e-01 1.62651753e+00 -2.46186376e+00 3.59277159e-01 8.48931909e-01 1.89093754e-01 1.31419644e-01 5.41396774e-02 4.79014784e-01 4.54007275e-03 1.34549022e-01 -5.96330583e-01 7.17295557e-02 -1.05070114e-01 5.12423888e-02 -2.84636244e-02 6.20412946e-01 -2.53378600e-01 3.43600899e-01 -8.68596554e-01 -1.21657598e+00 9.62508097e-02 3.98793340e-01 -3.25809032e-01 -5.93697578e-02 2.76174396e-01 2.42507339e-01 -5.62439799e-01 5.31310976e-01 8.23034644e-01 -6.45358026e-01 3.73351455e-01 -6.18493915e-01 -9.50121060e-02 -4.65546250e-01 -1.87337315e+00 1.57645917e+00 -4.33076806e-02 1.53638482e-01 3.00520301e-01 -1.23468173e+00 8.98474932e-01 9.60172415e-02 6.58321619e-01 -9.40881521e-02 1.09180301e-01 5.21607637e-01 -1.01781920e-01 -2.98828930e-01 1.12286024e-01 -4.50922698e-02 1.42431542e-01 4.52347010e-01 -3.23012233e-01 5.06649874e-02 5.61581433e-01 4.55623537e-01 8.29273462e-01 -3.34457397e-01 3.28131825e-01 -6.13015354e-01 8.33156645e-01 -4.66468818e-02 7.48816013e-01 4.84180063e-01 -2.63574481e-01 6.77090406e-01 4.87731695e-01 1.63014606e-02 -7.17608392e-01 -7.43545949e-01 -2.11271495e-01 6.62123501e-01 6.47337973e-01 -5.87658107e-01 -9.85418141e-01 -6.59691751e-01 -2.08800539e-01 1.43061783e-02 -3.69983137e-01 -2.15838514e-02 -4.23684388e-01 -1.14982808e+00 8.02828595e-02 3.03102821e-01 5.57288945e-01 -5.29049337e-01 -4.20526937e-02 1.52508616e-01 -5.09915888e-01 -9.47876990e-01 -7.99619853e-01 -7.07726628e-02 -8.71274769e-01 -1.23365378e+00 -6.30827069e-01 -9.69345331e-01 8.82014275e-01 7.60757208e-01 7.24816263e-01 4.07218158e-01 -1.77605599e-01 2.43346483e-01 -5.03046632e-01 1.55103639e-01 3.19878422e-02 2.39589572e-01 1.82921290e-01 5.02465248e-01 2.88739383e-01 -7.36601889e-01 -6.68769836e-01 7.89321184e-01 -8.28237295e-01 4.45785113e-02 7.64215708e-01 8.34492803e-01 1.00237954e+00 4.70606029e-01 4.38474387e-01 -7.43118346e-01 6.09352529e-01 -2.34710336e-01 -8.54726136e-01 4.11214054e-01 -8.07780683e-01 1.28818527e-01 7.01826155e-01 -5.43175876e-01 -6.24655783e-01 5.65980852e-01 2.52647847e-01 -5.44578969e-01 2.20892072e-01 6.41311169e-01 -2.32140720e-01 -2.39647344e-01 6.47852719e-01 3.29165220e-01 1.27225429e-01 -4.68008429e-01 2.98107654e-01 7.10412979e-01 5.15335858e-01 -6.50662363e-01 9.89997089e-01 4.91710216e-01 3.61824892e-02 -8.75551939e-01 -5.92719555e-01 -7.97512949e-01 -7.97727525e-01 -3.12570512e-01 8.54381204e-01 -5.56745589e-01 -8.60818446e-01 3.31880629e-01 -8.06061387e-01 -9.32873413e-02 4.59301740e-01 6.40210688e-01 -2.96170443e-01 8.28447461e-01 -4.16312426e-01 -7.79628992e-01 -1.21613406e-01 -1.29538178e+00 9.55894530e-01 3.13673355e-02 2.04977289e-01 -8.32834363e-01 3.62868011e-02 2.07554147e-01 4.97807525e-02 1.80109829e-01 7.83395767e-01 -2.30359033e-01 -5.17044544e-01 -7.33331814e-02 -3.82219285e-01 5.74929044e-02 2.82586128e-01 2.59940803e-01 -2.76349723e-01 -4.35583919e-01 -1.61793441e-01 -2.45812051e-02 6.66438580e-01 2.98817396e-01 1.40233302e+00 -2.25486413e-01 -6.30891263e-01 4.91630167e-01 1.50653088e+00 1.71503633e-01 4.53117937e-01 1.42728552e-01 7.15424538e-01 4.11065251e-01 6.61701977e-01 3.14873785e-01 1.86984167e-01 8.17859471e-01 3.05086840e-02 -1.03218831e-01 2.69683182e-01 1.74055863e-02 1.14944100e-01 1.39609635e+00 -1.15470625e-01 1.19705915e-01 -9.84621525e-01 4.03953999e-01 -2.28678560e+00 -9.71377313e-01 -5.39679110e-01 2.37334514e+00 1.07908630e+00 -7.39203915e-02 3.26204211e-01 3.70102793e-01 1.15598845e+00 -2.13731110e-01 -4.17700946e-01 3.74154419e-01 -1.08398542e-01 -1.93412200e-01 4.16436940e-01 5.11915267e-01 -1.12320781e+00 8.08422625e-01 6.23475933e+00 1.25730658e+00 -8.44772160e-01 -1.02466550e-02 4.68791306e-01 2.18527898e-01 -3.45647242e-03 7.65255615e-02 -5.69895804e-01 8.39307964e-01 4.44661140e-01 -1.72411501e-02 5.63584447e-01 7.66239762e-01 2.08446346e-02 -2.96601534e-01 -6.16312444e-01 1.47216213e+00 3.57185379e-02 -8.92476439e-01 -2.80274540e-01 7.66523555e-02 7.64401853e-01 -4.34579611e-01 -9.69969034e-02 -1.25089049e-01 2.06934869e-01 -4.65275347e-01 2.96326160e-01 3.56791079e-01 6.07709587e-01 -8.46119881e-01 5.91403306e-01 4.06880766e-01 -1.48328269e+00 9.70340818e-02 -6.05922580e-01 3.26633662e-01 -1.09872870e-01 1.05875289e+00 -3.93623233e-01 4.46509272e-01 7.90270984e-01 5.79482615e-01 -6.67096019e-01 1.14573157e+00 -6.88403472e-02 5.67762196e-01 -5.97273171e-01 -2.54880458e-01 1.17220171e-01 -9.28886712e-01 1.95469812e-01 1.07370138e+00 4.97516721e-01 3.46145272e-01 6.05723262e-01 6.50542736e-01 1.69995651e-01 7.29919195e-01 -3.09610337e-01 3.63629684e-02 7.47490942e-01 1.60016572e+00 -1.30020165e+00 -2.86093026e-01 -3.09368670e-01 1.00651014e+00 3.49303335e-01 3.38509768e-01 -8.35623384e-01 -5.43999434e-01 7.04090893e-02 6.26956858e-03 1.54987946e-01 -4.17269260e-01 -6.82932362e-02 -1.15581012e+00 1.13920063e-01 -8.66441965e-01 5.74571311e-01 -4.70013678e-01 -1.51655293e+00 4.18884754e-01 -6.42177910e-02 -1.33373582e+00 1.32377282e-01 -4.74814147e-01 -3.95467937e-01 5.10034561e-01 -9.40806210e-01 -5.48041701e-01 -5.01455247e-01 8.31967056e-01 2.30620056e-01 -7.52596259e-02 5.38510203e-01 5.46800077e-01 -9.12309825e-01 5.04258692e-01 4.66155529e-01 1.15322284e-01 8.97920012e-01 -1.30828202e+00 -3.01528126e-01 8.40268135e-01 1.86191276e-01 6.40446365e-01 6.13606632e-01 -5.69164693e-01 -1.36426437e+00 -8.49508643e-01 4.35195833e-01 -7.93581381e-02 8.04770947e-01 -3.57037693e-01 -1.06943142e+00 1.35120869e-01 -6.57831971e-03 -1.10491447e-01 8.53630662e-01 1.95150554e-01 -1.95088312e-01 -4.03267711e-01 -8.35201085e-01 6.71923459e-01 9.77217019e-01 -3.08110476e-01 -1.70236483e-01 4.69166994e-01 3.17192197e-01 1.19129635e-01 -8.04904103e-01 4.09228623e-01 2.40811735e-01 -1.10038602e+00 8.45681131e-01 3.69470753e-02 -2.26545200e-01 -9.32119489e-01 -6.37098476e-02 -1.16230810e+00 -5.79871655e-01 -7.02185094e-01 1.65358678e-01 1.33678615e+00 3.56180042e-01 -6.07175887e-01 7.27314353e-01 1.85574189e-01 1.61048979e-01 -7.17003047e-01 -7.49468029e-01 -8.43857884e-01 -3.36191088e-01 -1.44586399e-01 4.78874385e-01 1.20607305e+00 1.39395446e-01 6.12430274e-01 -3.59049320e-01 3.35447460e-01 8.97537529e-01 4.19117510e-01 7.47749269e-01 -1.54559386e+00 -2.36668482e-01 -5.16348720e-01 -3.66354793e-01 -9.79000807e-01 -2.09829565e-02 -7.59112239e-01 2.13514432e-01 -1.45379269e+00 5.82153201e-01 -6.75861239e-01 -1.89129278e-01 3.09736669e-01 -4.72796887e-01 2.63188660e-01 -7.54429996e-02 7.17377067e-01 -9.08130348e-01 5.31760335e-01 9.71214473e-01 -6.62955865e-02 -3.86988074e-01 -1.36427984e-01 -5.46921015e-01 6.93257511e-01 6.29344165e-01 -4.27544922e-01 -5.12215436e-01 6.20447695e-02 2.39239901e-01 -5.87898046e-02 2.76125856e-02 -1.08947432e+00 4.18158382e-01 -2.89829075e-01 2.10224539e-01 -7.40525901e-01 1.43510371e-01 -1.04537487e+00 3.25368494e-01 3.20174485e-01 5.76998033e-02 1.75452769e-01 -3.64744276e-01 7.77403891e-01 -2.57292300e-01 -3.14094663e-01 9.63755608e-01 -3.79609019e-02 -5.18568993e-01 3.26385081e-01 -1.78993911e-01 -3.39443460e-02 1.02853882e+00 -2.11127356e-01 6.85106590e-02 -3.37389588e-01 -6.46404922e-01 4.44747239e-01 6.48025453e-01 -1.28377050e-01 5.14613986e-01 -1.69904709e+00 -5.35400152e-01 6.16504289e-02 1.07593797e-01 -1.09883219e-01 1.21075727e-01 1.36342096e+00 -3.69853526e-01 -1.34209460e-02 1.83830023e-01 -9.41415012e-01 -1.38955212e+00 8.89376223e-01 1.03330262e-01 3.87480222e-02 -3.47470850e-01 4.88491654e-01 1.97964102e-01 -2.29790106e-01 2.11785302e-01 1.12883359e-01 -1.13330796e-01 2.01085865e-01 3.51885378e-01 4.45826381e-01 -1.31933600e-01 -6.84995234e-01 -4.26419944e-01 1.13567233e+00 1.39358357e-01 -3.10421251e-02 1.00849366e+00 -3.02903742e-01 -4.69221860e-01 4.46693599e-01 1.19755423e+00 4.05669585e-03 -9.06812608e-01 -2.46334538e-01 1.55949220e-01 -5.08752227e-01 -8.23403746e-02 -2.11396456e-01 -1.19466555e+00 7.44180202e-01 7.79829085e-01 5.88160396e-01 1.38188159e+00 1.76407881e-02 6.19738638e-01 5.39126694e-01 4.52950537e-01 -1.42352760e+00 2.65698135e-01 4.64613400e-02 6.47952676e-01 -1.22933352e+00 1.69147789e-01 -8.11981082e-01 -6.64068699e-01 9.70059812e-01 5.64615965e-01 1.09260440e-01 8.12276781e-01 -5.13217263e-02 1.20340735e-01 -2.15550497e-01 -2.20030457e-01 -3.41448009e-01 3.36601019e-01 4.06956404e-01 2.82620460e-01 6.04203204e-03 -8.23042095e-01 4.04567122e-01 -8.24788120e-03 -4.76832092e-01 2.82950342e-01 5.06259978e-01 -6.88472927e-01 -1.22269285e+00 -6.63454592e-01 2.59581923e-01 -1.38245538e-01 3.66579369e-02 -5.00427783e-01 5.66941500e-01 -1.89517234e-02 1.04125881e+00 -3.07096064e-01 -4.23785836e-01 3.10891997e-02 -6.69232830e-02 2.54031807e-01 -1.84716597e-01 7.90643692e-02 5.96274912e-01 -3.20113033e-01 -4.18779165e-01 -7.48781741e-01 -7.23771930e-01 -1.41370571e+00 -1.33218110e-01 -7.35292137e-01 7.03683734e-01 4.70679879e-01 5.83199441e-01 3.83570045e-01 -1.35802761e-01 8.67598295e-01 -6.21802270e-01 -3.93677354e-01 -7.43527472e-01 -8.57512712e-01 6.08478129e-01 -2.09376708e-01 -1.00139308e+00 -5.17184734e-01 9.88921151e-02]
[7.604463577270508, 4.675248146057129]
d102fa34-89e4-4be8-bb16-2750ec2fec3d
end-to-end-audio-visual-scene-aware-dialog
1806.08409
null
http://arxiv.org/abs/1806.08409v2
http://arxiv.org/pdf/1806.08409v2.pdf
End-to-End Audio Visual Scene-Aware Dialog using Multimodal Attention-Based Video Features
Dialog systems need to understand dynamic visual scenes in order to have conversations with users about the objects and events around them. Scene-aware dialog systems for real-world applications could be developed by integrating state-of-the-art technologies from multiple research areas, including: end-to-end dialog technologies, which generate system responses using models trained from dialog data; visual question answering (VQA) technologies, which answer questions about images using learned image features; and video description technologies, in which descriptions/captions are generated from videos using multimodal information. We introduce a new dataset of dialogs about videos of human behaviors. Each dialog is a typed conversation that consists of a sequence of 10 question-and-answer(QA) pairs between two Amazon Mechanical Turk (AMT) workers. In total, we collected dialogs on roughly 9,000 videos. Using this new dataset for Audio Visual Scene-aware dialog (AVSD), we trained an end-to-end conversation model that generates responses in a dialog about a video. Our experiments demonstrate that using multimodal features that were developed for multimodal attention-based video description enhances the quality of generated dialog about dynamic scenes (videos). Our dataset, model code and pretrained models will be publicly available for a new Video Scene-Aware Dialog challenge.
['Irfan Essa', 'Raphael Gontijo Lopes', 'Tim K. Marks', 'Vincent Cartillier', 'Huda Alamri', 'Takaaki Hori', 'Jue Wang', 'Dhruv Batra', 'Gordon Wichern', 'Devi Parikh', 'Chiori Hori', 'Anoop Cherian', 'Abhishek Das']
2018-06-21
null
null
null
null
['video-description']
['computer-vision']
[ 3.93279716e-02 -1.22633064e-02 3.00548404e-01 -8.98271263e-01 -1.00426519e+00 -8.42682660e-01 7.65248299e-01 -3.43046993e-01 -2.31474385e-01 5.26888311e-01 8.58659327e-01 1.06825896e-01 6.76889181e-01 -3.41481507e-01 -7.23527491e-01 -2.94823855e-01 4.33916271e-01 7.68170416e-01 3.61006230e-01 -4.39667791e-01 2.98872381e-01 1.56933784e-01 -1.41115487e+00 1.30823028e+00 7.16791078e-02 1.09935737e+00 4.84703153e-01 1.31714845e+00 -5.07945716e-01 1.65115023e+00 -8.41481745e-01 -5.29848456e-01 -1.61767736e-01 -5.31831026e-01 -1.05924165e+00 5.63782930e-01 7.14520276e-01 -1.07657325e+00 -6.47983015e-01 5.71401596e-01 1.91237882e-01 4.69311863e-01 6.21959448e-01 -1.71822095e+00 -8.38672280e-01 3.35088283e-01 8.92747846e-03 5.72665501e-03 1.29166412e+00 8.72849464e-01 6.98550582e-01 -1.20083058e+00 1.07953441e+00 1.98216307e+00 1.98130161e-02 1.02171540e+00 -1.07815337e+00 -4.21396136e-01 4.39425446e-02 5.36669195e-01 -9.62980926e-01 -6.95457518e-01 6.26401842e-01 -5.70303857e-01 1.06081367e+00 3.13557327e-01 5.51480770e-01 1.79977655e+00 1.93554834e-02 1.09939766e+00 6.52274728e-01 -7.49777928e-02 1.45961881e-01 4.52450395e-01 -1.97251379e-01 5.70661187e-01 -8.88666689e-01 -2.40128323e-01 -9.45776105e-01 -2.23660748e-02 8.15762341e-01 4.59587760e-02 -2.40132481e-01 -3.80448937e-01 -1.52688038e+00 1.09117949e+00 3.59251827e-01 4.60749045e-02 -2.00498149e-01 1.59432694e-01 4.69877988e-01 2.90143847e-01 6.39103279e-02 3.78449768e-01 -7.49627352e-02 -4.33298171e-01 -3.01989943e-01 3.93271118e-01 1.04788744e+00 1.29755402e+00 7.50791848e-01 -4.79687378e-02 -7.49736309e-01 7.40768194e-01 3.02315056e-01 6.99197054e-01 3.18304092e-01 -1.79027414e+00 5.22891283e-01 7.69097328e-01 3.74820560e-01 -8.73804569e-01 -1.62444249e-01 1.07281685e+00 -2.99652845e-01 -1.05718955e-01 6.85966671e-01 -2.90047437e-01 -6.08403504e-01 1.48056149e+00 4.67799634e-01 -1.22183733e-01 2.88317770e-01 1.35197389e+00 1.57894588e+00 1.18987226e+00 1.80579334e-01 2.61733476e-02 1.72268462e+00 -1.32428396e+00 -9.94687617e-01 -2.02596366e-01 2.78424919e-01 -9.93691742e-01 1.55992353e+00 1.01040043e-01 -1.21207905e+00 -8.13814223e-01 -3.77657354e-01 -4.69007432e-01 -4.33468282e-01 -4.33997735e-02 1.98586032e-01 2.17303649e-01 -1.29203165e+00 -3.51364940e-01 -2.98245966e-01 -7.68698752e-01 -5.39041981e-02 -1.58106610e-02 -5.86946905e-01 -1.33865535e-01 -1.07068264e+00 7.21752942e-01 -1.41712278e-01 -1.86159939e-01 -1.63090205e+00 -3.59764218e-01 -1.09087157e+00 2.38296036e-02 5.14865458e-01 -7.00273573e-01 1.83552647e+00 -1.24005234e+00 -1.83780801e+00 8.63952577e-01 -2.85092443e-01 -3.79210442e-01 2.57607490e-01 -2.13443056e-01 -1.83781773e-01 1.04938722e+00 7.12500094e-03 1.42064595e+00 9.19161260e-01 -1.50081611e+00 -5.56022465e-01 -1.37395620e-01 5.70268512e-01 4.46087211e-01 -2.69953191e-01 1.55410677e-01 -5.63966930e-01 -2.10409999e-01 -5.46738386e-01 -1.04400849e+00 7.00279996e-02 1.76287532e-01 -2.97067910e-01 -2.79793084e-01 1.39807534e+00 -6.78826630e-01 3.68077368e-01 -2.13756371e+00 2.08246246e-01 -5.42317092e-01 2.89455205e-02 -1.02309912e-01 -4.82636720e-01 9.15167212e-01 3.93499613e-01 -2.60602981e-01 3.61305058e-01 -5.85078418e-01 2.77245771e-02 1.39262244e-01 -6.53881609e-01 -7.43429875e-03 2.26021945e-01 9.41255748e-01 -8.61571908e-01 -7.06973016e-01 7.79392600e-01 5.63800335e-01 -5.54668963e-01 1.11803675e+00 -8.89390886e-01 7.95766532e-01 -3.22907418e-01 5.50926507e-01 2.71232605e-01 -3.48984629e-01 -2.56682336e-02 -3.83888453e-01 1.08123049e-01 -1.33820429e-01 -5.02953827e-01 1.94971848e+00 -5.71156502e-01 1.14260721e+00 3.09659690e-01 -4.74966049e-01 7.21146882e-01 7.03475177e-01 3.82398248e-01 -3.80774409e-01 1.74384668e-01 -4.14556026e-01 -6.76650882e-01 -1.30408382e+00 7.70297587e-01 2.83713073e-01 -3.96023989e-01 4.81592208e-01 3.81646514e-01 -5.80088139e-01 1.96040228e-01 7.92075276e-01 1.01783538e+00 -5.50047010e-02 -1.90024450e-01 3.73003513e-01 5.83111882e-01 6.30683482e-01 -2.27254540e-01 7.14148343e-01 -2.85478503e-01 7.16237128e-01 4.45236355e-01 -5.65060198e-01 -9.94426250e-01 -1.29349101e+00 4.89652634e-01 1.59626377e+00 3.19678575e-01 -2.68025339e-01 -7.89999366e-01 -5.55517137e-01 -2.89233923e-01 7.67957747e-01 -4.17807192e-01 1.96147248e-01 -3.95541757e-01 4.17291373e-01 2.92682678e-01 4.23395216e-01 7.49598384e-01 -1.60472429e+00 -6.92090333e-01 -1.52968407e-01 -8.82140040e-01 -1.83317554e+00 -9.25498664e-01 -5.97400188e-01 -4.11036253e-01 -1.02642488e+00 -9.57568407e-01 -9.37024057e-01 5.17618537e-01 5.60890257e-01 1.21902740e+00 -2.99631268e-01 -2.05163002e-01 1.46460724e+00 -6.72878504e-01 -5.80781251e-02 -7.40112662e-01 -4.18397218e-01 -1.56078607e-01 3.05521488e-01 4.79604602e-01 6.60428554e-02 -6.24861181e-01 7.02933669e-01 -8.10402215e-01 3.95368338e-01 1.76384881e-01 6.82153583e-01 4.44200225e-02 -8.98830891e-01 4.16768134e-01 -3.98656309e-01 6.52348757e-01 -3.84965718e-01 -3.35899025e-01 3.85597765e-01 5.51821947e-01 -2.11077735e-01 5.02898455e-01 -7.61853635e-01 -1.56176293e+00 5.62749624e-01 1.48365676e-01 -8.55572522e-01 -5.97056508e-01 -9.75704193e-02 -5.36765195e-02 1.59132689e-01 6.46380484e-01 1.22075453e-01 1.17615752e-01 1.28234282e-01 8.85943234e-01 9.83087659e-01 9.82819855e-01 -6.17717028e-01 3.05072665e-01 5.99508822e-01 -5.93868792e-01 -1.07074070e+00 -5.72444916e-01 -6.41181588e-01 -3.11555922e-01 -9.36634839e-01 1.53471458e+00 -1.16190410e+00 -1.48380494e+00 2.29495168e-01 -1.57200623e+00 -6.80981219e-01 -3.23390425e-03 2.72181153e-01 -9.42265272e-01 1.20136127e-01 -7.16454327e-01 -8.76303673e-01 -1.02671519e-01 -1.39190555e+00 1.56145525e+00 4.45191979e-01 -3.67684513e-01 -7.87675858e-01 -3.97393182e-02 1.17292488e+00 3.77087831e-01 4.66543101e-02 3.57780933e-01 -6.12151444e-01 -8.81143570e-01 5.42983003e-02 -3.42591822e-01 1.76992714e-01 -4.55626771e-02 -1.28299803e-01 -1.11053705e+00 -4.31247018e-02 4.06590253e-02 -1.13119876e+00 4.41033572e-01 3.62923414e-01 9.64969993e-01 -5.49691379e-01 -5.71254790e-02 2.68522240e-02 8.07947397e-01 4.54006106e-01 5.53982913e-01 -2.38380805e-01 6.74709737e-01 1.12763536e+00 8.70476663e-01 4.74479914e-01 7.27550626e-01 9.78985846e-01 5.74496090e-01 -3.95610705e-02 -2.27006316e-01 -3.25807422e-01 8.56086075e-01 3.33846748e-01 4.03767705e-01 -5.78402340e-01 -7.36528814e-01 6.91979766e-01 -1.95457888e+00 -1.10489511e+00 -4.21229191e-02 1.58214962e+00 6.39045537e-01 -5.17635524e-01 4.39327031e-01 -6.44177377e-01 7.63578594e-01 1.38890103e-01 -5.91308236e-01 -4.94623035e-01 7.30549321e-02 -5.44885457e-01 -1.87937230e-01 5.06159961e-01 -9.00022388e-01 1.23962700e+00 5.68772173e+00 2.12827235e-01 -9.21167135e-01 2.78598685e-02 6.75449967e-01 -1.73041224e-01 -7.35356137e-02 1.44067947e-02 -5.90394318e-01 1.14917241e-01 9.23785269e-01 5.57875447e-02 6.09784782e-01 1.01315224e+00 4.99284923e-01 -3.85146052e-01 -1.28257561e+00 1.15468240e+00 5.28551877e-01 -1.63093698e+00 3.03984016e-01 -4.52093512e-01 5.70896626e-01 -2.78824836e-01 1.41849428e-01 3.99662822e-01 3.88431638e-01 -8.91221941e-01 5.99037766e-01 5.35209179e-01 8.10532749e-01 -2.57474780e-01 5.49380302e-01 7.48110339e-02 -1.22821093e+00 -1.58625081e-01 -2.28054389e-01 1.75658204e-02 7.09666729e-01 -2.09725648e-01 -1.50580800e+00 -1.30539343e-01 9.48500156e-01 4.89276111e-01 -5.25714397e-01 3.47787887e-01 -4.06076834e-02 1.14370219e-01 1.56782314e-01 -4.81427163e-01 1.71996564e-01 8.37125629e-02 4.86330986e-01 1.28229237e+00 1.52579501e-01 3.64744335e-01 2.06426814e-01 9.05961394e-01 -1.13302000e-01 -7.31669292e-02 -9.19116318e-01 -1.94750667e-01 3.64474714e-01 1.37570429e+00 -5.12308002e-01 -5.82918227e-01 -5.58535218e-01 1.09543312e+00 -2.31425837e-02 5.33732474e-01 -8.53931129e-01 -4.60167415e-02 5.92401981e-01 1.13573922e-02 9.34486315e-02 -2.54220933e-01 4.01503801e-01 -1.21023834e+00 -2.71899730e-01 -1.17483687e+00 4.23873305e-01 -1.77892005e+00 -1.44062376e+00 8.69801879e-01 4.78398681e-01 -1.15564442e+00 -9.23502326e-01 -7.20948517e-01 -4.38209683e-01 2.00103879e-01 -6.15121245e-01 -1.30031550e+00 -1.04151416e+00 1.14447510e+00 1.57243919e+00 -3.91021222e-01 8.54983687e-01 -1.06814884e-01 1.96044277e-02 1.00073017e-01 -5.74432492e-01 2.75562018e-01 1.26178169e+00 -1.05060470e+00 1.73039898e-01 1.27144903e-01 2.23766759e-01 2.09012762e-01 8.72093737e-01 -3.95748585e-01 -1.82684910e+00 -8.67474794e-01 5.85507572e-01 -9.27817225e-01 5.87337554e-01 -6.20227814e-01 -6.27969086e-01 7.23454118e-01 9.90452170e-01 -4.12720352e-01 5.24258196e-01 -5.39866686e-01 -2.49643221e-01 5.54551855e-02 -9.90996361e-01 8.13001752e-01 7.88377047e-01 -8.23187649e-01 -6.76173925e-01 6.24415278e-01 9.70335543e-01 -5.58287203e-01 -6.23502910e-01 4.57750401e-03 6.60027981e-01 -1.01331544e+00 1.10251808e+00 -5.98397255e-01 7.94612288e-01 -1.56778753e-01 -4.71875221e-01 -9.59190428e-01 4.58357126e-01 -6.96855187e-01 1.22824080e-01 1.31137133e+00 2.36612514e-01 7.31201768e-02 5.27722478e-01 9.43760335e-01 -4.61064056e-02 -3.36373486e-02 -6.42716944e-01 -2.57326663e-01 -3.89316410e-01 -2.38702238e-01 3.69901419e-01 6.00347400e-01 3.32891643e-01 8.04489017e-01 -6.97066784e-01 -1.24262787e-01 7.20108151e-02 7.18402937e-02 1.64572787e+00 -6.06899559e-01 -1.00497529e-01 8.83089975e-02 -2.02750191e-01 -1.43499613e+00 4.04153079e-01 -3.35687310e-01 3.16896230e-01 -1.61847830e+00 4.55888540e-01 4.85870421e-01 7.72468269e-01 2.59687424e-01 1.71362072e-01 1.79086193e-01 6.06997073e-01 2.85667419e-01 -1.07234395e+00 7.50230014e-01 1.32448244e+00 -3.64510268e-01 -3.65147948e-01 -3.04916859e-01 -1.87421486e-01 6.40325546e-01 3.89024168e-01 -7.44384751e-02 -5.57442784e-01 -4.54230905e-01 -2.02346653e-01 9.61229563e-01 7.14260101e-01 -9.16825414e-01 4.95416433e-01 -3.66553038e-01 3.37171257e-01 -7.55133092e-01 1.24855077e+00 -8.08594108e-01 1.33885443e-01 1.41808599e-01 -8.33145142e-01 2.18227878e-01 2.74496526e-01 5.79981565e-01 -3.94008845e-01 9.89777297e-02 3.98709118e-01 -3.57585013e-01 -1.07072270e+00 -4.75047827e-02 -8.72489750e-01 -2.92157587e-02 1.15865767e+00 8.90825018e-02 -9.37302709e-01 -1.70777416e+00 -7.63920605e-01 5.50708532e-01 4.60599631e-01 8.88499498e-01 1.23433590e+00 -1.23300803e+00 -7.81707406e-01 -2.06813574e-01 4.11318958e-01 -2.08493158e-01 6.60363555e-01 3.64498407e-01 -6.27957761e-01 5.48315227e-01 -3.59377205e-01 -1.07616520e+00 -1.55116439e+00 6.64365947e-01 -2.17742771e-02 2.23268390e-01 -3.66780102e-01 7.28639185e-01 8.41097474e-01 -4.06251103e-01 4.30164963e-01 -8.63887519e-02 -2.15142503e-01 -8.19318816e-02 8.15711081e-01 -5.48556186e-02 -6.90592229e-01 -9.19487059e-01 -2.06882209e-01 3.10605228e-01 8.70927870e-02 -7.42337406e-01 8.12393904e-01 -4.32991445e-01 2.88056821e-01 6.02246344e-01 1.21496046e+00 -3.50195438e-01 -1.67929101e+00 3.27521609e-03 -6.29688501e-01 -5.09610474e-01 -6.11023962e-01 -7.36670196e-01 -8.21992338e-01 1.04569411e+00 5.46126068e-01 2.75991380e-01 9.73239362e-01 5.21407008e-01 8.82497013e-01 8.61599147e-01 3.00758541e-01 -1.01645660e+00 1.20082235e+00 5.40026009e-01 1.47082329e+00 -1.66838181e+00 -5.49528480e-01 -2.16912597e-01 -1.64310992e+00 1.23187697e+00 1.00134265e+00 5.34448922e-02 1.42310202e-01 7.89702311e-03 4.51654524e-01 -2.37760752e-01 -1.24441862e+00 -1.04983702e-01 1.16882093e-01 8.33003759e-01 1.15179941e-01 -1.82398185e-01 5.73089063e-01 4.66902345e-01 -1.44209236e-01 -1.40371814e-01 7.51133561e-01 5.65764546e-01 -3.13186467e-01 -6.81075037e-01 -6.16231084e-01 -1.99719127e-02 1.20756470e-01 1.65167525e-01 -1.02884126e+00 7.91394651e-01 -5.78974247e-01 1.79451275e+00 2.87433743e-01 -6.47374630e-01 2.59360671e-01 3.90502810e-02 2.41977140e-01 -6.45974100e-01 -5.96981049e-01 -6.59987405e-02 3.57885450e-01 -9.15452421e-01 -5.91604829e-01 -4.65360969e-01 -1.14466262e+00 -2.21258998e-01 1.00454763e-01 -1.23607423e-02 6.50233805e-01 8.23848009e-01 3.86645406e-01 2.03578815e-01 7.65653789e-01 -1.22796571e+00 -9.75038111e-02 -1.14168334e+00 8.09609219e-02 8.24027598e-01 2.87968844e-01 -4.30026442e-01 -4.50992256e-01 7.60825157e-01]
[10.879088401794434, 1.220104694366455]
3b1bb6b2-d3c6-4ee6-9b01-cefbe41e1f1e
latentgaze-cross-domain-gaze-estimation
2209.10171
null
https://arxiv.org/abs/2209.10171v1
https://arxiv.org/pdf/2209.10171v1.pdf
LatentGaze: Cross-Domain Gaze Estimation through Gaze-Aware Analytic Latent Code Manipulation
Although recent gaze estimation methods lay great emphasis on attentively extracting gaze-relevant features from facial or eye images, how to define features that include gaze-relevant components has been ambiguous. This obscurity makes the model learn not only gaze-relevant features but also irrelevant ones. In particular, it is fatal for the cross-dataset performance. To overcome this challenging issue, we propose a gaze-aware analytic manipulation method, based on a data-driven approach with generative adversarial network inversion's disentanglement characteristics, to selectively utilize gaze-relevant features in a latent code. Furthermore, by utilizing GAN-based encoder-generator process, we shift the input image from the target domain to the source domain image, which a gaze estimator is sufficiently aware. In addition, we propose gaze distortion loss in the encoder that prevents the distortion of gaze information. The experimental results demonstrate that our method achieves state-of-the-art gaze estimation accuracy in a cross-domain gaze estimation tasks. This code is available at https://github.com/leeisack/LatentGaze/.
['Seok Bong Yoo', 'Youngju Na', 'Hee Hyeon Kim', 'Jun-Seok Yun', 'Isack Lee']
2022-09-21
null
null
null
null
['gaze-estimation']
['computer-vision']
[ 4.33230549e-01 1.88744009e-01 -1.60728768e-01 -6.29562557e-01 -5.52701473e-01 -3.65684271e-01 4.01870668e-01 -6.79760754e-01 -1.27522811e-01 8.07322383e-01 1.12797379e-01 3.40953469e-02 -8.53226483e-02 -3.66540670e-01 -7.03311265e-01 -8.74785542e-01 4.01143104e-01 -2.84105450e-01 -2.59939194e-01 -7.21064284e-02 5.18162489e-01 -7.42567778e-02 -1.82688785e+00 -7.72327706e-02 1.20020628e+00 9.67017889e-01 7.64897466e-02 4.14118618e-01 1.89168096e-01 7.53527582e-01 -5.88293970e-01 -6.23464704e-01 1.10489197e-01 -7.77175069e-01 -4.91463363e-01 -1.29988287e-02 7.37185478e-01 -6.19779110e-01 -2.39922255e-02 1.24984944e+00 5.12215316e-01 -1.59843311e-01 7.10526824e-01 -1.79658246e+00 -1.02174342e+00 5.36271110e-02 -9.66618061e-01 1.87951162e-01 4.56207305e-01 4.49389338e-01 6.14148736e-01 -7.65199721e-01 3.70714992e-01 1.03498638e+00 2.06533447e-01 1.02698243e+00 -1.05803621e+00 -1.46705246e+00 7.98701271e-02 2.62062609e-01 -1.52735150e+00 -8.59353423e-01 1.21264207e+00 -4.80358511e-01 9.03604105e-02 2.93989867e-01 4.36873645e-01 1.44488549e+00 1.99238077e-01 6.38303757e-01 1.30123687e+00 -4.74812597e-01 -1.93474427e-01 2.91289061e-01 -2.71442473e-01 7.31083274e-01 1.12013638e-01 2.50623494e-01 -8.55564296e-01 1.81496352e-01 5.76012313e-01 1.02162451e-01 -5.88305891e-01 -3.35249186e-01 -9.81445312e-01 6.96281552e-01 4.43983465e-01 -1.74429953e-01 -1.57234609e-01 -4.17635553e-02 -1.02898106e-02 1.57161370e-01 6.40387177e-01 2.06907377e-01 -8.70487988e-02 -2.57040679e-01 -9.09798265e-01 6.88733160e-02 2.58448571e-01 1.11029994e+00 9.61252809e-01 1.33599760e-02 -3.09076875e-01 5.70165634e-01 7.02540874e-01 7.77939022e-01 3.86000484e-01 -9.04044569e-01 6.18933439e-01 6.90072358e-01 -1.66051425e-02 -1.03130102e+00 6.93738684e-02 -9.17702764e-02 -6.87426805e-01 4.97486055e-01 2.24401876e-01 -3.79573941e-01 -8.05982828e-01 2.31063223e+00 3.88459802e-01 1.77298769e-01 -1.88961193e-01 1.11035955e+00 6.97834313e-01 3.96412611e-01 6.68955892e-02 -4.06019330e-01 1.37877512e+00 -8.28037739e-01 -9.55876768e-01 -2.12736607e-01 1.89262301e-01 -7.43269861e-01 1.28082335e+00 1.88431829e-01 -1.05082190e+00 -5.12757659e-01 -1.11776114e+00 -3.32229823e-01 -1.07337527e-01 2.43872091e-01 4.72148389e-01 8.98596227e-01 -8.91848147e-01 7.69245327e-02 -5.24487495e-01 -1.13316059e-01 7.34329998e-01 5.57829320e-01 -3.68172318e-01 8.71865079e-02 -1.03551388e+00 6.61321402e-01 -1.12557776e-01 1.28997996e-01 -8.49732041e-01 -6.84617519e-01 -9.72518504e-01 4.85653616e-02 3.21896464e-01 -7.02975154e-01 1.09250510e+00 -1.46696901e+00 -1.68040514e+00 7.54946351e-01 -6.70641363e-01 1.27279118e-01 4.36630577e-01 -1.88228875e-01 -4.89272058e-01 3.76801044e-02 6.75097778e-02 8.53411496e-01 1.52003968e+00 -1.25639033e+00 -5.27652562e-01 -5.65765917e-01 1.30374327e-01 3.45031261e-01 -4.50060934e-01 9.95284468e-02 -2.59920746e-01 -3.81179333e-01 -2.44548127e-01 -9.35952485e-01 5.65904558e-01 2.99191386e-01 -6.64846301e-01 -1.11322828e-01 8.25176835e-01 -4.00650024e-01 1.43837965e+00 -2.35519481e+00 4.52062935e-02 -7.49028921e-02 6.47198260e-01 1.02400795e-01 6.27997294e-02 5.97175620e-02 -2.70926178e-01 1.98169351e-01 -1.90760091e-01 -6.59746110e-01 -7.88277015e-02 -2.61722833e-01 -3.96473080e-01 5.07920086e-01 5.00621021e-01 7.99914122e-01 -7.30631351e-01 -6.28849328e-01 5.03389873e-02 6.31341279e-01 -5.61006010e-01 4.14691806e-01 1.51377559e-01 6.99607790e-01 -5.39653718e-01 5.37923336e-01 1.02219486e+00 -3.20239842e-01 -3.31224382e-01 -1.98637173e-01 -6.43275380e-02 1.72920957e-01 -5.57564378e-01 1.68485796e+00 -3.24689090e-01 9.38511610e-01 -3.16702485e-01 -2.15789676e-01 8.56436968e-01 9.94941145e-02 4.65749726e-02 -6.27685249e-01 4.53856200e-01 -1.81801051e-01 1.54715434e-01 -6.78531587e-01 4.52939391e-01 6.86094612e-02 1.98497519e-01 6.47003710e-01 6.56631812e-02 2.76515543e-01 -1.43977731e-01 -3.36617008e-02 5.74800611e-01 4.35704798e-01 2.21578822e-01 -4.35834900e-02 6.96835756e-01 -4.47191477e-01 5.06287813e-01 4.79145050e-02 -5.07297456e-01 9.72908616e-01 6.69578910e-01 1.16012558e-01 -7.86501467e-01 -8.95314157e-01 -1.00677446e-01 1.10312402e+00 2.93479830e-01 -3.81043553e-01 -1.15751231e+00 -7.90746391e-01 -3.00143749e-01 7.21714437e-01 -1.16647029e+00 -5.05214632e-01 -2.31325343e-01 -2.74690509e-01 5.07594824e-01 7.02294037e-02 5.26198924e-01 -9.48397815e-01 -5.04408598e-01 -6.20625019e-01 -1.97609067e-01 -5.40788352e-01 -9.13308620e-01 -4.64052796e-01 -4.36091095e-01 -1.21708548e+00 -8.28467131e-01 -4.10785973e-01 1.04919589e+00 3.87046576e-01 7.41534889e-01 -5.53669855e-02 1.51282609e-01 4.88296226e-02 -2.18182445e-01 -5.83539367e-01 -1.10603236e-01 1.48837611e-01 -6.75096363e-02 4.74406391e-01 9.25312579e-01 -4.03955251e-01 -9.23772812e-01 3.70460212e-01 -5.70080936e-01 2.35412255e-01 4.35835361e-01 8.46347690e-01 1.72947496e-01 -5.32187760e-01 5.22279322e-01 -8.74414027e-01 6.95746064e-01 -6.45585477e-01 -5.16008496e-01 1.71513915e-01 -7.88853168e-01 1.16641931e-01 3.04593444e-01 -5.07560015e-01 -1.26974392e+00 -2.23572999e-01 2.01988727e-01 -7.94764638e-01 -1.22237734e-01 -1.27373189e-02 -3.34740102e-01 -4.78770398e-03 8.06688130e-01 3.77560556e-01 3.50949764e-01 -6.62660301e-02 1.26295984e-01 8.99466634e-01 1.69963628e-01 -3.13930005e-01 9.33000922e-01 3.54653716e-01 -1.35314122e-01 -3.85467738e-01 -8.48703384e-01 3.93988937e-02 -3.87538344e-01 -4.09853250e-01 8.75823140e-01 -1.00948572e+00 -1.02186716e+00 5.28290868e-01 -1.04599118e+00 3.20680887e-02 1.61296483e-02 2.95456082e-01 -5.48416495e-01 -5.35042472e-02 8.91999081e-02 -8.32609177e-01 -3.40477675e-01 -1.33884048e+00 1.17132878e+00 7.13385046e-01 -2.03448594e-01 -5.56735158e-01 1.04040459e-01 4.14939016e-01 3.00415039e-01 1.93169728e-01 4.12246436e-01 -2.51859784e-01 -7.97996819e-01 6.03634715e-02 -4.92001951e-01 2.85744667e-01 4.16160911e-01 2.10970819e-01 -1.34892845e+00 -3.03147227e-01 1.31296050e-02 -3.69688094e-01 4.66461957e-01 2.74084479e-01 1.16080821e+00 -3.85414034e-01 -3.89531076e-01 1.01753271e+00 1.12639618e+00 7.83705339e-02 6.68256640e-01 -6.82799658e-03 9.51734662e-01 6.53691053e-01 6.48919642e-01 3.86691034e-01 6.68983519e-01 5.74826479e-01 4.77997929e-01 1.16190173e-01 -1.23958342e-01 -5.57707727e-01 3.57550591e-01 3.92538190e-01 -1.71368480e-01 -3.81835729e-01 -6.02790654e-01 3.55786413e-01 -1.44694316e+00 -9.17325079e-01 9.17583779e-02 2.18110800e+00 1.14352846e+00 -1.03437230e-01 -3.49901319e-02 -4.53182906e-02 9.13583398e-01 2.11775914e-01 -7.13981569e-01 -1.71472564e-01 3.20267260e-01 -4.12740037e-02 2.65036225e-01 2.64525265e-01 -7.90710807e-01 7.21015036e-01 5.22275305e+00 6.16493464e-01 -1.48886383e+00 3.22219908e-01 4.65951145e-01 -4.58820701e-01 -4.24084097e-01 -8.61645192e-02 -7.63784230e-01 1.09296405e+00 7.84001946e-01 -3.37782949e-01 4.78127271e-01 7.33874142e-01 2.46992514e-01 -1.84482485e-01 -1.05912614e+00 1.19387829e+00 5.53506255e-01 -7.38298476e-01 -1.53970852e-01 3.38437170e-01 4.69413459e-01 -3.18388671e-01 7.12860763e-01 5.92222288e-02 -2.35340238e-01 -1.10031962e+00 6.61020994e-01 8.61273706e-01 1.42829192e+00 -8.60813320e-01 3.31792802e-01 1.49167970e-01 -6.42324686e-01 2.57698372e-02 -1.40700668e-01 -4.77016047e-02 -1.08169422e-01 -2.51445044e-02 -7.40222335e-01 2.49899670e-01 5.86267948e-01 7.20997512e-01 -7.76058137e-01 7.45772958e-01 -5.77303052e-01 5.40992737e-01 1.41438488e-02 3.93236578e-02 -2.96398014e-01 -7.15293363e-02 5.23578465e-01 4.62630212e-01 3.83451313e-01 1.55547336e-02 -6.59503281e-01 1.18659282e+00 -2.82184660e-01 -1.93974212e-01 -8.57470870e-01 2.53356665e-01 6.98871672e-01 1.10411370e+00 7.39351138e-02 1.33760139e-01 -3.16756129e-01 9.24267888e-01 3.88467699e-01 5.87395728e-01 -1.04122996e+00 -5.97335875e-01 1.04322219e+00 2.36227334e-01 1.03376575e-01 2.18097791e-01 -2.69659966e-01 -1.20621943e+00 1.72937542e-01 -8.13384414e-01 -9.52975377e-02 -1.29841793e+00 -9.33396280e-01 6.12881482e-01 4.64404821e-02 -1.70986736e+00 -5.13264000e-01 -1.62993565e-01 -6.12258852e-01 1.29921186e+00 -1.64390695e+00 -1.38784552e+00 -7.48589218e-01 8.43524337e-01 2.99986273e-01 -2.95453399e-01 6.28505707e-01 2.49085531e-01 -7.38233328e-01 1.21896064e+00 -1.04931451e-01 1.23687364e-01 1.22420526e+00 -9.79974508e-01 6.23104572e-02 9.45729017e-01 -2.83517599e-01 1.00552821e+00 6.12675011e-01 -3.03636879e-01 -1.11550653e+00 -7.81433642e-01 7.13044703e-01 -7.79021859e-01 2.61902153e-01 -5.74054480e-01 -8.36097658e-01 8.49622726e-01 6.37045801e-01 -7.55049735e-02 8.30197990e-01 -2.12617684e-02 -3.79100919e-01 -3.71374011e-01 -1.11661506e+00 7.55590141e-01 1.00999141e+00 -6.91110373e-01 -4.94987994e-01 -2.41010144e-01 6.14846110e-01 -4.76625115e-01 -4.15971994e-01 1.80135399e-01 7.30233431e-01 -1.20084286e+00 5.89183152e-01 -3.00021380e-01 8.12969446e-01 -4.26894665e-01 2.98694670e-01 -1.20464563e+00 -1.33422211e-01 -7.62861669e-01 -2.47993022e-01 1.48622572e+00 2.06618786e-01 -7.53381312e-01 6.49779022e-01 6.08445525e-01 1.99549198e-01 -6.10719681e-01 -7.23789096e-01 -1.50051191e-01 -1.14196382e-01 6.38170633e-03 9.34113979e-01 9.07769144e-01 -7.00598061e-02 4.53392625e-01 -6.28487647e-01 1.75811261e-01 7.43915021e-01 -3.07199582e-02 1.12850022e+00 -1.03053236e+00 1.47808567e-01 -3.36611897e-01 -3.34824920e-01 -8.89092028e-01 4.06203806e-01 -3.51722896e-01 9.57238395e-03 -7.83260286e-01 2.22059965e-01 -2.15050042e-01 -3.76488239e-01 3.81402999e-01 -4.85157907e-01 3.72983456e-01 2.02312171e-01 3.55516434e-01 -4.67536658e-01 7.13261545e-01 1.66454673e+00 1.59396534e-03 -7.28670806e-02 1.14483060e-02 -1.11437500e+00 3.51253837e-01 6.82603836e-01 -5.72478652e-01 -9.33066070e-01 -5.24741113e-01 3.66100401e-01 -1.37665108e-01 5.35808504e-01 -8.70808840e-01 3.44202489e-01 -2.92087913e-01 4.03777868e-01 -3.31785440e-01 5.15771389e-01 -7.01389134e-01 -7.50040114e-02 -9.09904912e-02 -4.24115092e-01 -1.18583888e-01 1.23425178e-01 4.73526061e-01 -2.55177915e-01 -2.44135454e-01 6.78064048e-01 3.98193687e-01 -3.31306517e-01 4.12119716e-01 2.00235918e-01 2.06222877e-01 1.09186924e+00 -4.07358497e-01 -6.72829330e-01 -4.90137517e-01 -2.75977105e-01 8.06318000e-02 8.50868583e-01 6.12884521e-01 5.42555213e-01 -1.30786598e+00 -5.88720798e-01 7.01904714e-01 4.95517075e-01 -1.79170407e-02 4.62037832e-01 8.98958564e-01 -1.08708724e-01 1.30700096e-01 -5.32375634e-01 -6.91384077e-01 -1.28805804e+00 5.99133015e-01 2.14629054e-01 4.64086443e-01 6.93854243e-02 9.45017219e-01 6.61597967e-01 6.88574091e-02 -1.82094932e-01 -4.71376255e-02 -3.76139134e-01 4.41610739e-02 6.66563094e-01 1.01625361e-01 -3.61736625e-01 -8.46761882e-01 -3.20015162e-01 6.65816486e-01 -1.74938157e-01 7.87346214e-02 8.50779057e-01 -7.28977919e-01 6.17734641e-02 3.87445301e-01 1.26319170e+00 3.12921911e-01 -1.73452926e+00 -7.95230493e-02 -6.88077807e-01 -9.55914855e-01 4.37765718e-02 -7.47144341e-01 -1.07945597e+00 1.07013249e+00 8.35477531e-01 1.01319954e-01 1.52597749e+00 -2.06698999e-01 4.55649167e-01 -1.85603395e-01 7.20403716e-02 -5.61894596e-01 -1.32535622e-01 8.98197293e-02 8.70269358e-01 -1.67711055e+00 -9.57525969e-02 -3.51412714e-01 -8.12615454e-01 7.80866325e-01 9.57012355e-01 -6.37848582e-03 6.95270419e-01 -2.04404101e-01 1.51252508e-01 -1.37187049e-01 -8.41140568e-01 -1.40192360e-01 5.12112677e-01 7.70178020e-01 3.58330160e-01 -1.87239602e-01 -5.27186617e-02 5.06522715e-01 -4.32054013e-01 2.53029466e-01 4.58585173e-01 6.17437720e-01 5.41576408e-02 -8.90856326e-01 -2.96703517e-01 3.30241442e-01 -4.64646101e-01 -3.08768243e-01 -1.17183529e-01 8.03322017e-01 2.53097385e-01 8.03612947e-01 3.02785844e-01 -5.32187879e-01 -1.70727707e-02 1.08068854e-01 5.51534891e-01 -4.71771389e-01 -1.62095234e-01 -2.11586699e-01 -2.69293875e-01 -5.05732119e-01 -5.69359839e-01 -7.16344953e-01 -6.10151291e-01 -5.61404765e-01 -5.03406763e-01 -1.10822827e-01 5.98873913e-01 8.88744652e-01 8.06206763e-01 5.13125896e-01 9.12276566e-01 -7.02773869e-01 -3.51150841e-01 -1.28896463e+00 -3.45151067e-01 3.26207995e-01 8.44143331e-01 -1.06656063e+00 -4.23838288e-01 3.32638144e-01]
[14.06788444519043, 0.017367567867040634]
0412b030-d299-4f03-a6ce-d3682027d08e
cross-encoder-for-unsupervised-gaze
null
null
http://openaccess.thecvf.com//content/ICCV2021/html/Sun_Cross-Encoder_for_Unsupervised_Gaze_Representation_Learning_ICCV_2021_paper.html
http://openaccess.thecvf.com//content/ICCV2021/papers/Sun_Cross-Encoder_for_Unsupervised_Gaze_Representation_Learning_ICCV_2021_paper.pdf
Cross-Encoder for Unsupervised Gaze Representation Learning
In order to train 3D gaze estimators without too many annotations, we propose an unsupervised learning framework, Cross-Encoder, to leverage the unlabeled data to learn suitable representation for gaze estimation. To address the issue that the feature of gaze is always intertwined with the appearance of the eye, Cross-Encoder disentangles the features using a latent-code-swapping mechanism on eye-consistent image pairs and gaze-similar ones. Specifically, each image is encoded as a gaze feature and an eye feature. Cross-Encoder is trained to reconstruct each image in the eye-consistent pair according to its gaze feature and the other's eye feature, but to reconstruct each image in the gaze-similar pair according to its eye feature and the other's gaze feature. Experimental results show the validity of our work. First, using the Cross-Encoder-learned gaze representation, the gaze estimator trained with very few samples outperforms the ones using other unsupervised learning methods, under both within-dataset and cross-dataset protocol. Second, ResNet18 pretrained by Cross-Encoder is competitive with state-of-the-art gaze estimation methods. Third, ablation study shows that Cross-Encoder disentangles the gaze feature and eye feature.
['Xilin Chen', 'Shiguang Shan', 'Jiabei Zeng', 'Yunjia Sun']
2021-01-01
null
null
null
iccv-2021-1
['gaze-estimation']
['computer-vision']
[ 6.58904836e-02 2.49152258e-01 -4.53481495e-01 -6.22735500e-01 -3.02888960e-01 -2.01974526e-01 3.99588048e-01 -5.17784715e-01 -2.49298915e-01 4.70374227e-01 1.97200701e-01 1.81822315e-01 -5.73151931e-02 -2.10377574e-01 -8.60354841e-01 -7.57912457e-01 2.44928554e-01 -1.26712114e-01 5.28980792e-02 1.44937068e-01 3.24920624e-01 -1.50508836e-01 -2.24264407e+00 1.65044551e-03 8.36294591e-01 1.27320814e+00 2.16066986e-01 4.43935245e-01 7.75168836e-02 9.47999895e-01 -3.15554827e-01 -3.19621712e-01 2.18794689e-01 -7.96635926e-01 -6.36176884e-01 1.01489864e-01 1.19110394e+00 -3.95587951e-01 -1.17793292e-01 1.06223619e+00 1.70549333e-01 -2.91301906e-01 7.55587816e-01 -1.66657317e+00 -1.10066640e+00 6.04846179e-02 -8.97258520e-01 1.42266676e-01 5.00954509e-01 1.46493435e-01 1.13751149e+00 -8.85545850e-01 6.65773213e-01 1.05379069e+00 3.74097079e-01 8.43145311e-01 -1.19128859e+00 -1.08387232e+00 2.73553491e-01 2.58030862e-01 -1.44289780e+00 -7.06349432e-01 7.82331884e-01 -5.61186910e-01 2.46089771e-01 5.33811152e-02 5.98059833e-01 1.23327959e+00 2.95755804e-01 7.83750832e-01 1.38000202e+00 -4.33408558e-01 -2.56034762e-01 2.36205190e-01 7.89418072e-02 9.30316210e-01 1.33575499e-01 5.62045753e-01 -1.14958131e+00 4.81864691e-01 5.60821056e-01 1.99976936e-01 -6.99871182e-01 -6.34355128e-01 -1.17040062e+00 6.36309564e-01 8.23225737e-01 2.93477569e-02 -1.01488024e-01 3.13660987e-02 -1.70592621e-01 4.43015426e-01 7.02592254e-01 3.43249291e-01 -4.41579312e-01 -1.29469395e-01 -1.00827038e+00 -7.74199367e-02 4.02459919e-01 1.06984460e+00 1.24723423e+00 -2.28651524e-01 -1.62267745e-01 3.67171943e-01 8.37968528e-01 6.84093297e-01 5.69635868e-01 -6.86707258e-01 2.47465760e-01 8.10524702e-01 -8.17163214e-02 -8.11486244e-01 -2.02301085e-01 -5.44301756e-02 -5.31765878e-01 4.90959734e-01 4.37895566e-01 -5.94542436e-02 -8.87827933e-01 2.22727585e+00 3.12590867e-01 3.50572079e-01 -1.12387739e-01 9.87438798e-01 9.74126577e-01 9.18684062e-03 -1.16837919e-01 -3.56824666e-01 1.53274548e+00 -1.20860457e+00 -7.82576501e-01 -2.50151008e-01 4.64614272e-01 -6.04350030e-01 9.63097215e-01 -1.89455152e-02 -9.11373198e-01 -7.35329151e-01 -1.08506286e+00 -3.44181299e-01 -1.24132864e-01 2.70675957e-01 3.28376323e-01 5.79817355e-01 -1.11538899e+00 1.01346999e-01 -5.31793594e-01 -2.92670190e-01 5.49582064e-01 5.24433136e-01 -7.12110281e-01 1.39590546e-01 -7.82058358e-01 9.24240768e-01 -9.12270993e-02 5.43129556e-02 -6.46388054e-01 -7.69919574e-01 -1.23779690e+00 1.31958768e-01 2.07297727e-01 -8.64347637e-01 9.77145970e-01 -1.54774892e+00 -1.52415991e+00 1.02167475e+00 -7.65684187e-01 -5.39805740e-02 2.23829955e-01 -2.18388930e-01 -5.80890775e-01 6.95282817e-02 2.10248306e-01 1.00732493e+00 1.54072833e+00 -1.22728121e+00 -7.33328640e-01 -6.81621432e-01 1.03782758e-01 2.07498372e-01 -2.46580362e-01 -1.20479926e-01 -3.55787188e-01 -6.86781183e-02 7.02133179e-02 -1.01178598e+00 7.08119810e-01 3.57881874e-01 -4.97394353e-01 -3.13854307e-01 9.34215784e-01 -1.63223937e-01 1.16914320e+00 -2.28923106e+00 2.43667811e-01 -1.85442433e-01 7.35268772e-01 -9.83496383e-02 -1.70800716e-01 -5.61506376e-02 -4.35375988e-01 -9.84581187e-02 -1.76144782e-02 -8.59434664e-01 -2.12867066e-01 5.04917130e-02 -9.86820236e-02 7.49638736e-01 6.25046909e-01 9.85228419e-01 -9.60666060e-01 -6.07875109e-01 -3.94499004e-02 5.22475898e-01 -5.83319306e-01 3.81416678e-01 2.06474543e-01 6.90067351e-01 -2.05064744e-01 4.93653387e-01 7.46293008e-01 -4.98138964e-01 -1.46483377e-01 -3.65078628e-01 -1.91636592e-01 2.31253028e-01 -5.50168395e-01 1.81395948e+00 -4.88203436e-01 1.19251561e+00 -4.19774264e-01 -3.19563776e-01 7.63851166e-01 1.34064972e-01 1.11031733e-01 -9.13255572e-01 2.15369865e-01 -6.89652562e-02 8.73909369e-02 -7.66201794e-01 3.82033497e-01 5.22069745e-02 2.08827689e-01 9.20621932e-01 5.61442256e-01 4.55501020e-01 -7.03054965e-02 7.37564522e-04 6.16811097e-01 4.88663316e-01 1.38341472e-01 -1.49005145e-01 5.82370698e-01 -7.97488272e-01 2.58941352e-01 2.63785303e-01 -4.07086581e-01 8.42426360e-01 5.93529165e-01 -2.47958079e-01 -5.71350098e-01 -7.45935559e-01 -2.90279925e-01 1.33056140e+00 4.47249562e-01 -4.10279602e-01 -8.26963186e-01 -1.14937460e+00 1.26528386e-02 4.70986366e-01 -1.45480478e+00 -2.60217339e-01 -8.72645974e-02 3.10351532e-02 1.72172487e-01 3.12210798e-01 3.16510737e-01 -9.16675925e-01 -7.74959445e-01 -6.82148874e-01 1.04344159e-01 -8.17325771e-01 -9.49895978e-01 -7.80851580e-03 -3.42612237e-01 -1.48909771e+00 -5.83118796e-01 -6.68280363e-01 1.09962034e+00 4.81722206e-01 1.02297807e+00 1.92854390e-01 1.43668458e-01 2.84662127e-01 -2.62644440e-01 -3.36778939e-01 5.34952022e-02 1.27726227e-01 2.86868075e-03 5.82270563e-01 7.62422442e-01 -5.00783622e-01 -6.68516457e-01 3.57579678e-01 -5.41198254e-01 7.97677189e-02 4.40345198e-01 9.39612627e-01 3.28704357e-01 -7.45365679e-01 2.01023668e-01 -9.55946267e-01 3.20101619e-01 -5.35973370e-01 -4.40007001e-01 3.84034544e-01 -8.58797491e-01 4.58330572e-01 -1.51632130e-01 -4.28567648e-01 -8.58163297e-01 1.52379096e-01 4.27912980e-01 -9.25569475e-01 -4.95547540e-02 2.04883859e-01 -5.16160913e-02 -6.49813637e-02 6.12577379e-01 2.33207062e-01 4.22099203e-01 -2.74880469e-01 3.78218651e-01 6.64992034e-01 3.62796128e-01 -1.73400417e-02 7.84523070e-01 3.99575114e-01 -8.44202563e-02 -3.49981010e-01 -1.34074903e+00 -4.12730843e-01 -9.13766861e-01 -3.50885451e-01 1.13022316e+00 -1.14223289e+00 -1.07165992e+00 5.18957436e-01 -1.09699357e+00 6.77849799e-02 -2.16547802e-01 5.96364498e-01 -4.16505188e-01 -1.53859168e-01 9.50965509e-02 -6.69167459e-01 -1.04948342e-01 -1.37581265e+00 1.56529784e+00 5.03505588e-01 -2.87942976e-01 -8.91980350e-01 1.99928582e-01 1.72710076e-01 1.00374624e-01 -4.44894880e-02 3.16277385e-01 -4.46204334e-01 -6.72214091e-01 -6.04999550e-02 -6.09130621e-01 1.40034646e-01 5.25951028e-01 1.71070904e-01 -1.34393871e+00 -2.77077585e-01 2.54581571e-02 -4.97832239e-01 8.25255632e-01 1.95468441e-01 9.11291242e-01 -1.43736377e-01 -3.31825882e-01 9.30080414e-01 1.06544650e+00 -3.15977842e-01 6.27903402e-01 1.52568454e-02 1.08680415e+00 6.13938630e-01 4.10384774e-01 2.44112443e-02 9.37056422e-01 6.31083727e-01 5.62293231e-01 9.67161655e-02 -3.08824390e-01 -4.45117414e-01 3.72514278e-01 6.38344347e-01 -9.05767903e-02 -8.44145566e-03 -6.19665861e-01 2.86893338e-01 -1.64317107e+00 -9.84460473e-01 -1.71528339e-01 2.13198876e+00 8.62724900e-01 -1.56410590e-01 1.47267371e-01 -1.28165141e-01 7.36739814e-01 2.18984947e-01 -7.00673640e-01 -2.02448945e-02 1.14142939e-01 1.57504171e-01 2.16828778e-01 3.78593326e-01 -8.38110328e-01 6.69406772e-01 5.81045437e+00 2.55225152e-01 -1.64188814e+00 2.85556108e-01 2.40028337e-01 -2.63401240e-01 -2.90744126e-01 3.91703285e-02 -8.63259256e-01 8.93350184e-01 8.85061800e-01 -2.46734284e-02 3.83717507e-01 6.17577434e-01 -1.94650471e-01 -3.35859686e-01 -1.63703954e+00 1.16931260e+00 8.37778330e-01 -9.09274995e-01 -2.50613362e-01 3.14218432e-01 6.24613822e-01 1.31472503e-03 4.79996949e-01 1.40093774e-01 -1.95426881e-01 -1.04363036e+00 9.00622368e-01 9.73948538e-01 1.36632419e+00 -3.46876353e-01 5.30938268e-01 1.18239567e-01 -9.69519496e-01 1.86188761e-02 1.62336186e-01 -1.75296918e-01 -1.79184213e-01 -1.44050419e-01 -5.46938777e-01 3.08819622e-01 8.31621528e-01 1.25516295e+00 -9.34192717e-01 7.67173350e-01 -5.72291672e-01 2.67099828e-01 1.28694519e-01 2.18681037e-01 -1.88268930e-01 -1.38193265e-01 3.69121343e-01 4.23062086e-01 1.33303300e-01 -2.64511049e-01 -5.21824837e-01 1.16212559e+00 -2.06055418e-01 -3.37951183e-01 -5.41263759e-01 1.97776392e-01 4.42410767e-01 9.52531517e-01 -1.62713844e-02 -1.01455078e-01 -4.82419252e-01 9.52718318e-01 6.25245988e-01 4.67643827e-01 -7.05687165e-01 -3.08848649e-01 8.93822372e-01 1.67878762e-01 6.04062974e-01 2.39193797e-01 -2.03639343e-01 -1.24898136e+00 -1.41624706e-02 -4.88774478e-01 2.95872800e-03 -1.31183887e+00 -1.21668887e+00 8.76754045e-01 -1.13990754e-01 -1.70654416e+00 -5.13235927e-01 -4.39861715e-01 -5.11517167e-01 1.19568563e+00 -1.91325223e+00 -1.42406237e+00 -6.68645680e-01 8.70075703e-01 2.82736063e-01 -2.68347204e-01 9.10517216e-01 3.40999514e-02 -7.86830068e-01 1.13534343e+00 -2.08636180e-01 7.29846656e-02 1.15694499e+00 -1.12266505e+00 -7.15158461e-03 6.01031423e-01 3.77611876e-01 8.91948581e-01 3.25754136e-01 -1.67564362e-01 -1.10615718e+00 -7.55238831e-01 1.15909922e+00 -8.52156341e-01 5.76839745e-01 -5.03206730e-01 -8.73945832e-01 8.25056195e-01 5.93569100e-01 5.19537449e-01 7.81823158e-01 4.70180154e-01 -8.21301579e-01 -3.06196392e-01 -8.95961463e-01 3.17391008e-01 1.01213145e+00 -1.16735423e+00 -6.56493247e-01 -2.47903258e-01 6.87304199e-01 -6.21041179e-01 -5.24999022e-01 1.49796307e-01 9.20034349e-01 -1.39543605e+00 5.17206967e-01 -5.16698182e-01 8.54266524e-01 -2.70343393e-01 2.38616988e-01 -1.20916188e+00 -1.13049053e-01 -4.20756549e-01 -4.57524061e-01 1.16462684e+00 3.52024645e-01 -7.56535411e-01 4.55072105e-01 4.83918399e-01 6.77317381e-02 -8.03890646e-01 -7.91317105e-01 -2.92461574e-01 -2.42272452e-01 -6.24192320e-02 9.04161215e-01 9.23322558e-01 -9.80231389e-02 7.61181951e-01 -5.03175437e-01 1.28680214e-01 5.60209095e-01 2.72483259e-01 9.59129632e-01 -1.49900997e+00 2.33423561e-02 -2.99771667e-01 -5.95604300e-01 -1.11400616e+00 6.54273629e-01 -5.48256338e-01 1.45200014e-01 -7.61490762e-01 3.23944807e-01 -2.22972393e-01 -4.91530299e-01 5.36234558e-01 -4.70931798e-01 5.54690123e-01 2.93678492e-01 4.41606313e-01 -7.80338049e-01 6.07907414e-01 1.57509327e+00 -1.09689888e-02 -1.37013897e-01 -9.86009911e-02 -8.93878579e-01 5.28249502e-01 1.13655649e-01 -6.86410010e-01 -5.80641687e-01 -5.91713369e-01 3.91819090e-01 -1.14906892e-01 4.37111855e-01 -7.49521077e-01 6.09451234e-01 2.01376483e-01 3.80744219e-01 -5.38653970e-01 2.85111725e-01 -9.14821684e-01 -3.02567035e-01 -5.26203886e-02 -4.27007318e-01 1.44154048e-02 -8.14773887e-02 7.11735487e-01 -3.69818091e-01 -1.31318450e-01 5.71205616e-01 4.15254265e-01 -4.57892239e-01 4.01937723e-01 2.80374169e-01 2.18924031e-01 9.38252211e-01 -6.18236184e-01 -4.68161523e-01 -3.39605153e-01 -4.59686488e-01 8.92305896e-02 8.18158984e-01 7.83421934e-01 4.96893585e-01 -1.28918719e+00 -4.06020194e-01 8.92089903e-01 6.68855369e-01 1.07572731e-02 1.57576203e-01 1.20070863e+00 8.43089595e-02 2.21985430e-01 -5.02374172e-01 -1.10540557e+00 -1.32654655e+00 5.89345992e-01 2.91345417e-01 3.65071625e-01 -5.68758883e-02 1.15680599e+00 6.98126674e-01 -9.54643488e-02 8.30704272e-02 -3.23439449e-01 -4.05104190e-01 3.29537481e-01 7.16487646e-01 -6.72849566e-02 -1.35269910e-01 -1.24450397e+00 -3.36069405e-01 9.02212799e-01 -1.98103920e-01 2.39015684e-01 1.17334425e+00 -5.88220716e-01 -1.11575186e-01 7.42138088e-01 1.59426284e+00 1.11910239e-01 -1.68486643e+00 -3.62185121e-01 -3.42911243e-01 -8.14975441e-01 2.27220401e-01 -6.45046055e-01 -1.31856394e+00 8.90069127e-01 8.31232250e-01 3.72104011e-02 1.25362349e+00 1.19227648e-01 2.59380430e-01 -3.76407132e-02 1.39575154e-01 -4.23452795e-01 1.03905819e-01 3.82425576e-01 7.42781818e-01 -1.83168972e+00 -9.80696976e-02 -2.27034524e-01 -8.42501700e-01 8.67168307e-01 8.13146114e-01 -1.64170623e-01 9.83043551e-01 -2.22546339e-01 1.75904036e-01 -5.57199419e-01 -8.18439126e-01 -4.62870985e-01 8.94322932e-01 6.10116482e-01 4.83747691e-01 -1.73463285e-01 2.22594410e-01 3.18573147e-01 -4.00585204e-01 1.68130025e-01 2.02970684e-01 4.76918072e-01 1.43282160e-01 -9.06202078e-01 3.51561010e-02 3.68434042e-01 -3.49212214e-02 -2.32788369e-01 -3.22479010e-01 8.82413566e-01 4.15497690e-01 6.31402850e-01 7.24982977e-01 -5.51397800e-01 -4.05038111e-02 3.61846350e-02 7.50497043e-01 -7.21341074e-01 -3.17114770e-01 -3.25913072e-01 -3.62624049e-01 -7.25388169e-01 -1.00091386e+00 -5.99733293e-01 -7.27374494e-01 -1.14787787e-01 -7.44077146e-01 -1.88530207e-01 6.02727056e-01 1.10693872e+00 6.73723102e-01 4.67113614e-01 9.11731005e-01 -9.34459567e-01 -2.62922972e-01 -1.08649004e+00 -6.85968220e-01 5.00896752e-01 1.02872932e+00 -1.27360260e+00 -5.41751325e-01 3.33587170e-01]
[14.117530822753906, 0.031076705083251]
25cb36a5-1879-42e6-ac40-59e524d154b5
leveraging-long-and-short-term-information-in
1712.09059
null
http://arxiv.org/abs/1712.09059v5
http://arxiv.org/pdf/1712.09059v5.pdf
Leveraging Long and Short-term Information in Content-aware Movie Recommendation
Movie recommendation systems provide users with ranked lists of movies based on individual's preferences and constraints. Two types of models are commonly used to generate ranking results: long-term models and session-based models. While long-term models represent the interactions between users and movies that are supposed to change slowly across time, session-based models encode the information of users' interests and changing dynamics of movies' attributes in short terms. In this paper, we propose an LSIC model, leveraging Long and Short-term Information in Content-aware movie recommendation using adversarial training. In the adversarial process, we train a generator as an agent of reinforcement learning which recommends the next movie to a user sequentially. We also train a discriminator which attempts to distinguish the generated list of movies from the real records. The poster information of movies is integrated to further improve the performance of movie recommendation, which is specifically essential when few ratings are available. The experiments demonstrate that the proposed model has robust superiority over competitors and sets the state-of-the-art. We will release the source code of this work after publication.
['Chen Xiaojun', 'Zhao Zhou', 'Yang Min', 'Ye Jianbo', 'Wang Benyou', 'Chai Haixia', 'Zhao Wei']
2018-06-26
null
null
null
null
['movie-recommendation']
['miscellaneous']
[-4.74722907e-02 -5.10373354e-01 -2.70931154e-01 -8.17454636e-01 -4.28185582e-01 -9.92837906e-01 6.88899815e-01 -3.75466049e-01 -2.86638111e-01 7.43768871e-01 5.25353611e-01 1.92293953e-02 -3.29925954e-01 -9.21243906e-01 -6.95096970e-01 -5.78368127e-01 -2.09254205e-01 3.68157059e-01 2.83489138e-01 -7.31255114e-01 3.98789316e-01 2.09258616e-01 -1.53914976e+00 9.10364389e-01 7.29663789e-01 1.03276825e+00 2.05483183e-01 7.72487342e-01 3.73001359e-02 9.78503406e-01 -6.31096780e-01 -7.94353962e-01 5.21895707e-01 -6.44327879e-01 -2.97665238e-01 -9.99644995e-02 2.70762086e-01 -6.14798009e-01 -7.18314826e-01 9.28957105e-01 5.69088876e-01 6.74782693e-01 6.56177223e-01 -1.03074610e+00 -1.20467520e+00 8.99167299e-01 1.74987033e-01 3.80160481e-01 4.39639926e-01 1.34363649e-02 1.11366487e+00 -7.46725798e-01 5.35690784e-01 7.99098432e-01 2.93866545e-01 9.38972652e-01 -6.75558746e-01 -5.83877027e-01 7.11908937e-01 2.12196589e-01 -7.90888608e-01 -2.72983015e-01 8.35137188e-01 -5.80733001e-01 4.15784001e-01 5.19848764e-01 5.27842164e-01 1.33774233e+00 3.19675386e-01 6.00205243e-01 7.37419248e-01 1.10492982e-01 2.41283804e-01 4.93006766e-01 2.54139584e-02 1.27835423e-01 -4.23746198e-01 3.83035362e-01 -4.23116684e-01 -2.90158451e-01 8.43707442e-01 4.88648087e-01 -8.93755481e-02 -1.84544697e-01 -9.14818347e-01 9.43264902e-01 4.06471401e-01 1.85621470e-01 -4.91148382e-01 -1.39640316e-01 2.13294774e-01 9.20494735e-01 5.28004467e-01 6.14881635e-01 -3.44252646e-01 1.10605597e-01 -7.29469240e-01 5.55479467e-01 5.69177985e-01 9.10354018e-01 -1.07598796e-01 7.74569809e-02 -5.96853077e-01 9.61789250e-01 1.81738228e-01 2.56264299e-01 7.64066577e-01 -7.95658529e-01 3.19113344e-01 -4.28774022e-02 6.63929939e-01 -9.48513210e-01 5.86601384e-02 -6.60170913e-01 -7.06219614e-01 -9.41222012e-02 2.01814488e-01 -2.76584208e-01 -5.58141708e-01 1.80045474e+00 5.27636558e-02 3.71377289e-01 -5.49590727e-03 1.09400582e+00 9.41912234e-01 9.63428378e-01 -2.54961282e-01 -6.40859962e-01 6.39631987e-01 -1.36439657e+00 -8.22792888e-01 1.48356825e-01 -1.65561691e-01 -6.39808536e-01 1.03123820e+00 6.73747361e-01 -1.41563666e+00 -1.03535843e+00 -8.55922699e-01 4.42720175e-01 -2.95480072e-01 2.39704251e-01 6.54312074e-01 2.80817211e-01 -9.88928616e-01 1.03392398e+00 -1.69960737e-01 2.27930564e-02 -8.75339471e-03 4.79914099e-01 1.35827258e-01 1.84288025e-01 -1.80350995e+00 2.65102059e-01 -3.12799335e-01 8.54098424e-02 -1.13585985e+00 -4.25787181e-01 -1.84396029e-01 7.41894841e-02 2.84063280e-01 -5.83212316e-01 1.37495828e+00 -1.17683554e+00 -1.96492553e+00 1.87092289e-01 3.93396020e-01 -1.69366300e-01 5.09916246e-01 -1.63105562e-01 -1.02143264e+00 -1.24023266e-01 -2.88876891e-01 -8.59221257e-03 9.86286402e-01 -1.14956486e+00 -9.27255034e-01 -1.01537064e-01 7.21910715e-01 3.79163325e-01 -6.61715150e-01 1.77133918e-01 -5.38493812e-01 -1.07884860e+00 -4.06016648e-01 -1.07441938e+00 -3.59287471e-01 -3.60252798e-01 -1.14545319e-02 -2.26178706e-01 1.36112243e-01 -4.73547727e-01 1.50996327e+00 -2.11053038e+00 2.97534734e-01 1.79363176e-01 -2.10337758e-01 2.25471467e-01 -3.61703336e-01 7.50644445e-01 2.20930889e-01 -1.32541284e-01 4.75647867e-01 -1.63411230e-01 -6.79636449e-02 -4.24286984e-02 -7.38640845e-01 6.48782402e-02 -5.24922609e-01 7.88165331e-01 -1.08392859e+00 2.25605041e-01 -5.71797304e-02 2.15445340e-01 -8.71755838e-01 8.36311936e-01 -3.94203156e-01 7.04441249e-01 -6.88439727e-01 1.07005164e-02 3.73227835e-01 -2.82930851e-01 2.48709694e-01 -8.53924081e-02 1.49165377e-01 3.87105316e-01 -1.14449275e+00 1.79401040e+00 -5.42655408e-01 -1.29943818e-01 -2.53466696e-01 -5.99519551e-01 8.38269174e-01 5.02565503e-01 4.14613187e-01 -5.06031692e-01 1.18431419e-01 -1.80744991e-01 -2.81275641e-02 -5.76081753e-01 6.92104578e-01 -8.43646973e-02 -1.64184645e-01 5.98615885e-01 5.62241413e-02 1.51584893e-01 1.57320067e-01 5.72093666e-01 8.70839715e-01 1.52060226e-01 -2.04134643e-01 7.23231509e-02 7.35354960e-01 -5.36145568e-01 5.50648987e-01 1.05895078e+00 9.37460810e-02 5.84769785e-01 7.51317814e-02 -4.64491844e-01 -7.72215426e-01 -9.84326005e-01 1.79068804e-01 1.61342847e+00 2.73528129e-01 -3.80695969e-01 -3.84751618e-01 -1.06633866e+00 -6.73379600e-02 8.16083848e-01 -8.23486447e-01 -4.31037247e-01 -2.46724665e-01 -4.24643219e-01 -1.74661234e-01 3.75844628e-01 -7.28215054e-02 -1.37297416e+00 3.14065605e-01 4.38725978e-01 -1.10722017e-02 -6.02285743e-01 -1.13202655e+00 -2.81007051e-01 -8.23793173e-01 -5.99849522e-01 -8.86864901e-01 -6.62899792e-01 7.23731756e-01 2.75018185e-01 1.05359411e+00 -6.90852702e-02 4.41610992e-01 2.51936197e-01 -8.27037394e-01 1.71091184e-01 -4.37775642e-01 -2.66458154e-01 4.73079532e-01 4.71163571e-01 -9.77210514e-03 -7.34604478e-01 -1.05708945e+00 5.79706728e-01 -9.10720646e-01 -1.83087990e-01 2.16824517e-01 9.12589431e-01 4.77640003e-01 2.52379149e-01 9.24078166e-01 -1.59279859e+00 9.57793891e-01 -8.21453631e-01 -2.22970083e-01 3.45598251e-01 -7.23248065e-01 -2.67033935e-01 1.51856577e+00 -8.22427273e-01 -1.25273657e+00 -2.89179444e-01 -2.55117655e-01 -3.62234801e-01 -1.30894249e-02 3.55600834e-01 -8.59572664e-02 4.00246173e-01 5.34380913e-01 2.17027843e-01 -4.35876250e-01 -8.15065145e-01 5.93401253e-01 8.33861351e-01 5.58659852e-01 -4.04164881e-01 6.55313551e-01 2.06765488e-01 -5.79654992e-01 2.65967458e-01 -1.34847724e+00 -4.74051654e-01 -3.45238596e-01 -4.83188719e-01 4.85377014e-01 -8.19285154e-01 -4.54175949e-01 4.07629162e-01 -7.15731084e-01 -2.25487307e-01 -3.95384997e-01 5.65326810e-01 -4.92618650e-01 -4.62103561e-02 -1.13064456e+00 -5.05593359e-01 -3.33860189e-01 -9.66818929e-01 4.53984678e-01 3.49020392e-01 2.60568321e-01 -1.05511057e+00 3.64105999e-01 3.64166737e-01 5.64104557e-01 -1.66350231e-01 4.30408835e-01 -9.15883541e-01 -2.61346042e-01 -4.77845699e-01 5.24099708e-01 6.75719678e-01 1.82150796e-01 -7.00529292e-03 -6.86332107e-01 -6.52786732e-01 -1.44896144e-02 -2.97945380e-01 4.75399584e-01 3.13900888e-01 1.42081034e+00 -6.94924653e-01 1.15500554e-01 5.34635246e-01 1.21324289e+00 6.06141031e-01 4.80607182e-01 -7.46731274e-03 5.67700505e-01 5.87084293e-01 1.00066483e+00 6.96169019e-01 2.71044165e-01 8.67484212e-01 3.30683619e-01 2.56431133e-01 1.44841835e-01 -6.27816856e-01 6.51909649e-01 1.35752642e+00 -2.50175208e-01 -6.83780134e-01 2.29591742e-01 4.88426313e-02 -2.08423448e+00 -1.44079924e+00 1.53423563e-01 2.41385031e+00 7.79898584e-01 3.64560306e-01 3.10347199e-01 -3.52990746e-01 7.24604607e-01 1.43583655e-01 -6.99378610e-01 -5.13542533e-01 2.16103047e-01 -1.77295040e-02 1.48937955e-01 4.93276358e-01 -1.02573240e+00 7.74708688e-01 6.11621189e+00 7.52756178e-01 -1.01883435e+00 1.79425225e-01 4.85163271e-01 -5.37050128e-01 -6.28222346e-01 -2.17605636e-01 -6.06732845e-01 9.73586619e-01 1.01514101e+00 -3.54953587e-01 8.64327133e-01 7.32452393e-01 3.24281693e-01 6.41217828e-01 -1.16564047e+00 5.66366196e-01 2.88685858e-01 -1.13759601e+00 2.69049704e-01 -1.02363274e-01 1.28576374e+00 -3.49130660e-01 4.40998971e-01 5.87933183e-01 5.50456583e-01 -7.47397780e-01 7.41210938e-01 1.08570051e+00 7.10704327e-01 -8.60262930e-01 6.19816244e-01 5.67967534e-01 -9.25780177e-01 -4.90590215e-01 -6.84653461e-01 -2.10987285e-01 3.93781722e-01 3.61351103e-01 -7.30257528e-03 6.27027452e-01 5.05967975e-01 1.09709096e+00 -3.66008639e-01 8.44102502e-01 -3.30070734e-01 7.44562387e-01 4.49086964e-01 -1.04781024e-01 4.77616303e-02 -5.89181066e-01 3.89954031e-01 8.52314651e-01 4.16983694e-01 4.86322165e-01 3.26159209e-01 2.32327178e-01 -2.13048145e-01 3.44186515e-01 -3.77698064e-01 1.71688780e-01 3.77921313e-01 1.40392137e+00 -1.81419075e-01 -4.28437859e-01 -4.66806144e-01 1.25784838e+00 2.49872476e-01 3.43328625e-01 -1.00782204e+00 -1.64526150e-01 3.95385325e-01 2.23163992e-01 5.03335834e-01 3.04116696e-01 4.16609615e-01 -1.30183446e+00 -2.22874627e-01 -1.16297340e+00 5.62314212e-01 -8.43363583e-01 -1.82107627e+00 1.01098084e+00 -3.37416381e-01 -1.67733729e+00 -2.92336017e-01 -1.78460255e-01 -8.28952670e-01 6.31069839e-01 -1.05004334e+00 -9.30180728e-01 -3.20258774e-02 8.77934635e-01 7.83791423e-01 -7.15659916e-01 7.66541600e-01 5.93562365e-01 -2.98515826e-01 9.97830153e-01 7.39395082e-01 -2.70822674e-01 1.08592510e+00 -1.20144916e+00 2.47508913e-01 5.94067097e-01 3.98268759e-01 8.73063445e-01 7.30394840e-01 -5.06457210e-01 -1.12730229e+00 -1.16212630e+00 6.49392188e-01 -4.95813489e-01 5.41580141e-01 -3.14057499e-01 -5.49245059e-01 5.45549512e-01 7.31174797e-02 -1.75197665e-02 9.69579816e-01 -3.80869396e-02 -2.35039771e-01 -4.51532215e-01 -1.08099902e+00 4.93736416e-01 1.05312085e+00 -3.69874954e-01 -4.89816666e-01 5.60973585e-01 8.66562843e-01 -2.57607341e-01 -1.00395334e+00 2.69825190e-01 8.03506494e-01 -9.76927459e-01 9.31986153e-01 -1.22065961e+00 7.67316341e-01 -1.77417800e-01 -8.10246915e-02 -1.69959939e+00 -9.92149293e-01 -6.82971060e-01 -3.80374581e-01 1.38124800e+00 4.19088513e-01 -1.83223471e-01 6.73815489e-01 5.43889761e-01 -5.24189733e-02 -8.30247641e-01 -2.54235685e-01 -5.69182158e-01 1.80683676e-02 1.14880309e-01 8.44090700e-01 9.45570886e-01 6.64116964e-02 4.70324248e-01 -1.22037399e+00 5.38424328e-02 2.41035447e-01 4.53707069e-01 5.60801148e-01 -9.50912476e-01 -1.07500458e+00 -9.59465131e-02 1.55919611e-01 -1.37303579e+00 4.25227843e-02 -9.46881652e-01 -2.33323611e-02 -1.34166098e+00 2.97572464e-01 -4.95546460e-01 -1.17618299e+00 -9.90123674e-02 -2.65925467e-01 3.28190833e-01 2.10939497e-01 2.80198842e-01 -9.42092061e-01 5.54771125e-01 1.63995278e+00 1.00438893e-01 -3.10977817e-01 8.74831259e-01 -8.85028660e-01 3.90640080e-01 7.30838358e-01 -5.10612786e-01 -1.01353467e+00 -2.23136112e-01 4.33349282e-01 5.02134860e-01 -2.63229072e-01 -7.63195634e-01 2.74119437e-01 -5.17260432e-01 3.08133513e-01 -3.60816807e-01 3.00025523e-01 -7.56181777e-01 3.71979684e-01 9.55171809e-02 -1.25557482e+00 1.29602015e-01 -4.81952280e-01 8.87420952e-01 -1.50997475e-01 -3.34228992e-01 4.90817994e-01 -2.45591149e-01 -3.73751640e-01 9.09906507e-01 -2.92924970e-01 -2.37036705e-01 8.59429061e-01 3.48177671e-01 -1.33860841e-01 -1.05796289e+00 -1.07858908e+00 1.68836296e-01 2.40962788e-01 9.00433600e-01 5.95591426e-01 -1.61006582e+00 -7.24287748e-01 -4.07542326e-02 2.99004819e-02 -7.78924167e-01 5.63441038e-01 -2.02999283e-02 6.22451082e-02 1.58860683e-01 -2.89595455e-01 3.08892161e-01 -1.14791298e+00 1.08829129e+00 1.18213154e-01 -5.27046502e-01 -1.12370290e-01 1.12224460e+00 2.36396194e-01 -5.59556544e-01 5.09059727e-01 3.58046234e-01 -9.70411181e-01 -4.64699306e-02 6.97162688e-01 2.24736378e-01 -2.41821364e-01 -4.61666733e-01 8.63007382e-02 1.58804417e-01 -5.65275490e-01 -1.24272190e-01 1.38132000e+00 -3.48254085e-01 2.23625347e-01 6.56229138e-01 8.60652089e-01 4.55956280e-01 -1.35179424e+00 -3.44077349e-01 -7.63074398e-01 -8.20521355e-01 -3.83920036e-02 -1.24774694e+00 -1.42790377e+00 5.56174576e-01 5.64451396e-01 4.62737083e-01 1.08526039e+00 -2.56754309e-01 9.67423499e-01 9.57379118e-03 4.85117763e-01 -1.13923478e+00 3.46924931e-01 2.96710908e-01 1.00352037e+00 -1.04325008e+00 4.63879388e-03 -8.47308338e-02 -9.68624175e-01 8.09898973e-01 8.00943971e-01 -4.84663069e-01 9.70337451e-01 -2.14247942e-01 1.27378464e-01 9.86205712e-02 -1.19104755e+00 1.64649114e-01 6.47946835e-01 2.70834029e-01 4.89526242e-01 1.28432676e-01 -7.00721085e-01 1.48320639e+00 -1.35374799e-01 1.40675500e-01 5.52268147e-01 4.90591347e-01 -2.23674506e-01 -1.56131291e+00 2.64860541e-01 8.17072690e-01 -7.37416685e-01 -1.11908004e-01 -1.60147011e-01 -2.16237023e-01 2.66808510e-01 1.24697816e+00 -4.48777005e-02 -9.49239671e-01 4.98466432e-01 -2.80218899e-01 2.58133531e-01 -7.54280150e-01 -9.82560754e-01 -4.93081398e-02 1.04763418e-01 -4.91599739e-01 -2.19300613e-01 -5.31758666e-01 -6.49553120e-01 -2.81776190e-01 -4.40847456e-01 6.94691420e-01 4.01789129e-01 5.38011491e-01 4.12417829e-01 6.91887915e-01 1.66271949e+00 -6.54773235e-01 -1.15966332e+00 -8.54075313e-01 -9.18645620e-01 9.59638655e-01 2.38414817e-02 -4.40656781e-01 -4.03614879e-01 2.41547972e-01]
[10.128472328186035, 5.630187511444092]
505ddd5e-0556-4044-922e-7a24b89eb131
unreal-unlabeled-nodes-retrieval-and-labeling
2303.10371
null
https://arxiv.org/abs/2303.10371v1
https://arxiv.org/pdf/2303.10371v1.pdf
UNREAL:Unlabeled Nodes Retrieval and Labeling for Heavily-imbalanced Node Classification
Extremely skewed label distributions are common in real-world node classification tasks. If not dealt with appropriately, it significantly hurts the performance of GNNs in minority classes. Due to its practical importance, there have been a series of recent research devoted to this challenge. Existing over-sampling techniques smooth the label distribution by generating ``fake'' minority nodes and synthesizing their features and local topology, which largely ignore the rich information of unlabeled nodes on graphs. In this paper, we propose UNREAL, an iterative over-sampling method. The first key difference is that we only add unlabeled nodes instead of synthetic nodes, which eliminates the challenge of feature and neighborhood generation. To select which unlabeled nodes to add, we propose geometric ranking to rank unlabeled nodes. Geometric ranking exploits unsupervised learning in the node embedding space to effectively calibrates pseudo-label assignment. Finally, we identify the issue of geometric imbalance in the embedding space and provide a simple metric to filter out geometrically imbalanced nodes. Extensive experiments on real-world benchmark datasets are conducted, and the empirical results show that our method significantly outperforms current state-of-the-art methods consistent on different datasets with different imbalance ratios.
['Zengfeng Huang', 'Min Zhou', 'Bisheng Li', 'Shengzhong Zhang', 'Liang Yan']
2023-03-18
null
null
null
null
['pseudo-label']
['miscellaneous']
[ 9.25315768e-02 2.55148411e-01 -5.16726375e-01 -4.19296533e-01 -3.84450912e-01 -7.00488627e-01 4.33840454e-01 1.41824275e-01 -2.30436884e-02 7.48075664e-01 -1.01888180e-01 -2.57451922e-01 -6.79362863e-02 -1.11483610e+00 -3.85299981e-01 -8.82811129e-01 -1.21850304e-01 5.91218054e-01 2.21006617e-01 -2.17422381e-01 2.78082341e-01 4.23001975e-01 -1.32721281e+00 -2.20239967e-01 1.06060004e+00 6.77160621e-01 -3.89515519e-01 2.54672676e-01 -2.66601592e-01 6.74371779e-01 -7.93963313e-01 -5.65539062e-01 5.26677608e-01 -4.41530883e-01 -6.71789646e-01 1.64180174e-01 5.04024804e-01 -1.96116820e-01 -4.68103856e-01 1.24046528e+00 6.57003105e-01 -2.62320191e-01 9.01292682e-01 -1.86889243e+00 -5.65147221e-01 8.41893315e-01 -9.77166414e-01 -1.09822050e-01 -5.33186421e-02 -1.47067204e-01 1.10615408e+00 -7.91420162e-01 7.45630085e-01 1.24644566e+00 9.32965338e-01 5.14672637e-01 -1.29007816e+00 -8.97711039e-01 1.82742849e-01 -4.12103683e-02 -1.76971853e+00 -2.27283537e-01 1.14557564e+00 -2.66228616e-01 3.98211181e-02 3.36771369e-01 4.33211058e-01 8.90974104e-01 -2.43284360e-01 6.17307901e-01 9.02790904e-01 -3.70531380e-01 2.79677719e-01 2.13109419e-01 -1.53461553e-03 7.19897926e-01 6.24108672e-01 -8.91327038e-02 -1.44768387e-01 -3.55994523e-01 4.99643236e-01 2.06417903e-01 -5.81281818e-02 -8.23745131e-01 -1.17186654e+00 9.48962331e-01 8.85153294e-01 1.56861797e-01 -2.70068180e-03 1.17795691e-01 4.46631312e-01 2.81418353e-01 8.01219463e-01 3.74814242e-01 -2.75993228e-01 4.18801576e-01 -8.54302764e-01 2.48895325e-02 8.03840399e-01 9.65653062e-01 9.86580789e-01 -3.21597233e-02 -1.64373562e-01 9.83153880e-01 4.15782332e-01 4.29322094e-01 1.81993932e-01 -6.58365011e-01 3.82278860e-01 1.00587976e+00 -1.57865271e-01 -1.39992619e+00 -4.82306987e-01 -8.52235973e-01 -1.05498934e+00 -3.28019783e-02 6.35086477e-01 -1.48503914e-01 -8.49797666e-01 1.69430017e+00 7.20628858e-01 3.07741940e-01 -2.26980701e-01 7.70199776e-01 7.74708629e-01 3.77964914e-01 -4.28182147e-02 -6.15689717e-02 1.13553762e+00 -1.21244216e+00 -5.40752470e-01 -4.19473685e-02 1.00234163e+00 -6.82003736e-01 9.20694053e-01 7.88592026e-02 -5.11131823e-01 -2.48038471e-01 -1.22415328e+00 3.92217368e-01 -4.75220770e-01 2.12282032e-01 8.11940312e-01 1.05499256e+00 -1.10116065e+00 7.15246141e-01 -4.89157557e-01 -2.70870984e-01 5.84003389e-01 3.68487328e-01 -1.01849541e-01 -1.54916957e-01 -1.19072533e+00 3.19211304e-01 2.75839090e-01 3.61251645e-02 -5.78657687e-01 -7.11696506e-01 -7.82703876e-01 -1.44573823e-01 5.11398256e-01 -2.39001200e-01 8.28856587e-01 -8.79927158e-01 -1.06144989e+00 8.19369018e-01 4.82793488e-02 8.06509107e-02 6.49480760e-01 5.88696480e-01 -4.84751821e-01 1.68320015e-01 2.62373030e-01 7.54773796e-01 7.45368183e-01 -1.54174066e+00 -4.72686499e-01 -5.13540685e-01 -9.57457647e-02 1.51539013e-01 -8.75826359e-01 -5.26446223e-01 -3.66941363e-01 -7.33486950e-01 6.20119810e-01 -1.00045133e+00 -5.19721806e-01 1.36879280e-01 -7.97573030e-01 -3.36917281e-01 9.69651282e-01 -1.14738513e-02 1.28860188e+00 -2.00041533e+00 -2.10181892e-01 7.61753261e-01 7.94380963e-01 -1.75678413e-02 -2.25363940e-01 3.91254723e-01 -7.60702565e-02 5.53873479e-01 -1.18976161e-02 -2.24954113e-01 -1.61364973e-02 3.37343253e-02 -1.58708543e-01 8.23930323e-01 5.76014929e-02 7.73088574e-01 -1.32441711e+00 -7.65782058e-01 4.52503003e-02 2.50412047e-01 -3.22774172e-01 -1.23617046e-01 -4.35235649e-02 2.86111325e-01 -6.31497264e-01 1.05962420e+00 1.04645360e+00 -5.08431494e-01 3.93820107e-01 -2.01341480e-01 5.34139454e-01 2.08920717e-01 -1.28931177e+00 1.22023296e+00 2.02149153e-02 2.29073748e-01 -1.51824176e-01 -1.12492430e+00 1.18482471e+00 -1.09760407e-02 5.51848054e-01 -4.55125034e-01 2.25028634e-01 3.67783725e-01 -8.37994516e-02 -1.06789783e-01 4.06383306e-01 3.42175663e-02 -1.82521045e-01 8.23362470e-01 -1.76078662e-01 1.18986741e-01 3.58108163e-01 5.88862598e-01 1.25997317e+00 -3.73134464e-01 1.84458692e-03 -4.91225213e-01 3.65014315e-01 -9.86229479e-02 6.36147082e-01 6.05134845e-01 -3.60347450e-01 6.92188323e-01 7.25984156e-01 -5.62787473e-01 -9.01218534e-01 -1.07414591e+00 -1.57267347e-01 1.13095272e+00 6.41810417e-01 -3.80257875e-01 -7.75494874e-01 -1.32227981e+00 9.60569754e-02 1.98183551e-01 -7.51080275e-01 -2.72247314e-01 -4.48715866e-01 -1.22628641e+00 7.30789244e-01 3.46417785e-01 1.80500656e-01 -8.57432604e-01 1.78930447e-01 3.23743112e-02 -2.57608742e-01 -8.92438173e-01 -3.77968043e-01 -5.26495539e-02 -9.36402798e-01 -1.26354778e+00 -4.77166295e-01 -1.00525069e+00 1.47042477e+00 6.19624257e-01 1.38525522e+00 6.53496683e-01 -4.30477142e-01 -1.80875227e-01 -5.95049322e-01 -9.10285413e-02 -3.15228224e-01 7.62142718e-01 2.66781989e-02 9.63105354e-03 4.01344091e-01 -5.96593261e-01 -6.92887664e-01 6.64460838e-01 -8.37749302e-01 -1.49529904e-01 6.32480800e-01 8.75981748e-01 4.85798627e-01 2.91920781e-01 9.14059699e-01 -1.56787658e+00 4.22036707e-01 -8.10783982e-01 -3.89409274e-01 2.58628994e-01 -8.12262714e-01 8.70018527e-02 8.32861602e-01 -4.68566954e-01 -5.07429838e-01 -1.23402998e-01 1.06126510e-01 -7.67835826e-02 1.79980263e-01 6.56440854e-02 -3.58683169e-01 -2.90105462e-01 7.82311380e-01 -2.27868870e-01 2.58724168e-02 -2.27079749e-01 1.83405071e-01 5.94490588e-01 1.39810190e-01 -5.65248549e-01 1.38891578e+00 7.79559135e-01 1.61474049e-01 -3.88818234e-01 -8.78794611e-01 -3.46245468e-01 -4.99075741e-01 -2.50726849e-01 -1.55682527e-02 -8.28679860e-01 -4.31324005e-01 4.96631354e-01 -6.31557882e-01 -2.49045640e-01 -2.32090384e-01 1.39049307e-01 -3.82215343e-02 3.29082608e-01 -6.00949228e-01 -5.90504289e-01 -2.25798041e-01 -1.08953547e+00 8.94383252e-01 2.78036982e-01 -1.53942080e-02 -1.14010596e+00 -1.14617370e-01 1.16081849e-01 4.34986353e-01 3.35230380e-01 8.21340024e-01 -7.87347436e-01 -6.37154281e-01 -3.35187584e-01 -5.25350630e-01 1.60257258e-02 2.21311837e-01 2.91962415e-01 -8.56712937e-01 -6.31436884e-01 -4.92367297e-01 -4.17869896e-01 7.19363332e-01 6.08023480e-02 1.38362145e+00 -5.49148954e-02 -7.51816332e-01 6.95069671e-01 1.32305336e+00 -4.21886504e-01 4.57829386e-01 1.14009082e-01 9.32184279e-01 7.51593232e-01 7.44068086e-01 3.33618492e-01 5.14948010e-01 2.96972901e-01 6.70692921e-01 -3.80874425e-01 -2.15724066e-01 -5.81773818e-01 -6.51467219e-02 9.35287952e-01 4.06675428e-01 -3.94563198e-01 -7.58629799e-01 6.11848295e-01 -1.57881939e+00 -5.73251128e-01 -1.84191853e-01 1.97745407e+00 8.01781714e-01 2.42900431e-01 1.79357395e-01 4.47923511e-01 1.22559261e+00 3.02281469e-01 -6.16291344e-01 2.40513846e-01 -3.39774005e-02 -1.07656039e-01 7.75068462e-01 3.56850326e-01 -1.13569462e+00 8.39265883e-01 5.76916742e+00 1.16564274e+00 -1.05036676e+00 -7.30293542e-02 1.08390641e+00 2.24618658e-01 -7.16854692e-01 1.94998737e-02 -9.12730992e-01 6.02938533e-01 3.87179106e-01 -9.11692064e-03 2.04102263e-01 9.24084604e-01 -1.84288010e-01 2.43062228e-01 -9.14862633e-01 7.82487452e-01 1.96793601e-01 -1.19348192e+00 -1.27402410e-01 2.61292160e-01 1.10427451e+00 -9.53116640e-02 1.13452084e-01 2.99642205e-01 5.12051880e-01 -1.10085118e+00 4.29697663e-01 -1.23579822e-01 1.05378222e+00 -7.72173405e-01 7.90211499e-01 2.47409880e-01 -1.29006743e+00 5.46172336e-02 -6.50874436e-01 7.42051005e-02 -3.21905404e-01 1.21374381e+00 -1.06587970e+00 3.26158434e-01 5.14827669e-01 6.73122883e-01 -9.55418706e-01 1.01988924e+00 -3.01765025e-01 6.43175364e-01 -5.65081656e-01 -2.39484772e-01 2.64374390e-02 -1.74320087e-01 3.10540915e-01 6.92271888e-01 2.47058451e-01 -3.82758975e-01 3.08391631e-01 6.65920377e-01 -6.49074137e-01 3.28749895e-01 -6.16143525e-01 2.03895792e-01 1.09230399e+00 1.72939718e+00 -1.45262170e+00 -1.90093011e-01 4.34581712e-02 5.81967473e-01 4.56580788e-01 3.05600703e-01 -8.74703109e-01 -5.94554305e-01 2.66970158e-01 3.25389802e-01 3.97699745e-03 2.87032664e-01 -3.58024120e-01 -1.06449914e+00 9.27874893e-02 -6.36414230e-01 2.57429332e-01 3.74678895e-03 -1.65448117e+00 4.92932826e-01 -2.80544281e-01 -1.53086698e+00 1.25212045e-02 -3.25751483e-01 -6.70428157e-01 3.76403540e-01 -1.45573342e+00 -1.03059804e+00 -6.48388684e-01 1.48946539e-01 1.84403211e-01 5.39974645e-02 5.39782882e-01 5.78477919e-01 -6.60653710e-01 1.08827364e+00 3.53670359e-01 3.89444262e-01 9.58352685e-01 -1.27014554e+00 5.59346020e-01 5.35896063e-01 8.29806998e-02 4.15382057e-01 4.83207345e-01 -6.23928666e-01 -1.07899666e+00 -1.40621006e+00 8.20683241e-01 -3.00733298e-01 4.43673700e-01 -7.24036455e-01 -6.72767878e-01 4.29511517e-01 -4.27837253e-01 6.30597889e-01 7.80028105e-01 -1.95687357e-02 -6.17864728e-01 -4.21511948e-01 -1.49100232e+00 7.09488750e-01 1.25686657e+00 -9.79294255e-02 1.27087057e-01 5.96221507e-01 6.92553580e-01 -1.93753555e-01 -7.59703815e-01 7.78245807e-01 3.76433492e-01 -9.60969687e-01 1.13634861e+00 -2.30847239e-01 2.61338592e-01 -5.52983284e-01 2.55393744e-01 -1.48939252e+00 -2.12094501e-01 -3.77137512e-01 4.81492095e-02 1.53521264e+00 5.69353938e-01 -7.76865423e-01 1.47889698e+00 1.18637875e-01 2.16668755e-01 -9.60651517e-01 -5.35480380e-01 -5.97903669e-01 -6.13739379e-02 1.48973078e-01 9.38109636e-01 1.27344322e+00 -2.40502968e-01 2.90302545e-01 -2.11362228e-01 -1.24700945e-02 1.01905048e+00 1.87416881e-01 8.72760952e-01 -1.74287534e+00 2.79491395e-02 -3.93390536e-01 -4.41961586e-01 -9.77569818e-01 1.71054646e-01 -9.75086749e-01 -1.12601861e-01 -1.21121967e+00 2.58255631e-01 -1.29556072e+00 -1.56772226e-01 3.33523810e-01 -3.57653528e-01 8.73863816e-01 -2.49674976e-01 2.99541831e-01 -9.33955073e-01 4.85334814e-01 1.36361146e+00 -1.71516806e-01 4.07561697e-02 -1.30634857e-02 -1.01992559e+00 7.16013253e-01 9.95542884e-01 -7.72263527e-01 -5.69700599e-01 -2.25751981e-01 4.16058809e-01 -5.21939635e-01 1.38120592e-01 -9.46106970e-01 3.95974703e-02 7.84087460e-04 5.54979980e-01 -6.65887237e-01 -5.74693419e-02 -8.70738268e-01 -1.87536135e-01 4.02004659e-01 -2.36234859e-01 -8.53427127e-02 -4.42931086e-01 6.41226411e-01 6.75159097e-02 -3.49809885e-01 7.95207858e-01 1.00597784e-01 -3.90187681e-01 5.93006670e-01 1.39050245e-01 5.36361873e-01 1.35323608e+00 -2.39688382e-01 -6.99046016e-01 -3.17343622e-01 -1.55910522e-01 5.27828813e-01 9.06063259e-01 1.90106407e-01 2.36966148e-01 -1.53156412e+00 -6.70965612e-01 3.76245528e-01 3.19212765e-01 2.74473608e-01 1.62873030e-01 6.03445172e-01 -7.15814292e-01 -1.78938165e-01 1.27570078e-01 -6.70031071e-01 -9.47559059e-01 4.01223212e-01 1.07540406e-01 -3.68358225e-01 -2.83108056e-01 9.36063290e-01 -6.30460354e-03 -1.13494921e+00 3.81219566e-01 2.72083223e-01 -1.44972160e-01 2.02701077e-01 2.29807511e-01 5.61607480e-01 1.78323030e-01 -6.61119878e-01 -2.70085663e-01 4.24863428e-01 -2.49377996e-01 3.88947517e-01 1.12630057e+00 -1.55174583e-01 -3.17509711e-01 1.91555411e-01 1.43694901e+00 2.19649181e-01 -9.87751126e-01 -3.09830517e-01 -2.16420833e-02 -6.98411703e-01 -2.96931893e-01 -3.63106132e-01 -1.54859483e+00 7.82847762e-01 2.78710067e-01 6.16562426e-01 8.80992830e-01 -6.52924255e-02 8.00090849e-01 2.47080982e-01 3.79139930e-01 -1.16112244e+00 3.61289769e-01 2.29999214e-01 1.51555508e-01 -1.32956529e+00 2.49073952e-01 -9.78039503e-01 -2.19545424e-01 8.01629007e-01 1.01503956e+00 -2.80268103e-01 6.87811911e-01 2.86320060e-01 3.80342841e-01 -1.30792141e-01 -3.53957266e-01 1.69666499e-01 -1.97717294e-01 6.68591261e-01 2.73568392e-01 2.07582980e-01 -3.53464276e-01 2.98043430e-01 -3.55819315e-01 -4.36171651e-01 5.08594513e-01 7.26147652e-01 -3.96375924e-01 -1.41542482e+00 -5.25091827e-01 8.57744694e-01 -3.86847079e-01 1.43434733e-01 -4.41124588e-01 6.98047340e-01 1.37412637e-01 7.88129210e-01 1.12196375e-02 -5.63424587e-01 2.10378729e-02 -2.12432608e-01 2.02252120e-01 -5.59225559e-01 -3.84622335e-01 -1.44699886e-01 -2.31933340e-01 -1.63807854e-01 -2.03226417e-01 -2.55918175e-01 -1.16894972e+00 -5.44067621e-01 -7.09803462e-01 3.98769498e-01 5.60998559e-01 5.19018233e-01 2.79119015e-01 4.68490660e-01 1.22342169e+00 -8.34248185e-01 -9.75977421e-01 -8.51363420e-01 -7.79188633e-01 6.98897481e-01 1.42184161e-02 -8.94006014e-01 -8.29255581e-01 -5.21932960e-01]
[7.304291248321533, 6.009034633636475]