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
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
19ede530-87c8-42b1-a9f4-fa1edb208b89
|
variational-transformer-a-framework-beyond
|
2205.14458
| null |
https://arxiv.org/abs/2205.14458v2
|
https://arxiv.org/pdf/2205.14458v2.pdf
|
Variational Transformer: A Framework Beyond the Trade-off between Accuracy and Diversity for Image Captioning
|
Accuracy and Diversity are two essential metrizable manifestations in generating natural and semantically correct captions. Many efforts have been made to enhance one of them with another decayed due to the trade-off gap. In this work, we will show that the inferior standard of accuracy draws from human annotations (leave-one-out) are not appropriate for machine-generated captions. To improve diversity with a solid accuracy performance, we exploited a novel Variational Transformer framework. By introducing the "Invisible Information Prior" and the "Auto-selectable GMM", we instruct the encoder to learn the precise language information and object relation in different scenes for accuracy assurance. By introducing the "Range-Median Reward" baseline, we retain more diverse candidates with higher rewards during the RL-based training process for diversity assurance. Experiments show that our method achieves the simultaneous promotion of accuracy (CIDEr) and diversity (self-CIDEr), up to 1.1 and 4.8 percent. Also, our method got the most similar performance of the semantic retrieval compared to human annotations, with 50.3 (50.6 of human) for R@1(i2t).
|
['Lianghua He', 'Yitao Peng', 'Yihang Liu', 'Longzhen Yang']
|
2022-05-28
| null | null | null | null |
['semantic-retrieval']
|
['natural-language-processing']
|
[-1.44306459e-02 1.91651180e-01 -9.52579007e-02 -6.19172573e-01
-1.45469928e+00 -4.87218201e-01 6.73434377e-01 -3.14511508e-01
-4.44658816e-01 9.00115728e-01 2.39797100e-01 6.34885281e-02
9.16740149e-02 -3.81182492e-01 -8.59191895e-01 -8.02239299e-01
3.75158429e-01 6.48596525e-01 9.61727276e-02 -3.53163630e-01
2.47400850e-01 2.52671719e-01 -1.72111094e+00 4.48285967e-01
1.31991613e+00 1.02661645e+00 4.09833223e-01 5.05985975e-01
-1.90592289e-01 7.22899377e-01 -7.79727221e-01 -9.89782155e-01
2.71484315e-01 -5.56811750e-01 -5.11276484e-01 -2.16084927e-01
5.33320308e-01 -2.74081677e-01 -2.88675100e-01 1.05088317e+00
7.91923642e-01 1.60949647e-01 8.38777125e-01 -1.23970437e+00
-1.18967283e+00 7.70860314e-01 -6.98770404e-01 6.56247139e-02
2.77016550e-01 3.69050741e-01 1.28899789e+00 -1.01754344e+00
4.64026541e-01 1.51794386e+00 3.74868631e-01 9.45872784e-01
-8.92808497e-01 -7.40724444e-01 2.11471200e-01 8.48789811e-02
-1.44616079e+00 -4.78726864e-01 4.62729931e-01 -1.61870226e-01
6.65988207e-01 5.59686959e-01 3.50031734e-01 1.42755795e+00
-3.47814150e-02 1.09655118e+00 9.46901560e-01 -3.59137565e-01
-5.83745353e-03 7.51024187e-01 3.23176719e-02 3.71691108e-01
2.38809749e-01 1.45099014e-02 -4.68167841e-01 -3.77810448e-02
6.52007937e-01 -4.43916023e-01 -2.83337802e-01 -2.06259921e-01
-1.05311692e+00 8.69556725e-01 4.59001333e-01 1.81820154e-01
-2.15602696e-01 2.14310110e-01 2.34666258e-01 1.58836782e-01
4.64567423e-01 7.73020208e-01 -3.35152417e-01 -1.41496986e-01
-8.05637717e-01 3.78109843e-01 4.57996994e-01 1.24493587e+00
4.43020791e-01 1.45424441e-01 -9.80223238e-01 1.04487538e+00
3.40685844e-01 8.97524893e-01 5.74867368e-01 -1.10064185e+00
5.37202597e-01 2.67296821e-01 4.34168100e-01 -7.68465340e-01
-1.92100182e-03 -7.96226978e-01 -8.10330570e-01 -1.70386910e-01
2.98319697e-01 -1.89873599e-03 -1.11167049e+00 2.01211882e+00
9.13457274e-02 -3.07451785e-01 7.83566162e-02 1.06525373e+00
6.56670988e-01 8.39059114e-01 4.18179095e-01 -1.12862550e-01
1.33267820e+00 -1.10519421e+00 -8.15325141e-01 -1.16972938e-01
5.17952323e-01 -7.85791576e-01 1.59962344e+00 8.61994773e-02
-1.09381115e+00 -5.93746245e-01 -8.09595883e-01 -7.20745698e-02
2.19239183e-02 3.38043988e-01 4.94391173e-01 6.57668889e-01
-1.14683580e+00 4.61658984e-01 -4.33823377e-01 6.86862469e-02
4.09699440e-01 1.39023274e-01 -5.48460931e-02 3.58495750e-02
-1.34439528e+00 1.08194911e+00 2.11203724e-01 -1.63702637e-01
-9.76512194e-01 -7.96589613e-01 -5.73387980e-01 1.50877684e-01
1.62856340e-01 -8.96531582e-01 1.36422598e+00 -1.17829084e+00
-1.55082822e+00 8.72006595e-01 -9.08995271e-02 -3.59934628e-01
8.32461119e-01 -6.46268308e-01 -1.65868729e-01 -1.39505537e-02
3.51524264e-01 1.19463241e+00 6.49230063e-01 -1.58951128e+00
-4.53794599e-01 -1.95326522e-01 2.24599745e-02 5.41046679e-01
-2.55049974e-01 5.23764975e-02 -4.85746384e-01 -6.79163694e-01
-1.15839101e-01 -9.37065542e-01 -9.06315297e-02 -2.23244727e-01
-3.98017734e-01 -4.38953549e-01 3.83243501e-01 -7.86609709e-01
1.16521370e+00 -2.10393786e+00 3.36192735e-02 -2.11567618e-02
6.67494535e-02 2.93300867e-01 -2.92170107e-01 1.62500218e-01
2.47938871e-01 3.36408406e-01 -1.77311152e-01 -4.84437615e-01
1.43851563e-01 6.95970133e-02 -4.25068855e-01 1.80865988e-01
3.29330087e-01 9.45824862e-01 -8.47931802e-01 -6.31236374e-01
-9.76957977e-02 5.35181642e-01 -8.46269846e-01 4.23403591e-01
-3.82268667e-01 5.92074156e-01 -5.62649846e-01 4.40645814e-01
6.07106090e-01 -4.54457730e-01 -6.64030313e-02 -1.05093136e-01
1.51519492e-01 1.46308661e-01 -9.36929584e-01 1.63396573e+00
-3.67953539e-01 4.13563341e-01 -3.53871226e-01 -5.44280112e-01
9.78206098e-01 2.38282531e-01 2.31944844e-01 -1.13759100e+00
9.19422582e-02 3.62861663e-01 -1.66019842e-01 -4.58678454e-01
8.07983041e-01 1.39073819e-01 -9.36627761e-02 6.19333647e-02
3.49635258e-02 -9.93827134e-02 7.54096508e-02 3.21840107e-01
5.81155241e-01 2.70446390e-01 -1.55255854e-01 -4.03972715e-01
3.64506334e-01 -1.38300031e-01 5.76298833e-01 1.00746500e+00
-1.49649352e-01 8.94608200e-01 5.26270688e-01 -9.77093801e-02
-1.43754387e+00 -1.05697799e+00 -1.17920063e-01 1.08814979e+00
3.38496417e-01 -7.13533312e-02 -8.85862172e-01 -6.68125927e-01
-2.00749651e-01 1.24089432e+00 -4.69089866e-01 -4.18512732e-01
-6.08596683e-01 -6.52397811e-01 7.20995605e-01 3.53496820e-01
7.11037874e-01 -8.95282865e-01 -3.04360956e-01 -4.06243503e-02
-4.54039097e-01 -1.02906191e+00 -5.65874159e-01 -2.36206427e-01
-5.99995136e-01 -4.52121377e-01 -1.13577425e+00 -6.39642596e-01
5.67350566e-01 1.96596920e-01 1.27419770e+00 -1.36064850e-02
1.46003485e-01 1.12006448e-01 -5.73616743e-01 -3.66041988e-01
-5.12222111e-01 2.19594970e-01 6.74069971e-02 -8.34944844e-02
9.79812965e-02 -3.19857568e-01 -6.39075398e-01 3.45132291e-01
-7.17845678e-01 1.52719870e-01 7.00381815e-01 1.00851941e+00
5.32377779e-01 -4.55154657e-01 7.28015244e-01 -5.39719760e-01
6.97791100e-01 -4.22389388e-01 -5.16660273e-01 5.43855965e-01
-8.58277500e-01 4.26731855e-01 4.44400191e-01 -6.41567290e-01
-1.24632275e+00 -9.70711857e-02 -2.74156004e-01 -5.75592279e-01
1.79105699e-01 -1.82163864e-01 -3.80872756e-01 3.37546378e-01
7.48680770e-01 2.59384930e-01 -1.78201824e-01 -3.55071336e-01
6.74872279e-01 7.74498701e-01 3.93669039e-01 -8.60898674e-01
4.78907019e-01 1.21859219e-02 -5.96369922e-01 -3.22643250e-01
-1.23406160e+00 -2.03969821e-01 -5.10610193e-02 4.98247938e-03
9.15055156e-01 -1.12346959e+00 -5.61343908e-01 2.71130204e-01
-1.40329945e+00 -2.95774844e-02 -4.73872066e-01 4.37807113e-01
-4.42960382e-01 2.05264255e-01 -4.95812535e-01 -1.03444374e+00
-5.94506025e-01 -1.37587714e+00 1.38442707e+00 3.48150998e-01
-1.56616330e-01 -4.27082211e-01 -2.03264449e-02 4.77978170e-01
6.03067994e-01 -1.01895899e-01 6.75695181e-01 -8.65225732e-01
-6.17388368e-01 3.72738391e-02 -3.44497293e-01 5.57536840e-01
-1.44530728e-01 -4.25599962e-02 -1.14713931e+00 -1.50855154e-01
-7.43983835e-02 -3.89346361e-01 7.09489763e-01 3.34983826e-01
1.20251560e+00 -5.59460461e-01 -1.68557540e-01 3.63323301e-01
1.20877182e+00 3.33789170e-01 8.87862861e-01 2.32465670e-01
7.06772149e-01 5.06297469e-01 6.58383012e-01 4.30385441e-01
3.72731209e-01 8.36901963e-01 2.46863231e-01 6.55612070e-03
-2.72496104e-01 -5.95835268e-01 3.05766881e-01 8.05541396e-01
7.19616711e-02 -6.50606215e-01 -7.08321214e-01 4.91653621e-01
-1.70679176e+00 -1.00248694e+00 3.54447924e-02 2.20049477e+00
1.22073841e+00 6.16630875e-02 -6.96091652e-02 -3.23254168e-01
9.17851567e-01 -4.96802777e-02 -4.41694498e-01 -2.43335038e-01
-3.55845004e-01 -1.64920405e-01 3.59603882e-01 6.21657014e-01
-7.36269772e-01 1.26317561e+00 6.66391325e+00 1.25106537e+00
-7.61529863e-01 1.24083236e-02 1.15096283e+00 -6.89527765e-02
-8.88746440e-01 -1.06879309e-01 -1.07435644e+00 7.05687881e-01
1.06059921e+00 -8.98443758e-02 2.08662927e-01 1.01710415e+00
3.46059464e-02 7.80536011e-02 -1.05106413e+00 1.11584044e+00
1.27432302e-01 -1.05673742e+00 6.14501953e-01 -1.46066785e-01
1.05089653e+00 -3.42621505e-01 3.36565495e-01 5.93350053e-01
4.30308849e-01 -1.12243390e+00 1.01124775e+00 6.85096383e-01
8.58967483e-01 -7.09689856e-01 7.68573821e-01 3.16710651e-01
-6.76645279e-01 3.46720889e-02 -3.98960620e-01 4.11193877e-01
3.30151111e-01 8.12867582e-01 -7.43344128e-01 5.48860610e-01
5.97220898e-01 1.73739806e-01 -4.33368355e-01 8.59483182e-01
-1.86664060e-01 3.42460006e-01 -1.42543763e-01 -3.19539964e-01
2.33850464e-01 -8.52304474e-02 6.07339442e-01 1.16890359e+00
4.24244672e-01 8.91763419e-02 -8.27389732e-02 1.06210804e+00
-3.59904945e-01 4.19453755e-02 -3.83263707e-01 -1.48281306e-02
7.28594661e-01 8.98201883e-01 -1.97796881e-01 -5.55162370e-01
1.07371904e-01 1.14943182e+00 2.91968733e-01 2.83049762e-01
-1.32652390e+00 -3.15626442e-01 5.28701067e-01 -8.08510557e-03
7.34172985e-02 1.24043576e-01 -4.56764370e-01 -1.12850249e+00
1.36667013e-01 -1.19219017e+00 6.53416365e-02 -9.91039038e-01
-1.42791605e+00 8.64106715e-01 1.20054059e-01 -1.07874501e+00
-3.03106457e-01 -3.54937017e-01 -8.19843635e-02 8.97016287e-01
-1.42185307e+00 -9.91580665e-01 -2.60726511e-01 2.90226996e-01
7.08092391e-01 -2.74115145e-01 5.49309492e-01 6.14535332e-01
-4.02486235e-01 1.11112607e+00 8.42309296e-02 -7.93266222e-02
7.11321533e-01 -1.08149016e+00 3.07453722e-01 6.38415873e-01
1.01645179e-01 4.51421678e-01 8.52141976e-01 -5.51600695e-01
-1.00833154e+00 -8.63753438e-01 1.16134715e+00 -7.84713924e-01
2.25323334e-01 -1.77015692e-01 -8.21171761e-01 4.18374181e-01
1.40221357e-01 -5.02920389e-01 3.67841929e-01 -1.49536446e-01
-5.25840938e-01 -2.20899791e-01 -1.22860289e+00 7.94978917e-01
1.10989285e+00 -3.07467103e-01 -5.12881100e-01 4.26834822e-01
1.44011140e+00 -5.11825085e-01 -6.29415572e-01 6.03072047e-01
4.51632947e-01 -9.60848689e-01 9.59253609e-01 -4.81958061e-01
5.89709044e-01 -9.20424983e-02 -4.68464911e-01 -1.13470411e+00
-3.40849221e-01 -4.82456625e-01 1.63928978e-02 1.43453801e+00
6.41278625e-01 -4.51670915e-01 4.97261733e-01 7.17721462e-01
-2.40220204e-01 -8.42061222e-01 -7.67011046e-01 -7.82358944e-01
2.54266322e-01 -1.74224764e-01 7.79551804e-01 7.81161726e-01
-6.07829213e-01 4.62908208e-01 -6.36267722e-01 3.12642120e-02
4.20763105e-01 -2.14401394e-01 5.91579795e-01 -8.87058973e-01
-3.41032803e-01 -5.79614162e-01 -4.01306301e-02 -1.35386622e+00
1.67203452e-02 -7.08998680e-01 1.98261976e-01 -1.36038709e+00
4.47020650e-01 -6.37988567e-01 -3.06511343e-01 2.60588855e-01
-5.05658805e-01 -2.68936753e-02 2.28926972e-01 4.72759277e-01
-8.08808923e-01 1.01501775e+00 1.50519717e+00 -5.33236079e-02
-1.91069782e-01 -1.17535137e-01 -9.56359863e-01 2.74886608e-01
9.02216792e-01 -5.46755135e-01 -5.80289781e-01 -7.59617150e-01
8.22963640e-02 1.22405486e-02 3.07362318e-01 -9.50735092e-01
-1.82011455e-01 -8.55918229e-02 3.06079000e-01 -3.49569231e-01
4.70548570e-01 -4.81413186e-01 4.44621257e-02 2.45377943e-01
-8.23745906e-01 1.50558263e-01 1.15725789e-02 3.92714083e-01
-1.19149759e-01 -2.10815817e-01 8.20549965e-01 -1.62627280e-01
-3.98164511e-01 2.26647362e-01 1.77335277e-01 3.52493584e-01
8.28329980e-01 1.22205978e-02 -3.14579844e-01 -5.53441763e-01
-2.97920853e-01 3.35282683e-01 2.29484290e-01 5.69474638e-01
5.12254357e-01 -1.48070216e+00 -1.14329839e+00 -1.53797537e-01
1.21713221e-01 -8.23856071e-02 2.28202626e-01 4.72478777e-01
-3.03131908e-01 4.16418701e-01 -7.85465166e-02 -7.59462357e-01
-7.68120885e-01 2.87889004e-01 3.98109406e-01 -3.57209146e-01
-4.00226623e-01 1.12165570e+00 2.55392879e-01 -2.72492796e-01
5.53975701e-01 1.24469563e-01 -8.18344653e-02 -7.38148093e-02
4.21731174e-01 2.24510267e-01 -1.48057505e-01 -3.98600131e-01
-2.09581330e-01 3.44386488e-01 -2.82577932e-01 -4.06285107e-01
8.93860757e-01 -2.30993018e-01 1.48908779e-01 2.21940741e-01
1.17445469e+00 1.83314793e-02 -1.39869714e+00 -5.40119186e-02
-4.52190675e-02 -6.20164394e-01 -3.03715110e-01 -1.08884311e+00
-8.97313654e-01 7.21447408e-01 7.47088611e-01 9.62410308e-03
8.68130147e-01 1.73311740e-01 8.75448465e-01 1.98110953e-01
3.46932977e-01 -1.04258037e+00 4.10224825e-01 3.89070779e-01
1.14173794e+00 -1.32350516e+00 -2.65573382e-01 -2.58698940e-01
-1.31699717e+00 4.82549995e-01 9.05090988e-01 2.89357509e-02
-9.79302153e-02 -1.06952302e-01 1.01909779e-01 1.00461259e-01
-7.65972018e-01 -9.64559615e-02 4.95867431e-01 4.97929871e-01
5.29125512e-01 1.70399621e-02 -3.63436013e-01 6.92910790e-01
-4.09940094e-01 -3.58247489e-01 7.39954039e-02 3.35608810e-01
-6.97621584e-01 -9.16500568e-01 -1.29658565e-01 3.00511420e-01
-5.29502630e-01 -3.29010814e-01 -2.25920945e-01 7.64009476e-01
-4.01287153e-02 8.80255222e-01 1.17776059e-01 -3.86730611e-01
3.53431046e-01 1.41715677e-02 3.68704289e-01 -3.00571352e-01
-4.47391361e-01 -4.76574227e-02 2.05539629e-01 -5.31357467e-01
-1.42192468e-01 -2.65242696e-01 -1.12509036e+00 -2.72822112e-01
-4.60449040e-01 4.41113442e-01 6.58740699e-01 7.28226602e-01
5.53140461e-01 5.37278175e-01 7.43559122e-01 -4.49016511e-01
-1.04224622e+00 -9.47270513e-01 -2.68559575e-01 5.09481072e-01
2.17410885e-02 -5.70827842e-01 -5.50247371e-01 -1.54410228e-01]
|
[11.041481018066406, 0.9209818243980408]
|
f4062297-a940-409c-af0f-d7b8f7b008fd
|
codex-hacks-hackerrank-memorization-issues
|
2212.02684
| null |
https://arxiv.org/abs/2212.02684v1
|
https://arxiv.org/pdf/2212.02684v1.pdf
|
Codex Hacks HackerRank: Memorization Issues and a Framework for Code Synthesis Evaluation
|
The Codex model has demonstrated extraordinary competence in synthesizing code from natural language problem descriptions. However, in order to reveal unknown failure modes and hidden biases, such large-scale models must be systematically subjected to multiple and diverse evaluation studies. In this work, we evaluate the code synthesis capabilities of the Codex model based on a set of 115 Python problem statements from a popular competitive programming portal: HackerRank. Our evaluation shows that Codex is indeed proficient in Python, solving 96% of the problems in a zero-shot setting, and 100% of the problems in a few-shot setting. However, Codex exhibits clear signs of generating memorized code based on our evaluation. This is alarming, especially since the adoption and use of such models could directly impact how code is written and produced in the foreseeable future. With this in mind, we further discuss and highlight some of the prominent risks associated with large-scale models of source code. Finally, we propose a framework for code-synthesis evaluation using variations of problem statements based on mutations.
|
['Romain Robbes', "Marco D'Ambros", 'Julian Aron Prenner', 'Anjan Karmakar']
|
2022-12-06
| null | null | null | null |
['memorization']
|
['natural-language-processing']
|
[ 1.72577634e-01 3.54054570e-01 4.32919115e-02 -2.89788485e-01
-9.15208101e-01 -8.36438417e-01 4.42186505e-01 2.82086402e-01
-5.49926944e-02 6.15899384e-01 2.23412976e-01 -7.02672839e-01
-3.10226399e-02 -6.58983290e-01 -1.03713751e+00 -2.29584634e-01
-1.48547217e-01 6.70459196e-02 -2.19502702e-01 -3.69653612e-01
7.26954639e-01 -8.55825171e-02 -1.75718307e+00 4.32816327e-01
1.17131138e+00 4.22405675e-02 2.55139530e-01 8.82364810e-01
-7.34772682e-02 1.21124208e+00 -8.50925624e-01 -6.59327686e-01
1.21113151e-01 -2.50754952e-01 -1.02468967e+00 -3.14112455e-01
4.40676212e-01 -1.73543528e-01 2.07380429e-01 1.26350486e+00
5.20415485e-01 -2.91268498e-01 1.32221282e-01 -1.21023500e+00
-6.14044070e-01 1.23254836e+00 -2.50341058e-01 5.92550039e-02
7.12128460e-01 8.01832318e-01 1.14084172e+00 -7.16485083e-01
7.05298543e-01 9.31467414e-01 7.75519788e-01 6.87344790e-01
-1.63444006e+00 -5.94019890e-01 -9.96131822e-02 -4.81812388e-01
-1.30003524e+00 -5.12308538e-01 3.56241405e-01 -7.75885522e-01
1.44216228e+00 4.94368881e-01 4.76151228e-01 1.30219328e+00
3.35134685e-01 5.76655269e-01 1.16559148e+00 -4.58032995e-01
3.11931551e-01 3.93067867e-01 2.45171696e-01 8.08933139e-01
5.17467737e-01 1.72533542e-01 -4.08460617e-01 -7.91017711e-01
1.27395419e-02 -5.71074963e-01 -2.97738254e-01 -1.09573133e-01
-1.15249169e+00 7.02230752e-01 1.66158989e-01 3.15120667e-01
-8.29512253e-03 4.78677869e-01 3.66411299e-01 2.44925320e-01
2.18962878e-02 1.40317869e+00 -4.85096127e-01 -7.22520888e-01
-9.89999175e-01 6.27046883e-01 1.02279866e+00 1.14505088e+00
6.72291100e-01 7.42945969e-02 -1.75518095e-01 4.81692851e-01
2.13714242e-01 2.14102045e-01 6.71675444e-01 -9.82418239e-01
8.38254213e-01 8.62006962e-01 5.76511100e-02 -9.02092516e-01
-1.97392628e-01 -4.28150445e-01 -1.08035766e-01 4.72763687e-01
3.10205996e-01 -3.26810062e-01 -3.66066366e-01 1.70337486e+00
-2.26481870e-01 -4.08275574e-01 1.03473485e-01 6.17007494e-01
5.46905756e-01 5.21452308e-01 1.34366408e-01 2.34480664e-01
1.11350465e+00 -6.76799238e-01 -1.87306404e-01 -6.88299119e-01
1.04440022e+00 -6.67983651e-01 1.50537539e+00 4.50677633e-01
-1.25043523e+00 -3.88025314e-01 -1.21468532e+00 9.26934630e-02
-3.66859809e-02 1.35364041e-01 7.20082283e-01 1.02505600e+00
-1.10692239e+00 5.90963125e-01 -7.58475840e-01 -3.69709641e-01
3.14433455e-01 1.13248937e-01 -9.13483743e-03 -1.67070135e-01
-7.62699008e-01 8.66191685e-01 3.31202447e-01 -2.12111607e-01
-1.05142510e+00 -9.94997978e-01 -7.70677269e-01 3.42875242e-01
1.36921525e-01 -4.52819318e-01 1.46922958e+00 -9.47899520e-01
-9.92785215e-01 8.39146554e-01 9.72398445e-02 -1.77412808e-01
6.32104039e-01 -5.98438606e-02 -1.15147702e-01 -5.18747807e-01
2.08441094e-01 3.21228445e-01 5.00019073e-01 -1.29895186e+00
-3.35674077e-01 -2.32821126e-02 5.01967788e-01 -2.52766728e-01
-2.55937278e-01 1.84552416e-01 4.30516042e-02 -5.37690818e-01
-5.76343536e-01 -1.02412617e+00 -2.99882472e-01 -2.54539937e-01
-3.99531662e-01 7.33511522e-02 -2.69596595e-02 -6.92840576e-01
1.51103127e+00 -2.26718545e+00 1.50977641e-01 1.62077472e-01
4.72588331e-01 -5.63523397e-02 -3.70818526e-01 6.83562398e-01
-2.11141005e-01 7.31308222e-01 -5.12598813e-01 5.29816002e-02
1.96211025e-01 -2.25754321e-01 -4.66979355e-01 1.50150880e-01
4.82364595e-01 7.68843055e-01 -1.03769124e+00 -2.53889620e-01
-3.41312647e-01 1.12722851e-01 -1.21118736e+00 2.53366858e-01
-5.64421356e-01 -8.84589739e-03 -2.32614487e-01 6.23942375e-01
4.02606070e-01 -4.25544322e-01 2.87696242e-01 5.49965501e-01
-3.06896061e-01 3.80395144e-01 -7.86214828e-01 1.74220407e+00
-6.33852661e-01 7.53326118e-01 -1.56682670e-01 -2.51501560e-01
8.10359359e-01 2.48340592e-01 -2.36194670e-01 -5.88040948e-01
-1.48123279e-01 4.28605199e-01 4.18283105e-01 -8.62500072e-01
7.56649971e-01 6.82439581e-02 -2.59383082e-01 8.06329906e-01
-1.33188054e-01 -4.51715738e-01 4.26337034e-01 4.56325799e-01
1.62071252e+00 -9.01493728e-02 1.83858290e-01 -5.28245747e-01
2.83256948e-01 3.14719558e-01 2.14991048e-01 1.08698988e+00
4.46441993e-02 6.73888445e-01 1.02644169e+00 -1.84269220e-01
-1.13967240e+00 -6.48508310e-01 -8.59625190e-02 1.01997089e+00
-3.98530245e-01 -9.60847914e-01 -8.87157500e-01 -6.48865461e-01
3.59774306e-02 1.19608963e+00 -6.78422153e-01 -4.24391001e-01
-3.28524321e-01 -8.65366518e-01 9.27406728e-01 2.73222536e-01
-3.56147066e-02 -1.15568388e+00 -9.69391942e-01 2.05537945e-01
-1.17292836e-01 -5.36299050e-01 -6.97797611e-02 2.06700131e-01
-5.93433976e-01 -1.08338201e+00 -4.23815846e-01 -5.56634963e-01
7.42465794e-01 -1.04452498e-01 1.55339146e+00 7.66690791e-01
-4.90742683e-01 4.29423213e-01 -3.54561061e-01 -2.54937500e-01
-1.21293449e+00 3.56967539e-01 -4.24440861e-01 -8.19314778e-01
2.92093903e-01 -3.86002719e-01 -2.06245497e-01 2.49600261e-02
-9.17413533e-01 6.13136515e-02 5.56317151e-01 8.15520644e-01
-3.44910026e-01 -7.34792575e-02 2.03186318e-01 -1.09959626e+00
1.06974554e+00 -8.50962281e-01 -8.56175900e-01 4.32886243e-01
-6.86150014e-01 2.15478957e-01 6.26444757e-01 -2.28277281e-01
-1.01098919e+00 -1.55525878e-01 -1.31576836e-01 4.01005179e-01
9.98652279e-02 7.99828947e-01 2.06886828e-01 -2.02007517e-01
1.51428246e+00 2.06291661e-01 -1.58814535e-01 -1.98351905e-01
1.65975422e-01 5.89600623e-01 2.24912316e-02 -1.18761039e+00
8.29797566e-01 -1.50292724e-01 -6.87110066e-01 -6.23296142e-01
-1.74259990e-01 1.96843207e-01 3.53832655e-02 -9.03722122e-02
3.84948164e-01 -9.29985464e-01 -5.47337174e-01 3.62850815e-01
-1.20787501e+00 -6.69871509e-01 -1.32285625e-01 8.46291718e-04
-4.25638586e-01 1.03237815e-01 -4.87710923e-01 -6.87934101e-01
3.81006207e-03 -1.73478913e+00 8.69766176e-01 6.83938563e-02
-9.44664896e-01 -7.13313520e-01 3.02447587e-01 4.40585196e-01
7.66462624e-01 7.60229677e-02 1.42068541e+00 -4.39621747e-01
-7.86375940e-01 -2.27317899e-01 -2.21362934e-02 1.91928267e-01
-2.20173731e-01 4.75744605e-01 -9.54675436e-01 -4.79229063e-01
4.08093520e-02 -6.33139968e-01 2.69315064e-01 -2.99291670e-01
1.03296447e+00 -3.42447937e-01 -2.32383870e-02 4.09221977e-01
1.67084730e+00 -1.40950158e-01 6.10072076e-01 4.65042800e-01
4.01037186e-01 5.57320058e-01 1.85126528e-01 6.89988196e-01
3.00405622e-01 2.58906096e-01 4.10176039e-01 3.80332947e-01
2.63293117e-01 -2.69373298e-01 6.46753550e-01 7.89585829e-01
2.65044063e-01 -1.13479257e-01 -1.69418252e+00 6.08453929e-01
-1.34635985e+00 -6.76752985e-01 -2.49898702e-01 2.14655828e+00
1.13626409e+00 2.73677796e-01 -1.59094453e-01 -2.58929357e-02
3.91245216e-01 -3.72098275e-02 -3.20582062e-01 -6.15500867e-01
7.59859309e-02 1.76274434e-01 1.93543866e-01 3.60834092e-01
-4.19189513e-01 5.64374506e-01 7.25151253e+00 4.53283787e-01
-1.21000195e+00 -1.35318726e-01 5.87134242e-01 -1.00300841e-01
-8.93533528e-01 3.51641059e-01 -5.12507081e-01 6.86144471e-01
1.12081909e+00 -6.75244987e-01 7.26095438e-01 1.03357673e+00
-1.64658591e-01 -3.29295099e-01 -1.40231240e+00 4.71427798e-01
-2.94993375e-03 -1.36857319e+00 -1.58237934e-01 -2.05385253e-01
9.74720657e-01 1.52249679e-01 1.61792666e-01 7.50424862e-01
7.28776097e-01 -1.26065743e+00 1.11343300e+00 1.88046768e-01
5.67974985e-01 -4.21385169e-01 5.13552785e-01 5.67780912e-01
-3.87530088e-01 -3.13614488e-01 -2.49354094e-01 -5.66868067e-01
-3.84441406e-01 4.38614488e-01 -1.04380286e+00 1.21440999e-01
6.60124183e-01 1.96018338e-01 -1.28070378e+00 1.07108581e+00
-2.40665495e-01 8.33362341e-01 2.17542108e-02 -3.72766465e-01
2.45579518e-02 4.26034331e-01 3.69079590e-01 1.27055669e+00
3.73345017e-01 -8.36102441e-02 -1.51338652e-01 1.49951518e+00
-5.52894287e-02 1.06093146e-01 -6.43873215e-01 -5.22875667e-01
4.06905174e-01 1.00654125e+00 -3.86651605e-01 -2.87560314e-01
-5.25264680e-01 4.62859392e-01 5.75938940e-01 3.28682989e-01
-8.02598655e-01 -4.39725339e-01 6.16860449e-01 1.39286891e-01
-6.24928176e-02 -8.18786323e-02 -6.27453923e-01 -1.26632690e+00
3.35934162e-01 -1.61314714e+00 -1.14749543e-01 -8.70463312e-01
-1.09998465e+00 5.22703528e-01 -4.06840406e-02 -8.99920881e-01
-9.40184295e-02 -4.56545144e-01 -6.58905447e-01 8.28274488e-01
-1.14418912e+00 -5.76961279e-01 -3.29249710e-01 -1.43828407e-01
3.08289945e-01 -1.97813064e-01 8.43761802e-01 2.29833916e-01
-5.06058097e-01 6.61230206e-01 3.31984796e-02 -9.05723369e-04
5.33482492e-01 -1.27346599e+00 1.00245273e+00 8.86477232e-01
-2.04934657e-01 1.35234082e+00 9.87962902e-01 -6.98399544e-01
-1.59313893e+00 -9.24027920e-01 8.79898131e-01 -8.83315384e-01
7.44496346e-01 -6.21511638e-01 -9.97633517e-01 7.86521733e-01
1.89677596e-01 -2.94005662e-01 6.21587157e-01 2.30826050e-01
-5.12142122e-01 3.26525658e-01 -1.08232951e+00 7.91922152e-01
8.02046120e-01 -7.62407124e-01 -6.07434869e-01 3.75858963e-01
5.42054057e-01 -4.62692887e-01 -7.30111837e-01 -7.90640563e-02
4.46136832e-01 -1.15777349e+00 5.16974270e-01 -6.29223645e-01
1.31556809e+00 -1.22676447e-01 -1.73603356e-01 -1.26660585e+00
-2.11635515e-01 -7.12290049e-01 4.39615935e-01 1.40300190e+00
7.54376650e-01 -5.35560548e-01 6.38610959e-01 1.15231359e+00
1.32779768e-02 -3.43857229e-01 -4.42293048e-01 -6.52790129e-01
4.11805898e-01 -5.25439084e-01 7.41671085e-01 1.12122655e+00
4.90679890e-01 -4.08791527e-02 -1.72451586e-01 -4.85899346e-03
3.68056536e-01 -8.18187967e-02 9.94958937e-01 -7.64742911e-01
-7.42551565e-01 -4.26380038e-01 -2.78653443e-01 -4.38689321e-01
2.23385081e-01 -1.20702052e+00 3.24838668e-01 -1.00038469e+00
5.83389401e-01 -4.77142543e-01 1.92988321e-01 4.15344656e-01
-4.48815882e-01 -2.12885812e-01 4.28192347e-01 1.04748987e-01
-4.55021173e-01 1.90179482e-01 7.66877770e-01 -2.62523472e-01
4.79142256e-02 -1.86627284e-01 -1.11972213e+00 5.47673047e-01
7.63871610e-01 -8.17004561e-01 -2.12586999e-01 -8.26045990e-01
1.01086402e+00 1.14921667e-01 3.58152032e-01 -1.21516705e+00
1.74117580e-01 -1.05593815e-01 -3.09348367e-02 1.67637140e-01
-2.68184185e-01 -5.65682352e-01 4.21770483e-01 7.16222167e-01
-8.09558511e-01 4.14605230e-01 4.43287909e-01 1.48206547e-01
-7.06945488e-04 -5.99434137e-01 5.22227526e-01 -4.94289011e-01
-5.57553232e-01 -4.50325906e-01 -7.66551018e-01 4.85526145e-01
1.04340780e+00 -4.50354405e-02 -6.44899368e-01 3.60317864e-02
-1.20188564e-01 1.71301305e-01 1.02004409e+00 6.91312551e-01
3.19386035e-01 -9.31676984e-01 -6.93444312e-01 3.85863900e-01
6.27589762e-01 -4.72061515e-01 1.23941340e-01 4.67800915e-01
-9.72874403e-01 2.20530197e-01 -2.78981626e-01 -4.36011374e-01
-1.02686357e+00 4.72734660e-01 1.21535830e-01 -1.00386724e-01
-2.09770232e-01 9.32397842e-01 4.03801538e-02 -7.68893361e-01
5.56593686e-02 -5.87544799e-01 3.75928372e-01 -3.56889814e-01
3.91646922e-01 1.84311122e-01 1.69895232e-01 -7.21848058e-03
-4.02614474e-01 3.26223038e-02 -1.82298511e-01 -4.62926738e-02
1.33076537e+00 4.39718634e-01 -4.71876323e-01 4.91824418e-01
1.01518095e+00 3.68216872e-01 -9.29627478e-01 1.87615111e-01
2.30094016e-01 -6.42329872e-01 -3.24859500e-01 -1.05056143e+00
-6.27144933e-01 7.57041037e-01 2.56185979e-01 2.15083405e-01
4.99678284e-01 -3.68327826e-01 1.35755882e-01 5.54965734e-01
7.28765607e-01 -8.10280144e-01 2.95202881e-02 4.80656832e-01
9.13842618e-01 -1.14118230e+00 -3.23981680e-02 -4.46584821e-02
-4.71697360e-01 1.09504056e+00 8.78425896e-01 -4.15237732e-02
1.21160679e-01 4.64232475e-01 6.74434146e-03 -2.27656856e-01
-1.21336997e+00 5.98913312e-01 -3.15484703e-01 4.00086969e-01
8.46085787e-01 -4.64363024e-02 -2.91014314e-01 5.60463369e-01
-4.47531611e-01 1.41893744e-01 1.16103375e+00 1.30140412e+00
-1.86275512e-01 -1.02912319e+00 -5.55748284e-01 3.98971140e-01
-4.52988803e-01 -4.81030464e-01 -4.60164279e-01 6.91536367e-01
-7.46741742e-02 7.47888923e-01 -2.10628435e-01 -4.38656896e-01
2.12105885e-01 1.50730044e-01 3.36079150e-01 -9.99711692e-01
-1.34903944e+00 -7.97593594e-01 2.99559027e-01 -5.21229029e-01
2.13506907e-01 -6.62427545e-01 -8.45847964e-01 -5.25350451e-01
-2.36436203e-01 1.46012157e-01 6.88342452e-01 5.78327894e-01
4.53134328e-01 4.72639680e-01 1.50789022e-01 -5.14692128e-01
-9.38211679e-01 -6.13672793e-01 -1.82361826e-01 4.83342648e-01
2.95719683e-01 -1.33819178e-01 -4.62518036e-01 6.62119687e-02]
|
[7.864603042602539, 7.745560646057129]
|
557ebe3b-aa59-4e7e-ae6c-c5db243bffd8
|
storyer-automatic-story-evaluation-via
|
2210.08459
| null |
https://arxiv.org/abs/2210.08459v2
|
https://arxiv.org/pdf/2210.08459v2.pdf
|
StoryER: Automatic Story Evaluation via Ranking, Rating and Reasoning
|
Existing automatic story evaluation methods place a premium on story lexical level coherence, deviating from human preference. We go beyond this limitation by considering a novel \textbf{Story} \textbf{E}valuation method that mimics human preference when judging a story, namely \textbf{StoryER}, which consists of three sub-tasks: \textbf{R}anking, \textbf{R}ating and \textbf{R}easoning. Given either a machine-generated or a human-written story, StoryER requires the machine to output 1) a preference score that corresponds to human preference, 2) specific ratings and their corresponding confidences and 3) comments for various aspects (e.g., opening, character-shaping). To support these tasks, we introduce a well-annotated dataset comprising (i) 100k ranked story pairs; and (ii) a set of 46k ratings and comments on various aspects of the story. We finetune Longformer-Encoder-Decoder (LED) on the collected dataset, with the encoder responsible for preference score and aspect prediction and the decoder for comment generation. Our comprehensive experiments result in a competitive benchmark for each task, showing the high correlation to human preference. In addition, we have witnessed the joint learning of the preference scores, the aspect ratings, and the comments brings gain in each single task. Our dataset and benchmarks are publicly available to advance the research of story evaluation tasks.\footnote{Dataset and pre-trained model demo are available at anonymous website \url{http://storytelling-lab.com/eval} and \url{https://github.com/sairin1202/StoryER}}
|
['Hideki Nakayama', 'Yusuke Miyao', 'Hiroya Takamura', 'Duc Minh Vo', 'Hong Chen']
|
2022-10-16
| null | null | null | null |
['comment-generation']
|
['natural-language-processing']
|
[ 1.51960552e-01 1.25800490e-01 -3.49197149e-01 -6.00566566e-01
-1.36429775e+00 -8.09321582e-01 6.80857837e-01 1.74940020e-01
-3.56768519e-01 7.18636990e-01 1.03127527e+00 -3.52692753e-02
2.05239534e-01 -6.79599881e-01 -6.58506334e-01 -4.23274875e-01
2.47183904e-01 5.81252337e-01 3.30009796e-02 -3.55255216e-01
5.41595459e-01 -5.14827967e-01 -1.49213529e+00 8.00182164e-01
6.69009566e-01 1.29234600e+00 4.19220001e-01 9.79540765e-01
3.62650484e-01 8.23868811e-01 -5.14392197e-01 -9.61549103e-01
-5.79765886e-02 -3.21289361e-01 -8.64880025e-01 8.80561396e-02
-2.65118051e-02 -5.74625134e-01 -4.19549346e-01 6.40356421e-01
7.53190339e-01 2.12261960e-01 1.09766793e+00 -1.04108584e+00
-1.02198291e+00 1.25472414e+00 -4.50229108e-01 -2.23606988e-03
7.74719417e-01 2.23488018e-01 1.80458486e+00 -1.22694290e+00
6.23454928e-01 6.55547559e-01 3.97224694e-01 5.06479383e-01
-7.85620809e-01 -3.70040298e-01 1.93964481e-01 2.89601713e-01
-1.24421442e+00 -5.29779971e-01 6.68298781e-01 -6.32217050e-01
8.71590137e-01 5.19864559e-01 6.02081060e-01 1.61579192e+00
-3.12388718e-01 1.33275127e+00 7.72189498e-01 5.24648279e-02
1.53825745e-01 2.50270933e-01 1.70032561e-01 1.44695297e-01
-3.03687036e-01 -5.16994186e-02 -9.79165256e-01 3.96925323e-02
5.30145824e-01 -4.30028230e-01 -5.94175696e-01 2.18863636e-01
-1.54581320e+00 8.49312544e-01 9.86314490e-02 -5.42236529e-02
-2.15217784e-01 -4.05612327e-02 7.65686870e-01 2.29295850e-01
5.45393467e-01 4.69426841e-01 -6.20870829e-01 -5.82058787e-01
-7.88254619e-01 4.37643468e-01 6.63645923e-01 1.35345173e+00
2.31515676e-01 -2.49635026e-01 -9.38540220e-01 1.21954775e+00
2.86022455e-01 4.06134635e-01 4.65202123e-01 -8.28105450e-01
8.41352224e-01 3.53863776e-01 1.66485742e-01 -5.49853802e-01
-1.37961760e-01 -3.83802056e-01 -9.25701797e-01 -2.66076535e-01
2.32376859e-01 -1.46152258e-01 -3.53607684e-01 1.88468313e+00
-2.46304907e-02 -2.96503901e-01 1.70336306e-01 9.52185035e-01
1.29135334e+00 9.12707865e-01 -2.35942259e-01 -1.90062925e-01
1.45660484e+00 -1.17965269e+00 -2.69179434e-01 -1.40547469e-01
6.84965491e-01 -9.37750816e-01 1.71205425e+00 5.08729100e-01
-1.23942745e+00 -3.53473485e-01 -9.93021548e-01 -2.05411583e-01
-1.21368386e-01 6.83708727e-01 3.53723586e-01 2.06408501e-01
-1.07291508e+00 4.04430330e-01 -9.80878323e-02 -3.53004038e-01
2.94163108e-01 4.46018912e-02 -1.60182416e-01 1.44758001e-01
-1.29661298e+00 5.45496225e-01 3.95799726e-01 -4.17301841e-02
-1.03154945e+00 -5.97821295e-01 -9.77785349e-01 -1.69805005e-01
1.97103813e-01 -5.96580565e-01 1.58247566e+00 -6.89691305e-01
-1.46617353e+00 1.07617569e+00 -4.66664694e-02 -2.30079100e-01
6.88055634e-01 -5.34061730e-01 -3.69361669e-01 -2.53407955e-01
4.49811637e-01 7.85287023e-01 6.08391464e-01 -1.08450830e+00
-6.40120924e-01 1.13384418e-01 2.32231185e-01 4.49390143e-01
-4.03406203e-01 2.67900169e-01 -6.04213536e-01 -1.17792690e+00
-3.79682094e-01 -6.79065526e-01 2.04799965e-01 -4.13351357e-01
-8.78141582e-01 -3.98348391e-01 2.38649726e-01 -7.70610869e-01
1.64207995e+00 -2.11028433e+00 1.36212528e-01 -1.87006623e-01
-5.76134250e-02 -1.56773314e-01 -3.51573169e-01 6.31426573e-01
-1.25403227e-02 1.83437914e-02 -3.40173960e-01 -5.18500209e-01
2.22634390e-01 -1.65370211e-01 -4.56335157e-01 2.20187947e-01
6.02950305e-02 8.84553552e-01 -9.46875691e-01 -3.95722300e-01
-5.69622703e-02 1.97823301e-01 -7.76813209e-01 4.36038703e-01
-4.04105127e-01 3.89710635e-01 -2.99332172e-01 5.21776438e-01
1.11379951e-01 -3.54767174e-01 -3.54333192e-01 -5.16673364e-02
4.75972667e-02 7.48667479e-01 -1.09671271e+00 1.67263639e+00
-3.95898283e-01 7.22124159e-01 -4.73938912e-01 -3.88515800e-01
9.76926744e-01 5.71869254e-01 1.89471439e-01 -3.93101692e-01
3.44266564e-01 1.87583238e-01 -5.10534346e-01 -4.75538552e-01
1.05495942e+00 9.48736444e-02 -6.02914155e-01 8.62533033e-01
7.01325238e-02 -3.51691395e-01 4.97287214e-01 4.41424161e-01
9.98409510e-01 1.90601066e-01 3.32101405e-01 -1.40543431e-01
4.21006560e-01 -8.28645080e-02 2.64463544e-01 4.97814596e-01
1.55336976e-01 1.17480350e+00 7.21259594e-01 -1.67912617e-01
-1.26489782e+00 -1.01672304e+00 -3.62083986e-02 1.37721622e+00
1.84004586e-02 -7.78668702e-01 -6.03339493e-01 -6.38202131e-01
-4.19999033e-01 1.32218862e+00 -7.92572618e-01 -3.95679474e-02
-4.25271273e-01 -6.32422268e-01 6.09961987e-01 6.95474148e-01
3.87219310e-01 -1.22201693e+00 -4.77988213e-01 -2.85223816e-02
-7.54687786e-01 -9.01433527e-01 -1.17213428e+00 7.76128918e-02
-2.35349983e-01 -7.16417730e-01 -9.01609004e-01 -5.94988108e-01
4.26194012e-01 -1.36192456e-01 1.49912953e+00 -1.94169939e-01
3.28622490e-01 2.28354216e-01 -9.32996809e-01 -3.23580861e-01
-2.16882631e-01 2.28110537e-01 -3.01677417e-02 2.27631796e-02
1.59224674e-01 -5.81034422e-01 -6.56086743e-01 4.69552875e-01
-7.76478529e-01 5.87664068e-01 4.92547840e-01 8.25790823e-01
7.18398035e-01 -2.99165428e-01 5.26878119e-01 -8.28601658e-01
9.57391322e-01 -6.12964928e-01 2.20536627e-02 1.69462442e-01
-5.16087353e-01 -1.87987491e-01 6.21167183e-01 -3.70199412e-01
-9.92094755e-01 -1.68033347e-01 -3.79436433e-01 8.89010653e-02
-5.07987961e-02 5.90373635e-01 -3.46300125e-01 1.09175777e+00
6.64648116e-01 3.33225608e-01 -7.40986764e-01 -1.72987372e-01
5.26635826e-01 8.60533476e-01 7.07933426e-01 -6.93372428e-01
4.59764451e-01 -2.18637157e-02 -7.40233183e-01 -3.91138673e-01
-1.26689637e+00 -5.01833022e-01 -4.22479987e-01 -6.32165611e-01
9.46297228e-01 -1.09696913e+00 -3.99968415e-01 4.11807954e-01
-1.37521923e+00 -5.79246640e-01 -5.85597098e-01 3.83742422e-01
-1.05168045e+00 8.87199044e-02 -6.97302639e-01 -6.76315308e-01
-5.02847672e-01 -1.05435574e+00 1.28783071e+00 1.17867604e-01
-9.37699497e-01 -6.15229845e-01 1.71692148e-01 7.44087577e-01
2.09895354e-02 4.63824458e-02 8.10416818e-01 -7.24715233e-01
-2.58070916e-01 -3.48530829e-01 -2.21154049e-01 2.60135204e-01
-3.67410243e-01 5.38423322e-02 -1.19994771e+00 -3.77493277e-02
-2.77143180e-01 -8.27205539e-01 8.55041742e-01 3.05184096e-01
1.06640708e+00 -5.36414146e-01 1.87167510e-01 3.18045408e-01
1.02508736e+00 -2.10212201e-01 6.54329121e-01 2.61564255e-01
5.13397157e-01 5.65871179e-01 8.28639209e-01 1.10217094e+00
6.91698551e-01 8.82834077e-01 2.53183812e-01 3.56761307e-01
-9.31478962e-02 -5.69780409e-01 8.10527325e-01 1.12121928e+00
-3.71612251e-01 -8.29777539e-01 -6.73036456e-01 7.10023403e-01
-2.06973624e+00 -1.00341928e+00 -2.34731853e-01 2.04831386e+00
1.11794686e+00 2.71683156e-01 1.95660621e-01 2.39342973e-01
5.71982682e-01 5.88849485e-01 -4.87797230e-01 -3.91969562e-01
-4.03866827e-01 -3.89353931e-01 1.17398694e-01 4.64648575e-01
-1.08758104e+00 8.37572992e-01 5.11647892e+00 1.11495280e+00
-6.29011095e-01 1.13896228e-01 1.00887525e+00 -4.25465077e-01
-8.21468592e-01 -2.14284077e-01 -7.00008273e-01 5.10606766e-01
6.08932018e-01 -3.67932409e-01 3.73685390e-01 9.62320685e-01
2.34573573e-01 -4.23971191e-02 -1.36899233e+00 1.13605630e+00
2.80147612e-01 -1.27765894e+00 2.72904932e-01 -3.22305441e-01
7.94678330e-01 -3.12340260e-03 1.41866073e-01 4.48063016e-01
5.04586518e-01 -1.08339417e+00 1.47246599e+00 4.32799518e-01
1.32873130e+00 -4.37489271e-01 7.84272432e-01 3.20268422e-01
-1.39557290e+00 1.61039516e-01 -1.17306478e-01 -1.92070588e-01
5.43255270e-01 6.97410166e-01 -5.72774887e-01 3.28380108e-01
7.43677557e-01 1.00547147e+00 -3.20836276e-01 8.15295875e-01
-7.53576875e-01 7.67216802e-01 7.74737075e-02 -2.26036534e-01
3.13530639e-02 -8.80184211e-03 7.27359772e-01 1.44162571e+00
4.90695029e-01 2.52675354e-01 -2.14690585e-02 8.34603071e-01
-2.40526974e-01 4.60069954e-01 -3.19152504e-01 -5.39290020e-03
4.00112569e-01 1.06696367e+00 -5.71530342e-01 -3.08474630e-01
-2.78557301e-01 1.19227624e+00 2.90652245e-01 3.14378917e-01
-1.04129755e+00 -7.73198679e-02 5.21522999e-01 1.35189980e-01
3.08209926e-01 1.64183229e-01 -6.91882193e-01 -1.21809971e+00
2.19726950e-01 -6.84351385e-01 5.13979077e-01 -1.11646688e+00
-1.35272682e+00 8.67177010e-01 -6.69425055e-02 -1.49267423e+00
-2.31035113e-01 -2.99588501e-01 -8.59190345e-01 6.62554920e-01
-1.09859216e+00 -1.12514579e+00 -1.23142585e-01 5.87405384e-01
1.18945622e+00 -1.96428984e-01 7.01596975e-01 2.97977507e-01
-5.11310101e-01 8.53223920e-01 -1.86032474e-01 2.38726437e-01
8.24330270e-01 -1.34647167e+00 3.98195654e-01 6.35407805e-01
1.64067522e-01 1.95587501e-02 9.56903875e-01 -5.42846262e-01
-7.37884164e-01 -1.17883873e+00 1.34519958e+00 -7.29636610e-01
6.69761181e-01 -4.51511502e-01 -3.66882592e-01 6.06293797e-01
3.51209849e-01 -6.25150144e-01 9.76029038e-01 1.13540351e-01
-4.25733954e-01 2.26447552e-01 -6.68821096e-01 8.23420882e-01
1.05826783e+00 -5.23280084e-01 -4.84728336e-01 4.11282897e-01
1.03647876e+00 -5.01308978e-01 -7.57448673e-01 3.08319181e-01
6.26745105e-01 -1.12633729e+00 6.18366480e-01 -4.04751152e-01
1.32072771e+00 -1.52620137e-01 -5.41503668e-01 -1.29540133e+00
-3.73704076e-01 -5.49898863e-01 1.20547814e-02 1.63139296e+00
1.15269935e+00 -1.39874294e-01 5.27264237e-01 5.68499029e-01
-4.87922639e-01 -1.30214489e+00 -7.00611353e-01 -3.81825626e-01
-7.84162581e-02 -1.03794098e+00 5.99224865e-01 5.67869842e-01
2.67329395e-01 6.61454678e-01 -7.68395782e-01 -1.51125774e-01
1.40378341e-01 1.87598586e-01 5.42169750e-01 -6.81425273e-01
-7.51484454e-01 -7.63851762e-01 2.01059461e-01 -1.33409595e+00
-1.12450309e-01 -9.34750080e-01 2.54531741e-01 -1.69590807e+00
5.78612149e-01 -2.78896689e-01 -8.75287652e-02 3.38537931e-01
-2.52687156e-01 3.51601511e-01 1.13712087e-01 3.05220872e-01
-1.00651407e+00 8.40758622e-01 1.19180346e+00 -1.20419212e-01
-5.31032830e-02 1.99671358e-01 -9.58642006e-01 6.59686565e-01
9.80862677e-01 -2.81820804e-01 -5.35468578e-01 -5.79657614e-01
8.09919655e-01 2.31202796e-01 1.32524624e-01 -7.35953033e-01
-1.11401360e-02 -5.77371046e-02 1.49321079e-01 -6.79723203e-01
4.68527585e-01 -1.51110142e-01 -4.36963998e-02 -2.06822634e-01
-8.54754567e-01 1.70327336e-01 -2.05413982e-01 3.35689157e-01
-2.81159222e-01 -2.15458795e-01 4.10640001e-01 -2.53448030e-03
-5.53235292e-01 3.84317100e-01 -3.57386380e-01 3.64606678e-01
9.78423715e-01 -1.53752953e-01 -5.10713756e-01 -8.99119735e-01
-6.13723576e-01 3.58311474e-01 4.84491408e-01 5.87645531e-01
7.31700718e-01 -1.61010504e+00 -1.27914548e+00 -1.67228535e-01
6.18364871e-01 4.29732688e-02 2.85531074e-01 7.55349338e-01
-5.02705984e-02 1.76403135e-01 1.75607160e-01 -1.82806596e-01
-1.20330417e+00 3.02750051e-01 -2.32173339e-01 -4.76504207e-01
-3.13387215e-01 1.34060085e+00 2.99519420e-01 -3.86912078e-01
3.78888488e-01 -2.52573103e-01 -4.51045632e-01 3.07255268e-01
6.52832210e-01 3.53823096e-01 -2.72147089e-01 -8.71672153e-01
-1.59097854e-02 3.68219078e-01 -7.67806321e-02 -4.30170417e-01
1.30811465e+00 -3.23556006e-01 7.95897320e-02 7.34343112e-01
8.78049910e-01 6.18131831e-02 -1.31428885e+00 -1.82439938e-01
-1.07024632e-01 -2.59834379e-01 -2.70935088e-01 -9.68419433e-01
-8.43065679e-01 6.63265705e-01 -1.81778416e-01 1.72790229e-01
1.15951383e+00 4.79403585e-01 1.09422588e+00 2.86161117e-02
2.56286055e-01 -1.27597713e+00 3.19817305e-01 7.13506103e-01
1.46503270e+00 -1.18229663e+00 -1.87778309e-01 -2.38733262e-01
-1.46605384e+00 8.07194531e-01 5.25966942e-01 2.71040127e-02
4.88110751e-01 9.67571661e-02 -4.56530713e-02 -1.00454524e-01
-1.11307240e+00 -8.95695612e-02 4.59107369e-01 4.08130795e-01
7.52375603e-01 4.73581225e-01 -3.45109046e-01 1.38925004e+00
-9.54146743e-01 -1.85853034e-01 4.87794548e-01 3.13788444e-01
-3.65784466e-01 -8.93827736e-01 -4.14191186e-02 7.30890512e-01
-1.60621703e-01 -3.95931125e-01 -4.60249156e-01 1.17852919e-01
1.39357567e-01 1.09807014e+00 -3.22649360e-01 -7.46638060e-01
3.64577621e-01 -1.14516191e-01 -5.25954319e-03 -6.69134378e-01
-6.54219985e-01 -1.27291575e-01 5.62661648e-01 -3.81139666e-01
1.99529119e-02 -9.50617611e-01 -1.03360128e+00 -3.55874032e-01
-1.55885860e-01 1.29748523e-01 7.78535604e-02 7.87986040e-01
-1.15857683e-02 3.69668543e-01 6.03478432e-01 -7.41581917e-01
-2.91719139e-01 -1.15754008e+00 -6.97215676e-01 5.26661992e-01
-5.80958053e-02 -2.06732213e-01 -2.53543824e-01 2.50500739e-01]
|
[11.627408981323242, 8.832967758178711]
|
f88eaf00-210b-4c11-8105-166f52b90ca1
|
dialogvcs-robust-natural-language
|
2305.14751
| null |
https://arxiv.org/abs/2305.14751v1
|
https://arxiv.org/pdf/2305.14751v1.pdf
|
DialogVCS: Robust Natural Language Understanding in Dialogue System Upgrade
|
In the constant updates of the product dialogue systems, we need to retrain the natural language understanding (NLU) model as new data from the real users would be merged into the existent data accumulated in the last updates. Within the newly added data, new intents would emerge and might have semantic entanglement with the existing intents, e.g. new intents that are semantically too specific or generic are actually subset or superset of some existing intents in the semantic space, thus impairing the robustness of the NLU model. As the first attempt to solve this problem, we setup a new benchmark consisting of 4 Dialogue Version Control dataSets (DialogVCS). We formulate the intent detection with imperfect data in the system update as a multi-label classification task with positive but unlabeled intents, which asks the models to recognize all the proper intents, including the ones with semantic entanglement, in the inference. We also propose comprehensive baseline models and conduct in-depth analyses for the benchmark, showing that the semantically entangled intents can be effectively recognized with an automatic workflow.
|
['Yunbo Cao', 'Baobao Chang', 'Binghuai Lin', 'Gang Yuan', 'Jiaqi Han', 'Haoran Meng', 'Xu Wang', 'Tianyu Liu', 'Xin Zheng', 'Zefan Cai']
|
2023-05-24
| null | null | null | null |
['intent-detection']
|
['natural-language-processing']
|
[ 2.25841030e-01 6.79432511e-01 -8.36926848e-02 -6.26320601e-01
-2.68002898e-01 -1.09812200e+00 5.77354312e-01 1.99460521e-01
-6.91767856e-02 5.94846249e-01 2.71716893e-01 -3.37170511e-01
7.20372796e-02 -5.48645318e-01 -6.40917480e-01 -2.59334326e-01
3.38063091e-01 8.44768345e-01 1.17099568e-01 -7.30396032e-01
9.57616419e-02 -2.33365729e-01 -1.44368041e+00 4.74254519e-01
8.94241393e-01 9.23070490e-01 2.07507849e-01 4.82980460e-01
-4.45750326e-01 6.55437469e-01 -7.24439859e-01 -7.55239010e-01
3.93952459e-01 -1.91144973e-01 -1.25922239e+00 2.13030010e-01
2.70467848e-01 -5.16231120e-01 3.94665189e-02 1.21000051e+00
-5.87157607e-02 1.34250745e-01 4.22980040e-01 -1.49674308e+00
-6.19246781e-01 7.58016527e-01 -3.28184180e-02 -4.43825305e-01
8.51660967e-01 2.47512296e-01 1.50689530e+00 -6.57898247e-01
7.68597901e-01 1.38765836e+00 4.52223271e-01 6.54570401e-01
-1.02873206e+00 -3.30602974e-01 3.49712014e-01 1.55301824e-01
-9.98361826e-01 -4.85657930e-01 6.41807914e-01 -4.01769221e-01
9.83842015e-01 5.11093259e-01 2.83433408e-01 1.27860188e+00
2.17034272e-03 8.62965465e-01 8.64736319e-01 -3.76651198e-01
4.19391096e-01 5.90628862e-01 8.30558896e-01 6.91449583e-01
1.38228163e-01 -2.81889811e-02 -3.90435278e-01 -2.31722653e-01
8.36393461e-02 -2.02399686e-01 -1.62405416e-01 -3.32727164e-01
-8.32390904e-01 9.36742902e-01 5.15239723e-02 3.54114264e-01
-1.04926944e-01 -5.02957106e-01 4.48319525e-01 4.08614010e-01
3.80049884e-01 9.96956110e-01 -1.06738758e+00 -1.61887631e-01
-3.49540293e-01 1.11380227e-01 1.38419318e+00 1.28702497e+00
1.06969547e+00 -7.11292148e-01 -9.51358080e-02 6.75081551e-01
5.14767170e-01 -3.22579108e-02 6.39530957e-01 -1.02658379e+00
4.20446485e-01 1.24527156e+00 3.80768687e-01 -7.20425129e-01
-5.47218978e-01 -2.06638649e-01 -4.38148141e-01 -2.48539343e-01
4.25370455e-01 -2.16649756e-01 -4.89143819e-01 1.82417107e+00
4.11564946e-01 -6.55117035e-02 4.99466270e-01 6.93066299e-01
9.00271356e-01 5.82822800e-01 -5.36204614e-02 -2.66401380e-01
1.78528726e+00 -1.03742397e+00 -1.15010965e+00 -3.94673109e-01
1.15127718e+00 -7.79566288e-01 1.22722948e+00 2.91777581e-01
-5.39707959e-01 -6.11661434e-01 -1.35399616e+00 -1.85738236e-01
-7.31741130e-01 -8.27550516e-02 8.85436058e-01 7.67895520e-01
-8.17475438e-01 4.32742834e-01 -2.09886119e-01 -4.59665924e-01
-1.01986237e-01 2.81404436e-01 -3.56142581e-01 -2.53788233e-02
-1.71698487e+00 9.91337717e-01 9.51563001e-01 1.96308568e-01
-6.66799486e-01 -5.13772905e-01 -1.01317096e+00 6.38388172e-02
1.20185578e+00 -5.81037819e-01 1.47416925e+00 -8.14146698e-01
-1.52339125e+00 7.27039993e-01 -5.92829548e-02 -2.16539398e-01
3.49315166e-01 -1.76270783e-01 -4.05717611e-01 -4.28865731e-01
1.48825198e-01 5.54453969e-01 4.70637321e-01 -1.43688166e+00
-7.59225011e-01 -5.54924726e-01 7.84044921e-01 4.21536297e-01
-1.86059296e-01 -2.27363408e-01 -3.62007380e-01 -9.91283432e-02
4.30832170e-02 -1.14619100e+00 3.14165279e-02 -4.89471048e-01
-7.83193111e-01 -5.40685236e-01 8.96450639e-01 -4.65648979e-01
1.15048110e+00 -2.04825115e+00 1.21254064e-01 -2.97905982e-01
2.95409381e-01 2.91781843e-01 -1.51995614e-01 3.09454858e-01
-5.90059627e-03 3.56089354e-01 -1.00071050e-01 -5.80421209e-01
2.90730864e-01 5.82906365e-01 -2.60667145e-01 -1.68498769e-01
3.04469764e-02 8.07000935e-01 -1.10687900e+00 -2.83517182e-01
1.45286992e-01 -6.73333228e-01 -4.10776496e-01 5.36979735e-01
-7.94679105e-01 3.75936538e-01 -5.21783829e-01 4.84231532e-01
7.39181578e-01 -3.33889067e-01 2.91180432e-01 -6.24076605e-01
8.37669894e-02 4.35778111e-01 -1.09312344e+00 1.90473950e+00
-5.29576719e-01 4.18157643e-03 -5.32656489e-03 -6.99473619e-01
5.67868471e-01 2.18306735e-01 4.24929589e-01 -4.53531563e-01
-1.46654576e-01 2.68096253e-02 1.41675612e-02 -8.33026886e-01
8.80858839e-01 -3.66074204e-01 -5.70827127e-01 4.93802130e-01
5.58074176e-01 -1.25969574e-01 1.38617709e-01 3.74775261e-01
1.00685573e+00 -9.76824462e-02 5.38942158e-01 -1.65060118e-01
6.71819389e-01 5.29922694e-02 5.03412426e-01 1.00319135e+00
-2.40748525e-01 1.65415928e-01 7.90561616e-01 -3.46349031e-01
-7.21229911e-01 -8.16401422e-01 -1.83418199e-01 1.27435195e+00
5.15298665e-01 -8.02362800e-01 -5.35313725e-01 -1.39679229e+00
-1.86123326e-01 1.19178391e+00 -4.59330857e-01 -2.44184896e-01
-2.02570647e-01 -5.75202703e-01 6.35142446e-01 -1.47955418e-02
7.36521959e-01 -7.48740137e-01 -3.37516040e-01 2.93905318e-01
-6.13937736e-01 -1.35181701e+00 -3.43777448e-01 2.43359730e-01
-3.42228681e-01 -1.26874411e+00 9.05978084e-02 -3.22126359e-01
4.46631998e-01 -1.10247940e-01 9.74255025e-01 -1.12077370e-02
1.18181393e-01 3.07904780e-01 -5.15887380e-01 -2.82508582e-01
-1.02161849e+00 1.28640786e-01 4.56799157e-02 1.52537405e-01
4.57163423e-01 9.86165851e-02 -6.62457049e-02 5.87415457e-01
-9.97639418e-01 2.87937611e-01 1.44041613e-01 9.59078729e-01
-7.31248483e-02 1.76136702e-01 6.20624721e-01 -1.10888338e+00
8.08682680e-01 -6.45829380e-01 -5.25538146e-01 7.10418701e-01
-6.70161963e-01 4.17720854e-01 8.53181899e-01 -2.25807056e-01
-1.56316245e+00 6.64673746e-02 -3.46583307e-01 1.63066119e-01
-3.99600714e-01 7.40612030e-01 -6.82614148e-01 3.85786057e-01
5.65807164e-01 -6.41916841e-02 -8.04757699e-02 -6.49421573e-01
8.66275012e-01 8.79473627e-01 2.95292616e-01 -6.62806451e-01
2.86974698e-01 1.66388050e-01 -3.89982611e-01 -6.70074403e-01
-1.17862678e+00 -7.15618610e-01 -8.18081975e-01 -1.16549537e-01
8.11825395e-01 -5.87544084e-01 -8.73090506e-01 5.39218426e-01
-1.57266295e+00 5.53540885e-02 -4.32504982e-01 1.22864559e-01
-3.37003618e-01 8.22513521e-01 -5.73667526e-01 -7.08313525e-01
2.13344190e-02 -1.49315798e+00 1.13094735e+00 1.87918171e-01
-6.43627346e-01 -1.11635888e+00 -4.97083515e-02 6.35658443e-01
5.14622778e-02 -3.27471405e-01 1.37671876e+00 -1.56695664e+00
-3.25978070e-01 -8.36283118e-02 -6.91139475e-02 5.16851187e-01
5.07594824e-01 -3.50643218e-01 -1.04432833e+00 1.96559578e-02
8.16699564e-01 -4.87355620e-01 4.63279277e-01 -3.64527673e-01
6.60535753e-01 -7.51384795e-01 -2.73440659e-01 -1.82010457e-02
1.04943585e+00 3.65552485e-01 3.19228292e-01 -1.72788054e-01
5.88163614e-01 9.85385537e-01 7.12383270e-01 3.33122373e-01
5.10309458e-01 8.91676307e-01 3.93969536e-01 1.66437268e-01
3.13045353e-01 -2.73438364e-01 3.25518966e-01 6.93053365e-01
4.41784650e-01 -4.94131088e-01 -5.88986933e-01 7.18224123e-02
-2.24102426e+00 -4.95756447e-01 -1.65100515e-01 1.98994172e+00
9.73186731e-01 1.14491880e-01 -3.76729876e-01 -4.36474442e-01
6.61032975e-01 1.92337945e-01 -7.76208043e-01 -4.11540866e-01
-1.86144616e-02 -2.14293540e-01 1.62266195e-01 6.70428753e-01
-1.03205466e+00 1.18867898e+00 6.09627008e+00 5.90427637e-01
-4.83002156e-01 2.08971277e-01 5.49988031e-01 2.51339763e-01
-4.66202468e-01 4.29171205e-01 -9.61335897e-01 4.04031366e-01
8.02492023e-01 -4.81767915e-02 3.13449264e-01 8.76810849e-01
-9.20222849e-02 -4.08254206e-01 -1.62666249e+00 5.59136212e-01
1.89818114e-01 -9.58041370e-01 1.19057797e-01 3.86118442e-02
4.98453647e-01 -1.80915862e-01 -2.42704481e-01 7.65980840e-01
5.62447727e-01 -6.21110320e-01 4.32068527e-01 3.50881845e-01
2.88908124e-01 -9.30985287e-02 1.04459631e+00 6.94948316e-01
-7.75818348e-01 -3.84223834e-02 -2.74924308e-01 -4.57301848e-02
2.87647307e-01 4.59085464e-01 -9.97143745e-01 8.76648664e-01
8.96673277e-02 6.09960139e-01 -7.00239480e-01 2.21666008e-01
-1.23604201e-01 1.09926596e-01 -4.15204257e-01 -8.50397572e-02
2.50528246e-01 -4.41435277e-01 6.84911609e-01 7.62321234e-01
3.52216922e-02 1.30034909e-01 2.44770527e-01 1.15878785e+00
-2.49638140e-01 3.11295502e-02 -7.54856884e-01 -8.66053849e-02
1.78876102e-01 1.37084103e+00 -3.41126293e-01 -7.05133021e-01
-6.20111108e-01 1.24673271e+00 1.63646087e-01 1.60315394e-01
-8.27363849e-01 7.99350217e-02 6.98299527e-01 -5.39787173e-01
-2.56310552e-01 1.11815222e-01 -1.40318066e-01 -1.53806973e+00
1.54185772e-01 -9.61019814e-01 5.38340449e-01 -6.61969304e-01
-1.45127809e+00 3.01804006e-01 4.96401312e-03 -8.64693403e-01
-3.15099984e-01 -6.74990416e-01 -2.85105258e-01 6.24692857e-01
-1.01172733e+00 -9.48207200e-01 -1.58169419e-01 2.84831017e-01
8.19208980e-01 -6.40351549e-02 1.02018523e+00 1.17253132e-01
-4.56099033e-01 3.04336399e-01 -2.61606902e-01 -1.60018370e-01
9.40027773e-01 -1.28015566e+00 2.33788788e-01 5.97877324e-01
5.98545335e-02 8.74683797e-01 7.85923302e-01 -8.63991857e-01
-1.40032876e+00 -7.92673886e-01 9.47548389e-01 -7.23234296e-01
8.96996975e-01 -7.15802312e-01 -1.06831300e+00 9.43011642e-01
3.59406829e-01 -5.65791488e-01 8.26026559e-01 4.52392310e-01
-3.70944828e-01 3.13251346e-01 -1.14843833e+00 6.53637528e-01
1.22455013e+00 -5.23513734e-01 -1.08300769e+00 6.31051302e-01
1.16886008e+00 -4.91300762e-01 -7.96197176e-01 4.95557964e-01
4.24728125e-01 -8.70350361e-01 6.43703520e-01 -1.17138565e+00
2.71958500e-01 -2.85895258e-01 -3.33150506e-01 -1.33887303e+00
1.68381289e-01 -6.37519181e-01 -4.10078317e-02 1.44401121e+00
6.15174472e-01 -6.94247544e-01 4.15032268e-01 1.23202789e+00
-2.03724235e-01 -3.26842934e-01 -1.11006403e+00 -6.35678589e-01
-1.69923738e-01 -4.31646764e-01 7.35261619e-01 9.89979088e-01
6.04254901e-01 1.13288975e+00 -3.51028562e-01 9.60932299e-03
2.14172497e-01 1.30531639e-01 6.68687046e-01 -1.32910132e+00
-4.33504820e-01 -1.70896068e-01 -3.71746123e-02 -1.54283726e+00
4.24425036e-01 -8.94437313e-01 2.73656011e-01 -1.18996310e+00
2.71312058e-01 -5.96834838e-01 1.52054906e-01 7.71542966e-01
-2.52730638e-01 -6.18117094e-01 2.98895359e-01 3.13909233e-01
-8.25659096e-01 7.00334013e-01 9.80771780e-01 -4.06626552e-01
-3.08806956e-01 1.30729035e-01 -9.55328524e-01 8.15270543e-01
4.58489060e-01 -2.77399898e-01 -5.23129463e-01 1.37656942e-01
5.53252339e-01 2.54068762e-01 1.45263553e-01 -4.92915601e-01
1.70827925e-01 4.79505491e-03 -3.10742766e-01 -3.62748384e-01
3.75800848e-01 -1.15054452e+00 1.16090089e-01 2.22507700e-01
-7.00863540e-01 -3.67108971e-01 1.99062396e-02 4.66437846e-01
7.85518438e-04 -9.48948979e-01 4.01246428e-01 -2.86809623e-01
-8.79770339e-01 -1.21147104e-01 -3.90460163e-01 4.98116799e-02
9.74909365e-01 1.11379847e-01 -6.88416064e-01 -2.53974110e-01
-9.49468136e-01 4.40969735e-01 4.34469968e-01 8.55552018e-01
7.46344328e-02 -8.74994814e-01 -2.95367211e-01 7.70355090e-02
5.45956969e-01 1.37682622e-02 3.62884253e-01 3.60854119e-01
8.60309079e-02 5.05847991e-01 6.30583838e-02 -4.02003318e-01
-1.03913927e+00 8.70774984e-01 4.02490824e-01 -7.00901210e-01
-1.69862092e-01 5.36464751e-01 6.11035049e-01 -1.01546347e+00
7.50083625e-02 -5.88663042e-01 -3.13240170e-01 2.43349150e-01
1.54742867e-01 1.26160178e-02 1.75767750e-01 -5.55001855e-01
-1.38585001e-01 1.11936766e-03 -3.77590507e-01 -4.12347689e-02
7.93499589e-01 -4.72097278e-01 -2.71423221e-01 8.23969424e-01
1.30779755e+00 -1.68623641e-01 -8.27397048e-01 -3.67139429e-01
3.08226764e-01 -5.37761375e-02 -4.12664622e-01 -1.09716952e+00
-5.13312101e-01 4.69670802e-01 3.69594127e-01 6.44504070e-01
6.08240843e-01 4.48537529e-01 7.52530217e-01 6.95110321e-01
4.98462617e-01 -1.24043202e+00 1.34805202e-01 8.61932337e-01
8.74137342e-01 -1.46052933e+00 -2.70071954e-01 -7.31773913e-01
-8.69439781e-01 1.13487256e+00 8.52051854e-01 7.47320890e-01
5.13699710e-01 4.45950404e-02 -5.81626147e-02 -4.64751035e-01
-7.70302713e-01 -1.50954038e-01 1.07684068e-01 1.96337476e-01
6.29517362e-02 1.48368731e-01 -3.37087214e-01 9.87474620e-01
-2.47693241e-01 -2.57280797e-01 6.21540844e-01 6.17550492e-01
-2.06464335e-01 -1.29133046e+00 1.27900198e-01 5.08638799e-01
2.35737804e-02 -1.07548118e-01 -6.61550701e-01 5.32502949e-01
3.60828489e-01 1.33695400e+00 -2.38538191e-01 -5.41929543e-01
4.62521344e-01 6.04273200e-01 1.71864688e-01 -8.83717716e-01
-6.72919035e-01 -3.06021631e-01 6.70506001e-01 -5.22416055e-01
-2.82792658e-01 -4.26239133e-01 -1.25395358e+00 1.42229095e-01
-8.44731390e-01 1.14641145e-01 6.75643682e-01 1.48447990e+00
4.35991466e-01 3.58462393e-01 6.62143648e-01 -1.31286487e-01
-8.18261385e-01 -1.12953305e+00 -3.95701796e-01 8.06055784e-01
1.51358157e-01 -6.17987812e-01 -6.03371322e-01 7.93807656e-02]
|
[12.612701416015625, 7.722309112548828]
|
8312986d-090a-466f-b4a0-366e1395aaa2
|
dp-kb-data-programming-with-knowledge-bases-1
|
2203.09598
| null |
https://arxiv.org/abs/2203.09598v1
|
https://arxiv.org/pdf/2203.09598v1.pdf
|
DP-KB: Data Programming with Knowledge Bases Improves Transformer Fine Tuning for Answer Sentence Selection
|
While transformers demonstrate impressive performance on many knowledge intensive (KI) tasks, their ability to serve as implicit knowledge bases (KBs) remains limited, as shown on several slot-filling, question-answering (QA), fact verification, and entity-linking tasks. In this paper, we implement an efficient, data-programming technique that enriches training data with KB-derived context and improves transformer utilization of encoded knowledge when fine-tuning for a particular QA task, namely answer sentence selection (AS2). Our method outperforms state of the art transformer approach on WikiQA and TrecQA, two widely studied AS2 benchmarks, increasing by 2.0% p@1, 1.3% MAP, 1.1% MRR, and 4.4% p@1, 0.9% MAP, 2.4% MRR, respectively. To demonstrate our improvements in an industry setting, we additionally evaluate our approach on a proprietary dataset of Alexa QA pairs, and show increase of 2.3% F1 and 2.0% MAP. We additionally find that these improvements remain even when KB context is omitted at inference time, allowing for the use of our models within existing transformer workflows without additional latency or deployment costs.
|
['Alessandro Moschitti', 'Manish Gupta', 'Thuy Vu', 'Nic Jedema']
|
2022-03-17
|
dp-kb-data-programming-with-knowledge-bases
|
https://openreview.net/forum?id=AN4xPK0F0Fs
|
https://openreview.net/pdf?id=AN4xPK0F0Fs
|
neurips-workshop-dbai-2021-12
|
['slot-filling']
|
['natural-language-processing']
|
[-1.36058526e-02 6.41246617e-01 -9.36550051e-02 -2.84943104e-01
-1.57662272e+00 -8.17299366e-01 5.55802286e-01 5.28130293e-01
-6.09280109e-01 1.27406871e+00 3.32717776e-01 -5.77110529e-01
-2.96625346e-01 -8.62974286e-01 -1.05115569e+00 1.69735644e-02
8.92230719e-02 7.97018170e-01 8.40066969e-01 -6.01858318e-01
2.21983299e-01 -1.69761106e-01 -1.20371294e+00 6.45580292e-01
1.38778818e+00 9.46686149e-01 -2.43263304e-01 5.92009723e-01
-3.45097572e-01 1.26589918e+00 -7.02481389e-01 -9.49706078e-01
-9.97853801e-02 2.98591137e-01 -1.47718358e+00 -8.44184041e-01
7.53738999e-01 -1.82801206e-02 -1.86284482e-01 6.26842797e-01
4.84586149e-01 -6.80890605e-02 1.84101194e-01 -8.75842214e-01
-7.23966718e-01 1.01422000e+00 -1.50592044e-01 2.03450888e-01
5.30941248e-01 1.02856837e-01 1.25753021e+00 -7.19021857e-01
8.34316313e-01 8.81318092e-01 7.93444932e-01 3.83319914e-01
-1.03373575e+00 -4.72271383e-01 -2.72382408e-01 5.25524616e-01
-1.20539951e+00 -6.98529422e-01 1.36852697e-01 -1.23527102e-01
1.62820017e+00 4.18779403e-01 1.87692806e-01 5.16618013e-01
-1.15724713e-01 9.67746735e-01 1.29866266e+00 -5.45300782e-01
1.21975884e-01 4.20439094e-01 5.57157457e-01 7.94376671e-01
-6.09883368e-02 -4.82966036e-01 -7.90388346e-01 -4.06619906e-01
1.11178011e-01 -8.58799875e-01 -3.71059328e-01 1.11283623e-01
-1.07420957e+00 6.03044450e-01 3.23609114e-01 -2.35442966e-02
-4.20814514e-01 2.33349055e-01 4.17883635e-01 4.98101383e-01
2.53372401e-01 8.62685025e-01 -1.02821898e+00 -6.62511766e-01
-8.30033541e-01 5.68933129e-01 1.05405557e+00 1.09964979e+00
8.46937418e-01 -4.54979986e-01 -6.88073695e-01 7.88883388e-01
-1.06483757e-01 7.73661077e-01 1.77053496e-01 -1.11463153e+00
9.41043079e-01 9.53478217e-01 3.14632058e-01 -7.29811251e-01
-2.66351193e-01 -5.73327959e-01 -2.45120510e-01 -5.73257089e-01
4.37340260e-01 1.49981230e-01 -1.09419823e+00 1.70150757e+00
3.71752709e-01 1.14996526e-02 4.16275322e-01 4.40315485e-01
1.08186638e+00 4.52913731e-01 3.62594545e-01 2.14484930e-01
1.66641986e+00 -1.10024118e+00 -7.39754677e-01 -2.87462384e-01
1.10615122e+00 -6.94949627e-01 1.24654973e+00 1.47521332e-01
-1.27040589e+00 -1.17845446e-01 -7.36570477e-01 -4.57435608e-01
-3.46176147e-01 3.59692872e-02 5.87520182e-01 5.82480788e-01
-1.09261656e+00 3.41915518e-01 -5.30175209e-01 -2.71572232e-01
4.09599066e-01 3.93016458e-01 -4.47926641e-01 -1.30271792e-01
-1.59898937e+00 1.19701338e+00 2.78578997e-01 -2.50724375e-01
-6.95690215e-01 -1.44711995e+00 -6.45008922e-01 1.99297801e-01
6.32425547e-01 -8.83188844e-01 1.44168568e+00 -1.48772389e-01
-1.40799701e+00 6.27303660e-01 -3.58050138e-01 -8.89365554e-01
2.46123493e-01 -7.32041836e-01 -5.78202605e-01 1.33060187e-01
3.70451629e-01 6.51771069e-01 1.21505231e-01 -7.58184552e-01
-8.03847134e-01 -1.00208506e-01 5.69206297e-01 2.17811555e-01
-2.27964237e-01 -7.78132901e-02 -7.00783730e-01 -6.92705438e-02
-1.56897455e-01 -7.35372245e-01 -2.06394140e-02 -4.52108532e-01
-3.39249432e-01 -3.72786582e-01 4.50130641e-01 -1.20276940e+00
1.49188209e+00 -1.70971882e+00 -5.04681692e-02 9.08857062e-02
1.08400039e-01 4.97091979e-01 -5.86176291e-02 5.50730884e-01
4.67748225e-01 2.21396491e-01 -4.86588895e-01 -1.05701171e-01
6.64876699e-02 1.29191533e-01 -3.94976526e-01 -2.20800504e-01
5.64148724e-01 1.39393687e+00 -8.00196409e-01 -6.17864311e-01
-2.81482100e-01 9.49944407e-02 -6.02942348e-01 -1.32553279e-01
-4.54204798e-01 8.56398419e-02 -3.83972019e-01 7.94957936e-01
4.54910517e-01 -5.19341886e-01 3.38286102e-01 -4.45507675e-01
1.13985471e-01 1.02369595e+00 -8.48170698e-01 1.93105590e+00
-5.46189547e-01 4.64546412e-01 -2.44884953e-01 -7.00308383e-01
7.59972692e-01 3.54878247e-01 -7.07630590e-02 -1.43273389e+00
-4.84214216e-01 4.61876601e-01 -1.68451890e-01 -4.31593657e-01
9.00908589e-01 1.36413574e-01 -2.14444444e-01 2.02440783e-01
1.79615438e-01 8.29981789e-02 2.42305875e-01 7.20932126e-01
1.64879751e+00 -8.87202844e-02 1.27291679e-01 -4.01587963e-01
6.87323749e-01 6.55036092e-01 4.55415636e-01 8.21194649e-01
1.09526075e-01 1.83026955e-01 5.10159910e-01 -2.24926814e-01
-7.73410201e-01 -9.40185189e-01 -1.94556460e-01 9.21633065e-01
-6.02644347e-02 -6.39321446e-01 -6.96770906e-01 -9.96047854e-01
1.56502157e-01 9.64761496e-01 -4.64765042e-01 -9.81026888e-02
-9.23009098e-01 -5.37122488e-01 1.12742388e+00 5.38870096e-01
7.00256824e-01 -8.18705380e-01 -4.46357816e-01 3.54730397e-01
-6.32381082e-01 -1.34232914e+00 1.19504698e-01 -1.35886014e-01
-8.10600042e-01 -1.10499239e+00 -2.59321719e-01 -3.80984187e-01
2.02312797e-01 -2.61684418e-01 1.84303641e+00 3.11742611e-02
6.78351447e-02 4.50414449e-01 -4.19112891e-01 -1.20027751e-01
-1.41498774e-01 4.99961406e-01 -2.85445184e-01 -5.73786438e-01
3.86772811e-01 -3.51490080e-01 -6.66150212e-01 3.39750677e-01
-5.38585365e-01 1.95596982e-02 5.08542597e-01 9.31502581e-01
5.02407610e-01 -3.07681978e-01 8.79826486e-01 -1.26851881e+00
5.02270103e-01 -4.76139426e-01 -4.93226260e-01 8.51184964e-01
-9.99217749e-01 3.14935774e-01 1.85613304e-01 7.30584934e-02
-1.30870759e+00 -3.21172982e-01 -2.26178795e-01 2.68583387e-01
1.82945639e-01 8.94750535e-01 7.88832977e-02 -1.35169283e-01
9.88432527e-01 1.31585807e-01 -2.51458853e-01 -4.68820840e-01
7.42055357e-01 7.15507865e-01 6.64805591e-01 -7.87205219e-01
6.19451642e-01 2.18188688e-01 -3.96793127e-01 -8.13272968e-02
-1.06747544e+00 -5.33793569e-01 -2.19209448e-01 3.59570593e-01
5.85627735e-01 -1.13263249e+00 -8.87132525e-01 -1.89175531e-02
-9.10521865e-01 -4.27401602e-01 -3.30301315e-01 -4.21265746e-03
-3.11647445e-01 3.29811424e-01 -7.13126421e-01 -5.81219852e-01
-1.10763693e+00 -9.76198494e-01 9.39581037e-01 2.90953308e-01
-2.56700277e-01 -8.76338482e-01 1.53293177e-01 1.22551394e+00
7.15507746e-01 -1.72035679e-01 1.19202387e+00 -6.41785979e-01
-9.07202780e-01 -1.05432235e-01 -3.72440845e-01 1.91092506e-01
-1.68417126e-01 -4.67451900e-01 -9.39147949e-01 -7.11681396e-02
-6.88328564e-01 -7.01359212e-01 6.54279113e-01 -2.56160855e-01
9.01601255e-01 -2.83317208e-01 -2.32861519e-01 1.08743958e-01
1.26595974e+00 -1.88209042e-02 8.99541557e-01 7.14673579e-01
3.86243194e-01 5.26364386e-01 8.96661699e-01 -4.89338860e-02
9.12426054e-01 8.34099054e-01 2.21053213e-01 3.96456331e-01
-3.41081679e-01 -2.88881093e-01 1.55168951e-01 8.84361386e-01
7.44373351e-02 -1.09951347e-01 -1.31109977e+00 1.11227727e+00
-1.97198403e+00 -5.04378021e-01 -3.57881069e-01 2.12636423e+00
1.39405823e+00 2.16271237e-01 -2.10845992e-01 -7.17648789e-02
1.13244109e-01 -3.77245456e-01 -4.16388988e-01 -1.53463334e-01
-2.69614577e-01 9.45714891e-01 6.22964263e-01 5.36380112e-01
-9.21828926e-01 1.30685258e+00 5.67505360e+00 1.00763142e+00
-7.46308923e-01 2.60232002e-01 3.89295250e-01 1.29734129e-02
-5.86670041e-01 3.02812397e-01 -9.45644677e-01 3.11853439e-01
1.29170656e+00 -2.08372667e-01 1.71489790e-01 5.74760497e-01
-4.09483880e-01 -3.20601791e-01 -8.46871853e-01 5.95231354e-01
-2.78586328e-01 -1.95839918e+00 3.57985869e-02 -2.44781762e-01
5.23096025e-01 1.51633233e-01 4.22117226e-02 9.72814500e-01
4.34627891e-01 -1.07320237e+00 4.81146753e-01 5.03241301e-01
7.96456158e-01 -5.84709644e-01 1.12405336e+00 1.75026625e-01
-8.05842340e-01 4.27106842e-02 -2.23825186e-01 4.80924100e-02
4.24786329e-01 7.49763191e-01 -1.27029538e+00 1.07110596e+00
9.23192024e-01 2.07194522e-01 -7.93179095e-01 7.97261894e-01
-3.35078925e-01 1.11984444e+00 -4.50532466e-01 5.58711700e-02
2.83266634e-01 3.96249354e-01 2.94942349e-01 1.26117206e+00
1.81046233e-01 2.90835619e-01 -3.22045416e-01 5.57022095e-01
-4.33795184e-01 1.66040286e-01 6.47033304e-02 7.85781145e-02
7.94056118e-01 1.31533933e+00 4.61393297e-02 -5.51729977e-01
-1.73198760e-01 9.39426839e-01 7.96790004e-01 1.54551253e-01
-8.45578611e-01 -5.85434973e-01 4.28015172e-01 -5.07757440e-02
3.03959191e-01 9.90744308e-02 -1.96512550e-01 -1.02525926e+00
5.14470577e-01 -9.83803689e-01 6.03075087e-01 -7.17337251e-01
-1.03437340e+00 5.45851231e-01 -9.27163884e-02 -5.59904456e-01
-2.82647401e-01 -3.91593069e-01 -1.36808649e-01 1.12228477e+00
-1.85323465e+00 -1.16655934e+00 -1.50149211e-01 5.36273658e-01
7.62944371e-02 -7.13501275e-02 9.64109600e-01 7.59659350e-01
-2.77272671e-01 9.47731614e-01 -1.15437560e-01 9.02362168e-02
8.49450827e-01 -1.51240122e+00 7.21149683e-01 5.31528294e-01
1.19501010e-01 9.42335367e-01 4.02512431e-01 -6.63969755e-01
-1.63945782e+00 -1.00210297e+00 1.43617713e+00 -1.11934710e+00
8.20492625e-01 -1.85830161e-01 -1.14371765e+00 7.21752048e-01
2.23597959e-01 1.79240257e-02 7.22048163e-01 7.29335129e-01
-6.36343122e-01 -2.56968290e-01 -1.28545070e+00 4.25427258e-01
9.33737516e-01 -7.40557075e-01 -7.62125671e-01 3.86211753e-01
1.20055735e+00 -8.36438835e-01 -1.44673383e+00 6.53357148e-01
3.68632376e-01 -5.77846825e-01 1.04563797e+00 -9.30874288e-01
3.44599098e-01 -4.94968891e-01 -2.79532850e-01 -1.02475023e+00
-2.91029483e-01 -3.18864048e-01 -5.05297363e-01 1.50727105e+00
1.09825778e+00 -6.27662063e-01 8.80864322e-01 8.98366570e-01
-3.23626935e-01 -8.43989074e-01 -1.11198497e+00 -5.13124764e-01
2.34816112e-02 -5.37509024e-01 7.28486896e-01 9.92015064e-01
4.16669585e-02 5.70716977e-01 -8.71925503e-02 3.78414333e-01
2.90585101e-01 5.53091690e-02 7.08638191e-01 -8.28953326e-01
-4.11487877e-01 5.83137162e-02 -1.94083437e-01 -1.01670694e+00
-1.48360789e-01 -8.24909031e-01 -2.14189067e-01 -1.60602176e+00
1.67330369e-01 -1.05411780e+00 -3.51288557e-01 8.89291644e-01
-3.93453091e-01 1.51533157e-01 8.00961927e-02 1.78013463e-02
-8.82199943e-01 4.52993542e-01 7.51511335e-01 -2.69068122e-01
4.42142002e-02 -3.64088267e-01 -7.60582328e-01 4.76280786e-02
7.33264983e-01 -4.33329761e-01 -5.65163016e-01 -9.02420998e-01
6.21915579e-01 2.86321677e-02 2.17901096e-01 -8.82638216e-01
4.74339515e-01 1.31641850e-01 -2.06996948e-01 -4.13622916e-01
4.03263330e-01 -3.61912072e-01 8.72772485e-02 2.61708766e-01
-3.56842130e-01 -2.31161229e-02 6.68008447e-01 3.80329847e-01
-3.64145130e-01 -2.31107980e-01 1.00429170e-01 -1.84119970e-01
-9.17153716e-01 -1.14400864e-01 1.20384790e-01 5.75311184e-01
6.10192299e-01 1.59891829e-01 -9.94047284e-01 -2.05439061e-01
-3.61036569e-01 6.92875266e-01 -2.60963775e-02 1.75781056e-01
4.04996127e-01 -9.67141509e-01 -8.48879993e-01 -4.17993903e-01
4.58591849e-01 1.14024326e-01 5.56220591e-01 1.01771259e+00
-5.54132342e-01 8.64823282e-01 7.70158470e-02 -5.32061040e-01
-1.15022242e+00 1.64832361e-02 1.95587113e-01 -8.70485842e-01
-4.07091379e-01 9.43509042e-01 -2.56482005e-01 -8.31671059e-01
-1.08934045e-01 -2.39127383e-01 -8.67173821e-02 -3.07070851e-01
5.25879502e-01 3.72231096e-01 6.65761769e-01 -1.65081024e-03
-6.41912580e-01 1.74800605e-01 -4.94134367e-01 -1.42323390e-01
1.21975732e+00 6.50909469e-02 -3.00787538e-01 3.02879582e-03
6.79595828e-01 2.48621956e-01 -5.61674476e-01 -5.58669806e-01
4.75750983e-01 -1.44430399e-01 -3.87985967e-02 -1.81890738e+00
-7.05106437e-01 6.12005591e-01 4.15859163e-01 -4.88493517e-02
1.11427915e+00 1.70118615e-01 9.61012602e-01 8.04455101e-01
5.82245767e-01 -7.79216111e-01 -2.52729774e-01 7.58591056e-01
7.17670381e-01 -1.05063939e+00 -1.07414745e-01 -6.93865001e-01
-7.07592547e-01 5.24114966e-01 6.96574271e-01 3.93423170e-01
1.24559440e-01 1.29739031e-01 5.83311357e-02 -4.86073136e-01
-1.08201063e+00 -2.46289372e-01 3.61210406e-01 5.04739821e-01
5.16012549e-01 1.34819329e-01 -3.21786582e-01 5.64137697e-01
-4.17311400e-01 1.18021779e-01 2.38777459e-01 1.00922799e+00
-2.76097417e-01 -1.16198373e+00 1.74316674e-01 5.51245570e-01
-5.69777250e-01 -6.42010510e-01 -1.37045547e-01 8.25747907e-01
-2.19458923e-01 8.26930463e-01 -2.00009763e-01 -2.87104964e-01
7.00340509e-01 3.30870092e-01 3.97152126e-01 -5.73065996e-01
-1.02116036e+00 -9.30797517e-01 1.00913012e+00 -5.98725736e-01
-2.89444804e-01 -3.18056613e-01 -1.32726419e+00 -3.16745728e-01
-4.04636979e-01 5.81458569e-01 4.90277797e-01 9.51714635e-01
9.79075253e-01 3.11518461e-01 -2.56147951e-01 6.57244742e-01
-5.73486328e-01 -9.92646575e-01 2.20948439e-02 3.35409909e-01
-9.96065587e-02 -5.79371393e-01 1.09062366e-01 -4.52847034e-02]
|
[10.526408195495605, 8.099804878234863]
|
3db57c38-c19a-47ff-85c1-cb5d67196b58
|
towards-unsupervised-sketch-based-image
|
2105.08237
| null |
https://arxiv.org/abs/2105.08237v4
|
https://arxiv.org/pdf/2105.08237v4.pdf
|
Towards Unsupervised Sketch-based Image Retrieval
|
The practical value of existing supervised sketch-based image retrieval (SBIR) algorithms is largely limited by the requirement for intensive data collection and labeling. In this paper, we present the first attempt at unsupervised SBIR to remove the labeling cost (both category annotations and sketch-photo pairings) that is conventionally needed for training. Existing single-domain unsupervised representation learning methods perform poorly in this application, due to the unique cross-domain (sketch and photo) nature of the problem. We therefore introduce a novel framework that simultaneously performs sketch-photo domain alignment and semantic-aware representation learning. Technically this is underpinned by introducing joint distribution optimal transport (JDOT) to align data from different domains, which we extend with trainable cluster prototypes and feature memory banks to further improve scalability and efficacy. Extensive experiments show that our framework achieves excellent performance in the new unsupervised setting, and performs comparably or better than state-of-the-art in the zero-shot setting.
|
['Yi-Zhe Song', 'Timothy M. Hospedales', 'Yunpeng Li', 'Yongxin Yang', 'Conghui Hu']
|
2021-05-18
| null | null | null | null |
['sketch-based-image-retrieval']
|
['computer-vision']
|
[ 4.56504673e-01 -3.35018784e-01 -4.87326771e-01 -4.02969390e-01
-1.37162256e+00 -7.84433782e-01 9.93086815e-01 5.84902912e-02
-3.03095579e-01 3.19019020e-01 2.84400821e-01 1.04491875e-01
-3.85039240e-01 -3.84419590e-01 -4.19788063e-01 -6.56737804e-01
2.43354410e-01 8.72060418e-01 2.64178216e-01 5.99635392e-02
4.07359689e-01 5.15419066e-01 -1.86314487e+00 4.33277041e-01
4.47961956e-01 1.02108908e+00 3.32681417e-01 5.03189802e-01
-4.50993568e-01 7.21244812e-01 -2.57391930e-01 -2.69527286e-01
4.15941834e-01 -3.67652178e-01 -9.42395449e-01 4.01264966e-01
8.40080976e-01 -5.09942949e-01 -4.85611260e-01 8.18117976e-01
5.02227068e-01 4.45580244e-01 1.06414247e+00 -1.42134845e+00
-1.02241015e+00 1.54005647e-01 -4.21157986e-01 -1.55006960e-01
1.63829595e-01 -3.69907647e-01 1.31571436e+00 -1.08276212e+00
8.68086219e-01 1.04272044e+00 4.83692169e-01 6.22474790e-01
-1.40531576e+00 -4.68354464e-01 1.50138199e-01 2.94814199e-01
-1.74090827e+00 -7.17941999e-01 9.89137173e-01 -4.76599753e-01
8.86880219e-01 -5.71423583e-02 6.16804957e-02 1.13544059e+00
-7.52673030e-01 1.08025753e+00 9.48263288e-01 -7.58790314e-01
3.69254529e-01 1.79569498e-01 1.12528764e-01 3.75040799e-01
1.73220399e-03 -1.35242835e-01 -5.90461075e-01 -2.02143073e-01
9.61861134e-01 1.92175418e-01 2.34451145e-01 -1.18473196e+00
-1.15127826e+00 8.26287508e-01 3.11214089e-01 3.03657889e-01
-1.45848945e-01 3.47892374e-01 3.75100434e-01 4.25770104e-01
4.09142315e-01 3.01345050e-01 -2.68661350e-01 5.71528859e-02
-1.15861166e+00 1.27892256e-01 5.69496572e-01 1.39422143e+00
9.36207831e-01 -2.84153104e-01 -1.44998848e-01 1.38741541e+00
1.23005353e-01 4.04842675e-01 2.84760416e-01 -9.97646093e-01
2.66126752e-01 4.01594520e-01 1.53282553e-01 -8.54673088e-01
9.74000990e-02 1.66636363e-01 -5.81325710e-01 -1.33524284e-01
3.24931383e-01 5.61691999e-01 -1.12764490e+00 1.51109469e+00
1.14830293e-01 2.03664929e-01 3.31679508e-02 8.86412084e-01
8.07020247e-01 4.26245630e-01 3.51767898e-01 5.22175580e-02
1.20207107e+00 -1.06481862e+00 -4.84943479e-01 -1.89384535e-01
5.00070572e-01 -8.00936341e-01 1.08304632e+00 9.06270146e-02
-8.50562990e-01 -5.47361016e-01 -9.60407317e-01 -4.86987293e-01
-6.42766654e-01 3.82515848e-01 5.82986116e-01 4.83797878e-01
-1.03613222e+00 3.90018463e-01 -5.51683664e-01 -8.12952459e-01
7.06988752e-01 3.61578524e-01 -6.14775538e-01 -5.83376050e-01
-6.60459340e-01 7.02838957e-01 1.63769603e-01 -4.55609381e-01
-7.67374456e-01 -7.30815172e-01 -8.05778027e-01 1.13364658e-03
4.39935178e-01 -4.98091638e-01 1.28159225e+00 -1.17629766e+00
-1.27212679e+00 1.14201772e+00 -1.68390796e-01 -1.53209358e-01
3.14336598e-01 -9.59948450e-02 -4.43436801e-02 4.89965975e-01
1.32975191e-01 1.01933706e+00 1.05527794e+00 -1.47537422e+00
-4.66199368e-01 -3.54855895e-01 4.16294709e-02 3.29567105e-01
-7.46577680e-01 2.50563353e-01 -8.44935656e-01 -7.31359959e-01
1.16223715e-01 -9.64495897e-01 -9.41152945e-02 5.24259448e-01
4.44216132e-02 -4.90890652e-01 7.73414016e-01 -4.82494891e-01
8.88179958e-01 -2.17066526e+00 1.96992934e-01 2.01749176e-01
-1.48580760e-01 3.31869990e-01 -5.05857646e-01 8.21950793e-01
-3.39256739e-03 -1.46172479e-01 -3.07071984e-01 -6.81832552e-01
1.82415381e-01 5.27514637e-01 -4.99203354e-01 4.85236406e-01
3.46107960e-01 8.82303059e-01 -1.01326954e+00 -7.57261932e-01
2.82561690e-01 4.63815987e-01 -4.69022095e-01 2.97601491e-01
-2.35172912e-01 3.02852273e-01 -3.30679595e-01 8.04626167e-01
7.14307189e-01 -4.74673748e-01 4.95691717e-01 -1.89726710e-01
1.54590711e-01 2.43803144e-01 -1.16260684e+00 2.28443980e+00
-3.51952702e-01 5.76303482e-01 -1.08269557e-01 -1.45161808e+00
1.02922010e+00 2.39108607e-01 7.36142278e-01 -7.59049535e-01
-3.47673327e-01 1.38643473e-01 -6.44129097e-01 -2.35148668e-01
6.03667796e-01 -2.08679199e-01 -1.80454373e-01 6.62173271e-01
3.91544580e-01 5.94122242e-03 -5.42678945e-02 3.55337352e-01
9.29492891e-01 3.29249173e-01 2.15312839e-01 -2.50490606e-01
3.91968250e-01 1.70686260e-01 1.19430423e-01 7.96929419e-01
-1.18627891e-01 9.98849154e-01 1.28989622e-01 -3.64197642e-01
-1.41891134e+00 -1.20729244e+00 -5.62272444e-02 1.51858437e+00
3.59179050e-01 -2.67788351e-01 -3.86245191e-01 -8.19707155e-01
1.17031023e-01 4.11532342e-01 -4.07275110e-01 6.19176328e-02
-4.26248133e-01 -1.92384675e-01 3.68647218e-01 7.62435853e-01
1.85452402e-01 -1.00302315e+00 -1.09070577e-01 5.25669083e-02
-1.30661502e-01 -1.17445850e+00 -4.08815742e-01 -1.88614745e-02
-7.93549657e-01 -9.49108899e-01 -9.30510521e-01 -1.09102511e+00
7.62369037e-01 8.33100140e-01 1.15391660e+00 7.48700202e-02
-5.66686332e-01 8.69464934e-01 -5.02071977e-01 2.48207781e-03
-6.07273839e-02 2.23661691e-01 -8.85968730e-02 -6.62591606e-02
5.87208509e-01 -5.67467153e-01 -7.07085669e-01 4.32517380e-01
-1.04636502e+00 -1.35161996e-01 6.65264368e-01 9.35298383e-01
7.00671613e-01 -3.29266638e-01 6.88055873e-01 -8.11513484e-01
3.43418121e-01 -4.07485336e-01 -5.51287532e-01 6.07539296e-01
-5.60918093e-01 8.37589577e-02 2.72872448e-01 -5.41144133e-01
-1.05496442e+00 4.70717669e-01 3.19135457e-01 -8.08215499e-01
-3.14471900e-01 2.63470322e-01 -1.04409635e-01 -9.93663892e-02
6.41156673e-01 3.82957548e-01 1.28994554e-01 -7.81699359e-01
7.55575299e-01 1.01989961e+00 5.39914668e-01 -8.53320360e-01
7.85464466e-01 6.65237188e-01 -2.09221736e-01 -8.95876288e-01
-8.38872075e-01 -1.11651933e+00 -1.17433214e+00 2.67368425e-02
6.46399260e-01 -1.14866245e+00 -1.27435207e-01 7.11098835e-02
-1.09205711e+00 -2.49750599e-01 -3.63034099e-01 1.20436884e-01
-9.20765758e-01 6.26557887e-01 -3.71410757e-01 -8.91576409e-01
-2.48659790e-01 -7.68263519e-01 1.45657003e+00 -1.47530615e-01
2.52699945e-02 -8.54064107e-01 1.73326060e-01 3.69044960e-01
2.85038233e-01 -2.26682112e-01 8.78503084e-01 -7.59248257e-01
-7.90057719e-01 -2.49857724e-01 -8.26762199e-01 4.76089358e-01
1.53892878e-02 -3.21581364e-01 -1.12023461e+00 -3.24262828e-01
-6.42584741e-01 -7.24022865e-01 8.80486012e-01 3.38366139e-03
1.18036354e+00 -1.15165807e-01 -3.64768893e-01 3.58444512e-01
1.69500732e+00 -2.18801036e-01 7.97324061e-01 2.29026750e-01
7.06288695e-01 7.66924560e-01 7.81210601e-01 2.86381245e-01
3.03548008e-01 1.01049125e+00 -4.26865108e-02 -4.98014390e-02
-6.17688715e-01 -4.62789744e-01 1.17518991e-01 6.67377412e-01
1.28577828e-01 -8.28333646e-02 -8.11004162e-01 1.08522046e+00
-2.18964982e+00 -1.05684626e+00 2.47496992e-01 2.33271503e+00
7.03070879e-01 -5.91592729e-01 3.36652398e-01 -6.80431724e-02
6.18097365e-01 6.73383623e-02 -2.84477949e-01 -8.90555903e-02
2.74157058e-02 4.39757288e-01 4.47989881e-01 1.66575536e-01
-1.43746018e+00 1.18559718e+00 6.97115326e+00 1.06594634e+00
-6.39555216e-01 2.99376577e-01 2.35536456e-01 4.75582629e-02
-1.27418265e-01 3.62410694e-02 -5.68718076e-01 1.09690197e-01
6.65858984e-01 1.40012622e-01 6.09021068e-01 1.01748061e+00
-4.65582281e-01 1.76159531e-01 -1.41333318e+00 1.37427306e+00
2.86996275e-01 -1.34373069e+00 9.00267214e-02 -2.06040777e-02
9.17922974e-01 4.59718220e-02 7.47617111e-02 2.06773683e-01
4.16607529e-01 -8.37203145e-01 4.49180514e-01 3.90500218e-01
1.19460452e+00 -5.21435738e-01 3.18564594e-01 5.29483147e-02
-1.24710274e+00 -1.34881780e-01 -7.67361641e-01 2.23763362e-01
-1.77315697e-01 1.56137973e-01 -7.60830641e-01 5.07375836e-01
4.63441104e-01 9.16418552e-01 -4.70081329e-01 8.97443950e-01
-8.65633115e-02 1.97475389e-01 -2.89570153e-01 2.12727249e-01
1.63488030e-01 -7.60261565e-02 1.92008421e-01 1.40682364e+00
2.53332108e-01 1.87930167e-02 3.96864742e-01 6.13967359e-01
-2.71895945e-01 5.59393354e-02 -9.19855058e-01 -2.23215654e-01
7.03326643e-01 1.27264309e+00 -7.02292442e-01 -5.16851723e-01
-5.87067485e-01 1.26660955e+00 5.63057244e-01 3.93614799e-01
-3.74032766e-01 -1.70998260e-01 7.59132385e-01 3.57151553e-02
7.91645586e-01 -4.07395035e-01 -1.90508470e-01 -1.21667457e+00
1.13029834e-02 -5.64703822e-01 6.26669765e-01 -7.24276006e-01
-1.90760434e+00 1.60818353e-01 1.63265467e-01 -1.45658100e+00
-3.15160513e-01 -6.72944188e-01 -1.05320930e-01 5.62308908e-01
-1.79869545e+00 -1.66775835e+00 -1.87055647e-01 7.47593403e-01
5.99302769e-01 -1.74897373e-01 1.10719466e+00 5.07051289e-01
-1.11585192e-01 8.56220663e-01 4.98898655e-01 2.81167567e-01
1.07409573e+00 -1.12733018e+00 2.55346358e-01 5.60029805e-01
6.01433754e-01 5.67948461e-01 2.73617595e-01 -3.76750082e-01
-1.63212800e+00 -1.03819668e+00 9.11465704e-01 -7.30073273e-01
7.55855620e-01 -6.47349715e-01 -7.07583368e-01 5.39380789e-01
-1.34991482e-01 3.56112778e-01 8.76769960e-01 2.49528646e-01
-1.06762290e+00 -1.13882013e-01 -1.09940398e+00 3.71819466e-01
1.38872981e+00 -1.15028691e+00 -5.79898775e-01 5.29970765e-01
4.18071359e-01 2.27594361e-01 -1.01207530e+00 8.09065253e-02
7.79871166e-01 -5.66481769e-01 1.45466161e+00 -7.66885459e-01
5.13677120e-01 -2.57516503e-01 -6.09127223e-01 -7.73894966e-01
-2.62702167e-01 -3.00604790e-01 -4.82043438e-02 1.41690302e+00
2.59197969e-02 -1.32178798e-01 8.78230453e-01 6.06159508e-01
1.65645853e-01 -1.25972196e-01 -8.04514110e-01 -1.02646852e+00
4.13163938e-02 -3.24776441e-01 3.73516560e-01 1.09349465e+00
-3.51458043e-02 3.88401508e-01 -4.81838226e-01 6.29181936e-02
8.03101659e-01 4.58472997e-01 9.36315715e-01 -1.29902923e+00
-1.94666356e-01 -3.60056043e-01 -5.80235004e-01 -1.20787442e+00
2.29709730e-01 -9.71683741e-01 2.75127262e-01 -1.60846853e+00
4.77632165e-01 -8.01841438e-01 -5.88188708e-01 6.09330714e-01
5.10792993e-02 6.30156815e-01 2.91082084e-01 7.54781306e-01
-1.09735715e+00 4.74266171e-01 6.59319878e-01 -1.96701229e-01
-4.18136120e-02 -5.12877703e-01 -5.41154683e-01 2.73302495e-01
4.02430862e-01 -4.00306076e-01 -6.81123614e-01 -6.96650863e-01
-2.55903721e-01 -2.87604094e-01 4.78210241e-01 -7.62456059e-01
3.26556057e-01 -1.31492645e-01 2.38139406e-01 -5.11159837e-01
5.08794069e-01 -9.82172132e-01 -8.78185853e-02 -2.80261338e-01
-6.95771515e-01 -4.31363165e-01 -9.43618789e-02 8.99539113e-01
-1.65838048e-01 -1.94765911e-01 7.28390157e-01 -2.56303046e-02
-1.07556355e+00 3.88977826e-01 -2.47512199e-02 -1.00653172e-01
9.13922668e-01 -1.95197985e-01 -2.38358870e-01 -3.66909564e-01
-4.77223992e-01 -1.46751210e-01 8.32136035e-01 5.41418672e-01
5.95449030e-01 -1.50044930e+00 -4.26232159e-01 1.52542859e-01
8.06684792e-01 -1.50289863e-01 2.15538070e-01 4.93554860e-01
-1.55694276e-01 3.73124212e-01 -1.99888244e-01 -6.57214344e-01
-1.20420694e+00 8.71645391e-01 -1.87434748e-01 -2.49501750e-01
-6.96358025e-01 6.52028680e-01 2.14764163e-01 -6.06146514e-01
3.37331980e-01 3.48423272e-01 7.77054578e-02 1.30308896e-01
5.81063688e-01 2.38113195e-01 1.29460514e-01 -6.71167731e-01
-4.13708776e-01 6.51952386e-01 -2.55199671e-01 -2.30372682e-01
1.40971875e+00 -1.09237649e-01 3.28589045e-02 2.67561227e-01
1.48935401e+00 -5.13333499e-01 -1.28843296e+00 -5.98610461e-01
3.98013294e-01 -6.53188288e-01 4.41425331e-02 -6.53090000e-01
-6.62553132e-01 9.40298676e-01 5.87273180e-01 -3.33740674e-02
9.80063200e-01 3.97974223e-01 7.56912947e-01 6.29411519e-01
4.24082577e-01 -1.26133537e+00 3.04781884e-01 2.78668821e-01
6.23317480e-01 -1.27357090e+00 1.77036643e-01 -3.75328869e-01
-6.21629000e-01 9.29175556e-01 1.63167715e-01 -4.93747860e-01
5.67626655e-01 -2.12824762e-01 3.77753936e-02 -1.86747476e-01
-4.94024128e-01 -5.25411427e-01 5.28443098e-01 1.02251899e+00
3.01966697e-01 -3.98126096e-02 -5.51898479e-02 2.82793283e-01
5.03442228e-01 7.53976330e-02 -3.36791016e-02 1.25836504e+00
-4.21645015e-01 -1.52177691e+00 -8.48179758e-02 2.39376381e-01
-5.14373928e-02 -6.12210892e-02 -6.34558737e-01 7.23692000e-01
-2.94466883e-01 8.23546290e-01 4.23986197e-01 -6.51730895e-02
5.24060987e-02 3.57248724e-01 6.92309260e-01 -7.05636919e-01
-1.38377473e-01 1.40177131e-01 3.85473110e-02 -5.57553411e-01
-9.01204348e-01 -5.98378420e-01 -8.56572986e-01 2.76559256e-02
-1.28165096e-01 -1.57128960e-01 8.07357788e-01 8.75299394e-01
6.08176947e-01 -1.69819713e-01 6.35017574e-01 -1.00825107e+00
-6.57430589e-01 -8.11691582e-01 -5.29621959e-01 7.56761491e-01
1.49813071e-01 -9.48476970e-01 -5.87396659e-02 3.51416826e-01]
|
[11.587098121643066, 0.7107003927230835]
|
7447b6a9-0951-4fd5-a51b-19bdb5edef76
|
a-transformer-framework-for-data-fusion-and
|
2211.10506
| null |
https://arxiv.org/abs/2211.10506v1
|
https://arxiv.org/pdf/2211.10506v1.pdf
|
A Transformer Framework for Data Fusion and Multi-Task Learning in Smart Cities
|
Rapid global urbanization is a double-edged sword, heralding promises of economical prosperity and public health while also posing unique environmental and humanitarian challenges. Smart and connected communities (S&CCs) apply data-centric solutions to these problems by integrating artificial intelligence (AI) and the Internet of Things (IoT). This coupling of intelligent technologies also poses interesting system design challenges regarding heterogeneous data fusion and task diversity. Transformers are of particular interest to address these problems, given their success across diverse fields of natural language processing (NLP), computer vision, time-series regression, and multi-modal data fusion. This begs the question whether Transformers can be further diversified to leverage fusions of IoT data sources for heterogeneous multi-task learning in S&CC trade spaces. In this paper, a Transformer-based AI system for emerging smart cities is proposed. Designed using a pure encoder backbone, and further customized through interchangeable input embedding and output task heads, the system supports virtually any input data and output task types present S&CCs. This generalizability is demonstrated through learning diverse task sets representative of S&CC environments, including multivariate time-series regression, visual plant disease classification, and image-time-series fusion tasks using a combination of Beijing PM2.5 and Plant Village datasets. Simulation results show that the proposed Transformer-based system can handle various input data types via custom sequence embedding techniques, and are naturally suited to learning a diverse set of tasks. The results also show that multi-task learners increase both memory and computational efficiency while maintaining comparable performance to both single-task variants, and non-Transformer baselines.
|
['Dwi Novitasari', 'Mochamad Donny Koerniawan', 'Rachmawan Budiarto', 'Wangda Zuo', 'Walid Saad', 'Alexander C. DeRieux']
|
2022-11-18
| null | null | null | null |
['time-series-regression']
|
['time-series']
|
[ 1.83621854e-01 -3.73693943e-01 -2.03809798e-01 -8.48701075e-02
-7.37209499e-01 -4.30733144e-01 8.44760537e-01 7.42526576e-02
-2.81440139e-01 6.82672441e-01 3.22677284e-01 -4.99986112e-01
-3.59102339e-01 -9.49721456e-01 -3.17970276e-01 -8.87129247e-01
-1.95857197e-01 4.41210270e-01 -2.53090769e-01 -4.46190119e-01
-3.23327541e-01 2.93963194e-01 -1.62948740e+00 3.19498003e-01
1.31039643e+00 1.07360840e+00 5.10783553e-01 4.88757372e-01
-1.16812631e-01 6.78156793e-01 -3.40356737e-01 -3.99101317e-01
3.66198123e-01 3.38214248e-01 -4.65591878e-01 -1.63095370e-01
1.00799993e-01 1.07254945e-01 -2.14380592e-01 6.74003065e-01
8.99549484e-01 -1.60595596e-01 5.87647557e-01 -1.64166152e+00
-9.96510029e-01 5.25778055e-01 -3.63278955e-01 1.88944921e-01
6.83519691e-02 3.40498656e-01 9.41489100e-01 -7.74007082e-01
5.41658327e-02 1.27421343e+00 7.78903782e-01 1.09424785e-01
-1.17287052e+00 -1.03189397e+00 2.45971382e-01 3.77121210e-01
-1.36065578e+00 -3.53623748e-01 7.47623861e-01 -6.64058864e-01
1.27380681e+00 2.19843194e-01 7.24980235e-01 1.31436038e+00
3.85656297e-01 8.92757475e-01 9.66143787e-01 1.17390387e-01
2.18542844e-01 -9.34393983e-03 -2.36339733e-01 4.23928171e-01
2.61858135e-01 2.24735588e-02 -3.74391079e-01 1.19596809e-01
8.52630958e-02 4.86041099e-01 7.49625042e-02 -5.95273636e-02
-1.63788807e+00 7.68270969e-01 7.16976881e-01 4.73887146e-01
-5.32858729e-01 1.96935326e-01 6.06229544e-01 4.43501264e-01
7.16136754e-01 1.50396317e-01 -7.40496397e-01 7.26260897e-03
-8.04288566e-01 -2.21905205e-03 5.04451096e-01 8.00944865e-01
6.34699464e-01 5.80463231e-01 -1.30272254e-01 6.34017646e-01
2.28672326e-01 1.31311679e+00 4.87638950e-01 -6.55252814e-01
9.54915106e-01 6.90649986e-01 -2.33950958e-01 -1.06687891e+00
-7.33614802e-01 -3.76713693e-01 -1.39859211e+00 -1.39408574e-01
-1.70407519e-02 -2.40752414e-01 -8.52809727e-01 1.80236709e+00
3.77133936e-01 1.11310102e-01 2.53628522e-01 4.61423367e-01
8.97142172e-01 7.82792330e-01 4.20935273e-01 -1.86732616e-02
1.53615212e+00 -5.22015333e-01 -6.47115052e-01 -3.10451031e-01
6.54276431e-01 -4.14740384e-01 9.61624622e-01 1.72721162e-01
-5.80010891e-01 -6.15204751e-01 -9.15346265e-01 -6.43729568e-02
-9.98627603e-01 2.71083247e-02 5.57800233e-01 5.25689602e-01
-9.87415612e-01 2.41222262e-01 -7.68624246e-01 -6.64353848e-01
5.87914586e-01 4.87584144e-01 -2.14272544e-01 -5.58773875e-02
-1.33012176e+00 9.45582747e-01 3.61190975e-01 -1.37400758e-02
-6.09596491e-01 -9.70756829e-01 -8.42228532e-01 1.10327296e-01
4.68050269e-03 -1.04377532e+00 8.09617937e-01 -6.21767998e-01
-1.08327079e+00 3.93885493e-01 7.71559328e-02 -5.48490465e-01
2.11956829e-01 -6.36152998e-02 -8.91677141e-01 -2.14941055e-01
4.16743517e-01 8.40575159e-01 1.02435923e+00 -8.36626232e-01
-8.32115471e-01 -4.61546570e-01 -2.81384766e-01 1.61163017e-01
-9.95666325e-01 -5.74795082e-02 2.95897335e-01 -7.00464904e-01
-3.19931120e-01 -9.16224658e-01 -2.64875889e-01 -1.54862761e-01
-1.19853862e-01 -3.77202600e-01 1.53171229e+00 -7.44357884e-01
1.14005625e+00 -2.18867874e+00 -7.04980716e-02 -2.01994613e-01
4.18753803e-01 3.91438782e-01 -5.01865268e-01 6.10495985e-01
1.24237768e-01 2.67031670e-01 -1.91118360e-01 -2.35253453e-01
2.21421883e-01 3.44163239e-01 -4.11012948e-01 2.87976861e-01
3.60489517e-01 1.34178042e+00 -1.00348854e+00 -3.20771992e-01
5.76337337e-01 6.96174145e-01 -1.16725177e-01 -1.88566670e-01
-2.42253512e-01 6.35784566e-01 -7.54047453e-01 7.74926841e-01
4.14695889e-01 -5.06815910e-01 1.19188674e-01 -5.11974931e-01
-2.84016371e-01 1.38972759e-01 -9.76217330e-01 1.49233150e+00
-8.57415557e-01 6.61658287e-01 9.11121629e-03 -1.26581299e+00
8.41793597e-01 5.58121443e-01 1.00428855e+00 -1.22282064e+00
-5.29768765e-02 2.47989133e-01 -1.02615125e-01 -7.43931293e-01
3.25544119e-01 -5.04970104e-02 -4.15308058e-01 3.03812534e-01
-1.02166548e-01 -3.60049784e-01 3.40323374e-02 -3.28174010e-02
1.15769124e+00 -1.46706954e-01 3.01705241e-01 -2.65712410e-01
4.79576528e-01 -8.79832134e-02 5.86031437e-01 2.55054593e-01
-2.46532232e-01 -3.03009283e-02 -2.34541133e-01 -8.03408146e-01
-1.18183005e+00 -1.11497080e+00 -7.83648118e-02 1.18217182e+00
-1.62855774e-01 -3.13828111e-01 -4.23762053e-02 -3.85880053e-01
2.92781115e-01 5.02107739e-01 -2.13120267e-01 -2.23184843e-02
-4.82097656e-01 -8.80841672e-01 5.32083273e-01 4.85723674e-01
7.58564293e-01 -8.71766627e-01 -7.63797045e-01 2.66518921e-01
-5.29921830e-01 -1.45569980e+00 -1.78009689e-01 2.97652274e-01
-7.82070696e-01 -7.99646258e-01 -4.89564002e-01 -7.05939472e-01
1.59745365e-01 6.15478575e-01 1.09651589e+00 -5.62245607e-01
-1.16931289e-01 5.64253509e-01 -2.41238669e-01 -7.07593918e-01
-2.32725218e-01 3.28016996e-01 4.35936421e-01 8.50285888e-02
3.69722068e-01 -8.60255599e-01 -5.08385599e-01 2.77942210e-01
-9.04357433e-01 1.56774953e-01 5.48213363e-01 6.78397059e-01
2.63908714e-01 4.03948843e-01 1.06878114e+00 -1.08344391e-01
5.00982404e-01 -8.45973194e-01 -4.15731877e-01 3.71336430e-01
-6.14143968e-01 -1.75925288e-02 9.86905932e-01 -5.08449197e-01
-9.53903854e-01 9.55032781e-02 1.33365884e-01 -3.91966403e-01
-4.61763050e-03 6.10384524e-01 -3.51667672e-01 1.91398159e-01
5.38117170e-01 2.90616423e-01 -1.15586653e-01 -2.82498032e-01
6.05421126e-01 1.10163236e+00 3.37012380e-01 -5.33095837e-01
1.07222629e+00 6.58565700e-01 -1.38619468e-01 -1.13004506e+00
-6.16562903e-01 -4.49074835e-01 -3.79586697e-01 -1.80802956e-01
1.07142544e+00 -1.49561393e+00 -7.57857919e-01 4.46502268e-01
-1.07622397e+00 -1.95635036e-01 -3.65342885e-01 4.95078772e-01
-3.07808936e-01 5.37114814e-02 -1.09362662e-01 -6.64706409e-01
-6.83895350e-01 -1.18257642e+00 1.35611093e+00 -2.64737364e-02
5.54179307e-03 -1.03469765e+00 -4.15935330e-02 6.49300933e-01
7.61730731e-01 3.68164778e-01 9.86664534e-01 -5.31808972e-01
-4.97280091e-01 2.24884935e-02 -3.03467065e-01 3.12709481e-01
5.22529066e-01 -4.03316021e-01 -1.08675766e+00 -4.96452838e-01
-1.83314800e-01 -2.93271333e-01 7.56559432e-01 2.30820015e-01
1.09587276e+00 -5.50257623e-01 -5.14276505e-01 6.30984187e-01
1.24086952e+00 1.41318321e-01 3.29335213e-01 2.04049662e-01
9.47417498e-01 4.04512405e-01 2.34684557e-01 5.43169141e-01
1.06041467e+00 6.34331167e-01 5.24784505e-01 -1.94182441e-01
-1.39239719e-02 -1.58007815e-02 6.70097053e-01 1.21439075e+00
7.86319152e-02 -3.34550947e-01 -1.24231064e+00 9.12541389e-01
-1.95845366e+00 -1.34613287e+00 -1.13092914e-01 1.97941613e+00
5.96006751e-01 -3.69968593e-01 1.39002502e-01 3.55381787e-01
4.35994089e-01 3.74740422e-01 -8.33887637e-01 -1.75673977e-01
-3.69886875e-01 2.54623801e-01 6.50890946e-01 -8.46610069e-02
-1.31570590e+00 7.30474412e-01 5.37226391e+00 8.81961763e-01
-1.54713011e+00 5.39042354e-01 4.01726782e-01 -6.73732162e-02
-4.41916794e-01 -4.21977073e-01 -5.18475294e-01 5.82297146e-01
1.22010505e+00 -3.39962244e-01 5.58848262e-01 5.19288361e-01
4.58909094e-01 5.79041719e-01 -9.13262069e-01 1.21811318e+00
-3.19035679e-01 -1.31430292e+00 -4.38107885e-02 1.25454515e-01
8.18344235e-01 8.44891787e-01 4.51824218e-01 4.77872193e-01
6.26859963e-01 -1.01838064e+00 6.60709202e-01 1.66189522e-01
9.80948985e-01 -3.31053078e-01 3.99532467e-01 3.60117942e-01
-1.92132604e+00 -8.75091374e-01 1.31928042e-01 -9.49361250e-02
1.73207477e-01 7.32444108e-01 -8.03610325e-01 8.51607263e-01
1.07149076e+00 1.13868892e+00 -4.96005684e-01 6.45275712e-01
2.14794919e-01 4.84563410e-01 -4.82553542e-01 1.21675134e-01
1.21404000e-01 -9.66462493e-02 4.84113604e-01 1.20724928e+00
6.39920056e-01 -6.17985465e-02 4.03872997e-01 5.62619448e-01
6.02477714e-02 -5.35330735e-02 -1.15571475e+00 -2.95434237e-01
7.08405018e-01 1.32256591e+00 -2.67349005e-01 -3.23994011e-01
-6.85141087e-01 6.02395892e-01 1.04097135e-01 4.81380016e-01
-9.49576676e-01 -4.94146682e-02 1.08590424e+00 -3.67869101e-02
3.84519517e-01 -4.37141687e-01 -5.16874492e-01 -1.34391344e+00
-8.06645155e-02 -1.03488529e+00 3.97781461e-01 -6.01254702e-01
-1.55030799e+00 3.68014425e-01 2.78526056e-03 -1.51657939e+00
1.89527776e-02 -4.73536730e-01 -5.73581278e-01 5.26303768e-01
-1.77235198e+00 -1.77680469e+00 -3.21714014e-01 8.63341749e-01
5.74451566e-01 -3.33023250e-01 6.87166750e-01 7.23987758e-01
-6.12658501e-01 2.64608592e-01 2.97632158e-01 -1.52688473e-01
3.85426491e-01 -9.92025495e-01 4.82138455e-01 8.27514946e-01
8.52210745e-02 1.09467030e-01 1.77376285e-01 -5.57218015e-01
-1.78148425e+00 -1.80215144e+00 1.09502912e+00 -3.47511530e-01
8.52119505e-01 -4.84708458e-01 -4.32273716e-01 5.58960259e-01
3.32913250e-01 -9.09854770e-02 5.99387228e-01 -6.25539348e-02
-5.75986028e-01 -7.56853223e-01 -1.21920168e+00 6.22519135e-01
1.01822901e+00 -7.96836972e-01 -2.12120280e-01 5.77915192e-01
1.09213829e+00 2.92217165e-01 -1.17815995e+00 7.09299505e-01
3.85811090e-01 -3.61794055e-01 1.40072870e+00 -4.00039673e-01
1.81814402e-01 -4.60107207e-01 -5.83080113e-01 -1.40863252e+00
-5.13896525e-01 -6.91799164e-01 -1.70228645e-01 1.14278400e+00
2.80372709e-01 -8.83979678e-01 2.82384783e-01 2.92090565e-01
-3.13965678e-01 -4.28765625e-01 -1.22883677e+00 -8.02382708e-01
2.23038375e-01 -7.39603996e-01 9.71681595e-01 1.08902907e+00
-8.80829897e-03 5.35672963e-01 -4.63738859e-01 2.32280225e-01
6.31155670e-01 -2.25111395e-02 7.37889528e-01 -1.49807680e+00
2.22811356e-01 -5.62573373e-01 -4.44873571e-01 -5.40771306e-01
9.71319154e-02 -1.15966237e+00 -4.02756572e-01 -1.63255596e+00
-2.16296583e-01 -6.59834445e-01 -3.69044542e-01 7.84016430e-01
1.21398084e-01 1.82909593e-01 4.10671860e-01 1.29445314e-01
-4.96112615e-01 9.17153180e-01 1.09300315e+00 -8.52042556e-01
-1.57504559e-01 -1.21312283e-01 -6.90365076e-01 3.97158742e-01
9.45643365e-01 -1.79393589e-01 -6.90412760e-01 -7.94221163e-01
4.38309550e-01 -4.46471535e-02 5.17500997e-01 -1.30378366e+00
2.47122720e-01 -3.25331777e-01 2.27726102e-01 -5.26164532e-01
4.32085186e-01 -1.28601873e+00 5.19010603e-01 6.18910313e-01
1.10583760e-01 4.35192436e-01 2.02659294e-01 5.32297850e-01
5.53056523e-02 8.10151637e-01 5.39666235e-01 1.22407287e-01
-7.81680167e-01 5.84789455e-01 -4.35049504e-01 -3.75183485e-02
1.23619926e+00 -2.32309073e-01 -6.53331935e-01 -1.84960395e-01
-4.54996437e-01 5.97803593e-01 2.50754841e-02 8.55220318e-01
5.51968038e-01 -1.54875576e+00 -9.16573644e-01 3.11058462e-01
2.80509800e-01 -1.11343838e-01 1.91763550e-01 1.01009810e+00
1.72827438e-01 4.75557715e-01 -9.56882760e-02 -8.12137783e-01
-8.79958212e-01 6.93241119e-01 1.11109488e-01 -4.18350279e-01
-6.66690886e-01 1.48670271e-01 1.35018617e-01 -8.82065177e-01
5.47335967e-02 -6.23181164e-01 -1.04963280e-01 3.60901386e-01
2.68384427e-01 5.04439175e-01 2.37299338e-01 -6.44488692e-01
-4.36755151e-01 5.83441079e-01 4.44478482e-01 2.74111003e-01
1.56669402e+00 -2.97325999e-01 1.14319056e-01 4.21593755e-01
1.21674418e+00 -4.30064261e-01 -6.99007869e-01 -4.64363337e-01
2.08415184e-02 7.18905532e-04 3.45124416e-02 -9.05277252e-01
-1.21166337e+00 9.08494413e-01 8.06583345e-01 4.18211967e-01
1.39600945e+00 -1.25559911e-01 1.20436633e+00 6.34592891e-01
5.65060437e-01 -1.05204821e+00 1.29324973e-01 5.47948956e-01
7.45936930e-01 -1.34869361e+00 -1.98218271e-01 1.11874424e-01
-7.02563167e-01 6.29114747e-01 4.02279347e-01 3.02410811e-01
9.61012959e-01 3.82993996e-01 -3.65031250e-02 -1.99088588e-01
-1.02056885e+00 -6.35063291e-01 1.28700227e-01 9.34192419e-01
1.94117650e-01 3.92180771e-01 9.01601613e-02 3.77327204e-01
-6.19873069e-02 6.65840581e-02 -2.05252562e-02 7.51034975e-01
-3.43767464e-01 -1.06369984e+00 -4.56921428e-01 6.35833621e-01
-1.02922380e-01 -1.77697673e-01 1.51787624e-01 4.98943180e-01
4.69357431e-01 1.20790398e+00 3.39244045e-02 -7.45734572e-01
1.48971304e-01 2.03239881e-02 4.11491506e-02 -1.87822610e-01
-8.82861912e-01 -3.83452684e-01 1.36283655e-02 -4.51768875e-01
-6.19841397e-01 -7.21828043e-01 -1.01274681e+00 -5.26258349e-01
-9.10680145e-02 -2.01551110e-01 8.20497632e-01 9.45273936e-01
8.72783005e-01 5.43826103e-01 9.25923884e-01 -9.81357396e-01
-5.70513606e-01 -8.19401562e-01 -2.10109919e-01 3.02601784e-01
4.82724667e-01 -6.04360044e-01 2.04333458e-02 7.69547671e-02]
|
[6.782973766326904, 2.413879156112671]
|
522c97d4-702b-4077-b581-7fc275918805
|
predictive-and-contrastive-dual-auxiliary
|
2203.03982
| null |
https://arxiv.org/abs/2203.03982v2
|
https://arxiv.org/pdf/2203.03982v2.pdf
|
Predictive and Contrastive: Dual-Auxiliary Learning for Recommendation
|
Self-supervised learning (SSL) recently has achieved outstanding success on recommendation. By setting up an auxiliary task (either predictive or contrastive), SSL can discover supervisory signals from the raw data without human annotation, which greatly mitigates the problem of sparse user-item interactions. However, most SSL-based recommendation models rely on general-purpose auxiliary tasks, e.g., maximizing correspondence between node representations learned from the original and perturbed interaction graphs, which are explicitly irrelevant to the recommendation task. Accordingly, the rich semantics reflected by social relationships and item categories, which lie in the recommendation data-based heterogeneous graphs, are not fully exploited. To explore recommendation-specific auxiliary tasks, we first quantitatively analyze the heterogeneous interaction data and find a strong positive correlation between the interactions and the number of user-item paths induced by meta-paths. Based on the finding, we design two auxiliary tasks that are tightly coupled with the target task (one is predictive and the other one is contrastive) towards connecting recommendation with the self-supervision signals hiding in the positive correlation. Finally, a model-agnostic DUal-Auxiliary Learning (DUAL) framework which unifies the SSL and recommendation tasks is developed. The extensive experiments conducted on three real-world datasets demonstrate that DUAL can significantly improve recommendation, reaching the state-of-the-art performance.
|
['Xu Wang', 'Qingyu Xiong', 'Zongwei Wang', 'Junliang Yu', 'Min Gao', 'Yinghui Tao']
|
2022-03-08
| null | null | null | null |
['auxiliary-learning']
|
['methodology']
|
[ 2.04459384e-01 3.01893800e-01 -8.37737858e-01 -3.64438266e-01
-2.12158605e-01 -3.29458475e-01 6.47421658e-01 -5.37028797e-02
1.76126853e-01 6.44472063e-01 6.14168286e-01 -2.14500561e-01
-5.02686143e-01 -7.92607367e-01 -7.55137742e-01 -8.23012769e-01
-1.12910964e-01 3.57660890e-01 1.36971414e-01 -5.27551711e-01
-1.30107170e-02 -2.75362045e-01 -1.20212293e+00 4.21281844e-01
8.07825029e-01 8.53671670e-01 1.08709984e-01 1.13403603e-01
1.66142769e-02 5.64085603e-01 -2.75873840e-02 -2.14953795e-01
1.82875410e-01 -5.48549891e-01 -5.23339748e-01 8.48354176e-02
8.65196362e-02 1.02110110e-01 -5.81019998e-01 9.03451025e-01
1.66393176e-01 3.00611913e-01 6.39515519e-01 -1.37143087e+00
-1.03393459e+00 1.21386957e+00 -5.17482102e-01 1.32676929e-01
4.42147702e-01 -5.56278974e-02 1.69239485e+00 -8.08421612e-01
4.52376336e-01 1.16645467e+00 5.26880860e-01 2.12607995e-01
-1.43175375e+00 -7.69791067e-01 7.46396303e-01 -8.23265091e-02
-1.13950908e+00 -2.53530562e-01 1.01713908e+00 -3.77503604e-01
5.12996733e-01 1.84355780e-01 4.86352026e-01 1.57872117e+00
-2.64486611e-01 8.87548029e-01 8.49547684e-01 -8.68878141e-02
-6.30092695e-02 4.46572334e-01 6.50987327e-01 7.43650317e-01
4.52508211e-01 2.02584922e-01 -6.92050457e-01 -1.29716620e-01
7.21189559e-01 4.69337106e-01 -5.70653856e-01 -3.36185426e-01
-1.43058872e+00 8.52122724e-01 7.00448632e-01 3.11770648e-01
-3.08208346e-01 -3.30899000e-01 3.54709208e-01 6.13805175e-01
5.92692554e-01 3.79685998e-01 -5.81707597e-01 4.16184872e-01
-4.63529050e-01 -2.16467142e-01 7.66520858e-01 1.06016862e+00
8.08143318e-01 -6.93618655e-02 -3.94286662e-01 6.34992063e-01
7.26235151e-01 2.98791617e-01 4.14594173e-01 -2.93692470e-01
5.55731654e-01 8.12199533e-01 -9.44185704e-02 -1.41582930e+00
-4.25986081e-01 -1.19260526e+00 -1.13766730e+00 -6.77656829e-01
3.63053322e-01 -1.16030291e-01 -3.39903891e-01 1.89892268e+00
3.29950511e-01 6.04644001e-01 -7.15547875e-02 9.43886280e-01
1.16523397e+00 4.98447716e-01 -1.15113184e-01 -4.58991319e-01
1.01636684e+00 -1.30623460e+00 -6.45958126e-01 -4.78952527e-02
7.91120708e-01 -1.91921130e-01 1.43174243e+00 2.36949578e-01
-5.12858689e-01 -6.41024590e-01 -1.05735564e+00 2.14331806e-01
-9.58458930e-02 2.03475788e-01 9.04355764e-01 2.17999682e-01
-6.10751271e-01 7.41333663e-01 -4.16136652e-01 -2.59674668e-01
3.32303554e-01 3.63899559e-01 -3.92059326e-01 -4.05382477e-02
-1.52808595e+00 1.23267077e-01 5.54563552e-02 1.77476779e-01
-4.93047535e-01 -5.26514530e-01 -5.65320134e-01 1.92138568e-01
8.44189882e-01 -6.00682378e-01 6.70376420e-01 -1.07416630e+00
-1.47951519e+00 3.85573775e-01 9.19323228e-03 -7.71611184e-02
2.07275018e-01 -1.21389106e-01 -7.84422040e-01 -3.12492073e-01
6.47931993e-02 2.70057302e-02 8.93218577e-01 -1.37005711e+00
-5.80638766e-01 -4.12288547e-01 2.59802967e-01 1.55378535e-01
-8.17313969e-01 -5.52248538e-01 -6.33354783e-01 -8.59024167e-01
4.91510481e-02 -9.73047197e-01 -1.17000178e-01 -3.58609587e-01
-6.54610932e-01 -5.04396558e-01 6.75170243e-01 -4.08393115e-01
1.52519226e+00 -2.18019104e+00 4.60464507e-01 6.04836047e-01
3.74788940e-01 1.25419706e-01 -6.20464265e-01 3.25690567e-01
-5.81354871e-02 -4.37485017e-02 2.05030307e-01 -3.33500773e-01
-8.94830748e-02 3.96334440e-01 -5.17976582e-01 4.24520791e-01
-3.00409585e-01 1.10577416e+00 -1.34294975e+00 -1.56586453e-01
-1.12616926e-01 2.51790941e-01 -6.77598178e-01 3.94382268e-01
-2.51595140e-01 8.23477685e-01 -8.27591419e-01 3.77726763e-01
3.53404611e-01 -8.29908431e-01 4.65610474e-01 -6.26813054e-01
3.70998204e-01 6.06566966e-01 -1.12948358e+00 1.65347373e+00
-3.82183433e-01 -6.82149157e-02 6.95994645e-02 -1.19070995e+00
1.01905978e+00 1.72352210e-01 4.66644078e-01 -5.98662198e-01
-1.51219517e-01 -5.99457473e-02 2.13326588e-02 -3.20487261e-01
2.79312342e-01 2.25384355e-01 4.12090383e-02 8.07294488e-01
2.65433311e-01 7.42847323e-01 -4.85433377e-02 4.91479665e-01
1.00877285e+00 2.15696335e-01 2.81494647e-01 -3.21081668e-01
6.48935854e-01 -3.96900088e-01 6.86849594e-01 7.43080735e-01
2.48336822e-01 1.31641656e-01 6.30329847e-01 -6.67696446e-02
-4.29867744e-01 -7.50835836e-01 2.15989336e-01 1.40316367e+00
4.25986141e-01 -7.32271671e-01 -1.13412470e-01 -1.25621998e+00
6.29851520e-02 5.43041468e-01 -8.04416120e-01 -5.28031468e-01
-2.73285866e-01 -7.41176665e-01 -1.28116682e-01 4.43236351e-01
2.79681206e-01 -9.55563307e-01 6.40518129e-01 1.01760447e-01
-1.06375508e-01 -9.55737531e-01 -7.67188787e-01 -1.79338772e-02
-9.61964667e-01 -1.09875441e+00 -3.14922959e-01 -5.80169261e-01
8.68111730e-01 9.52664137e-01 1.22105205e+00 4.06677455e-01
5.08532703e-01 3.91280413e-01 -7.94156075e-01 3.04050088e-01
-2.22988967e-02 3.09436858e-01 1.47813156e-01 4.62448508e-01
4.05807048e-01 -1.03989887e+00 -6.33141577e-01 5.92529833e-01
-5.83428621e-01 2.57161647e-01 7.85542488e-01 1.05047345e+00
5.26010513e-01 -1.60869002e-01 9.76728976e-01 -1.56354964e+00
7.49345481e-01 -1.14970899e+00 -9.83442292e-02 2.58670211e-01
-9.85571623e-01 6.37608096e-02 9.82874393e-01 -6.04075015e-01
-8.65197957e-01 -2.57059932e-01 1.02579236e-01 -3.19460958e-01
-1.48636419e-02 8.32342088e-01 -3.16511065e-01 1.84779212e-01
5.90321481e-01 1.33216456e-01 -9.80336443e-02 -6.39276505e-01
5.62959135e-01 7.42859781e-01 1.72173917e-01 -4.48906988e-01
9.34384763e-01 4.08235162e-01 -1.05612546e-01 -5.15146792e-01
-1.64789796e+00 -7.18177736e-01 -5.40506780e-01 -1.20930590e-01
2.24030629e-01 -9.19228494e-01 -5.72360039e-01 -1.06279127e-01
-5.30386388e-01 -2.68188745e-01 -3.59630138e-01 6.13195598e-01
-2.07432479e-01 4.47343677e-01 -5.47710240e-01 -4.53321546e-01
-4.04254258e-01 -8.23578238e-01 8.34946871e-01 5.20758405e-02
-1.56015649e-01 -1.20260072e+00 1.59119517e-01 4.93301123e-01
2.19391838e-01 -1.53558820e-01 1.03406322e+00 -9.87351656e-01
-4.98367369e-01 -1.35997877e-01 -3.91030550e-01 1.39685795e-01
4.80891347e-01 -4.60341603e-01 -6.72044575e-01 -4.81000155e-01
-3.03013444e-01 -2.39885703e-01 9.03309345e-01 7.02444315e-02
1.01858282e+00 -5.22983372e-01 -6.05893970e-01 4.82666731e-01
9.19716835e-01 -3.77123535e-01 1.32837385e-01 -7.02710450e-02
1.14318395e+00 5.71720064e-01 6.64466560e-01 3.46484393e-01
5.79473853e-01 7.85027504e-01 3.77100497e-01 -1.62547305e-01
-1.28623664e-01 -9.64326084e-01 5.59326291e-01 1.19100475e+00
-1.78819284e-01 -3.32461357e-01 -2.55434215e-01 2.32291460e-01
-2.46936274e+00 -7.91710973e-01 -3.53313535e-01 2.30615544e+00
5.94419241e-01 1.77902058e-01 2.75615454e-01 -6.79013580e-02
7.10034430e-01 3.26692045e-01 -7.32734799e-01 2.90520936e-01
-5.42761870e-02 -2.24462584e-01 1.25536472e-01 3.42908144e-01
-1.01086140e+00 8.02718699e-01 5.21041155e+00 7.89570391e-01
-8.93524587e-01 1.63171232e-01 3.64885747e-01 6.25554472e-02
-8.03287029e-01 2.30001612e-03 -7.67255187e-01 6.14647985e-01
6.71750486e-01 -9.45893116e-03 6.05250776e-01 7.42683887e-01
2.74270624e-01 4.92619634e-01 -1.38196671e+00 7.39471793e-01
1.00311406e-01 -1.21299744e+00 2.84403563e-01 2.44508833e-01
8.63652349e-01 -4.14350405e-02 4.49021161e-02 6.59643352e-01
5.23422837e-01 -7.13154376e-01 3.34937811e-01 6.67914093e-01
5.78507900e-01 -4.64516699e-01 7.99351275e-01 5.75475931e-01
-1.34717512e+00 -2.36770853e-01 -3.40088636e-01 -1.40559837e-01
7.99462199e-02 8.80931973e-01 -2.45888531e-01 8.79783630e-01
4.90480721e-01 1.80636191e+00 -3.93309921e-01 4.87702072e-01
-6.42140746e-01 1.07060695e+00 4.23542149e-02 8.38725641e-03
7.93064609e-02 -8.11483204e-01 6.54795229e-01 1.01910162e+00
7.45693520e-02 1.38628259e-01 6.24617457e-01 7.19881177e-01
-2.94412732e-01 3.78383279e-01 -6.11236989e-01 -1.42667428e-01
3.66419554e-01 1.46231115e+00 -4.92109537e-01 -1.24213673e-01
-7.91762888e-01 7.33007371e-01 6.37101054e-01 5.72179317e-01
-5.78251600e-01 1.33082837e-01 5.01534402e-01 1.75699085e-01
2.89471775e-01 2.81103817e-03 -1.01172134e-01 -1.42456293e+00
-7.01170787e-02 -7.65660703e-01 5.07077992e-01 -2.92792588e-01
-1.85783803e+00 2.73957759e-01 -3.33114505e-01 -1.37878919e+00
3.95076908e-02 -2.23480865e-01 -8.06889415e-01 5.23615777e-01
-1.53743255e+00 -1.34483743e+00 -3.67765546e-01 8.46018553e-01
2.35552505e-01 -4.27716106e-01 7.46259570e-01 5.09740293e-01
-9.12791848e-01 9.75453019e-01 2.23614588e-01 1.36241596e-02
7.72362709e-01 -1.04221594e+00 -6.30581155e-02 5.90072334e-01
4.69950736e-01 8.62318575e-01 3.07313681e-01 -5.95299661e-01
-1.65602028e+00 -1.35195184e+00 4.68976557e-01 -3.72730345e-01
9.10918057e-01 -4.78520244e-01 -9.16905880e-01 9.07440782e-01
-2.82980591e-01 1.35973349e-01 9.66150999e-01 7.50844717e-01
-6.02368951e-01 -1.96686491e-01 -6.14949405e-01 6.11284435e-01
1.65707505e+00 -5.80597818e-01 -5.33105075e-01 4.62923795e-01
1.05388761e+00 2.04773650e-01 -8.13154459e-01 4.68734473e-01
4.14456278e-01 -7.44967043e-01 9.90020990e-01 -1.07953155e+00
5.00338614e-01 -2.13291675e-01 -5.68254702e-02 -1.54587352e+00
-8.81306708e-01 -6.93556547e-01 -7.81921744e-01 1.46098030e+00
5.60504317e-01 -7.03516185e-01 6.62126660e-01 9.63707939e-02
-1.02758445e-01 -8.77631187e-01 -2.97216028e-01 -6.98287129e-01
-4.10208642e-01 1.20099792e-02 4.55806345e-01 1.28065312e+00
2.56773293e-01 1.04192185e+00 -7.59525239e-01 6.47689402e-02
6.33975804e-01 7.13989198e-01 7.91803479e-01 -1.62725663e+00
-7.41952658e-01 -1.94285661e-01 4.33530621e-02 -1.61081791e+00
4.97623056e-01 -1.38940382e+00 -2.58150220e-01 -1.34455895e+00
4.14374113e-01 -8.15173984e-01 -8.40140998e-01 5.28890371e-01
-3.53001565e-01 4.97731790e-02 -2.22911909e-01 5.33554375e-01
-9.73752856e-01 7.89812028e-01 1.45077884e+00 -1.53571218e-01
-3.68771434e-01 3.76253664e-01 -1.40182531e+00 6.19656622e-01
4.57496107e-01 -4.96014506e-01 -8.17510188e-01 -7.66593814e-02
5.38452685e-01 -1.90717801e-02 -5.31610623e-02 -2.28761420e-01
2.28637606e-01 -1.59881726e-01 -6.82924539e-02 -2.00145021e-01
-4.41202261e-02 -8.17282557e-01 -5.79635277e-02 1.51122034e-01
-7.35432804e-01 -4.64505970e-01 -5.24640441e-01 1.34381080e+00
-8.96195993e-02 1.20109707e-01 3.16450596e-01 7.64121711e-02
-4.55787867e-01 7.20733404e-01 2.67563075e-01 2.10186094e-02
7.12926567e-01 1.46618590e-01 -3.53548318e-01 -5.43166816e-01
-8.44889462e-01 3.78149182e-01 1.07460476e-01 6.42889977e-01
4.11016434e-01 -1.52876461e+00 -6.50117993e-01 1.38614669e-01
4.16360795e-01 -1.87780887e-01 2.28867963e-01 1.04075730e+00
6.25951350e-01 3.98609519e-01 1.02588303e-01 -4.27846670e-01
-1.00194252e+00 8.99296403e-01 -8.97348672e-02 -6.79490745e-01
-7.70495474e-01 7.51252472e-01 3.69375229e-01 -5.82797408e-01
3.04883599e-01 5.20871095e-02 -6.92591548e-01 1.51273295e-01
3.99184346e-01 3.30221474e-01 -1.87268749e-01 -5.01451194e-01
-2.44146530e-02 3.51211756e-01 -2.13312060e-01 5.38961053e-01
1.60964882e+00 -3.51659268e-01 -1.42912015e-01 5.50291777e-01
1.06226015e+00 1.49131998e-01 -1.06425691e+00 -1.01203191e+00
-8.03805739e-02 -4.74583179e-01 2.53451139e-01 -4.95269209e-01
-1.42397833e+00 5.89099705e-01 5.23790568e-02 3.97837967e-01
8.07088196e-01 1.62440151e-01 7.75644004e-01 5.50531685e-01
5.12656391e-01 -7.75818408e-01 3.71243775e-01 4.08150315e-01
8.30937028e-01 -1.42554033e+00 6.94335699e-02 -6.96346223e-01
-6.15069032e-01 7.82428145e-01 7.58335412e-01 -2.36066625e-01
1.10575032e+00 -2.03642756e-01 -4.76472229e-01 -2.76691437e-01
-8.00876915e-01 -1.28928050e-01 6.59401774e-01 3.93873066e-01
5.43143451e-01 2.51687050e-01 -3.27968478e-01 1.26540589e+00
1.80159211e-01 -1.10240452e-01 1.35568038e-01 2.64736116e-01
-2.01256692e-01 -1.11552751e+00 2.49924093e-01 9.66524541e-01
9.81984288e-02 -1.29641384e-01 -3.87076020e-01 3.75342965e-01
-1.15405940e-01 1.14098525e+00 -1.66891903e-01 -8.42281818e-01
2.86078036e-01 -3.28769147e-01 -3.24985012e-02 -9.01156783e-01
-5.90461671e-01 2.37948559e-02 2.17949420e-01 -6.84567034e-01
-5.07678807e-01 -3.27879012e-01 -9.23559487e-01 -1.74343809e-01
-6.20623708e-01 4.50509816e-01 1.19569385e-02 1.17440987e+00
5.80249786e-01 6.23477757e-01 1.08262360e+00 -7.28864193e-01
-6.78821981e-01 -9.64703023e-01 -8.05026531e-01 6.13146305e-01
1.73372328e-01 -9.10682738e-01 -6.22581005e-01 -3.41750383e-01]
|
[10.22691822052002, 5.612393379211426]
|
5f9c0f31-b053-407e-901b-3884a2b3c1b2
|
unsupervised-source-hierarchies-for-low
| null | null |
https://aclanthology.org/W18-2902
|
https://aclanthology.org/W18-2902.pdf
|
Unsupervised Source Hierarchies for Low-Resource Neural Machine Translation
|
Incorporating source syntactic information into neural machine translation (NMT) has recently proven successful (Eriguchi et al., 2016; Luong et al., 2016). However, this is generally done using an outside parser to syntactically annotate the training data, making this technique difficult to use for languages or domains for which a reliable parser is not available. In this paper, we introduce an unsupervised tree-to-sequence (tree2seq) model for neural machine translation; this model is able to induce an unsupervised hierarchical structure on the source sentence based on the downstream task of neural machine translation. We adapt the Gumbel tree-LSTM of Choi et al. (2018) to NMT in order to create the encoder. We evaluate our model against sequential and supervised parsing baselines on three low- and medium-resource language pairs. For low-resource cases, the unsupervised tree2seq encoder significantly outperforms the baselines; no improvements are seen for medium-resource translation.
|
['Kenneth Heafield', 'Anna Currey']
|
2018-07-01
| null | null | null |
ws-2018-7
|
['low-resource-neural-machine-translation']
|
['natural-language-processing']
|
[ 4.77422029e-01 4.18035746e-01 -2.57459670e-01 -6.04194164e-01
-1.42526281e+00 -8.89794588e-01 4.38955307e-01 -1.10369489e-01
-4.99594122e-01 9.83691394e-01 4.21837002e-01 -9.53937709e-01
7.20129609e-01 -6.02145791e-01 -1.02555156e+00 -3.10827494e-01
2.63201803e-01 6.63608551e-01 -3.14124793e-01 -2.15523064e-01
-1.63358271e-01 -2.44683400e-02 -6.90110266e-01 6.87623918e-01
9.69997227e-01 4.18147266e-01 5.03762424e-01 6.96103275e-01
-1.55620888e-01 7.17233717e-01 -4.71046865e-01 -6.75277531e-01
3.83336723e-01 -8.75433207e-01 -9.61378694e-01 -3.62362146e-01
4.96251464e-01 -3.18075627e-01 1.32516682e-01 8.83277655e-01
4.91985321e-01 -3.41567755e-01 5.26264548e-01 -4.37651515e-01
-8.83207798e-01 1.10503078e+00 -4.36803728e-01 1.08442880e-01
1.80476844e-01 -2.75586769e-02 1.23069906e+00 -1.02609599e+00
9.94549572e-01 1.27156615e+00 5.44954240e-01 8.13148677e-01
-1.40611863e+00 -5.93682766e-01 -1.43156111e-01 -1.51584774e-01
-8.52680802e-01 -6.53306365e-01 3.83455396e-01 -2.51538277e-01
1.35799742e+00 -1.48389310e-01 2.38288268e-01 1.43127406e+00
4.88562703e-01 7.45215595e-01 1.29904294e+00 -7.57074177e-01
1.39801934e-01 -1.64963648e-01 -2.31347993e-01 5.12131453e-01
-6.12736493e-02 9.41525176e-02 -6.11450255e-01 1.04863405e-01
6.77369773e-01 -6.39859438e-01 2.51857311e-01 2.25203797e-01
-1.31936610e+00 8.65622103e-01 2.64025480e-01 5.09522378e-01
-4.02461857e-01 1.19781584e-01 6.63979948e-01 6.36071563e-01
8.53652358e-01 5.17769575e-01 -7.63404846e-01 -3.91520232e-01
-9.35310364e-01 -1.53476804e-01 9.16988432e-01 1.07591426e+00
5.70035696e-01 1.06856957e-01 -2.67067909e-01 9.05851901e-01
6.65722415e-02 5.00340223e-01 5.25980890e-01 -9.39359665e-01
9.89722967e-01 7.29699209e-02 -2.05084920e-01 -2.32142672e-01
-1.24540702e-01 -3.50459754e-01 -7.43220508e-01 -1.93488389e-01
6.07746065e-01 -5.90318739e-01 -9.15044248e-01 1.86294293e+00
-6.89615961e-03 -1.27857938e-01 4.52304691e-01 8.17242146e-01
6.17871702e-01 8.38347793e-01 2.40033835e-01 -2.07118183e-01
1.16602015e+00 -1.28486347e+00 -6.63386464e-01 -5.09106398e-01
1.10690212e+00 -9.53890145e-01 1.11284113e+00 5.26956208e-02
-1.19461930e+00 -4.80937809e-01 -7.44902253e-01 -3.82034540e-01
-6.10127673e-02 3.57997507e-01 5.51299334e-01 3.89520109e-01
-1.40544128e+00 7.46994078e-01 -9.58140433e-01 -5.29195070e-01
2.44359404e-01 3.55406404e-01 -4.91590083e-01 -3.89770865e-02
-1.38421464e+00 1.27719831e+00 2.56734967e-01 2.26074830e-01
-8.30047786e-01 -4.42662358e-01 -8.97420526e-01 -2.71276176e-01
-1.68877214e-01 -8.25169325e-01 1.72331762e+00 -1.41553700e+00
-1.97243381e+00 1.10722053e+00 -5.45088530e-01 -7.38334358e-01
3.33077699e-01 -3.73840630e-01 1.66418940e-01 -6.27736971e-02
3.40592831e-01 9.99948561e-01 5.61901629e-01 -7.37165391e-01
-4.57000047e-01 -2.26606071e-01 -1.55632973e-01 3.56006086e-01
-3.89144570e-02 6.03091061e-01 -3.09359804e-02 -6.52448893e-01
-2.47693479e-01 -1.00147939e+00 -2.85817176e-01 -6.07417464e-01
-3.95758510e-01 -2.52986044e-01 2.80259997e-01 -1.28898251e+00
9.20868933e-01 -1.85228026e+00 4.54911083e-01 -4.06541228e-01
-3.66036832e-01 2.52583653e-01 -5.13129711e-01 6.01022363e-01
3.70827802e-02 3.90265048e-01 -5.86916685e-01 -6.40072227e-01
-2.40331829e-01 4.50169683e-01 -3.47738922e-01 9.88548100e-02
6.13537312e-01 1.34066510e+00 -7.63013482e-01 -4.57170248e-01
-3.01064670e-01 3.16575646e-01 -4.79218900e-01 4.25921172e-01
-3.36406857e-01 9.06967521e-01 -1.62092537e-01 5.59842229e-01
1.94951743e-01 1.68624759e-01 3.77386093e-01 3.59509736e-01
-3.45346838e-01 1.13234270e+00 -1.07864246e-01 2.09148550e+00
-9.97489750e-01 7.62367249e-01 1.67072564e-01 -7.44240999e-01
8.29127491e-01 5.77999592e-01 -2.43118708e-03 -6.63015664e-01
4.30360138e-02 7.81569958e-01 4.36655432e-01 -3.26789141e-01
2.45242923e-01 -3.46821696e-01 -2.56949514e-01 4.94647294e-01
2.62293339e-01 -1.33091763e-01 1.92396045e-01 -4.60758656e-02
1.23712158e+00 7.58258224e-01 1.41829431e-01 -3.06251079e-01
1.73146203e-01 2.97677726e-01 5.65326333e-01 4.72720861e-01
5.45203835e-02 5.86606383e-01 4.64581132e-01 -1.73531383e-01
-1.36864865e+00 -8.24139237e-01 8.17743689e-03 1.25809717e+00
-7.61951745e-01 -2.86515415e-01 -1.20829618e+00 -8.59855592e-01
-4.97012436e-01 8.31673563e-01 -4.72771615e-01 1.08091876e-01
-1.12489414e+00 -5.38062871e-01 8.36311340e-01 4.29836690e-01
3.56087029e-01 -1.20436215e+00 -2.41531968e-01 5.65353632e-01
-4.93073285e-01 -1.25815725e+00 -5.09151995e-01 4.81657118e-01
-1.15371859e+00 -3.34422916e-01 -6.77829921e-01 -1.06306422e+00
4.52197254e-01 -1.89888299e-01 1.29562318e+00 -2.81262368e-01
3.64193648e-01 -2.66322136e-01 -3.94968957e-01 -3.89744312e-01
-9.63758647e-01 6.05742276e-01 -4.29842509e-02 -3.27750176e-01
4.78398621e-01 -5.52541673e-01 -1.52996242e-01 -1.44186407e-01
-5.37640512e-01 4.66922462e-01 9.61479962e-01 9.17793334e-01
5.62579393e-01 -7.93717444e-01 7.38180101e-01 -9.70807314e-01
5.68935156e-01 -4.53130573e-01 -4.13715750e-01 2.81130262e-02
-3.07002306e-01 2.06732005e-01 1.07401621e+00 -1.82955548e-01
-1.01561153e+00 1.09208643e-01 -4.95108187e-01 -1.51374228e-02
-2.18014583e-01 5.90620875e-01 -2.17784837e-01 2.48595834e-01
3.70187253e-01 1.41664743e-01 -2.23261446e-01 -7.28817523e-01
3.99024129e-01 9.34865892e-01 3.38047802e-01 -6.69183195e-01
7.08198369e-01 -6.53298572e-02 -7.52966329e-02 -3.88880163e-01
-1.02097452e+00 -4.80734594e-02 -1.18363881e+00 2.54993707e-01
1.19350302e+00 -1.08819950e+00 1.70811996e-01 2.58376569e-01
-1.73070312e+00 -8.62828374e-01 -2.51909345e-02 5.32619298e-01
-6.64415300e-01 3.66529040e-02 -1.12350917e+00 -3.87699217e-01
-5.93607724e-01 -1.23277998e+00 1.31639326e+00 -2.50538111e-01
-3.71001214e-01 -1.05711186e+00 1.24424689e-01 3.83750916e-01
4.92250979e-01 4.22623120e-02 1.01454782e+00 -8.07002246e-01
-4.34000939e-01 2.96223253e-01 -8.97663236e-02 5.55095315e-01
1.46837190e-01 -2.09354103e-01 -7.84626186e-01 -1.70255408e-01
3.18889916e-02 -4.86137986e-01 7.45923281e-01 1.87256321e-01
6.25667334e-01 -4.96217966e-01 7.97707811e-02 5.35936415e-01
1.02758992e+00 -2.01869067e-02 5.41611850e-01 4.35804337e-01
7.91063070e-01 9.43986714e-01 4.03987616e-01 -4.56793904e-01
5.34482419e-01 7.12564707e-01 1.08385226e-02 -1.49983048e-01
-2.30288908e-01 -4.90978926e-01 1.12579632e+00 1.46385431e+00
-7.44879693e-02 -1.63464487e-01 -8.92843068e-01 5.75922370e-01
-1.67543125e+00 -5.39828777e-01 -4.30530787e-01 2.01141143e+00
1.36374605e+00 5.02640940e-03 -1.44286543e-01 -5.74429452e-01
6.60095394e-01 -1.51965216e-01 -1.68831885e-01 -1.16839492e+00
-8.50336105e-02 5.90265989e-01 5.11118472e-01 6.25254393e-01
-8.68719459e-01 1.58008909e+00 5.80749130e+00 6.17288709e-01
-1.15168941e+00 6.42573178e-01 6.97692513e-01 2.18852106e-02
-2.23332196e-01 3.79042983e-01 -8.63524616e-01 4.20666099e-01
1.78841221e+00 1.05175793e-01 5.00689864e-01 4.72298324e-01
4.01036292e-01 9.07417759e-02 -1.36308873e+00 4.33010012e-01
-2.17390247e-02 -1.03730679e+00 -1.48858607e-01 1.03138387e-01
6.67395473e-01 6.45816207e-01 -2.32074618e-01 3.90349895e-01
6.24845266e-01 -1.18180561e+00 6.04014099e-01 -2.64501385e-03
1.09680712e+00 -5.48862219e-01 8.25296938e-01 5.53779781e-01
-7.15548694e-01 3.54757458e-01 -3.57011259e-01 -2.39640936e-01
4.50271219e-01 4.87151861e-01 -9.70196784e-01 5.12082338e-01
5.26434302e-01 7.12469935e-01 -4.56240296e-01 3.67561430e-01
-8.59848320e-01 1.25257039e+00 -2.89404035e-01 8.11856836e-02
5.05813420e-01 -4.12173659e-01 5.13019741e-01 1.43212605e+00
3.54195058e-01 -2.61279345e-01 7.57534429e-02 7.85503209e-01
-5.15281737e-01 4.50170249e-01 -7.78510749e-01 -2.59570777e-01
1.81609213e-01 1.18435526e+00 -4.94528413e-01 -3.48650068e-01
-5.07000029e-01 1.36293423e+00 7.92654753e-01 2.61623114e-01
-4.61117893e-01 -6.43599182e-02 4.61955845e-01 -1.66808128e-01
1.79022700e-01 -5.07597566e-01 -4.75428075e-01 -1.29439819e+00
6.38331473e-02 -1.17481840e+00 7.89651200e-02 -6.10126913e-01
-1.19941878e+00 8.33935380e-01 -2.32294306e-01 -9.14122939e-01
-7.30959833e-01 -5.56528151e-01 -5.70516527e-01 1.34703553e+00
-1.51601779e+00 -1.51438129e+00 5.93544900e-01 -4.11310308e-02
8.21070850e-01 -1.66131899e-01 1.13521183e+00 2.11314112e-01
-5.08549750e-01 5.58282077e-01 1.85536921e-01 4.31468427e-01
9.06053185e-01 -1.25471783e+00 1.12066317e+00 1.05051196e+00
3.22226405e-01 7.42333889e-01 4.89344090e-01 -7.59577692e-01
-1.50030208e+00 -1.33243215e+00 1.58652723e+00 -7.06188798e-01
7.52783537e-01 -9.64648426e-01 -8.71249259e-01 1.02205491e+00
9.26903665e-01 -3.73253524e-01 7.28937209e-01 1.04172364e-01
-3.47733617e-01 2.72379369e-01 -7.64921725e-01 5.79362035e-01
1.04544079e+00 -5.43775201e-01 -5.33152819e-01 4.48485941e-01
1.05782425e+00 -4.64253426e-01 -9.88979697e-01 3.89386207e-01
3.60628515e-01 -5.77755034e-01 2.76852518e-01 -7.73175418e-01
1.03273571e+00 9.61106177e-03 -8.57542455e-02 -1.60313261e+00
-1.10474072e-01 -8.49060059e-01 1.41105428e-01 1.36638582e+00
1.03623950e+00 -6.72738731e-01 3.77271563e-01 9.86690149e-02
-4.32164758e-01 -7.10911870e-01 -1.17042780e+00 -8.55828226e-01
8.86897743e-01 -2.00987011e-01 2.61152029e-01 7.19238102e-01
-2.61214543e-02 8.98324907e-01 -3.00825238e-01 -1.48359314e-01
3.20998102e-01 6.68024085e-03 7.06488490e-01 -8.07129502e-01
-6.84272587e-01 -2.84475178e-01 1.26318768e-01 -1.10758424e+00
6.54957056e-01 -1.39474654e+00 2.88474858e-01 -1.75383937e+00
1.87034220e-01 -2.76449114e-01 5.02196588e-02 7.21504211e-01
-1.13215812e-01 3.47736716e-01 2.28348047e-01 3.84740591e-01
-1.27937138e-01 4.19462621e-01 1.16237307e+00 8.13956484e-02
-2.55856849e-02 -1.96400896e-01 -6.07679725e-01 3.77093345e-01
1.06110060e+00 -7.40656137e-01 8.69274437e-02 -1.10527134e+00
2.48527169e-01 1.04347363e-01 -4.95831817e-02 -4.35095072e-01
-7.47446641e-02 1.12159453e-01 1.00719415e-01 -2.38319665e-01
-9.24756229e-02 -4.37863767e-01 -2.28123173e-01 3.78560305e-01
-5.50463796e-01 4.36711043e-01 3.13570231e-01 2.00848728e-02
-3.61161202e-01 -4.36114877e-01 5.63130021e-01 -3.33771050e-01
-1.05720937e-01 -4.02268134e-02 -6.43543541e-01 1.27861843e-01
5.36493838e-01 1.81985244e-01 -1.80091023e-01 -4.58815873e-01
-3.56719404e-01 6.57159537e-02 5.26755273e-01 5.37324548e-01
1.79217860e-01 -1.16623127e+00 -1.14786887e+00 3.35163833e-03
-1.50370076e-01 1.36686072e-01 -4.77089286e-01 1.05840611e+00
-4.69076157e-01 5.76416671e-01 -7.27395490e-02 -3.69513392e-01
-9.75737751e-01 3.10028136e-01 1.58857733e-01 -5.64671457e-01
-4.08784628e-01 7.53777325e-01 8.28569606e-02 -8.62461448e-01
-2.45979398e-01 -3.71427089e-01 6.17580973e-02 -1.92265317e-01
1.92831233e-01 -1.23979986e-01 2.29354322e-01 -8.87262702e-01
-1.23305723e-01 2.92311370e-01 -2.16081768e-01 -4.94660646e-01
1.32592738e+00 -1.31883904e-01 -5.98415911e-01 5.86066365e-01
1.25135195e+00 9.65010822e-02 -8.87088716e-01 -8.29931423e-02
3.78228903e-01 1.92451775e-01 -2.12495714e-01 -8.59143078e-01
-5.77966213e-01 1.23505902e+00 2.79224832e-02 -2.66300768e-01
8.88651252e-01 -8.40910450e-02 1.28105116e+00 4.12107378e-01
5.36691189e-01 -9.99823511e-01 -5.44132829e-01 1.21837091e+00
7.80871868e-01 -1.19561505e+00 -6.64160550e-01 -3.42867136e-01
-4.93708968e-01 1.21513975e+00 4.92453456e-01 -2.67600920e-02
9.48151425e-02 4.46148366e-01 3.64773065e-01 3.05379897e-01
-9.99862194e-01 -1.79663241e-01 1.16603345e-01 4.18335408e-01
9.86723781e-01 1.64431751e-01 -7.16432512e-01 5.14848888e-01
-4.94422942e-01 -1.02455541e-01 4.61233258e-01 7.92166591e-01
-7.31242597e-02 -1.73606110e+00 -1.51226416e-01 2.89080739e-01
-9.73582149e-01 -8.06616843e-01 -8.30876172e-01 4.25915688e-01
-6.78246915e-02 1.05740619e+00 4.51609083e-02 -2.14827389e-01
1.15069941e-01 4.49837148e-01 5.44932187e-01 -1.24959254e+00
-9.36289668e-01 2.50068575e-01 6.21072292e-01 -3.61019522e-01
-3.55438620e-01 -8.09010625e-01 -1.25695348e+00 1.60054937e-01
-1.10023163e-01 2.79216945e-01 8.70783985e-01 1.05820835e+00
5.28081536e-01 3.69704068e-01 5.07183969e-01 -8.42001319e-01
-5.56827724e-01 -1.46445370e+00 1.53765455e-01 8.23790431e-02
8.35657418e-02 6.80911774e-03 -2.04366893e-01 3.84099156e-01]
|
[11.602338790893555, 10.246121406555176]
|
8ded1157-92d5-4293-97c7-a108024b7cbd
|
enriching-object-detection-with-2d-3d
| null | null |
http://openaccess.thecvf.com/content_cvpr_2015/html/Choy_Enriching_Object_Detection_2015_CVPR_paper.html
|
http://openaccess.thecvf.com/content_cvpr_2015/papers/Choy_Enriching_Object_Detection_2015_CVPR_paper.pdf
|
Enriching Object Detection With 2D-3D Registration and Continuous Viewpoint Estimation
|
A large body of recent work on object detection has focused on exploiting 3D CAD model databases to improve detection performance. Many of these approaches work by aligning exact 3D models to images using templates generated from renderings of the 3D models at a set of discrete viewpoints. However, the training procedures for these approaches are computationally expensive and require gigabytes of memory and storage, while the viewpoint discretization hampers pose estimation performance. We propose an efficient method for synthesizing templates from 3D models that runs on the fly -- that is, it quickly produces detectors for an arbitrary viewpoint of a 3D model without expensive dataset-dependent training or template storage. Given a 3D model and an arbitrary continuous detection viewpoint, our method synthesizes a discriminative template by extracting features from a rendered view of the object and decorrelating spatial dependences among the features. Our decorrelation procedure relies on a gradient-based algorithm that is more numerically stable than standard decomposition-based procedures, and we efficiently search for candidate detections by computing FFT-based template convolutions. Due to the speed of our template synthesis procedure, we are able to perform joint optimization of scale, translation, continuous rotation, and focal length using Metropolis-Hastings algorithm. We provide an efficient GPU implementation of our algorithm, and we validate its performance on 3D Object Classes and PASCAL3D+ datasets.
|
['Sam Corbett-Davies', 'Christopher Bongsoo Choy', 'Silvio Savarese', 'Michael Stark']
|
2015-06-01
| null | null | null |
cvpr-2015-6
|
['viewpoint-estimation']
|
['computer-vision']
|
[ 2.14564443e-01 -4.48380172e-01 2.51722038e-01 -1.87705040e-01
-8.53000760e-01 -7.67982423e-01 6.36562586e-01 -9.71627533e-02
-3.60682696e-01 -1.60086453e-01 -4.04086441e-01 -2.53932774e-01
1.61087364e-01 -7.57760227e-01 -6.80806518e-01 -4.53767627e-01
1.23471744e-01 9.48099494e-01 8.60628068e-01 1.83681056e-01
4.42952991e-01 1.27373946e+00 -1.77038801e+00 1.23718768e-01
1.60301626e-01 1.03780007e+00 2.87736952e-01 9.13331389e-01
2.91083772e-02 2.03916579e-02 -5.74134409e-01 -2.87908107e-01
7.02036440e-01 -2.50232786e-01 -3.23469520e-01 6.82124734e-01
7.35159516e-01 -3.27517688e-01 -2.61247247e-01 8.23647559e-01
4.44721460e-01 -9.40740556e-02 8.45072806e-01 -9.13103819e-01
-3.54546160e-02 -5.74837267e-01 -6.71199083e-01 6.99678510e-02
3.68862689e-01 1.23884983e-01 6.76769137e-01 -1.38666844e+00
9.09410894e-01 1.38585591e+00 6.93235099e-01 3.84145945e-01
-1.37191319e+00 -3.04301828e-01 -1.35097966e-01 -1.59171186e-02
-1.64201832e+00 -3.19245249e-01 7.85755157e-01 -5.56887984e-01
1.11948192e+00 5.20819604e-01 9.24090922e-01 6.84160829e-01
2.04067320e-01 6.02671146e-01 1.14547694e+00 -7.26247251e-01
2.49753475e-01 1.81482539e-01 -2.77559161e-02 9.87958193e-01
3.31472605e-01 3.40522319e-01 -4.50727552e-01 -4.67020094e-01
1.16409147e+00 -9.72601995e-02 9.99989212e-02 -1.06087935e+00
-1.14045703e+00 7.69237280e-01 7.43634030e-02 -2.72658527e-01
-3.10613602e-01 -1.81850623e-02 1.10459089e-01 6.19894341e-02
5.44127882e-01 1.54825300e-01 -1.89932048e-01 1.53721437e-01
-8.94835413e-01 5.24556577e-01 8.05829942e-01 1.01688397e+00
8.11561525e-01 -3.06293190e-01 1.25463903e-01 4.08041000e-01
1.93810880e-01 8.51694882e-01 -1.53507469e-02 -9.89904583e-01
6.69184923e-02 6.22899115e-01 2.55951852e-01 -9.96525407e-01
-4.73861814e-01 -4.31935906e-01 -3.53617698e-01 6.61487162e-01
4.53140467e-01 3.65236640e-01 -7.60457695e-01 1.11148441e+00
9.95857060e-01 -6.96915481e-03 -3.54919881e-01 9.75697935e-01
3.46667677e-01 3.57049465e-01 -6.75283372e-01 -9.32554230e-02
1.32585073e+00 -6.27859890e-01 -5.69364205e-02 -1.33748442e-01
5.13694406e-01 -1.18929660e+00 5.96142054e-01 3.23389500e-01
-1.07991397e+00 -6.41602576e-01 -1.18809497e+00 -6.53575957e-02
-1.68914899e-01 4.66800332e-01 3.72136682e-01 6.91295028e-01
-8.00856292e-01 3.59858811e-01 -9.94770288e-01 -3.82868081e-01
4.59376872e-02 4.71861571e-01 -3.01072806e-01 5.20010777e-02
-1.86199471e-01 1.18697906e+00 8.63573477e-02 -2.39974812e-01
-7.80520976e-01 -3.77728790e-01 -6.49245620e-01 -2.24192679e-01
2.76074231e-01 -6.97716713e-01 1.23600495e+00 -3.47197294e-01
-1.34497273e+00 1.40503073e+00 -3.28908443e-01 -2.65649259e-01
5.60551047e-01 -2.17759311e-02 -5.62823974e-02 1.91576794e-01
1.76255386e-02 5.29829085e-01 1.16536295e+00 -9.86620784e-01
-5.00136435e-01 -7.11431682e-01 -2.63644934e-01 3.65736336e-01
6.13349602e-02 1.90340802e-01 -9.01106417e-01 -2.68861622e-01
7.36758232e-01 -1.13174820e+00 -3.32143813e-01 4.29863960e-01
-9.98022333e-02 -4.59374078e-02 1.01232255e+00 -3.11964303e-01
6.16693616e-01 -2.33765268e+00 -1.95472408e-02 3.45437676e-01
2.15791732e-01 2.36323372e-01 -4.59316820e-02 3.10876846e-01
2.74673969e-01 -3.73583078e-01 1.46779254e-01 -4.06547666e-01
-2.21929587e-02 1.16464466e-01 -3.67124617e-01 8.59259307e-01
2.30738968e-01 5.73553741e-01 -6.80028319e-01 -5.60049891e-01
5.66978812e-01 5.31098485e-01 -7.47908890e-01 8.96436796e-02
-1.63808048e-01 1.41958192e-01 -5.01323760e-01 5.60725272e-01
9.81620967e-01 -1.60340786e-01 7.03165084e-02 -2.17571616e-01
-3.76949489e-01 4.74852294e-01 -1.49656570e+00 1.49562907e+00
-3.19152504e-01 4.32763308e-01 4.13043685e-02 -8.25505495e-01
1.18472791e+00 5.22780698e-03 4.47455466e-01 -3.08722913e-01
1.68698043e-01 3.90715301e-01 -2.33593345e-01 -2.03403886e-02
3.97057444e-01 1.86061487e-01 8.38288199e-03 5.33721387e-01
-5.23074120e-02 -8.09713125e-01 4.22863849e-02 -6.65370449e-02
1.11344111e+00 2.99540967e-01 3.07385147e-01 6.99813887e-02
3.26711744e-01 3.31421494e-01 1.68561265e-01 8.02251697e-01
6.13827482e-02 6.64794564e-01 2.14107424e-01 -7.01577902e-01
-1.22215843e+00 -1.19389856e+00 -2.70597965e-01 4.70829427e-01
2.55639046e-01 -3.66327286e-01 -6.95419431e-01 -5.87889612e-01
2.12074280e-01 3.36021185e-01 -3.86164665e-01 3.14911157e-02
-7.30166137e-01 -5.71257532e-01 2.84731500e-02 2.78980076e-01
2.67671913e-01 -4.80158418e-01 -1.18385029e+00 2.59582341e-01
3.48667800e-01 -1.17178917e+00 -3.76348823e-01 2.92856663e-01
-1.19093096e+00 -1.22426295e+00 -3.58640105e-01 -6.94357455e-01
9.59842026e-01 7.96953380e-01 1.06944966e+00 3.86338420e-02
-7.44231641e-01 4.13719326e-01 4.50607799e-02 -4.29790109e-01
-3.81905854e-01 -2.85254836e-01 1.85526282e-01 -1.06413350e-01
5.43337584e-01 -4.61320668e-01 -1.88976645e-01 7.07184613e-01
-4.57800984e-01 1.36582822e-01 6.22139335e-01 7.98738182e-01
9.83463764e-01 -1.65821403e-01 -4.18325961e-01 -4.53467280e-01
1.28286973e-01 2.35588208e-01 -1.42116833e+00 6.94137290e-02
-1.61818475e-01 -2.11256021e-03 1.06414169e-01 -5.36788523e-01
-7.00716972e-01 8.45176339e-01 8.52394029e-02 -8.95160794e-01
-1.56243324e-01 -1.94515765e-01 7.60258362e-02 -4.18028057e-01
9.74155009e-01 3.26930493e-01 9.68968123e-02 -5.57694018e-01
2.73797244e-01 4.78836715e-01 5.54192543e-01 -5.44911444e-01
1.21542525e+00 7.74509430e-01 3.10871094e-01 -9.50800598e-01
-7.44234383e-01 -6.83976293e-01 -9.41094816e-01 -3.57236981e-01
7.53135860e-01 -7.10525274e-01 -6.37435913e-01 4.03526425e-01
-1.46591973e+00 1.19014040e-01 -2.02145666e-01 6.98130429e-01
-6.53711796e-01 2.66625613e-01 -2.45077014e-01 -9.00038183e-01
-7.88661186e-03 -1.22434020e+00 1.45086789e+00 -1.76535413e-01
-2.12816924e-01 -4.29206252e-01 1.11927278e-01 8.66290703e-02
8.99561271e-02 1.90087706e-01 6.95700526e-01 -3.14601541e-01
-1.05498850e+00 -7.75370121e-01 -2.87198365e-01 -4.04090583e-02
-1.63637176e-01 1.02471016e-01 -9.59100723e-01 -1.63675785e-01
3.17236125e-01 5.45608588e-02 5.07619977e-01 3.09311301e-01
8.39004874e-01 1.20119646e-01 -6.98610306e-01 5.82834661e-01
1.16122127e+00 1.02082416e-01 4.44119722e-02 1.33398578e-01
3.60835850e-01 3.20702583e-01 7.57176042e-01 5.01136959e-01
2.52949651e-02 1.05915892e+00 4.01843518e-01 -6.52382299e-02
-2.09629625e-01 -2.00597256e-01 1.52931541e-01 5.97286344e-01
-5.24836406e-02 2.22974956e-01 -7.75046527e-01 3.27348471e-01
-1.58119214e+00 -7.78412879e-01 -3.16597760e-01 2.52581549e+00
3.69720072e-01 3.45079094e-01 1.66336060e-01 7.94101804e-02
6.26716435e-01 -1.71603158e-01 -6.14077866e-01 -1.57981783e-01
1.32053673e-01 3.80594879e-01 5.60074091e-01 4.82643515e-01
-1.14224422e+00 7.32583821e-01 6.58796549e+00 6.18594468e-01
-1.09215295e+00 5.13142236e-02 8.05541873e-02 -1.78776607e-01
1.61354184e-01 3.84255014e-02 -1.23312163e+00 -7.86697716e-02
4.27550912e-01 4.95225079e-02 2.32187524e-01 1.12632811e+00
-1.98849011e-02 -2.23521873e-01 -1.24608278e+00 1.12994456e+00
1.26991183e-01 -1.39071143e+00 -1.51749969e-01 3.02587807e-01
4.62442636e-01 1.31299719e-01 -1.68945007e-02 -1.77328095e-01
1.28962725e-01 -4.55815613e-01 9.14006948e-01 9.83187109e-02
4.80945170e-01 -5.03396332e-01 2.68269390e-01 7.45194077e-01
-1.23453963e+00 2.71337152e-01 -6.57897472e-01 -1.13913536e-01
-8.65631402e-02 6.40263855e-01 -1.13499939e+00 1.34984136e-01
5.33045948e-01 2.09209934e-01 -5.23059785e-01 1.14910138e+00
-8.42395648e-02 1.30169079e-01 -7.81056046e-01 -1.99966982e-01
2.14277077e-02 -3.67859066e-01 7.72668421e-01 1.07962608e+00
5.55579782e-01 1.37142405e-01 3.45164031e-01 8.50701928e-01
2.63661504e-01 -9.31934416e-02 -7.82865345e-01 3.69591415e-01
5.56652904e-01 1.24850726e+00 -1.06978750e+00 -2.66970575e-01
-3.83053958e-01 9.13833678e-01 2.21769586e-01 -7.03807697e-02
-7.44875968e-01 -2.44769067e-01 5.11486411e-01 2.70950109e-01
7.26647675e-01 -8.22850168e-01 -2.60535747e-01 -1.04551888e+00
2.91830927e-01 -6.70555830e-01 1.03581257e-01 -6.60610497e-01
-9.01044607e-01 5.94879389e-01 7.24435821e-02 -1.51182675e+00
-2.92499572e-01 -8.28133404e-01 -3.06855291e-01 8.36526215e-01
-1.04880643e+00 -9.38733459e-01 -1.96411565e-01 4.56403017e-01
5.09228408e-01 8.80037248e-02 8.67061853e-01 -1.72291119e-02
-7.11519271e-02 1.58426955e-01 -7.24021271e-02 -1.19708449e-01
4.10265148e-01 -8.84441257e-01 8.64606559e-01 7.14017570e-01
6.27857804e-01 3.57159883e-01 6.17343068e-01 -5.08251250e-01
-1.74631178e+00 -8.27542722e-01 9.57466602e-01 -8.03811133e-01
3.77593637e-01 -7.51383483e-01 -5.92561662e-01 5.15683532e-01
-3.04993391e-01 2.39245519e-01 2.67386556e-01 -1.29299477e-01
-3.67463112e-01 3.05082407e-02 -1.03718114e+00 4.18611139e-01
1.24118149e+00 -4.61604595e-01 -5.06645143e-01 6.86074674e-01
1.61269456e-01 -9.78974462e-01 -4.18171465e-01 2.87723839e-01
6.10161006e-01 -1.06726420e+00 1.26264071e+00 -1.79470673e-01
-2.63731539e-01 -6.50842607e-01 -2.39070728e-01 -8.19251657e-01
-2.78512180e-01 -4.36400086e-01 -1.31844282e-01 5.38824737e-01
2.13850755e-02 -4.41068590e-01 8.22898030e-01 2.68489093e-01
-8.41702297e-02 -6.97317421e-01 -1.11059773e+00 -9.62236166e-01
-5.12112677e-01 -5.60405374e-01 3.58692706e-01 4.13585603e-01
-5.51577806e-01 2.29368329e-01 -1.07254265e-02 6.02426350e-01
9.94635522e-01 6.74207151e-01 1.29270399e+00 -1.27892435e+00
-4.50836450e-01 -2.83262730e-01 -8.31227541e-01 -1.48964274e+00
-1.63888752e-01 -8.22329521e-01 -1.35744601e-01 -1.01547968e+00
8.28663930e-02 -4.39157605e-01 4.18902129e-01 2.89193630e-01
3.30489188e-01 4.54413176e-01 1.13690674e-01 2.63744533e-01
-3.37842882e-01 1.76371053e-01 1.06782150e+00 1.12838537e-01
-8.22795779e-02 2.06550077e-01 1.28672823e-01 9.47154999e-01
3.00929099e-01 -5.62245786e-01 -2.72440799e-02 -4.65655208e-01
-1.26982450e-01 1.43275872e-01 6.40755951e-01 -1.10265851e+00
1.36696935e-01 2.61617918e-03 7.97154546e-01 -1.23156476e+00
8.29248667e-01 -7.67730832e-01 1.99482694e-01 6.62095010e-01
1.47450149e-01 1.76802903e-01 1.34924218e-01 3.43230575e-01
9.66512263e-02 -2.95193374e-01 8.60223770e-01 -2.51523435e-01
-4.23846841e-01 1.82360560e-01 -4.68513310e-01 -2.43151873e-01
1.05940104e+00 -4.64905113e-01 1.96610585e-01 -2.83352453e-02
-6.45233989e-01 -2.59444088e-01 8.96114528e-01 2.07658142e-01
7.25212395e-01 -1.37653840e+00 -5.92316329e-01 7.36043930e-01
-8.37055296e-02 4.45750654e-02 -2.14298502e-01 6.94355369e-01
-7.03101099e-01 4.78066862e-01 -6.16149092e-03 -1.22641826e+00
-1.53944409e+00 6.76578104e-01 3.66986036e-01 -9.36551094e-02
-6.44865453e-01 8.82935286e-01 2.48534441e-01 -4.07823503e-01
1.75552234e-01 -3.33671033e-01 4.61013913e-01 -1.76041558e-01
4.64493394e-01 3.83994788e-01 4.65132356e-01 -5.87378979e-01
-5.27474642e-01 9.84340072e-01 -3.76565121e-02 -2.63809800e-01
1.14406192e+00 2.86553621e-01 2.01242521e-01 -7.27993995e-03
1.14334619e+00 1.37881786e-01 -1.38444090e+00 -2.81674296e-01
-3.72205861e-02 -8.81280243e-01 5.30331992e-02 -2.19109088e-01
-6.89586997e-01 9.19620991e-01 7.05017686e-01 1.01620898e-01
8.68072152e-01 3.00825089e-01 4.14173126e-01 5.11249781e-01
6.81572318e-01 -7.73416758e-01 1.15149423e-01 5.70115209e-01
7.58209825e-01 -8.30296338e-01 2.29118124e-01 -7.65526772e-01
1.11421973e-01 1.36259317e+00 4.25882578e-01 -4.43536550e-01
4.30227757e-01 4.76140738e-01 3.72290537e-02 -4.12725359e-01
-5.66217661e-01 -2.28995830e-01 5.14257908e-01 5.30938268e-01
-6.70306683e-02 -3.36639322e-02 5.55201881e-02 -3.67115103e-02
-1.70944467e-01 -3.13314438e-01 2.16611683e-01 1.07509255e+00
-4.83254701e-01 -1.28391945e+00 -7.89351225e-01 2.12172776e-01
1.10368207e-01 2.21746981e-01 -5.40276587e-01 8.64952564e-01
1.36891864e-02 5.04984260e-01 2.19215393e-01 -3.04829717e-01
4.70046490e-01 5.83376922e-02 1.08201110e+00 -7.98811972e-01
-1.39682055e-01 4.86043930e-01 -1.78023890e-01 -7.02505708e-01
-3.18948418e-01 -1.00886786e+00 -9.53590214e-01 -2.53520906e-02
-6.42722070e-01 -1.21454366e-01 9.16441381e-01 7.52352893e-01
4.76765484e-01 -7.66097307e-02 8.54803503e-01 -1.39060879e+00
-7.88504601e-01 -4.45485771e-01 -3.57251823e-01 7.41905198e-02
6.12044036e-02 -9.44329321e-01 -3.17647725e-01 -1.43634796e-01]
|
[7.732873439788818, -2.658701181411743]
|
7bf0551f-952e-486f-87f2-0f3d7df53e11
|
effseg-efficient-fine-grained-instance
|
2307.01545
| null |
https://arxiv.org/abs/2307.01545v1
|
https://arxiv.org/pdf/2307.01545v1.pdf
|
EffSeg: Efficient Fine-Grained Instance Segmentation using Structure-Preserving Sparsity
|
Many two-stage instance segmentation heads predict a coarse 28x28 mask per instance, which is insufficient to capture the fine-grained details of many objects. To address this issue, PointRend and RefineMask predict a 112x112 segmentation mask resulting in higher quality segmentations. Both methods however have limitations by either not having access to neighboring features (PointRend) or by performing computation at all spatial locations instead of sparsely (RefineMask). In this work, we propose EffSeg performing fine-grained instance segmentation in an efficient way by using our Structure-Preserving Sparsity (SPS) method based on separately storing the active features, the passive features and a dense 2D index map containing the feature indices. The goal of the index map is to preserve the 2D spatial configuration or structure between the features such that any 2D operation can still be performed. EffSeg achieves similar performance on COCO compared to RefineMask, while reducing the number of FLOPs by 71% and increasing the FPS by 29%. Code will be released.
|
['Tinne Tuytelaars', 'Cédric Picron']
|
2023-07-04
| null | null | null | null |
['instance-segmentation']
|
['computer-vision']
|
[ 2.59255737e-01 1.18221782e-01 2.61419248e-02 -3.44957441e-01
-7.41531968e-01 -4.49326277e-01 2.44259670e-01 3.75936002e-01
-2.67190188e-01 5.15496850e-01 -1.01027958e-01 6.40875846e-02
-2.09593847e-02 -9.51860964e-01 -6.50595069e-01 -6.00921929e-01
-1.14333987e-01 4.55474973e-01 9.24586952e-01 3.36603552e-01
6.27562881e-01 8.05999756e-01 -1.75891769e+00 5.31394660e-01
8.42259586e-01 1.26603258e+00 3.85935903e-01 6.47601664e-01
-2.20950112e-01 2.99225360e-01 -6.34720802e-01 -1.26666978e-01
7.07222044e-01 -1.02592878e-01 -7.46548533e-01 1.72038674e-01
8.88998866e-01 -3.43642056e-01 -2.03666501e-02 7.32787430e-01
2.06195399e-01 2.71372527e-01 3.92163068e-01 -9.22862172e-01
2.42719099e-01 2.77505189e-01 -9.24105644e-01 1.50860935e-01
-2.74825022e-02 1.32384762e-01 8.89823556e-01 -9.68232036e-01
5.08377194e-01 9.98381853e-01 6.17445052e-01 5.32740653e-02
-1.14842856e+00 -5.79311788e-01 -3.23073789e-02 2.42733993e-02
-1.58333611e+00 -3.82704556e-01 5.32243133e-01 -1.63369760e-01
1.01349938e+00 6.99314237e-01 8.34801018e-01 -1.07805662e-01
2.86376715e-01 4.77076352e-01 1.04142153e+00 -2.58577317e-01
3.42498124e-01 -2.46228933e-01 1.97066873e-01 8.84683788e-01
1.51538238e-01 -8.42478648e-02 -7.51328528e-01 -1.64058983e-01
1.32481337e+00 1.29583701e-01 -1.42087907e-01 -2.55738556e-01
-1.25091743e+00 6.63587928e-01 5.22185028e-01 2.35736385e-01
-4.21703696e-01 1.22441620e-01 1.25905901e-01 4.97170836e-02
4.48617011e-01 5.89073062e-01 -4.68933821e-01 -2.86496431e-01
-1.71728826e+00 3.08858722e-01 6.17323399e-01 8.69401753e-01
1.51494801e+00 -2.26986259e-01 -3.57116044e-01 5.98528326e-01
-1.69783980e-02 3.90970170e-01 2.17000112e-01 -1.40140295e+00
4.78252769e-01 9.29810882e-01 -4.24676836e-02 -1.10900164e+00
-2.76018262e-01 -2.99461573e-01 -7.54425347e-01 2.43752629e-01
3.64572853e-01 4.01144922e-02 -1.32387662e+00 8.57125461e-01
6.19047344e-01 5.03096461e-01 -4.11773831e-01 7.55300939e-01
7.65179217e-01 1.03366315e+00 -3.04781705e-01 1.95605382e-02
1.21323144e+00 -1.33287895e+00 -3.54389012e-01 -2.42433473e-01
5.75488508e-01 -9.78142381e-01 8.76530647e-01 3.22291940e-01
-1.28964972e+00 -5.14969349e-01 -9.93079901e-01 -4.39900130e-01
-1.40128031e-01 3.83096039e-02 5.70042849e-01 3.89193326e-01
-1.01073349e+00 6.24715865e-01 -1.23679984e+00 2.63116241e-01
5.96175194e-01 7.32813179e-01 -3.38966876e-01 -9.33954641e-02
-4.69618052e-01 3.69368106e-01 3.61101776e-01 -9.37321708e-02
-5.18296778e-01 -1.19942355e+00 -1.01369512e+00 5.01622438e-01
3.86983782e-01 -2.25527599e-01 8.32517684e-01 -5.50677359e-01
-1.16243696e+00 5.75922906e-01 -4.30638194e-01 -4.26227868e-01
3.37045759e-01 -1.81917027e-01 2.40518242e-01 3.70165795e-01
3.47177237e-01 9.98412371e-01 7.82992423e-01 -1.03413868e+00
-8.77740741e-01 -4.53507036e-01 -1.69794694e-01 3.27782750e-01
-1.30950719e-01 -3.60615551e-01 -1.03403330e+00 -6.30705953e-01
4.31334019e-01 -8.79129589e-01 -5.58591843e-01 6.92509338e-02
-4.86930162e-01 1.70032308e-01 1.20785689e+00 -5.73002100e-01
1.36284864e+00 -2.45675349e+00 -2.48631895e-01 5.50024450e-01
3.46899331e-01 4.17991132e-01 6.78600594e-02 1.60040095e-01
3.42796624e-01 -1.62828807e-03 -4.93153840e-01 -4.63671893e-01
-4.01838571e-01 3.72546971e-01 -7.67542869e-02 3.91941726e-01
1.35734186e-01 9.47059631e-01 -5.03673434e-01 -7.19883978e-01
5.51322699e-01 5.61029971e-01 -8.70720744e-01 1.75733268e-02
-8.42862763e-03 2.99169868e-01 -5.11351883e-01 4.94735658e-01
1.11505008e+00 -3.26409727e-01 -9.98470485e-02 -3.45040947e-01
-5.22870600e-01 2.30400294e-01 -1.49214554e+00 1.65409470e+00
-2.47962683e-01 5.13694823e-01 3.52397233e-01 -7.34520435e-01
9.69878554e-01 -9.64485854e-02 7.32989430e-01 -7.00593829e-01
-2.10111678e-01 1.88163310e-01 -2.92090863e-01 2.77859509e-01
7.03687608e-01 7.89005607e-02 -1.28233582e-01 2.92530417e-01
-3.00345302e-01 -4.30816948e-01 2.22584218e-01 1.35289147e-01
1.10668564e+00 -2.02577978e-01 1.78689405e-01 -5.36874175e-01
2.95561284e-01 3.45299542e-01 8.36360872e-01 7.07789898e-01
2.29989424e-01 1.07998884e+00 3.90455008e-01 -4.58312601e-01
-9.49453413e-01 -9.05827880e-01 -2.55463600e-01 6.50798917e-01
4.68694836e-01 -7.19500244e-01 -8.62406731e-01 -5.47292531e-01
-4.09383699e-02 4.30915982e-01 -4.72935110e-01 2.95221299e-01
-1.01661003e+00 -4.32012260e-01 9.10417438e-02 5.27266383e-01
7.86585450e-01 -8.54791403e-01 -1.14108956e+00 2.72986501e-01
1.02023356e-01 -9.07287836e-01 -7.46828735e-01 3.93838823e-01
-1.13548779e+00 -9.21972811e-01 -5.70045590e-01 -7.26750016e-01
8.96374166e-01 2.86276311e-01 1.06015122e+00 2.95378894e-01
-4.36955094e-01 -1.74367785e-01 -2.95549214e-01 3.07518169e-02
1.87725499e-01 2.65790820e-01 -5.77466249e-01 -1.85311109e-01
-4.87503596e-02 -5.30066133e-01 -7.32892692e-01 4.58792508e-01
-1.01886511e+00 4.37648743e-01 5.85022807e-01 6.12282217e-01
1.29201412e+00 2.84694254e-01 -3.10876351e-02 -1.08061683e+00
-1.55639157e-01 -1.23210307e-02 -7.21454203e-01 4.82841916e-02
-3.33476990e-01 -1.90510489e-02 4.56995100e-01 -1.17727369e-01
-9.64768469e-01 4.15358633e-01 -2.04747066e-01 -4.29793447e-01
-1.09054230e-01 1.03307277e-01 -8.04179534e-02 -1.74662217e-01
1.22003451e-01 3.19802463e-01 -2.47838385e-02 -7.39305854e-01
9.63176638e-02 3.05291504e-01 4.04228270e-01 -5.10441899e-01
5.70579588e-01 6.85956180e-01 1.62669808e-01 -9.21557724e-01
-8.15967858e-01 -6.30566835e-01 -7.96648622e-01 2.09012598e-01
7.49423087e-01 -7.41693497e-01 -3.55762988e-01 4.23669666e-01
-7.85122216e-01 -4.58243221e-01 -7.23647714e-01 1.33329228e-01
-4.59868699e-01 3.27525109e-01 -6.81617379e-01 -2.31072903e-01
-3.26919138e-01 -1.29277909e+00 1.43234825e+00 3.00278068e-01
-1.06651209e-01 -5.07738709e-01 -2.38376722e-01 3.43455672e-01
3.74726206e-01 2.22846314e-01 7.43329406e-01 -2.46192992e-01
-1.10074747e+00 -2.34845039e-02 -3.75599593e-01 5.98287582e-03
1.01343371e-01 -1.08915903e-02 -5.23512125e-01 -2.48222888e-01
-1.00130402e-01 1.24257401e-01 9.03041542e-01 7.11944580e-01
1.29030764e+00 -2.13653013e-01 -5.72141767e-01 8.09025049e-01
1.44118762e+00 6.69894367e-02 8.16627383e-01 4.23473455e-02
8.14150035e-01 3.85356486e-01 8.84177089e-01 5.73979259e-01
2.25167751e-01 7.88396239e-01 2.92379349e-01 -4.30217326e-01
-4.17044878e-01 -1.47595257e-01 -1.06498681e-01 6.09309196e-01
1.46146238e-01 -2.40729544e-02 -8.61796439e-01 7.32197642e-01
-1.69561505e+00 -4.85141367e-01 -3.85366827e-01 2.25566626e+00
6.95492625e-01 6.89893961e-02 -1.04816861e-01 1.69623792e-01
6.78272069e-01 1.62520975e-01 -5.36210239e-01 -4.14600581e-01
1.69061393e-01 6.40932322e-01 7.13120699e-01 6.60941720e-01
-1.03331554e+00 1.03519332e+00 5.77077579e+00 1.19296837e+00
-1.24415743e+00 -6.72001243e-02 9.33192790e-01 -3.53635401e-01
-1.23904780e-01 1.07179269e-01 -1.15987837e+00 5.51978707e-01
4.47622418e-01 3.54209542e-01 8.33160058e-03 7.28996396e-01
8.03650320e-02 -6.34311080e-01 -6.80790782e-01 9.95845973e-01
-2.37510443e-01 -1.72886169e+00 2.33532172e-02 2.75610209e-01
9.32398796e-01 -4.78288420e-02 -3.30579318e-02 -1.09285094e-01
-1.57137901e-01 -1.01976395e+00 6.12067103e-01 2.83040017e-01
9.34569955e-01 -8.38598371e-01 4.37741548e-01 4.05317158e-01
-1.45552337e+00 2.33918726e-01 -3.80837232e-01 -7.21870884e-02
1.93985373e-01 1.11121714e+00 -7.99834609e-01 3.09200823e-01
9.02597487e-01 5.80032110e-01 -5.48673809e-01 1.18980956e+00
2.86485106e-01 5.82403123e-01 -8.67310166e-01 3.28004569e-01
4.19946194e-01 -6.00454099e-02 3.62868875e-01 1.08893311e+00
3.48497003e-01 2.54087746e-01 4.34040815e-01 8.01840603e-01
2.25156367e-01 3.22377160e-02 -2.05345862e-02 2.73408145e-01
5.50135076e-01 1.21808028e+00 -1.36602557e+00 -4.44552690e-01
-3.76137197e-01 1.10219252e+00 1.77396432e-01 -4.45104726e-02
-5.31005859e-01 -4.09063935e-01 6.72174335e-01 5.42778671e-01
9.27373290e-01 -4.82865810e-01 -7.30933666e-01 -9.45077360e-01
1.31737724e-01 -5.41006923e-01 1.35072365e-01 -9.79044735e-02
-7.20067799e-01 4.28455710e-01 -6.23122677e-02 -9.93511736e-01
-7.70520279e-03 -2.35085890e-01 -4.29038167e-01 8.67597938e-01
-1.41294813e+00 -1.05479610e+00 -2.83200085e-01 5.61385751e-01
7.03270912e-01 4.78653252e-01 5.08740664e-01 2.16066018e-01
-3.78770411e-01 4.33520943e-01 7.45310411e-02 -5.56722768e-02
3.35936159e-01 -1.10942209e+00 4.16556567e-01 6.47256315e-01
1.34597018e-01 4.76607144e-01 3.22907656e-01 -7.99974859e-01
-9.65680659e-01 -1.09829450e+00 8.89569581e-01 -1.17726056e-02
-1.42588183e-01 -3.06017429e-01 -9.91990864e-01 4.00686741e-01
-1.33836180e-01 4.49008346e-01 4.37621772e-01 -2.19004020e-01
1.01111636e-01 -2.17767417e-01 -1.41602755e+00 2.86673605e-01
8.73293281e-01 8.12429115e-02 -1.90987378e-01 -1.07586183e-01
5.68456054e-01 -7.29281485e-01 -1.03486550e+00 4.12992090e-01
4.08114970e-01 -1.29501057e+00 9.81572688e-01 2.19033748e-01
3.04813504e-01 -5.88756382e-01 -1.41327247e-01 -8.14669251e-01
-3.57756346e-01 -4.64204222e-01 -1.23909190e-01 1.04050291e+00
3.89657557e-01 -4.81233031e-01 1.29106009e+00 7.22625375e-01
-3.00841033e-01 -1.18109179e+00 -1.08977306e+00 -3.69564533e-01
-2.61500567e-01 -2.96896726e-01 8.67579520e-01 6.35183930e-01
-4.82242733e-01 -9.03463662e-02 3.21248919e-02 3.03797930e-01
5.23590863e-01 6.55705452e-01 6.20559216e-01 -1.02633297e+00
-1.70318589e-01 -1.40654191e-01 -4.78405178e-01 -1.46251047e+00
-2.75326103e-01 -5.05523026e-01 9.39885434e-03 -1.40945804e+00
1.56099871e-01 -1.00279605e+00 7.67898187e-02 6.99601710e-01
-1.44309759e-01 8.28182757e-01 2.46562332e-01 2.72342682e-01
-5.69051504e-01 2.08938599e-01 1.54468405e+00 1.17355436e-01
-3.69124383e-01 -1.12284511e-01 -5.53731143e-01 5.78995705e-01
5.95869243e-01 -4.49736059e-01 -2.71109492e-01 -5.03765583e-01
-3.27741742e-01 8.12530518e-02 2.31150001e-01 -1.26011014e+00
2.60720193e-01 -8.09801295e-02 6.97247565e-01 -1.11077797e+00
8.21986377e-01 -7.72938609e-01 2.35216513e-01 3.75077397e-01
5.29144853e-02 -6.41721264e-02 2.93314546e-01 2.23527342e-01
-4.23252285e-01 -2.92949826e-01 8.81117284e-01 -2.50431329e-01
-8.02463591e-01 4.72777098e-01 -9.27564129e-02 -7.52264634e-02
1.20768034e+00 -7.65218973e-01 9.01124626e-02 1.34181995e-02
-7.74331570e-01 1.71464548e-01 8.78948450e-01 -1.70904964e-01
7.01621890e-01 -9.31426942e-01 -5.23976266e-01 6.76130235e-01
-4.85715181e-01 8.28745067e-01 5.68577409e-01 6.39422178e-01
-9.62407887e-01 5.21704912e-01 -1.26296952e-01 -7.82819390e-01
-1.42953753e+00 9.40995514e-02 6.35595433e-03 -5.52834213e-01
-1.05762005e+00 1.03860414e+00 4.92100865e-01 -2.02575013e-01
-1.22100249e-01 -3.45757812e-01 3.68822701e-02 -9.48049873e-02
6.34604394e-01 6.12242639e-01 3.10335308e-01 -6.37770772e-01
-3.78365129e-01 9.25212443e-01 -3.15160483e-01 3.80821414e-02
1.40713155e+00 -5.68275005e-02 -2.90673643e-01 -1.24993317e-01
1.16333973e+00 3.61865669e-01 -1.69086552e+00 -4.84541850e-03
-1.84062243e-01 -9.04039979e-01 4.49129879e-01 -5.98445594e-01
-1.47876799e+00 7.20904171e-01 4.58106339e-01 1.40431657e-01
1.24061084e+00 1.02978893e-01 1.24550450e+00 -9.76665914e-02
4.71470058e-01 -9.74892616e-01 -1.98881775e-01 3.74023020e-01
4.81759340e-01 -7.49268532e-01 2.93979049e-01 -1.11910915e+00
-4.65971231e-01 8.64504099e-01 7.19112635e-01 -4.30089116e-01
7.55752206e-01 6.56480432e-01 -1.64856985e-01 -3.84123415e-01
-4.07630563e-01 -5.70290387e-02 4.14824873e-01 3.28699023e-01
1.59438968e-01 9.25616175e-02 -2.33695835e-01 2.22789213e-01
-3.42295766e-01 -3.05273741e-01 1.93525374e-01 1.04681742e+00
-5.26258171e-01 -1.11011314e+00 -4.49962080e-01 8.68477106e-01
-3.76304418e-01 -1.58509195e-01 -1.27840146e-01 6.00015163e-01
2.74806947e-01 7.58884609e-01 6.62195504e-01 -4.20632325e-02
2.20844120e-01 -3.83568972e-01 3.11942756e-01 -1.00398624e+00
-6.99723125e-01 3.10510010e-01 -1.89545199e-01 -9.34882820e-01
-1.57166958e-01 -6.82477713e-01 -1.54090059e+00 -2.05585018e-01
-3.06280285e-01 2.03248352e-01 4.80384409e-01 7.41283238e-01
7.90619135e-01 3.38642001e-01 4.07125533e-01 -1.06295514e+00
8.55006427e-02 -5.73659480e-01 -7.70410717e-01 5.34284636e-02
2.52135575e-01 -5.01767755e-01 -7.91785195e-02 -4.86054830e-02]
|
[9.41862678527832, -0.05342837795615196]
|
15f5e3ad-568d-49ca-9191-a6336319c251
|
multi-view-imputation-and-cross-attention
|
2206.08019
| null |
https://arxiv.org/abs/2206.08019v2
|
https://arxiv.org/pdf/2206.08019v2.pdf
|
Multi-View Imputation and Cross-Attention Network Based on Incomplete Longitudinal and Multimodal Data for Conversion Prediction of Mild Cognitive Impairment
|
Predicting whether subjects with mild cognitive impairment (MCI) will convert to Alzheimer's disease is a significant clinical challenge. Longitudinal variations and complementary information inherent in longitudinal and multimodal data are crucial for MCI conversion prediction, but persistent issue of missing data in these data may hinder their effective application. Additionally, conversion prediction should be achieved in the early stages of disease progression in clinical practice, specifically at baseline visit (BL). Therefore, longitudinal data should only be incorporated during training to capture disease progression information. To address these challenges, a multi-view imputation and cross-attention network (MCNet) was proposed to integrate data imputation and MCI conversion prediction in a unified framework. First, a multi-view imputation method combined with adversarial learning was presented to handle various missing data scenarios and reduce imputation errors. Second, two cross-attention blocks were introduced to exploit the potential associations in longitudinal and multimodal data. Finally, a multi-task learning model was established for data imputation, longitudinal classification, and conversion prediction tasks. When the model was appropriately trained, the disease progression information learned from longitudinal data can be leveraged by BL data to improve MCI conversion prediction at BL. MCNet was tested on two independent testing sets and single-modal BL data to verify its effectiveness and flexibility in MCI conversion prediction. Results showed that MCNet outperformed several competitive methods. Moreover, the interpretability of MCNet was demonstrated. Thus, our MCNet may be a valuable tool in longitudinal and multimodal data analysis for MCI conversion prediction. Codes are available at https://github.com/Meiyan88/MCNET.
|
['Xiaoling Zhang', 'Xiumei Chen', 'Qianjin Feng', 'Shuoling Zhou', 'Tao Wang', 'Meiyan Huang']
|
2022-06-16
| null | null | null | null |
['disease-prediction']
|
['medical']
|
[ 3.30171958e-02 -3.59612197e-01 -4.65205908e-01 -7.49572575e-01
-1.09730816e+00 -1.67303368e-01 2.65976101e-01 -1.72343701e-01
-2.54576832e-01 9.83901203e-01 6.30097389e-01 -2.14623213e-01
-1.58809289e-01 -6.92251384e-01 -6.68729424e-01 -4.33511585e-01
1.85114052e-02 4.30355757e-01 -1.89669460e-01 -1.22265913e-01
-2.66306698e-01 1.80657618e-02 -1.06903291e+00 7.22861886e-01
1.37754703e+00 8.58038068e-01 2.64446288e-01 3.32097739e-01
6.73923418e-02 6.01933122e-01 -1.73851207e-01 -3.98956120e-01
7.43726268e-02 -2.20935225e-01 -4.74190503e-01 -1.84073985e-01
2.14545429e-01 -6.89662576e-01 -3.01096857e-01 7.06937790e-01
8.66215825e-01 -1.99866712e-01 5.87880611e-01 -1.31497753e+00
-7.67856240e-01 3.46462339e-01 -4.83404875e-01 1.78940982e-01
3.10991585e-01 3.94352764e-01 6.71375275e-01 -1.01666224e+00
4.76548135e-01 8.95487607e-01 1.02744544e+00 6.48348510e-01
-1.37433755e+00 -8.05100024e-01 2.80548245e-01 5.21677494e-01
-8.39711189e-01 -4.26891595e-01 6.86069191e-01 -8.43882859e-01
4.56212908e-01 2.29117945e-01 8.04099798e-01 1.49749076e+00
5.42132497e-01 7.23198473e-01 1.26948607e+00 1.56358153e-01
2.31799409e-02 -1.80513546e-01 3.97436708e-01 3.83732796e-01
-2.52057873e-02 2.32935011e-01 -5.66415228e-02 -2.90080994e-01
6.09081566e-01 5.81427157e-01 -2.85730124e-01 5.29233459e-03
-1.38531530e+00 8.50149632e-01 6.60976529e-01 -3.34886648e-02
-4.47317034e-01 -1.97838604e-01 6.00376546e-01 2.18478128e-01
4.57075715e-01 -2.19111398e-01 -4.82740611e-01 1.53780952e-01
-8.71870100e-01 9.39856097e-02 2.63207674e-01 7.87086368e-01
1.27421260e-01 -5.12262806e-02 -3.87280315e-01 1.10785484e+00
3.96831781e-01 5.65067589e-01 7.08278716e-01 -9.67727602e-01
9.25718486e-01 9.36878204e-01 -1.70511782e-01 -5.86616993e-01
-6.46539330e-01 -8.19414914e-01 -1.37027597e+00 2.30380416e-01
3.01320940e-01 -2.68791348e-01 -8.59181404e-01 1.85365093e+00
7.20322737e-03 -5.83095141e-02 -5.05058840e-02 9.38813269e-01
7.24418879e-01 2.38809824e-01 4.23718899e-01 -1.33439645e-01
1.51228833e+00 -5.81716120e-01 -6.42378747e-01 -3.11440945e-01
5.83869874e-01 -4.47849602e-01 1.20661736e+00 5.32233305e-02
-1.14443815e+00 -7.25312352e-01 -9.01650906e-01 -2.77674347e-01
-5.90529665e-02 4.63025838e-01 5.78352511e-01 3.66704732e-01
-5.70617557e-01 1.59683019e-01 -1.11709797e+00 -1.03846192e-01
8.89240682e-01 4.83844966e-01 -3.48057747e-01 -4.47530836e-01
-1.36867583e+00 7.00269222e-01 1.68564126e-01 5.61843812e-01
-6.84009671e-01 -1.17703629e+00 -5.44523656e-01 -1.75567612e-01
-2.77314186e-01 -1.39676988e+00 9.56027210e-01 -9.28143263e-01
-8.49295497e-01 5.27015805e-01 -2.99374700e-01 -2.95147985e-01
1.00682116e+00 -2.86248595e-01 -6.53968334e-01 -1.16330713e-01
4.17324990e-01 7.22760737e-01 3.69494736e-01 -1.08678365e+00
-5.10029316e-01 -1.02655613e+00 -3.89529854e-01 1.53111413e-01
-2.17811167e-02 -1.45120308e-01 2.38522235e-02 -7.49269068e-01
1.69543084e-02 -7.86984265e-01 -2.39618987e-01 7.78209120e-02
-4.31424916e-01 8.43151584e-02 4.39867228e-01 -1.30998445e+00
1.04823816e+00 -1.85792184e+00 6.69998899e-02 -8.96553174e-02
3.83865684e-01 8.38087499e-02 -2.13440750e-02 -1.49737466e-02
-2.76206136e-01 1.96396306e-01 -3.73302877e-01 -3.98485214e-01
-1.60672843e-01 -9.91909429e-02 -1.75275840e-02 3.04182917e-01
2.10837826e-01 1.30519319e+00 -3.97315800e-01 -2.35035092e-01
6.40681088e-02 6.78079545e-01 -7.80522048e-01 2.19056830e-01
-2.96850763e-02 1.08284616e+00 -4.81233150e-01 8.27749550e-01
8.44560564e-01 -2.71125019e-01 9.34726670e-02 -4.91823286e-01
1.18952371e-01 -1.14677191e-01 -5.13384402e-01 1.48404503e+00
-4.69532967e-01 2.47135624e-01 -1.04779139e-01 -9.37636614e-01
6.75759494e-01 3.35394979e-01 5.47972143e-01 -9.14698243e-01
7.16844946e-02 2.16459334e-01 1.63485378e-01 -6.90843940e-01
-3.45018178e-01 -2.08653733e-01 6.43220395e-02 2.62538612e-01
-4.05603051e-01 8.75589550e-01 -1.28857628e-01 -1.75345197e-01
8.50705087e-01 1.70110270e-01 3.33897881e-02 1.71495199e-01
6.05690002e-01 8.56298432e-02 1.09568965e+00 5.43758512e-01
-4.07512367e-01 7.44842052e-01 3.94485563e-01 -5.80592752e-01
-9.36791182e-01 -1.25052297e+00 -4.07754540e-01 8.40823352e-01
-4.37716067e-01 -3.60888354e-02 -4.07660693e-01 -7.26809919e-01
-3.01649254e-02 5.29633641e-01 -5.35847247e-01 -3.72668982e-01
-5.47734678e-01 -1.43859923e+00 3.62933189e-01 1.00943840e+00
7.68045485e-01 -8.49244893e-01 -4.05618139e-02 3.41354460e-01
-6.32602990e-01 -7.08660841e-01 -7.05399752e-01 -7.59522691e-02
-1.34482396e+00 -1.34961319e+00 -8.05563271e-01 -7.18363464e-01
7.75499046e-01 -1.04852952e-01 9.48372185e-01 -1.38922304e-01
-1.90900429e-03 2.53439486e-01 -1.71229675e-01 -1.43398494e-01
-5.10463297e-01 1.21901214e-01 -3.65594253e-02 2.36756373e-02
4.67223465e-01 -9.25841808e-01 -1.09804893e+00 2.58262575e-01
-5.61498940e-01 2.62235612e-01 7.88898289e-01 1.12394345e+00
8.13425720e-01 -4.18627769e-01 1.06577885e+00 -7.92176306e-01
4.29841042e-01 -7.43509233e-01 -1.94176823e-01 3.42813939e-01
-6.49904907e-01 -3.37015569e-01 5.96622407e-01 -4.46531028e-01
-1.19331634e+00 1.87184691e-01 -4.39094186e-01 -9.74257737e-02
-3.46181333e-01 7.09671915e-01 -6.55230165e-01 4.28358346e-01
2.40672633e-01 3.47429812e-01 3.01716089e-01 -5.46979666e-01
8.55641887e-02 7.98833013e-01 3.42277586e-01 -2.79487401e-01
2.08922938e-01 2.78359741e-01 -1.28155068e-01 -3.12222362e-01
-7.61309922e-01 -1.51991164e-02 -8.30634058e-01 -6.28929306e-03
1.01406944e+00 -1.38438678e+00 -6.03915453e-01 5.91264844e-01
-9.44058359e-01 -2.91620135e-01 1.41037494e-01 7.60549903e-01
-5.52143037e-01 1.63877249e-01 -5.57825804e-01 -2.45927840e-01
-7.57289827e-01 -1.30289137e+00 5.96070349e-01 -2.04038359e-02
-1.58838212e-01 -1.09757197e+00 1.92864947e-02 1.08728790e+00
3.56700718e-01 3.45175683e-01 1.36977208e+00 -6.42001569e-01
-5.24354517e-01 -2.64990330e-01 -4.85312730e-01 3.61390531e-01
2.81027079e-01 -5.33267856e-01 -5.77928483e-01 -1.87548429e-01
3.18637900e-02 -2.08179817e-01 9.10271883e-01 7.20013738e-01
1.02225852e+00 -1.25465721e-01 -2.46675238e-01 7.10278928e-01
1.24740052e+00 3.26354504e-01 7.66262054e-01 4.39989686e-01
8.21245790e-01 2.37049446e-01 2.53701676e-02 1.61104366e-01
9.65289712e-01 7.65022099e-01 3.21744144e-01 -2.18453214e-01
-2.31773347e-01 -1.45648271e-01 4.78856981e-01 7.77831674e-01
-1.36648923e-01 1.47361726e-01 -9.11140919e-01 1.36625633e-01
-1.77855837e+00 -9.50479507e-01 -5.28924167e-01 2.14557385e+00
7.69198418e-01 1.59923322e-02 2.75917113e-01 -2.29131266e-01
7.17674255e-01 -3.68686646e-01 -8.80290568e-01 5.31788766e-02
-2.44095787e-01 -2.33276293e-01 6.72711357e-02 2.15544805e-01
-1.17537522e+00 2.28097990e-01 5.76169872e+00 2.58432955e-01
-9.30385888e-01 6.94256783e-01 1.01661217e+00 -5.09750322e-02
-2.11912125e-01 -2.48777837e-01 -7.35292792e-01 7.92290986e-01
8.83304477e-01 2.47904673e-01 7.57333487e-02 3.60651344e-01
6.92292213e-01 2.34299421e-01 -9.52100992e-01 5.80616474e-01
-2.20862329e-01 -1.08688009e+00 -1.63302440e-02 9.47249979e-02
4.97739613e-01 2.50161141e-01 4.04629186e-02 5.88072717e-01
1.10462375e-01 -9.09261703e-01 1.46169662e-01 1.08435225e+00
8.00527930e-01 -6.18968010e-01 1.13755763e+00 2.24020258e-01
-1.01875758e+00 -2.89902449e-01 -1.19777843e-01 1.38643131e-01
3.82479489e-01 5.63427567e-01 -4.67983484e-01 7.15014398e-01
4.48441505e-01 9.55506504e-01 -9.26354468e-01 1.03113937e+00
5.41444495e-02 6.43213212e-01 8.45932141e-02 6.89588547e-01
-2.70767987e-01 -4.15615439e-01 2.46049196e-01 8.75883996e-01
5.65287590e-01 2.14031134e-02 2.81911641e-01 1.22930014e+00
1.16992146e-01 -1.36407139e-02 -2.13728607e-01 3.11221153e-01
1.30167395e-01 9.39711034e-01 -1.95780396e-01 -7.55900368e-02
-8.81974816e-01 7.40416050e-01 2.18354359e-01 4.65224594e-01
-1.04997981e+00 1.83490768e-01 3.89884919e-01 2.67681569e-01
-4.07089032e-02 -2.49043982e-02 -5.47906697e-01 -1.45443928e+00
1.24320656e-01 -7.30783165e-01 6.86854661e-01 -7.30796099e-01
-1.87705779e+00 6.28239989e-01 -2.57303417e-01 -1.49028790e+00
3.26177734e-03 -1.93353876e-01 -8.22235584e-01 9.40875649e-01
-1.30216920e+00 -1.58735549e+00 -4.07303184e-01 6.33128047e-01
3.87930691e-01 -4.60314244e-01 7.36106932e-01 6.78488553e-01
-1.05342269e+00 5.94334543e-01 3.33886862e-01 2.45911002e-01
8.93675804e-01 -1.14815128e+00 -9.07089859e-02 5.44598341e-01
-6.49875522e-01 6.45533025e-01 1.16650857e-01 -1.09359562e+00
-1.08867919e+00 -1.75099921e+00 7.69213080e-01 -3.71167153e-01
4.56362695e-01 -1.51939228e-01 -1.00152206e+00 1.01694620e+00
-2.09787846e-01 -4.26414311e-02 1.05009556e+00 1.17233559e-01
-1.65517971e-01 -3.46304744e-01 -9.46677208e-01 6.19803250e-01
9.40562725e-01 -4.01504457e-01 -6.37383640e-01 3.44045699e-01
5.11698544e-01 -3.27565014e-01 -1.37044454e+00 5.35558105e-01
7.84612298e-01 -8.10051978e-01 1.36205137e+00 -7.49428689e-01
7.47964501e-01 -1.88288569e-01 -2.49947354e-01 -1.17870331e+00
-3.94751877e-01 1.97761044e-01 -1.37522236e-01 1.32698929e+00
4.71800894e-01 -6.86928868e-01 6.12007141e-01 9.88823175e-01
-3.53911579e-01 -8.04009676e-01 -1.08229041e+00 -5.08685112e-01
4.24532473e-01 -4.59260970e-01 4.69084263e-01 5.67707837e-01
-3.69735211e-01 3.48298103e-01 -4.39807057e-01 2.53046066e-01
7.50017583e-01 1.26985103e-01 3.27371657e-01 -1.28286302e+00
4.80713472e-02 -2.58617133e-01 -2.98033744e-01 -3.65775883e-01
3.29140335e-01 -1.58645296e+00 -6.83783114e-01 -1.77887738e+00
6.15203440e-01 -4.71978426e-01 -4.63014126e-01 4.65755790e-01
-4.19687033e-01 2.42257729e-01 1.47852167e-01 4.70215142e-01
-2.19161749e-01 9.73587930e-01 1.34044230e+00 -3.90577108e-01
-5.56330979e-01 6.49331152e-01 -7.76523829e-01 4.93985295e-01
1.13051212e+00 -3.28686565e-01 -3.98195803e-01 -2.90243000e-01
-1.11042209e-01 2.60425717e-01 9.91579533e-01 -9.08979118e-01
1.11136943e-01 1.49576142e-01 1.16401124e+00 -6.82799697e-01
2.91002840e-01 -9.66646254e-01 5.45883298e-01 6.13476515e-01
-2.46224090e-01 2.02220172e-01 2.38282643e-02 6.52378261e-01
-1.11597054e-01 1.98628262e-01 4.89778697e-01 -3.98987681e-02
-3.06050777e-01 5.93824506e-01 -3.74224484e-01 1.32950349e-02
7.60801733e-01 -2.79854923e-01 -3.52419227e-01 -9.88612622e-02
-1.44924581e+00 4.52939332e-01 1.20286616e-02 4.30950165e-01
6.82476699e-01 -1.77533448e+00 -8.37588966e-01 3.56626928e-01
5.42906411e-02 -2.80431539e-01 8.40161502e-01 1.54078078e+00
-1.07332945e-01 2.33078808e-01 -5.33230603e-01 -5.67444682e-01
-9.13810134e-01 4.27315533e-01 4.90411848e-01 -2.66731769e-01
-8.29057455e-01 -4.46737232e-03 1.63040340e-01 -6.34060979e-01
1.42321751e-01 -2.52752572e-01 -2.60641456e-01 2.48548806e-01
5.21160424e-01 5.32244205e-01 9.09801796e-02 -5.19215882e-01
-3.01539451e-01 4.17126715e-01 -2.24008471e-01 1.53661504e-01
1.60401106e+00 -3.92193019e-01 -1.03598259e-01 4.20298666e-01
1.10377276e+00 -3.89128029e-01 -1.41761231e+00 -4.64918874e-02
-2.98684351e-02 4.19973172e-02 1.13849863e-01 -1.20735133e+00
-1.22933030e+00 8.68931830e-01 1.23981512e+00 -5.87005138e-01
1.20305097e+00 -1.90080628e-01 9.18333352e-01 -8.98108259e-02
4.50268798e-02 -5.61381400e-01 -3.26541930e-01 1.73988268e-01
1.12234581e+00 -1.61273587e+00 -1.28814220e-01 1.69399176e-02
-7.35318303e-01 1.14583325e+00 6.85641289e-01 1.53520897e-01
7.02850521e-01 -4.71885912e-02 7.19262064e-02 3.42499577e-02
-5.74530005e-01 2.72193074e-01 4.89009738e-01 8.19768190e-01
5.53383291e-01 2.05229536e-01 -3.31223249e-01 1.54068851e+00
6.30636215e-02 5.37939906e-01 3.15016478e-01 6.11275494e-01
3.34330238e-02 -1.31216133e+00 -3.57057095e-01 9.98201072e-01
-4.47811723e-01 -9.88377333e-02 -4.71453294e-02 7.68524647e-01
4.35513943e-01 7.98265040e-01 -2.08588645e-01 -2.76642352e-01
4.83266860e-01 4.02701944e-01 1.37685180e-01 -3.02595466e-01
-6.23833418e-01 1.43957868e-01 -5.68073913e-02 -4.64339226e-01
-4.69706625e-01 -9.17038083e-01 -1.01782799e+00 -1.36979803e-01
2.33300790e-01 -3.79614770e-01 2.00567007e-01 9.87137079e-01
7.41976738e-01 8.62226248e-01 4.46071476e-01 -5.30470967e-01
-4.79704499e-01 -9.49537933e-01 -2.81068087e-01 2.47652933e-01
2.81489640e-01 -5.30438900e-01 -9.03953239e-02 3.59984607e-01]
|
[14.27318000793457, -1.729067087173462]
|
40f0a3e7-1663-41c9-94f5-db5ac5e2d391
|
rignerf-fully-controllable-neural-3d-1
|
2206.06481
| null |
https://arxiv.org/abs/2206.06481v1
|
https://arxiv.org/pdf/2206.06481v1.pdf
|
RigNeRF: Fully Controllable Neural 3D Portraits
|
Volumetric neural rendering methods, such as neural radiance fields (NeRFs), have enabled photo-realistic novel view synthesis. However, in their standard form, NeRFs do not support the editing of objects, such as a human head, within a scene. In this work, we propose RigNeRF, a system that goes beyond just novel view synthesis and enables full control of head pose and facial expressions learned from a single portrait video. We model changes in head pose and facial expressions using a deformation field that is guided by a 3D morphable face model (3DMM). The 3DMM effectively acts as a prior for RigNeRF that learns to predict only residuals to the 3DMM deformations and allows us to render novel (rigid) poses and (non-rigid) expressions that were not present in the input sequence. Using only a smartphone-captured short video of a subject for training, we demonstrate the effectiveness of our method on free view synthesis of a portrait scene with explicit head pose and expression controls. The project page can be found here: http://shahrukhathar.github.io/2022/06/06/RigNeRF.html
|
['Zhixin Shu', 'Eli Shechtman', 'Kalyan Sunkavalli', 'Zexiang Xu', 'ShahRukh Athar']
|
2022-06-13
|
rignerf-fully-controllable-neural-3d
|
http://openaccess.thecvf.com//content/CVPR2022/html/Athar_RigNeRF_Fully_Controllable_Neural_3D_Portraits_CVPR_2022_paper.html
|
http://openaccess.thecvf.com//content/CVPR2022/papers/Athar_RigNeRF_Fully_Controllable_Neural_3D_Portraits_CVPR_2022_paper.pdf
|
cvpr-2022-1
|
['face-model']
|
['computer-vision']
|
[ 1.86096609e-01 4.30758387e-01 2.93490946e-01 -5.78035414e-01
-4.51600403e-01 -3.57036769e-01 5.59936404e-01 -8.46750438e-01
1.04306765e-01 5.84421456e-01 4.44167584e-01 1.91883087e-01
4.28369641e-01 -5.93988299e-01 -1.06693125e+00 -5.40296495e-01
2.03091577e-01 2.22894818e-01 -1.21400863e-01 -3.17415059e-01
-1.32171541e-01 8.30916703e-01 -1.85073340e+00 2.92466909e-01
5.95565796e-01 7.06210256e-01 1.39591068e-01 8.67138743e-01
3.85589540e-01 5.84695280e-01 -4.57166761e-01 -4.02996987e-01
4.63796079e-01 -3.18585783e-01 -4.03652996e-01 3.69269282e-01
1.10355830e+00 -7.00854897e-01 -7.01420188e-01 6.33582532e-01
6.35293663e-01 4.28859770e-01 4.20514882e-01 -1.14564908e+00
-5.03938794e-01 -1.78101018e-01 -4.20584708e-01 -4.80132014e-01
8.58529449e-01 1.46253094e-01 4.99789357e-01 -1.01359177e+00
1.00247133e+00 1.38675094e+00 4.99872208e-01 1.26911306e+00
-1.22671783e+00 -7.48250306e-01 2.88407952e-01 -1.57414809e-01
-1.36883175e+00 -8.51124942e-01 8.98068726e-01 -4.34819847e-01
5.91600597e-01 6.65403187e-01 1.12592363e+00 1.34443188e+00
3.55443507e-01 6.32589996e-01 1.03458965e+00 -2.66220599e-01
1.75524086e-01 -9.78467166e-02 -4.22884256e-01 8.08519900e-01
-3.80547434e-01 7.87333101e-02 -7.13760972e-01 -2.16601655e-01
1.15121889e+00 -5.75860171e-03 -7.19230175e-01 -4.04132843e-01
-7.59072185e-01 6.02907360e-01 3.09639096e-01 -9.93306115e-02
-1.90234020e-01 1.97834328e-01 -1.03567205e-01 -6.54889410e-03
7.12699890e-01 6.53829277e-02 -3.57690722e-01 3.45529728e-02
-9.32211578e-01 5.41380465e-01 7.46345997e-01 1.01493287e+00
7.43771791e-01 2.82931924e-01 -4.78561744e-02 6.82884693e-01
3.54236424e-01 6.63674057e-01 2.43851289e-01 -1.51804960e+00
-6.13539070e-02 2.64682591e-01 3.05033121e-02 -9.54701543e-01
-3.31760645e-01 -5.78414388e-02 -6.18251741e-01 4.45823133e-01
1.32766932e-01 -1.88394979e-01 -1.17382276e+00 1.92824602e+00
8.75624359e-01 3.34781826e-01 -2.85215527e-01 1.09919095e+00
1.04326928e+00 6.21950865e-01 -3.88941616e-01 -1.92747727e-01
1.09606445e+00 -5.94208121e-01 -6.97503984e-01 -2.68155038e-01
3.49394560e-01 -6.59495950e-01 1.20083177e+00 3.99564922e-01
-1.41271198e+00 -2.94971883e-01 -7.04232872e-01 -3.77239406e-01
1.32437587e-01 -3.75861704e-01 5.39107382e-01 3.20755363e-01
-1.31850576e+00 4.92443174e-01 -8.80674541e-01 -3.73621911e-01
2.96832055e-01 4.83061194e-01 -5.60844183e-01 -2.02130303e-01
-8.89917612e-01 6.68691397e-01 -3.58135641e-01 2.26830512e-01
-9.24513280e-01 -8.21064413e-01 -9.64869261e-01 -2.34642729e-01
3.46211374e-01 -9.58621979e-01 1.19031656e+00 -1.22836256e+00
-1.97611678e+00 1.06858230e+00 -2.94506371e-01 1.66473180e-01
6.47886097e-01 -3.19826245e-01 -1.13985483e-02 2.24334568e-01
-2.90699422e-01 9.22854543e-01 9.92806733e-01 -1.43217754e+00
2.13780314e-01 -5.08305907e-01 1.80341899e-01 5.44005334e-01
9.91965607e-02 4.66076806e-02 -6.00795329e-01 -5.65300524e-01
6.50865957e-02 -1.23057926e+00 -5.64010208e-03 5.63633084e-01
-4.39739257e-01 3.73827994e-01 9.02806461e-01 -1.01265574e+00
6.70315206e-01 -2.02346396e+00 3.29007119e-01 1.74300775e-01
6.84192404e-02 3.40170413e-02 7.14046732e-02 1.22900747e-01
-1.55101568e-01 -1.57112733e-01 -2.48163149e-01 -7.19100475e-01
-2.62358226e-02 3.38945270e-01 -1.60315633e-01 6.23618186e-01
-2.20893681e-01 8.58471155e-01 -6.38333023e-01 -2.84219563e-01
2.33613089e-01 1.27118242e+00 -8.28802049e-01 3.91049355e-01
-4.17201251e-01 1.05641818e+00 -2.25514799e-01 5.74671328e-01
9.23971057e-01 2.38014329e-02 1.75171360e-01 -2.72669405e-01
4.00190987e-02 2.01158710e-02 -1.04680300e+00 1.84440172e+00
-3.87126803e-01 4.44278657e-01 4.06753540e-01 -1.79271787e-01
6.17641509e-01 4.11836326e-01 5.05169749e-01 -5.90705276e-01
1.92460045e-01 -2.25620851e-01 -4.51400936e-01 -5.29134154e-01
4.11237895e-01 -2.15573758e-01 2.43866026e-01 1.67102605e-01
-1.39151081e-01 -5.86049139e-01 -4.08713549e-01 1.25192208e-02
8.00218046e-01 7.41247356e-01 -3.24679203e-02 5.54405339e-02
2.44273365e-01 -5.74904501e-01 6.80758238e-01 5.49024269e-02
2.32238144e-01 1.00535274e+00 1.53688520e-01 -3.46229553e-01
-9.87089992e-01 -1.07712436e+00 -1.88512936e-01 8.99244130e-01
-1.76653743e-01 -3.83881956e-01 -1.11127245e+00 -2.83433229e-01
-1.55769929e-01 6.61511838e-01 -6.55945599e-01 2.04830486e-02
-7.91332483e-01 -4.07923847e-01 2.69525170e-01 1.98459327e-01
2.96631962e-01 -8.87891233e-01 -7.91075885e-01 -2.42725909e-01
-2.37057433e-01 -1.05819976e+00 -7.88109601e-01 -3.92834127e-01
-8.58430922e-01 -8.40172410e-01 -8.26403856e-01 -4.51466918e-01
9.70022261e-01 -4.81141694e-02 9.62365568e-01 1.01473898e-01
-3.52934450e-01 7.41726100e-01 4.23937924e-02 -1.95448592e-01
-3.19619149e-01 -4.62641627e-01 1.24468803e-01 2.80882239e-01
-4.02715266e-01 -9.11814451e-01 -8.60400379e-01 2.63974220e-01
-8.54627192e-01 5.64195693e-01 -1.63598374e-01 5.90129256e-01
5.30932903e-01 -6.74085617e-01 -7.26742446e-02 -7.17748582e-01
1.29571661e-01 6.27146615e-03 -6.09289825e-01 5.10968314e-03
-1.68219998e-01 -3.53597134e-01 2.71625459e-01 -5.96654475e-01
-1.31316173e+00 2.01702982e-01 -4.17595237e-01 -8.24543834e-01
-2.21325070e-01 -7.48854503e-02 -5.23850977e-01 -2.30495095e-01
5.28657377e-01 4.69267890e-02 9.10106078e-02 -4.73680079e-01
3.49417061e-01 3.08984250e-01 6.01293266e-01 -5.87258816e-01
8.12005937e-01 7.23363101e-01 1.60872892e-01 -1.08379984e+00
-4.86117363e-01 1.96471855e-01 -8.36481214e-01 -6.10103607e-01
8.80277693e-01 -1.02560866e+00 -9.08115804e-01 6.57108665e-01
-1.03152609e+00 -8.35381389e-01 -3.24595362e-01 3.95933568e-01
-6.80460095e-01 1.58209354e-01 -6.01728499e-01 -7.74843991e-01
-3.20592821e-01 -9.70243990e-01 1.36663556e+00 3.14849049e-01
-3.74236494e-01 -9.80788827e-01 4.40447964e-03 5.93693197e-01
2.50190377e-01 8.60179365e-01 6.11970246e-01 3.03702593e-01
-7.36460030e-01 -2.67406311e-02 4.74333465e-01 3.07682931e-01
6.03459962e-02 2.63838381e-01 -1.33513701e+00 -5.08267522e-01
1.09526142e-01 -2.45874077e-01 2.09485427e-01 4.44998860e-01
1.04864645e+00 -6.18654490e-01 -9.73636210e-02 1.24916005e+00
1.03489435e+00 1.47018015e-01 8.31379294e-01 -1.38735399e-01
1.15457606e+00 6.62727356e-01 1.65100008e-01 6.79040253e-01
5.75039566e-01 1.05218410e+00 5.34430861e-01 -1.13589287e-01
-2.34813496e-01 -5.61502993e-01 5.22002041e-01 7.23014116e-01
-5.45406699e-01 -1.66010991e-01 -5.64571261e-01 5.75239118e-03
-1.37053430e+00 -7.79878318e-01 1.04041055e-01 2.31172967e+00
7.65747666e-01 -4.43964750e-01 -1.00855738e-01 -2.41366729e-01
4.22841042e-01 2.51992464e-01 -7.58274913e-01 -4.61487591e-01
2.65824515e-02 2.37819850e-01 9.61187854e-02 8.04300129e-01
-5.46809316e-01 9.28778768e-01 5.42015219e+00 1.73060507e-01
-1.43985474e+00 9.17597115e-02 4.62311715e-01 -6.29288554e-01
-7.81423390e-01 -1.18909106e-01 -6.13693118e-01 1.29568398e-01
7.19316661e-01 4.81863879e-02 6.99438870e-01 5.27070940e-01
5.37381947e-01 6.34618700e-02 -1.00552773e+00 1.03764820e+00
5.46712637e-01 -1.04333568e+00 7.80546442e-02 3.43080401e-01
6.85318470e-01 -2.09478393e-01 1.88268304e-01 2.17149407e-02
5.97592536e-03 -9.91819322e-01 8.77202213e-01 8.80332649e-01
1.07444525e+00 -5.54684281e-01 -8.58750343e-02 3.18044454e-01
-6.92005813e-01 2.29708210e-01 -1.49250682e-02 8.45419019e-02
3.14510345e-01 1.73285335e-01 -6.14895523e-01 2.23561004e-01
6.84807539e-01 4.75366592e-01 -2.43175358e-01 5.50327063e-01
-2.27977693e-01 2.84826785e-01 -3.18412095e-01 6.43310905e-01
-3.20014268e-01 -3.19565237e-01 8.06805253e-01 6.93325877e-01
4.00424778e-01 5.17303109e-01 -1.30269840e-01 6.83658242e-01
-2.60190725e-01 8.61188024e-02 -7.98779905e-01 3.61853629e-01
9.14411806e-03 1.17627490e+00 -2.76894242e-01 5.20069264e-02
-2.70816892e-01 1.21462619e+00 1.99427977e-01 5.71294725e-01
-9.31075692e-01 3.59459907e-01 9.10207212e-01 7.30172932e-01
9.27785188e-02 -2.65129179e-01 1.85053825e-01 -1.47666538e+00
2.14196723e-02 -9.83005285e-01 -1.15914486e-01 -1.41912639e+00
-6.80764556e-01 6.53986275e-01 2.25828663e-01 -1.00796986e+00
-4.60754961e-01 -4.04218763e-01 -5.76688826e-01 7.71362066e-01
-1.01408887e+00 -1.46977770e+00 -5.92236817e-01 8.27629089e-01
4.52805191e-01 2.61699229e-01 9.16938484e-01 2.26366967e-01
-4.29105878e-01 7.72211730e-01 -2.17402115e-01 -7.86423981e-02
7.56834686e-01 -8.04853380e-01 4.30925995e-01 5.18678188e-01
-3.82257365e-02 5.59690654e-01 7.67547488e-01 -7.25505888e-01
-1.72328424e+00 -9.60084379e-01 4.43737328e-01 -5.69360375e-01
-9.85293090e-02 -8.31070781e-01 -9.78800595e-01 1.12077332e+00
7.13054240e-02 2.40318909e-01 5.67924857e-01 -3.33082885e-01
-1.36116952e-01 -1.02842726e-01 -1.39555538e+00 8.60354245e-01
1.34647143e+00 -6.20232224e-01 -1.57385513e-01 3.27303022e-01
6.15442872e-01 -1.13668287e+00 -9.68144894e-01 3.82713318e-01
1.01140726e+00 -9.82573628e-01 8.73740792e-01 -3.44309092e-01
1.94406852e-01 -2.10392028e-01 -3.66428375e-01 -1.34470153e+00
-9.03463885e-02 -9.87648070e-01 -2.85514832e-01 9.73324478e-01
-1.39371887e-01 -6.39422357e-01 8.12149525e-01 1.10392761e+00
-2.13972718e-01 -8.41653407e-01 -9.21235263e-01 -3.91166240e-01
-4.73876171e-05 -3.46632779e-01 6.97472870e-01 8.56355131e-01
-3.93132836e-01 7.93369934e-02 -6.19309545e-01 3.46282601e-01
5.73617637e-01 -1.44870371e-01 1.08761668e+00 -1.09981740e+00
-4.09360558e-01 2.31754050e-01 -3.25043678e-01 -9.99156237e-01
4.03885692e-01 -7.57196724e-01 -8.17068070e-02 -1.36858749e+00
-2.24603210e-02 -1.89413056e-02 4.71246302e-01 4.97017086e-01
2.14697629e-01 2.87419438e-01 4.54900384e-01 1.91879317e-01
4.16356623e-02 7.35286295e-01 1.69955540e+00 2.24920765e-01
-2.69743770e-01 -3.86541374e-02 -4.43890393e-01 1.04251945e+00
5.09031773e-01 -2.03935474e-01 -5.56929946e-01 -5.79104483e-01
1.12028509e-01 4.97393817e-01 6.04214311e-01 -7.62760460e-01
-7.87932128e-02 -2.19639510e-01 6.91359341e-01 -3.39425921e-01
1.03876102e+00 -7.20107198e-01 9.35024142e-01 1.05676480e-01
-1.04685754e-01 1.92821383e-01 2.15926692e-01 1.94237009e-01
2.72182226e-01 2.73581177e-01 8.21260154e-01 -3.08153570e-01
-3.02081883e-01 6.47044778e-01 -1.14319965e-01 8.80138855e-03
7.44565904e-01 -4.18374121e-01 -4.83276136e-02 -7.61963367e-01
-9.23848569e-01 -1.02581836e-01 1.02429032e+00 3.53785783e-01
8.46406043e-01 -1.28716719e+00 -6.76749051e-01 6.55646265e-01
-1.41848907e-01 4.29297656e-01 5.55677176e-01 6.03075147e-01
-4.99770015e-01 -1.86611027e-01 -1.77101389e-01 -5.83932698e-01
-1.59745121e+00 1.63570672e-01 8.04925561e-01 2.93374568e-01
-1.02404511e+00 7.56218374e-01 7.82848597e-01 -6.14537120e-01
1.04571767e-01 -2.70432323e-01 9.12721679e-02 -1.85048759e-01
4.25294578e-01 1.98624685e-01 -4.67100255e-02 -1.18324423e+00
-2.13605613e-01 8.64191711e-01 2.20926434e-01 -2.95213491e-01
1.33314800e+00 -2.52710521e-01 1.45045400e-01 4.47183728e-01
1.24471200e+00 1.37938872e-01 -1.76667368e+00 1.20318763e-01
-8.39569151e-01 -7.17920303e-01 2.91731339e-02 -6.32848799e-01
-1.25364363e+00 6.31764591e-01 6.62524879e-01 -6.99652612e-01
1.22439015e+00 -7.48008043e-02 9.29980934e-01 1.23747692e-01
5.06066561e-01 -8.83056402e-01 -1.59529205e-02 3.91187489e-01
1.43280721e+00 -7.84380674e-01 6.27473369e-02 -4.27542090e-01
-5.43784499e-01 9.16885138e-01 6.26882255e-01 1.30771166e-02
6.58484578e-01 3.85766447e-01 1.68523908e-01 -1.73906252e-01
-6.86385214e-01 3.31671774e-01 4.31964040e-01 6.79835439e-01
4.11970198e-01 1.45411462e-01 2.64124960e-01 6.64935971e-04
-5.42183936e-01 2.07331643e-01 5.94865620e-01 8.88195395e-01
9.54157040e-02 -7.99242318e-01 -5.77390730e-01 1.91220254e-01
-3.14472169e-01 1.40864760e-01 -2.22303271e-01 7.28321314e-01
2.74723321e-01 4.30122524e-01 2.52986401e-01 -3.39732885e-01
6.31464839e-01 1.48802534e-01 9.29504156e-01 -6.22457325e-01
-3.55947196e-01 3.52290511e-01 8.42542201e-03 -8.76787424e-01
-2.81838179e-01 -7.72451341e-01 -1.22576594e+00 -5.64361513e-01
1.19452022e-01 -3.20680857e-01 6.21082127e-01 6.19753540e-01
3.57972085e-01 2.69166619e-01 7.01522589e-01 -1.54183817e+00
-1.00547977e-01 -6.45535111e-01 -6.30081415e-01 4.60880041e-01
5.28340220e-01 -6.81961596e-01 -4.63188142e-01 3.47800195e-01]
|
[12.7859525680542, -0.42001432180404663]
|
32ce11ed-6017-4a96-b880-81dd32b238c7
|
kecp-knowledge-enhanced-contrastive-prompting
|
2205.03071
| null |
https://arxiv.org/abs/2205.03071v1
|
https://arxiv.org/pdf/2205.03071v1.pdf
|
KECP: Knowledge Enhanced Contrastive Prompting for Few-shot Extractive Question Answering
|
Extractive Question Answering (EQA) is one of the most important tasks in Machine Reading Comprehension (MRC), which can be solved by fine-tuning the span selecting heads of Pre-trained Language Models (PLMs). However, most existing approaches for MRC may perform poorly in the few-shot learning scenario. To solve this issue, we propose a novel framework named Knowledge Enhanced Contrastive Prompt-tuning (KECP). Instead of adding pointer heads to PLMs, we introduce a seminal paradigm for EQA that transform the task into a non-autoregressive Masked Language Modeling (MLM) generation problem. Simultaneously, rich semantics from the external knowledge base (KB) and the passage context are support for enhancing the representations of the query. In addition, to boost the performance of PLMs, we jointly train the model by the MLM and contrastive learning objectives. Experiments on multiple benchmarks demonstrate that our method consistently outperforms state-of-the-art approaches in few-shot settings by a large margin.
|
['Ming Gao', 'Jun Huang', 'Hongbin Wang', 'Qiuhui Shi', 'Minghui Qiu', 'Chengyu Wang', 'Jianing Wang']
|
2022-05-06
| null | null | null | null |
['machine-reading-comprehension']
|
['natural-language-processing']
|
[ 4.08092529e-01 3.28353405e-01 -9.56552923e-02 -2.08186015e-01
-1.31671321e+00 -2.45152101e-01 6.71427071e-01 3.41378689e-01
-6.25577986e-01 6.94637418e-01 6.43901289e-01 -4.65497643e-01
-3.47839184e-02 -8.93979728e-01 -8.89232457e-01 -4.05743927e-01
4.90074158e-01 3.93703043e-01 5.59304893e-01 -5.55155873e-01
3.26627254e-01 -7.02620894e-02 -1.48006213e+00 4.70645040e-01
1.18767118e+00 6.75344944e-01 5.45649648e-01 8.36480916e-01
-6.05485141e-01 1.24630725e+00 -5.10419428e-01 -4.91533220e-01
-2.40613863e-01 -5.87221980e-01 -1.22476256e+00 -2.10358366e-01
2.32497722e-01 -3.62886012e-01 -4.73076373e-01 7.08715618e-01
6.52317822e-01 7.58162439e-01 5.70185900e-01 -8.24961245e-01
-8.60300899e-01 8.27462733e-01 -4.00859386e-01 5.69194138e-01
5.09804189e-01 1.39486909e-01 1.14092660e+00 -1.11831927e+00
3.95592481e-01 1.40661442e+00 1.57057822e-01 6.41802609e-01
-9.74892080e-01 -2.00239763e-01 4.04464990e-01 7.55003095e-01
-9.96670902e-01 -6.75322175e-01 6.15402102e-01 -1.21017985e-01
1.15154147e+00 2.26299912e-01 1.20340109e-01 1.17414355e+00
-1.65517792e-01 1.30965567e+00 1.03333759e+00 -9.36741948e-01
3.64327282e-01 -1.29679665e-01 6.35295987e-01 7.87178099e-01
-2.34403670e-01 -3.90138716e-01 -8.90549302e-01 -2.97075808e-01
2.34421641e-01 -2.58974880e-01 -4.22686607e-01 -9.73903686e-02
-1.00737989e+00 8.49703074e-01 4.48310822e-02 6.15594871e-02
-5.39370358e-01 -5.07517979e-02 2.00894907e-01 2.39069521e-01
3.40118855e-01 6.56450391e-01 -5.63511848e-01 -1.92996889e-01
-8.41848373e-01 3.68699461e-01 7.89435923e-01 9.34033871e-01
6.08155131e-01 -2.89478093e-01 -1.22324681e+00 9.27186489e-01
1.53611779e-01 2.51381487e-01 5.40811419e-01 -8.96331668e-01
7.10849047e-01 4.04683560e-01 2.76438028e-01 -3.69316310e-01
-2.01473832e-01 -4.39659178e-01 -5.28541863e-01 -4.31153446e-01
4.12478983e-01 -1.17329676e-02 -9.03450429e-01 1.84191239e+00
4.45597321e-01 4.02805597e-01 2.80710787e-01 6.11135960e-01
8.31596434e-01 1.09755802e+00 3.82823229e-01 -3.65542740e-01
1.64319909e+00 -1.53926206e+00 -9.52809036e-01 -5.17938495e-01
5.90721190e-01 -5.24937212e-01 1.60666251e+00 6.98166713e-02
-1.43478894e+00 -4.34534967e-01 -8.56693089e-01 -5.82173467e-01
-1.68661803e-01 -3.15610647e-01 1.29913226e-01 3.80852185e-02
-6.58065259e-01 3.27958643e-01 -7.44016409e-01 -1.73152342e-01
3.22130531e-01 -1.90371618e-01 1.96952075e-01 -4.05785620e-01
-1.63836229e+00 1.02388930e+00 5.59722126e-01 -2.35053375e-01
-9.46107268e-01 -8.70263040e-01 -8.01448166e-01 4.78862345e-01
8.79322708e-01 -9.83838737e-01 1.70202839e+00 -4.47306156e-01
-1.81003284e+00 6.33456767e-01 -6.72932148e-01 -5.59603930e-01
9.88277867e-02 -7.05536187e-01 -9.93150100e-02 3.71847093e-01
5.63730970e-02 4.19465750e-01 1.03029335e+00 -9.95152950e-01
-6.24372900e-01 -1.77599892e-01 3.01091075e-01 3.96283925e-01
-2.85581768e-01 2.49576673e-01 -4.75668848e-01 -5.81454217e-01
-2.45107368e-01 -3.09739739e-01 -2.01310173e-01 -4.89915788e-01
-3.13058048e-01 -6.82666957e-01 2.59171635e-01 -1.19903100e+00
1.61931264e+00 -1.96040320e+00 4.43482727e-01 -2.55850554e-01
1.55812055e-01 4.70854610e-01 -4.82739270e-01 5.33915520e-01
3.10922205e-01 -1.94284543e-01 -3.65889460e-01 -4.55765128e-01
1.29215598e-01 1.45384461e-01 -7.23667800e-01 -1.75244048e-01
3.29021275e-01 1.32214308e+00 -1.26172078e+00 -5.86591363e-01
-1.79798976e-01 7.56467134e-02 -4.64868784e-01 7.37664282e-01
-7.59168327e-01 2.99220741e-01 -5.29701829e-01 2.69599646e-01
2.82295972e-01 -5.94485223e-01 -2.14573234e-01 2.03760386e-01
2.24568486e-01 6.35429800e-01 -8.93049419e-01 1.92291641e+00
-4.60535467e-01 4.46143150e-02 -2.81342655e-01 -8.86162221e-01
5.87432027e-01 2.85941809e-01 -1.52611390e-01 -8.54164958e-01
-6.48882762e-02 -4.44730595e-02 -7.69696757e-02 -8.02765608e-01
7.43454695e-01 -6.33638799e-02 4.72039916e-02 7.08939016e-01
4.67263222e-01 2.15144679e-01 2.11298347e-01 6.34620309e-01
1.22669089e+00 1.11769371e-01 6.00008130e-01 1.71311311e-02
5.42765141e-01 -1.96725264e-01 3.22239727e-01 1.24747610e+00
-1.45525739e-01 4.19743240e-01 3.78515899e-01 1.81169823e-01
-8.52845669e-01 -1.19275105e+00 3.52876902e-01 1.91531968e+00
6.19544573e-02 -4.55509514e-01 -1.00630498e+00 -6.77218974e-01
-4.16427612e-01 1.41648448e+00 -5.05560160e-01 -5.54175615e-01
-6.57363772e-01 -5.66357791e-01 5.01720667e-01 6.44965112e-01
4.38823819e-01 -1.21980834e+00 -6.55279934e-01 5.41225195e-01
-6.46219671e-01 -1.17380452e+00 -4.98020977e-01 -2.08469890e-02
-6.27177835e-01 -8.09131682e-01 -9.16941226e-01 -7.09573328e-01
2.44967550e-01 4.73060757e-01 1.45403731e+00 1.05697930e-01
-7.42370635e-02 5.42350054e-01 -6.81208432e-01 -5.43852031e-01
-3.96607846e-01 4.01197225e-01 -2.46722043e-01 9.07370895e-02
5.13441622e-01 -4.13727254e-01 -5.07659197e-01 -8.54059160e-02
-9.59643066e-01 3.70508939e-01 5.27090609e-01 9.23821509e-01
4.46639568e-01 -3.59491587e-01 9.76541340e-01 -8.46058667e-01
9.56987262e-01 -7.17599094e-01 -2.31183156e-01 9.64202285e-01
-3.93019676e-01 4.45079148e-01 5.24170101e-01 -5.08940756e-01
-1.65964758e+00 -5.12565255e-01 -1.29867986e-01 -1.40908986e-01
-6.01506308e-02 6.53713465e-01 -2.06761479e-01 4.72096056e-01
7.58567810e-01 5.70004821e-01 -3.64523441e-01 -5.75606763e-01
7.79899359e-01 6.07300460e-01 6.46023333e-01 -6.72722101e-01
6.31524324e-01 1.04538225e-01 -4.12262827e-01 -7.92511284e-01
-1.73114014e+00 -7.27992952e-01 -4.34483647e-01 4.66541275e-02
9.02741909e-01 -8.65074039e-01 -2.11391136e-01 5.00212431e-01
-1.32307696e+00 -3.89293641e-01 -3.97651166e-01 2.33486876e-01
-6.65598750e-01 4.23344433e-01 -6.37285829e-01 -1.01584852e+00
-6.30793869e-01 -7.30995834e-01 9.75731134e-01 5.43103337e-01
-2.11537525e-01 -8.70921314e-01 1.02630667e-01 7.30613053e-01
5.82084477e-01 -3.48157376e-01 1.49740005e+00 -8.92667234e-01
-6.39401793e-01 1.66774079e-01 -1.05087094e-01 1.47437915e-01
-2.72583187e-01 -5.73712766e-01 -1.19718742e+00 -3.71230356e-02
2.22484127e-01 -7.44970798e-01 1.17291653e+00 1.50898233e-01
1.22286177e+00 -3.90915245e-01 3.37214097e-02 2.25678697e-01
1.01148140e+00 -1.64526284e-01 6.68858230e-01 3.06346774e-01
5.22558630e-01 7.81481743e-01 7.48989820e-01 4.72307295e-01
7.47992933e-01 4.39342380e-01 9.66702700e-02 2.98084617e-01
-1.64775640e-01 -7.07527399e-01 3.49278063e-01 1.09744751e+00
7.24331886e-02 -2.86683470e-01 -9.31533575e-01 6.82400465e-01
-2.00671172e+00 -9.02817667e-01 1.36613831e-01 2.04441786e+00
1.34652233e+00 6.36523031e-03 -1.77617267e-01 -1.54165506e-01
4.92457837e-01 3.68641913e-01 -6.11704290e-01 -2.44321048e-01
-1.37564421e-01 6.36053264e-01 -1.50800452e-01 6.56688809e-01
-7.61115015e-01 1.19547915e+00 5.60915136e+00 1.04029441e+00
-3.91129494e-01 4.36213374e-01 3.49169314e-01 -4.82015237e-02
-3.90043348e-01 1.43324837e-01 -1.02390277e+00 1.87328443e-01
1.15411270e+00 -4.30993170e-01 5.64953089e-01 6.87021255e-01
4.37027737e-02 -2.15065017e-01 -1.09893775e+00 6.86499894e-01
3.79336059e-01 -1.18819022e+00 2.65098691e-01 -4.62313443e-01
7.81325340e-01 -9.23672840e-02 -1.79580837e-01 9.96504247e-01
1.16111435e-01 -8.87900114e-01 4.57897156e-01 9.88048315e-01
3.12653542e-01 -4.85329598e-01 4.35029089e-01 9.72610474e-01
-6.64119840e-01 -2.20018029e-01 -5.57143152e-01 -1.30615607e-01
5.19144356e-01 2.51036406e-01 -5.23763180e-01 4.55196559e-01
4.83432055e-01 6.54338598e-02 -6.00521863e-01 9.86507535e-01
-7.21416414e-01 1.01889646e+00 5.89987002e-02 -1.20331340e-01
2.48589411e-01 2.54556328e-01 6.27279878e-01 1.07268107e+00
-1.58508234e-02 4.75739300e-01 -3.80621701e-02 1.08156848e+00
-4.76156712e-01 3.27749223e-01 -6.99879695e-03 -1.48115292e-01
5.86618483e-01 9.78148818e-01 -3.08115557e-02 -7.09086120e-01
-5.95473111e-01 1.13293910e+00 8.53729665e-01 4.98548627e-01
-5.56593239e-01 -4.27356362e-01 1.05288118e-01 -2.53646122e-03
3.13094735e-01 -2.61033163e-03 1.54494569e-01 -1.50355780e+00
1.71008438e-01 -1.03952706e+00 5.60001373e-01 -9.17136490e-01
-1.36243057e+00 1.73404157e-01 1.20032907e-01 -5.55272758e-01
-3.41912061e-01 -3.27964216e-01 -7.95668006e-01 1.05953002e+00
-2.04896855e+00 -1.00137389e+00 -2.55735636e-01 5.14598310e-01
1.01924503e+00 -1.29902712e-03 1.02819014e+00 -1.63012698e-01
-4.99774426e-01 5.38517237e-01 6.72185868e-02 -1.47655636e-01
7.26134837e-01 -1.29181695e+00 5.52256286e-01 1.10841048e+00
2.11482599e-01 6.42237842e-01 5.93564034e-01 -3.96220565e-01
-1.36459672e+00 -9.88753259e-01 1.29768991e+00 -4.81782258e-01
6.85389400e-01 -3.03084552e-01 -1.59724975e+00 6.30608618e-01
4.12638694e-01 -3.16062897e-01 7.08162785e-01 1.57513335e-01
-4.49156076e-01 3.28676045e-01 -5.52145123e-01 6.37876809e-01
7.58889794e-01 -7.11246669e-01 -1.41840494e+00 4.47499245e-01
1.23437381e+00 -3.90897274e-01 -6.93571866e-01 1.80276796e-01
1.06626399e-01 -4.85459477e-01 8.83906066e-01 -1.28110278e+00
7.28526831e-01 1.37137214e-03 -1.31442025e-01 -1.28826904e+00
-3.70504141e-01 -7.13040233e-01 -1.05603266e+00 1.22814536e+00
4.09670264e-01 -2.34122783e-01 2.98635244e-01 6.55792236e-01
-1.63297251e-01 -8.60184610e-01 -8.06488276e-01 -6.45424128e-01
2.26975009e-01 -3.37431014e-01 5.62732160e-01 6.19610250e-01
9.16844904e-02 1.04969382e+00 -4.43580270e-01 6.77298829e-02
5.59658587e-01 -1.09274946e-02 7.02351749e-01 -8.89831245e-01
-7.57687986e-01 -2.91699618e-01 4.61841166e-01 -1.50686932e+00
3.21935862e-01 -8.41666579e-01 3.12845767e-01 -1.73155129e+00
5.89644253e-01 9.43502188e-02 -5.21124780e-01 3.37802738e-01
-1.10331094e+00 -6.52947664e-01 2.47066021e-01 3.32218818e-02
-1.01601183e+00 9.09731388e-01 1.21077359e+00 4.56230491e-02
-1.84424445e-01 -2.27013770e-02 -8.65333259e-01 6.71993673e-01
6.31661594e-01 -3.09315145e-01 -5.20118654e-01 -5.75274706e-01
2.13127062e-01 2.63812751e-01 3.14428091e-01 -5.19629598e-01
5.71126759e-01 -2.17288002e-01 -1.05384797e-01 -4.90117043e-01
2.82184482e-01 -4.88482453e-02 -7.91691899e-01 8.04727226e-02
-8.70418787e-01 -2.74980128e-01 1.17900902e-02 7.58745849e-01
-1.73793346e-01 -6.84048235e-01 6.65437877e-01 -2.65305042e-01
-9.31223035e-01 2.04972133e-01 -2.45458022e-01 7.67249763e-01
5.65819800e-01 3.55559945e-01 -6.52402580e-01 -4.67470139e-01
-5.07253826e-01 5.84491789e-01 -1.44861802e-01 6.30406559e-01
7.18496382e-01 -9.47236419e-01 -8.02500308e-01 -9.99713093e-02
4.62031692e-01 2.45678484e-01 5.34962177e-01 8.29054773e-01
1.53536528e-01 5.87503910e-01 2.64239073e-01 -7.43266940e-02
-9.92419660e-01 7.27352083e-01 1.29949346e-01 -7.59988725e-01
-5.95144689e-01 9.70627189e-01 2.20196784e-01 -3.83701026e-01
5.21115124e-01 -4.12523113e-02 -4.97125059e-01 4.05580774e-02
1.08081186e+00 5.35831392e-01 -1.13783274e-02 -2.10060790e-01
1.66040584e-01 3.63306627e-02 -4.74570721e-01 -2.96671331e-01
1.19555438e+00 -3.19744229e-01 -4.36880300e-03 6.63556159e-01
8.01197886e-01 -1.51200503e-01 -1.07907283e+00 -8.82746994e-01
3.79310966e-01 2.56392336e-03 2.67242920e-02 -1.01569843e+00
-9.57363471e-02 1.12231243e+00 2.11950839e-02 -1.82053313e-01
1.03966188e+00 2.05627963e-01 1.17970014e+00 8.14671218e-01
1.06518589e-01 -1.19632864e+00 3.73800844e-01 8.54260564e-01
8.32154512e-01 -1.21571946e+00 -3.87486130e-01 -1.80191278e-01
-7.48019099e-01 7.88270354e-01 7.65987217e-01 1.40528619e-01
2.81936526e-01 -2.24283054e-01 -1.06927313e-01 -1.68522466e-02
-1.14774919e+00 -2.52519727e-01 4.14014250e-01 1.87971875e-01
2.55971164e-01 -1.42773300e-01 -3.32613856e-01 1.13722575e+00
-3.16363834e-02 -1.68428645e-02 3.11614484e-01 1.21820641e+00
-9.24911141e-01 -9.64703381e-01 -1.97567135e-01 5.11422157e-01
-4.96067613e-01 -5.80911875e-01 -7.36092329e-02 3.08931142e-01
-4.83516872e-01 9.73163068e-01 -2.75597721e-01 2.91304085e-02
3.78591865e-01 6.98892117e-01 4.15194124e-01 -8.14074218e-01
-4.58135515e-01 -2.21057579e-01 7.57133216e-02 -5.00815153e-01
-2.52654463e-01 -3.60778749e-01 -1.18708861e+00 2.52410740e-01
-4.60792959e-01 2.83361286e-01 2.00261056e-01 1.48542500e+00
4.25758302e-01 6.37296498e-01 3.12709361e-01 -1.20313734e-01
-1.17005575e+00 -1.32582438e+00 -3.22134733e-01 4.58284855e-01
2.37321109e-01 -4.50406551e-01 -2.24217013e-01 5.50101027e-02]
|
[11.168326377868652, 8.04631519317627]
|
400fcc3b-f13e-45d2-a18d-224573aaa071
|
no-metrics-are-perfect-adversarial-reward
|
1804.0916
| null |
http://arxiv.org/abs/1804.09160v2
|
http://arxiv.org/pdf/1804.09160v2.pdf
|
No Metrics Are Perfect: Adversarial Reward Learning for Visual Storytelling
|
Though impressive results have been achieved in visual captioning, the task
of generating abstract stories from photo streams is still a little-tapped
problem. Different from captions, stories have more expressive language styles
and contain many imaginary concepts that do not appear in the images. Thus it
poses challenges to behavioral cloning algorithms. Furthermore, due to the
limitations of automatic metrics on evaluating story quality, reinforcement
learning methods with hand-crafted rewards also face difficulties in gaining an
overall performance boost. Therefore, we propose an Adversarial REward Learning
(AREL) framework to learn an implicit reward function from human
demonstrations, and then optimize policy search with the learned reward
function. Though automatic eval- uation indicates slight performance boost over
state-of-the-art (SOTA) methods in cloning expert behaviors, human evaluation
shows that our approach achieves significant improvement in generating more
human-like stories than SOTA systems.
|
['Yuan-Fang Wang', 'William Yang Wang', 'Wenhu Chen', 'Xin Wang']
|
2018-04-24
|
no-metrics-are-perfect-adversarial-reward-1
|
https://aclanthology.org/P18-1083
|
https://aclanthology.org/P18-1083.pdf
|
acl-2018-7
|
['visual-storytelling']
|
['natural-language-processing']
|
[ 2.68254161e-01 3.41802448e-01 -3.08704644e-01 -1.75842881e-01
-8.23491275e-01 -4.82845366e-01 8.61690938e-01 -3.71201873e-01
-2.67408371e-01 1.10878038e+00 4.31514263e-01 6.02159053e-02
3.85999203e-01 -4.61690992e-01 -1.18226564e+00 -3.18657845e-01
9.44954827e-02 4.53714103e-01 -5.15015163e-02 -2.91847080e-01
1.94344014e-01 1.15332007e-01 -1.61423504e+00 4.61871594e-01
6.57452047e-01 5.91839492e-01 4.15364616e-02 1.00971866e+00
-3.09567843e-02 1.44600141e+00 -1.00436890e+00 -5.35297930e-01
4.23908457e-02 -8.91668618e-01 -7.24607706e-01 7.96405450e-02
5.97868145e-01 -7.14140534e-01 -4.39524144e-01 9.29663718e-01
4.41169769e-01 -3.34713086e-02 7.10125566e-01 -1.82835543e+00
-1.02633369e+00 6.53419554e-01 -3.98158640e-01 -1.39821246e-01
8.28681052e-01 6.09440565e-01 8.18547666e-01 -5.36225140e-01
9.39727843e-01 1.28398812e+00 4.77697253e-01 1.15238762e+00
-1.52185047e+00 -7.56671667e-01 8.07604417e-02 1.59387305e-01
-9.54981208e-01 -3.38639081e-01 6.76224172e-01 -5.45680285e-01
9.79668558e-01 3.33379030e-01 9.00974035e-01 1.96446502e+00
-2.58389622e-01 1.19882798e+00 1.12141526e+00 -1.46198764e-01
3.02762955e-01 4.05070812e-01 -5.20500004e-01 6.23899817e-01
-2.37950962e-02 4.26430106e-01 -5.78703105e-01 -8.36313367e-02
1.01965439e+00 -3.91864866e-01 -3.70304622e-02 -6.03850305e-01
-1.46732605e+00 7.14070737e-01 4.24646825e-01 5.84877729e-02
-3.92293483e-01 8.06706131e-01 4.12565619e-01 2.29650423e-01
1.03143968e-01 8.85804057e-01 1.16485566e-01 -7.32589006e-01
-8.91488373e-01 8.12328756e-01 6.03718221e-01 1.19820154e+00
3.28492373e-01 5.29479861e-01 -6.60894096e-01 7.16218352e-01
-2.79314518e-01 5.86867571e-01 3.63443434e-01 -1.29640949e+00
3.83108974e-01 3.41321975e-01 5.13198733e-01 -7.85160959e-01
1.15943454e-01 6.06790418e-04 -3.18418562e-01 9.22947228e-01
5.02602696e-01 -2.32937172e-01 -9.04011846e-01 1.85209250e+00
-2.13891901e-02 1.73417911e-01 1.88807413e-01 1.02259159e+00
6.05476677e-01 8.32131624e-01 2.93739706e-01 1.20438553e-01
9.55677569e-01 -1.30717099e+00 -5.45243740e-01 -3.32796514e-01
2.43386388e-01 -3.93101245e-01 1.55623317e+00 3.34439009e-01
-1.28385782e+00 -4.96904731e-01 -1.10857630e+00 2.01468840e-01
-7.46331364e-02 1.19539395e-01 5.68663299e-01 6.02640450e-01
-8.76884639e-01 4.83884752e-01 -6.16009355e-01 -3.13073337e-01
6.72357857e-01 1.45064844e-02 -2.50753134e-01 -3.46494070e-03
-6.44978225e-01 1.07848132e+00 2.31238335e-01 -6.13999963e-01
-1.54698491e+00 -7.25959301e-01 -8.28907907e-01 -4.34510559e-02
5.25200427e-01 -7.15927780e-01 1.62241757e+00 -1.52206242e+00
-1.68669569e+00 7.13845849e-01 2.27924019e-01 -6.48515224e-01
9.73170280e-01 -3.80596250e-01 -2.33165056e-01 2.56208718e-01
1.98650807e-01 1.38131571e+00 1.06380987e+00 -1.59993637e+00
-5.09328425e-01 4.17322636e-01 4.42062885e-01 2.70240605e-01
-2.56071746e-01 1.19650766e-01 -8.02088305e-02 -8.60578716e-01
-9.13947582e-01 -9.02063787e-01 -1.98344767e-01 3.43942285e-01
-2.26151958e-01 2.18434073e-03 8.29809427e-01 -6.30644619e-01
8.83051038e-01 -1.96303082e+00 1.71688393e-01 -3.52554291e-01
-7.46912956e-02 2.68112361e-01 -3.07023793e-01 3.65080893e-01
9.20466632e-02 2.33736753e-01 -1.07199736e-01 -1.54573694e-01
2.19243646e-01 2.52268612e-01 -5.29218256e-01 -3.21544595e-02
5.70215523e-01 1.13273442e+00 -1.31306338e+00 -5.46827555e-01
1.38915300e-01 2.31268272e-01 -6.49547696e-01 6.42253220e-01
-7.48535514e-01 4.31891203e-01 -2.35634550e-01 7.06989825e-01
6.87349141e-02 -1.64709538e-01 -1.42192349e-01 4.52166349e-01
2.50094384e-01 -8.81313458e-02 -7.50849545e-01 1.70022500e+00
-2.32141808e-01 8.99190545e-01 -3.78423810e-01 -6.07212186e-01
6.50844455e-01 4.38432634e-01 3.25793177e-01 -8.02391768e-01
8.40654448e-02 -2.18530774e-01 -8.57840329e-02 -7.68984437e-01
7.68996060e-01 -7.06630051e-02 -1.84193403e-01 3.60164016e-01
3.20759527e-02 -4.54160869e-01 2.85892248e-01 2.57414192e-01
1.26459527e+00 8.74577343e-01 2.70548481e-02 3.31917316e-01
5.86407930e-02 6.89455807e-01 4.00337249e-01 7.82338381e-01
-2.52026379e-01 9.59492147e-01 8.04662526e-01 -3.07951808e-01
-1.79214036e+00 -1.16729915e+00 4.28741574e-01 9.47756410e-01
1.77768171e-01 -2.13199928e-01 -1.10573411e+00 -8.42544317e-01
1.91025082e-02 1.18691230e+00 -8.51413488e-01 -1.70613766e-01
-6.20076776e-01 -1.26792744e-01 9.21135664e-01 7.90759027e-01
2.97502995e-01 -1.62879550e+00 -1.02792001e+00 3.25451314e-01
-1.49508044e-01 -1.23256958e+00 -5.61934531e-01 -5.83706975e-01
-4.82713073e-01 -7.52112329e-01 -1.19477546e+00 -5.55140674e-01
7.35735416e-01 6.15076013e-02 1.26578152e+00 -7.33528584e-02
-4.58256602e-01 6.32296026e-01 -5.04182696e-01 -5.81313908e-01
-9.33635175e-01 -3.87819558e-01 -2.85628170e-01 -3.87030154e-01
-7.77444988e-02 -3.23488563e-01 -5.20015597e-01 3.11290503e-01
-8.39146972e-01 5.81863940e-01 5.52595258e-01 1.02995086e+00
2.14165285e-01 -5.13225138e-01 7.01480031e-01 -5.90611517e-01
1.01219785e+00 -2.30861306e-01 -4.79068995e-01 1.41286552e-01
-6.09967768e-01 2.48503610e-01 6.52035713e-01 -1.03040528e+00
-1.09726155e+00 -2.03863555e-03 1.57811895e-01 -7.41763473e-01
-2.84049332e-01 -7.04943165e-02 3.04012656e-01 2.28133559e-01
9.79354441e-01 1.46593228e-01 2.21456826e-01 4.15078178e-02
6.54387951e-01 2.21786708e-01 6.98031604e-01 -7.96106875e-01
7.87319720e-01 3.20918441e-01 -4.04732853e-01 -4.01034534e-01
-4.20205772e-01 1.47176921e-01 5.11945970e-02 -6.74312115e-01
1.06068587e+00 -8.59109938e-01 -7.94966519e-01 2.69434512e-01
-1.14816511e+00 -6.42157137e-01 -7.30018973e-01 2.60692686e-01
-1.23019695e+00 -3.80659886e-02 -4.16235805e-01 -1.06421137e+00
-1.52239621e-01 -1.15810466e+00 9.94698167e-01 3.21773350e-01
-6.57356799e-01 -3.67223382e-01 1.50854602e-01 4.10323471e-01
3.63511294e-01 7.53410697e-01 8.83535504e-01 -1.80036306e-01
-8.27438295e-01 -2.24811926e-01 -2.28523627e-01 2.93156594e-01
-2.95761764e-01 -5.78904711e-02 -8.22569728e-01 -1.46706820e-01
-6.50689363e-01 -9.38689172e-01 3.28888446e-01 7.26190582e-02
1.00783014e+00 -5.18117070e-01 -2.25634739e-01 2.31403202e-01
1.21274245e+00 2.37352833e-01 8.42211425e-01 4.76036906e-01
5.36229491e-01 5.06582141e-01 8.92953813e-01 5.52880704e-01
2.03433305e-01 6.62694871e-01 6.22832775e-01 -1.00254513e-01
-4.29775923e-01 -9.31960344e-01 7.78215408e-01 -1.27279237e-01
-2.38278285e-01 -3.06436718e-01 -6.49889827e-01 7.33660400e-01
-2.02784634e+00 -1.44477665e+00 2.59722650e-01 1.89894950e+00
9.81676757e-01 2.31334165e-01 5.91358781e-01 -1.78973943e-01
5.87421536e-01 1.44641966e-01 -8.60444784e-01 -6.67742252e-01
-2.25235801e-02 1.46724973e-02 4.10551608e-01 6.27810210e-02
-6.64528370e-01 1.02458727e+00 7.11977386e+00 7.45622039e-01
-1.10116148e+00 -1.12507910e-01 6.41319096e-01 -3.99658710e-01
-2.50085056e-01 -8.50859433e-02 -2.87693113e-01 4.04268086e-01
6.84039533e-01 -4.11183715e-01 8.38420391e-01 1.27831459e+00
1.61199033e-01 -2.26390678e-02 -1.30290544e+00 1.00827682e+00
2.35014752e-01 -1.40261698e+00 9.11252648e-02 -5.94832078e-02
9.47051764e-01 -3.93384725e-01 2.11281031e-01 6.46627069e-01
7.40135491e-01 -1.30286992e+00 1.31641090e+00 5.08455217e-01
1.02339697e+00 -7.72032142e-01 1.70006990e-01 3.88857007e-01
-5.85828543e-01 -1.27770588e-01 -3.43338251e-01 -9.30455476e-02
2.35954002e-01 -2.42522493e-01 -1.29592228e+00 7.26200789e-02
6.53861761e-01 4.53175306e-01 -3.37993532e-01 1.03510022e+00
-5.11112273e-01 6.07936203e-01 3.00139040e-01 -6.41223669e-01
2.84866363e-01 1.28559619e-01 7.05527008e-01 1.15605271e+00
3.29278827e-01 -1.69500977e-01 -3.71564217e-02 1.37129796e+00
-1.95573837e-01 -8.39592367e-02 -9.00479853e-01 -6.31938934e-01
7.69461244e-02 1.01854718e+00 -3.58416677e-01 -5.81140935e-01
-1.70850739e-01 1.15895188e+00 3.63256246e-01 4.75870520e-01
-1.37531781e+00 -8.41265693e-02 8.65556061e-01 2.70825207e-01
2.78066933e-01 -1.81755990e-01 -1.99197888e-01 -8.85466218e-01
-3.92693058e-02 -1.34387350e+00 -8.48647580e-02 -1.51257181e+00
-1.02638102e+00 5.91973424e-01 1.95221156e-01 -1.54269660e+00
-7.22790539e-01 -4.01556402e-01 -6.40562415e-01 3.83421451e-01
-1.01023114e+00 -1.29871690e+00 -2.65746444e-01 2.63766706e-01
8.12580645e-01 -3.41779143e-01 7.41580546e-01 -1.51863918e-01
-2.07703292e-01 7.16350794e-01 -9.54861641e-02 1.14772588e-01
6.49458826e-01 -1.29779613e+00 4.71426487e-01 6.75989568e-01
1.04372509e-01 -2.27960069e-02 1.08438683e+00 -6.49504542e-01
-1.22379637e+00 -8.81231010e-01 2.24545211e-01 -6.43689692e-01
5.64241111e-01 -4.78201658e-01 -5.32422721e-01 5.70428967e-01
7.36231983e-01 -4.17592078e-01 1.91221491e-01 -3.71073633e-01
-5.05873322e-01 2.82058686e-01 -1.20812333e+00 1.12641978e+00
1.22172451e+00 -3.92555416e-01 -5.47304630e-01 1.34377927e-01
9.64588583e-01 -4.46053833e-01 -5.21633744e-01 1.51955858e-02
7.43458748e-01 -8.74291480e-01 1.15551388e+00 -9.11051393e-01
1.13924122e+00 -1.63454369e-01 1.99188322e-01 -1.40279019e+00
-2.30631635e-01 -6.35005832e-01 -2.99639013e-02 1.13895082e+00
4.61020023e-01 -1.67600930e-01 1.18781590e+00 8.61097693e-01
2.93968413e-02 -5.55762827e-01 -4.98357415e-01 -1.09254551e+00
4.59664725e-02 -2.17964336e-01 5.58936536e-01 8.50874603e-01
1.92158401e-01 3.38084579e-01 -9.91165876e-01 -2.74110585e-01
5.24332464e-01 -9.69500374e-03 1.29225409e+00 -5.00400960e-01
-4.88612980e-01 -5.72414994e-01 -2.81968206e-01 -7.95205295e-01
2.15941757e-01 -5.63393652e-01 3.32508713e-01 -1.45202005e+00
3.70140076e-01 -1.24045648e-01 8.76846910e-02 5.44724941e-01
-2.07945928e-01 1.57190651e-01 5.02487779e-01 -4.07333449e-02
-5.73897243e-01 6.75653756e-01 1.59557414e+00 -4.13172126e-01
-2.56134244e-03 -4.33337331e-01 -5.94819963e-01 4.49835747e-01
7.99231768e-01 -7.50999987e-01 -5.22109568e-01 -4.05121267e-01
1.36907831e-01 3.72912914e-01 7.10651278e-01 -1.28176200e+00
-9.93092135e-02 -5.81308305e-01 4.10085976e-01 -3.07525694e-01
7.10743427e-01 -5.60204208e-01 3.06873739e-01 5.26801884e-01
-5.93195021e-01 1.67012751e-01 2.33706579e-01 5.70400357e-01
-7.57089406e-02 -3.21942329e-01 6.50221407e-01 -3.16073000e-01
-6.71111882e-01 2.53116395e-02 -5.90859592e-01 2.41406441e-01
1.23644459e+00 -5.06543279e-01 -4.22123641e-01 -8.69490683e-01
-4.05786574e-01 1.67796403e-01 8.99757683e-01 7.38699615e-01
7.14920878e-01 -1.60332870e+00 -8.87118101e-01 -2.41656333e-01
3.32241237e-01 -3.80188972e-01 2.25404017e-02 1.97729960e-01
-5.39389968e-01 -6.76180944e-02 -7.67761707e-01 -3.78071219e-01
-1.30255175e+00 8.39598596e-01 7.01119974e-02 -1.85647994e-01
-5.84475696e-01 8.10120761e-01 2.22359061e-01 -5.51341586e-02
3.70678037e-01 -1.49466574e-01 1.05026970e-02 -4.24197704e-01
4.70498323e-01 4.47033018e-01 -7.12688386e-01 -3.87045294e-01
2.47531943e-02 4.04748358e-02 7.29717985e-02 -6.49250925e-01
1.26421463e+00 4.34742808e-01 5.83466470e-01 1.88648328e-01
8.56203973e-01 -1.77874610e-01 -1.80102611e+00 2.08131179e-01
-1.93603992e-01 -6.69014156e-01 -4.86347288e-01 -1.11162293e+00
-5.56676805e-01 7.90145814e-01 4.62046236e-01 3.97666097e-02
8.53389680e-01 -4.68144193e-02 8.05176258e-01 3.22967231e-01
4.63394076e-01 -1.27552080e+00 8.92197073e-01 1.03340074e-01
1.33007526e+00 -1.16198599e+00 -2.07806423e-01 2.17446089e-02
-1.37796593e+00 7.17242956e-01 9.72280443e-01 -3.26322079e-01
-3.41804147e-01 2.94034988e-01 1.19878046e-01 1.83623031e-01
-8.97864401e-01 -1.44794777e-01 -4.43159975e-02 9.66742754e-01
1.19398221e-01 1.09907150e-01 -5.29689901e-02 2.23832160e-01
-2.68769145e-01 1.37594610e-01 6.92808807e-01 8.91116858e-01
-2.91810036e-01 -1.13592362e+00 -4.73856747e-01 1.98746040e-01
-3.02191466e-01 1.87608868e-01 -5.20414472e-01 9.05667365e-01
-2.23335519e-01 7.78307378e-01 -6.57069236e-02 -4.20069188e-01
4.95802641e-01 5.99481575e-02 7.63579190e-01 -3.51754427e-01
-7.41873443e-01 -4.43261296e-01 5.16478240e-01 -5.87145984e-01
-1.75984770e-01 -6.98486984e-01 -1.23834538e+00 -2.89689600e-01
2.83446182e-02 -1.37323812e-01 5.87866306e-01 4.84715074e-01
3.28887194e-01 6.38115346e-01 3.87194276e-01 -8.77017021e-01
-5.32416582e-01 -7.52168238e-01 -4.82966602e-02 9.67290342e-01
4.61368710e-02 -6.45073116e-01 1.21673830e-01 3.39775205e-01]
|
[11.159467697143555, 0.6141761541366577]
|
001972c2-82bc-4306-a46e-c235a7d5a2c5
|
codeie-large-code-generation-models-are
|
2305.05711
| null |
https://arxiv.org/abs/2305.05711v2
|
https://arxiv.org/pdf/2305.05711v2.pdf
|
CodeIE: Large Code Generation Models are Better Few-Shot Information Extractors
|
Large language models (LLMs) pre-trained on massive corpora have demonstrated impressive few-shot learning ability on many NLP tasks. A common practice is to recast the task into a text-to-text format such that generative LLMs of natural language (NL-LLMs) like GPT-3 can be prompted to solve it. However, it is nontrivial to perform information extraction (IE) tasks with NL-LLMs since the output of the IE task is usually structured and therefore is hard to be converted into plain text. In this paper, we propose to recast the structured output in the form of code instead of natural language and utilize generative LLMs of code (Code-LLMs) such as Codex to perform IE tasks, in particular, named entity recognition and relation extraction. In contrast to NL-LLMs, we show that Code-LLMs can be well-aligned with these IE tasks by designing code-style prompts and formulating these IE tasks as code generation tasks. Experiment results on seven benchmarks show that our method consistently outperforms fine-tuning moderate-size pre-trained models specially designed for IE tasks (e.g., UIE) and prompting NL-LLMs under few-shot settings. We further conduct a series of in-depth analyses to demonstrate the merits of leveraging Code-LLMs for IE tasks.
|
['Xipeng Qiu', 'Xuanjing Huang', 'Yuanbin Wu', 'Hang Yan', 'Qiong Tang', 'Tianxiang Sun', 'Peng Li']
|
2023-05-09
| null | null | null | null |
['uie', 'relation-extraction']
|
['computer-vision', 'natural-language-processing']
|
[ 4.75066483e-01 4.76433635e-01 -2.23792404e-01 -3.26194912e-01
-1.14896238e+00 -4.40443963e-01 6.18557632e-01 -7.97921047e-02
-6.11280501e-02 5.47192633e-01 2.50342488e-01 -8.87148976e-01
3.86322916e-01 -8.86637270e-01 -8.83759618e-01 -1.50476933e-01
8.54251534e-02 4.79427218e-01 7.43633360e-02 -2.51933664e-01
1.50206700e-01 -2.65021384e-01 -1.31979847e+00 4.46500897e-01
1.17666924e+00 5.37021697e-01 5.67718267e-01 8.94085884e-01
-7.40392864e-01 1.21376431e+00 -6.97889507e-01 -5.58152258e-01
-4.27884981e-02 -6.37566924e-01 -9.19492245e-01 1.33007718e-02
1.01203034e-02 -2.04106867e-01 -1.62514314e-01 9.61825490e-01
2.40298823e-01 2.70293266e-01 7.12105274e-01 -1.27763188e+00
-9.37421918e-01 9.87175167e-01 -4.23028529e-01 -8.12990889e-02
6.59272969e-01 2.99687922e-01 1.23208976e+00 -1.08994293e+00
7.89391041e-01 1.15551496e+00 4.69622433e-01 8.63300264e-01
-1.35264087e+00 -5.56765437e-01 -9.24044922e-02 -7.13551044e-02
-1.14738762e+00 -4.42316890e-01 4.02694315e-01 -5.69670975e-01
1.54064202e+00 6.54577613e-02 6.64077848e-02 1.44858730e+00
3.37725908e-01 1.02021790e+00 5.77483594e-01 -6.37324154e-01
4.24809128e-01 8.31536874e-02 2.12516263e-01 8.58402550e-01
1.40631095e-01 -1.84536815e-01 -2.45396882e-01 -3.68744403e-01
4.86128926e-01 -5.34421997e-03 -1.80601940e-01 -1.63219035e-01
-1.27227747e+00 1.17019343e+00 1.82095215e-01 1.89305276e-01
-6.28795475e-02 1.40856236e-01 6.15316510e-01 2.53016055e-01
3.69963706e-01 7.61558950e-01 -5.79219043e-01 -5.52502990e-01
-8.65143657e-01 9.83958170e-02 1.26873231e+00 1.64222443e+00
8.37669373e-01 4.97719161e-02 -6.31093323e-01 1.06942260e+00
-8.03305022e-03 2.24596605e-01 8.05446804e-01 -4.92322862e-01
1.09445858e+00 5.34871221e-01 -4.44831178e-02 -4.41114098e-01
-2.90614441e-02 -5.43136410e-02 -8.34671795e-01 -8.41637924e-02
9.79259759e-02 -6.69073403e-01 -1.08297801e+00 1.51993358e+00
-1.48351580e-01 1.80362418e-01 4.32344109e-01 3.85818034e-01
9.50365961e-01 1.20720100e+00 1.68911323e-01 -5.66751808e-02
1.30384958e+00 -1.29762816e+00 -3.73985767e-01 -7.51383245e-01
1.02534819e+00 -5.22136867e-01 1.52544200e+00 2.19909400e-02
-7.76620209e-01 -6.41422153e-01 -8.59887898e-01 -1.09744564e-01
-1.93324998e-01 1.45600736e-01 6.90957665e-01 2.86606431e-01
-8.16083372e-01 3.90333503e-01 -7.50714481e-01 -4.91669536e-01
2.86370337e-01 -2.36213710e-02 -7.41748139e-02 -9.14575979e-02
-1.19464755e+00 6.01319909e-01 4.76878941e-01 -4.48193282e-01
-1.08859611e+00 -7.51989126e-01 -1.47543883e+00 4.04809058e-01
5.90881228e-01 -5.08900166e-01 1.78007782e+00 -6.07148707e-01
-1.47222078e+00 7.41331697e-01 -3.70590597e-01 -5.01305819e-01
2.11107478e-01 -1.83299348e-01 -3.36861104e-01 -2.29091123e-01
3.09682548e-01 6.67807221e-01 6.94359660e-01 -9.67124820e-01
-4.71423447e-01 5.14467597e-01 2.10876882e-01 -1.57231927e-01
-3.81461293e-01 3.03660691e-01 -4.45171624e-01 -6.28457844e-01
-4.92679834e-01 -8.54395926e-01 -4.22395319e-01 -4.18267578e-01
-7.21784592e-01 -4.87815678e-01 5.60116172e-01 -4.45869952e-01
1.43758023e+00 -2.09418869e+00 4.16344125e-03 -2.05684587e-01
2.37065196e-01 5.17541707e-01 -5.22737086e-01 5.90973616e-01
-1.17287122e-01 2.60533810e-01 -4.58680540e-01 -3.61978740e-01
1.52648732e-01 4.04293925e-01 -7.41056204e-01 -2.27230415e-01
6.13721073e-01 1.38062978e+00 -1.16884434e+00 -5.45992315e-01
-1.30074427e-01 3.01067121e-02 -7.42021978e-01 8.57511818e-01
-7.49090374e-01 1.29723419e-02 -5.70814431e-01 4.83567774e-01
1.06718272e-01 -6.13594949e-01 -7.80231580e-02 3.69053751e-01
2.40105279e-02 5.04191279e-01 -7.92248607e-01 1.84674180e+00
-9.99832451e-01 6.63224459e-01 -3.96869332e-01 -8.79543245e-01
1.09323406e+00 4.17781770e-01 -1.21261731e-01 -4.40533102e-01
-1.06804192e-01 -9.28355660e-03 1.00573328e-04 -8.01223338e-01
4.47748810e-01 4.68824990e-02 -6.62337482e-01 8.35105658e-01
5.02910316e-01 -2.40998432e-01 6.22058868e-01 6.00899518e-01
1.51007545e+00 8.69243443e-02 6.47990525e-01 1.01925716e-01
1.85334086e-01 1.56709075e-01 4.57337469e-01 9.70411062e-01
1.99991375e-01 6.72953606e-01 6.92148387e-01 -2.01631993e-01
-1.11010766e+00 -8.83519113e-01 9.80214998e-02 1.49408305e+00
-3.41240168e-01 -8.23015451e-01 -7.65204847e-01 -9.37373579e-01
-2.88693666e-01 1.22262871e+00 -5.34265220e-01 -2.22373590e-01
-3.42017502e-01 -6.97480440e-01 6.68679535e-01 6.54772580e-01
2.74953127e-01 -1.34658015e+00 -6.15225017e-01 5.74666917e-01
-1.70676142e-01 -1.08330011e+00 -8.06403339e-01 5.21164834e-01
-4.99628991e-01 -9.04523849e-01 -7.59222090e-01 -1.02701128e+00
8.50455642e-01 8.07812586e-02 1.45485616e+00 1.00721449e-01
-3.05985779e-01 1.13904454e-01 -7.80105412e-01 -3.21934700e-01
-1.01786304e+00 4.11932826e-01 -3.45335931e-01 -3.61222953e-01
4.13723648e-01 -5.52020013e-01 -3.83983888e-02 -2.93056667e-02
-1.02797532e+00 4.99718994e-01 8.63019049e-01 9.31200445e-01
6.03740998e-02 -3.27132642e-01 4.62950945e-01 -1.40403962e+00
9.38897491e-01 -6.86235547e-01 -4.59162056e-01 6.69098675e-01
-4.70414460e-01 4.96790469e-01 8.88376653e-01 -7.75099933e-01
-1.31556273e+00 1.12004034e-01 -3.62315089e-01 -1.22477323e-01
-3.33821893e-01 7.98342168e-01 -9.76296235e-03 2.15743661e-01
8.91000450e-01 5.13400912e-01 -5.15383065e-01 -3.50165248e-01
5.81365943e-01 1.05472445e+00 5.11452973e-01 -9.00900364e-01
9.92289245e-01 -3.29000384e-01 -6.03696525e-01 -5.93518734e-01
-1.02151632e+00 -4.08987880e-01 -4.16475892e-01 3.51756722e-01
8.60761762e-01 -9.36433196e-01 -9.69281346e-02 1.38998196e-01
-1.43170834e+00 -6.83582187e-01 -1.68046445e-01 2.18249555e-03
-6.40299141e-01 2.71743059e-01 -9.89124358e-01 -6.09146059e-01
-6.61133826e-01 -1.11723590e+00 1.12478435e+00 3.04631770e-01
-6.42196715e-01 -1.07359099e+00 3.43726218e-01 -1.22786406e-02
4.44996715e-01 -9.57461372e-02 1.29793811e+00 -9.98806655e-01
-2.51260191e-01 -7.14475960e-02 -2.22361118e-01 3.73674035e-02
4.61643375e-02 1.40075222e-01 -7.82989740e-01 -7.96576682e-03
-1.27640024e-01 -8.02700043e-01 5.84522903e-01 -1.48800701e-01
1.04599643e+00 -5.61592996e-01 -3.86812598e-01 6.74899757e-01
1.34542298e+00 3.59041274e-01 6.33296609e-01 1.09946646e-01
6.68311238e-01 3.12833518e-01 5.57274282e-01 5.73748648e-01
4.39388603e-01 4.69236434e-01 -3.22998911e-02 1.10590495e-01
-1.23836873e-02 -6.27466738e-01 6.76254094e-01 1.04150307e+00
4.75990117e-01 -1.60480678e-01 -1.23433149e+00 5.93581080e-01
-1.88534975e+00 -7.96389580e-01 -2.56595723e-02 1.72151732e+00
1.53401983e+00 1.47223711e-01 -3.27249080e-01 -5.22455037e-01
6.05148375e-01 4.65608276e-02 -4.63678241e-01 -5.51431656e-01
3.96514893e-01 3.91177922e-01 3.76615115e-02 2.52582341e-01
-9.96935546e-01 1.03543389e+00 5.66910839e+00 9.36270654e-01
-8.93290341e-01 1.60870343e-01 6.23072743e-01 2.11068660e-01
-4.08070385e-01 2.95100272e-01 -1.04955375e+00 5.55958748e-01
1.30424070e+00 -6.99541748e-01 5.16053796e-01 1.38912022e+00
-1.75979525e-01 9.91965234e-02 -1.41790295e+00 8.95017028e-01
-3.94796915e-02 -1.34168506e+00 6.65052459e-02 -2.60533601e-01
1.10697627e+00 2.13186756e-01 -4.65372294e-01 1.27314198e+00
1.00554693e+00 -8.82459998e-01 5.69707811e-01 1.82548642e-01
1.10077167e+00 -4.26782876e-01 4.65355754e-01 7.81215549e-01
-1.34505737e+00 -1.00818716e-01 -6.67499483e-01 -1.05177552e-01
2.36092463e-01 5.78814089e-01 -1.09207058e+00 3.05989146e-01
3.17103714e-01 5.73328376e-01 -7.01933146e-01 1.01127183e+00
-5.88109732e-01 6.88512623e-01 8.92332494e-02 -2.23753974e-01
4.08102781e-01 1.60129666e-01 2.08755627e-01 1.56483841e+00
6.20184362e-01 5.52219376e-02 2.91582614e-01 1.44663274e+00
-3.22268248e-01 1.05139308e-01 -7.69062281e-01 -6.61937475e-01
3.85973394e-01 1.52042067e+00 -5.01152873e-01 -8.04735303e-01
-9.19704854e-01 9.80919063e-01 5.53566575e-01 5.00918746e-01
-8.33922327e-01 -9.67737556e-01 5.34546137e-01 -1.94522336e-01
3.43183249e-01 -5.59795275e-02 -2.04568207e-02 -1.65476632e+00
-2.18611583e-01 -9.43695605e-01 2.44567767e-01 -1.00984240e+00
-1.24117351e+00 9.44630384e-01 -3.89669463e-02 -1.24813819e+00
-8.24467182e-01 -4.55027580e-01 -1.22022116e+00 7.99617290e-01
-1.13862526e+00 -8.21371019e-01 -1.86072379e-01 3.84446472e-01
1.19675612e+00 -1.81052551e-01 1.02520359e+00 -1.19615095e-02
-6.81111813e-01 6.10857666e-01 -8.08371231e-02 6.37986243e-01
6.07570350e-01 -1.35684872e+00 1.17349017e+00 9.98856962e-01
4.18759376e-01 8.61730516e-01 6.27371073e-01 -7.76688755e-01
-1.41408384e+00 -1.53578115e+00 1.09846127e+00 -4.83476877e-01
1.02384758e+00 -8.85272264e-01 -9.59209919e-01 8.33759487e-01
3.66739243e-01 -3.24971275e-03 7.02231169e-01 1.76916644e-02
-4.45661634e-01 4.96328801e-01 -5.87501168e-01 6.68405473e-01
9.83157575e-01 -6.87360466e-01 -1.01947010e+00 3.89138520e-01
1.26701856e+00 -5.71983993e-01 -6.43660903e-01 4.93293628e-02
-2.59356033e-02 -3.57651353e-01 6.71611965e-01 -8.68006110e-01
1.06299937e+00 -1.83231160e-02 1.36060908e-01 -1.54763198e+00
-1.61732301e-01 -7.95888603e-01 -4.16921198e-01 1.54518318e+00
6.44720614e-01 -1.68830350e-01 4.64471072e-01 8.44363332e-01
-2.03833356e-01 -6.78647697e-01 -4.09824103e-01 -8.34049225e-01
-5.98333403e-02 -5.76700866e-01 3.34440380e-01 8.42527628e-01
5.30381083e-01 8.43724966e-01 -4.24230129e-01 -3.37385923e-01
2.87989229e-01 5.00729263e-01 9.77180958e-01 -1.03204918e+00
-7.93267488e-01 -1.68466330e-01 2.34495491e-01 -1.18458319e+00
3.99685591e-01 -1.15066695e+00 6.85273945e-01 -1.63237000e+00
6.10882938e-01 -3.18954647e-01 2.16013312e-01 8.71703565e-01
-5.13589263e-01 -3.52983922e-01 2.17993200e-01 1.60118833e-01
-8.08838606e-01 4.80430365e-01 9.48472321e-01 -3.41625959e-01
-2.46493027e-01 5.33604957e-02 -7.13105261e-01 3.96174878e-01
4.99058217e-01 -6.70631051e-01 -5.23615599e-01 -2.42729574e-01
3.71954501e-01 4.32799786e-01 -7.72978663e-02 -9.33512270e-01
3.46015751e-01 -3.22474658e-01 -7.45522678e-02 -1.23762846e-01
-2.78874576e-01 -3.07100296e-01 -2.44290859e-01 2.54331470e-01
-5.70544839e-01 -2.21962452e-01 1.06734738e-01 4.16753590e-01
-2.27480575e-01 -7.44748652e-01 3.62352043e-01 -4.42296445e-01
-1.06186986e+00 3.41137201e-01 -5.04575908e-01 5.71742654e-01
9.69710171e-01 2.58115143e-01 -5.11590123e-01 -4.61432606e-01
-4.23022687e-01 2.28461072e-01 3.97960722e-01 6.11849070e-01
6.54191971e-01 -1.14727950e+00 -5.41396499e-01 2.07506254e-01
6.55609727e-01 1.48916170e-01 -9.23332796e-02 6.29362583e-01
-2.14810476e-01 5.49666226e-01 5.50693311e-02 -2.24435687e-01
-9.74931061e-01 8.25772107e-01 -1.08600162e-01 -6.48311317e-01
-7.02981651e-01 9.38783824e-01 4.49652821e-01 -5.69500685e-01
2.23673686e-01 -5.07824659e-01 7.21266940e-02 -1.66357934e-01
7.69253850e-01 -2.92356908e-01 -1.10058211e-01 -9.22319666e-02
-1.34657482e-02 5.85576408e-02 -2.88916111e-01 6.61135167e-02
1.51734149e+00 1.75296605e-01 -6.29551932e-02 3.14786077e-01
1.33638382e+00 -2.24175975e-01 -1.42275798e+00 -4.53036338e-01
3.82733524e-01 -1.24908641e-01 -4.52321798e-01 -6.06761336e-01
-7.21554279e-01 1.09013915e+00 -2.79530406e-01 4.34617652e-03
7.86136448e-01 3.25601220e-01 1.08896112e+00 8.11293244e-01
5.87983370e-01 -9.11939025e-01 3.57783049e-01 9.36367810e-01
7.35804379e-01 -1.18300796e+00 -6.02288246e-01 -1.26320854e-01
-8.81649137e-01 1.27933812e+00 8.81576300e-01 6.52743271e-03
2.09601656e-01 6.97718740e-01 -1.26548588e-01 1.04893092e-02
-1.36263549e+00 -2.81850904e-01 2.15504274e-01 5.44556201e-01
6.06148005e-01 -1.69790059e-01 4.48191464e-02 1.01867592e+00
-1.64449111e-01 1.05048269e-01 7.13662505e-01 1.16529548e+00
-5.01167238e-01 -1.28206599e+00 -1.20017104e-01 7.89312482e-01
-2.28637531e-01 -6.30557239e-01 -1.52959988e-01 4.45506752e-01
-2.35931724e-01 8.74979019e-01 -1.43292993e-02 -5.24290562e-01
6.93350360e-02 4.77932721e-01 5.86015359e-02 -1.52751529e+00
-4.95572537e-01 -2.28923634e-01 1.82794586e-01 -4.58893001e-01
1.65634841e-01 -3.60213280e-01 -1.43041456e+00 -7.01745553e-03
-1.58643275e-01 2.21949294e-01 1.97326735e-01 1.05274045e+00
2.75305241e-01 6.15347147e-01 3.53626907e-01 -7.19708264e-01
-6.79777205e-01 -1.11792254e+00 -2.07150921e-01 4.99831975e-01
8.22476745e-02 -2.66836286e-01 -2.68550158e-01 4.05924380e-01]
|
[7.883000373840332, 7.918665409088135]
|
96df1cad-304e-43a2-97a9-0b125d734ebf
|
on-learning-universal-representations-across
|
2007.1596
| null |
https://arxiv.org/abs/2007.15960v4
|
https://arxiv.org/pdf/2007.15960v4.pdf
|
On Learning Universal Representations Across Languages
|
Recent studies have demonstrated the overwhelming advantage of cross-lingual pre-trained models (PTMs), such as multilingual BERT and XLM, on cross-lingual NLP tasks. However, existing approaches essentially capture the co-occurrence among tokens through involving the masked language model (MLM) objective with token-level cross entropy. In this work, we extend these approaches to learn sentence-level representations and show the effectiveness on cross-lingual understanding and generation. Specifically, we propose a Hierarchical Contrastive Learning (HiCTL) method to (1) learn universal representations for parallel sentences distributed in one or multiple languages and (2) distinguish the semantically-related words from a shared cross-lingual vocabulary for each sentence. We conduct evaluations on two challenging cross-lingual tasks, XTREME and machine translation. Experimental results show that the HiCTL outperforms the state-of-the-art XLM-R by an absolute gain of 4.2% accuracy on the XTREME benchmark as well as achieves substantial improvements on both of the high-resource and low-resource English-to-X translation tasks over strong baselines.
|
['Yue Hu', 'Rongxiang Weng', 'Weihua Luo', 'Heng Yu', 'Luxi Xing', 'Xiangpeng Wei']
|
2020-07-31
| null |
https://openreview.net/forum?id=Uu1Nw-eeTxJ
|
https://openreview.net/pdf?id=Uu1Nw-eeTxJ
|
iclr-2021-1
|
['cross-lingual-natural-language-inference']
|
['natural-language-processing']
|
[ 1.34781480e-01 -2.89091736e-01 -5.77640891e-01 -5.73781312e-01
-2.03841233e+00 -5.58983028e-01 7.53123045e-01 -1.07998520e-01
-3.29761147e-01 1.00022912e+00 4.12029564e-01 -7.54223049e-01
5.32279372e-01 -3.61427575e-01 -1.04804516e+00 -2.85891473e-01
9.53923240e-02 5.80121040e-01 -6.31119430e-01 -3.10872018e-01
-1.98259860e-01 -5.85114993e-02 -8.29639494e-01 6.49760306e-01
1.34935534e+00 6.64743781e-01 3.88916224e-01 2.03008801e-01
-3.39260578e-01 5.70609868e-01 -4.00293797e-01 -7.22230196e-01
1.91344753e-01 -6.08981013e-01 -9.43646371e-01 -2.70059854e-01
5.12419403e-01 1.03781402e-01 -2.40941328e-04 8.92195344e-01
5.16004682e-01 -1.26627058e-01 6.90322280e-01 -8.55662286e-01
-1.21862745e+00 1.27149796e+00 -7.76802242e-01 2.08636045e-01
1.57322094e-01 1.22574955e-01 1.58084822e+00 -1.37191653e+00
5.60653865e-01 1.52861035e+00 5.92165589e-01 5.21765411e-01
-1.37078249e+00 -9.99786258e-01 2.25415498e-01 2.18685538e-01
-1.64596117e+00 -6.68876350e-01 4.83812392e-01 -1.77225426e-01
1.48183048e+00 5.42393439e-02 -2.54123420e-01 1.33267534e+00
4.17673290e-01 1.22267246e+00 1.52025068e+00 -7.90774703e-01
-2.90173143e-01 1.50909290e-01 -2.33496413e-01 5.76169789e-01
-1.10668175e-01 7.16282576e-02 -7.29797959e-01 -2.21403148e-02
2.79138744e-01 -5.24927735e-01 -2.36059725e-01 2.55955756e-01
-1.51280773e+00 1.03214693e+00 3.29639792e-01 5.91520369e-01
-3.16657275e-01 3.46952751e-02 7.25861549e-01 4.74278122e-01
9.36261058e-01 5.37883878e-01 -9.74953651e-01 8.36942121e-02
-1.00290716e+00 -1.79309294e-01 5.10716558e-01 1.05282414e+00
7.62966812e-01 2.52900332e-01 -3.98689955e-01 1.07485235e+00
2.77073920e-01 7.22568154e-01 9.99910116e-01 -2.09470093e-01
1.20324314e+00 2.27472514e-01 -3.82029861e-01 -3.65969986e-01
1.57945037e-01 -5.64437687e-01 -7.50410020e-01 -5.25271118e-01
-1.25148982e-01 -2.27988228e-01 -7.76096702e-01 2.10043859e+00
-1.13981605e-01 2.01005209e-02 6.76373363e-01 4.47321385e-01
7.39846408e-01 1.10044014e+00 2.57500499e-01 -1.85224071e-01
1.29416883e+00 -1.24542117e+00 -7.06368268e-01 -5.97484767e-01
1.04722846e+00 -1.04582298e+00 1.37859368e+00 -1.34721205e-01
-1.13858080e+00 -8.74543905e-01 -1.07090020e+00 -2.86282271e-01
-4.38114017e-01 6.92123413e-01 3.47517937e-01 3.49856287e-01
-8.14833760e-01 5.47503769e-01 -8.04222703e-01 -6.28870502e-02
3.78258042e-02 3.20472680e-02 -2.85490155e-01 -3.81853700e-01
-1.74917471e+00 1.14688706e+00 5.11288524e-01 -1.90719198e-02
-7.55267441e-01 -9.38856184e-01 -1.13508987e+00 -2.89207548e-02
2.31086332e-02 -4.60204840e-01 1.25602758e+00 -1.04099858e+00
-1.43389297e+00 1.11915541e+00 -4.80209559e-01 -5.30296743e-01
2.93076813e-01 -6.33554518e-01 -5.60910583e-01 -1.75252631e-01
6.66933358e-01 6.85031354e-01 3.22627097e-01 -1.28128147e+00
-5.11447370e-01 -8.34037438e-02 -3.39643627e-01 4.83559102e-01
-1.61887467e-01 2.88337916e-01 -3.39824289e-01 -8.35891008e-01
-2.10532069e-01 -9.28747475e-01 -1.45003293e-02 -7.45667934e-01
-6.00478232e-01 -5.36503851e-01 4.16507304e-01 -1.12373316e+00
1.02380085e+00 -1.82468617e+00 1.60717607e-01 -4.26719189e-01
-5.98615289e-01 2.31869325e-01 -4.55305278e-01 7.09104955e-01
-4.38999608e-02 2.49390796e-01 -2.76765615e-01 -9.74294543e-01
1.87196538e-01 3.72579992e-01 -7.73130000e-01 2.16250300e-01
4.73532468e-01 1.25327456e+00 -8.19593310e-01 -5.19793510e-01
9.33069065e-02 4.77972150e-01 -1.57612309e-01 1.96526393e-01
-2.06421927e-01 5.41105390e-01 -1.12531930e-01 5.17768979e-01
3.35467547e-01 -1.33804351e-01 4.41494256e-01 -4.63459976e-02
-1.58292323e-01 1.25523221e+00 -3.89496922e-01 1.99074924e+00
-1.22474861e+00 4.09499466e-01 -3.40324998e-01 -7.67210722e-01
8.54272962e-01 6.39264643e-01 1.91416070e-01 -7.31872976e-01
-7.97999725e-02 6.14712417e-01 -1.83245569e-01 -6.44949824e-02
6.01681650e-01 -3.15841466e-01 -5.10744393e-01 5.17783463e-01
3.95738393e-01 -4.51241322e-02 5.48315197e-02 6.06714785e-02
4.82705653e-01 4.18790013e-01 5.84197998e-01 -4.78024900e-01
5.80730617e-01 -8.13885033e-02 6.12262964e-01 5.32072783e-01
2.24615589e-01 4.56043303e-01 -1.80198923e-02 -6.11917898e-02
-7.76820481e-01 -1.16431940e+00 -1.39783978e-01 1.21665430e+00
-8.65302831e-02 -5.58142900e-01 -5.31943440e-01 -8.42861652e-01
-1.41603619e-01 1.24933696e+00 -3.55391979e-01 -7.99150094e-02
-8.27165306e-01 -9.03662741e-01 7.80637860e-01 5.24771273e-01
4.42162365e-01 -1.10142088e+00 3.29283506e-01 3.21059227e-01
-7.75462091e-01 -1.46239507e+00 -8.84491742e-01 2.18033940e-01
-6.00785255e-01 -3.84282053e-01 -4.45210844e-01 -1.07924366e+00
3.62439275e-01 9.54950601e-02 1.56299722e+00 -4.12654966e-01
2.57227812e-02 -4.73198034e-02 -4.55716908e-01 -1.07147425e-01
-6.66751325e-01 5.50197184e-01 2.08821878e-01 6.35818616e-02
5.92859864e-01 -2.37316430e-01 -9.92627293e-02 4.84896526e-02
-4.33614641e-01 1.52965054e-01 8.21357250e-01 9.99245524e-01
1.01762617e+00 -4.53067809e-01 9.09076512e-01 -8.28384638e-01
6.52909458e-01 -6.57566845e-01 -3.99217576e-01 6.23944819e-01
-6.09269857e-01 2.61833161e-01 9.65125978e-01 -2.78020680e-01
-1.15930021e+00 -1.81617841e-01 -1.96272284e-01 -2.75703013e-01
1.79200053e-01 9.03169334e-01 -3.11634034e-01 5.08895397e-01
3.32165956e-01 6.52679324e-01 -5.82450688e-01 -6.11920238e-01
7.86616445e-01 8.13094020e-01 5.16420782e-01 -8.93304884e-01
6.71016574e-01 -4.79323044e-02 -5.63400149e-01 -5.21414757e-01
-1.22125137e+00 -3.12239677e-01 -8.03677261e-01 3.10906172e-01
9.21758294e-01 -1.47709513e+00 5.73253967e-02 6.31904155e-02
-1.41914845e+00 -2.64441580e-01 -6.77303672e-02 7.46336222e-01
-5.00439107e-01 2.24360570e-01 -8.37547243e-01 -3.84207904e-01
-7.34565496e-01 -1.24945402e+00 1.41109848e+00 -3.30671579e-01
-2.11153105e-01 -1.24820447e+00 2.46040046e-01 4.90970820e-01
3.70474577e-01 -2.39285350e-01 1.15040421e+00 -7.03166723e-01
-5.23967981e-01 -7.34840694e-04 -7.81434700e-02 6.40505552e-01
3.36159587e-01 -3.55653346e-01 -8.98761511e-01 -5.56646407e-01
-1.09897584e-01 -7.54701376e-01 8.33455563e-01 1.98124260e-01
8.55540335e-01 -4.32453632e-01 -2.66433656e-01 6.71873450e-01
1.51468050e+00 -2.36093178e-01 3.59442025e-01 6.81761727e-02
7.07479179e-01 5.44878125e-01 5.57605386e-01 -1.64346933e-01
7.26527154e-01 8.89914930e-01 -1.61510244e-01 -2.78833568e-01
-3.18835586e-01 -6.25623405e-01 9.72727716e-01 1.80617845e+00
3.79587680e-01 -2.43894905e-01 -8.81881833e-01 8.59009027e-01
-1.66953301e+00 -6.63892269e-01 9.69672501e-02 2.06349158e+00
1.53846085e+00 -8.87344629e-02 -4.92735773e-01 -5.77057004e-01
7.19436765e-01 1.90470472e-01 -3.63033921e-01 -5.93659282e-01
-4.04597938e-01 4.72986102e-01 4.55343068e-01 7.28638589e-01
-1.09399056e+00 1.55762196e+00 5.92631149e+00 1.24036300e+00
-1.33547032e+00 5.49326301e-01 7.14019954e-01 3.98821644e-02
-3.74902308e-01 2.87211849e-04 -1.29994762e+00 3.32352728e-01
1.25243199e+00 -5.33245325e-01 3.05121541e-01 8.52974176e-01
-5.30756526e-02 5.22540271e-01 -1.24720478e+00 8.25922966e-01
3.40070426e-01 -1.07744062e+00 2.06953242e-01 -6.17436022e-02
1.26902580e+00 8.03802729e-01 1.31908089e-01 8.93625140e-01
6.96366549e-01 -1.14155412e+00 5.86989880e-01 -7.48734325e-02
1.16614604e+00 -7.61786163e-01 6.41987860e-01 3.54396433e-01
-1.40515709e+00 6.01644218e-01 -3.70317012e-01 2.67555714e-01
4.16435570e-01 3.19273978e-01 -9.49747384e-01 1.05085087e+00
4.08076227e-01 8.66622448e-01 -3.53331923e-01 2.18728513e-01
-5.51640630e-01 1.01939082e+00 3.28448452e-02 1.73377663e-01
5.82119346e-01 -1.09675497e-01 2.87024289e-01 1.58423734e+00
4.06484872e-01 -4.89253372e-01 4.11184698e-01 9.66660798e-01
-6.40065908e-01 6.88057840e-01 -5.36893427e-01 -1.71947122e-01
4.91207480e-01 1.09646630e+00 -8.17423034e-03 -6.20790422e-01
-8.27342808e-01 1.26901197e+00 6.53533041e-01 2.55839646e-01
-7.24164367e-01 -1.37509137e-01 8.92336905e-01 -4.19161558e-01
4.43227649e-01 -3.21454167e-01 -2.62090623e-01 -1.45917797e+00
2.18801737e-01 -1.04147160e+00 1.87969953e-01 -4.93399143e-01
-1.90335715e+00 1.19244802e+00 -2.03847751e-01 -1.32588863e+00
-6.93411827e-01 -5.41765690e-01 -3.76144469e-01 1.54490435e+00
-2.08330870e+00 -1.77274132e+00 6.07125938e-01 4.69066709e-01
1.01085246e+00 -4.12603080e-01 1.17835152e+00 2.88298041e-01
-4.45674747e-01 9.25996065e-01 5.54396451e-01 3.95021558e-01
9.23872232e-01 -1.19749177e+00 8.86132658e-01 9.42710578e-01
6.02758884e-01 7.64619291e-01 1.28987357e-01 -7.02237010e-01
-1.18863893e+00 -1.55324650e+00 1.75317430e+00 -5.82825005e-01
7.94280887e-01 -6.93726420e-01 -9.25088584e-01 1.07541871e+00
7.46158838e-01 -1.47666901e-01 9.21359718e-01 3.59276086e-01
-6.64770663e-01 1.25363341e-03 -6.19903982e-01 6.86352432e-01
6.35371089e-01 -9.94089901e-01 -8.04224610e-01 6.44540370e-01
1.01669037e+00 -2.09969684e-01 -9.40490007e-01 5.06205380e-01
2.36633018e-01 -3.33243400e-01 7.94182718e-01 -8.60946715e-01
7.26297617e-01 -1.51308533e-03 -6.42505884e-01 -1.73489952e+00
-6.75009936e-02 -5.67760944e-01 2.81868309e-01 1.55712628e+00
9.42876458e-01 -5.28070092e-01 -7.32069230e-03 -5.69937117e-02
-3.59286129e-01 -8.79950404e-01 -1.11560917e+00 -1.21148527e+00
8.04095447e-01 -4.54837501e-01 5.56343138e-01 1.19701624e+00
2.33681053e-02 9.49851692e-01 -6.16185486e-01 2.00852484e-01
4.11190718e-01 4.74952161e-01 4.89006877e-01 -5.42136371e-01
-5.19634724e-01 -3.01002860e-01 1.26434028e-01 -1.34374666e+00
8.37620139e-01 -1.64804387e+00 3.33820760e-01 -1.30962133e+00
1.59558028e-01 -3.79437029e-01 -4.25772309e-01 4.59349811e-01
-6.02431595e-01 1.51413545e-01 6.51950240e-02 2.44092718e-01
-4.13281560e-01 9.97927785e-01 8.66575539e-01 -1.64178342e-01
8.72590318e-02 -3.59706193e-01 -7.57669151e-01 3.52773160e-01
7.60122716e-01 -6.18601322e-01 -1.98581651e-01 -1.08514631e+00
-1.67341784e-01 -4.79598157e-02 -2.58140266e-01 -5.25034368e-01
-2.40654022e-01 5.51006682e-02 1.17992543e-01 -6.31898820e-01
3.60859245e-01 -2.45470420e-01 -3.42155010e-01 2.93676198e-01
-7.46543944e-01 3.69030893e-01 3.42132419e-01 1.49421275e-01
-5.86621404e-01 -1.31997373e-02 5.87381542e-01 -2.61137754e-01
-4.32313710e-01 2.29334161e-01 2.63498630e-02 4.05998677e-01
6.26747608e-01 4.76722836e-01 -2.45898664e-01 -2.52913773e-01
-1.35068431e-01 2.63229817e-01 1.19831488e-01 9.63745773e-01
3.43511432e-01 -1.67623031e+00 -1.38649464e+00 2.54345387e-01
4.46438253e-01 -4.26729381e-01 -5.89663908e-02 5.74139953e-01
3.06376182e-02 8.19448471e-01 1.07783325e-01 -5.15779972e-01
-1.05737710e+00 3.88813734e-01 1.93839997e-01 -7.79287755e-01
-2.77041316e-01 8.88166904e-01 5.31907439e-01 -9.79100525e-01
-2.17774883e-01 -1.70545518e-01 2.34744787e-01 -2.94045091e-01
4.05754030e-01 -1.96096733e-01 1.72135949e-01 -1.11748886e+00
-4.81821686e-01 5.78164279e-01 -4.99145508e-01 -3.17924887e-01
1.11865091e+00 -3.01313341e-01 -1.60326198e-01 7.23447204e-01
1.53819442e+00 1.57995179e-01 -8.98376584e-01 -8.37902129e-01
1.52666628e-01 -1.08265251e-01 -6.03200309e-02 -9.83467579e-01
-8.29507709e-01 1.08366418e+00 7.68875033e-02 -4.18108463e-01
7.74693131e-01 2.54540741e-01 1.12326598e+00 2.22184718e-01
3.81301641e-01 -1.00734305e+00 -1.58088297e-01 7.51953542e-01
1.07553804e+00 -1.48344588e+00 -2.62655526e-01 -2.45747983e-01
-1.02742445e+00 7.74353206e-01 5.86508155e-01 -8.89435336e-02
3.66526037e-01 1.73670262e-01 3.94198895e-01 1.96253225e-01
-1.11219800e+00 -1.90098658e-01 6.12841904e-01 2.05274850e-01
1.02000606e+00 4.54942763e-01 -3.35865766e-01 6.17849350e-01
-4.31848049e-01 -4.12992954e-01 -1.28336444e-01 5.57316840e-01
4.86849109e-03 -1.41754043e+00 -4.78668511e-03 8.32263231e-02
-7.20354855e-01 -9.66704905e-01 -4.36062217e-01 7.15482950e-01
-1.38696330e-02 9.43464100e-01 -9.06471983e-02 -2.69358397e-01
4.69161980e-02 4.02246445e-01 4.82796639e-01 -8.03235650e-01
-6.18600607e-01 7.30720628e-03 1.62832439e-01 -3.01632494e-01
-1.88565418e-01 -8.46447289e-01 -9.82533276e-01 -4.27552685e-03
-3.10889602e-01 2.41936445e-01 8.12986135e-01 1.11347210e+00
4.48238015e-01 4.44887489e-01 7.99317181e-01 -6.04076684e-01
-8.58330905e-01 -1.43872476e+00 -1.83541551e-01 4.07771647e-01
1.16478719e-01 -2.15597436e-01 -2.07358524e-01 1.50717765e-01]
|
[11.270623207092285, 9.88017463684082]
|
35cfe365-f747-4b29-a02f-0f45b45fffed
|
villandiffusion-a-unified-backdoor-attack
|
2306.06874
| null |
https://arxiv.org/abs/2306.06874v2
|
https://arxiv.org/pdf/2306.06874v2.pdf
|
VillanDiffusion: A Unified Backdoor Attack Framework for Diffusion Models
|
Diffusion Models (DMs) are state-of-the-art generative models that learn a reversible corruption process from iterative noise addition and denoising. They are the backbone of many generative AI applications, such as text-to-image conditional generation. However, recent studies have shown that basic unconditional DMs (e.g., DDPM and DDIM) are vulnerable to backdoor injection, a type of output manipulation attack triggered by a maliciously embedded pattern at model input. This paper presents a unified backdoor attack framework (VillanDiffusion) to expand the current scope of backdoor analysis for DMs. Our framework covers mainstream unconditional and conditional DMs (denoising-based and score-based) and various training-free samplers for holistic evaluations. Experiments show that our unified framework facilitates the backdoor analysis of different DM configurations and provides new insights into caption-based backdoor attacks on DMs.
|
['Tsung-Yi Ho', 'Pin-Yu Chen', 'Sheng-Yen Chou']
|
2023-06-12
| null | null | null | null |
['backdoor-attack']
|
['adversarial']
|
[ 6.85817838e-01 -9.96868312e-02 -1.90909822e-02 3.04945081e-01
-1.19552565e+00 -1.24487019e+00 1.29858053e+00 -5.49862921e-01
1.10068098e-01 5.31862855e-01 1.44169638e-02 -7.57740319e-01
3.18905979e-01 -1.05022633e+00 -1.02209044e+00 -1.07730389e+00
-4.66963463e-03 2.68281430e-01 2.27002650e-01 -1.13835521e-01
7.18202144e-02 3.96502256e-01 -8.48244786e-01 3.11439335e-01
6.43155575e-01 3.54288340e-01 -2.29627848e-01 1.13332057e+00
3.90426964e-01 7.81442404e-01 -1.02659011e+00 -9.71770883e-01
2.11041778e-01 -7.65244126e-01 -4.70824778e-01 -2.07946986e-01
2.67839849e-01 -5.14491737e-01 -7.32231021e-01 1.21338856e+00
7.62989938e-01 -4.07927036e-01 8.32020700e-01 -1.40290761e+00
-8.74872863e-01 8.64755094e-01 -3.20837826e-01 1.70927063e-01
2.13741899e-01 7.43351698e-01 7.47050762e-01 -8.75629663e-01
8.40017557e-01 1.37126172e+00 3.85744900e-01 9.27375436e-01
-1.65092480e+00 -9.08461452e-01 -2.78740168e-01 -1.25183553e-01
-1.35721529e+00 -4.75085169e-01 7.50841916e-01 -1.74106792e-01
6.73207700e-01 7.04004407e-01 3.73800427e-01 2.14044452e+00
4.56117868e-01 9.50187147e-01 1.30343544e+00 -1.51882127e-01
3.28351557e-01 -3.68412584e-02 -3.58080566e-01 5.64943671e-01
5.57219207e-01 5.37726820e-01 -7.50753164e-01 -7.45382547e-01
9.73782122e-01 -4.42592233e-01 -2.10152045e-01 -1.78892404e-01
-9.60011661e-01 9.76508021e-01 3.35392095e-02 -1.00511843e-02
3.66945006e-02 8.20372105e-01 1.50877476e-01 1.90568119e-01
2.26552129e-01 1.41438335e-01 1.05574116e-01 -2.32285246e-01
-1.13207865e+00 6.86830461e-01 1.00825858e+00 8.89993906e-01
4.49618667e-01 5.82264662e-01 -6.09927535e-01 9.77980867e-02
4.53486413e-01 1.12496471e+00 -5.38420193e-02 -7.10552514e-01
6.33469224e-01 -3.33862990e-01 -2.43149564e-01 -8.41876030e-01
3.50831479e-01 -7.00675786e-01 -7.99898684e-01 1.16652369e-01
1.38878271e-01 -4.12195146e-01 -1.27064776e+00 1.98337519e+00
1.04742125e-01 4.11766946e-01 5.08104917e-04 4.17357713e-01
5.03591061e-01 7.33265221e-01 -1.19266048e-01 7.65155479e-02
1.17620349e+00 -5.26271403e-01 -6.47661269e-01 -1.49176329e-01
3.43771577e-01 -7.06034601e-01 9.19011235e-01 4.10961479e-01
-1.11317134e+00 8.09070319e-02 -1.20274949e+00 1.89341128e-01
-1.70592949e-01 -3.70450616e-01 4.33859050e-01 1.51833510e+00
-9.15225625e-01 3.90732229e-01 -1.04742503e+00 2.73559451e-01
6.93189263e-01 2.25294709e-01 8.48233551e-02 -7.43713826e-02
-1.35660243e+00 5.76426029e-01 -2.72432063e-02 -4.35096323e-02
-2.14702272e+00 -5.22107840e-01 -7.65916526e-01 -1.13133624e-01
2.82219440e-01 -1.07298136e+00 1.21413291e+00 -3.84847224e-01
-1.72628856e+00 7.47672260e-01 -5.13193011e-02 -8.56935978e-01
1.01213694e+00 -2.36042976e-01 -3.39601755e-01 1.95728570e-01
-1.61958575e-01 3.82533699e-01 1.51979423e+00 -1.48592889e+00
3.42380077e-01 -4.54527400e-02 -1.61374465e-01 -2.07113028e-01
-1.70468926e-01 1.35997757e-01 -2.30789527e-01 -1.02917695e+00
-2.26584628e-01 -1.07091880e+00 -1.25358239e-01 -4.11718309e-01
-1.36589181e+00 2.98259258e-01 1.07262528e+00 -3.94374520e-01
1.28738582e+00 -2.04104209e+00 1.65365994e-01 4.50102657e-01
1.52229831e-01 4.34809953e-01 -1.55545160e-01 8.39865327e-01
7.88425803e-02 6.61061823e-01 -4.77042437e-01 -6.51582181e-01
3.66322011e-01 2.42970362e-01 -1.09906697e+00 4.72555429e-01
2.28353262e-01 1.25658011e+00 -6.59205675e-01 -1.61233395e-01
-2.77550630e-02 7.52148271e-01 -6.55040741e-01 5.07874675e-02
-5.56287885e-01 3.52523565e-01 -1.94457963e-01 7.62491524e-01
7.32623875e-01 -8.52498040e-02 1.16155474e-02 2.12265134e-01
4.50428694e-01 3.96262020e-01 -9.60913301e-01 1.36391461e+00
-1.06773190e-01 7.42623150e-01 -2.38447301e-02 -1.30762517e-01
4.37329024e-01 3.92361373e-01 -3.63508880e-01 -1.04966179e-01
2.16372073e-01 2.51270711e-01 -3.54203209e-03 8.30958262e-02
5.61871946e-01 -1.01961978e-01 -2.19486892e-01 7.66119003e-01
1.55786902e-01 -4.95233595e-01 8.95832255e-02 7.45571554e-01
1.19070470e+00 -2.73975953e-02 -2.10203856e-01 3.13387029e-02
8.77056569e-02 -4.77829188e-01 2.02341661e-01 1.31059825e+00
2.70940721e-01 8.63291800e-01 7.79670000e-01 2.04025775e-01
-1.08946300e+00 -1.69636929e+00 -2.98482757e-02 4.80062544e-01
-1.04309030e-01 -8.38462114e-01 -1.16980255e+00 -5.50214648e-01
-2.32125819e-01 8.83990943e-01 -5.38203776e-01 -3.81881684e-01
-4.33385700e-01 -9.35076475e-01 1.63300574e+00 4.17693585e-01
5.78499496e-01 -7.23377347e-01 -2.24652693e-01 2.50728913e-02
-9.04754326e-02 -9.77066815e-01 -5.22461414e-01 9.01226625e-02
-8.68753850e-01 -8.38634849e-01 -7.94670761e-01 -3.04071516e-01
6.54475093e-01 -1.05801262e-01 1.01347125e+00 -2.92207420e-01
-1.82622969e-01 5.03936172e-01 1.78961486e-01 -3.29692811e-01
-1.14784014e+00 1.40518174e-01 2.06159502e-02 2.74292827e-02
-2.20512226e-01 -8.02689612e-01 -3.55099082e-01 2.71286905e-01
-1.54766369e+00 -1.50862515e-01 5.29942274e-01 6.62115812e-01
3.91340107e-01 -2.10871492e-02 -1.59723554e-02 -1.05200446e+00
1.17063236e+00 -3.92390043e-01 -7.23244488e-01 1.31124303e-01
-4.13423777e-01 3.21339130e-01 2.75026798e-01 -6.98199809e-01
-9.55612242e-01 -3.67493093e-01 -3.63062322e-01 -6.08870983e-01
1.33627921e-01 3.33246380e-01 -3.72207552e-01 -1.01398528e-01
9.20460880e-01 8.57579410e-01 -2.65925199e-01 -2.39417568e-01
7.90018916e-01 2.96846569e-01 7.59240448e-01 -7.43067086e-01
1.50258386e+00 6.59732401e-01 1.58849627e-01 -7.11823702e-01
-1.40223220e-01 5.22216022e-01 2.61695504e-01 3.33679877e-02
7.51094103e-01 -8.58236253e-01 -5.65807343e-01 1.01057243e+00
-1.34440660e+00 -3.30147475e-01 -1.99080035e-01 -1.41518921e-01
-5.98323464e-01 4.33670282e-01 -1.06316650e+00 -1.01316154e+00
-4.92563695e-01 -1.26735473e+00 9.40751195e-01 2.69498304e-02
-2.08865806e-01 -9.68132734e-01 3.31075609e-01 2.22392231e-01
5.45805216e-01 3.71367067e-01 9.74194586e-01 -6.14093184e-01
-1.19190884e+00 -2.83969045e-01 4.99729723e-01 5.41714549e-01
-2.77658045e-01 2.62538433e-01 -1.20514214e+00 -3.10802579e-01
2.02466935e-01 -2.14837968e-01 1.09164786e+00 1.45225942e-01
7.23888338e-01 -6.81280911e-01 -3.60068351e-01 7.43023276e-01
1.24768686e+00 -1.12207130e-01 1.12413859e+00 3.99565473e-02
5.63044369e-01 -1.73823684e-01 -3.96915257e-01 3.52634341e-01
-2.02173650e-01 3.29179585e-01 7.60831058e-01 2.12206066e-01
-5.83600737e-02 -8.63275707e-01 1.06858206e+00 5.12972355e-01
1.88605145e-01 -7.65074193e-01 -5.82815349e-01 3.04766417e-01
-1.30425453e+00 -1.25070524e+00 -9.26350728e-02 2.12676406e+00
1.15144241e+00 4.97852057e-01 -3.81829813e-02 1.96850196e-01
7.93861628e-01 4.78661388e-01 -4.59559351e-01 -3.88423771e-01
-4.36093926e-01 6.22291386e-01 5.88637352e-01 5.13711810e-01
-1.05964839e+00 9.49984372e-01 7.30875158e+00 1.28898728e+00
-7.62061775e-01 3.44433010e-01 6.59321070e-01 -1.29015565e-01
-9.63653386e-01 1.06681466e-01 -9.60020125e-01 5.21718979e-01
8.84347916e-01 -2.29765981e-01 5.41102886e-01 6.83388829e-01
-2.35217184e-01 1.61981031e-01 -1.09663117e+00 7.28062510e-01
2.23050624e-01 -1.67389560e+00 3.05489868e-01 5.28821290e-01
1.02358592e+00 -7.65572563e-02 9.44827616e-01 -2.22835084e-03
7.77013958e-01 -1.11050332e+00 7.98339069e-01 1.57201305e-01
8.09561431e-01 -8.90759230e-01 1.89647049e-01 3.18503439e-01
-6.43881977e-01 1.87728569e-01 4.16597724e-02 4.31588441e-01
3.98576051e-01 7.09079564e-01 -6.69042766e-01 3.74896228e-01
1.58867151e-01 7.55234361e-02 -2.95970291e-01 5.95342457e-01
-9.24577594e-01 1.19632387e+00 -5.15458226e-01 1.79464117e-01
1.52772158e-01 4.44825832e-03 1.22692573e+00 1.14357817e+00
4.09658134e-01 -4.14817929e-01 -4.64081496e-01 1.57243133e+00
-3.54418129e-01 -7.42428184e-01 -9.44057226e-01 -4.31396246e-01
4.93278086e-01 8.41013730e-01 -4.74797696e-01 -1.24171644e-01
2.52762854e-01 1.28236580e+00 -3.15266401e-01 5.14868438e-01
-1.15200436e+00 -3.07939321e-01 8.65671933e-01 1.85347274e-01
5.47273517e-01 -5.78295052e-01 -1.92977712e-01 -1.24616456e+00
-1.20153479e-01 -1.25245893e+00 9.52786207e-02 -6.36348486e-01
-1.18587816e+00 2.92136401e-01 -3.55236530e-02 -9.42169368e-01
-4.85669971e-01 -3.19957078e-01 -9.45423961e-01 7.43437648e-01
-1.14578664e+00 -1.16228998e+00 3.70027483e-01 6.65570319e-01
5.38007766e-02 -6.00127950e-02 9.17920291e-01 3.19787189e-02
-5.74208677e-01 8.66665661e-01 2.65152961e-01 3.67453009e-01
2.75937736e-01 -8.88020396e-01 9.65124488e-01 1.47252071e+00
6.96627975e-01 1.19375312e+00 8.99819076e-01 -8.25972080e-01
-1.74852824e+00 -8.82144332e-01 4.88315314e-01 -7.43271708e-01
5.98099828e-01 -8.87508690e-01 -3.48518074e-01 8.10043573e-01
4.17186171e-01 -3.77650350e-01 7.17881620e-01 -5.47703624e-01
-6.64171994e-01 3.24501008e-01 -1.06965411e+00 1.08071578e+00
9.62588668e-01 -9.70475674e-01 -9.70744789e-02 3.38220298e-01
7.57297754e-01 -7.63624668e-01 -3.29413831e-01 -9.05955583e-02
2.87955433e-01 -9.24378812e-01 1.21445882e+00 -5.00720322e-01
6.70581460e-01 -3.26408595e-01 -1.46269217e-01 -1.12777805e+00
1.04607970e-01 -1.67197216e+00 -7.35401630e-01 1.38165522e+00
4.20807868e-01 -6.94893956e-01 6.82663620e-01 3.11289459e-01
2.49232426e-01 -4.12910342e-01 -1.13719046e+00 -1.05110061e+00
3.87413323e-01 -8.67288232e-01 5.31268775e-01 5.43751359e-01
-5.16145825e-01 1.29647225e-01 -7.29556918e-01 3.70995253e-01
9.28381622e-01 -3.00182164e-01 9.06087816e-01 -5.76714933e-01
-6.71407521e-01 -3.66483986e-01 -2.65998840e-01 -1.12009943e+00
-2.06052661e-01 -1.04418063e+00 -2.59006292e-01 -1.05470753e+00
1.04776867e-01 9.52341780e-02 -3.64458822e-02 1.27250567e-01
-7.72512406e-02 5.10239482e-01 4.46959108e-01 2.33635962e-01
-1.24658629e-01 6.50539398e-01 9.44186151e-01 -3.48769903e-01
1.92635968e-01 2.67335117e-01 -6.62744641e-01 4.07714605e-01
6.83960676e-01 -1.07680309e+00 -5.98232687e-01 -3.44839126e-01
8.02453458e-01 -2.59162277e-01 8.25314701e-01 -9.35466707e-01
1.76024631e-01 2.82821227e-02 1.53592184e-01 -4.88268197e-01
4.26982433e-01 -4.49971497e-01 4.78061855e-01 7.38633156e-01
-2.60168761e-01 3.54917673e-03 5.01988567e-02 8.71597052e-01
1.72789246e-01 -4.84993272e-02 5.53351402e-01 -1.40435740e-01
6.55218139e-02 2.78931767e-01 -7.37005413e-01 4.14206743e-01
8.16870093e-01 -4.36641090e-02 -7.28605568e-01 -8.36003065e-01
-6.93114400e-01 -3.44164759e-01 4.02572870e-01 1.24326400e-01
7.49018908e-01 -1.25057364e+00 -8.16311002e-01 4.16168630e-01
-3.30838203e-01 -1.61489621e-01 -3.81114744e-02 6.23996019e-01
-4.95833874e-01 2.45057717e-01 2.85416335e-01 -5.54514289e-01
-1.24242258e+00 3.28277707e-01 3.51368636e-01 -6.47588253e-01
-3.35744321e-01 1.09080064e+00 1.87272280e-01 -1.19017355e-01
6.64841086e-02 -8.39630589e-02 7.42493033e-01 -3.34004819e-01
4.34601575e-01 4.31353837e-01 -9.97776017e-02 -2.44568333e-01
-1.95613444e-01 -5.29823452e-03 -3.46256420e-02 -8.82246315e-01
9.63061869e-01 3.85846734e-01 -3.20337527e-02 -2.53049135e-02
9.10194516e-01 2.97235072e-01 -1.26731360e+00 -1.05691426e-01
-5.73542118e-01 -4.06641006e-01 -2.88730055e-01 -7.76421547e-01
-8.69996071e-01 1.17463005e+00 2.77609259e-01 1.86406046e-01
7.00388610e-01 -1.36885405e-01 9.45221961e-01 2.86629468e-01
2.96410263e-01 -8.06507587e-01 2.36707449e-01 3.62118930e-01
7.77676344e-01 -2.51749992e-01 -6.18242286e-02 -4.07701403e-01
-5.90019166e-01 6.12054050e-01 6.50196075e-02 -1.58559307e-02
4.92734104e-01 8.14970732e-01 -2.26988792e-01 9.03525352e-02
-7.75288641e-01 1.81793272e-01 -2.25751102e-02 8.29113722e-01
-1.98325291e-01 -5.72183505e-02 1.07428826e-01 5.00522554e-01
-2.87371337e-01 -1.84710085e-01 6.60406172e-01 1.22519171e+00
-1.58743355e-02 -1.53083336e+00 -6.09636068e-01 6.07365556e-02
-8.31663191e-01 -6.09907627e-01 -5.99435985e-01 4.66585785e-01
-1.29849598e-01 9.63776112e-01 -3.08266073e-01 -5.03795326e-01
-1.71885535e-01 3.06788772e-01 6.66591227e-01 -3.53332520e-01
-8.00890505e-01 6.55589551e-02 -6.26799017e-02 -3.27579826e-01
3.63709144e-02 -5.26218534e-01 -7.77487457e-01 -6.75156713e-01
-4.80290860e-01 -1.59975827e-01 6.44980669e-01 6.36800885e-01
6.29858971e-01 3.05980057e-01 5.53473175e-01 -5.80604017e-01
-1.08278525e+00 -6.99548900e-01 -4.93840337e-01 2.00770348e-01
3.08713377e-01 -1.87405154e-01 -5.73247552e-01 1.18774854e-01]
|
[5.760780334472656, 7.859830379486084]
|
5478252d-4d76-4016-9b95-fe613f6f8fa7
|
hybrur-a-hybrid-physical-neural-solution-for
|
2107.0266
| null |
https://arxiv.org/abs/2107.02660v1
|
https://arxiv.org/pdf/2107.02660v1.pdf
|
HybrUR: A Hybrid Physical-Neural Solution for Unsupervised Underwater Image Restoration
|
Robust vision restoration for an underwater image remains a challenging problem. For the lack of aligned underwater-terrestrial image pairs, the unsupervised method is more suited to this task. However, the pure data-driven unsupervised method usually has difficulty in achieving realistic color correction for lack of optical constraint. In this paper, we propose a data- and physics-driven unsupervised architecture that learns underwater vision restoration from unpaired underwater-terrestrial images. For sufficient domain transformation and detail preservation, the underwater degeneration needs to be explicitly constructed based on the optically unambiguous physics law. Thus, we employ the Jaffe-McGlamery degradation theory to design the generation models, and use neural networks to describe the process of underwater degradation. Furthermore, to overcome the problem of invalid gradient when optimizing the hybrid physical-neural model, we fully investigate the intrinsic correlation between the scene depth and the degradation factors for the backscattering estimation, to improve the restoration performance through physical constraints. Our experimental results show that the proposed method is able to perform high-quality restoration for unconstrained underwater images without any supervision. On multiple benchmarks, we outperform several state-of-the-art supervised and unsupervised approaches. We also demonstrate that our methods yield encouraging results on real-world applications.
|
['Junzhi Yu', 'Min Tan', 'Yue Lu', 'Jian Wang', 'Zhengxing Wu', 'Xingyu Chen', 'Shuaizheng Yan']
|
2021-07-06
| null | null | null | null |
['underwater-image-restoration']
|
['computer-vision']
|
[ 3.43896836e-01 -2.90720463e-01 5.79524457e-01 -5.93241751e-01
-6.82715714e-01 -3.11129224e-02 1.39362872e-01 -2.18743965e-01
-6.32705092e-01 7.24996805e-01 1.84549823e-01 -2.81543117e-02
-2.05651611e-01 -7.99918890e-01 -8.98814797e-01 -1.26453173e+00
1.03381097e-01 2.36509070e-02 1.67857155e-01 -2.58542359e-01
2.73353577e-01 2.01745138e-01 -1.68474770e+00 -3.31525594e-01
1.53465593e+00 8.58506680e-01 6.29329503e-01 5.10649264e-01
-8.48397389e-02 7.41015851e-01 -1.91123530e-01 -4.17276397e-02
5.06200969e-01 -4.86382157e-01 -2.05565289e-01 4.38095152e-01
6.37907088e-01 -7.49275863e-01 -4.31850016e-01 1.51813388e+00
6.31084979e-01 3.67730200e-01 6.91412091e-01 -6.04576170e-01
-7.90749073e-01 2.06473619e-01 -5.21033406e-01 -1.73987925e-01
-1.21788695e-01 2.47327358e-01 7.37712264e-01 -9.73887682e-01
2.97370315e-01 1.16055930e+00 4.98833477e-01 5.65149307e-01
-9.89389479e-01 -3.25245500e-01 3.12210500e-01 3.37710619e-01
-9.79701757e-01 -5.34546912e-01 7.93577373e-01 -4.06797111e-01
3.35770279e-01 -9.93961692e-02 6.62708163e-01 6.76924288e-01
1.72061756e-01 4.70160276e-01 1.35431027e+00 -4.71241862e-01
2.69707203e-01 -2.58410126e-01 1.41794086e-01 7.11207628e-01
3.29442382e-01 2.16727465e-01 -4.94470328e-01 2.11583182e-01
6.18167937e-01 1.88148722e-01 -7.35854626e-01 -2.46604010e-01
-6.08565032e-01 5.26225805e-01 6.60933375e-01 -3.22309345e-01
-2.30829775e-01 -3.55854910e-03 -6.59391808e-04 1.67319596e-01
5.34632385e-01 1.47011742e-01 -3.06076348e-01 3.36459935e-01
-7.79799283e-01 -8.96188319e-02 6.31245792e-01 5.54738939e-01
1.04842734e+00 3.60661626e-01 3.28060657e-01 1.06460536e+00
9.08024609e-01 9.25635576e-01 3.37315410e-01 -1.12536621e+00
1.60821870e-01 2.73065835e-01 3.53841722e-01 -7.56824851e-01
-2.36349523e-01 -2.72119403e-01 -1.05249846e+00 6.05519593e-01
3.04996610e-01 -1.34905130e-02 -1.30379379e+00 1.41343296e+00
1.55155197e-01 2.21282676e-01 7.15787053e-01 1.31387615e+00
8.30270827e-01 8.29576492e-01 -2.19751164e-01 -5.77520430e-01
9.23723638e-01 -8.98686945e-01 -8.28714788e-01 -5.53341508e-01
8.83599073e-02 -5.23078740e-01 1.10243809e+00 3.84613156e-01
-8.71995509e-01 -3.68607700e-01 -1.28375518e+00 -1.46021366e-01
4.16449690e-03 1.67847779e-02 4.06451672e-01 4.36253637e-01
-1.05867720e+00 6.51986361e-01 -8.77253592e-01 -4.27659065e-01
1.07953483e-02 9.05188322e-02 -2.48374701e-01 -5.15560985e-01
-1.06553769e+00 7.14988291e-01 2.33839512e-01 8.50474834e-01
-1.30915725e+00 -3.78439903e-01 -9.17806208e-01 -1.72424406e-01
-6.75336346e-02 -5.16601682e-01 8.34285080e-01 -7.90008962e-01
-1.69921815e+00 4.68199700e-01 1.14999771e-01 -3.17896634e-01
4.88918275e-01 -3.02338123e-01 -1.65859032e-02 2.91217595e-01
-1.41515270e-01 3.26807082e-01 7.95512557e-01 -1.86762321e+00
-4.84516114e-01 -4.08087164e-01 4.70829234e-02 5.72808862e-01
-6.51993692e-01 -3.55330259e-01 -6.41153693e-01 -3.08757871e-01
6.12244785e-01 -5.74224949e-01 -3.36844712e-01 4.97945487e-01
-1.69514611e-01 5.04401565e-01 6.49440825e-01 -9.07852173e-01
4.65584546e-01 -2.31106782e+00 3.17131519e-01 -1.42789409e-01
-1.28706142e-01 5.54627627e-02 -1.31936371e-01 2.67799467e-01
4.13567424e-01 -2.27966726e-01 -9.57731426e-01 -6.87081814e-01
-2.43310064e-01 7.29297936e-01 -2.28467435e-01 8.19171846e-01
-1.92573488e-01 1.63522616e-01 -9.28394437e-01 -6.09022260e-01
2.25994617e-01 4.19265628e-01 -4.95965272e-01 6.64466560e-01
-1.38512984e-01 6.99392259e-01 -2.74662673e-01 7.16859043e-01
1.13792169e+00 2.51806974e-01 1.38277888e-01 -3.26619238e-01
-3.01270783e-01 -1.42073706e-01 -1.00688839e+00 1.74677324e+00
-5.76790214e-01 5.72478831e-01 6.28140807e-01 -9.56330597e-01
1.12696135e+00 -1.87739849e-01 1.72950044e-01 -7.77324677e-01
-1.28852025e-01 4.86349672e-01 -2.64117122e-01 -9.44880009e-01
3.49673212e-01 -5.16249418e-01 6.11253619e-01 -5.04874624e-02
-2.52378941e-01 -3.49265754e-01 -4.84099798e-02 4.51731868e-02
6.70612693e-01 3.74990702e-01 -2.53626227e-01 -3.74226511e-01
4.49662864e-01 -1.76970974e-01 9.73539650e-01 6.79055750e-01
-4.17875312e-02 9.02121842e-01 -1.24548368e-01 -2.12194189e-01
-7.80377746e-01 -1.01808071e+00 -2.61093527e-01 6.83090627e-01
7.15163648e-01 4.39249724e-01 -5.70227444e-01 -1.07519187e-01
-2.43972570e-01 3.10207874e-01 -3.61621767e-01 -1.65341347e-01
-2.64844179e-01 -1.29433048e+00 2.58149296e-01 2.91036844e-01
8.20477784e-01 -9.03738618e-01 -2.71507978e-01 2.01635033e-01
-3.24172527e-01 -1.23539960e+00 -7.74789304e-02 2.67790437e-01
-1.16711462e+00 -8.68644476e-01 -7.30960131e-01 -9.69796896e-01
9.39658761e-01 6.05206549e-01 5.95595777e-01 3.61071616e-01
-9.07615349e-02 2.71977246e-01 -5.54638624e-01 -1.12856850e-01
-2.07459480e-01 -7.22316921e-01 2.41234794e-01 3.67633611e-01
-2.43541956e-01 -8.59412968e-01 -9.01625156e-01 3.48252773e-01
-1.24363565e+00 -2.61367783e-02 5.88315606e-01 1.07503414e+00
5.09308159e-01 9.71414000e-02 1.04891054e-01 -4.01403934e-01
1.81635708e-01 -1.88333347e-01 -7.18907952e-01 8.02270398e-02
-7.41499543e-01 6.00158386e-02 5.43101788e-01 -2.26897284e-01
-1.47170115e+00 2.83954501e-01 -2.09235951e-01 -2.77483523e-01
8.52226391e-02 7.87986219e-01 -4.79291081e-01 -4.30881709e-01
6.05046511e-01 7.01281488e-01 2.36019209e-01 -8.07894111e-01
4.26819995e-02 7.98264623e-01 7.26777852e-01 -5.18744051e-01
1.05730081e+00 9.75463510e-01 -1.66304290e-01 -1.28227711e+00
-7.87962973e-01 -5.09702206e-01 -3.79481465e-01 -4.30938631e-01
8.53404403e-01 -1.21780992e+00 -3.40761781e-01 9.60555494e-01
-9.97390985e-01 -5.87578654e-01 1.43456191e-01 6.70326650e-01
-2.52302378e-01 1.13202631e+00 -7.70140588e-01 -1.15902638e+00
-4.37940031e-01 -1.18583810e+00 7.67183781e-01 5.18919945e-01
1.05888546e+00 -8.77417505e-01 5.11907525e-02 4.16562200e-01
2.05359176e-01 6.95579201e-02 5.70508957e-01 3.04940104e-01
-8.03709626e-01 2.27679282e-01 -4.63834792e-01 8.90230238e-01
2.36028343e-01 2.36423507e-01 -1.00586557e+00 -3.76195848e-01
2.63090849e-01 -3.35445702e-01 1.41475701e+00 3.52646053e-01
9.11427379e-01 -8.24405923e-02 3.06636333e-01 1.17566669e+00
1.77469337e+00 -5.03656417e-02 1.03836405e+00 6.13627017e-01
9.40912426e-01 8.69570971e-01 6.91855788e-01 4.45187479e-01
5.81265211e-01 2.26965114e-01 1.07163215e+00 -2.27538317e-01
-8.66239369e-02 -4.51151021e-02 6.54540420e-01 9.99893725e-01
-3.17314595e-01 -2.62757868e-01 -6.35910571e-01 7.36286998e-01
-1.91351867e+00 -5.65529287e-01 -4.66158807e-01 2.22577453e+00
7.99501359e-01 -5.88673055e-02 -6.73585713e-01 -4.36202921e-02
5.97184539e-01 1.87244609e-01 -6.40269518e-01 2.06098542e-01
-4.03210908e-01 -2.85693228e-01 8.23888898e-01 7.41518140e-01
-8.17020118e-01 8.76574814e-01 5.66338682e+00 3.03489655e-01
-8.63870323e-01 -5.88387400e-02 2.85338730e-01 4.16887939e-01
-4.65807110e-01 9.86966789e-02 -6.05081439e-01 3.46647173e-01
4.06190634e-01 5.31804025e-01 4.69205081e-01 4.08156067e-01
7.73008764e-01 -2.70714134e-01 -6.10132933e-01 9.19373035e-01
8.78218338e-02 -7.44409442e-01 6.89482912e-02 8.92659351e-02
8.66121113e-01 1.21454403e-01 -2.33189106e-01 -1.78122222e-01
2.74352074e-01 -5.63378632e-01 6.25747442e-01 9.57784235e-01
5.32035708e-01 -2.03387976e-01 7.82667458e-01 3.20038915e-01
-7.81873703e-01 -1.83777824e-01 -8.62523675e-01 -2.26146579e-01
2.78967589e-01 9.32217479e-01 1.66658871e-02 5.98672509e-01
9.78626847e-01 8.75016689e-01 -2.56671488e-01 1.41232479e+00
-5.44364631e-01 5.52773654e-01 -4.12035674e-01 2.22320646e-01
1.06933795e-01 -6.87957406e-01 5.21200597e-01 9.86493349e-01
4.75002736e-01 2.97128886e-01 1.90028638e-01 4.06889260e-01
1.73095223e-02 2.16911793e-01 -2.65469939e-01 2.26498753e-01
4.72268574e-02 1.15668750e+00 -3.69242758e-01 -2.86789183e-02
-3.13352048e-01 9.97192979e-01 1.31683037e-01 5.94455242e-01
-4.80934650e-01 -9.62348655e-02 7.12121069e-01 -1.87005952e-01
1.03083722e-01 -5.52057981e-01 -2.03506351e-01 -1.37865448e+00
1.70888692e-01 -6.78263664e-01 7.73180947e-02 -1.12425983e+00
-1.46204114e+00 3.57130647e-01 -2.92192936e-01 -1.50622153e+00
4.55767274e-01 -8.35742772e-01 -6.99869514e-01 8.06772172e-01
-2.39118910e+00 -9.96931791e-01 -9.61716771e-01 5.31300485e-01
5.24721682e-01 2.05628470e-01 5.51728189e-01 3.84630650e-01
-6.47817612e-01 1.45787103e-02 7.29724586e-01 -2.86105294e-02
9.77608502e-01 -1.32367969e+00 -1.22833788e-01 1.34201217e+00
-3.09204280e-01 3.30033213e-01 1.05152667e+00 -6.25333130e-01
-1.76854467e+00 -9.55667078e-01 1.53965950e-01 3.26429456e-01
7.25345194e-01 9.50595960e-02 -1.28675914e+00 1.13834456e-01
4.52700183e-02 1.37962386e-01 4.52263176e-01 -4.00884509e-01
-2.71917820e-01 -6.48228467e-01 -1.06344831e+00 5.55706799e-01
9.95191872e-01 -3.75186890e-01 -4.81831759e-01 4.40611362e-01
6.62116885e-01 -3.30216289e-01 -7.55289853e-01 5.18438458e-01
5.62304318e-01 -1.11533475e+00 8.50185692e-01 2.50879433e-02
8.27979922e-01 -7.05930650e-01 -4.37218189e-01 -1.33814716e+00
-2.08937041e-02 -3.30790043e-01 1.53680354e-01 1.22395694e+00
2.82884747e-01 -6.45903468e-01 5.08882403e-01 3.37154984e-01
-4.94952142e-01 -1.09584101e-01 -6.89724028e-01 -5.44447422e-01
-7.99065754e-02 -1.29517332e-01 -9.38047841e-03 6.33420408e-01
-5.31777918e-01 2.15848908e-02 -8.59772861e-01 1.00434256e+00
1.34314716e+00 1.59340307e-01 6.27798378e-01 -1.20456588e+00
-3.84663731e-01 8.04086551e-02 -2.11879343e-01 -1.36578190e+00
-5.51526584e-02 -2.28576034e-01 1.05412674e+00 -2.06071568e+00
1.55979425e-01 -1.77603826e-01 -3.52385402e-01 2.70701557e-01
-3.12128365e-01 4.08193350e-01 -7.15374947e-03 5.14432967e-01
-2.00274885e-01 1.06391358e+00 1.43883121e+00 -4.17614192e-01
-1.15007989e-01 -3.16313028e-01 -5.72555482e-01 6.65712118e-01
6.86208069e-01 -2.91380078e-01 -2.37791136e-01 -1.11015654e+00
3.71321999e-02 4.68651056e-02 3.70216608e-01 -1.02611172e+00
4.34174359e-01 -3.90999377e-01 1.45574048e-01 -2.60979921e-01
3.69699657e-01 -7.99942255e-01 -2.88671613e-01 3.84181499e-01
9.52927172e-02 -7.37632930e-01 -1.84827089e-01 9.20101702e-01
-3.58405054e-01 -6.03901148e-01 1.10434783e+00 -1.77262381e-01
-9.86005783e-01 3.32795262e-01 -2.59977967e-01 -3.26363266e-01
2.60286391e-01 -1.78662330e-01 -3.98948878e-01 -3.60338300e-01
-4.18908119e-01 5.51122367e-01 6.83738589e-01 -4.66665849e-02
1.00849736e+00 -6.97330534e-01 -9.27151918e-01 3.85213457e-02
3.05219918e-01 3.36803317e-01 5.27231693e-01 7.01936662e-01
-1.07994831e+00 -6.77697718e-01 -1.88007116e-01 -6.42487824e-01
-8.84706974e-01 8.43623281e-02 6.55703485e-01 1.59028202e-01
-8.70881259e-01 6.38941169e-01 3.97173375e-01 -5.40475547e-01
2.38189548e-01 -1.24327756e-01 -3.24865043e-01 -1.83454677e-01
6.18584752e-01 2.22720057e-01 -6.94251209e-02 -5.78163683e-01
-6.00819662e-02 8.77437890e-01 2.57503301e-01 -8.29991251e-02
1.64071310e+00 -6.30787015e-01 -3.19729060e-01 2.22338527e-01
9.47393775e-01 -2.00222954e-01 -1.96338010e+00 -2.47168571e-01
-5.11653423e-01 -4.54997629e-01 4.94970083e-01 -5.34772098e-01
-1.31215692e+00 9.70785797e-01 8.70373189e-01 9.35210139e-02
1.36375904e+00 -6.04983807e-01 7.02895820e-01 6.51774883e-01
2.79815853e-01 -1.03865540e+00 -2.11262647e-02 6.36059582e-01
8.07661057e-01 -1.60593224e+00 2.45095044e-01 -4.57860023e-01
-3.95731181e-01 1.26855493e+00 6.37131691e-01 -1.11585192e-01
4.83488619e-01 2.97919791e-02 5.51692903e-01 -5.38727548e-03
-8.98345336e-02 -4.50609744e-01 -1.14675045e-01 5.83456397e-01
-1.15786865e-01 -1.69275582e-01 -3.91368091e-01 1.65514797e-01
3.07733387e-01 -2.48836875e-01 8.56010497e-01 7.66943693e-01
-8.42951179e-01 -8.25869083e-01 -6.04997873e-01 3.99211012e-02
-1.76563591e-01 -2.96522290e-01 6.41399175e-02 2.92583436e-01
-6.11409321e-02 1.09511161e+00 -2.45850042e-01 -1.65979341e-01
2.69111902e-01 -3.89021903e-01 3.23908299e-01 -5.31558394e-01
1.67522997e-01 3.57736081e-01 -1.12495624e-01 -9.27396938e-02
-9.70783949e-01 -5.03131211e-01 -1.34287870e+00 1.62349507e-01
-2.90412128e-01 2.14197621e-01 8.56549919e-01 1.01139617e+00
-2.38670915e-01 3.28954220e-01 9.33064103e-01 -1.32407820e+00
-5.02402961e-01 -1.10599160e+00 -9.61578846e-01 2.52443343e-01
6.25707865e-01 -7.31120586e-01 -1.03750110e+00 3.34721148e-01]
|
[10.699172973632812, -3.5258805751800537]
|
9443ea03-3e34-46ff-8254-6d9c1685c3d9
|
not-all-lotteries-are-made-equal-1
|
2206.08175
| null |
https://arxiv.org/abs/2206.08175v1
|
https://arxiv.org/pdf/2206.08175v1.pdf
|
Not All Lotteries Are Made Equal
|
The Lottery Ticket Hypothesis (LTH) states that for a reasonably sized neural network, a sub-network within the same network yields no less performance than the dense counterpart when trained from the same initialization. This work investigates the relation between model size and the ease of finding these sparse sub-networks. We show through experiments that, surprisingly, under a finite budget, smaller models benefit more from Ticket Search (TS).
|
['Somya Suhans Mahapatra', 'Sai Mitheran', 'Surya Kant Sahu']
|
2022-06-16
| null | null | null | null |
['ticket-search']
|
['methodology']
|
[-9.82455090e-02 4.82701838e-01 -5.54966748e-01 -4.59958524e-01
-3.09601039e-01 -1.00261152e-01 5.81736743e-01 -3.50751162e-01
-6.89147711e-01 9.05569911e-01 1.54011399e-01 -3.86283606e-01
-1.22212335e-01 -7.17457294e-01 -1.09248352e+00 -5.22913873e-01
-3.55009466e-01 8.52636397e-01 5.21843806e-02 6.43573748e-03
3.76504473e-02 4.29904103e-01 -1.38192332e+00 1.36688516e-01
4.73863155e-01 1.20740199e+00 1.11532055e-01 5.29237278e-02
1.47189021e-01 1.02887976e+00 -5.58123648e-01 -3.72257769e-01
8.16145599e-01 -2.84556389e-01 -7.82217205e-01 -1.90582052e-01
7.11912215e-01 -4.99674559e-01 -6.46169722e-01 9.57427800e-01
2.24562854e-01 5.14655828e-01 5.28335512e-01 -1.19703460e+00
-1.63760424e-01 1.25839746e+00 -3.88350964e-01 4.10579711e-01
-2.83317268e-01 2.12956533e-01 1.04004121e+00 -8.60309839e-01
4.69307661e-01 1.14098859e+00 1.06319129e+00 2.35770196e-01
-1.31404722e+00 -1.07153106e+00 1.26734540e-01 8.43352452e-03
-1.63122416e+00 -8.65919173e-01 4.64385897e-01 1.81902602e-01
1.42130542e+00 -2.58257855e-02 6.05017841e-01 1.00737166e+00
2.08290204e-01 6.28983855e-01 9.43094730e-01 -4.11724836e-01
3.44615906e-01 3.61593157e-01 2.79313028e-01 6.06849611e-01
8.19976032e-01 4.42429900e-01 -4.82854873e-01 -1.24080010e-01
1.04519987e+00 -7.31255561e-02 -1.80949885e-02 -3.64472806e-01
-6.33870780e-01 1.24420655e+00 5.39692521e-01 5.33644974e-01
-5.84959745e-01 3.21534723e-01 3.37929726e-01 5.68334103e-01
3.47750425e-01 7.25072145e-01 -3.48787814e-01 -5.89085603e-03
-1.50567102e+00 -2.73335502e-02 1.15629268e+00 9.22268033e-01
9.33768094e-01 4.63483661e-01 1.40786871e-01 8.10184896e-01
-3.16521376e-02 3.47305760e-02 6.15090966e-01 -1.01950788e+00
5.92390239e-01 2.05353573e-01 2.63464786e-02 -7.95923173e-01
-3.11483771e-01 -7.62209654e-01 -1.10809851e+00 -4.39297482e-02
4.01618868e-01 -4.60118294e-01 -8.38220537e-01 1.84031034e+00
-1.05508402e-01 4.46399391e-01 -1.45603076e-01 5.16247034e-01
5.97129107e-01 6.64543986e-01 5.54947997e-04 -8.61138478e-02
9.82213676e-01 -9.12815690e-01 -9.53601748e-02 -8.74165237e-01
4.78746086e-01 -3.50476146e-01 8.09655190e-01 3.87195945e-01
-1.45202923e+00 -5.05385756e-01 -1.10298753e+00 2.14747041e-01
-3.55768681e-01 -1.94312766e-01 8.94002259e-01 8.29063773e-01
-1.30617225e+00 7.19131291e-01 -7.35787213e-01 -2.62331039e-01
5.24562418e-01 7.18528986e-01 -2.86160469e-01 -4.08948869e-01
-1.35465002e+00 1.35562658e+00 6.39266074e-01 1.02521904e-01
-1.05653858e+00 -6.86234236e-01 -7.32187986e-01 6.85050011e-01
2.45623097e-01 -7.33390212e-01 1.09534371e+00 -9.94343877e-01
-1.34215069e+00 6.19066536e-01 -4.55641039e-02 -1.31473422e+00
1.21766597e-01 -6.93421392e-03 7.57386982e-02 2.42397431e-02
-5.40619455e-02 1.03350389e+00 9.80667830e-01 -1.00264609e+00
-5.02146900e-01 -4.19214457e-01 1.97954044e-01 2.55820841e-01
-4.21449602e-01 -6.94202632e-02 -4.21292007e-01 -4.76646036e-01
-2.80667782e-01 -9.11909461e-01 -6.02727175e-01 -5.27229011e-01
-3.25296372e-01 -3.31420064e-01 3.95977139e-01 -1.22847132e-01
1.08317256e+00 -2.13688040e+00 -3.26470733e-01 6.14619017e-01
3.74214530e-01 7.51172677e-02 -3.46905202e-01 1.80467293e-02
-3.66928816e-01 1.37189314e-01 2.62083858e-01 -2.86436498e-01
1.93918422e-01 4.21850175e-01 -1.82214782e-01 4.64430541e-01
-1.44022211e-01 8.95969987e-01 -3.54309916e-01 -4.53632563e-01
2.50704978e-02 7.28217140e-02 -7.55588591e-01 -1.09902360e-01
1.70395628e-01 -6.13696933e-01 -1.50200859e-01 2.40782112e-01
6.16333067e-01 -5.96334815e-01 3.25492889e-01 -4.50010449e-02
1.36440337e-01 3.97584051e-01 -1.13705862e+00 1.10178530e+00
-4.38299030e-01 7.20159233e-01 3.18630457e-01 -1.39198184e+00
6.88185871e-01 2.06516266e-01 1.36709973e-01 -9.32359636e-01
3.53476912e-01 4.31010760e-02 1.77708551e-01 2.19294474e-01
4.51142132e-01 -3.56579393e-01 1.07815899e-02 7.46162653e-01
1.55661598e-01 3.19196731e-01 2.10739240e-01 2.21085355e-01
1.25661778e+00 -9.01937008e-01 2.99878776e-01 -3.24527323e-01
-2.96215892e-01 -6.55672466e-03 3.55173707e-01 1.52237749e+00
-3.48725356e-02 9.20785517e-02 4.30905491e-01 -7.97110975e-01
-1.11261368e+00 -8.30210090e-01 -2.56379038e-01 1.28957450e+00
-6.06649294e-02 -4.97418717e-02 -5.91838479e-01 -3.56306106e-01
1.17411248e-01 6.57968044e-01 -7.21121550e-01 -1.92383990e-01
-6.83765829e-01 -7.80118585e-01 7.42438555e-01 6.02834284e-01
7.44624436e-01 -1.09377038e+00 -6.19383454e-01 1.20543413e-01
2.76295453e-01 -1.10261798e+00 -3.17726493e-01 7.70466924e-01
-1.30937195e+00 -5.67601621e-01 -9.07223642e-01 -1.05844975e+00
7.05654383e-01 3.32720846e-01 1.51770937e+00 2.78815448e-01
3.50402668e-02 -1.56587288e-01 5.27313165e-02 -9.17809606e-02
-5.60139194e-02 4.10351366e-01 1.26997545e-01 -5.20586371e-01
4.38611954e-01 -6.60850465e-01 -3.95853251e-01 3.13529253e-01
-7.51494825e-01 -7.86606893e-02 9.45884526e-01 1.12192059e+00
2.10198939e-01 5.73306799e-01 4.91813004e-01 -9.36083198e-01
9.39247727e-01 -5.50095558e-01 -5.94450474e-01 1.23827003e-01
-8.67645323e-01 3.13513398e-01 7.70236075e-01 -5.81947923e-01
-8.64665568e-01 6.50407672e-02 2.24743247e-01 -6.81299150e-01
-2.55925536e-01 7.17946053e-01 1.79875806e-01 -3.27196181e-01
7.11030900e-01 1.40421838e-01 4.62770686e-02 -3.84776622e-01
2.86763698e-01 1.44940317e-01 2.30931774e-01 -4.50319588e-01
7.37225771e-01 2.74784952e-01 -2.45387435e-01 -7.17726231e-01
-8.57241511e-01 -2.65610188e-01 -2.26799473e-01 1.60763308e-01
2.66074389e-01 -9.22349155e-01 -6.00458503e-01 3.69973816e-02
-9.90401149e-01 -5.67091644e-01 -4.36497092e-01 5.50491691e-01
-3.08471560e-01 -2.66280118e-03 -5.25739312e-01 -5.01520514e-01
-2.57023811e-01 -8.21560740e-01 4.27206635e-01 2.89949745e-01
-2.16529205e-01 -8.90346229e-01 -1.32814636e-02 1.75534070e-01
5.00105739e-01 -4.27733600e-01 9.32624280e-01 -1.03041601e+00
-5.86465180e-01 -3.93656850e-01 -4.33567703e-01 3.34573507e-01
-3.57551008e-01 -7.08027184e-01 -7.14206517e-01 -3.71056139e-01
1.58844471e-01 -6.11038029e-01 1.20185304e+00 9.70841348e-01
8.97015214e-01 -8.95309806e-01 -4.47369009e-01 7.72797167e-01
1.62076914e+00 1.76544368e-01 8.06306839e-01 4.40900028e-01
4.41409051e-01 3.96880329e-01 1.04066543e-01 4.28684592e-01
4.10121232e-02 3.64466131e-01 2.79289275e-01 -2.60989338e-01
1.90114424e-01 -2.70596683e-01 1.33124709e-01 3.14694822e-01
-6.52155131e-02 3.46084014e-02 -8.11724067e-01 7.03095853e-01
-1.59237957e+00 -1.24845147e+00 6.43861055e-01 2.19592595e+00
1.12713623e+00 6.08947039e-01 1.97028011e-01 -9.10670310e-02
6.66118324e-01 2.92236686e-01 -6.13586009e-01 -4.67341363e-01
-3.62479314e-02 7.28753448e-01 9.83375132e-01 4.66364354e-01
-8.01184118e-01 1.16791582e+00 8.30901718e+00 1.26416993e+00
-8.94604921e-01 9.41153243e-02 1.01465225e+00 -5.30755341e-01
1.42196584e-02 -9.60901380e-02 -1.27564561e+00 3.31117600e-01
1.43012953e+00 -3.43886286e-01 6.97769940e-01 1.31737995e+00
1.28003970e-01 -1.56957760e-01 -1.23332477e+00 6.78325236e-01
6.06685132e-02 -1.75322247e+00 1.57126412e-01 2.72080243e-01
8.34701002e-01 4.68436241e-01 1.36714458e-01 7.62070656e-01
8.40711772e-01 -1.49822021e+00 4.51058209e-01 -7.87738040e-02
5.21821618e-01 -8.90641332e-01 7.62794673e-01 5.41622996e-01
-1.04572856e+00 -2.22903550e-01 -8.95565867e-01 -1.13917701e-01
-5.99010400e-02 6.63506806e-01 -1.21551013e+00 -1.71116754e-01
5.92180729e-01 1.92135468e-01 -6.19365096e-01 1.24690461e+00
2.64048755e-01 7.48177350e-01 -7.95032024e-01 -1.04318641e-01
7.12617695e-01 -2.51055416e-02 -1.08850226e-02 1.13648200e+00
1.36649325e-01 -2.23144054e-01 4.14312445e-02 9.21627164e-01
-3.02915841e-01 -2.99402863e-01 -8.87665749e-01 -3.34627666e-02
9.18747127e-01 7.69532204e-01 -7.77038991e-01 -5.95016837e-01
-2.52419025e-01 3.40908796e-01 5.78640044e-01 3.76336068e-01
-5.65359950e-01 -4.04602140e-01 3.38194132e-01 1.01884581e-01
7.75898397e-01 5.91686033e-02 -6.53503835e-01 -8.96401346e-01
-3.36324751e-01 -1.05107367e+00 3.88431013e-01 -6.00243747e-01
-1.21044004e+00 6.91227019e-01 2.18607888e-01 -7.09618688e-01
-5.59817791e-01 -3.00269961e-01 -6.86853468e-01 6.76611543e-01
-1.27066326e+00 -5.37071645e-01 7.26890042e-02 8.52968156e-01
4.31642771e-01 -2.89451897e-01 7.46538639e-01 3.38323146e-01
-7.35375404e-01 9.22305167e-01 2.02523798e-01 2.84143746e-01
1.12435855e-01 -8.48324656e-01 4.12625968e-01 5.45848310e-01
3.51570517e-01 1.08527458e+00 7.92883158e-01 -4.26862329e-01
-1.09414995e+00 -8.27545404e-01 9.87439573e-01 -3.75040583e-02
6.50111735e-01 -1.61619380e-01 -6.74024403e-01 8.45761895e-01
3.69463772e-01 -1.63008347e-01 4.62659448e-01 6.67962611e-01
-4.13345516e-01 -2.06898153e-01 -1.10788250e+00 3.57009262e-01
8.10346305e-01 -4.46914852e-01 -8.66411567e-01 2.96946138e-01
4.46559161e-01 -7.96074495e-02 -7.89466262e-01 2.16944024e-01
6.29904866e-01 -1.14343309e+00 1.35854518e+00 -6.70288682e-01
3.53947401e-01 5.04252732e-01 -2.68057454e-02 -1.31453729e+00
-3.83053541e-01 -5.42858958e-01 -1.04520442e-02 5.13479412e-01
5.78556657e-01 -5.47089398e-01 1.46183872e+00 7.92910278e-01
2.18030572e-01 -7.22326815e-01 -1.26591372e+00 -1.28555942e+00
1.91840023e-01 -2.41629496e-01 6.50269449e-01 9.69812691e-01
-2.58717928e-02 6.06212199e-01 -5.33060193e-01 -1.80144310e-01
6.44679010e-01 7.68879279e-02 5.10506928e-01 -1.28697360e+00
-4.10314947e-01 -6.23070776e-01 -1.42305270e-01 -1.41546834e+00
3.83237213e-01 -6.84004188e-01 6.47227466e-02 -1.14770627e+00
2.69347042e-01 -6.84341073e-01 -4.03671056e-01 7.15640604e-01
3.04986507e-01 4.29167300e-01 1.22468971e-01 2.28704780e-01
-5.93166828e-01 2.56925195e-01 8.38293135e-01 -6.47722259e-02
-1.75732389e-01 -5.21918423e-02 -9.77985859e-01 7.05652416e-01
8.88429284e-01 -6.92426980e-01 -4.96131808e-01 -4.55270350e-01
1.66764334e-01 1.82832435e-01 2.07614720e-01 -9.68304694e-01
5.12245953e-01 3.29836421e-02 3.28914702e-01 -5.57549000e-01
4.82258767e-01 -8.83740664e-01 1.38846442e-01 5.55354238e-01
-5.69994152e-01 1.03558823e-01 2.37007663e-01 3.60970974e-01
-1.63801253e-01 -7.51164734e-01 8.98758113e-01 -4.75046843e-01
-3.44227791e-01 3.36274058e-01 -3.82395774e-01 2.38604754e-01
6.76945329e-01 -6.63194597e-01 -4.10007685e-01 -5.42279184e-01
-4.60446924e-01 3.28331292e-02 2.11730227e-01 -1.63256571e-01
4.54636663e-01 -1.28457034e+00 -3.33546162e-01 3.07605833e-01
-5.02333820e-01 -2.31057387e-02 4.53319773e-03 7.51822650e-01
-3.26086760e-01 8.83928120e-01 -3.64619702e-01 -1.46968365e-01
-9.08987045e-01 1.74779370e-01 3.68248850e-01 -6.72688127e-01
-5.48783183e-01 1.11724341e+00 2.56571501e-01 -1.53165460e-01
6.66487277e-01 -2.37936944e-01 4.17656787e-02 2.16168184e-02
3.65619540e-01 2.84008861e-01 1.33119687e-01 -1.67459905e-01
-1.01382576e-01 -3.20631675e-02 -5.79223573e-01 1.60895735e-02
1.63476872e+00 3.72138590e-01 1.37631536e-01 -5.56022823e-02
1.25880063e+00 -6.31208122e-01 -1.13229954e+00 -3.59542727e-01
1.77433826e-02 -4.70278919e-01 4.85146970e-01 -4.23321307e-01
-1.31539440e+00 4.89099711e-01 1.60962880e-01 5.05330920e-01
8.16223145e-01 -9.35084652e-03 7.41224766e-01 1.02263272e+00
5.29445052e-01 -1.18655944e+00 -2.15301737e-01 8.35292578e-01
6.40302181e-01 -9.84121978e-01 1.70430243e-01 1.69274017e-01
-4.96525675e-01 5.64578533e-01 6.11922681e-01 -5.79095602e-01
8.72258663e-01 4.94971007e-01 -5.49204707e-01 -2.75033563e-01
-1.18357289e+00 1.09802522e-01 -1.14495441e-01 5.71169853e-01
2.69347504e-02 -1.22490041e-01 1.40170619e-01 5.76165736e-01
-5.72162569e-01 2.02034369e-01 2.96227276e-01 6.42286301e-01
-7.66457736e-01 -6.32758021e-01 -1.60021260e-01 1.00704610e+00
-3.19943756e-01 -7.02682674e-01 -2.82925695e-01 1.00547802e+00
-2.30751306e-01 7.67791927e-01 4.48752224e-01 -3.41836125e-01
-4.97467034e-02 3.19999233e-02 5.73293626e-01 -7.12766945e-01
-9.23683107e-01 -3.10120638e-02 4.92112607e-01 -7.61180282e-01
-3.31207484e-01 -3.35296690e-01 -7.28453636e-01 -8.99751008e-01
-6.57268941e-01 4.03026789e-01 2.67871082e-01 9.27440584e-01
2.33004883e-01 -4.64962162e-02 6.39359117e-01 -8.94769967e-01
-1.00136459e+00 -1.08013594e+00 -1.05836344e+00 -1.29277661e-01
1.68646462e-02 -5.76188445e-01 -8.88758779e-01 -3.43249798e-01]
|
[8.476202964782715, 3.3509488105773926]
|
9fbca879-edbf-4e30-a5d0-33a7a5b08a2b
|
nerfinvertor-high-fidelity-nerf-gan-inversion
|
2211.17235
| null |
https://arxiv.org/abs/2211.17235v1
|
https://arxiv.org/pdf/2211.17235v1.pdf
|
NeRFInvertor: High Fidelity NeRF-GAN Inversion for Single-shot Real Image Animation
|
Nerf-based Generative models have shown impressive capacity in generating high-quality images with consistent 3D geometry. Despite successful synthesis of fake identity images randomly sampled from latent space, adopting these models for generating face images of real subjects is still a challenging task due to its so-called inversion issue. In this paper, we propose a universal method to surgically fine-tune these NeRF-GAN models in order to achieve high-fidelity animation of real subjects only by a single image. Given the optimized latent code for an out-of-domain real image, we employ 2D loss functions on the rendered image to reduce the identity gap. Furthermore, our method leverages explicit and implicit 3D regularizations using the in-domain neighborhood samples around the optimized latent code to remove geometrical and visual artifacts. Our experiments confirm the effectiveness of our method in realistic, high-fidelity, and 3D consistent animation of real faces on multiple NeRF-GAN models across different datasets.
|
['Yun Fu', 'Xin Tong', 'Jiaolong Yang', 'HsiangTao Wu', 'Kamran Ghasedi', 'Yu Yin']
|
2022-11-30
| null |
http://openaccess.thecvf.com//content/CVPR2023/html/Yin_NeRFInvertor_High_Fidelity_NeRF-GAN_Inversion_for_Single-Shot_Real_Image_Animation_CVPR_2023_paper.html
|
http://openaccess.thecvf.com//content/CVPR2023/papers/Yin_NeRFInvertor_High_Fidelity_NeRF-GAN_Inversion_for_Single-Shot_Real_Image_Animation_CVPR_2023_paper.pdf
|
cvpr-2023-1
|
['image-animation']
|
['computer-vision']
|
[ 3.34727466e-01 4.65049654e-01 2.57621258e-01 -3.05444598e-01
-9.77855444e-01 -5.93032598e-01 7.70945191e-01 -8.90583754e-01
7.27190748e-02 6.82127476e-01 2.80522674e-01 1.89357102e-01
3.23420852e-01 -6.12062395e-01 -8.28716338e-01 -5.83315551e-01
2.70446151e-01 5.18780112e-01 -5.31839848e-01 -1.47542274e-02
-1.14652857e-01 5.43211102e-01 -1.33759260e+00 5.44103123e-02
7.90097475e-01 5.85636079e-01 -2.44631633e-01 6.52852058e-01
2.87458330e-01 4.47871774e-01 -6.25680625e-01 -8.28678548e-01
6.70758009e-01 -1.00652671e+00 -4.19579655e-01 3.37710857e-01
9.64154601e-01 -7.71328390e-01 -2.94071287e-01 1.07005918e+00
4.89043444e-01 -2.13639252e-02 8.89247954e-01 -1.34677327e+00
-8.55206072e-01 -5.92421405e-02 -7.23133862e-01 -5.23735881e-01
5.27066767e-01 4.03153598e-01 5.67966521e-01 -7.26373971e-01
1.03647912e+00 1.51066446e+00 6.76630318e-01 1.07510471e+00
-1.58167970e+00 -1.00574207e+00 -3.99347275e-01 -3.80772263e-01
-1.40190625e+00 -7.02184498e-01 9.75030065e-01 -4.56895411e-01
3.09288889e-01 1.66473076e-01 7.64626622e-01 1.55758023e+00
3.65864262e-02 1.93332732e-01 1.29817057e+00 -3.05558324e-01
-5.78704523e-04 1.04140323e-02 -8.51648390e-01 8.58125329e-01
9.70120952e-02 2.61093348e-01 -4.95147973e-01 -4.15652633e-01
1.43718004e+00 -2.93751478e-01 -2.06966206e-01 -3.11185747e-01
-9.81619239e-01 1.01028597e+00 2.69205809e-01 -1.52687117e-01
-3.80601674e-01 4.57157165e-01 -1.06407642e-01 -2.06781775e-01
6.44505322e-01 6.15083456e-01 1.29747540e-01 -1.88711941e-01
-1.04511559e+00 4.89169210e-01 4.11077917e-01 8.70761335e-01
6.61331654e-01 4.96937960e-01 -3.69168483e-02 7.19806492e-01
2.07391277e-01 6.20276451e-01 1.94586992e-01 -1.45852697e+00
1.03900500e-01 2.52188444e-01 2.30464086e-01 -1.11992013e+00
3.13064694e-01 -3.45292449e-01 -9.60825682e-01 5.71792662e-01
3.98594558e-01 -9.59223211e-02 -1.09156108e+00 2.07011080e+00
5.44050336e-01 5.52799523e-01 -8.32734406e-02 8.92626107e-01
4.82356638e-01 6.23182297e-01 -4.27318141e-02 -5.98784946e-02
1.08984733e+00 -7.34314501e-01 -5.84988594e-01 -1.85985163e-01
1.03659861e-01 -9.02386546e-01 1.18499851e+00 8.64466131e-02
-1.35898042e+00 -3.61312211e-01 -8.46967876e-01 -2.53865570e-01
3.15159827e-01 1.13667905e-01 5.13387799e-01 5.80520689e-01
-1.00375080e+00 5.76207697e-01 -6.15455389e-01 -7.07780123e-02
8.13434124e-01 1.90876007e-01 -5.15619099e-01 -3.82694416e-03
-9.26354289e-01 4.59126234e-01 -2.09971756e-01 6.52648658e-02
-1.19790292e+00 -9.76315737e-01 -8.85039985e-01 -1.83978185e-01
-1.09011540e-02 -1.04859638e+00 8.73640954e-01 -1.11674416e+00
-1.81952810e+00 1.11887491e+00 -1.86992273e-01 -2.08833337e-01
9.07147467e-01 -4.39014435e-02 -1.09172523e-01 2.77035654e-01
1.97718054e-01 9.23443079e-01 1.35788178e+00 -1.69366610e+00
1.78256333e-01 -1.55238315e-01 -1.16393626e-01 1.27815306e-01
-1.31715685e-01 -1.76419273e-01 -3.21277380e-01 -9.52109694e-01
-9.78210419e-02 -1.07234108e+00 -7.84166232e-02 2.41460755e-01
-5.29121637e-01 3.90918970e-01 8.18474233e-01 -8.33500981e-01
6.79395080e-01 -2.15808487e+00 1.30600572e-01 2.31645733e-01
4.16906953e-01 4.19160947e-02 -2.89857745e-01 3.79157104e-02
-4.34852019e-02 2.55663007e-01 -2.45920420e-01 -7.70688295e-01
-6.60444722e-02 1.26754656e-01 -5.80489814e-01 5.02771795e-01
3.25917155e-01 1.12061799e+00 -8.29150140e-01 -5.94655097e-01
1.00419745e-01 1.14080763e+00 -8.34719300e-01 4.19891953e-01
-2.00451806e-01 1.14988279e+00 -3.73749405e-01 3.69643360e-01
7.80594587e-01 -3.16843808e-01 1.33167863e-01 -3.03685993e-01
3.74961495e-01 1.19696953e-03 -8.28033626e-01 1.70533288e+00
-5.14211476e-01 5.51663041e-01 1.49528995e-01 -4.34226632e-01
7.52611279e-01 2.88442641e-01 4.05861437e-01 -5.84636867e-01
1.53046727e-01 1.69196814e-01 -1.64492413e-01 -1.53456181e-01
3.17536920e-01 -4.74779725e-01 -1.15612652e-02 6.04479015e-01
-7.59670585e-02 -4.94966149e-01 -2.79586822e-01 2.41001904e-01
5.62981725e-01 4.72187340e-01 -1.85703278e-01 -5.36935963e-03
1.94448411e-01 -2.95214802e-01 4.77493793e-01 4.08520222e-01
1.09481074e-01 1.21629810e+00 4.42189366e-01 -3.52640569e-01
-1.66393161e+00 -1.10407865e+00 3.97812203e-02 2.90320724e-01
-4.30264100e-02 -3.75062734e-01 -1.19083142e+00 -4.45022821e-01
-1.80057332e-01 6.45847321e-01 -9.41645145e-01 -1.73795626e-01
-7.55756199e-01 -4.53265429e-01 7.49183834e-01 9.15600434e-02
6.32237732e-01 -8.13619196e-01 -4.49366510e-01 -1.14074238e-02
-3.56028199e-01 -1.20701885e+00 -8.31668258e-01 -7.85300016e-01
-6.00695133e-01 -8.29040051e-01 -9.16619182e-01 -5.59248447e-01
1.14207506e+00 -1.88504364e-02 1.16866601e+00 1.08084001e-01
-3.25379699e-01 2.31310040e-01 7.64127001e-02 9.80964378e-02
-8.17580640e-01 -4.16407377e-01 3.80203985e-02 3.40072870e-01
-4.10510480e-01 -8.11145067e-01 -7.81958342e-01 3.56842339e-01
-7.52203703e-01 3.83942515e-01 9.74399969e-02 1.02718794e+00
5.94709456e-01 -1.79982439e-01 3.08913022e-01 -7.64780343e-01
4.45049375e-01 -3.18496227e-02 -7.55664408e-01 7.18163103e-02
-3.95300955e-01 -7.11009197e-04 6.54065967e-01 -5.87063134e-01
-1.09499109e+00 3.27204503e-02 -2.71337088e-02 -8.22870672e-01
4.65634465e-02 -2.72540510e-01 -2.60491461e-01 -3.21681380e-01
4.97231543e-01 1.84337631e-01 2.17974141e-01 -2.31497079e-01
5.32600760e-01 1.55854955e-01 7.22913384e-01 -7.67464995e-01
1.22458363e+00 7.65048563e-01 2.87636727e-01 -6.52206957e-01
-6.26973033e-01 4.48017746e-01 -4.24679190e-01 -2.60449171e-01
1.09034359e+00 -1.00175869e+00 -7.57437587e-01 6.65453434e-01
-1.19232905e+00 -5.16738415e-01 -4.26133752e-01 2.17758447e-01
-7.33102441e-01 3.11007529e-01 -5.55548131e-01 -7.48146057e-01
-3.02981585e-01 -1.27353060e+00 1.49167073e+00 1.90017372e-01
-3.60432148e-01 -8.44137132e-01 -1.27213830e-02 5.23220241e-01
3.42867166e-01 7.90009797e-01 8.00064027e-01 2.73286968e-01
-8.20144713e-01 3.02405935e-02 -1.32451117e-01 2.96711892e-01
2.44377449e-01 1.21274635e-01 -1.02518058e+00 -3.32333922e-01
2.12560408e-02 -4.81749624e-01 3.91020715e-01 3.36523473e-01
1.07289565e+00 -6.19093001e-01 -1.98545456e-01 1.12787867e+00
1.19177210e+00 -3.89513001e-02 7.40301967e-01 -2.73638427e-01
1.06157219e+00 5.46417117e-01 2.32385397e-01 3.36793900e-01
1.41531870e-01 9.01405156e-01 2.92079777e-01 -1.88796669e-01
-5.52779198e-01 -7.29526222e-01 3.21173966e-01 4.77408856e-01
-1.69383898e-01 -1.43698618e-01 -5.16400039e-01 3.67129087e-01
-1.29991090e+00 -1.16680264e+00 1.33691296e-01 2.24932098e+00
1.00145638e+00 -2.06313506e-01 1.08400986e-01 -3.98173541e-01
7.46401250e-01 -1.40460730e-02 -5.87400258e-01 -1.53251693e-01
-2.19507366e-01 3.58365774e-01 2.70493358e-01 7.57824779e-01
-7.68132985e-01 1.03579414e+00 6.19948387e+00 9.13928926e-01
-1.25472939e+00 2.40854502e-01 9.94276226e-01 -2.79459983e-01
-7.69329846e-01 5.93398511e-02 -5.70123255e-01 5.60118139e-01
6.40875936e-01 -6.80022240e-02 6.64194465e-01 6.58027887e-01
2.68113315e-01 2.54118949e-01 -8.28707039e-01 1.26426160e+00
2.11749092e-01 -1.64038908e+00 2.02543303e-01 5.09856582e-01
1.19194984e+00 -6.31530583e-01 4.44471896e-01 -2.65599668e-01
4.02461857e-01 -1.39760435e+00 8.99163485e-01 3.69834542e-01
1.67542517e+00 -7.77512550e-01 3.38968597e-02 -5.14211282e-02
-9.06109035e-01 4.89472419e-01 -1.09841347e-01 4.07127082e-01
4.52691287e-01 4.51567292e-01 -8.89427364e-01 2.49600217e-01
3.84474695e-01 4.87382084e-01 -4.83253479e-01 3.57075989e-01
-2.92046160e-01 4.95678484e-01 -4.13987756e-01 6.41805232e-01
5.45234494e-02 -5.39440095e-01 5.85788429e-01 6.21884227e-01
5.79740584e-01 7.67489076e-02 -2.40849003e-01 1.37986636e+00
-3.14658225e-01 -2.81276911e-01 -8.33159089e-01 -1.20928727e-01
3.83732706e-01 1.17333615e+00 -4.81719196e-01 -1.49279550e-01
1.85604304e-01 1.37436569e+00 1.46804199e-01 4.23131853e-01
-1.12761581e+00 4.36321460e-02 8.96413207e-01 4.38981891e-01
2.30505109e-01 -3.69895786e-01 -3.53528857e-01 -1.19466698e+00
3.37966233e-02 -1.03410125e+00 -1.52314574e-01 -1.06972980e+00
-1.25720322e+00 8.88663769e-01 -1.05721086e-01 -1.29351091e+00
-5.96134245e-01 -1.16645776e-01 -5.21856010e-01 1.10135317e+00
-1.11291385e+00 -1.58495653e+00 -3.61375391e-01 6.50299907e-01
1.23251081e-01 -1.37892991e-01 8.94871771e-01 2.56678522e-01
-3.39455664e-01 9.40593660e-01 1.93117149e-02 1.94844365e-01
8.37624133e-01 -8.45176220e-01 7.02134192e-01 7.65976191e-01
1.03993706e-01 7.67677784e-01 6.48911595e-01 -8.63461554e-01
-1.31741548e+00 -1.07396698e+00 4.52882797e-01 -5.56981564e-01
1.16119206e-01 -6.63518012e-01 -7.94030845e-01 6.87233984e-01
2.89017558e-01 9.55602527e-02 4.81922448e-01 -5.20529449e-01
-6.17036223e-01 1.67028368e-01 -1.50109208e+00 9.18426037e-01
1.29363203e+00 -6.68053508e-01 8.36034492e-02 3.41156542e-01
5.19236207e-01 -5.87582588e-01 -7.91491151e-01 2.04852119e-01
7.46291637e-01 -9.19731438e-01 1.16820705e+00 -4.33389157e-01
7.32249975e-01 -3.83679032e-01 1.04729056e-01 -1.29706275e+00
-4.94694039e-02 -1.27374268e+00 2.06805915e-01 1.37561429e+00
1.05269682e-02 -5.26879251e-01 1.02128923e+00 6.72592640e-01
2.10293978e-01 -5.21701396e-01 -9.01274860e-01 -6.71106219e-01
7.29954466e-02 -1.33273840e-01 7.56421328e-01 9.78331089e-01
-6.91872895e-01 2.35875502e-01 -9.28104341e-01 7.58492202e-02
1.11103320e+00 1.98757380e-01 1.16556692e+00 -8.84186625e-01
-3.51895422e-01 -2.69820273e-01 -3.31578732e-01 -9.89262044e-01
5.20421028e-01 -7.32859135e-01 -9.04302672e-02 -1.00227714e+00
8.67691934e-02 -5.61915278e-01 7.22957075e-01 2.53960162e-01
-9.24595352e-03 8.90689313e-01 2.05706149e-01 3.20606321e-01
1.23695113e-01 8.30083549e-01 1.75249779e+00 1.21252291e-01
1.45964175e-01 -4.69935834e-01 -6.61649048e-01 7.15804279e-01
5.34685850e-01 -5.87600291e-01 -5.56126356e-01 -4.52366173e-01
1.50033772e-01 5.67404833e-03 8.74339521e-01 -7.52973139e-01
-2.29098126e-01 -2.68895358e-01 6.43541276e-01 -1.41152054e-01
8.63222301e-01 -6.54128432e-01 8.73447180e-01 2.11315468e-01
-2.21946001e-01 -9.57299247e-02 -1.15162125e-02 4.09020066e-01
3.06194788e-03 6.63259998e-02 1.22861886e+00 -1.65659547e-01
-3.88252437e-02 5.26235521e-01 2.12471828e-01 3.36089283e-01
7.74304509e-01 -3.28868151e-01 -1.29848540e-01 -9.37363386e-01
-4.14524227e-01 -3.58106256e-01 1.17615807e+00 1.81072429e-01
6.89488947e-01 -1.62597561e+00 -8.17814469e-01 5.82532465e-01
-1.27214938e-01 1.65145025e-01 3.95105332e-01 3.07835251e-01
-6.59246802e-01 -8.87678713e-02 -3.79831910e-01 -5.56248665e-01
-1.24757564e+00 2.23177612e-01 4.90618914e-01 -1.11548744e-01
-7.20147729e-01 9.28500831e-01 6.33104384e-01 -4.17018920e-01
-2.25148946e-01 2.75072873e-01 4.44458306e-01 -3.45030814e-01
3.19711208e-01 5.33449687e-02 -2.41980493e-01 -1.10192978e+00
-1.35930896e-01 9.83112097e-01 1.67945430e-01 -5.61843574e-01
1.20392728e+00 -1.46565847e-02 -1.17881699e-02 -1.48519024e-01
1.43155384e+00 4.74880755e-01 -1.85423231e+00 3.12720478e-01
-8.02644014e-01 -1.04672980e+00 -2.66139090e-01 -5.85541606e-01
-1.32545507e+00 7.09450960e-01 3.45205277e-01 -3.46338689e-01
9.29715395e-01 -1.11890405e-01 1.08145952e+00 -3.88345450e-01
3.27001423e-01 -6.06713414e-01 3.08782995e-01 8.56194273e-02
1.06233943e+00 -9.96579051e-01 3.61334980e-02 -5.49742043e-01
-6.53030396e-01 8.48008454e-01 5.61648905e-01 -3.37198883e-01
3.48024637e-01 4.86626923e-02 6.06938004e-02 -1.82260662e-01
-3.06405902e-01 4.27926987e-01 2.81354249e-01 7.83534765e-01
2.63704896e-01 -1.71765555e-02 1.19578563e-01 7.64904767e-02
-6.43346190e-01 -8.80792663e-02 6.54203355e-01 4.79377776e-01
4.51967537e-01 -1.19941318e+00 -4.78394836e-01 3.46015021e-02
-4.36828226e-01 -5.14075607e-02 -2.95031250e-01 6.63850665e-01
9.63137969e-02 6.21729553e-01 2.06021443e-02 -2.28618070e-01
9.96043789e-04 3.07491776e-02 9.88102198e-01 -3.39578480e-01
-3.94105881e-01 2.07886264e-01 -5.60099632e-02 -6.11320198e-01
-2.04044387e-01 -6.52173042e-01 -1.05213225e+00 -5.40828168e-01
2.43241135e-02 3.05006700e-03 6.20834231e-01 5.32808602e-01
7.88247108e-01 1.13101803e-01 8.00125659e-01 -1.01164806e+00
-4.38524276e-01 -5.95795691e-01 -3.36228579e-01 9.32899058e-01
3.35062802e-01 -6.64112031e-01 -5.62378943e-01 3.86846900e-01]
|
[12.547532081604004, -0.3553749918937683]
|
0550eae1-87d2-47c6-a30e-5d03acdc6cd7
|
effective-resistance-for-pandemics-mobility
|
2111.02449
| null |
https://arxiv.org/abs/2111.02449v3
|
https://arxiv.org/pdf/2111.02449v3.pdf
|
Effective Resistance for Pandemics: Mobility Network Sparsification for High-Fidelity Epidemic Simulation
|
Network science has increasingly become central to the field of epidemiology and our ability to respond to infectious disease threats. However, many networks derived from modern datasets are not just large, but dense, with a high ratio of edges to nodes. This includes human mobility networks where most locations have a large number of links to many other locations. Simulating large-scale epidemics requires substantial computational resources and in many cases is practically infeasible. One way to reduce the computational cost of simulating epidemics on these networks is sparsification, where a representative subset of edges is selected based on some measure of their importance. We test several sparsification strategies, ranging from naive thresholding to random sampling of edges, on mobility data from the U.S. Following recent work in computer science, we find that the most accurate approach uses the effective resistances of edges, which prioritizes edges that are the only efficient way to travel between their endpoints. The resulting sparse network preserves many aspects of the behavior of an SIR model, including both global quantities, like the epidemic size, and local details of stochastic events, including the probability each node becomes infected and its distribution of arrival times. This holds even when the sparse network preserves fewer than $10\%$ of the edges of the original network. In addition to its practical utility, this method helps illuminate which links of a weighted, undirected network are most important to disease spread.
|
['Cristopher Moore', 'Samuel V. Scarpino', 'Alexander M. Mercier']
|
2021-11-03
| null | null | null | null |
['epidemiology']
|
['medical']
|
[ 9.76287201e-02 1.17798418e-01 -2.41014466e-01 2.52096832e-01
1.14685535e-01 -6.75187051e-01 4.92686361e-01 3.65287334e-01
-5.67054152e-01 9.85109091e-01 1.36354879e-01 -4.21796948e-01
-6.59951746e-01 -1.22098243e+00 -3.13940257e-01 -6.70665383e-01
-1.01588786e+00 9.53307867e-01 4.03724134e-01 -4.88256454e-01
3.05677447e-02 5.85037649e-01 -6.72669888e-01 -5.00748873e-01
6.19997323e-01 3.34355623e-01 -1.31827578e-01 5.92153788e-01
1.17292115e-02 2.46331394e-01 -7.05930054e-01 -3.67909938e-01
1.17074415e-01 -3.43693882e-01 -5.39864063e-01 -1.87673524e-01
-5.29482663e-01 -8.80722478e-02 -6.60804629e-01 8.30263257e-01
3.66547674e-01 -1.08963482e-01 8.03379536e-01 -1.47534800e+00
-6.64871708e-02 6.92332983e-01 -9.15899456e-01 6.28409266e-01
3.32420588e-01 -4.95044924e-02 6.80827618e-01 -1.13436813e-02
1.04427028e+00 1.08441687e+00 9.76328135e-01 6.21609874e-02
-1.37044168e+00 -6.23339117e-01 -1.63956098e-02 -4.44262445e-01
-1.54820883e+00 -3.96240771e-01 4.91187662e-01 -3.04976374e-01
6.30407333e-01 4.24270481e-01 8.52191210e-01 8.47647429e-01
3.69682819e-01 -1.39555067e-01 5.67132175e-01 1.40527159e-01
1.68347016e-01 -1.81432575e-01 -1.74878150e-01 4.69597459e-01
1.00680196e+00 -1.37214407e-01 -5.14437445e-02 -8.86466503e-01
8.96824479e-01 4.37644124e-01 -3.94954056e-01 -2.20773906e-01
-1.13690114e+00 8.72581959e-01 4.61569518e-01 3.47426772e-01
-5.11210620e-01 2.01261520e-01 1.36197418e-01 5.00043929e-01
5.45469940e-01 3.18004042e-01 -4.09714788e-01 -4.96125855e-02
-8.53776693e-01 1.71729103e-01 1.14957559e+00 5.89756906e-01
5.89683175e-01 -3.60797733e-01 5.12047827e-01 5.04041612e-01
1.86274171e-01 8.79701555e-01 -3.94792229e-01 -9.76831019e-01
4.14902866e-01 6.59350634e-01 1.26901388e-01 -1.48519111e+00
-7.02100873e-01 -5.32158077e-01 -1.46949279e+00 -2.74009258e-01
4.45874602e-01 -6.33703589e-01 -5.61820269e-01 1.93757641e+00
4.15257812e-01 1.56962678e-01 -2.59872168e-01 3.00737441e-01
1.44746572e-01 8.27092469e-01 8.07303116e-02 -5.63230574e-01
1.05851519e+00 5.90848178e-02 -4.32463825e-01 -2.24297434e-01
5.65690517e-01 -4.93606567e-01 7.00902417e-02 -6.22546040e-02
-1.10325050e+00 4.47488219e-01 -3.93995017e-01 8.04066658e-01
-3.41384858e-01 -8.16483200e-01 6.07429147e-01 5.63446939e-01
-1.34275913e+00 4.31100607e-01 -7.96898723e-01 -8.20955992e-01
4.89019036e-01 5.08175075e-01 -3.15257370e-01 -9.56574604e-02
-1.37776673e+00 5.55479944e-01 -7.62159824e-02 -7.93953761e-02
-3.30770433e-01 -6.25925601e-01 -6.00733578e-01 1.24181271e-01
4.79235381e-01 -7.15065360e-01 5.36905766e-01 -4.67386365e-01
-4.75437582e-01 3.47072780e-01 -2.58864969e-01 -4.43343341e-01
3.75493288e-01 5.02503574e-01 -4.00635958e-01 4.56020206e-01
2.20503554e-01 2.50054657e-01 5.22000194e-01 -1.21233821e+00
-4.41166759e-01 -3.73685181e-01 -7.45581388e-02 -3.83387529e-03
-3.82152349e-01 1.75021738e-01 -5.76156795e-01 -5.35810828e-01
9.57864821e-02 -1.11749327e+00 -7.44998097e-01 -4.46293913e-02
-5.40287793e-01 5.48335686e-02 6.81669176e-01 -3.59091401e-01
1.38781512e+00 -1.75917065e+00 1.75454259e-01 1.08812344e+00
7.94630885e-01 4.28675078e-02 -1.31140009e-01 1.03868008e+00
4.14735287e-01 5.23925245e-01 -4.92630929e-01 2.76405793e-02
-4.23189133e-01 1.57208443e-01 9.67116654e-03 7.05656707e-01
-1.27921373e-01 6.52565300e-01 -1.08528805e+00 -4.36787158e-01
-1.41002715e-01 6.99272096e-01 -5.26582778e-01 -2.98910618e-01
2.56541044e-01 2.93145478e-01 -7.81250536e-01 2.71981448e-01
5.56713939e-01 -7.22048521e-01 7.64141798e-01 3.57579023e-01
1.74334660e-01 8.68586153e-02 -1.14952326e+00 6.13886654e-01
-1.94535077e-01 4.60647762e-01 5.81849694e-01 -8.75477970e-01
5.18390179e-01 2.53371835e-01 1.07025170e+00 -1.96037620e-01
1.41169325e-01 3.22243683e-02 2.08038047e-01 -2.09681958e-01
1.85112745e-01 7.06654266e-02 -4.17880602e-02 9.52743292e-01
-5.57446897e-01 5.55970483e-02 5.51992893e-01 9.98335898e-01
1.82376337e+00 -9.67779458e-01 3.66653621e-01 -3.90507460e-01
-1.81021586e-01 2.05046713e-01 5.42743027e-01 8.41471195e-01
-9.06096250e-02 1.61794677e-01 9.56174552e-01 -1.40935898e-01
-1.09678292e+00 -1.34141779e+00 -6.00091740e-02 5.69407701e-01
3.43525797e-01 -3.90261084e-01 -6.20073140e-01 -2.77575582e-01
2.25494817e-01 -4.37099971e-02 -7.60647774e-01 -1.44584164e-01
-4.71661121e-01 -1.27418160e+00 5.11144042e-01 -8.65237564e-02
1.80432081e-01 -9.48243380e-01 -2.57855386e-01 3.04195583e-01
-2.55231291e-01 -9.49598134e-01 -4.83998686e-01 -5.25035039e-02
-1.00782669e+00 -1.45804012e+00 -7.15015411e-01 -5.96323252e-01
1.17390072e+00 5.19497991e-01 1.10181308e+00 7.03004658e-01
-3.65777880e-01 3.78048122e-01 -1.65703997e-01 -2.63191462e-02
-3.57441902e-01 1.93282336e-01 2.38010183e-01 -2.33312979e-01
4.38833572e-02 -8.33211541e-01 -7.42311358e-01 3.69185299e-01
-8.88407707e-01 -5.94500959e-01 4.67046618e-01 2.49163792e-01
1.69846356e-01 5.73110640e-01 7.05421865e-01 -1.09122765e+00
9.19175208e-01 -1.37253678e+00 -2.85965174e-01 8.96089748e-02
-2.20437691e-01 -2.97905862e-01 5.59490740e-01 -3.72440904e-01
-5.08606493e-01 -3.74691874e-01 2.30674908e-01 3.88972372e-01
2.76534706e-01 6.22983098e-01 3.09967756e-01 -1.03981391e-01
4.11358029e-01 -9.41114649e-02 1.88698545e-01 -1.25161350e-01
3.57930660e-02 4.66095954e-01 -6.80203736e-02 -1.35668367e-01
9.22166109e-01 8.48283410e-01 3.20070326e-01 -1.43882704e+00
3.81324552e-02 -4.26525116e-01 -2.34950408e-01 -1.72457963e-01
2.39036724e-01 -5.33119559e-01 -1.01291180e+00 4.64834094e-01
-9.72566128e-01 -3.70887637e-01 -1.84444338e-01 3.89348477e-01
-3.09385601e-02 2.89880753e-01 -9.52102542e-01 -9.30419266e-01
-7.11107701e-02 -7.17725158e-01 5.45049429e-01 -6.99143112e-02
-5.01705825e-01 -1.40132439e+00 1.84256360e-01 -1.27228394e-01
8.68250668e-01 4.46690679e-01 8.90520334e-01 -5.39752185e-01
-3.38747144e-01 -4.27673459e-01 -3.21179390e-01 -5.61719954e-01
3.53448510e-01 2.46985674e-01 6.82115108e-02 -5.32413423e-01
-4.23308611e-01 2.28003189e-01 7.98112988e-01 8.40669274e-01
5.54421484e-01 -4.73529249e-01 -1.08847940e+00 2.65885293e-01
1.33014750e+00 1.22801721e-01 6.02005482e-01 -5.59658408e-02
4.15858299e-01 7.98179805e-01 -2.83152517e-02 5.54232836e-01
4.76031631e-01 4.19811219e-01 3.11555594e-01 -2.82790542e-01
2.10198358e-01 -7.67143518e-02 2.35095620e-01 7.48112261e-01
-2.48334378e-01 -7.86933064e-01 -1.11796844e+00 6.17775798e-01
-1.48685813e+00 -1.18840539e+00 -1.45865932e-01 2.35494232e+00
6.67879164e-01 4.73608017e-01 5.31760573e-01 9.70509648e-02
1.01580441e+00 2.04115182e-01 -4.36470687e-01 -6.71882480e-02
-1.45532429e-01 -1.50977019e-02 8.43781114e-01 8.06027293e-01
-6.39623404e-01 3.73296916e-01 7.16263247e+00 5.22599697e-01
-7.01059937e-01 -1.65245488e-01 8.27040851e-01 -2.11795256e-01
-4.48866308e-01 -1.46867156e-01 -4.65106159e-01 6.16276562e-01
1.11620188e+00 -3.56282145e-01 5.98585784e-01 2.99913675e-01
6.58783853e-01 -2.64170647e-01 -2.70814925e-01 3.76087517e-01
-3.12349081e-01 -1.27908123e+00 1.65566953e-03 6.41795754e-01
8.35872412e-01 2.41324991e-01 -1.29031718e-01 -4.00312960e-01
8.95052075e-01 -9.87660885e-01 -1.09141827e-01 3.82495970e-01
7.86769509e-01 -7.91644633e-01 6.84354544e-01 2.96219856e-01
-1.42105162e+00 -4.32332791e-03 -1.31348714e-01 -3.39115448e-02
6.65431678e-01 1.08468473e+00 -4.82496023e-01 5.46056405e-02
6.54329300e-01 5.50098658e-01 -1.36718243e-01 1.18149531e+00
2.25910649e-01 6.10908568e-01 -1.03097546e+00 -1.34283334e-01
1.10701360e-01 -4.32972193e-01 8.64821732e-01 9.22760010e-01
3.10460120e-01 3.05456161e-01 1.58066787e-02 3.47746313e-01
-3.47088546e-01 -1.58485532e-01 -1.12388289e+00 -1.07951730e-01
1.03688979e+00 1.06152844e+00 -1.28149652e+00 -1.78571135e-01
-6.90685362e-02 4.72940803e-01 5.16882204e-02 6.69453979e-01
-5.17400622e-01 -6.57332480e-01 8.93891752e-01 8.02109241e-01
1.49809748e-01 -5.34870863e-01 1.93574920e-01 -6.41824841e-01
-3.18832308e-01 -6.79636300e-01 2.91282594e-01 -1.66789979e-01
-1.16935349e+00 6.66234314e-01 1.40216248e-02 -7.20297754e-01
-1.93485111e-01 9.42406505e-02 -8.53810489e-01 4.72393006e-01
-1.03295135e+00 -4.39089864e-01 8.86688232e-02 5.92693746e-01
-8.38605240e-02 2.60815531e-01 3.72392863e-01 5.06256342e-01
-5.93641281e-01 1.71644643e-01 4.72178459e-01 8.23062509e-02
2.02884793e-01 -7.26559639e-01 6.35865808e-01 7.09519207e-01
-1.42365649e-01 8.71839583e-01 7.78264880e-01 -1.23540151e+00
-1.36358166e+00 -9.33136106e-01 1.05986249e+00 -3.30987394e-01
8.79311681e-01 -2.59900391e-01 -5.86458027e-01 6.43704534e-01
-1.07939258e-01 -1.12850040e-01 3.49798530e-01 1.05308451e-01
9.46609527e-02 3.54943844e-03 -1.39603555e+00 8.84339094e-01
1.33120036e+00 -6.27548024e-02 8.98231789e-02 3.95094872e-01
5.02309799e-01 3.30217272e-01 -8.56260240e-01 1.60177931e-01
4.73519742e-01 -7.02572048e-01 1.13174045e+00 -5.67725420e-01
-9.40348208e-03 -3.95228676e-02 3.28746855e-01 -1.51247132e+00
-3.28276932e-01 -1.00141633e+00 3.63937840e-02 9.31345999e-01
6.18112564e-01 -1.03452265e+00 9.31621313e-01 3.39597523e-01
7.88897395e-01 -8.38111699e-01 -8.66591096e-01 -7.44056344e-01
-4.66660708e-02 6.89401664e-03 4.42291498e-01 9.75233555e-01
-1.54075967e-02 2.11102769e-01 -2.91965902e-01 1.60487071e-02
1.00123155e+00 -3.05723935e-01 6.49166286e-01 -1.73453248e+00
1.48680955e-01 -4.85507339e-01 -2.83290684e-01 -7.71546423e-01
-3.06316525e-01 -4.21908468e-01 -3.18286300e-01 -1.62237751e+00
2.31459722e-01 -9.92616951e-01 1.63800254e-01 9.55232680e-02
9.06327441e-02 4.79843467e-01 -2.17061102e-01 4.03131157e-01
-6.35017097e-01 9.17127077e-03 1.06709099e+00 1.19683938e-02
-5.71761549e-01 2.31025264e-01 -7.26883590e-01 8.53376985e-01
1.14662862e+00 -8.34908485e-01 -5.11601925e-01 -1.36562780e-01
6.66263938e-01 3.37450475e-01 2.12131292e-01 -6.80720210e-01
3.57760876e-01 -3.30768257e-01 1.18006550e-01 -1.94428787e-01
3.70783746e-01 -9.04058576e-01 6.52844846e-01 9.22688544e-01
-1.23902164e-01 4.50032026e-01 -1.58604741e-01 1.01238966e+00
3.13048184e-01 2.32332408e-01 6.26416624e-01 -7.66415522e-02
2.69145574e-02 5.59400141e-01 -7.39890575e-01 5.58974564e-01
1.16916263e+00 -1.95088059e-01 -5.93980610e-01 -8.47294807e-01
-5.33915639e-01 3.37228358e-01 6.94485128e-01 3.27781285e-03
4.31501985e-01 -8.13963115e-01 -7.70843804e-01 -1.29936516e-01
-3.61098409e-01 -4.44896579e-01 1.09049916e-01 1.21479416e+00
-5.93995333e-01 2.23230273e-01 -1.19424872e-01 -1.08455442e-01
-1.16507578e+00 3.31062764e-01 8.04945752e-02 -3.67872655e-01
-5.94805479e-01 6.25476003e-01 -1.52047202e-01 -1.69704720e-01
1.14021435e-01 1.38171092e-01 -1.07965611e-01 1.34415880e-01
4.37244922e-01 7.31235087e-01 -3.86493295e-01 -7.02771366e-01
-6.17644131e-01 5.83706558e-01 -2.44517643e-02 -9.16417763e-02
1.61520028e+00 -4.10093367e-01 -5.56495070e-01 2.01486461e-02
1.05678225e+00 3.15094531e-01 -8.33317280e-01 -4.80738096e-02
-1.40415579e-01 -3.34220022e-01 -2.33384714e-01 -2.19451636e-01
-1.46196651e+00 3.66830558e-01 -2.28281170e-01 1.09618878e+00
8.85157824e-01 1.72212914e-01 9.56528723e-01 3.69249731e-01
6.47830546e-01 -6.05716646e-01 -3.24973553e-01 2.99623162e-01
2.84249932e-01 -9.21783030e-01 2.31727853e-01 -6.82685912e-01
-1.89819619e-01 6.01241708e-01 1.04941182e-01 -2.34589502e-01
1.25819921e+00 4.93756026e-01 -3.06389511e-01 -5.96015036e-01
-7.72105634e-01 -4.55370694e-02 -5.33307731e-01 6.77167892e-01
1.47549972e-01 5.38037755e-02 -4.55424428e-01 -5.85690998e-02
1.31855160e-01 -2.93280005e-01 7.09357440e-01 7.69959927e-01
-5.73819757e-01 -1.01202476e+00 -2.53043413e-01 1.16056943e+00
-4.02364343e-01 -2.62279004e-01 -5.92001855e-01 8.85196805e-01
-2.70326823e-01 1.09412813e+00 2.57662773e-01 -2.30280876e-01
-4.12863791e-02 -6.68570220e-01 5.46363704e-02 -3.49975973e-01
-4.63066638e-01 -4.12403256e-01 1.35788515e-01 -3.32170218e-01
-2.30834663e-01 -5.86592615e-01 -1.13749266e+00 -1.34424913e+00
-2.21481338e-01 5.36688566e-01 3.88212740e-01 6.45816743e-01
5.31898856e-01 1.58394471e-01 7.85429597e-01 -6.69675648e-01
-4.00119908e-02 -5.09714007e-01 -9.72811043e-01 3.17841411e-01
3.95880342e-01 -5.72893143e-01 -8.31514537e-01 -4.60262924e-01]
|
[6.484703063964844, 5.001572608947754]
|
b5ffff13-14f1-4c8e-b521-90d8514c234a
|
deep-gaussian-mixture-ensembles
|
2306.07235
| null |
https://arxiv.org/abs/2306.07235v1
|
https://arxiv.org/pdf/2306.07235v1.pdf
|
Deep Gaussian Mixture Ensembles
|
This work introduces a novel probabilistic deep learning technique called deep Gaussian mixture ensembles (DGMEs), which enables accurate quantification of both epistemic and aleatoric uncertainty. By assuming the data generating process follows that of a Gaussian mixture, DGMEs are capable of approximating complex probability distributions, such as heavy-tailed or multimodal distributions. Our contributions include the derivation of an expectation-maximization (EM) algorithm used for learning the model parameters, which results in an upper-bound on the log-likelihood of training data over that of standard deep ensembles. Additionally, the proposed EM training procedure allows for learning of mixture weights, which is not commonly done in ensembles. Our experimental results demonstrate that DGMEs outperform state-of-the-art uncertainty quantifying deep learning models in handling complex predictive densities.
|
['Svitlana Vyetrenko', 'Elizabeth Fons', 'Niccolò Dalmasso', 'Yousef El-Laham']
|
2023-06-12
| null | null | null | null |
['probabilistic-deep-learning']
|
['computer-vision']
|
[-5.03465474e-01 1.35110244e-01 2.13471442e-01 -3.65825802e-01
-9.82158601e-01 -4.13615167e-01 9.08917010e-01 -6.21855818e-02
-5.15312016e-01 8.56376350e-01 -3.59674059e-02 -5.25186360e-01
-2.23100051e-01 -9.92128372e-01 -8.21776748e-01 -9.54065144e-01
-5.51682115e-02 9.11280751e-01 -1.67444438e-01 4.46723759e-01
-1.56512231e-01 4.44773316e-01 -1.46596825e+00 -1.13851000e-02
1.06789744e+00 1.13517475e+00 -2.77015120e-01 7.57881939e-01
-2.67187148e-01 9.45824564e-01 -7.82600880e-01 -8.55418742e-01
-2.49726638e-01 4.98233130e-03 -3.37472945e-01 -3.41771632e-01
1.68280423e-01 -5.78184962e-01 -2.60701060e-01 1.10961807e+00
4.15092647e-01 4.03935194e-01 1.75519264e+00 -1.28674221e+00
-4.52226520e-01 1.04476130e+00 -3.17027569e-01 5.16158715e-02
-2.56508052e-01 -1.20805427e-01 7.65786171e-01 -8.79310608e-01
-1.99346006e-01 1.36721647e+00 8.18592489e-01 3.43752354e-01
-1.29163778e+00 -7.58652091e-01 -1.98625535e-01 -1.03255004e-01
-1.71714652e+00 -3.43667686e-01 5.08428991e-01 -5.63514292e-01
9.69780445e-01 -1.72279552e-01 2.30435148e-01 1.36470783e+00
4.71328646e-01 9.51151609e-01 1.04314375e+00 -3.12805861e-01
7.17835724e-01 1.87427789e-01 -2.76227333e-02 2.56439894e-01
3.52023751e-01 2.40497395e-01 -2.70975828e-01 -3.75393957e-01
8.08376849e-01 -3.05276960e-01 1.91451028e-01 -3.45536202e-01
-5.99554121e-01 9.23295915e-01 -1.19444124e-01 1.90848634e-02
-4.35233533e-01 6.44816697e-01 1.58129364e-01 -3.29488069e-01
7.14621842e-01 -1.65446416e-01 -3.40250462e-01 -3.88746768e-01
-1.33791625e+00 4.37103361e-01 1.10824847e+00 8.08189929e-01
5.46627879e-01 3.02464545e-01 -2.05127597e-01 7.03483939e-01
1.00826991e+00 8.77965808e-01 -3.91947292e-02 -1.20348597e+00
2.00737670e-01 1.02059789e-01 5.47253430e-01 -4.34944600e-01
-2.96317190e-01 -7.20814884e-01 -9.57825661e-01 4.22124594e-01
4.99160588e-01 -5.42914450e-01 -9.49100256e-01 1.83060002e+00
1.12962045e-01 3.76263112e-01 2.50435889e-01 2.75497466e-01
6.68226779e-01 5.98312199e-01 4.87124175e-01 1.85376108e-01
1.06783986e+00 -3.48026991e-01 -7.71071494e-01 -7.15027228e-02
2.82298118e-01 -3.33261043e-01 5.04086435e-01 4.39643294e-01
-1.07792389e+00 -3.59825611e-01 -9.50813413e-01 3.55246544e-01
-3.14377457e-01 4.58551524e-03 7.22395480e-01 1.21318901e+00
-9.35267389e-01 7.99587250e-01 -1.22393870e+00 4.06331897e-01
6.07230783e-01 1.76274449e-01 -1.40485615e-01 9.51973125e-02
-1.25346303e+00 1.11802924e+00 8.22031915e-01 5.39544933e-02
-1.32918715e+00 -7.72518277e-01 -1.14194691e+00 4.08905357e-01
-2.93925893e-03 -7.02778220e-01 1.54489839e+00 -4.57752287e-01
-1.79750431e+00 5.05120397e-01 3.56105901e-02 -5.48337758e-01
5.50445795e-01 -5.17930388e-01 -6.23580635e-01 -1.34494737e-01
-4.84126866e-01 4.24125344e-01 8.37319314e-01 -1.37496495e+00
-3.53507370e-01 -4.49511670e-02 -1.83591142e-01 -6.44189343e-02
5.62544242e-02 -1.39331132e-01 -1.68526396e-01 -4.80706513e-01
-3.10038716e-01 -6.25728130e-01 -1.65674612e-01 -4.19824779e-01
-3.84923369e-01 -2.93229818e-01 2.63232321e-01 -6.21965230e-01
1.34347403e+00 -1.95376432e+00 7.23077804e-02 3.58545721e-01
1.35016218e-01 1.24884158e-01 2.80596316e-01 4.34815079e-01
2.06181735e-01 1.12715006e-01 -5.67251205e-01 -1.09007180e+00
7.59805679e-01 3.63875359e-01 -4.11121935e-01 3.87643874e-01
2.51913875e-01 7.49021947e-01 -7.66409159e-01 -5.58925509e-01
3.90737206e-01 7.65239775e-01 -3.15430522e-01 1.85692877e-01
-4.28720117e-01 2.70298839e-01 -1.16397459e-02 2.70504177e-01
1.06664419e+00 -2.52231836e-01 7.96147361e-02 8.75585675e-02
1.26674294e-01 3.28212708e-01 -1.33058357e+00 1.42607713e+00
-5.91383636e-01 3.14061701e-01 1.46732479e-02 -5.47804892e-01
6.68093145e-01 3.44172984e-01 5.58036417e-02 2.32972756e-01
5.85022330e-01 1.49269059e-01 -1.11010686e-01 2.34965116e-01
4.79869694e-01 -4.75835472e-01 -1.91157326e-01 6.26621544e-01
7.65724659e-01 -3.35422844e-01 1.62548840e-01 1.92315191e-01
5.05536735e-01 4.46764052e-01 -2.36364808e-02 -3.59918982e-01
2.48560458e-01 -6.61447883e-01 2.62790501e-01 1.10856915e+00
1.54914223e-02 3.20834666e-01 5.82730174e-01 2.08790690e-01
-9.40883398e-01 -1.77840102e+00 -6.54688597e-01 1.02738583e+00
-3.62660021e-01 -3.46904039e-01 -9.23766911e-01 -4.27868724e-01
8.59719589e-02 1.44270289e+00 -5.73759437e-01 -2.35791892e-01
1.93074048e-01 -1.18840301e+00 8.77623260e-01 8.19541276e-01
3.03810686e-01 -6.05139434e-01 -9.11803171e-02 2.05951214e-01
7.35301524e-02 -9.52205122e-01 7.31179565e-02 3.40392798e-01
-8.28078330e-01 -6.63404942e-01 -6.22034669e-01 2.15099454e-02
1.24641068e-01 -8.17583382e-01 1.30409825e+00 -7.20369458e-01
3.27335566e-01 6.16139293e-01 2.76532117e-02 -7.80455172e-01
-8.37790012e-01 -2.58338064e-01 3.22734296e-01 -1.99091777e-01
6.56782031e-01 -8.34327579e-01 -3.67872417e-01 -2.46226881e-02
-1.06412172e+00 -1.91223666e-01 6.28910780e-01 5.87011635e-01
4.65313643e-01 4.25732285e-01 6.01845145e-01 -5.38521886e-01
8.75417173e-01 -7.51207292e-01 -6.95429862e-01 2.90083736e-01
-5.61513782e-01 4.70753163e-01 2.35834047e-01 -4.81451362e-01
-1.49958622e+00 -4.08642054e-01 -4.86308962e-01 -4.53820735e-01
-4.57556963e-01 6.87314153e-01 -2.20016941e-01 4.13794994e-01
5.24022102e-01 2.22374976e-01 -1.36872798e-01 -4.72970217e-01
6.26262724e-01 7.77394533e-01 8.59819174e-01 -1.01415122e+00
3.58516157e-01 3.09679776e-01 1.23894438e-01 -4.40358698e-01
-9.17700827e-01 -9.94395018e-02 -6.53820336e-01 -1.63432986e-01
7.79145956e-01 -1.05989349e+00 -6.72563553e-01 1.00831330e+00
-1.06548655e+00 -3.10610890e-01 -3.41296852e-01 6.44159675e-01
-7.60927498e-01 4.96185154e-01 -8.29029977e-01 -1.55395067e+00
-3.85013402e-01 -1.01642656e+00 1.06667554e+00 3.47129524e-01
-8.60782787e-02 -1.53487098e+00 3.31219286e-01 2.46604364e-02
5.24859011e-01 1.27072394e-01 8.07947576e-01 -1.02100039e+00
-7.66179562e-02 -3.16443950e-01 -1.82557711e-03 8.05556357e-01
-3.74861330e-01 3.54493529e-01 -1.41315019e+00 1.12560160e-01
-8.81157070e-02 -3.15497905e-01 1.06329417e+00 7.03598320e-01
1.03718412e+00 1.90676283e-02 -2.92060524e-01 4.80616122e-01
1.05589259e+00 -1.96364790e-01 8.83448899e-01 -1.62716269e-01
3.06399614e-01 2.76147395e-01 5.67461215e-02 7.63407290e-01
5.20596802e-01 2.62573827e-02 5.71250916e-01 5.44987619e-01
4.07629848e-01 -3.76643956e-01 4.34704781e-01 5.87961972e-01
-1.19672090e-01 -4.16238219e-01 -1.01366735e+00 3.28963459e-01
-1.83769965e+00 -1.05217373e+00 -1.04096666e-01 2.27314377e+00
1.09852266e+00 2.36787125e-01 5.00439554e-02 2.10587657e-03
5.83095908e-01 -7.11570829e-02 -4.77411777e-01 -2.59891361e-01
-1.28793865e-01 5.70130706e-01 4.17621940e-01 5.89568138e-01
-1.51134312e+00 7.37191975e-01 7.58228159e+00 1.30673337e+00
-4.29564327e-01 2.24289030e-01 4.76014435e-01 6.50590137e-02
-5.36115527e-01 -2.66397715e-01 -7.68965125e-01 6.46328628e-01
1.59187531e+00 -1.07856400e-01 9.40096974e-02 9.00300801e-01
-3.25832754e-01 -3.52406114e-01 -1.14753675e+00 9.22034085e-01
-2.87482649e-01 -1.05450380e+00 -1.24764174e-01 1.70053601e-01
7.99316823e-01 1.66807726e-01 4.08929080e-01 7.51732171e-01
1.26006746e+00 -1.40266633e+00 8.76479745e-01 1.06474483e+00
5.71957350e-01 -1.23683512e+00 1.11781490e+00 6.76820040e-01
-5.48966825e-01 1.77905917e-01 -5.20549834e-01 2.31714129e-01
3.20423484e-01 1.15901744e+00 -6.53544545e-01 5.38943887e-01
5.79054594e-01 1.47772819e-01 -1.22543141e-01 1.00063622e+00
-2.87281513e-01 9.16692436e-01 -7.48892128e-01 1.08578257e-01
2.26107482e-02 -3.65059465e-01 5.11446059e-01 1.51261151e+00
6.20419443e-01 -3.65686625e-01 -4.14562076e-01 1.50668943e+00
-1.21531248e-01 -4.71983373e-01 -2.72613525e-01 -4.30923939e-01
7.33540237e-01 1.13503098e+00 -2.25777015e-01 -2.65548974e-01
-1.11635111e-01 5.79731762e-01 4.33857888e-01 4.97221649e-01
-1.01548052e+00 -2.04200801e-02 7.58059561e-01 -4.32889819e-01
3.86468261e-01 -3.87500197e-01 -2.45047599e-01 -9.79299009e-01
-5.03583133e-01 -3.94756943e-01 3.41995090e-01 -6.79486692e-01
-1.80597854e+00 5.04683077e-01 5.97612619e-01 -8.29334140e-01
-6.31379604e-01 -9.33827758e-01 -8.43907535e-01 1.29339540e+00
-1.14283705e+00 -1.16555607e+00 1.20724171e-01 3.59219432e-01
-3.14813435e-01 -2.52764016e-01 1.01305139e+00 4.15175594e-02
-5.49897730e-01 5.56644320e-01 6.24071121e-01 -2.41583195e-02
4.50952470e-01 -1.62574267e+00 1.97837606e-01 6.71910286e-01
1.53515898e-02 6.62386537e-01 1.05878794e+00 -4.72699910e-01
-8.66449475e-01 -8.91714931e-01 2.03448907e-01 -7.83808529e-01
8.79613876e-01 -2.33018205e-01 -8.81208777e-01 7.81438053e-01
2.17635825e-01 -2.29630798e-01 1.22985899e+00 3.15572232e-01
-5.29747665e-01 4.01484758e-01 -1.25927889e+00 3.27523530e-01
3.72990102e-01 -5.82397282e-01 -6.91783071e-01 -5.28695844e-02
3.63186717e-01 -4.12928700e-01 -1.23324656e+00 4.83281434e-01
5.93859255e-01 -1.15231466e+00 8.43702495e-01 -6.66465759e-01
4.87911761e-01 -2.70358741e-01 -4.04729903e-01 -1.61485648e+00
-2.17115372e-01 -3.27762425e-01 -1.16549850e+00 1.34729350e+00
4.66107816e-01 -5.48551023e-01 4.25003350e-01 9.40104246e-01
-1.29935101e-01 -5.51627398e-01 -1.05520177e+00 -6.22415900e-01
7.74903655e-01 -1.19211376e+00 8.00554633e-01 4.35232610e-01
-2.19852522e-01 -1.84645101e-01 -2.72317141e-01 3.92968625e-01
1.05753815e+00 -4.10637736e-01 4.34661478e-01 -1.50134659e+00
-4.27390605e-01 -6.15465939e-01 -3.80279213e-01 -9.21432495e-01
6.29195869e-01 -7.44547963e-01 5.23145348e-02 -1.57676423e+00
1.50727361e-01 -2.57220000e-01 -5.15254319e-01 1.37581918e-02
-1.79870769e-01 -7.73904324e-02 -2.77840823e-01 -1.40459239e-01
-8.02205145e-01 9.79843497e-01 6.08763337e-01 6.55076606e-03
3.48455787e-01 2.96034813e-01 -5.33418477e-01 9.05324340e-01
7.03383088e-01 -3.93013626e-01 -3.71455580e-01 -9.07886475e-02
4.98358816e-01 -1.15628831e-01 4.32987839e-01 -1.09371638e+00
1.23918533e-01 7.85889179e-02 6.23431027e-01 -8.88896823e-01
5.02694905e-01 -6.93149805e-01 3.69097203e-01 -1.73463911e-01
-5.43693118e-02 -4.70232069e-01 6.08076811e-01 8.28383625e-01
-1.38819709e-01 -5.94719052e-01 7.15729713e-01 9.29120705e-02
-3.21352690e-01 9.83781219e-02 -6.65988803e-01 4.14031595e-02
7.39826083e-01 2.22331837e-01 -3.81816298e-01 -7.11655736e-01
-9.08866763e-01 2.00167209e-01 1.98366255e-01 -1.27962530e-01
3.88188183e-01 -1.37841952e+00 -7.91721702e-01 -2.90117443e-01
-1.57972500e-01 3.45097840e-01 5.78005254e-01 8.40677977e-01
-4.08679515e-01 1.75670326e-01 1.09408088e-01 -5.59227705e-01
-4.14961666e-01 3.44088674e-01 7.55699217e-01 -4.89862770e-01
-1.50025785e-01 9.73373592e-01 1.47021130e-01 -7.66773582e-01
2.90435493e-01 -1.44078508e-01 -3.45123969e-02 -1.46402761e-01
6.52121127e-01 5.23830295e-01 -6.16486594e-02 -3.93438399e-01
-1.25135079e-01 -2.06619799e-01 1.26831770e-01 -4.87251461e-01
1.18770647e+00 -6.68941587e-02 -2.53756404e-01 7.75273025e-01
6.74447179e-01 -1.25167847e-01 -1.56273735e+00 -1.11680306e-01
-5.77001870e-02 8.33223537e-02 3.25423628e-01 -1.05276585e+00
-5.11883199e-01 1.16816425e+00 5.10645747e-01 2.18691435e-02
7.49200225e-01 2.28652954e-02 3.17617476e-01 2.39993110e-01
3.75629872e-01 -9.93875146e-01 -3.72551650e-01 7.97422171e-01
6.05763137e-01 -1.14687574e+00 -8.70671347e-02 1.51572749e-01
-7.47595787e-01 9.21379089e-01 4.46266174e-01 -7.18592899e-03
1.13949227e+00 8.26215506e-01 -2.50400543e-01 3.92697677e-02
-6.26062334e-01 -1.04710326e-01 7.44421244e-01 8.51979673e-01
5.12730896e-01 3.49537015e-01 3.28442305e-01 1.43052840e+00
-2.83975571e-01 -1.65226713e-01 2.52082497e-01 4.82685179e-01
-4.82390672e-01 -8.48235786e-01 -4.50518459e-01 5.25166929e-01
-6.94569528e-01 -2.79544383e-01 4.78511900e-02 5.47560394e-01
-1.61307640e-02 8.97650778e-01 4.78853554e-01 -2.31884286e-01
-2.03572765e-01 7.15472817e-01 7.13293314e-01 -3.35928708e-01
-1.46606499e-02 -1.58917420e-02 1.11687198e-01 -1.25899520e-02
-3.25118303e-01 -5.22166073e-01 -1.23758781e+00 -4.89239424e-01
-4.20562983e-01 3.03757101e-01 8.43393922e-01 1.22488141e+00
2.59632170e-01 6.25139654e-01 -1.36272728e-01 -1.18207240e+00
-1.07536530e+00 -1.61168444e+00 -1.03063762e+00 1.05325431e-01
4.30090018e-02 -1.03115821e+00 -7.49162853e-01 -2.65176326e-01]
|
[7.2506632804870605, 3.8083291053771973]
|
fb700380-58c8-48ef-a29b-5a7044a188eb
|
comment-on-water-sources-and-kidney-function
|
2201.00399
| null |
https://arxiv.org/abs/2201.00399v3
|
https://arxiv.org/pdf/2201.00399v3.pdf
|
Comment on "Water sources and kidney function: investigating chronic kidney disease of unknown etiology in a prospective study", by P. Vlahos et al
|
Vlahos et al., Ref. 1, NPJ Clean water. 4, 50 (2021) have reported the presence of pesticide contamination above safe levels in a "single time-point analysis" of well water in a region in Sri Lanka where chronic kidney disease of unknown etiology (CKDu) is endemic. They conclude "that agrochemical use in paddy and other agricultural practices ... of the Green Revolution in Sri Lanka may now be contributing to ill health, rapid progression of disease, and mortality". The authors also propose "reducing ... agrochemical contaminants in Sri Lanka and other tropical countries to reduce ... CKDu". These conclusions, based on what they call a "single time-point analysis", tantamount to an identification of the etiology of CKDu are unsupported by the evidence presented by Vlahos et al. They do not satisfy, say, even the simplest of Bradford-Hill criteria for causation. In particular, (i) similar but non-persistent pesticide excesses have been detected sporadically in most parts of the country including where there is no CKDu; (ii) the pesticides reported in (1) cause both hepatotoxicity and nephrotoxicity; the latter with glomerular damage while CKDu is associated with tubulo-interstitial damage where no hepatotoxic symptoms have been reported; (iii) the pesticides detected have short half-lives and are used over short periods during farming; so the one time-point analysis is inadequate and misleading; (iv) farming communities that use pesticides in the same way but remain essentially without CKDu are found to exist adjacent to communities with CKDu; (v) the CKDu prevalence seems to correlate with local geomorphology but without correlation to agriculture which is practiced in most parts of the country. .
|
['M. W. C. Dharma-wardana']
|
2021-12-28
| null | null | null | null |
['kidney-function']
|
['medical']
|
[ 1.40761718e-01 -2.45268136e-01 3.38218398e-02 3.75812143e-01
-2.56690621e-01 -6.68622315e-01 5.08867562e-01 7.57388234e-01
-1.38749048e-01 9.18671668e-01 3.57463211e-02 -1.03088152e+00
-6.29474282e-01 -1.00753164e+00 -8.88824522e-01 -9.37925875e-01
-8.29455107e-02 2.38632713e-03 7.46984631e-02 -3.11031014e-01
1.47781193e-01 7.06225812e-01 -1.08923650e+00 -6.82942495e-02
1.02283394e+00 -1.68432593e-01 4.34504658e-01 4.50776547e-01
1.38038754e-01 4.07632172e-01 -2.40296498e-01 2.47679297e-02
2.56834745e-01 -5.27630687e-01 -5.47025204e-01 -2.09842891e-01
2.91795641e-01 -3.64033818e-01 -1.73793331e-01 1.09479368e+00
3.38885814e-01 -5.16446769e-01 9.86545146e-01 -7.32672095e-01
-6.55672133e-01 5.72954595e-01 -8.95118415e-01 -4.09972109e-03
2.81744480e-01 3.74102980e-01 4.97169644e-02 -5.21365166e-01
4.45029527e-01 1.28107548e+00 8.17515552e-01 -5.68574574e-03
-1.58262861e+00 -1.03563023e+00 2.86460686e-02 -2.66729385e-01
-1.43059695e+00 -2.53010184e-01 -2.12122381e-01 -6.33401275e-01
1.02964199e+00 2.49921203e-01 7.23752499e-01 4.03733611e-01
1.01002872e+00 -3.04658085e-01 1.13782716e+00 -3.49964470e-01
2.59490997e-01 -3.33335966e-01 -2.30368495e-01 2.31298357e-01
1.26046097e+00 2.67058134e-01 -7.62857720e-02 -5.19470036e-01
6.45199716e-01 1.29832461e-01 -2.19933972e-01 -5.29084206e-02
-8.80576015e-01 6.20956004e-01 1.95768535e-01 4.66598719e-01
-7.13817477e-01 4.94473100e-01 4.58909243e-01 4.39601898e-01
4.47453350e-01 1.84410900e-01 -6.93634093e-01 3.68214518e-01
-4.79395449e-01 4.19946402e-01 6.85463846e-01 7.24837899e-01
8.07619393e-01 -9.24072042e-02 3.91022950e-01 6.45663261e-01
7.26521373e-01 1.32638013e+00 -3.59804630e-01 -7.66228259e-01
4.27296191e-01 3.94734263e-01 4.78085995e-01 -7.12221801e-01
-1.71613634e-01 8.98334458e-02 -3.28483343e-01 4.56869900e-01
6.44497454e-01 -4.94557381e-01 -1.13868403e+00 1.62518358e+00
2.57027261e-02 -3.90875518e-01 3.10507804e-01 2.92201310e-01
2.51282960e-01 8.96064341e-01 9.83291149e-01 -4.59032565e-01
1.46837604e+00 3.40230495e-01 -6.85195744e-01 6.29887432e-02
4.81965512e-01 -8.59678030e-01 6.95156276e-01 3.07391047e-01
-7.98074126e-01 2.47034773e-01 -1.07122684e+00 7.20259428e-01
-6.18909657e-01 -3.67907584e-01 7.45150805e-01 1.02991307e+00
-8.41981113e-01 7.41389990e-01 -1.06022346e+00 -1.17376506e+00
1.57814145e-01 2.82789022e-01 -4.77023154e-01 -3.87755603e-01
-1.04106092e+00 1.21280921e+00 -1.59430355e-01 3.27081323e-01
-7.89172709e-01 -7.87269950e-01 -7.14181066e-01 3.07266358e-02
1.12252362e-01 -2.63809234e-01 7.11386979e-01 -3.11786998e-02
-7.52438009e-01 3.31181228e-01 -1.91538274e-01 4.05594818e-02
5.42324781e-01 -4.52836066e-01 -4.52770889e-01 1.44848257e-01
7.65389800e-01 8.80106986e-02 -3.00884724e-01 -1.25079060e+00
-5.86346149e-01 -8.61761212e-01 -2.42619142e-01 1.92179561e-01
1.03807271e-01 5.19218981e-01 4.96382684e-01 -2.98417211e-01
5.15728533e-01 -7.25767136e-01 -6.91079497e-01 -4.56910163e-01
-1.55128911e-01 -8.68930146e-02 7.77155876e-01 -5.38448632e-01
9.83481348e-01 -1.70471156e+00 -7.27863014e-01 2.69056648e-01
-2.46412173e-01 1.20646648e-01 7.54264891e-02 1.08964622e+00
-4.37248573e-02 5.95426977e-01 -5.97850919e-01 9.10000443e-01
9.99330543e-03 2.52633095e-01 -7.02230772e-03 1.07578063e+00
5.02534151e-01 3.72407407e-01 -1.27278304e+00 -7.17193931e-02
3.63202274e-01 3.73140246e-01 8.95951465e-02 -5.04061162e-01
9.84607264e-02 1.80639490e-01 -4.15546894e-01 9.42461431e-01
1.21890068e+00 3.72309476e-01 6.95044518e-01 5.51430136e-02
-9.41551983e-01 -5.41596524e-02 -1.14691806e+00 8.58733773e-01
9.04066488e-02 1.38620481e-01 1.35564804e-01 -4.60013419e-01
6.12976670e-01 8.03622782e-01 2.37003997e-01 -7.07774103e-01
-4.63471323e-01 6.62751555e-01 1.99879810e-01 -6.98548377e-01
1.41742915e-01 -4.77119833e-01 1.15544423e-01 2.11547032e-01
-3.70813608e-01 -7.51483813e-02 3.26185107e-01 -4.93224002e-02
1.27391231e+00 4.36873287e-01 3.08963776e-01 -1.08194005e+00
2.99543738e-01 3.12406033e-01 6.87453628e-01 4.83561486e-01
-5.84553406e-02 -3.97007465e-02 7.32818246e-01 -3.34127573e-03
-1.04110682e+00 -1.26918757e+00 -8.11240971e-01 7.85115063e-01
2.27208391e-01 1.72151387e-01 -3.12073112e-01 1.42866239e-01
3.73764247e-01 5.66418171e-01 -5.17617524e-01 3.39495718e-01
-3.57370138e-01 -1.36504853e+00 8.24289799e-01 1.80528998e-01
4.83961284e-01 -7.74163187e-01 -5.69864511e-01 7.68320858e-01
4.09524381e-01 -3.77900362e-01 2.96656638e-01 7.54253685e-01
-1.19968307e+00 -1.29434955e+00 -9.98789012e-01 -4.33705777e-01
7.10158467e-01 3.47500771e-01 6.38724029e-01 2.09976472e-02
-2.19556883e-01 3.41650955e-02 -8.97099972e-02 -8.18215907e-01
-6.17623866e-01 -6.20539188e-01 5.65781854e-02 -9.44512486e-01
8.45469296e-01 -6.76035464e-01 -5.66390932e-01 2.62030542e-01
-8.76885772e-01 -8.54219198e-01 9.28609788e-01 7.19919726e-02
2.93885767e-01 4.41948399e-02 1.26963627e+00 -1.07619441e+00
2.54403532e-01 -1.10926235e+00 -5.33870757e-01 3.44964415e-01
-7.98899770e-01 -4.57702637e-01 4.73375525e-03 -2.59468168e-01
-9.11282182e-01 -7.68556967e-02 1.84835613e-01 4.31233108e-01
-7.48464882e-01 7.47416615e-01 -3.59323025e-01 1.53840393e-01
8.26087475e-01 -1.67459980e-01 -8.92142504e-02 -3.85445565e-01
-2.69278407e-01 4.14663792e-01 4.30798143e-01 -6.35667503e-01
6.45312488e-01 2.85394520e-01 1.39126286e-01 -1.33687139e+00
1.32728264e-01 -4.83351976e-01 -2.91646987e-01 6.00812025e-02
1.00298631e+00 -1.21963561e+00 -5.73002756e-01 5.51719487e-01
-1.00651836e+00 -5.72398424e-01 3.06237310e-01 1.05168629e+00
-5.04580177e-02 4.34528857e-01 -7.14468598e-01 -1.37379110e+00
-1.57969706e-02 -7.72364616e-01 3.75895709e-01 1.57435372e-01
-2.52308756e-01 -8.13638926e-01 3.01556736e-01 -2.53480703e-01
3.38855535e-01 8.52030933e-01 1.32313013e+00 -1.70183882e-01
-2.67267108e-01 7.93447942e-02 -4.28994894e-01 -9.56640765e-03
7.11460471e-01 2.57882118e-01 -7.97073901e-01 -2.09221572e-01
-1.52636766e-01 2.26026475e-01 4.20088351e-01 6.05651855e-01
-1.37703791e-01 -2.07991153e-01 -4.68560964e-01 -5.10420948e-02
1.99032617e+00 1.02636874e+00 8.56909335e-01 3.38933110e-01
2.32790187e-01 8.86169910e-01 5.22092581e-01 -2.04410329e-02
-5.79891428e-02 -1.05696030e-01 3.66279900e-01 -1.71749160e-01
9.40237120e-02 -5.88473585e-03 7.19568133e-01 1.02524191e-01
-3.00511062e-01 -5.44213615e-02 -1.13648117e+00 1.18213964e+00
-1.59423828e+00 -1.12515926e+00 -1.18718708e+00 2.29289031e+00
6.09361053e-01 -2.46529326e-01 -2.13899881e-01 -1.66271627e-01
8.38373423e-01 -4.67802405e-01 -2.97083050e-01 -6.78145766e-01
1.26917781e-02 5.54423749e-01 1.35340297e+00 3.82024646e-01
-7.61845827e-01 6.19267285e-01 6.94173145e+00 7.17006475e-02
-8.66022050e-01 -1.46695495e-01 1.56805292e-01 5.43642938e-01
-3.95307094e-01 5.37079871e-01 -4.55979437e-01 2.95784384e-01
9.19602871e-01 -2.34628335e-01 -2.52840370e-01 -1.99622232e-02
7.52578855e-01 -8.20862949e-01 -8.32339168e-01 -1.54821485e-01
-4.97696549e-01 -7.49965966e-01 -1.70692831e-01 6.06656969e-01
6.47891164e-01 1.70392752e-01 -4.03621078e-01 -1.94912627e-02
6.73130035e-01 -1.08891058e+00 4.42713737e-01 2.20818579e-01
8.65450203e-01 -1.00232506e+00 8.63966465e-01 1.66157186e-01
-1.11574817e+00 -3.25196260e-03 -6.01704299e-01 -2.63665110e-01
1.64703548e-01 7.96348453e-01 -5.63952208e-01 7.09169865e-01
8.49618912e-01 1.02058753e-01 -3.30597818e-01 9.42193270e-01
-4.93492365e-01 8.45838428e-01 -5.66473603e-01 4.68452796e-02
4.41882700e-01 -5.36817193e-01 3.81385952e-01 1.06317127e+00
5.58642030e-01 4.63975847e-01 -3.69949222e-01 9.18242693e-01
5.80887079e-01 2.79261112e-01 -1.13217402e+00 -1.50814816e-01
6.48689926e-01 4.86480474e-01 -9.07645941e-01 8.08037743e-02
-5.66435754e-01 4.09007519e-01 -7.87887394e-01 5.55161059e-01
-1.68556213e-01 -7.83115327e-01 6.69268608e-01 3.57428044e-01
-1.05493469e-02 5.52361384e-02 -2.95948595e-01 -5.33001781e-01
-3.46505374e-01 -5.04609823e-01 3.02549638e-02 -2.10191339e-01
-1.10840809e+00 -3.15762162e-01 4.85100180e-01 -7.12013781e-01
1.59404516e-01 -3.46399367e-01 -8.39789927e-01 1.50020528e+00
-1.26543188e+00 -9.79649484e-01 1.82976216e-01 -6.39056787e-03
4.11224477e-02 2.78426796e-01 9.29314733e-01 2.87867218e-01
-2.83699483e-01 -1.31774787e-03 5.16728222e-01 -3.35744143e-01
6.65392220e-01 -1.13670552e+00 2.16672733e-01 7.69595087e-01
-9.44446981e-01 1.04998517e+00 8.92313421e-01 -1.59773970e+00
-1.23693645e+00 -1.08082020e+00 1.14569509e+00 -2.20971361e-01
7.02543259e-01 -9.04153660e-03 -8.41643691e-01 4.61749375e-01
4.15780544e-01 -7.66440094e-01 6.54663622e-01 1.54211491e-01
3.37908305e-02 4.47356217e-02 -1.62271571e+00 4.71301675e-01
3.96143258e-01 -2.78315246e-01 -4.94510293e-01 4.40297663e-01
4.25255001e-01 2.91470766e-01 -1.09175003e+00 3.45507354e-01
9.35739577e-01 -3.66848201e-01 6.12537742e-01 -5.64109206e-01
4.78265107e-01 -6.11428618e-01 -3.14030081e-01 -8.65084529e-01
-4.15180624e-01 -2.50225067e-01 8.18343937e-01 1.39417195e+00
6.12621665e-01 -8.72266889e-01 4.77838278e-01 6.18396580e-01
-2.70345854e-03 -1.59373507e-01 -5.00010669e-01 -8.96774650e-01
6.98299289e-01 1.95350666e-02 5.01784444e-01 1.23907006e+00
9.15500820e-02 5.10563292e-02 6.41677454e-02 7.35286593e-01
1.00143361e+00 -4.82811540e-01 3.33347887e-01 -1.31541896e+00
2.03635246e-01 -1.68306530e-02 -1.91954792e-01 1.61005557e-01
-5.14757335e-01 -3.34234238e-01 1.15872942e-01 -1.91758919e+00
3.69954884e-01 -5.58506191e-01 -9.40901935e-02 6.66731000e-01
-5.47655709e-02 -1.72094882e-01 -1.56673670e-01 8.64122733e-02
6.74075007e-01 -2.64832824e-01 7.32214808e-01 -8.61423835e-02
-5.58991969e-01 -1.38197035e-01 -1.05483949e+00 5.11344790e-01
7.42550910e-01 -6.32382929e-01 -5.05505264e-01 -4.02271986e-01
5.25245667e-01 3.16367924e-01 3.95304710e-01 -7.52687216e-01
-2.83143282e-01 -8.87031019e-01 5.94163179e-01 -6.08283341e-01
-5.63973010e-01 -9.79599118e-01 9.53325629e-01 1.40870905e+00
1.97265252e-01 -7.58872926e-02 4.79114801e-01 6.37507439e-01
5.31047404e-01 -4.53174800e-01 7.44163632e-01 -3.63988876e-01
-2.27245256e-01 -2.51938514e-02 -1.21709096e+00 -6.10255539e-01
9.68411863e-01 -3.30968022e-01 -6.10552371e-01 9.88650844e-02
-6.58342004e-01 5.05275011e-01 8.49181950e-01 -1.94785103e-01
-7.81135494e-03 -1.16697276e+00 -9.53253508e-01 -5.40617943e-01
4.58129775e-03 -1.06866345e-01 1.48635283e-01 9.40213025e-01
-1.23008537e+00 4.62205827e-01 -4.24200088e-01 -4.21186268e-01
-7.30082572e-01 3.91242504e-01 1.74152240e-01 1.66260034e-01
-6.76315904e-01 1.56434163e-01 4.38910961e-01 -2.54505157e-01
-1.88340515e-01 -4.18295890e-01 -8.19199905e-02 2.71689892e-01
5.24840057e-01 6.32112741e-01 6.53303117e-02 -3.53304267e-01
-5.45472801e-01 1.93635806e-01 -1.03363413e-02 4.75471541e-02
1.57765746e+00 -8.65407810e-02 -5.29886842e-01 5.36175847e-01
6.94915354e-01 2.19026387e-01 -8.19468677e-01 3.76654148e-01
2.76217610e-01 -2.59684652e-01 -4.26065296e-01 -1.08656490e+00
-3.92721504e-01 5.98545551e-01 9.92616713e-01 -4.66471240e-02
8.23337615e-01 -1.50099695e-01 3.31650913e-01 5.99859953e-01
3.28144699e-01 -9.78825092e-01 -8.04043174e-01 2.81054854e-01
8.09422493e-01 -4.02721077e-01 3.70892227e-01 -4.44325417e-01
-1.44650051e-02 8.62405121e-01 2.06294611e-01 -1.59773171e-01
3.94406796e-01 2.98630804e-01 1.33587778e-01 -5.35256684e-01
-5.87328255e-01 -7.16512948e-02 -9.08196032e-01 8.87435436e-01
8.32178831e-01 2.78932869e-01 -1.34103799e+00 1.58181176e-01
4.99517471e-01 9.87800956e-02 1.01308024e+00 1.28333271e+00
-7.42108464e-01 -1.12444925e+00 -8.32765639e-01 7.51016557e-01
-8.68726611e-01 -4.11034822e-02 -3.63793880e-01 1.14464593e+00
4.32432771e-01 1.16006958e+00 -2.29577571e-01 3.09507251e-01
7.30474174e-01 2.78235823e-01 2.94570953e-01 -6.54881001e-01
-7.75670111e-01 5.45197904e-01 2.15581805e-01 -7.27087930e-02
-6.37441695e-01 -9.44762707e-01 -1.24252486e+00 -5.69001794e-01
-5.00372231e-01 3.85909617e-01 8.62558067e-01 7.83826649e-01
-3.76357287e-01 8.53243023e-02 4.71075177e-01 -3.97671193e-01
-5.72781622e-01 -1.16587055e+00 -1.34037066e+00 -2.14542702e-01
1.40596822e-01 -4.99957055e-01 -3.37281555e-01 8.75453185e-03]
|
[5.869502544403076, 4.130763530731201]
|
bf56e2ce-636f-420c-91e3-1e3ba000e0c3
|
multi-view-graph-structure-learning-using
|
2204.05258
| null |
https://arxiv.org/abs/2204.05258v1
|
https://arxiv.org/pdf/2204.05258v1.pdf
|
Multi-view graph structure learning using subspace merging on Grassmann manifold
|
Many successful learning algorithms have been recently developed to represent graph-structured data. For example, Graph Neural Networks (GNNs) have achieved considerable successes in various tasks such as node classification, graph classification, and link prediction. However, these methods are highly dependent on the quality of the input graph structure. One used approach to alleviate this problem is to learn the graph structure instead of relying on a manually designed graph. In this paper, we introduce a new graph structure learning approach using multi-view learning, named MV-GSL (Multi-View Graph Structure Learning), in which we aggregate different graph structure learning methods using subspace merging on Grassmann manifold to improve the quality of the learned graph structures. Extensive experiments are performed to evaluate the effectiveness of the proposed method on two benchmark datasets, Cora and Citeseer. Our experiments show that the proposed method has promising performance compared to single and other combined graph structure learning methods.
|
['Alireza Bosaghzadeh', 'Hossein Amirkhani', 'Razieh Ghiasi']
|
2022-04-11
| null | null | null | null |
['multi-view-learning', 'graph-structure-learning']
|
['computer-vision', 'graphs']
|
[-4.88346070e-02 1.83404703e-02 -2.41101116e-01 -1.65680483e-01
-2.99361795e-01 -4.93234277e-01 5.99356413e-01 4.64670271e-01
6.00521937e-02 5.40390790e-01 1.14436433e-01 -2.35521331e-01
-3.47458661e-01 -1.02133524e+00 -5.14270484e-01 -6.63489401e-01
-7.55310282e-02 4.25992280e-01 2.63975233e-01 -1.78538114e-01
2.07987711e-01 3.11776221e-01 -1.05746400e+00 -8.98058787e-02
8.71546626e-01 4.70306158e-01 -9.24093947e-02 3.93840432e-01
-2.01545894e-01 7.94707417e-01 -9.00290459e-02 -3.79715413e-01
2.90303588e-01 -3.78951699e-01 -6.23385668e-01 3.53036851e-01
3.00823838e-01 2.49935448e-01 -4.29803759e-01 1.21301711e+00
3.29684585e-01 6.81773666e-03 6.47690296e-01 -1.25950062e+00
-3.98511589e-01 5.94942868e-01 -9.18232620e-01 1.23355752e-02
4.65818375e-01 -3.48814368e-01 1.36553955e+00 -6.81551814e-01
7.11568952e-01 1.35696709e+00 5.23677230e-01 -5.86181227e-03
-1.11148846e+00 -5.25672734e-01 2.65884370e-01 5.24294078e-01
-1.38432479e+00 -6.29428774e-02 1.28061891e+00 -4.56560493e-01
4.57969248e-01 -7.29419589e-02 6.80671692e-01 5.87303698e-01
3.12025666e-01 5.82503796e-01 1.30176306e+00 -3.47022772e-01
8.00093189e-02 -1.23515381e-02 4.01062459e-01 1.21851897e+00
4.11105782e-01 -1.82516471e-01 -2.35666096e-01 -2.52414554e-01
5.60408711e-01 -1.23323016e-01 -3.33339006e-01 -1.04963267e+00
-1.08501542e+00 8.42320383e-01 7.15507746e-01 2.69223273e-01
-2.10586056e-01 -1.37271211e-01 4.48533922e-01 3.25442731e-01
5.29626727e-01 1.74176544e-02 6.80807680e-02 4.07982290e-01
-4.46770370e-01 -1.37646019e-01 7.91948855e-01 8.39943230e-01
8.69396389e-01 1.75630644e-01 2.67441690e-01 7.25844383e-01
4.97632772e-01 2.07344845e-01 2.26772040e-01 -2.11178899e-01
7.19095290e-01 1.17514944e+00 -3.48382741e-01 -1.57828641e+00
-6.47021115e-01 -4.85001266e-01 -1.23705029e+00 -2.82263570e-02
1.28059685e-01 -9.20669809e-02 -9.46222126e-01 1.30269849e+00
4.24284190e-01 4.01249498e-01 1.93159446e-01 6.20877802e-01
1.13186097e+00 6.57319069e-01 -7.90599361e-02 -4.93426137e-02
7.94692993e-01 -9.95312810e-01 -4.55594361e-01 8.52272063e-02
7.65397847e-01 -4.58273709e-01 6.91026092e-01 3.69916588e-01
-6.47838354e-01 -4.34907615e-01 -1.14777744e+00 4.02044743e-01
-3.42427433e-01 4.19720411e-02 6.63901687e-01 4.35608238e-01
-9.26514328e-01 7.29604661e-01 -6.90407038e-01 -5.84398091e-01
2.81333476e-01 2.82791585e-01 -8.22153151e-01 -3.27046871e-01
-9.11843240e-01 4.55274671e-01 8.89864504e-01 7.18186572e-02
-7.29034722e-01 -6.35568276e-02 -9.15122986e-01 2.09274366e-01
7.52411306e-01 -6.66809380e-01 2.98929751e-01 -7.64999688e-01
-1.08013272e+00 7.15675175e-01 2.22598135e-01 -3.62316608e-01
3.58199328e-02 9.15019363e-02 -4.50418323e-01 1.44138634e-01
-2.44777501e-01 -1.21560097e-02 7.04422951e-01 -1.38040125e+00
-2.09943041e-01 -7.85823226e-01 1.46678239e-01 3.66250008e-01
-3.18541437e-01 -3.33616346e-01 -4.58434910e-01 -2.97507465e-01
4.85044032e-01 -1.00283289e+00 -2.78597206e-01 -4.08690304e-01
-5.87880611e-01 -4.96841639e-01 9.82702196e-01 -6.24592423e-01
1.21061528e+00 -1.80766928e+00 6.26555145e-01 6.05326772e-01
5.15762091e-01 3.46316397e-01 -2.04592973e-01 7.91151762e-01
-1.79490700e-01 4.90263812e-02 -1.69282883e-01 4.32463810e-02
-3.87594879e-01 1.51238531e-01 2.55875379e-01 6.15129530e-01
-2.67340899e-01 7.35565484e-01 -8.21230769e-01 -6.27929211e-01
2.81660616e-01 2.38143310e-01 -3.88214767e-01 2.30748832e-01
-1.71904847e-01 4.62973803e-01 -5.95185220e-01 4.55459118e-01
5.54565251e-01 -6.00526392e-01 6.37169182e-01 -3.50394160e-01
2.55779475e-01 -2.25539953e-01 -1.48972189e+00 1.68216789e+00
-2.06075549e-01 1.49730697e-01 1.86791316e-01 -1.53616548e+00
1.13369298e+00 1.94060847e-01 5.10705173e-01 -1.92020461e-01
1.89469278e-01 -6.64629042e-02 3.32051992e-01 -4.04779792e-01
-4.59811017e-02 9.35141593e-02 2.76092559e-01 3.10907245e-01
2.83792734e-01 2.26940885e-01 3.80347878e-01 5.52698553e-01
1.11399841e+00 1.27160072e-01 6.29090071e-01 -2.65660882e-01
1.10212660e+00 -2.36810744e-01 6.16569340e-01 2.04521179e-01
5.08472733e-02 2.08089754e-01 7.20605731e-01 -4.87821370e-01
-7.52548933e-01 -9.38041866e-01 4.09869403e-01 6.11138284e-01
2.75539607e-01 -8.50551546e-01 -5.90492606e-01 -1.05026281e+00
-1.19930714e-01 1.14884660e-01 -2.47476622e-01 -1.54791459e-01
-5.86599469e-01 -6.57573760e-01 1.08300798e-01 2.77073354e-01
6.53384805e-01 -7.71989405e-01 2.56236821e-01 1.09158710e-01
-4.02824730e-02 -1.18239284e+00 -2.04834759e-01 -4.21678007e-01
-1.10685873e+00 -1.50365400e+00 -4.15220708e-01 -9.86786067e-01
8.69226754e-01 5.83737969e-01 9.88525331e-01 4.59016323e-01
-1.07990988e-01 4.81567919e-01 -4.64999050e-01 1.25797819e-02
-4.08561409e-01 3.62138510e-01 -1.30770028e-01 4.32036608e-01
-7.90525675e-02 -9.36321259e-01 -2.30280489e-01 1.42170623e-01
-7.12750196e-01 2.04221889e-01 7.78276145e-01 7.67595708e-01
6.00233197e-01 2.11148635e-01 6.57051504e-01 -1.37942111e+00
8.04141283e-01 -4.73564059e-01 -8.13291669e-01 6.27380967e-01
-8.34217250e-01 3.78193021e-01 9.27662134e-01 1.81064934e-01
-7.33902812e-01 4.79425006e-02 1.33118898e-01 -5.04893661e-01
-1.33862436e-01 9.89618659e-01 -4.10422266e-01 -4.46643531e-01
3.00656676e-01 3.05881470e-01 1.77665785e-01 -5.23452044e-01
4.14971858e-01 3.85341883e-01 1.81153685e-01 -3.14311892e-01
1.08064258e+00 3.33403677e-01 6.38458312e-01 -8.75193000e-01
-6.06665432e-01 -6.63363636e-01 -8.57631624e-01 -3.26748550e-01
7.35902250e-01 -9.03674841e-01 -5.76701224e-01 4.43564028e-01
-7.94544160e-01 2.45902285e-01 6.50999427e-01 5.32940090e-01
-3.14897746e-01 9.17485476e-01 -3.41686547e-01 -4.94903475e-01
-5.27370095e-01 -8.27486575e-01 6.98851347e-01 1.38236374e-01
5.22828102e-01 -1.38173950e+00 2.41785824e-01 3.58901531e-01
9.64529291e-02 6.36201918e-01 1.30387616e+00 -8.09667945e-01
-7.89566338e-01 -1.99438840e-01 -3.84628862e-01 1.94748878e-01
3.19061428e-01 -1.13388672e-01 -2.95214534e-01 -6.79126561e-01
-3.81779701e-01 -2.26545215e-01 7.88397610e-01 1.05350658e-01
9.57219005e-01 -1.02717884e-01 -6.71023250e-01 5.45237303e-01
1.70309985e+00 -2.44634654e-02 3.39977026e-01 3.08145974e-02
1.36516297e+00 5.04230976e-01 3.83493125e-01 2.13234514e-01
4.00847614e-01 5.64019263e-01 5.90171516e-01 -2.94924881e-02
1.09183416e-01 -3.85507822e-01 8.14500228e-02 1.35125947e+00
-2.89711863e-01 -3.39721352e-01 -1.07894301e+00 2.09117725e-01
-2.06933975e+00 -7.93068230e-01 -3.13950568e-01 2.07725406e+00
-1.26351461e-01 3.56744975e-01 2.62412786e-01 2.47155298e-02
8.82999122e-01 5.29341340e-01 -3.97554845e-01 -3.75708789e-02
1.32625282e-03 -7.73990750e-02 3.24972689e-01 4.01244789e-01
-1.23080385e+00 9.87251103e-01 4.90542841e+00 6.06890976e-01
-8.38103414e-01 -2.27539793e-01 2.73568004e-01 6.77088618e-01
-3.05795372e-01 2.74650574e-01 -3.74495655e-01 -1.19707547e-02
6.53840542e-01 -3.34297299e-01 4.98026103e-01 8.36971521e-01
-1.43973138e-02 2.29037166e-01 -7.85766900e-01 1.13670826e+00
2.83109426e-01 -1.23077953e+00 3.82006228e-01 1.17314987e-01
6.25463724e-01 -5.98365255e-02 -4.21622187e-01 3.35728943e-01
1.69627845e-01 -8.61989558e-01 -2.20090821e-01 5.14048874e-01
2.03349739e-01 -9.15080786e-01 7.95593560e-01 4.92353201e-01
-1.69643891e+00 1.28013134e-01 -3.14910978e-01 5.51060475e-02
-2.26620212e-02 4.57496196e-01 -7.69174278e-01 1.39924204e+00
2.20665544e-01 1.22669673e+00 -9.01806056e-01 1.12956321e+00
-2.57649899e-01 6.60148919e-01 -2.23125219e-02 -9.76491198e-02
3.01917821e-01 -8.71002674e-01 7.70150542e-01 4.71129894e-01
1.36355832e-01 2.42724065e-02 6.00847244e-01 5.98282814e-01
-2.43796498e-01 6.52326703e-01 -1.15610623e+00 -2.68679708e-01
1.52565524e-01 1.60994077e+00 -9.10565794e-01 -1.36040226e-01
-7.99050868e-01 6.11420989e-01 7.03136921e-01 3.17867577e-01
-5.69457769e-01 -3.45843941e-01 1.67368574e-03 3.36988010e-02
1.03822954e-01 -4.86382753e-01 3.58528614e-01 -1.34103715e+00
7.26463320e-03 -8.83076131e-01 6.84092343e-01 -5.31091988e-01
-1.18112731e+00 5.78381300e-01 7.48329088e-02 -1.25844049e+00
-1.78448241e-02 -5.90393662e-01 -7.22346544e-01 5.30938983e-01
-1.31975055e+00 -1.38966894e+00 -5.40690303e-01 7.08851039e-01
1.61523551e-01 -4.81943101e-01 6.16035879e-01 1.96842194e-01
-5.01491845e-01 2.51311392e-01 2.59721607e-01 3.51773918e-01
5.27600050e-01 -1.33548939e+00 2.30438963e-01 7.85434663e-01
6.21217668e-01 4.49940324e-01 2.43261173e-01 -7.89274395e-01
-1.79688704e+00 -1.10367656e+00 3.43169749e-01 5.16360961e-02
6.21538520e-01 -2.30724335e-01 -9.30859745e-01 8.46477449e-01
3.31374764e-01 2.23203331e-01 5.70842981e-01 3.36832911e-01
-3.22508991e-01 -3.03052753e-01 -8.81185234e-01 3.69056374e-01
9.48660672e-01 -2.10697696e-01 -3.29471946e-01 3.61492962e-01
4.63209003e-01 -3.06786925e-01 -9.85057771e-01 7.06878960e-01
3.21146190e-01 -1.04601121e+00 7.71353066e-01 -7.56177545e-01
1.32982405e-02 -4.97420400e-01 7.76154473e-02 -1.65303707e+00
-3.89061302e-01 -2.63813913e-01 -1.12985529e-01 1.20233583e+00
1.25742644e-01 -6.12924159e-01 1.00815701e+00 -3.88080962e-02
7.23919049e-02 -8.42959642e-01 -6.22377276e-01 -6.76471233e-01
-2.71730930e-01 3.52104515e-01 3.74269605e-01 1.09224236e+00
-1.58211499e-01 1.13595724e+00 -4.15588260e-01 3.77774328e-01
9.98755693e-01 5.39488792e-01 1.05954564e+00 -1.74672103e+00
-3.30502927e-01 -1.58324331e-01 -9.85507667e-01 -4.83528048e-01
3.40929717e-01 -1.40070260e+00 -5.97385466e-01 -1.95323646e+00
3.63682806e-01 -2.35485971e-01 -2.48653844e-01 1.05096623e-01
-2.95819372e-01 -3.17566425e-01 2.14350522e-01 6.77637458e-02
-8.86091650e-01 6.32913172e-01 1.32773089e+00 -1.77726865e-01
-1.61046773e-01 7.08849877e-02 -3.92578989e-01 8.32401872e-01
7.79045165e-01 -2.20861509e-01 -6.74767733e-01 -1.68204382e-01
1.02121249e-01 3.07753980e-01 3.29437822e-01 -1.19528496e+00
2.44495615e-01 5.37334941e-02 -3.87278050e-02 -6.43939674e-01
3.57563794e-02 -8.11675191e-01 3.66573244e-01 4.96720254e-01
-6.95745051e-02 3.05820793e-01 -1.57349974e-01 1.15284383e+00
-4.04387265e-01 -5.02885357e-02 7.28421628e-01 -3.32021564e-01
-8.59104753e-01 5.59534848e-01 2.03105807e-01 -2.65146196e-02
1.03585887e+00 1.47361010e-01 -3.24594826e-01 -4.90590602e-01
-8.53982389e-01 4.02729452e-01 2.75534332e-01 4.61928755e-01
6.29926026e-01 -1.41122341e+00 -6.75237477e-01 -4.92167175e-02
3.27217251e-01 -1.48370057e-01 8.79118070e-02 7.43473649e-01
-5.63927591e-01 3.91518861e-01 -1.56099901e-01 -6.12506688e-01
-1.61763060e+00 7.00589955e-01 1.51615992e-01 -7.32404292e-01
-7.11002767e-01 3.56672436e-01 1.16691552e-01 -6.84166431e-01
1.57364219e-01 1.02175742e-01 -5.96341610e-01 -1.76173538e-01
-1.08608223e-01 3.74717414e-01 -2.24892478e-02 -7.74463415e-01
-1.95804939e-01 8.61515760e-01 -1.82806671e-01 3.78941804e-01
1.35811222e+00 -4.05646525e-02 -3.91497552e-01 3.57319444e-01
1.09668362e+00 -1.40468091e-01 -5.37207961e-01 -4.99572515e-01
2.77458370e-01 -2.74656981e-01 -6.06706999e-02 -1.72442898e-01
-1.13982534e+00 9.20474112e-01 3.98102582e-01 3.20357949e-01
1.02584589e+00 -1.17702037e-01 6.13407195e-01 6.52164519e-01
6.96892738e-01 -7.49929726e-01 2.04848751e-01 3.15620959e-01
7.30659187e-01 -1.39808476e+00 3.33414286e-01 -7.01499641e-01
-3.49150002e-01 1.26607144e+00 8.28053713e-01 -4.07200783e-01
9.20413256e-01 -5.42468965e-01 -2.50773221e-01 -5.47215700e-01
-5.49613893e-01 -3.14434469e-01 4.41368222e-01 4.32768762e-01
3.47499967e-01 7.83241019e-02 -4.74249482e-01 7.10140094e-02
2.12866485e-01 -3.86644572e-01 4.89346474e-01 6.87114656e-01
-4.02627170e-01 -1.39377093e+00 -1.22683853e-01 6.22018278e-01
-1.57047749e-01 1.92137763e-01 -7.25164473e-01 9.69346881e-01
-3.54400188e-01 8.82799029e-01 -5.63742220e-01 -5.78939021e-01
2.05682144e-01 -3.45213749e-02 4.90743399e-01 -7.26530671e-01
-1.56512439e-01 5.53697124e-02 2.13279232e-01 -4.89314139e-01
-8.23718429e-01 -3.84482771e-01 -1.08480215e+00 -7.49434382e-02
-4.75967526e-01 4.09264207e-01 3.73133808e-01 8.93991649e-01
4.16051447e-01 4.03620094e-01 7.71537483e-01 -5.29005587e-01
-3.00618529e-01 -7.78059185e-01 -8.44345331e-01 5.28551042e-01
-5.13488092e-02 -7.74544656e-01 -2.43976265e-01 -2.66312897e-01]
|
[7.344528675079346, 6.138421058654785]
|
a942c6e1-3ed9-4596-8c51-0ddff2c5767c
|
beeds-large-scale-biomedical-event-extraction
| null | null |
https://aclanthology.org/2022.bionlp-1.28
|
https://aclanthology.org/2022.bionlp-1.28.pdf
|
BEEDS: Large-Scale Biomedical Event Extraction using Distant Supervision and Question Answering
|
Automatic extraction of event structures from text is a promising way to extract important facts from the evergrowing amount of biomedical literature. We propose BEEDS, a new approach on how to mine event structures from PubMed based on a question-answering paradigm. Using a three-step pipeline comprising a document retriever, a document reader, and an entity normalizer, BEEDS is able to fully automatically extract event triples involving a query protein or gene and to store this information directly in a knowledge base. BEEDS applies a transformer-based architecture for event extraction and uses distant supervision to augment the scarce training data in event mining. In a knowledge base population setting, it outperforms a strong baseline in finding post-translational modification events consisting of enzyme-substrate-site triples while achieving competitive results in extracting binary relations consisting of protein-protein and protein-site interactions.
|
['Leon Weber', 'Ulf Leser', 'Xing David Wang']
| null | null | null | null |
bionlp-acl-2022-5
|
['knowledge-base-population']
|
['natural-language-processing']
|
[ 6.03996277e-01 2.99912483e-01 -4.89509255e-01 -2.24319324e-01
-1.27375889e+00 -5.09966791e-01 5.58978081e-01 1.23899055e+00
-7.72837043e-01 1.21684861e+00 4.02616262e-01 -6.11193180e-01
-2.03608856e-01 -9.11506772e-01 -9.83324468e-01 -5.29892445e-01
1.78959548e-01 7.68446982e-01 3.95646542e-01 -6.53410181e-02
1.82024166e-01 4.79079217e-01 -1.22538424e+00 7.19238281e-01
7.00158358e-01 6.73845470e-01 -5.39804585e-02 4.80038375e-01
-3.43003362e-01 8.98105860e-01 -4.69925016e-01 -6.84936702e-01
-2.85960793e-01 -6.32274568e-01 -9.06163454e-01 -5.89579344e-01
-1.62690446e-01 8.26780871e-02 -9.42598209e-02 5.97433627e-01
7.67842650e-01 -1.95712835e-01 7.10160553e-01 -8.26964140e-01
-3.15736718e-02 7.73915172e-01 -5.21257401e-01 6.47882283e-01
6.01467192e-01 -6.49178028e-02 1.42256510e+00 -1.00192440e+00
1.27702618e+00 7.83415139e-01 4.92011398e-01 1.99583352e-01
-1.33447921e+00 -4.77118671e-01 -3.27876776e-01 2.25972608e-01
-1.39109278e+00 -3.57965797e-01 2.16358319e-01 -2.04341143e-01
1.85229921e+00 4.29717302e-01 5.37794948e-01 1.06638682e+00
3.84065956e-01 5.55624664e-01 4.43289131e-01 -7.20064402e-01
3.01234066e-01 -1.97186545e-01 3.37618202e-01 7.58049130e-01
4.20924723e-01 -2.34273523e-01 -9.91186678e-01 -7.07475126e-01
1.25872120e-01 7.14715049e-02 -7.85047412e-02 8.28917101e-02
-1.24488795e+00 5.52488863e-01 1.30354017e-01 4.39428657e-01
-7.71941781e-01 -1.38866812e-01 7.29055762e-01 2.29912743e-01
3.46518606e-01 6.69250309e-01 -9.49227810e-01 1.01192035e-01
-8.93127799e-01 2.74720907e-01 8.46564949e-01 7.37660587e-01
4.29629058e-01 -1.00407445e+00 -4.57904190e-01 5.03789306e-01
4.54015993e-02 -3.59614007e-02 6.84786379e-01 -1.57410145e-01
3.63052607e-01 9.34502840e-01 4.86255661e-02 -6.56215668e-01
-7.63942957e-01 -2.05965519e-01 -3.73247296e-01 -4.32762504e-01
3.08693886e-01 -1.36215478e-01 -8.96571517e-01 1.59077394e+00
6.63873374e-01 1.96910366e-01 1.84697986e-01 4.30934966e-01
1.01362896e+00 3.19161057e-01 4.04616058e-01 -5.27717292e-01
2.04296088e+00 -5.99506021e-01 -8.47370267e-01 -1.24860900e-02
8.55526924e-01 -4.60107952e-01 3.99656087e-01 2.94987828e-01
-1.07655597e+00 1.70425013e-01 -9.42387998e-01 -6.30469441e-01
-7.18052268e-01 -1.55342952e-03 7.14764178e-01 2.39313077e-02
-5.44865489e-01 4.48971480e-01 -1.10081661e+00 -5.78932405e-01
5.47987342e-01 3.59109372e-01 -6.51616812e-01 8.96382798e-03
-1.27855742e+00 1.12660789e+00 8.07849646e-01 -3.21431428e-01
-4.78894264e-01 -1.12113953e+00 -8.35076034e-01 2.22774625e-01
6.20878875e-01 -1.06704402e+00 1.06213534e+00 -2.21312083e-02
-9.79879558e-01 1.08030844e+00 -4.63535219e-01 -7.68587530e-01
-5.81351891e-02 -8.46454650e-02 -5.84126532e-01 3.91424417e-01
3.62422705e-01 2.15068415e-01 1.61747411e-01 -1.27013043e-01
-8.16366136e-01 -7.50380635e-01 -3.56221706e-01 1.46037817e-01
-9.19822603e-02 3.14442396e-01 -4.46042001e-01 -5.44705987e-01
6.00239262e-02 -5.16541779e-01 -4.64114249e-01 -1.99540973e-01
-5.99210143e-01 -5.84408700e-01 3.27950507e-01 -6.32073998e-01
1.26965511e+00 -1.74703193e+00 9.45541635e-02 3.07601035e-01
2.98853338e-01 1.43571973e-01 2.34526113e-01 7.33532727e-01
-5.03773987e-01 4.77409028e-02 -1.62771538e-01 1.37057766e-01
-2.34040141e-01 1.10604368e-01 -2.56728321e-01 1.74160346e-01
7.96463907e-01 1.17885542e+00 -1.02557230e+00 -7.15890229e-01
-2.27485478e-01 3.83554697e-01 -4.75485146e-01 -3.02289687e-02
-6.81566238e-01 -7.66306072e-02 -6.20288551e-01 6.79779530e-01
1.88445330e-01 -6.12985253e-01 6.37315869e-01 -2.05116943e-01
2.16239631e-01 8.36348951e-01 -8.22008371e-01 1.81085801e+00
-2.68133115e-02 1.57394290e-01 -3.08023602e-01 -1.14231050e+00
6.44414902e-01 6.83068573e-01 6.23730779e-01 -5.91451287e-01
1.24753393e-01 2.41725489e-01 -2.02597991e-01 -5.18879771e-01
9.99092087e-02 -4.87483352e-01 -7.68109933e-02 4.22329545e-01
5.28484106e-01 5.09500146e-01 5.80078244e-01 5.37349284e-01
1.98795092e+00 -4.72637117e-02 8.95257235e-01 -1.02521166e-01
5.07234216e-01 2.23414212e-01 7.25298822e-01 5.21337628e-01
3.29693347e-01 2.03872636e-01 8.99418712e-01 -2.90445149e-01
-7.06620514e-01 -9.29661393e-01 -3.53579432e-01 9.06380594e-01
-4.17471856e-01 -9.36216116e-01 -2.81422943e-01 -9.04175282e-01
4.71445546e-02 7.11986244e-01 -7.46651351e-01 -2.51725644e-01
-4.13544506e-01 -1.21001291e+00 7.11763442e-01 3.43746036e-01
-5.98403104e-02 -9.15364087e-01 -4.26214725e-01 5.28450668e-01
-4.58425194e-01 -1.11155844e+00 -2.72252858e-01 1.00759566e+00
-6.39171481e-01 -1.44592083e+00 -3.79424542e-01 -6.11535907e-01
5.91966867e-01 -3.25045973e-01 1.27221537e+00 -1.80980742e-01
-6.27544701e-01 -4.87792879e-01 -8.21079761e-02 -9.70550835e-01
-3.84355426e-01 2.33179361e-01 -2.97240347e-01 -3.62084746e-01
9.13794219e-01 -5.73840261e-01 -5.36414206e-01 -4.16162983e-02
-9.76400554e-01 -3.12093645e-02 8.15847516e-01 7.55029321e-01
9.74870861e-01 8.48941579e-02 7.12988555e-01 -1.10997176e+00
3.81052345e-01 -8.41780961e-01 -5.22982717e-01 3.50361824e-01
-8.35240066e-01 3.79144877e-01 4.33866024e-01 -1.02795050e-01
-9.65633094e-01 2.92602479e-01 -3.27272266e-01 3.02268922e-01
-1.23462990e-01 1.02112126e+00 -3.16886276e-01 7.00068772e-01
9.24129725e-01 2.44625717e-01 -1.38469487e-01 -4.88991112e-01
2.47465432e-01 3.94773602e-01 6.55227005e-01 -3.73711944e-01
3.13757062e-01 4.47800845e-01 3.06001842e-01 -4.10640091e-01
-9.94330466e-01 -7.61399686e-01 -4.40101713e-01 5.77500105e-01
7.89054573e-01 -1.04519308e+00 -7.07128525e-01 3.78645808e-02
-8.95544946e-01 1.23132877e-01 -5.88123024e-01 5.76433718e-01
-2.10313320e-01 -4.01913701e-03 -8.48073184e-01 -1.97307929e-01
-5.87305665e-01 -5.91554821e-01 1.32920098e+00 1.21632554e-01
-3.49932224e-01 -5.96572220e-01 4.21639830e-01 1.83528513e-01
-2.56121099e-01 3.26316386e-01 1.19316626e+00 -1.31329143e+00
-3.23609740e-01 -3.18347812e-01 -6.12622313e-02 -5.52452803e-01
1.57244846e-01 -1.92125067e-01 -4.71018702e-01 2.67385900e-01
-4.71708596e-01 -2.33042076e-01 1.06541312e+00 1.04847752e-01
7.52645791e-01 -3.25770527e-01 -9.80225265e-01 2.57317901e-01
1.06258333e+00 2.45605744e-02 5.81279874e-01 2.64996022e-01
3.08291078e-01 2.93192059e-01 4.88937467e-01 2.86865324e-01
2.69404382e-01 6.46134555e-01 -1.72476813e-01 -1.25571638e-01
1.46912858e-01 -3.25913966e-01 9.91182700e-02 1.48968846e-01
1.54916421e-01 -4.52296257e-01 -9.75163162e-01 5.69111347e-01
-1.66126764e+00 -9.71704304e-01 -2.90118337e-01 1.89791453e+00
1.67936730e+00 2.69458175e-01 5.70875630e-02 1.89514786e-01
5.44360518e-01 -4.10482109e-01 -6.08526766e-01 -9.27124694e-02
-2.45052516e-01 6.91164255e-01 2.80508906e-01 7.29057342e-02
-7.66760170e-01 8.60483885e-01 6.30661058e+00 6.46260142e-01
-7.56827235e-01 -5.60138151e-02 5.22931159e-01 -2.13455126e-01
-1.57162502e-01 1.12634804e-02 -1.07429016e+00 3.67080480e-01
1.47591662e+00 -4.02019739e-01 -3.73461962e-01 4.50777113e-01
1.15619220e-01 -2.93971241e-01 -1.26320672e+00 7.91152239e-01
-2.23059684e-01 -1.90572739e+00 1.48433782e-02 1.61702409e-01
1.48187518e-01 2.38617197e-01 -6.20055914e-01 -4.68662530e-02
5.03925145e-01 -8.05814505e-01 1.42892644e-01 7.02493489e-01
4.82826322e-01 -6.11345649e-01 9.50678945e-01 2.46715099e-01
-8.31548631e-01 2.47538954e-01 -1.09990224e-01 3.01804066e-01
1.14069089e-01 1.33172083e+00 -1.50233746e+00 8.28285694e-01
5.88822246e-01 7.35170186e-01 -4.47262168e-01 1.05233598e+00
-3.44607502e-01 7.45889664e-01 -6.44471288e-01 2.18457077e-02
-2.80492038e-01 2.80674756e-01 3.62692565e-01 1.45410347e+00
1.04081526e-01 6.06556952e-01 -1.57930717e-01 6.38834834e-01
-3.84021342e-01 3.39474797e-01 -4.83012736e-01 -2.36387566e-01
3.82240474e-01 1.44311249e+00 -8.79102290e-01 -5.16859591e-01
-2.49615282e-01 9.66244876e-01 4.46906060e-01 2.38730293e-02
-7.06589222e-01 -6.60148501e-01 3.84589165e-01 -3.98501791e-02
4.93805915e-01 3.34921330e-01 1.25125214e-01 -1.02899432e+00
1.11945637e-01 -9.36009288e-01 1.16506398e+00 -7.06701815e-01
-9.71466601e-01 2.48822153e-01 -1.93840340e-01 -6.20155632e-01
-2.88395226e-01 -3.76934856e-01 -2.65473425e-01 7.33444810e-01
-1.25570428e+00 -1.14175391e+00 1.96073532e-01 4.89028305e-01
1.94616839e-02 2.09692791e-01 1.02514267e+00 4.84101295e-01
-7.71405220e-01 4.60027456e-01 -1.33341774e-01 2.03198597e-01
9.51664448e-01 -1.37787461e+00 4.99830902e-01 5.01965165e-01
2.89847672e-01 8.05734158e-01 6.31101608e-01 -8.72755349e-01
-1.37677085e+00 -1.17408514e+00 1.58756053e+00 -7.63093412e-01
6.82182074e-01 -4.35074806e-01 -9.28602755e-01 9.22552466e-01
-6.70533255e-02 1.09964907e-01 1.28950596e+00 3.23989093e-01
-4.11370844e-01 1.68732509e-01 -1.12339103e+00 3.09132159e-01
1.11413634e+00 -5.22347152e-01 -1.08603656e+00 5.99879086e-01
7.04240799e-01 -4.00389373e-01 -1.19219506e+00 2.08473742e-01
2.81274199e-01 -3.17496926e-01 1.03395009e+00 -1.19644284e+00
5.55155396e-01 -3.84789586e-01 2.02993557e-01 -8.53444517e-01
-7.10203201e-02 -7.22364306e-01 -4.55274075e-01 1.20484638e+00
9.93562400e-01 -4.38360333e-01 6.88190579e-01 4.53578681e-01
-1.25692293e-01 -6.04991794e-01 -9.36775684e-01 -2.04226255e-01
-2.51759470e-01 -2.06076559e-02 4.44761485e-01 8.98712695e-01
6.68004811e-01 1.09557760e+00 1.99663118e-01 1.13725327e-01
2.62484729e-01 6.80698529e-02 3.64965469e-01 -1.41575313e+00
-3.89669031e-01 4.23469841e-02 -3.62318456e-01 -3.29973370e-01
-3.48608010e-02 -1.29046106e+00 -1.12150371e-01 -1.76790106e+00
6.50911570e-01 1.49401098e-01 -7.21287131e-01 8.76465917e-01
-4.06025976e-01 1.36716023e-01 -7.41260827e-01 -8.56227204e-02
-8.05021763e-01 2.21641019e-01 4.80694234e-01 -5.47605392e-04
-2.85661697e-01 -1.84460521e-01 -1.08235562e+00 6.02570891e-01
3.70035976e-01 -9.10740018e-01 -1.57915220e-01 2.71346420e-01
7.89361060e-01 2.36548841e-01 2.78137386e-01 -6.15792513e-01
3.72894824e-01 9.77790281e-02 5.14001787e-01 -8.54370117e-01
-9.13751796e-02 -5.64680457e-01 2.60987818e-01 4.72800046e-01
-5.01384318e-01 1.35487807e-03 3.87755662e-01 8.20041418e-01
-1.06164113e-01 -4.15469036e-02 3.44281703e-01 -7.40355551e-02
-1.58535033e-01 8.40443224e-02 -5.78435838e-01 1.12924397e-01
9.46605980e-01 3.39767069e-01 -4.87943321e-01 2.99097538e-01
-1.01964831e+00 2.33390495e-01 -1.17123956e-02 3.03511441e-01
2.42256656e-01 -8.94105077e-01 -7.53210187e-01 -1.04440175e-01
3.56190532e-01 -4.15021405e-02 -1.12445429e-01 1.00576305e+00
-3.22147816e-01 6.61339462e-01 1.25700250e-01 -2.22491547e-01
-1.48225427e+00 7.09222257e-01 -3.03771417e-03 -9.12262499e-01
-6.34775460e-01 9.22647893e-01 -8.24811775e-03 -2.88905025e-01
-9.27443653e-02 -4.05184269e-01 -1.63336411e-01 3.13806117e-01
9.15779471e-01 -2.65916716e-02 8.12742770e-01 -5.11168204e-02
-8.15303743e-01 -3.21072757e-01 -5.66850781e-01 -1.11328594e-01
1.62400997e+00 3.17131251e-01 -4.05677021e-01 2.63180733e-01
1.20717072e+00 1.05388194e-01 -4.89580452e-01 -3.72482210e-01
6.95839942e-01 1.68037787e-01 2.54894085e-02 -1.22702157e+00
-6.60539448e-01 1.76218480e-01 1.52842551e-01 -2.54499197e-01
9.35679376e-01 4.40343887e-01 5.61930835e-01 6.85806751e-01
1.14014201e-01 -8.06224883e-01 -1.92139298e-01 3.82510245e-01
6.09562755e-01 -1.03444541e+00 2.83747464e-01 -5.66985548e-01
-1.20814309e-01 9.23460543e-01 1.49358451e-01 4.49923217e-01
6.17863655e-01 4.86845791e-01 -4.00355607e-01 -6.93808496e-01
-1.20199752e+00 -2.28842929e-01 3.90881300e-01 1.98129565e-01
6.44783139e-01 -4.05360818e-01 -7.12769806e-01 9.76965964e-01
-3.00122909e-02 4.16154921e-01 8.69259983e-02 1.10823345e+00
-4.61216494e-02 -1.56011105e+00 2.10920852e-02 8.72602105e-01
-1.12113440e+00 -5.85757196e-01 -5.22345185e-01 6.08379066e-01
1.87720023e-02 8.24267387e-01 -4.85706318e-04 9.50259939e-02
4.31696862e-01 6.17091000e-01 3.17905575e-01 -7.29834378e-01
-8.40623200e-01 2.43096501e-01 4.65686351e-01 -6.37962222e-01
-4.13688481e-01 -9.64809120e-01 -1.84154391e+00 2.22336486e-01
-5.87546825e-01 6.47515953e-01 3.12338889e-01 1.07211542e+00
1.08153200e+00 8.02313387e-01 -4.04071808e-02 1.82956442e-01
3.35872546e-02 -7.18835533e-01 -3.94838929e-01 2.00230271e-01
1.61406800e-01 -4.57455397e-01 1.26579762e-01 2.52588093e-01]
|
[8.551033020019531, 8.81286334991455]
|
e0beae2a-d7cf-4ef5-906c-0cd444224764
|
paused-agent-replay-refresh
|
2209.13398
| null |
https://arxiv.org/abs/2209.13398v1
|
https://arxiv.org/pdf/2209.13398v1.pdf
|
Paused Agent Replay Refresh
|
Reinforcement learning algorithms have become more complex since the invention of target networks. Unfortunately, target networks have not kept up with this increased complexity, instead requiring approximate solutions to be computationally feasible. These approximations increase noise in the Q-value targets and in the replay sampling distribution. Paused Agent Replay Refresh (PARR) is a drop-in replacement for target networks that supports more complex learning algorithms without this need for approximation. Using a basic Q-network architecture, and refreshing the novelty values, target values, and replay sampling distribution, PARR gets 2500 points in Montezuma's Revenge after only 30.9 million Atari frames. Finally, interpreting PARR in the context of carbon-based learning offers a new reason for sleep.
|
['Benjamin Parr']
|
2022-09-26
| null | null | null | null |
['montezumas-revenge']
|
['playing-games']
|
[ 2.15810478e-01 9.34634581e-02 -5.59770405e-01 3.48082110e-02
-4.92440194e-01 -7.16520786e-01 7.04595685e-01 -1.32894879e-02
-8.48656714e-01 1.38627303e+00 -1.89359456e-01 -4.48051363e-01
-4.46259409e-01 -9.46882606e-01 -6.74243569e-01 -9.15142179e-01
-3.99715245e-01 3.39221001e-01 2.51235276e-01 -2.70273477e-01
3.30049157e-01 5.00238419e-01 -1.32569385e+00 -1.66718751e-01
4.40467387e-01 6.21277511e-01 1.02141164e-01 9.44655716e-01
1.65120721e-01 9.58955407e-01 -1.23697793e+00 1.83042455e-02
4.28337932e-01 -8.65584850e-01 -2.39608347e-01 -5.15193164e-01
-2.40625754e-01 -5.78247190e-01 -4.60168153e-01 8.36727142e-01
6.11608446e-01 3.76357406e-01 3.85440260e-01 -1.57036006e+00
-1.98558345e-01 9.31893826e-01 -5.87871373e-01 4.57351208e-01
8.93187374e-02 8.12294424e-01 7.33325660e-01 4.93943170e-02
5.31382561e-01 1.10994828e+00 5.86868703e-01 7.36110270e-01
-1.25549853e+00 -7.46134818e-01 4.72662747e-02 2.76451826e-01
-1.11558890e+00 -2.95148283e-01 3.56538504e-01 6.48690835e-02
1.17550004e+00 3.11525255e-01 1.21349645e+00 1.38659537e+00
5.80822825e-01 2.91840762e-01 1.11029422e+00 -1.75720051e-01
8.07975531e-01 -1.59280017e-01 -5.82947969e-01 4.23181802e-01
6.30695283e-01 9.27009165e-01 -5.85392356e-01 -3.56591582e-01
1.08566570e+00 1.84140112e-02 -5.78343086e-02 -6.15983233e-02
-1.07889569e+00 1.09067500e+00 3.06494415e-01 -6.47611767e-02
-6.36855602e-01 9.33661520e-01 4.71900731e-01 4.68507946e-01
1.53269604e-01 8.69191945e-01 -5.44148982e-01 -4.67386782e-01
-7.07875729e-01 4.22124475e-01 7.14701116e-01 5.37141025e-01
5.23126304e-01 7.61416852e-01 1.61831588e-01 1.31946519e-01
2.08872214e-01 6.31559014e-01 4.78050947e-01 -1.72518981e+00
-7.41548166e-02 -1.11385975e-02 3.85654271e-01 -4.38644797e-01
-6.07741237e-01 -5.33511221e-01 -4.43711191e-01 7.37099349e-01
8.40944111e-01 -7.95271635e-01 -7.01416612e-01 1.59323573e+00
2.88608372e-01 1.87482461e-01 4.06209618e-01 6.43059254e-01
-1.23307273e-01 9.54654455e-01 -4.23563644e-03 -4.87179160e-01
8.43970120e-01 -8.12067091e-01 -5.09055078e-01 2.10382253e-01
1.79416180e-01 -1.63252398e-01 1.05695105e+00 8.18724990e-01
-9.76470292e-01 -9.33058709e-02 -1.21934569e+00 8.72949898e-01
-3.38465363e-01 -7.84937561e-01 8.07734430e-01 9.41949725e-01
-8.31190169e-01 9.95844960e-01 -8.13251495e-01 -3.07645917e-01
3.74025404e-01 5.78579128e-01 2.20900476e-01 3.28278065e-01
-1.15213299e+00 9.95299220e-01 5.16038954e-01 -3.15877885e-01
-1.43952441e+00 -7.95868576e-01 -3.54770482e-01 1.10547520e-01
7.62167752e-01 -6.02116108e-01 1.39443898e+00 -9.82727826e-01
-2.10944557e+00 -3.03893268e-01 5.62963426e-01 -1.01116955e+00
4.13258731e-01 1.06898241e-01 -4.09668744e-01 2.58858204e-01
-3.19343239e-01 8.01262319e-01 9.60520267e-01 -1.06048298e+00
-6.66338325e-01 -6.98395371e-02 6.24178872e-02 5.57531536e-01
8.82109776e-02 -2.45032996e-01 5.70959806e-01 -1.79571420e-01
-5.44256628e-01 -7.39914656e-01 -6.46127582e-01 -1.86551750e-01
-7.48994201e-02 -1.93125725e-01 5.87874889e-01 6.43968284e-02
7.56363988e-01 -1.65916312e+00 -2.65303195e-01 2.56099075e-01
1.34612605e-01 -1.37424484e-01 -4.45137084e-01 5.49083531e-01
-1.39690144e-02 2.33199969e-01 1.05203614e-01 6.59141719e-01
1.00482769e-01 4.31773633e-01 -2.39465237e-01 4.47063774e-01
-2.08159769e-03 6.79757774e-01 -1.36606705e+00 6.72876015e-02
2.56649882e-01 2.06700832e-01 -4.99252796e-01 7.13369623e-02
-5.82623482e-01 1.14800230e-01 -1.85651183e-01 5.66856623e-01
1.48279354e-01 -7.01728687e-02 4.44906622e-01 2.36521974e-01
-2.74645805e-01 3.92173007e-02 -8.09892476e-01 1.44051921e+00
6.07734248e-02 3.28998685e-01 -1.42678395e-01 -8.21079493e-01
6.75018609e-01 1.89481378e-01 1.04266608e+00 -9.89669025e-01
7.71635175e-02 1.79214045e-01 4.40185457e-01 -2.84395397e-01
3.96153927e-01 -3.96090090e-01 9.21961442e-02 6.12239838e-01
-1.85676083e-01 -6.35870874e-01 2.17534453e-01 9.72218886e-02
1.40402591e+00 3.50220948e-01 8.98941085e-02 -2.16630295e-01
-1.63171485e-01 3.38583767e-01 5.32201350e-01 1.00674975e+00
-4.31815237e-01 -1.04576573e-01 6.52289331e-01 -6.60081148e-01
-1.20215786e+00 -1.31999087e+00 3.41147780e-01 9.38171566e-01
-1.40805244e-01 -2.32355639e-01 -6.11124992e-01 -6.62319183e-01
1.82671770e-01 9.29626942e-01 -4.26619321e-01 -2.68282503e-01
-3.66683990e-01 -1.02948153e+00 8.40491891e-01 -2.56920774e-02
3.14081788e-01 -9.99846637e-01 -1.05989790e+00 6.17553413e-01
2.26742834e-01 -2.21683353e-01 -2.07399458e-01 8.23622763e-01
-8.60521674e-01 -1.09381866e+00 -5.64646304e-01 1.63503557e-01
1.53563485e-01 8.17066878e-02 1.08798289e+00 -1.63243026e-01
-2.25211412e-01 3.36360067e-01 -1.22592002e-01 -8.03114414e-01
-6.35261655e-01 1.58950333e-02 3.16079944e-01 -8.92265975e-01
1.43911570e-01 -7.05577195e-01 -6.56426191e-01 6.24692347e-03
-8.15172791e-01 -4.36418086e-01 5.34085512e-01 9.09034669e-01
4.09427434e-01 3.31835032e-01 1.31238866e+00 -5.46448588e-01
7.92474151e-01 -5.98070979e-01 -1.04620731e+00 -5.96163571e-02
-1.02340353e+00 -2.16143541e-02 6.85967803e-01 -6.90725803e-01
-7.48825550e-01 -1.28422335e-01 1.40342221e-01 -2.99980223e-01
5.18308133e-02 2.72164971e-01 4.13672268e-01 2.65259266e-01
1.11699891e+00 1.77401062e-02 5.50933599e-01 1.57610387e-01
4.88282025e-01 1.17403455e-01 2.50442296e-01 -4.76983339e-01
7.10311472e-01 1.61413059e-01 1.53188139e-01 -6.77899003e-01
-4.91238624e-01 2.21024871e-01 2.86761910e-01 -4.66955304e-01
5.41125476e-01 -8.78393590e-01 -1.35973394e+00 3.26143146e-01
-6.52015150e-01 -8.98350418e-01 -1.27127469e+00 6.32624626e-01
-7.81732559e-01 7.18122199e-02 -4.12274390e-01 -1.13575900e+00
-1.10008651e-02 -8.34972203e-01 -5.69498837e-02 6.88753188e-01
-8.47795233e-02 -6.92068398e-01 4.06770796e-01 -2.29341298e-01
8.05774391e-01 4.38689411e-01 6.33687615e-01 -6.42724752e-01
-6.47378206e-01 2.69452423e-01 1.32700518e-01 4.63771261e-02
3.84063780e-01 1.56353071e-01 -8.45640540e-01 -4.30447489e-01
-1.06700296e-02 -4.19565588e-01 5.13650715e-01 6.79621041e-01
7.59959519e-01 -6.79748058e-01 8.77779424e-02 2.89231598e-01
1.41289890e+00 9.26159441e-01 6.36002779e-01 5.65060258e-01
-8.15508887e-03 2.44105786e-01 4.53251600e-01 8.61507714e-01
-1.37196137e-02 1.86946243e-01 8.05688083e-01 9.33085606e-02
2.35727251e-01 -2.62382954e-01 8.58816087e-01 4.28523660e-01
-1.05528496e-01 -4.85527873e-01 -5.27665377e-01 1.75606847e-01
-1.48894238e+00 -1.42198873e+00 3.39098215e-01 2.20239305e+00
1.07853115e+00 6.08855844e-01 7.63525605e-01 -2.44139016e-01
4.73299801e-01 5.32132424e-02 -1.48604250e+00 -5.61256826e-01
-7.30606392e-02 2.60341197e-01 1.02525461e+00 3.81972820e-01
-3.38163376e-01 6.53822243e-01 7.94698715e+00 8.13626885e-01
-1.01122653e+00 -1.07490465e-01 5.74469209e-01 -5.23467362e-01
-3.86189848e-01 9.68809202e-02 -3.61091554e-01 5.42057812e-01
1.77366662e+00 -4.97407138e-01 1.13677895e+00 5.77203572e-01
6.22631252e-01 -6.87258899e-01 -7.92477965e-01 7.11368501e-01
-4.23895478e-01 -1.51865923e+00 -3.86128984e-02 2.05950841e-01
4.55563396e-01 3.94221395e-01 7.10669309e-02 5.17543972e-01
9.69768047e-01 -1.11190045e+00 5.80196321e-01 5.42860687e-01
5.97505987e-01 -1.23103321e+00 4.54844981e-01 1.80504471e-01
-6.07022703e-01 -3.64793777e-01 -5.66441894e-01 -3.94505113e-01
-1.12293623e-01 4.26458269e-01 -1.00134552e+00 2.60577887e-01
3.76075804e-01 2.08545342e-01 -8.12140405e-02 8.05650353e-01
1.20989792e-01 7.89817870e-01 -6.79516137e-01 -4.18077707e-01
3.63468111e-01 -3.85742217e-01 4.91312504e-01 6.48494363e-01
1.77764028e-01 -1.67040437e-01 2.20926516e-02 8.65998507e-01
1.36600867e-01 -4.44502413e-01 -6.86959147e-01 -3.72794569e-01
8.84291828e-01 1.04958558e+00 -8.34290326e-01 -2.98871338e-01
-3.81719992e-02 4.82523441e-01 -3.83604199e-01 4.66014534e-01
-1.06532431e+00 -6.23539150e-01 8.25408280e-01 -8.12916532e-02
-5.55502474e-02 -2.25528255e-01 2.10777476e-01 -6.30138636e-01
-8.37868750e-01 -1.20014656e+00 2.49993414e-01 -6.89913869e-01
-1.15189815e+00 1.04027838e-01 1.37244642e-01 -8.40152442e-01
-5.51197112e-01 -3.43925834e-01 -5.37249446e-01 4.46792126e-01
-1.49898660e+00 -3.18127483e-01 2.16273457e-01 5.17539084e-01
5.91843367e-01 -2.12936103e-01 7.09558427e-01 -3.40559602e-01
-4.70815122e-01 2.92745024e-01 4.43907917e-01 -6.11810684e-01
4.31821883e-01 -1.42487645e+00 4.15794820e-01 3.77818614e-01
-1.73997402e-01 1.57989994e-01 9.40541267e-01 -6.15312636e-01
-1.50579274e+00 -7.57336438e-01 -3.25701177e-01 -2.67906398e-01
8.76726210e-01 1.67726316e-02 -3.19973201e-01 3.68682027e-01
5.41527748e-01 -4.38660830e-01 5.20927846e-01 -2.88711756e-01
-1.17433123e-01 -2.37912849e-01 -1.30387211e+00 8.62343669e-01
5.79576492e-01 -2.52497375e-01 -2.91592907e-02 4.65899318e-01
1.03215718e+00 -2.17649758e-01 -9.85937059e-01 -2.01952234e-01
5.53340435e-01 -9.11086559e-01 8.55443895e-01 -6.67957962e-01
-1.36698708e-01 -3.89785171e-01 -9.47245508e-02 -1.46629381e+00
-5.14963977e-02 -1.29984772e+00 -1.38622627e-01 7.05592752e-01
3.39272439e-01 -7.48450279e-01 1.14453781e+00 3.74697834e-01
4.56102081e-02 -6.57140851e-01 -1.19326949e+00 -1.22339725e+00
1.00735694e-01 -1.13860063e-01 6.78925753e-01 7.96913862e-01
2.15538651e-01 2.44913206e-01 -3.46988708e-01 -1.93469793e-01
9.20446694e-01 -5.15083492e-01 5.30205905e-01 -1.00675356e+00
-5.27800679e-01 -4.32225019e-01 2.31444798e-02 -5.65118074e-01
-3.78333509e-01 -4.69040304e-01 1.70820709e-02 -1.17995393e+00
9.76825058e-02 -3.19558084e-01 -4.40202832e-01 5.07244408e-01
2.61958957e-01 -5.09917885e-02 4.04562980e-01 8.56458843e-02
-4.33127522e-01 6.85978293e-01 1.06092238e+00 -1.43876940e-01
-4.81876165e-01 -1.05654709e-01 -5.38535476e-01 5.79554677e-01
1.30985320e+00 -6.79751575e-01 -8.34199965e-01 1.44557074e-01
5.03026366e-01 3.32186759e-01 2.93197960e-01 -1.15904617e+00
-1.94414635e-03 -7.66075552e-01 6.07291281e-01 -4.11028236e-01
3.28593373e-01 -8.97797227e-01 5.42568982e-01 1.10359526e+00
-3.63755852e-01 5.16984224e-01 1.89431846e-01 7.45881677e-01
6.92910373e-01 -4.22414213e-01 9.06365275e-01 -4.13811743e-01
-2.47529581e-01 -1.54961068e-02 -1.17683232e+00 5.50508387e-02
1.09536242e+00 -3.99344623e-01 -7.27656662e-01 -4.36920762e-01
-6.12301648e-01 2.23643646e-01 5.59580624e-01 1.57081913e-02
6.19685471e-01 -1.01261675e+00 -3.51088554e-01 -1.34493291e-01
-5.78172266e-01 -8.03846791e-02 1.22154772e-01 3.25314224e-01
-5.07177472e-01 -4.29374836e-02 -6.24326766e-01 -3.79103899e-01
-4.95353997e-01 6.44981980e-01 7.53499687e-01 -2.97026098e-01
-3.77282113e-01 4.16699558e-01 -5.53447306e-01 -1.96254298e-01
1.38441131e-01 -2.39129469e-01 2.23139822e-01 7.64625818e-02
4.84806657e-01 6.67646050e-01 -4.37343717e-01 5.01116693e-01
-1.00889832e-01 -1.67884439e-01 1.55697325e-02 -4.54910576e-01
1.39669347e+00 -7.68201128e-02 2.64749825e-01 6.24191344e-01
3.92776698e-01 -1.92737773e-01 -1.89039838e+00 4.00663137e-01
1.00639038e-01 -1.45537034e-01 -2.11342983e-02 -1.08738768e+00
-8.13566923e-01 3.08369339e-01 8.71859193e-01 4.77295220e-01
8.21536899e-01 -5.17380536e-01 5.57333946e-01 9.06142712e-01
3.48911911e-01 -1.41475880e+00 4.75022227e-01 4.36604112e-01
4.25547808e-01 -5.73958695e-01 4.89174515e-01 5.29668808e-01
-4.79500055e-01 1.13491309e+00 6.37117624e-01 -1.03237830e-01
3.05722475e-01 3.54420006e-01 -3.21533494e-02 -2.03970313e-01
-1.21321273e+00 2.26084571e-02 -1.06533933e+00 1.12902653e+00
-1.63403943e-01 1.58340912e-02 -5.94627596e-02 9.40058008e-03
3.09358984e-02 3.76395974e-03 1.16771650e+00 9.43649292e-01
-9.89111423e-01 -9.80987549e-01 -3.73824894e-01 5.11385560e-01
-2.79428095e-01 -4.30209446e-04 2.12626392e-03 1.11265051e+00
-1.73606142e-01 9.67688322e-01 2.41797477e-01 -2.27816448e-01
1.86131299e-01 -4.24539112e-02 5.19935846e-01 -1.67502820e-01
-7.53538549e-01 1.89410672e-01 1.04978658e-01 -5.93117952e-01
-4.25634772e-01 -4.47027028e-01 -1.62513244e+00 -6.59211397e-01
-5.79888046e-01 5.29785275e-01 7.77309537e-01 4.38279718e-01
2.59137183e-01 6.40107453e-01 8.29551041e-01 -8.74776661e-01
-1.12060666e+00 -6.16433263e-01 -9.15481865e-01 -5.50073087e-01
2.24796757e-01 -4.23849165e-01 -3.78718436e-01 -3.45464617e-01]
|
[4.103999614715576, 2.2927353382110596]
|
18a2f57d-28d8-44bd-bee5-a41fa298e79e
|
towards-a-better-understanding-of-burrowss
| null | null |
https://aclanthology.org/W15-0709
|
https://aclanthology.org/W15-0709.pdf
|
Towards a better understanding of Burrows's Delta in literary authorship attribution
| null |
['Steffen Pielstr{\\"o}m', 'Fotis Jannidis', 'Christof Sch{\\"o}ch', 'Thorsten Vitt', 'Stefan Evert', 'Thomas Proisl']
|
2015-06-01
| null | null | null |
ws-2015-6
|
['text-clustering']
|
['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]
|
[-7.3436384201049805, 3.6421449184417725]
|
9f835edb-1795-4c6f-9f8e-707619c704b8
|
unsupervised-learning-of-monocular-depth-1
|
1803.03893
| null |
http://arxiv.org/abs/1803.03893v3
|
http://arxiv.org/pdf/1803.03893v3.pdf
|
Unsupervised Learning of Monocular Depth Estimation and Visual Odometry with Deep Feature Reconstruction
|
Despite learning based methods showing promising results in single view depth
estimation and visual odometry, most existing approaches treat the tasks in a
supervised manner. Recent approaches to single view depth estimation explore
the possibility of learning without full supervision via minimizing photometric
error. In this paper, we explore the use of stereo sequences for learning depth
and visual odometry. The use of stereo sequences enables the use of both
spatial (between left-right pairs) and temporal (forward backward) photometric
warp error, and constrains the scene depth and camera motion to be in a common,
real-world scale. At test time our framework is able to estimate single view
depth and two-view odometry from a monocular sequence. We also show how we can
improve on a standard photometric warp loss by considering a warp of deep
features. We show through extensive experiments that: (i) jointly training for
single view depth and visual odometry improves depth prediction because of the
additional constraint imposed on depths and achieves competitive results for
visual odometry; (ii) deep feature-based warping loss improves upon simple
photometric warp loss for both single view depth estimation and visual
odometry. Our method outperforms existing learning based methods on the KITTI
driving dataset in both tasks. The source code is available at
https://github.com/Huangying-Zhan/Depth-VO-Feat
|
['Chamara Saroj Weerasekera', 'Ravi Garg', 'Huangying Zhan', 'Harsh Agarwal', 'Kejie Li', 'Ian Reid']
|
2018-03-11
|
unsupervised-learning-of-monocular-depth-2
|
http://openaccess.thecvf.com/content_cvpr_2018/html/Zhan_Unsupervised_Learning_of_CVPR_2018_paper.html
|
http://openaccess.thecvf.com/content_cvpr_2018/papers/Zhan_Unsupervised_Learning_of_CVPR_2018_paper.pdf
|
cvpr-2018-6
|
['depth-and-camera-motion']
|
['computer-vision']
|
[-3.54404189e-02 4.22643796e-02 -3.10899705e-01 -5.13760448e-01
-7.60705292e-01 -5.47571540e-01 8.69545519e-01 -4.52835321e-01
-4.85797375e-01 7.00008392e-01 8.19166154e-02 -4.02378589e-02
3.31544340e-01 -6.65846765e-01 -9.05562460e-01 -6.81682408e-01
2.07225978e-01 5.68383396e-01 3.41580361e-01 -2.11720876e-02
3.58933359e-01 4.90497619e-01 -1.64081824e+00 -4.85171117e-02
4.16576773e-01 8.78804147e-01 3.98457587e-01 8.78851235e-01
3.33856493e-01 9.40180719e-01 5.59570529e-02 -1.52690098e-01
8.18772376e-01 -1.68789148e-01 -7.62630522e-01 2.19643325e-01
9.47628736e-01 -8.42663050e-01 -7.26836503e-01 8.73914659e-01
6.37725592e-01 2.69444585e-01 5.06269693e-01 -1.27550983e+00
3.40301283e-02 -3.09826493e-01 -5.82562149e-01 5.88458497e-03
7.11950302e-01 3.51446062e-01 1.01130533e+00 -9.42519128e-01
1.00348079e+00 1.10771120e+00 5.59555590e-01 5.22239327e-01
-1.30979443e+00 -4.70567822e-01 1.47735819e-01 3.21398437e-01
-1.12177265e+00 -4.86135811e-01 7.29111254e-01 -7.53633320e-01
1.31224453e+00 -2.76663095e-01 6.97500110e-01 9.18393850e-01
3.38026941e-01 7.04397500e-01 1.13655317e+00 -3.58477920e-01
1.96438413e-02 -6.46949038e-02 -4.17534590e-01 9.25036192e-01
-1.22440502e-01 6.10681236e-01 -7.12683737e-01 2.62559235e-01
8.62541139e-01 -6.65399479e-03 -4.65691119e-01 -1.01927876e+00
-1.25272512e+00 9.61077154e-01 5.51147878e-01 -3.43456179e-01
-6.52820244e-02 3.52135807e-01 3.01898569e-01 3.79961699e-01
5.75870514e-01 1.67941332e-01 -5.14073551e-01 -2.77352750e-01
-6.60337627e-01 2.83728898e-01 6.74645901e-01 9.04783189e-01
1.30847013e+00 9.59510857e-04 4.48002666e-01 5.81800640e-01
3.40571314e-01 6.17141724e-01 4.59152609e-01 -1.57136190e+00
6.90456271e-01 2.41137162e-01 2.61287075e-02 -5.23758829e-01
-4.63377535e-01 1.31525397e-01 -3.97842705e-01 8.92973006e-01
5.03371298e-01 -9.36825424e-02 -9.76746678e-01 1.69406128e+00
3.82477134e-01 1.42867520e-01 4.12865318e-02 9.47228432e-01
4.61686730e-01 4.31174725e-01 -6.36477888e-01 4.21810299e-02
9.47064638e-01 -1.07612741e+00 -2.81275302e-01 -7.37442195e-01
6.77106977e-01 -6.94954276e-01 7.74056196e-01 6.19751155e-01
-1.12749410e+00 -4.80654269e-01 -1.27294970e+00 -7.44416177e-01
-7.71502033e-02 -2.12663740e-01 4.72251713e-01 3.78310561e-01
-1.23965585e+00 5.54151833e-01 -1.03154194e+00 -3.46958667e-01
7.37710148e-02 4.02859837e-01 -6.83885634e-01 -4.50078189e-01
-9.52681065e-01 1.04215896e+00 2.01667815e-01 -2.21856892e-01
-1.14202619e+00 -6.89777315e-01 -1.42246032e+00 -5.20752966e-01
2.14188337e-01 -1.16904759e+00 1.18708599e+00 -7.95285881e-01
-1.72558939e+00 1.22590339e+00 -3.45219284e-01 -4.13312286e-01
9.46846962e-01 -4.10604686e-01 3.22894067e-01 2.84254253e-01
1.08772554e-01 1.11801887e+00 8.08012605e-01 -1.27778459e+00
-7.61760771e-01 -6.23811901e-01 3.35628748e-01 7.48860121e-01
1.64334908e-01 -7.29409993e-01 -4.24448371e-01 -5.84362783e-02
4.77050781e-01 -1.22899628e+00 -7.77619332e-02 5.05054116e-01
-2.27620766e-01 3.52896422e-01 7.01827228e-01 -4.88449872e-01
3.28507125e-01 -1.77452886e+00 4.87898946e-01 -2.02114314e-01
2.26440102e-01 -1.60225153e-01 9.02430639e-02 3.55349094e-01
-4.50468697e-02 -6.04762554e-01 -1.66750088e-01 -8.72010291e-01
-2.17104033e-01 4.89418089e-01 -1.30084231e-01 9.08250034e-01
-2.26606280e-01 6.79585874e-01 -9.45838749e-01 -2.36525029e-01
7.81654239e-01 5.95461249e-01 -7.86283314e-01 3.32974076e-01
-2.98151392e-02 7.47868180e-01 1.40554249e-01 4.69092101e-01
6.23805285e-01 8.64357799e-02 -5.57762794e-02 -9.66126174e-02
-2.57738411e-01 6.10216081e-01 -1.05214906e+00 2.23875427e+00
-7.85435259e-01 1.03491724e+00 -7.46443421e-02 -6.70222878e-01
8.01362813e-01 7.10051954e-02 4.80245024e-01 -7.76169837e-01
-4.42759767e-02 3.32768112e-01 -2.71379471e-01 -4.02865618e-01
4.34672445e-01 -4.51309919e-01 2.63071179e-01 1.93090022e-01
1.45272076e-01 -9.08306241e-01 -6.07642159e-02 2.14612167e-02
9.32449222e-01 6.25024617e-01 4.87505674e-01 -9.21042694e-04
5.57272077e-01 -2.08387226e-01 4.49768931e-01 2.24765033e-01
-2.62561500e-01 9.52322543e-01 3.89458209e-01 -4.05783057e-01
-1.38741398e+00 -1.36063981e+00 -1.65324971e-01 4.92532194e-01
2.80141473e-01 -3.37257311e-02 -2.48132333e-01 -4.65467781e-01
2.21278951e-01 3.83636981e-01 -5.18379807e-01 5.11949360e-02
-5.57460785e-01 -2.22506627e-01 2.16838777e-01 5.32536805e-01
4.97556448e-01 -5.03949344e-01 -7.73937106e-01 -1.13009073e-01
-3.11684102e-01 -1.45178533e+00 -4.24633145e-01 3.90094370e-01
-1.14881361e+00 -9.08960581e-01 -7.67401278e-01 -6.04017079e-01
2.85666168e-01 5.29545665e-01 9.04510736e-01 -3.95652771e-01
-1.11407079e-01 5.19539595e-01 5.02297916e-02 -1.23373941e-01
-7.26062581e-02 9.74065661e-02 1.33844987e-01 -3.08848441e-01
2.54803926e-01 -9.58694816e-01 -7.69729614e-01 3.33945721e-01
-5.97082853e-01 1.98878750e-01 1.24516658e-01 8.98746490e-01
4.90299702e-01 -7.66616404e-01 -2.83897460e-01 -5.80343544e-01
-3.16674620e-01 -2.00336367e-01 -9.86188591e-01 -5.06172955e-01
-7.37383962e-01 2.02343673e-01 2.42188245e-01 2.59330589e-02
-8.80453885e-01 4.12991792e-01 -2.17032805e-01 -8.44260752e-01
4.04618233e-02 1.09888744e-02 -3.20084095e-02 -3.37250948e-01
6.29270494e-01 2.81415254e-01 2.96755970e-01 -1.34318501e-01
2.21209005e-01 2.26110443e-01 3.99224877e-01 -2.32176140e-01
8.81972253e-01 1.24912024e+00 3.93749654e-01 -7.34974265e-01
-7.72871912e-01 -6.91891253e-01 -1.05218315e+00 -1.90361246e-01
1.01888072e+00 -1.54958630e+00 -7.45979190e-01 5.18514037e-01
-1.05514705e+00 -5.57866275e-01 -9.24648643e-02 8.10090661e-01
-1.17573905e+00 6.86434746e-01 -5.73721588e-01 -5.86537421e-01
1.61356345e-01 -1.26460934e+00 1.35018992e+00 -1.85631305e-01
-9.78372693e-02 -1.28987825e+00 4.14099604e-01 5.12399435e-01
-1.88721552e-01 3.81623626e-01 3.23224217e-01 1.88071132e-01
-1.00691438e+00 7.04857111e-02 -1.73126131e-01 5.07460058e-01
8.72226283e-02 -3.13615382e-01 -1.43214118e+00 -5.53312302e-01
1.08811595e-01 -5.51024079e-01 1.14511549e+00 4.66637045e-01
4.77287799e-01 2.31911227e-01 -1.48733808e-02 1.32908928e+00
1.80829906e+00 1.28904819e-01 8.25154066e-01 7.52967179e-01
1.07005036e+00 8.32981229e-01 6.13563180e-01 5.12364209e-01
7.29221225e-01 1.00894344e+00 8.54260683e-01 4.65497747e-02
-1.14610985e-01 -4.13866729e-01 5.49993277e-01 5.46025515e-01
-1.19547799e-01 -1.61757451e-02 -8.75389695e-01 8.22065294e-01
-1.71225989e+00 -8.69739234e-01 -6.91635208e-03 2.35126758e+00
5.67232847e-01 2.13100344e-01 1.13215698e-02 1.07185841e-01
7.00052232e-02 5.01533151e-01 -7.08079696e-01 -4.60944086e-01
-7.43643716e-02 -7.51320943e-02 1.00175071e+00 1.21149445e+00
-9.51468289e-01 7.28408813e-01 5.17927694e+00 1.04570277e-01
-1.16236913e+00 1.96630865e-01 2.85150230e-01 -5.74142456e-01
-3.01451683e-01 8.96303356e-02 -9.04495597e-01 2.00517222e-01
5.49006939e-01 2.21710235e-01 3.46121043e-01 8.08137000e-01
1.77602276e-01 -4.35604393e-01 -1.48497856e+00 1.38348055e+00
2.43749857e-01 -1.09845424e+00 -1.99928135e-01 4.15736318e-01
1.02299666e+00 6.83135986e-01 -5.25170565e-02 -1.32770479e-01
3.31901133e-01 -8.27255130e-01 7.49648511e-01 2.55696237e-01
7.93116629e-01 -6.36790454e-01 6.21019185e-01 4.48968530e-01
-1.04025638e+00 -4.40209806e-02 -3.20768178e-01 -4.66810048e-01
3.84681433e-01 4.54656810e-01 -7.53275692e-01 6.05084419e-01
5.54144263e-01 1.32557344e+00 -2.29832351e-01 6.76242232e-01
-4.69611079e-01 -2.07355261e-01 -3.75038952e-01 5.09602666e-01
4.21772629e-01 -3.21903646e-01 5.61744750e-01 7.51865387e-01
2.90535361e-01 -1.58307880e-01 8.45119432e-02 4.35803831e-01
1.94887429e-01 -3.49019945e-01 -1.07695782e+00 6.93939745e-01
6.63618445e-02 8.07134628e-01 -1.48147777e-01 -1.05969481e-01
-5.35291135e-01 1.26263094e+00 3.95677924e-01 2.81958938e-01
-4.97241795e-01 -4.86667939e-02 1.26920235e+00 3.08156908e-01
4.82247978e-01 -5.14867783e-01 -5.01767874e-01 -1.49166560e+00
2.47554153e-01 -3.47604603e-01 1.34429872e-01 -9.78322446e-01
-7.72971392e-01 2.97828823e-01 -1.03960361e-03 -1.57758188e+00
-7.26068735e-01 -7.95796812e-01 -2.95655280e-01 8.98741901e-01
-2.01022196e+00 -9.57339704e-01 -5.95811009e-01 6.89171195e-01
7.17150033e-01 8.76447037e-02 3.76071423e-01 2.55833417e-01
7.50749558e-02 1.83998898e-01 1.70713574e-01 -2.22356424e-01
1.00071526e+00 -1.39412546e+00 5.85540116e-01 7.04018414e-01
9.84525606e-02 3.97133306e-02 8.29885244e-01 -2.91821897e-01
-1.27976453e+00 -7.32292593e-01 9.78491485e-01 -7.51500189e-01
4.33117777e-01 -4.01295930e-01 -6.35890126e-01 1.01138771e+00
1.05975598e-01 1.63415506e-01 1.05782211e-01 -3.10414553e-01
-4.88155544e-01 -1.97578058e-01 -1.12134445e+00 3.69807243e-01
1.22434366e+00 -8.71902347e-01 -2.51722783e-01 2.74270743e-01
4.98742402e-01 -8.25423300e-01 -6.61174059e-01 2.22909853e-01
8.34876120e-01 -1.59253883e+00 1.15914524e+00 -2.80218329e-02
7.49882758e-01 -2.51596034e-01 -4.42751437e-01 -1.26323068e+00
8.27969760e-02 -4.69375879e-01 7.78034255e-02 5.20968378e-01
6.40117750e-02 -7.97984064e-01 1.02873504e+00 2.02658176e-01
-2.01619104e-01 -6.10897899e-01 -1.15434492e+00 -8.43705475e-01
1.52759314e-01 -5.05972028e-01 -1.34392545e-01 6.77268624e-01
-1.82410225e-01 2.87681520e-01 -5.62159777e-01 1.94994509e-01
8.82997334e-01 -5.71374185e-02 1.09861350e+00 -9.58850205e-01
-5.46325147e-01 -1.68487579e-01 -9.46418464e-01 -1.56148636e+00
1.77086234e-01 -6.79593265e-01 1.44292980e-01 -1.41242623e+00
-1.31654829e-01 1.16123311e-01 3.11560601e-01 6.33974820e-02
2.17237219e-01 4.25611198e-01 1.58880040e-01 2.26272777e-01
-3.40946615e-02 5.07523358e-01 1.25747478e+00 1.38354927e-01
1.64789185e-02 -1.61227196e-01 1.52236208e-01 8.36923182e-01
5.50178766e-01 -3.34353596e-01 -6.40965402e-01 -6.36629879e-01
2.99049228e-01 3.78417999e-01 6.12540066e-01 -9.62792695e-01
1.89850286e-01 -6.30917847e-02 2.51596481e-01 -5.36714911e-01
1.04751670e+00 -6.33143604e-01 -5.37926108e-02 6.79489851e-01
-1.05832852e-02 1.07858524e-01 5.09195551e-02 5.12433827e-01
-2.07073510e-01 -2.97975377e-03 1.15984929e+00 -2.97071010e-01
-1.04612529e+00 3.67304683e-01 -8.60704258e-02 1.24142326e-01
8.51646483e-01 -5.75014651e-01 -1.11247204e-01 -7.07446277e-01
-5.07134199e-01 1.70783326e-01 1.08345282e+00 3.17639202e-01
7.23081231e-01 -1.20738947e+00 -5.70471108e-01 3.94149691e-01
3.14019889e-01 2.59118855e-01 2.10857257e-01 9.82479155e-01
-9.54393685e-01 5.35803139e-01 -4.07278150e-01 -1.03615642e+00
-1.32655156e+00 3.37183684e-01 6.95706487e-01 -1.70623004e-01
-7.19684601e-01 9.56793904e-01 7.37022579e-01 -6.50891840e-01
2.00775102e-01 -3.01483810e-01 3.18624288e-01 -7.17615187e-02
3.53649527e-01 4.59480017e-01 7.86478445e-02 -6.53785348e-01
-2.98590541e-01 1.05042470e+00 6.00385442e-02 -4.67654377e-01
1.21628308e+00 -6.57556117e-01 2.37919480e-01 7.37562954e-01
1.81217468e+00 -1.62752643e-01 -2.06149101e+00 -3.02249432e-01
-3.12757343e-01 -7.88083375e-01 1.63720876e-01 -3.35172802e-01
-9.71782744e-01 1.27573943e+00 6.86867893e-01 -3.97551566e-01
9.59880829e-01 -1.16414629e-01 7.39399552e-01 3.15303653e-01
5.10330200e-01 -7.88627326e-01 1.84880018e-01 7.39931285e-01
6.24498844e-01 -1.73298514e+00 2.12036923e-01 -2.71729827e-01
-4.80347335e-01 1.17660153e+00 5.60856879e-01 -3.25820744e-01
5.05662322e-01 2.52089471e-01 2.34643266e-01 -1.10794129e-02
-8.77772808e-01 -2.89198339e-01 1.50243551e-01 7.66332626e-01
2.08517388e-01 -2.47215539e-01 2.21475229e-01 -7.10156560e-01
-2.59215951e-01 -1.22752734e-01 7.40966260e-01 7.94503748e-01
-5.08545458e-01 -1.12325227e+00 -1.68677136e-01 -8.15619975e-02
-5.15468381e-02 -3.99000607e-02 -1.61155328e-01 9.58576143e-01
4.88109775e-02 5.24716735e-01 1.71417087e-01 -3.15565169e-01
2.40174651e-01 -1.13609053e-01 9.71409976e-01 -6.68512881e-01
-4.94052134e-02 -9.20370221e-02 1.66764811e-01 -9.39500988e-01
-6.23230219e-01 -9.70485866e-01 -1.10474396e+00 -4.60565358e-01
1.61353275e-01 -6.04282320e-01 7.74979711e-01 9.81072485e-01
-1.54500008e-01 6.89716190e-02 7.27686942e-01 -1.37186837e+00
-4.35205638e-01 -5.63290954e-01 -6.36108160e-01 3.22198689e-01
9.43630755e-01 -8.59897494e-01 -7.42785692e-01 6.40206486e-02]
|
[8.561073303222656, -2.3413827419281006]
|
67cb7c2a-4eaf-40ef-94c2-9336c135abd9
|
pretraining-for-conditional-generation-with
| null | null |
https://openreview.net/forum?id=H1eFXO0WpV
|
https://openreview.net/pdf?id=H1eFXO0WpV
|
Pretraining for Conditional Generation with Pseudo Self Attention
|
Large pretrained language representation models have changed the way researchers approach discriminative natural language understanding tasks, leading to the dominance of approaches that finetune a pretrained model. However, such transfer learning approaches have not seen the same success for natural language generation. In this work, we explore transfer learning for conditional generation with large pretrained language models. We propose a simple modification to a pretrained unconditional transformer model to inject arbitrary conditioning into the self attention layer, an approach we call pseudo self attention. Through experiments on four long-form conditional text generation tasks, we show that this technique outperforms strong baselines and other transfer learning approaches, and produces coherent generations.
|
['Anonymous']
|
2019-05-21
| null | null | null | null |
['conditional-text-generation']
|
['natural-language-processing']
|
[ 4.15567845e-01 5.73782504e-01 -2.67641693e-01 -5.17527461e-01
-1.16599405e+00 -6.85525954e-01 1.13290977e+00 -1.89231932e-01
-4.24969226e-01 1.18582940e+00 7.37571061e-01 -5.07493377e-01
5.84800839e-01 -1.00864804e+00 -9.19898093e-01 -3.80432576e-01
4.70249623e-01 8.88292313e-01 7.81489983e-02 -3.91736478e-01
9.58440825e-02 -1.05683558e-01 -8.93487155e-01 5.28202176e-01
8.62939775e-01 3.01010013e-01 7.49578476e-02 7.67831385e-01
-2.58968294e-01 9.55750585e-01 -7.49506116e-01 -6.86540306e-01
-1.63044840e-01 -8.21380138e-01 -1.12876081e+00 -3.25142086e-01
3.51875931e-01 -4.14158851e-01 -6.32922053e-02 6.39608920e-01
4.04803157e-01 1.99660242e-01 9.32752907e-01 -8.75065446e-01
-1.30433083e+00 1.32497573e+00 -5.79168379e-01 3.08862664e-02
4.02239174e-01 6.20229021e-02 1.13939393e+00 -7.62933195e-01
6.55847251e-01 1.75644362e+00 5.63905358e-01 9.82244313e-01
-1.67216587e+00 -7.16926575e-01 2.19888657e-01 -1.93935856e-01
-1.04709518e+00 -3.22646260e-01 6.71613455e-01 -2.79847234e-01
1.23655081e+00 -4.00755644e-01 2.72300512e-01 1.51754189e+00
1.61921561e-01 9.45119560e-01 1.08750105e+00 -8.47915351e-01
-8.10292214e-02 1.37678325e-01 8.61656144e-02 6.71775579e-01
2.74272531e-01 2.48324826e-01 -3.87869179e-01 -1.70012936e-01
6.72268450e-01 -4.28050935e-01 -7.81680718e-02 -3.28257866e-02
-1.18414080e+00 1.38441563e+00 4.62577999e-01 5.03341019e-01
-9.75003093e-03 6.26849174e-01 3.05621982e-01 3.74971896e-01
7.85652578e-01 6.47002220e-01 -5.48149943e-01 -8.86349306e-02
-7.21971095e-01 3.32397938e-01 6.38242245e-01 9.21499312e-01
9.51108336e-01 3.75717789e-01 -5.74934959e-01 7.76521862e-01
2.01011777e-01 3.47046137e-01 8.10618341e-01 -5.15608668e-01
4.22813147e-01 4.08912778e-01 -1.91698074e-01 -3.39208245e-01
1.32621199e-01 -1.61552519e-01 -7.09294200e-01 -1.64782614e-01
2.74736017e-01 -6.29513741e-01 -1.27781832e+00 2.18725538e+00
-1.67028606e-01 1.31608933e-01 3.77369761e-01 1.75515980e-01
3.97884995e-01 1.05780196e+00 4.29123729e-01 1.14540242e-01
8.77480447e-01 -9.87342954e-01 -5.07015586e-01 -4.18034434e-01
6.46496475e-01 -7.89956510e-01 1.26973033e+00 2.50203818e-01
-1.16751933e+00 -6.68638408e-01 -8.77181351e-01 -4.31408614e-01
-5.41972756e-01 8.80314037e-03 9.83642340e-01 7.68601000e-01
-1.24373603e+00 5.50521374e-01 -8.06115508e-01 -2.63357759e-01
4.55281854e-01 2.30341554e-01 -1.37008280e-01 -1.98196128e-01
-1.37577558e+00 9.81159210e-01 8.21672022e-01 -2.16993913e-01
-1.10259175e+00 -9.64448035e-01 -9.05162334e-01 -7.30632395e-02
3.88222560e-02 -1.12847698e+00 1.55742431e+00 -1.11414003e+00
-1.77663422e+00 8.25371623e-01 -2.35873148e-01 -8.23004842e-01
2.81508237e-01 -5.19497931e-01 -6.14261031e-02 -2.89125383e-01
1.63629800e-01 1.28514910e+00 9.42843378e-01 -1.24108005e+00
-4.18378413e-01 2.95397080e-02 9.01430622e-02 -5.96532896e-02
-3.14903378e-01 4.69488166e-02 5.10224700e-03 -9.39700782e-01
-7.49149859e-01 -7.88098395e-01 -4.53529596e-01 -5.97870469e-01
-3.18471491e-01 -7.28042603e-01 5.66193998e-01 -4.13087755e-01
9.82212543e-01 -1.71509278e+00 3.64965677e-01 -1.98602542e-01
-1.72762513e-01 2.70637929e-01 -4.36772794e-01 5.39783597e-01
-1.59434259e-01 4.55275565e-01 -4.66639429e-01 -5.88616073e-01
2.31461883e-01 3.07767004e-01 -1.11295962e+00 -2.77727187e-01
6.27616107e-01 1.38735402e+00 -1.01741838e+00 -3.59333038e-01
-1.71495542e-01 6.69834554e-01 -9.44086373e-01 4.63039219e-01
-7.58065343e-01 1.88279524e-01 -3.58889639e-01 1.45878926e-01
2.57094085e-01 -2.66172886e-01 1.03510089e-01 1.98756903e-01
1.83969125e-01 7.58727014e-01 -4.83224690e-01 1.93794262e+00
-6.93830431e-01 5.44791579e-01 -4.71447915e-01 -8.83122623e-01
8.80245090e-01 5.05729198e-01 -1.53220609e-01 -3.06710631e-01
-1.64568648e-02 7.32894316e-02 1.77412387e-02 1.55952021e-01
7.25831509e-01 -7.75335848e-01 -3.47188830e-01 7.87184298e-01
6.01020455e-01 -6.73056722e-01 2.70386100e-01 6.59071028e-01
7.66286314e-01 6.06109977e-01 1.56727135e-01 -1.61074549e-01
2.54098088e-01 -1.29838571e-01 1.21557526e-01 7.87553966e-01
3.36785913e-01 6.94582283e-01 3.90140295e-01 -1.72344849e-01
-8.67092967e-01 -1.35360742e+00 2.09019557e-01 1.40574861e+00
-4.56428111e-01 -5.97060978e-01 -9.26404774e-01 -7.72927105e-01
-1.31694794e-01 1.12006283e+00 -8.65361691e-01 -4.63300526e-01
-8.27098489e-01 -8.68629456e-01 9.10910726e-01 8.81162226e-01
2.60166049e-01 -1.39765489e+00 -1.37173980e-01 2.48503864e-01
-2.74363071e-01 -8.87830377e-01 -5.50813496e-01 2.24438816e-01
-8.97237659e-01 -4.00922358e-01 -9.97226596e-01 -1.04293585e+00
5.69034457e-01 -3.51955831e-01 1.64472198e+00 -6.51107877e-02
-1.96586356e-01 4.24575359e-01 -3.15293759e-01 -4.92515206e-01
-7.64125586e-01 5.89094102e-01 -2.36082494e-01 -1.98810637e-01
4.17832106e-01 -5.24426103e-01 -5.90647310e-02 -3.91754717e-01
-1.01731837e+00 1.55364037e-01 6.10161185e-01 1.07527387e+00
2.86591321e-01 -4.73323345e-01 9.00835514e-01 -1.07511890e+00
1.10101891e+00 -3.64328265e-01 -3.15343559e-01 3.49896163e-01
-6.38173282e-01 6.55042350e-01 6.73088431e-01 -4.17891443e-01
-1.53346574e+00 -2.11683646e-01 -1.17851354e-01 -1.52923331e-01
-2.17315689e-01 7.36130774e-01 2.88990345e-02 5.17344117e-01
6.36297882e-01 2.80823708e-01 -3.38000685e-01 -4.10041749e-01
8.88232946e-01 2.45170832e-01 5.03527761e-01 -9.02445376e-01
1.01832700e+00 1.54181063e-01 -4.93419141e-01 -5.16721010e-01
-1.10047734e+00 1.16514444e-01 -5.98581612e-01 4.07948166e-01
1.21188235e+00 -1.07058752e+00 1.21993467e-01 2.67959595e-01
-1.48328233e+00 -9.57343221e-01 -5.83489358e-01 3.05939496e-01
-7.36656487e-01 1.53944597e-01 -8.84053409e-01 -5.30998826e-01
-4.62419271e-01 -6.99427783e-01 1.20393407e+00 9.67190862e-02
-4.39598590e-01 -1.35097861e+00 6.80398405e-01 1.00657918e-01
7.12451994e-01 -3.56882066e-02 1.06528103e+00 -4.47571933e-01
-6.00668252e-01 2.02782918e-02 -6.29637316e-02 3.40232402e-01
2.51397520e-01 -5.67804165e-02 -9.52164531e-01 -2.62141913e-01
-4.29724693e-01 -1.00007033e+00 1.38945508e+00 2.68882841e-01
8.40865433e-01 -2.53140777e-01 -4.97920364e-01 5.17381251e-01
1.12208366e+00 -8.85539055e-02 8.19787681e-01 -2.32326120e-01
6.60069942e-01 4.47856218e-01 1.53672740e-01 -2.27440782e-02
3.02397072e-01 2.97337472e-01 -2.35267580e-01 -2.84913659e-01
-4.44223017e-01 -9.00537670e-01 7.93269157e-01 7.58338571e-01
-1.55280605e-01 -6.87986314e-01 -7.22331762e-01 7.30270982e-01
-1.51705372e+00 -1.12457693e+00 4.44036514e-01 1.77013993e+00
1.33348095e+00 1.84910834e-01 -2.16121390e-01 -3.79009992e-01
5.29099584e-01 1.81326166e-01 -2.47146726e-01 -7.60488153e-01
-1.23476259e-01 1.00952864e+00 5.69537431e-02 8.84222925e-01
-1.00407135e+00 1.63934481e+00 7.11256933e+00 7.99820960e-01
-1.03668714e+00 1.40822724e-01 7.37821817e-01 2.41212219e-01
-8.90031338e-01 1.64223850e-01 -9.83665943e-01 3.32168080e-02
1.24169457e+00 -5.03381193e-01 3.05035532e-01 7.32929409e-01
-2.46101901e-01 2.54151046e-01 -1.41053772e+00 4.77146626e-01
3.68529767e-01 -1.28970683e+00 7.84964144e-01 -1.00342467e-01
1.07187629e+00 1.27152711e-01 1.10952169e-01 8.66772830e-01
1.26520157e+00 -1.33868039e+00 3.75739187e-01 3.14609230e-01
7.97460139e-01 -7.62119651e-01 2.88427055e-01 9.71657559e-02
-9.17870402e-01 2.51120359e-01 -4.11427349e-01 -1.03196599e-01
5.63399494e-01 2.90445805e-01 -1.08274651e+00 2.61977494e-01
2.37584546e-01 6.28314316e-01 -6.34651840e-01 3.90287340e-01
-6.70188189e-01 1.05995405e+00 -1.00499419e-02 -1.30438313e-01
3.49199802e-01 -8.49633366e-02 1.34616777e-01 1.46456528e+00
2.78971493e-01 -1.35567546e-01 1.08390294e-01 1.33950758e+00
-5.50058663e-01 -8.47749971e-03 -7.90185571e-01 -4.19029832e-01
-1.25071526e-01 9.86276805e-01 -3.91347677e-01 -7.28384495e-01
-3.23745251e-01 1.25128567e+00 6.44280672e-01 4.61124390e-01
-8.39721262e-01 -3.76588076e-01 4.16346669e-01 -5.63848810e-03
4.85007882e-01 -3.20211291e-01 -4.77893502e-02 -1.33367252e+00
-3.86026442e-01 -8.17442894e-01 4.35862690e-01 -8.34506750e-01
-1.53435481e+00 7.57505000e-01 1.69357181e-01 -6.64444566e-01
-9.52383816e-01 -5.18267572e-01 -9.89163935e-01 1.03883231e+00
-1.59718406e+00 -1.51014233e+00 1.92180112e-01 6.50695086e-01
6.69221401e-01 -2.24269092e-01 1.26174581e+00 -7.66822323e-02
-2.25481465e-01 7.35016584e-01 -2.54222065e-01 2.52910107e-01
1.03482425e+00 -1.47581160e+00 8.52210164e-01 9.04692113e-01
4.46776688e-01 1.01177883e+00 4.77480263e-01 -6.90264702e-01
-1.00257099e+00 -1.22555506e+00 1.16804838e+00 -6.82611883e-01
7.06300318e-01 -6.40739739e-01 -7.68325329e-01 1.33558798e+00
1.06687903e+00 -3.68987799e-01 7.58908927e-01 1.67255029e-01
-6.30583525e-01 2.53880829e-01 -6.80027068e-01 7.10239232e-01
7.09876180e-01 -5.67749560e-01 -1.00062311e+00 2.36948788e-01
1.28434718e+00 7.76467696e-02 -5.55979729e-01 3.74895096e-01
2.49698207e-01 -3.71563762e-01 8.69467616e-01 -1.04962289e+00
7.98373222e-01 4.18630354e-02 4.05472629e-02 -1.65609741e+00
-3.27249140e-01 -7.82942891e-01 -2.72471718e-02 1.51512694e+00
9.62052882e-01 -5.40655553e-01 7.33234882e-01 3.55480075e-01
-2.26298153e-01 -2.78329045e-01 -5.93040943e-01 -6.59149766e-01
1.09660697e+00 -5.96535355e-02 3.86253625e-01 7.65535772e-01
3.74003388e-02 1.09151781e+00 -4.33124244e-01 -4.61010963e-01
2.57601500e-01 2.69346297e-01 8.59219134e-01 -8.83200705e-01
-7.38464236e-01 -5.00639915e-01 1.85559303e-01 -1.34472144e+00
6.11454010e-01 -1.19121850e+00 2.86654204e-01 -1.64720201e+00
3.77129704e-01 -2.06715345e-01 -1.40943006e-01 5.79036355e-01
-5.24890065e-01 4.09859240e-01 -5.17892558e-03 -3.04578096e-01
-3.91116679e-01 8.46024930e-01 1.00596905e+00 -3.97209853e-01
-1.35208875e-01 -2.92323738e-01 -9.76829767e-01 5.15679121e-01
1.03198957e+00 -3.82449329e-01 -6.68881476e-01 -8.61963153e-01
2.54930794e-01 -3.60796243e-01 2.13986188e-01 -7.76504874e-01
-1.83588058e-01 -7.15586245e-02 4.13371116e-01 -3.76650363e-01
2.73918241e-01 -5.40207587e-02 -3.42836082e-01 3.61487538e-01
-9.02944624e-01 7.27967098e-02 3.93558443e-01 4.60603148e-01
-3.36730152e-01 -1.87601998e-01 6.37976348e-01 -4.03906047e-01
-3.82974058e-01 2.83933580e-01 -5.95211506e-01 4.17522699e-01
7.28665650e-01 2.10662007e-01 -2.99308062e-01 -5.98958790e-01
-6.07711852e-01 -1.11010727e-02 2.28458956e-01 6.17450595e-01
5.50903320e-01 -1.23098385e+00 -1.10981226e+00 1.02321334e-01
-2.47457568e-02 -1.34630278e-01 -2.25053519e-01 1.47937194e-01
-1.78030998e-01 6.12101376e-01 2.81619728e-02 -1.93318024e-01
-6.50323212e-01 7.39191830e-01 2.58679390e-01 -7.32180178e-01
-2.85316348e-01 1.05513716e+00 5.54493070e-01 -6.76989913e-01
-7.50317797e-02 -5.29204071e-01 6.20874204e-02 -1.18243337e-01
5.95126629e-01 -1.65594280e-01 -2.32460573e-01 -3.41295481e-01
4.24132235e-02 3.35414141e-01 -4.34130639e-01 -5.30560791e-01
1.24867475e+00 1.41426176e-01 -1.62396684e-01 5.33218980e-01
1.11733329e+00 1.63492620e-01 -1.01374257e+00 -1.13211177e-01
3.53597291e-02 1.16891794e-01 -1.83924958e-01 -9.36194897e-01
-7.25094080e-01 1.22673798e+00 1.23049244e-01 7.33470370e-04
7.99840987e-01 1.40470371e-01 8.79971802e-01 5.54352045e-01
2.77863413e-01 -7.17816830e-01 4.41157013e-01 8.88241351e-01
1.00795186e+00 -1.11550117e+00 -3.43591124e-01 -1.50122404e-01
-6.41896248e-01 9.18442726e-01 8.92900169e-01 -3.41437727e-01
6.10301197e-01 4.12960917e-01 1.40966922e-02 2.30096340e-01
-1.15571451e+00 -1.79655492e-01 1.12220593e-01 7.07307041e-01
1.11001098e+00 8.58464092e-02 -2.57683545e-01 5.23910940e-01
-5.90686440e-01 1.37653172e-01 4.40738052e-01 6.18548393e-01
-2.84615904e-01 -1.68606567e+00 -9.09986347e-02 2.08458155e-01
-5.44354677e-01 -6.98155224e-01 -7.74050355e-01 6.21183038e-01
-4.21312414e-02 8.52280140e-01 2.14655697e-01 -4.52670865e-02
-9.79387313e-02 5.24716377e-01 7.98872769e-01 -1.10539877e+00
-6.52493596e-01 -1.63228158e-02 1.05697259e-01 -8.00314844e-02
-3.35459054e-01 -3.78526449e-01 -1.21209836e+00 1.52087426e-02
-3.23542446e-01 4.64786261e-01 9.61038247e-02 7.62266099e-01
2.98456073e-01 5.47668338e-01 2.85214931e-01 -8.56798828e-01
-5.66949785e-01 -1.30106950e+00 -1.29388645e-02 3.96011710e-01
2.81636953e-01 -1.97841138e-01 -8.44076872e-02 5.71758926e-01]
|
[11.612360954284668, 9.020149230957031]
|
0ea3f7ee-128c-42c5-9e08-6f72ac46640e
|
cgt-clustered-graph-transformer-for-urban
| null | null |
https://openreview.net/forum?id=H1eJAANtvr
|
https://openreview.net/pdf?id=H1eJAANtvr
|
CGT: Clustered Graph Transformer for Urban Spatio-temporal Prediction
|
Deep learning based approaches have been widely used in various urban spatio-temporal forecasting problems, but most of them fail to account for the unsmoothness issue of urban data in their architecture design, which significantly deteriorates their prediction performance. The aim of this paper is to develop a novel clustered graph transformer framework that integrates both graph attention network and transformer under an encoder-decoder architecture to address such unsmoothness issue. Specifically, we propose two novel structural components to refine the architectures of those existing deep learning models. In spatial domain, we propose a gradient-based clustering method to distribute different feature extractors to regions in different contexts. In temporal domain, we propose to use multi-view position encoding to address the periodicity and closeness of urban time series data. Experiments on real datasets obtained from a ride-hailing business show that our method can achieve 10\%-25\% improvement than many state-of-the-art baselines.
|
['Jieping Ye', 'Hongtu Zhu', 'Qiang Yang', 'Leye Wang', 'Lulu Zhang', 'Yuanbo Zhang', 'Shulin Li', 'Lingyu Zhang', 'Xu Geng']
|
2019-09-25
| null | null | null | null |
['spatio-temporal-forecasting']
|
['time-series']
|
[-2.38863334e-01 -1.70247570e-01 -2.53884315e-01 -5.17414033e-01
-5.46963811e-01 -1.03031926e-01 6.82576656e-01 -2.25276694e-01
1.31663769e-01 4.03580576e-01 5.98613977e-01 -4.67928886e-01
-2.08811253e-01 -1.08661425e+00 -7.44978905e-01 -6.11936867e-01
-2.43046701e-01 3.43057096e-01 3.90304416e-01 -5.55024087e-01
9.08953995e-02 2.17995539e-01 -1.40828145e+00 3.96448314e-01
9.74463582e-01 9.08146501e-01 3.77042741e-01 4.33453560e-01
-1.67324692e-02 1.20747995e+00 -5.34506887e-02 -2.13110954e-01
1.20701171e-01 -3.18653405e-01 -5.24005413e-01 2.64675111e-01
4.27694917e-01 -2.83707589e-01 -8.24186981e-01 7.59549081e-01
4.82572526e-01 2.39578798e-01 6.68555558e-01 -1.31886363e+00
-9.38111782e-01 4.33485210e-01 -7.23596811e-01 4.12156105e-01
-1.38731226e-01 -8.45079198e-02 1.20385981e+00 -8.31841171e-01
3.13860267e-01 1.21516919e+00 1.02923989e+00 5.52893952e-02
-9.11407650e-01 -6.39710367e-01 5.19417882e-01 6.27152145e-01
-1.60692406e+00 -3.55151325e-01 1.08735502e+00 -5.06021976e-01
1.19607687e+00 -4.31653336e-02 7.24012136e-01 7.59205699e-01
2.72869617e-01 8.71251702e-01 6.78477228e-01 1.98131934e-01
-1.84293881e-01 -2.95367390e-01 -2.12119654e-01 7.33389914e-01
-2.25153342e-01 2.85465666e-03 -2.59071082e-01 2.94598639e-01
5.61955690e-01 3.32415462e-01 -7.65176937e-02 -3.82271171e-01
-9.77436662e-01 1.10691845e+00 1.18373990e+00 3.85493219e-01
-3.37925732e-01 4.22402173e-01 4.04409111e-01 1.44168481e-01
1.00240886e+00 -1.22868694e-01 -2.56433278e-01 1.36890054e-01
-9.48040366e-01 2.49455243e-01 2.85635352e-01 1.05455744e+00
7.17795908e-01 2.85605252e-01 -4.86607440e-02 8.04496229e-01
4.08369958e-01 4.02680457e-01 3.59005243e-01 -4.51885670e-01
9.33720112e-01 6.29289329e-01 -3.43374670e-01 -1.66820121e+00
-6.79802299e-01 -6.43552482e-01 -1.20942938e+00 -3.53493929e-01
-1.02258086e-01 -7.15521351e-02 -9.89519835e-01 1.66200149e+00
2.44198635e-01 7.28906274e-01 -2.11242393e-01 8.77624035e-01
9.14058387e-01 9.20749664e-01 -1.02860957e-01 1.27098218e-01
8.60955060e-01 -1.29279280e+00 -5.69979072e-01 -2.15280786e-01
6.94836259e-01 -4.67961997e-01 1.11415577e+00 8.87816586e-03
-7.09710896e-01 -6.44011915e-01 -1.01069605e+00 -1.55058861e-01
-6.97048843e-01 5.19754477e-02 5.45385361e-01 2.79923081e-01
-1.30788028e+00 5.12728810e-01 -7.70765543e-01 -4.79959995e-01
4.19680595e-01 6.28851727e-02 -1.14740372e-01 -1.41729951e-01
-1.28895795e+00 6.14136279e-01 2.19111711e-01 2.79738635e-01
-5.85784018e-01 -5.50269842e-01 -9.70310807e-01 2.14594901e-01
2.49323875e-01 -5.38295686e-01 8.34092200e-01 -6.87287807e-01
-1.19079745e+00 4.88438308e-01 -2.39952013e-01 -6.72415912e-01
2.06176952e-01 -1.24885201e-01 -7.50239551e-01 -2.11913943e-01
4.16496247e-01 4.49065596e-01 6.70263588e-01 -1.09167421e+00
-7.10842609e-01 -2.83105314e-01 -1.47527486e-01 3.56856912e-01
-2.98583508e-01 -3.24266255e-01 -5.94722867e-01 -8.82794976e-01
1.02830820e-01 -8.63715231e-01 -4.38712090e-01 -4.07215565e-01
-3.06259960e-01 -3.38649064e-01 1.16623783e+00 -7.25181758e-01
1.62064028e+00 -2.06132770e+00 -1.53802391e-02 1.66161507e-01
4.68373269e-01 1.65815860e-01 -1.89955935e-01 6.98046327e-01
-6.59290552e-02 6.63721412e-02 -1.82542026e-01 -3.48807365e-01
9.82058719e-02 3.13924432e-01 -5.00595808e-01 5.47969222e-01
1.75203472e-01 1.15006256e+00 -9.59086418e-01 -4.31264877e-01
4.45578039e-01 6.03211582e-01 -6.05365515e-01 2.95459032e-02
-4.21930701e-02 3.55753988e-01 -4.56854075e-01 3.20850134e-01
7.22539485e-01 -5.33568680e-01 1.85179010e-01 -1.28924310e-01
-1.00178741e-01 6.25222266e-01 -8.96190822e-01 1.47405624e+00
-5.70776403e-01 8.46371651e-01 -1.92139819e-01 -1.40083182e+00
7.94609904e-01 8.04278404e-02 7.52124846e-01 -1.15027666e+00
-2.32048810e-01 -1.11980364e-02 -1.06353588e-01 -4.47413266e-01
7.42259443e-01 2.64539365e-02 -1.02596484e-01 6.54368699e-02
-2.87189186e-01 6.27400875e-02 -5.73944822e-02 1.47071689e-01
1.11559582e+00 -7.91521445e-02 -1.06200920e-02 -3.08687687e-01
4.49697137e-01 -2.18414646e-02 5.93493342e-01 3.71300608e-01
-2.11638018e-01 8.10828388e-01 3.45445067e-01 -8.69340003e-01
-9.40211415e-01 -1.03533685e+00 1.48832962e-01 1.17202008e+00
6.70250803e-02 -4.82086748e-01 -3.88059497e-01 -7.74802268e-01
-1.94589049e-01 6.36019945e-01 -5.97510278e-01 1.49270166e-02
-8.71710896e-01 -9.38763857e-01 4.48362350e-01 6.45984709e-01
8.26485455e-01 -6.24271631e-01 -3.56636167e-01 5.00144660e-01
-5.48686624e-01 -1.12340188e+00 -5.70287883e-01 -3.65710780e-02
-6.98314071e-01 -9.03036296e-01 -5.65760851e-01 -8.44696224e-01
2.00450435e-01 8.48551035e-01 1.35865426e+00 7.51473084e-02
3.02037776e-01 -7.11926818e-02 -5.52507877e-01 -1.95227321e-02
2.46860072e-01 5.24265528e-01 -3.49574417e-01 2.36325234e-01
3.91144663e-01 -8.52105975e-01 -8.02921891e-01 4.52754050e-01
-5.66662073e-01 2.00769499e-01 3.50885272e-01 7.71648347e-01
4.70014572e-01 3.92155379e-01 7.72746027e-01 -8.99289072e-01
5.38285851e-01 -9.65776980e-01 -4.49883461e-01 3.66802700e-02
-8.25886488e-01 -9.29409266e-02 8.41678381e-01 1.74236053e-03
-6.46131158e-01 -1.17929205e-02 -2.50528216e-01 -5.08106291e-01
2.07609776e-02 9.11619604e-01 1.02537069e-02 2.27493480e-01
1.88031882e-01 3.48295301e-01 -4.16741431e-01 -4.23597097e-01
4.45132554e-01 5.02327800e-01 3.20266038e-01 -5.93829378e-02
7.92085171e-01 5.24480402e-01 -1.37707487e-01 -7.80494809e-01
-7.40414321e-01 -6.30314529e-01 -5.73242128e-01 -3.18949282e-01
9.19945121e-01 -1.34074152e+00 -2.49093726e-01 3.88105333e-01
-9.34337378e-01 -6.25886142e-01 1.42437026e-01 3.22567314e-01
-5.48509300e-01 1.82690561e-01 -4.62532729e-01 -5.39252222e-01
-1.82499781e-01 -1.10629451e+00 1.35365844e+00 -1.61204264e-01
3.37208420e-01 -1.14754128e+00 7.11405873e-02 3.20994020e-01
7.99624562e-01 5.69221795e-01 8.78810406e-01 -2.67921418e-01
-6.11472428e-01 2.14924337e-03 -5.71485221e-01 -1.14662252e-01
6.02387451e-02 -1.41364276e-01 -7.95095444e-01 -2.47363478e-01
-2.92539746e-01 1.03947766e-01 1.24907684e+00 5.50945222e-01
1.31859875e+00 -6.86169386e-01 -5.37141740e-01 7.94579566e-01
1.49001336e+00 7.32108429e-02 8.29299390e-01 7.57985786e-02
1.27186942e+00 5.32561839e-01 3.37522447e-01 3.28824997e-01
1.25548995e+00 8.05061042e-01 7.08684444e-01 -2.25509763e-01
-3.28470051e-01 -4.39453691e-01 3.90561044e-01 1.19870782e+00
-6.03878573e-02 -6.57326698e-01 -1.15068173e+00 1.12226951e+00
-2.25943732e+00 -1.31291997e+00 -4.20570642e-01 1.75030899e+00
2.75654763e-01 1.84578016e-01 3.12597185e-01 5.41595630e-02
5.76706409e-01 8.74906063e-01 -3.46086264e-01 -1.87613249e-01
4.66306321e-02 -1.26496613e-01 8.66829515e-01 4.92083102e-01
-1.42853034e+00 1.12001252e+00 5.67863607e+00 1.01415634e+00
-1.40712476e+00 2.44773388e-01 6.50240242e-01 7.66798621e-04
-3.57660532e-01 -2.58100808e-01 -4.54370677e-01 5.57609558e-01
1.17064059e+00 4.46386449e-02 5.01744747e-01 9.01183903e-01
4.63229120e-01 3.66188914e-01 -7.00489283e-01 8.93486142e-01
5.53824008e-02 -1.55117524e+00 -9.62891653e-02 1.83223113e-01
8.46420348e-01 8.37063730e-01 3.00465763e-01 5.50497115e-01
4.93822813e-01 -1.20140243e+00 6.95591271e-01 4.65580374e-01
4.70922768e-01 -8.28640938e-01 6.33978963e-01 3.48195970e-01
-2.10426545e+00 -4.14363928e-02 -3.77054483e-01 -1.99100837e-01
2.77996778e-01 6.10447288e-01 -9.28208828e-01 7.63460100e-01
8.86045337e-01 1.42553234e+00 -6.36115968e-01 8.52802515e-01
3.84767503e-02 7.72652507e-01 -2.32136324e-01 5.32349199e-02
8.27629268e-01 -1.81102380e-01 2.91816860e-01 1.18782854e+00
5.11474907e-01 -4.05906290e-02 4.07143950e-01 6.80756569e-01
-4.84060273e-02 -1.16075678e-02 -1.14645529e+00 1.71389163e-01
3.23340356e-01 8.35634232e-01 -5.80783606e-01 -3.91180396e-01
-7.23222494e-01 7.72626936e-01 5.23575246e-01 5.44402421e-01
-1.09883726e+00 -2.57276058e-01 5.50967276e-01 5.50529897e-01
7.80128419e-01 -5.45548975e-01 -3.59258533e-01 -1.10056603e+00
-8.58750120e-02 -6.47978008e-01 4.25038695e-01 -7.13424265e-01
-1.34115982e+00 7.04453468e-01 1.78939831e-02 -1.52025890e+00
-3.12811017e-01 -1.40867010e-01 -8.72711957e-01 5.23217261e-01
-1.85425580e+00 -1.72874665e+00 -2.33811378e-01 7.45081604e-01
7.89158940e-01 -1.90508723e-01 2.51125872e-01 8.58592510e-01
-5.74413955e-01 4.72770959e-01 2.11773172e-01 2.68561125e-01
3.29567730e-01 -1.19939327e+00 8.74984801e-01 1.11572242e+00
2.67180383e-01 1.37929723e-01 4.05713797e-01 -5.21769226e-01
-1.16921616e+00 -1.91773951e+00 1.28055835e+00 -1.82220921e-01
6.13817036e-01 -3.42763752e-01 -8.17326188e-01 8.71499002e-01
4.41826016e-01 6.72706366e-02 2.45970026e-01 1.07950442e-01
-3.55704010e-01 -4.62323517e-01 -7.38268495e-01 6.75857067e-01
1.12937212e+00 -6.01389587e-01 -2.41203189e-01 4.72279787e-01
9.15522814e-01 -2.63141364e-01 -6.98615491e-01 4.92862135e-01
1.01611026e-01 -1.09335268e+00 1.01629889e+00 -3.47116262e-01
4.15249825e-01 -4.16185439e-01 -4.17189121e-01 -1.36586881e+00
-8.90872300e-01 -3.28675479e-01 -1.85270771e-01 1.01951659e+00
3.77884388e-01 -6.76129758e-01 6.60842359e-01 1.11435816e-01
-6.00514174e-01 -8.72430921e-01 -1.07667696e+00 -5.12409866e-01
1.73719391e-01 -5.78923285e-01 9.12712634e-01 1.08469510e+00
-1.62124932e-01 7.41419733e-01 -6.98241949e-01 3.37330431e-01
4.59165990e-01 3.62655908e-01 8.42820883e-01 -9.56837058e-01
6.12453781e-02 -5.76972485e-01 -5.15082955e-01 -1.30655587e+00
1.09143570e-01 -9.43624258e-01 -1.47336209e-02 -1.80994570e+00
-1.49762586e-01 -4.50244695e-01 -3.73478740e-01 3.85733247e-01
2.38907021e-02 2.70671457e-01 3.07210069e-02 1.15492597e-01
-8.00830603e-01 1.04923868e+00 9.65388536e-01 -4.40014333e-01
-9.90686864e-02 9.98999365e-03 -4.59237993e-01 5.86910367e-01
9.58168328e-01 -2.61949092e-01 -7.89266646e-01 -9.03166711e-01
3.44838351e-01 -2.47847307e-02 5.26970744e-01 -1.37065423e+00
2.83835500e-01 -2.25125015e-01 6.58123940e-02 -1.27655017e+00
1.53361991e-01 -1.01671064e+00 1.52850419e-01 2.85032600e-01
-3.88894677e-02 6.52858019e-01 3.77253480e-02 9.81231034e-01
-4.63292897e-01 6.71281636e-01 5.82803547e-01 3.91346142e-02
-1.09913361e+00 7.02009737e-01 -4.84607309e-01 1.15137681e-01
8.81700635e-01 5.86372614e-02 -2.63137847e-01 -7.09866643e-01
-3.74655694e-01 5.21333516e-01 2.30647936e-01 5.44926047e-01
7.16847181e-01 -1.78440928e+00 -6.71839893e-01 2.26163968e-01
1.54267311e-01 2.40297783e-02 3.35750371e-01 1.00416863e+00
-5.47023416e-01 5.24333358e-01 1.91666320e-01 -6.02112114e-01
-7.06457019e-01 7.46423900e-01 4.43781018e-01 -6.86295033e-01
-8.96154583e-01 6.82025015e-01 2.55134672e-01 -6.38707340e-01
4.79569696e-02 -7.00100780e-01 -2.80509681e-01 5.03590889e-02
5.33105806e-02 3.60635519e-01 1.93583831e-01 -1.02530348e+00
-4.43999231e-01 7.47156262e-01 2.64602721e-01 1.86506867e-01
1.59154725e+00 -6.07500076e-01 2.45489836e-01 3.98584396e-01
1.47256339e+00 -2.22600505e-01 -1.11459792e+00 -4.55805004e-01
1.53442286e-02 -3.44842702e-01 2.81710595e-01 -2.06635907e-01
-1.55966043e+00 1.14535820e+00 5.28864145e-01 5.64758122e-01
1.21558452e+00 -2.03946486e-01 1.34419715e+00 1.99124187e-01
1.95242360e-01 -1.19521666e+00 -1.54535368e-01 8.03822577e-01
8.47207487e-01 -1.50484359e+00 -9.50727388e-02 -2.41581336e-01
-7.47940123e-01 7.92387307e-01 4.73479301e-01 -3.55566829e-01
1.18540049e+00 -1.03097036e-01 -1.70527399e-01 -5.96375048e-01
-9.35774565e-01 -7.62286901e-01 5.45535207e-01 6.28462911e-01
4.98211324e-01 2.39410043e-01 -6.45244494e-02 2.49584258e-01
-1.27321377e-01 -1.93048820e-01 1.21542223e-01 4.89359885e-01
-5.36826670e-01 -6.23330414e-01 9.62916985e-02 4.94201422e-01
-1.54772758e-01 -2.89388776e-01 6.98751258e-03 8.04099619e-01
1.25116557e-01 1.06456470e+00 2.49088362e-01 -9.32693779e-01
3.00031602e-01 -3.22818309e-01 -2.34561786e-01 -3.51533413e-01
-4.26496238e-01 2.78423518e-01 1.90568000e-01 -7.50469208e-01
-5.76796353e-01 -5.48777759e-01 -1.01201797e+00 -5.98561764e-01
-2.30198000e-02 -3.17454450e-02 1.27664730e-01 8.61523390e-01
6.97896063e-01 7.06841588e-01 9.18381870e-01 -9.38724637e-01
3.12753431e-02 -9.03097570e-01 -5.57439506e-01 3.43237698e-01
5.75187862e-01 -7.54610837e-01 -7.07085356e-02 -1.07519336e-01]
|
[6.5302510261535645, 2.0902459621429443]
|
25faf756-7ca2-4657-bc18-ccf69ab74c15
|
creativegan-editing-generative-adversarial
|
2103.06242
| null |
https://arxiv.org/abs/2103.06242v1
|
https://arxiv.org/pdf/2103.06242v1.pdf
|
CreativeGAN: Editing Generative Adversarial Networks for Creative Design Synthesis
|
Modern machine learning techniques, such as deep neural networks, are transforming many disciplines ranging from image recognition to language understanding, by uncovering patterns in big data and making accurate predictions. They have also shown promising results for synthesizing new designs, which is crucial for creating products and enabling innovation. Generative models, including generative adversarial networks (GANs), have proven to be effective for design synthesis with applications ranging from product design to metamaterial design. These automated computational design methods can support human designers, who typically create designs by a time-consuming process of iteratively exploring ideas using experience and heuristics. However, there are still challenges remaining in automatically synthesizing `creative' designs. GAN models, however, are not capable of generating unique designs, a key to innovation and a major gap in AI-based design automation applications. This paper proposes an automated method, named CreativeGAN, for generating novel designs. It does so by identifying components that make a design unique and modifying a GAN model such that it becomes more likely to generate designs with identified unique components. The method combines state-of-art novelty detection, segmentation, novelty localization, rewriting, and generative models for creative design synthesis. Using a dataset of bicycle designs, we demonstrate that the method can create new bicycle designs with unique frames and handles, and generalize rare novelties to a broad set of designs. Our automated method requires no human intervention and demonstrates a way to rethink creative design synthesis and exploration.
|
['Faez Ahmed', 'Muhammad Fathy Rashad', 'Amin Heyrani Nobari']
|
2021-03-10
| null | null | null | null |
['design-synthesis']
|
['adversarial']
|
[ 3.67401540e-01 7.95634091e-02 -7.41341338e-02 1.51184127e-01
-4.21810269e-01 -1.00352669e+00 5.79199016e-01 -4.43021506e-01
3.46243739e-01 6.19498134e-01 5.53045757e-02 -3.84033173e-01
5.83488978e-02 -1.06877434e+00 -9.84804213e-01 -5.27635813e-01
3.16767335e-01 4.74254429e-01 -2.40219072e-01 -3.59442949e-01
2.37140924e-01 7.85352051e-01 -1.65199554e+00 5.88962853e-01
8.79388511e-01 7.96222806e-01 1.49870664e-01 7.25544751e-01
-4.36075926e-01 3.25218141e-01 -6.94274366e-01 -4.57152277e-01
4.15322095e-01 -8.33249867e-01 -6.11518145e-01 5.29187471e-02
2.21821755e-01 -1.17143169e-01 -3.24905128e-03 9.42148507e-01
4.55925882e-01 -8.21922719e-03 6.71182930e-01 -1.14981496e+00
-1.31720698e+00 8.43830466e-01 -2.62990028e-01 -4.58227009e-01
4.79627192e-01 7.26344764e-01 9.95384395e-01 -9.44223583e-01
9.28328812e-01 1.23935366e+00 9.17992055e-01 9.27187145e-01
-1.44559324e+00 -8.56403291e-01 -2.83331484e-01 -8.12140629e-02
-1.18109322e+00 -3.84830117e-01 1.18133819e+00 -5.67588389e-01
9.11129713e-01 3.68819058e-01 1.09376371e+00 1.40498435e+00
1.88443497e-01 9.47275758e-01 9.99657869e-01 -4.20477659e-01
6.25223577e-01 -3.57036963e-02 -7.60772347e-01 6.68190956e-01
9.16204378e-02 3.27498645e-01 -2.31321797e-01 1.43506423e-01
9.62345839e-01 8.44099298e-02 1.11437611e-01 -1.14422902e-01
-1.14274406e+00 9.09606218e-01 4.58951056e-01 4.54238743e-01
-4.75850552e-01 5.71153700e-01 8.89957100e-02 2.93363661e-01
6.53978363e-02 1.41162896e+00 -2.87881494e-01 -2.02336192e-01
-9.99693155e-01 6.33867919e-01 8.40851665e-01 1.14570212e+00
7.13757455e-01 5.26766598e-01 -1.25894055e-01 5.95276892e-01
-1.13351390e-01 5.75523555e-01 4.87679958e-01 -9.98783648e-01
-1.32718340e-01 7.24017859e-01 -1.84155598e-01 -9.72350657e-01
-2.54080266e-01 -3.33874583e-01 -8.34918499e-01 3.85387987e-01
-8.63603428e-02 -2.26869226e-01 -1.27761650e+00 1.54737890e+00
2.03732461e-01 -1.03718318e-01 -3.45497653e-02 5.90472221e-01
9.19313073e-01 7.61939943e-01 -2.28243425e-01 1.90658286e-01
1.01359451e+00 -9.60548103e-01 -5.27197659e-01 -5.57006337e-02
3.82541478e-01 -9.85484004e-01 1.28767955e+00 5.83472013e-01
-1.15518785e+00 -6.59817636e-01 -9.60945964e-01 2.50316501e-01
-5.71821511e-01 -2.25880593e-01 9.54131663e-01 8.48880827e-01
-8.80707502e-01 7.37757683e-01 -3.88086379e-01 -1.40928954e-01
1.06915760e+00 3.23178411e-01 1.32226422e-02 -1.05639994e-01
-8.34119141e-01 5.94863832e-01 3.64531487e-01 -1.45084709e-02
-1.15478516e+00 -1.03756535e+00 -7.70390213e-01 -1.01339743e-02
5.03870666e-01 -1.13119042e+00 1.31347775e+00 -1.31729782e+00
-1.78361094e+00 2.80248404e-01 2.48741493e-01 -3.39053601e-01
4.10683364e-01 -1.02918837e-02 -5.72303295e-01 -1.57022789e-01
6.29323954e-03 1.12914991e+00 1.28900957e+00 -1.52897513e+00
-3.95789832e-01 1.87330127e-01 -1.73588455e-01 -4.18052882e-01
-4.02970091e-02 -3.62852156e-01 -1.42378146e-02 -1.24037957e+00
-9.30320323e-02 -1.08284187e+00 -5.42575359e-01 2.40484122e-02
-6.37032330e-01 -3.47567461e-02 1.06188202e+00 -5.12627244e-01
1.00213456e+00 -1.84599900e+00 1.90141022e-01 4.53407526e-01
2.19753355e-01 3.46433014e-01 -2.38963440e-01 3.66513997e-01
1.62419602e-01 6.26030624e-01 -1.96407124e-01 -7.91858062e-02
1.82337984e-01 2.66102165e-01 -4.62597549e-01 -3.04140925e-01
5.96035480e-01 1.77825117e+00 -8.52556586e-01 2.39396282e-02
2.34714776e-01 1.62004769e-01 -8.54149282e-01 -5.35046449e-03
-1.09651673e+00 6.29664004e-01 -2.88938314e-01 9.43669856e-01
2.75359541e-01 -1.34813309e-01 -1.64842740e-01 -8.77618864e-02
-4.46009375e-02 -1.23247318e-01 -1.05180788e+00 1.88199210e+00
-6.35097563e-01 8.10730815e-01 -5.08671463e-01 -8.12378883e-01
1.13666439e+00 1.06897295e-01 2.49029726e-01 -7.03314960e-01
2.55921066e-01 2.71760374e-01 -1.14672318e-01 -5.87239742e-01
5.92431605e-01 -2.02085927e-01 -3.96610975e-01 6.58157527e-01
-1.22803926e-01 -7.87171841e-01 1.31316140e-01 -1.60202309e-01
1.20821953e+00 2.07653522e-01 -1.05779488e-02 2.39786625e-01
-1.41093820e-01 2.48120010e-01 5.34267902e-01 8.08650017e-01
4.81388628e-01 8.12434435e-01 2.68548876e-01 -7.24189997e-01
-1.57181871e+00 -1.12799668e+00 4.04332966e-01 3.86333555e-01
-1.77020386e-01 -2.02484488e-01 -6.71501517e-01 -5.12909949e-01
4.57477942e-02 9.03676569e-01 -5.82133532e-01 -4.26433295e-01
-6.15581453e-01 -1.45249084e-01 5.94824195e-01 5.09158611e-01
4.49298859e-01 -1.49889576e+00 -4.86398786e-01 3.30839366e-01
2.59184927e-01 -8.37690175e-01 -3.85493368e-01 -2.15666026e-01
-7.09461510e-01 -6.66703343e-01 -8.61831427e-01 -9.54738319e-01
8.07467699e-01 -1.21854261e-01 1.17367566e+00 -3.09115723e-02
-8.31360281e-01 2.95866787e-01 -2.37898588e-01 -4.20551389e-01
-9.93911922e-01 1.25413567e-01 -1.30233899e-01 -2.59346753e-01
-2.24196196e-01 -9.05125320e-01 -5.66476464e-01 2.97110319e-01
-1.16450655e+00 4.77208644e-01 9.59945619e-01 9.01962638e-01
7.33187616e-01 4.01520938e-01 9.38560128e-01 -8.59501302e-01
7.99175382e-01 -3.42099190e-01 -4.38250989e-01 2.51299739e-01
-7.27744937e-01 1.85706392e-01 9.60734010e-01 -8.49976599e-01
-9.21068847e-01 1.03942961e-01 -1.28296986e-01 -6.46637499e-01
-3.16766500e-01 3.98188293e-01 -3.27344626e-01 -1.91639975e-01
8.59598577e-01 2.92873196e-02 -1.86345596e-02 -5.01716793e-01
1.03690326e+00 2.62960911e-01 5.07793784e-01 -6.61516905e-01
9.72547412e-01 1.98361248e-01 8.15865993e-02 -7.17917979e-01
-1.16566002e-01 2.83321470e-01 -3.45675290e-01 -2.16373265e-01
6.74320877e-01 -4.89769787e-01 -6.13248646e-01 3.22556674e-01
-1.21870422e+00 -2.48148844e-01 -1.09471929e+00 -1.06925964e-01
-6.86852515e-01 -2.37853099e-02 -3.24818164e-01 -4.34349239e-01
-3.99442673e-01 -1.27916968e+00 8.84940863e-01 5.44770002e-01
-7.28429496e-01 -7.26361096e-01 -5.12317494e-02 1.88242123e-01
5.67463815e-01 7.12688088e-01 1.20239091e+00 -1.69662029e-01
-1.00327194e+00 -1.89685121e-01 1.55809581e-01 3.88421178e-01
4.67529714e-01 1.41116455e-01 -7.68233955e-01 1.79350927e-01
-4.41122442e-01 5.77243343e-02 6.36932194e-01 2.60634691e-01
1.24663496e+00 -7.63004005e-01 -3.37651253e-01 6.12379193e-01
1.13524711e+00 6.47186935e-01 8.83195341e-01 1.26201838e-01
1.05333567e+00 3.00656587e-01 2.28577945e-02 3.14087480e-01
-4.79188785e-02 4.66767907e-01 2.52228230e-01 -1.29941955e-01
-4.17806953e-01 -5.24098873e-01 2.64423966e-01 4.68157351e-01
6.52697682e-02 -2.23362967e-01 -8.31333816e-01 6.53699219e-01
-1.67783117e+00 -1.15484142e+00 3.19589227e-01 1.55251586e+00
8.55957091e-01 1.13496661e-01 2.12348863e-01 9.65364128e-02
5.93092442e-01 -3.04571211e-01 -8.90459597e-01 -8.65153909e-01
-2.54526526e-01 9.41230357e-01 2.42011175e-01 -1.68210298e-01
-6.58214390e-01 1.24867380e+00 6.57770824e+00 9.57358360e-01
-1.11596096e+00 -1.19231261e-01 6.01347506e-01 -1.85917377e-01
-1.11514223e+00 9.58691090e-02 -3.49787384e-01 5.13365448e-01
4.91640151e-01 -6.29033521e-02 9.13536966e-01 1.06762981e+00
1.35141566e-01 1.92526549e-01 -1.10423076e+00 1.03972220e+00
1.55548435e-02 -2.26460338e+00 5.43345749e-01 1.01785464e-02
1.66605997e+00 -8.25928986e-01 5.19829512e-01 1.25594601e-01
4.86822605e-01 -1.39594185e+00 8.90696585e-01 6.88948989e-01
7.29586720e-01 -9.85175371e-01 2.46432707e-01 -9.50651392e-02
-1.00953555e+00 -2.26513878e-01 1.31257772e-01 3.97047661e-02
2.05290005e-01 6.99832797e-01 -1.14438713e+00 1.59781888e-01
5.72420001e-01 5.65013945e-01 -4.87354487e-01 9.78978872e-01
-4.32769716e-01 3.70408177e-01 -1.56664237e-01 -2.36880749e-01
1.29262865e-01 1.25666149e-02 6.93287969e-01 7.79159784e-01
7.37392485e-01 -2.64863074e-01 -3.22222933e-02 1.90989494e+00
-4.46495235e-01 -2.45708629e-01 -9.29941177e-01 -7.89915562e-01
4.06110704e-01 9.65106308e-01 -1.07452238e+00 -1.21460773e-01
3.89753953e-02 9.86928523e-01 -3.96118015e-01 3.63174945e-01
-8.52665544e-01 -5.47065437e-01 6.72841311e-01 4.24991623e-02
6.86560452e-01 -3.94475043e-01 -8.87668610e-01 -6.50877059e-01
-4.05863151e-02 -1.36004996e+00 -2.33442813e-01 -9.02427316e-01
-1.17437315e+00 3.20071518e-01 -2.86348373e-01 -1.22008717e+00
-4.63257611e-01 -4.03206557e-01 -7.70793319e-01 4.36790437e-01
-7.26266325e-01 -1.62099433e+00 -9.54211205e-02 9.98839438e-02
8.50935221e-01 -4.16105121e-01 5.71742773e-01 2.20908031e-01
-4.27248567e-01 7.39393055e-01 3.94816250e-02 -1.99276432e-01
2.07109168e-01 -1.00192022e+00 1.08261204e+00 7.68549621e-01
3.67071956e-01 5.15666604e-01 7.03483582e-01 -9.64871883e-01
-1.70790231e+00 -1.16106594e+00 2.38705471e-01 -2.89469302e-01
3.00153285e-01 -3.77994150e-01 -5.45659602e-01 3.89757484e-01
-6.79640397e-02 -4.82200921e-01 5.84196687e-01 -1.96594283e-01
-2.74508983e-01 1.93168372e-01 -1.29110456e+00 1.07641959e+00
1.46588957e+00 -3.20479602e-01 -4.54720646e-01 1.17251150e-01
1.26988339e+00 -2.50709057e-01 -7.28563547e-01 3.49121243e-01
7.96098769e-01 -7.54688561e-01 1.07067919e+00 -6.06149375e-01
7.96027720e-01 -3.03628862e-01 1.73817962e-01 -1.41790414e+00
-2.35442758e-01 -1.29457855e+00 -3.94986123e-01 1.39020681e+00
6.23991072e-01 -3.55675995e-01 8.60115230e-01 5.53214133e-01
-5.13552547e-01 -8.73473942e-01 -5.17739296e-01 -9.49184418e-01
-1.31574929e-01 -6.30598664e-01 1.31484103e+00 8.26427400e-01
-4.88144904e-01 6.56045079e-02 -3.09742630e-01 -3.05855036e-01
1.87051967e-01 4.31045890e-01 1.03480804e+00 -9.22143757e-01
-4.68251854e-01 -1.03428626e+00 -4.96091694e-01 -7.05108345e-01
2.57073939e-02 -8.26026618e-01 6.54450953e-02 -1.60705900e+00
-3.18865955e-01 -6.82346225e-01 2.55672067e-01 4.37565684e-01
3.21957529e-01 3.41300786e-01 1.14873841e-01 2.68725436e-02
-1.04667090e-01 5.34054697e-01 1.63550603e+00 -6.29707873e-01
-7.47275472e-01 -6.51692525e-02 -1.04684091e+00 5.67014039e-01
8.09794605e-01 -3.62910926e-01 -2.11998954e-01 -2.41932452e-01
7.67883360e-01 -6.12613380e-01 5.59702218e-01 -1.08195210e+00
1.06669804e-02 -2.21830159e-01 6.25500977e-01 -4.31277931e-01
-4.36133742e-02 -5.96894801e-01 9.60242093e-01 6.05643094e-01
-2.01557398e-01 1.37982637e-01 4.35262084e-01 4.30754930e-01
1.80031240e-01 -1.50020301e-01 6.07774019e-01 -3.84604812e-01
-1.05235946e+00 1.47218809e-01 -3.25038433e-01 -6.06509522e-02
1.12194538e+00 -5.05635023e-01 -1.30816802e-01 -2.88253307e-01
-7.66970277e-01 -1.99239284e-01 7.50536084e-01 7.84084737e-01
8.35193515e-01 -1.73743570e+00 -2.52207339e-01 4.98788714e-01
-1.56056762e-01 2.27924302e-01 2.42491066e-01 1.07765734e-01
-7.25026190e-01 2.54213475e-02 -3.29070687e-01 -5.43134570e-01
-7.47741520e-01 7.29384243e-01 7.66100064e-02 1.02844156e-01
-7.55908072e-01 8.14431608e-01 2.31370609e-02 -2.14411393e-01
-1.51219815e-01 -6.13297045e-01 3.71630400e-01 -1.62069932e-01
2.74228185e-01 4.38414693e-01 -8.34419653e-02 -1.59728259e-01
1.04510859e-01 7.59163320e-01 5.37335612e-02 -3.32384631e-02
1.42245448e+00 3.78854692e-01 6.00348320e-03 1.82087362e-01
1.03784657e+00 -2.23902658e-01 -1.07243669e+00 3.97621877e-02
-1.28548995e-01 -3.99728864e-01 -1.48650602e-01 -9.06934619e-01
-1.14993143e+00 6.62517071e-01 6.37498021e-01 1.95920467e-01
1.05182087e+00 2.14983821e-01 1.56661654e+00 4.22884315e-01
3.44855607e-01 -1.17278075e+00 6.71391606e-01 2.12522507e-01
1.09480476e+00 -9.93994653e-01 -2.78419226e-01 -5.24499267e-02
-4.80102122e-01 1.09587646e+00 3.83330256e-01 -7.66748711e-02
3.34145099e-01 3.29052567e-01 -3.58222336e-01 -2.54546821e-01
-2.95186758e-01 -1.53831974e-01 5.23172379e-01 6.90243900e-01
-1.53006196e-01 2.32885659e-01 2.08261698e-01 7.43614435e-01
-5.67790151e-01 9.63985026e-02 1.89471722e-01 1.05946207e+00
-2.55388260e-01 -1.25733912e+00 -3.92842948e-01 4.90132123e-01
-1.47975078e-02 -8.06684792e-02 -5.75313032e-01 5.98943472e-01
6.31259024e-01 6.94650710e-01 -4.27942388e-02 -7.18839765e-01
3.90583068e-01 -7.81722888e-02 6.58508182e-01 -6.13042951e-01
-6.05288923e-01 -1.73910573e-01 -8.77560154e-02 -4.06727582e-01
1.05570452e-02 -4.19544816e-01 -1.14473593e+00 -5.78009486e-01
-1.92129076e-01 -1.79359779e-01 8.74829412e-01 8.45508814e-01
8.74502420e-01 8.39106441e-01 7.75876403e-01 -1.04704833e+00
2.53687818e-02 -3.15176040e-01 -6.35746941e-02 3.75909626e-01
3.69138196e-02 -4.83539015e-01 -4.82744649e-02 3.68154854e-01]
|
[5.809299468994141, 3.322704792022705]
|
b2616525-0510-458f-9060-aa1f54f69941
|
open-domain-question-answering-over-virtual
|
2110.08417
| null |
https://arxiv.org/abs/2110.08417v2
|
https://arxiv.org/pdf/2110.08417v2.pdf
|
Open Domain Question Answering with A Unified Knowledge Interface
|
The retriever-reader framework is popular for open-domain question answering (ODQA) due to its ability to use explicit knowledge. Although prior work has sought to increase the knowledge coverage by incorporating structured knowledge beyond text, accessing heterogeneous knowledge sources through a unified interface remains an open question. While data-to-text generation has the potential to serve as a universal interface for data and text, its feasibility for downstream tasks remains largely unknown. In this work, we bridge this gap and use the data-to-text method as a means for encoding structured knowledge for ODQA. Specifically, we propose a verbalizer-retriever-reader framework for ODQA over data and text where verbalized tables from Wikipedia and graphs from Wikidata are used as augmented knowledge sources. We show that our Unified Data and Text QA, UDT-QA, can effectively benefit from the expanded knowledge index, leading to large gains over text-only baselines. Notably, our approach sets the single-model state-of-the-art on Natural Questions. Furthermore, our analyses indicate that verbalized knowledge is preferred for answer reasoning for both adapted and hot-swap settings.
|
['Jianfeng Gao', 'Eric Nyberg', 'Xiaodong Liu', 'Hao Cheng', 'Kaixin Ma']
|
2021-10-16
| null |
https://aclanthology.org/2022.acl-long.113
|
https://aclanthology.org/2022.acl-long.113.pdf
|
acl-2022-5
|
['data-to-text-generation']
|
['natural-language-processing']
|
[-1.55578479e-01 6.53497040e-01 -1.84608430e-01 -1.71264365e-01
-1.51982594e+00 -1.03463316e+00 6.61723554e-01 3.66072297e-01
-4.57425237e-01 7.79762506e-01 8.00568938e-01 -6.54384732e-01
-3.13516021e-01 -1.06808102e+00 -8.61220837e-01 1.24730349e-01
4.96075243e-01 1.19964647e+00 5.83009899e-01 -7.80644119e-01
-2.81447787e-02 -1.81788772e-01 -1.26688266e+00 7.33844459e-01
1.44400835e+00 7.50269294e-01 1.34873226e-01 4.55240577e-01
-8.76275063e-01 1.30860519e+00 -4.79195714e-01 -9.99429286e-01
6.00635968e-02 -1.69953659e-01 -1.75535798e+00 -5.43010473e-01
7.76574075e-01 -4.43922758e-01 -4.76466328e-01 4.84294325e-01
5.57031751e-01 1.01204798e-01 6.00210845e-01 -8.81124318e-01
-1.18051600e+00 1.07100415e+00 -2.21492555e-02 1.23397991e-01
7.93035924e-01 1.43062711e-01 1.51391900e+00 -7.06781328e-01
9.33649004e-01 1.42313123e+00 3.48248214e-01 5.44848323e-01
-1.15177786e+00 -1.25999689e-01 4.89299484e-02 4.47692484e-01
-1.06658685e+00 -5.11422098e-01 2.28322744e-01 -2.25330189e-01
1.52012253e+00 3.05138022e-01 2.10906193e-01 8.09519172e-01
-4.74263966e-01 1.07872200e+00 6.83379948e-01 -6.34156048e-01
-1.74468473e-01 1.41172007e-01 5.14723718e-01 8.07233512e-01
3.59498650e-01 -4.57023025e-01 -6.89222693e-01 -1.22830719e-01
3.12762648e-01 -5.81988215e-01 -4.67746884e-01 -2.79254079e-01
-1.02870011e+00 7.99698293e-01 4.17638600e-01 -1.23333789e-01
-4.31835264e-01 1.29927531e-01 4.08214033e-01 5.94241142e-01
4.07116681e-01 1.07060266e+00 -9.23457861e-01 -3.51872087e-01
-5.28085291e-01 7.15915442e-01 1.42279828e+00 1.11093116e+00
7.45027721e-01 -5.07333398e-01 -6.73826456e-01 9.10628080e-01
1.57294989e-01 7.98520446e-01 2.08063439e-01 -1.18528295e+00
9.99005437e-01 9.67966855e-01 1.86115786e-01 -6.43686533e-01
-2.69837618e-01 -2.94663370e-01 -7.80100450e-02 -4.61634308e-01
8.64954233e-01 -1.84405789e-01 -7.64006793e-01 1.70399022e+00
4.08223182e-01 -4.17456746e-01 5.93305171e-01 6.64943576e-01
1.41681015e+00 6.66267455e-01 2.93105748e-02 2.99236655e-01
1.66350508e+00 -1.03813362e+00 -8.60432029e-01 -3.96747798e-01
1.10797727e+00 -4.27066088e-01 1.39211106e+00 9.17822346e-02
-1.20650840e+00 -6.73290268e-02 -5.32256424e-01 -9.02227104e-01
-7.27915525e-01 -2.78920263e-01 4.63284671e-01 4.89861190e-01
-1.20333123e+00 -6.42602071e-02 -6.70556426e-01 -5.31429350e-01
3.58120263e-01 -1.76213771e-01 -2.28609264e-01 -6.79824710e-01
-1.62333322e+00 1.30872917e+00 4.77038652e-01 -1.66269898e-01
-5.52854836e-01 -1.10757005e+00 -9.96583879e-01 1.11972362e-01
9.57444787e-01 -1.35447741e+00 1.67657840e+00 -8.19151625e-02
-1.33021712e+00 6.88902617e-01 -2.86388040e-01 -5.96846700e-01
8.25075060e-02 -5.12954950e-01 -1.90127119e-01 3.16418052e-01
2.35725462e-01 5.32870471e-01 3.34796011e-01 -1.16231525e+00
-3.99956226e-01 -4.09353137e-01 7.72003591e-01 4.50263083e-01
-2.98990279e-01 -6.15803488e-02 -7.80043364e-01 -3.51696938e-01
1.06398836e-02 -5.16602278e-01 2.01894224e-01 -1.81956097e-01
-3.77857298e-01 -5.24537623e-01 4.80089247e-01 -1.17821670e+00
1.71626604e+00 -1.47238886e+00 2.88149178e-01 4.78393212e-02
4.24643666e-01 2.00867176e-01 -1.45585269e-01 8.45007479e-01
5.50347626e-01 2.00185388e-01 -2.15649828e-01 -3.39730419e-02
4.80415881e-01 3.83566320e-01 -6.30446732e-01 -4.73113239e-01
3.60781491e-01 1.57474411e+00 -1.00693440e+00 -7.37649083e-01
-3.64039540e-01 2.81253215e-02 -8.70010138e-01 1.62991241e-01
-9.64991093e-01 -1.04920372e-01 -6.56397521e-01 6.29639447e-01
1.97045550e-01 -5.21353483e-01 1.26980210e-03 -2.15738460e-01
2.67573118e-01 9.61265802e-01 -9.00634885e-01 1.98612022e+00
-6.25110567e-01 3.25615197e-01 -1.41655738e-02 -2.29181483e-01
5.01067877e-01 2.59526789e-01 -1.28354177e-01 -9.36237514e-01
-2.94281781e-01 3.97602618e-01 -1.82100669e-01 -8.22927237e-01
8.68570566e-01 2.71126777e-02 -1.78650588e-01 7.16448188e-01
1.94433153e-01 -5.01394212e-01 5.08360744e-01 1.09750581e+00
1.46522391e+00 2.62193561e-01 -4.22377959e-02 2.34741587e-02
2.56469667e-01 5.69295883e-01 -1.11390933e-01 9.01148260e-01
4.05998975e-01 8.38456377e-02 4.03704643e-01 2.34247968e-01
-5.60770273e-01 -1.16704845e+00 -4.32627127e-02 1.36063063e+00
-3.32979485e-02 -6.52184129e-01 -7.19338596e-01 -8.24727774e-01
2.69474238e-01 1.21078873e+00 -3.53176773e-01 -1.04871765e-01
-5.94585478e-01 -1.67705342e-01 9.96285021e-01 6.91539168e-01
4.81289387e-01 -9.17424917e-01 -3.35337698e-01 2.65637875e-01
-8.35914016e-01 -1.13386941e+00 -3.08153719e-01 -9.86602064e-03
-7.28666663e-01 -1.14884782e+00 -5.42793989e-01 -5.84059536e-01
1.50663882e-01 1.58374414e-01 1.89074647e+00 5.18343318e-03
1.74201220e-01 9.97035623e-01 -6.80354893e-01 -4.14966851e-01
-4.81767893e-01 4.81578529e-01 -5.57142735e-01 -5.96160233e-01
4.84479308e-01 -2.33920559e-01 -4.92692769e-01 2.81576766e-03
-1.08551240e+00 8.16500708e-02 4.11566705e-01 6.42425418e-01
2.16616049e-01 -5.62723398e-01 8.34341586e-01 -1.05634975e+00
9.94962215e-01 -7.29988515e-01 -3.98100466e-01 6.66451633e-01
-5.60703456e-01 7.06050515e-01 2.87123889e-01 2.18206257e-01
-1.49836481e+00 -6.27200961e-01 -1.99565873e-01 4.55052927e-02
3.07801574e-01 1.19562995e+00 -2.44462550e-01 3.19615752e-01
9.45678830e-01 -4.29219194e-02 -3.61042842e-02 -6.71852291e-01
1.30511618e+00 6.31917596e-01 6.67217910e-01 -1.24904203e+00
7.96224117e-01 2.10608810e-01 -4.10122335e-01 -4.23241496e-01
-1.13096380e+00 -6.67741120e-01 -3.64490658e-01 1.10733375e-01
9.50376451e-01 -1.04385340e+00 -4.37740743e-01 1.92787752e-01
-1.27632558e+00 -5.45368612e-01 -4.60091561e-01 -1.94505408e-01
-3.20355177e-01 4.18593615e-01 -6.86623871e-01 -4.79888469e-01
-6.13306046e-01 -7.18270719e-01 1.10266161e+00 -7.31762648e-02
-3.76866639e-01 -1.27621758e+00 1.84744358e-01 9.89857078e-01
5.43613672e-01 -1.95206806e-01 1.52966869e+00 -1.01951456e+00
-9.59503472e-01 -1.71165895e-02 -4.20722365e-01 2.71064639e-02
-1.59100115e-01 -4.40886736e-01 -8.33048344e-01 7.64221400e-02
-5.91901541e-01 -1.08089578e+00 9.54493999e-01 -1.30515218e-01
8.06546092e-01 -4.90644068e-01 -2.08958000e-01 1.98706061e-01
1.23623276e+00 -3.13908607e-01 6.97148800e-01 4.45632547e-01
7.26015925e-01 8.02644908e-01 4.04641122e-01 1.06214799e-01
1.40418625e+00 6.14612162e-01 2.86067367e-01 2.38915145e-01
-4.99661565e-01 -5.66816926e-01 1.04050204e-01 9.36425626e-01
3.41753989e-01 -4.91423398e-01 -1.30170274e+00 9.12406087e-01
-1.65718496e+00 -8.39700699e-01 -2.64971793e-01 1.92804289e+00
1.40572071e+00 -1.57008544e-01 -9.27013531e-02 -3.14932108e-01
7.89867640e-02 -4.17876150e-03 -5.38611889e-01 -6.19978495e-02
-3.00473988e-01 5.22466958e-01 2.71712005e-01 7.85309255e-01
-4.47019905e-01 1.11228096e+00 5.65801001e+00 8.76269341e-01
-3.66199493e-01 9.64366421e-02 -8.53930265e-02 -4.95120957e-02
-9.08996165e-01 1.49178669e-01 -9.85752404e-01 -1.52832689e-02
9.61726367e-01 -3.61840487e-01 4.73514587e-01 4.57280606e-01
-3.94565225e-01 -1.56044483e-01 -1.22112989e+00 6.45027637e-01
1.25278920e-01 -1.54992306e+00 5.96513510e-01 -2.85081297e-01
5.68235993e-01 9.91960764e-02 -1.17805690e-01 9.77070093e-01
8.04249287e-01 -9.59454238e-01 6.64097428e-01 6.46175206e-01
7.37489641e-01 -3.34946573e-01 6.10409617e-01 3.72342050e-01
-1.16657066e+00 -6.17255922e-03 -5.77406697e-02 9.74552259e-02
2.67763168e-01 1.77634865e-01 -9.36394334e-01 9.38016832e-01
5.02633512e-01 3.46246123e-01 -9.18366849e-01 6.64701283e-01
-3.25266272e-01 6.68955982e-01 -4.14376587e-01 -1.62773073e-01
2.22554788e-01 2.54922748e-01 4.26837176e-01 1.09025288e+00
-5.28223522e-04 3.09705198e-01 9.86861363e-02 9.85932529e-01
-4.56254750e-01 1.93085536e-01 -4.86224413e-01 -2.64286667e-01
7.96556771e-01 8.08154941e-01 3.78743522e-02 -5.45466840e-01
-6.57630682e-01 9.01829123e-01 8.60530019e-01 4.92463678e-01
-3.54798526e-01 -6.76530361e-01 4.03536677e-01 1.10118687e-01
3.15296382e-01 -3.39014605e-02 -1.68991834e-01 -1.34427834e+00
3.44131559e-01 -1.24415648e+00 1.05023801e+00 -1.19691885e+00
-1.39563012e+00 3.64518642e-01 3.12999308e-01 -5.46916842e-01
-6.27479315e-01 -6.36124730e-01 5.44290664e-03 1.03102851e+00
-1.77697897e+00 -1.26822996e+00 -2.39224330e-01 7.49037802e-01
4.98317361e-01 1.04982167e-01 8.66188645e-01 2.28031650e-01
-1.40760183e-01 6.02188945e-01 1.26190484e-01 3.71613204e-01
7.98611641e-01 -1.63336885e+00 5.44612110e-01 7.16748297e-01
3.22931468e-01 9.72038686e-01 5.16670048e-01 -8.11142206e-01
-1.88784528e+00 -9.65115428e-01 1.00128913e+00 -1.38403845e+00
9.95420158e-01 -1.95293292e-01 -1.30040514e+00 9.51343000e-01
5.05151868e-01 -4.00449246e-01 6.00685060e-01 5.19228101e-01
-7.76285946e-01 9.69338566e-02 -7.91783929e-01 6.96662724e-01
1.06345761e+00 -8.52798998e-01 -1.28170204e+00 2.56370246e-01
1.36626422e+00 -7.34670877e-01 -1.09360445e+00 9.34745297e-02
2.43721753e-01 -4.08265263e-01 9.45962131e-01 -8.56927991e-01
5.58963299e-01 -3.28528583e-01 -2.86677271e-01 -1.50557065e+00
8.05113167e-02 -5.63977957e-01 -6.43001854e-01 1.33573306e+00
9.85291421e-01 -7.30014741e-01 4.80954826e-01 9.09522176e-01
-4.59429950e-01 -5.74470520e-01 -8.66207421e-01 -6.64017320e-01
4.34274971e-01 -5.37447393e-01 7.31764793e-01 8.21800411e-01
4.95208621e-01 8.26207101e-01 2.78445840e-01 1.10366672e-01
2.55715072e-01 -1.68975629e-02 9.49301362e-01 -1.10185540e+00
-4.03097689e-01 -2.08767921e-01 2.32171506e-01 -1.59340417e+00
1.93576187e-01 -1.38985956e+00 -1.36284038e-01 -2.49025464e+00
1.38764367e-01 -3.75173599e-01 2.29599789e-01 6.65973842e-01
-4.21607971e-01 -2.81166106e-01 1.56660601e-01 -8.94736126e-02
-9.98575211e-01 6.87070727e-01 1.37073958e+00 -4.05435003e-02
-5.18492423e-04 -6.71876311e-01 -1.08722293e+00 2.21110225e-01
3.13296497e-01 -1.56359926e-01 -7.60815322e-01 -1.09734595e+00
8.00819993e-01 3.53962630e-01 3.71287167e-01 -5.93959153e-01
5.39470017e-01 1.02660351e-01 -3.55202436e-01 -4.81339931e-01
3.53077739e-01 -5.89086652e-01 -4.54084873e-01 -2.80436743e-02
-6.10300183e-01 1.42635971e-01 5.20896375e-01 5.51512480e-01
-3.23088616e-01 -2.61389732e-01 5.39635569e-02 -2.32505143e-01
-6.54886544e-01 4.43003960e-02 -2.31988445e-01 1.25299311e+00
2.09233001e-01 9.08076167e-02 -1.18953550e+00 -7.29906797e-01
-2.19210953e-01 7.95339108e-01 2.15414941e-01 3.97267997e-01
4.27806497e-01 -1.08768022e+00 -8.54609013e-01 -3.34513724e-01
6.27646446e-01 2.34539717e-01 2.07719266e-01 6.46126747e-01
-3.75663728e-01 8.67050588e-01 2.91446388e-01 -2.42291361e-01
-9.11018372e-01 3.86956334e-01 3.23898047e-01 -6.64739132e-01
-3.42238724e-01 8.71004522e-01 -2.02363729e-01 -9.26244080e-01
1.12829342e-01 -6.49204135e-01 -1.95079252e-01 1.21350899e-01
5.19702733e-01 4.65820044e-01 4.56423134e-01 -7.08135515e-02
-1.92836933e-02 1.08125702e-01 -2.42779195e-01 -3.30903858e-01
1.02474582e+00 -3.83533835e-01 -3.30249310e-01 2.32517436e-01
8.19978297e-01 1.81235939e-01 -5.91013312e-01 -7.63543129e-01
3.30765039e-01 -1.45924628e-01 -2.13658214e-02 -1.46425855e+00
-4.57924515e-01 6.22822881e-01 -1.22389078e-01 2.56613016e-01
8.16868961e-01 4.13959414e-01 1.08744514e+00 1.20908129e+00
4.69899833e-01 -7.14864850e-01 1.77259922e-01 1.12329054e+00
1.00791371e+00 -1.20364940e+00 -3.08906496e-01 -3.73447865e-01
-7.79181719e-01 7.44001329e-01 8.01032722e-01 6.93337321e-01
1.63269103e-01 -3.46707888e-02 2.78595001e-01 -7.01199353e-01
-1.13735282e+00 -6.91140056e-01 6.10012472e-01 7.96001077e-01
4.68618453e-01 -3.61001730e-01 1.11769505e-01 7.32006669e-01
-5.57999194e-01 -6.10848889e-02 3.26502502e-01 1.10174942e+00
-4.90311772e-01 -9.52974439e-01 -2.00876862e-01 7.31852889e-01
-2.57266313e-01 -6.78093910e-01 -4.43699300e-01 8.81496131e-01
-4.65679616e-01 1.21926820e+00 -1.57969385e-01 8.80994573e-02
6.53545558e-01 6.09112859e-01 6.42042100e-01 -8.71472597e-01
-7.39594221e-01 -7.00937629e-01 9.23713386e-01 -4.53606367e-01
2.09913100e-03 -3.65235835e-01 -1.34781718e+00 -2.51788765e-01
-2.99932986e-01 4.02086288e-01 2.43196324e-01 1.05263019e+00
6.38640046e-01 4.69677538e-01 -3.08942854e-01 1.40554130e-01
-7.16594100e-01 -9.70727265e-01 -1.52423501e-01 3.75481814e-01
2.33301118e-01 -4.84799325e-01 -1.43197283e-01 1.19396247e-01]
|
[10.735709190368652, 7.931951522827148]
|
2063082e-90ed-41c5-b08c-9d5cbf48f3c1
|
character-centric-story-visualization-via
|
2210.08465
| null |
https://arxiv.org/abs/2210.08465v4
|
https://arxiv.org/pdf/2210.08465v4.pdf
|
Character-Centric Story Visualization via Visual Planning and Token Alignment
|
Story visualization advances the traditional text-to-image generation by enabling multiple image generation based on a complete story. This task requires machines to 1) understand long text inputs and 2) produce a globally consistent image sequence that illustrates the contents of the story. A key challenge of consistent story visualization is to preserve characters that are essential in stories. To tackle the challenge, we propose to adapt a recent work that augments Vector-Quantized Variational Autoencoders (VQ-VAE) with a text-tovisual-token (transformer) architecture. Specifically, we modify the text-to-visual-token module with a two-stage framework: 1) character token planning model that predicts the visual tokens for characters only; 2) visual token completion model that generates the remaining visual token sequence, which is sent to VQ-VAE for finalizing image generations. To encourage characters to appear in the images, we further train the two-stage framework with a character-token alignment objective. Extensive experiments and evaluations demonstrate that the proposed method excels at preserving characters and can produce higher quality image sequences compared with the strong baselines. Codes can be found in https://github.com/sairin1202/VP-CSV
|
['Nanyun Peng', 'Hideki Nakayama', 'Te-Lin Wu', 'Rujun Han', 'Hong Chen']
|
2022-10-16
| null | null | null | null |
['story-visualization']
|
['computer-vision']
|
[ 3.42538297e-01 1.98978826e-01 1.03268586e-01 -1.26758412e-01
-6.12537503e-01 -5.09946406e-01 7.70747721e-01 -3.31361920e-01
2.50053480e-02 6.21981144e-01 4.39514220e-01 -2.51588970e-01
5.37416220e-01 -7.10825384e-01 -1.01346445e+00 -5.91194570e-01
4.36570317e-01 2.58645624e-01 4.53377627e-02 -2.14122415e-01
2.84935325e-01 3.01798694e-02 -1.59718645e+00 7.35577941e-01
9.07333970e-01 6.77516103e-01 6.40949070e-01 1.00511408e+00
-3.40757489e-01 1.26133490e+00 -6.07381582e-01 -5.74693918e-01
7.09399953e-02 -6.97613299e-01 -6.66877508e-01 2.40488559e-01
3.48381788e-01 -7.25880563e-01 -3.31639677e-01 8.17675173e-01
3.93913388e-01 2.40546972e-01 7.44667768e-01 -1.53171182e+00
-1.06255567e+00 8.50899339e-01 -7.30287910e-01 -3.54647905e-01
3.02516878e-01 5.50467610e-01 9.49142754e-01 -1.12277985e+00
9.95977819e-01 1.28882265e+00 2.78832406e-01 7.93110371e-01
-1.11122894e+00 -5.26742816e-01 1.16445988e-01 2.41787642e-01
-1.28000927e+00 -3.40898067e-01 7.96420991e-01 -6.66311085e-01
8.51604939e-01 3.48183841e-01 9.57368672e-01 1.38839400e+00
4.27584313e-02 1.26187849e+00 7.66298413e-01 -3.23541760e-01
-6.25985442e-04 1.62893757e-01 -3.04484308e-01 6.26768112e-01
-3.79930824e-01 1.86346248e-02 -7.59134173e-01 2.82510400e-01
1.00735855e+00 -1.18143477e-01 -2.63674259e-01 -2.69067228e-01
-1.30965960e+00 8.59076023e-01 2.89029211e-01 1.40419036e-01
-6.08120143e-01 5.35616815e-01 3.88866246e-01 -2.18249802e-02
2.53640294e-01 4.64909911e-01 1.24860577e-01 -2.46507540e-01
-1.19111586e+00 6.45626664e-01 2.43629426e-01 1.21256852e+00
4.28671986e-01 4.77627188e-01 -9.06535983e-01 7.79580712e-01
1.27391145e-01 4.42278683e-01 2.47314364e-01 -1.06836653e+00
4.23248887e-01 3.49998027e-01 2.15589836e-01 -6.42203391e-01
2.24823862e-01 -3.12288493e-01 -8.58052254e-01 4.77995515e-01
1.11436613e-01 -1.32202625e-01 -1.10337794e+00 1.69002569e+00
3.93770248e-01 2.83685178e-01 1.68869957e-01 1.00228608e+00
1.08950913e+00 1.22669244e+00 3.13611440e-02 1.09764889e-01
1.35946262e+00 -1.28837645e+00 -9.02777255e-01 -2.89938655e-02
2.19232902e-01 -8.16575825e-01 1.40159857e+00 2.14382246e-01
-1.46551085e+00 -6.74726784e-01 -1.09376156e+00 -5.55387259e-01
-8.88994038e-02 3.06480885e-01 2.00149357e-01 1.69458568e-01
-1.20001626e+00 4.63348657e-01 -6.09629750e-01 1.76758338e-02
5.17387629e-01 -2.68710017e-01 2.35997457e-02 2.42041685e-02
-1.09604776e+00 6.13708556e-01 5.46569347e-01 -2.53450070e-02
-1.33109486e+00 -8.67916405e-01 -1.01366413e+00 2.22444788e-01
1.79887667e-01 -1.05643415e+00 1.46380901e+00 -1.13548934e+00
-1.53824329e+00 5.43449283e-01 -2.54238278e-01 -2.01974988e-01
9.82827961e-01 -2.36662492e-01 1.62111878e-01 1.78284928e-01
1.99413851e-01 1.43653107e+00 1.00121975e+00 -1.74065375e+00
-6.45033062e-01 3.64936084e-01 -1.58323348e-01 4.10554528e-01
-1.96462408e-01 -1.96123961e-02 -8.18185031e-01 -9.85604405e-01
-3.47869486e-01 -6.99576497e-01 -7.05845654e-02 1.25506446e-01
-9.23612475e-01 -1.56641185e-01 9.30451751e-01 -9.79291081e-01
8.68692875e-01 -2.04705334e+00 2.92020798e-01 -2.12958515e-01
4.33497608e-01 3.76664251e-02 -3.92970502e-01 6.20456874e-01
-2.34968767e-01 1.09454818e-01 -1.87958017e-01 -1.04863870e+00
1.51650861e-01 -7.05670491e-02 -6.98648214e-01 -1.44925907e-01
3.79993170e-01 1.31575918e+00 -7.60643303e-01 -7.41111696e-01
4.47109431e-01 6.97933376e-01 -4.66607302e-01 4.07521546e-01
-6.54239714e-01 5.10606527e-01 -1.22292317e-01 5.37724793e-01
5.76165199e-01 -4.03775930e-01 -1.18643798e-01 8.90602544e-03
-2.03172773e-01 -1.52602434e-01 -9.63688731e-01 1.72733498e+00
-7.71030709e-02 1.04600501e+00 -2.70511240e-01 -5.31497419e-01
6.87156200e-01 4.35365915e-01 3.36964756e-01 -6.50387168e-01
2.25456089e-01 -2.85549104e-01 -5.18349886e-01 -5.00614226e-01
1.18418896e+00 1.04056589e-01 -3.30651086e-03 4.31241840e-01
-9.64521021e-02 -3.68472248e-01 4.07942474e-01 6.59369528e-01
5.06007373e-01 6.59061730e-01 -2.76819646e-01 2.61066407e-01
8.59789178e-02 1.71955600e-01 4.47787374e-01 6.30208075e-01
1.24055862e-01 1.10050321e+00 7.88232982e-01 -2.73385763e-01
-1.75397730e+00 -1.21828878e+00 3.43166947e-01 9.16259289e-01
1.23980440e-01 -4.36090887e-01 -9.73749816e-01 -3.13075036e-01
-3.00172120e-01 1.30932987e+00 -8.37650120e-01 1.35855094e-01
-4.55551505e-01 -1.41747847e-01 4.17756915e-01 6.64878607e-01
3.80608320e-01 -1.39583755e+00 -8.73178840e-01 2.54471768e-02
-4.96070206e-01 -1.05383384e+00 -6.07595742e-01 -2.69683599e-01
-3.19351792e-01 -5.80653012e-01 -1.20334911e+00 -8.27178359e-01
6.00927293e-01 1.76642105e-01 1.01752853e+00 5.12374565e-02
-2.68190831e-01 1.09316617e-01 -5.91793299e-01 -4.89496589e-01
-6.69698358e-01 -3.13076645e-01 -3.48679960e-01 -1.19874906e-02
-3.70658487e-01 -4.55189049e-01 -6.47652984e-01 -1.47815540e-01
-1.07194626e+00 1.20928264e+00 5.20337760e-01 7.88075507e-01
7.46655464e-01 -2.65573561e-01 8.53012502e-02 -7.01750159e-01
6.86822832e-01 -2.54923433e-01 -4.22416896e-01 3.15300256e-01
-2.99789727e-01 -1.45141883e-02 7.47761607e-01 -5.95596194e-01
-1.15753019e+00 2.03319028e-01 -2.70306747e-02 -9.84408796e-01
1.92610443e-01 3.13328952e-01 -1.50255531e-01 6.77021265e-01
3.37492496e-01 7.71446586e-01 -2.21448004e-01 -7.87388608e-02
8.76767099e-01 3.80132318e-01 9.16240335e-01 -3.71978402e-01
8.84281695e-01 2.88354009e-01 -4.73258555e-01 -6.93425715e-01
-2.35966310e-01 7.14682266e-02 -4.33666438e-01 -5.54767907e-01
1.22396457e+00 -1.02179503e+00 -5.06078303e-01 6.12551212e-01
-1.46751952e+00 -8.16721201e-01 -5.80157399e-01 1.07854292e-01
-8.14695895e-01 3.46557558e-01 -6.60632491e-01 -7.81210065e-01
-5.52236676e-01 -1.14062965e+00 1.22652507e+00 4.77979749e-01
-2.14151248e-01 -5.55575311e-01 -2.42834054e-02 2.39451751e-01
2.29554757e-01 5.17859280e-01 8.68798673e-01 5.63173741e-02
-7.93479800e-01 2.70484775e-01 -3.24208796e-01 1.01311937e-01
-2.80297309e-01 4.71431822e-01 -9.55361605e-01 -1.88115433e-01
-4.83424425e-01 -5.58930576e-01 1.01634610e+00 6.43933833e-01
1.17119563e+00 -3.56262982e-01 -1.36752486e-01 7.68152893e-01
1.26390898e+00 1.91508994e-01 9.03139412e-01 2.21954316e-01
1.06449652e+00 6.35666311e-01 4.99163598e-01 7.08135784e-01
6.37634158e-01 5.48574626e-01 5.49178004e-01 -4.75227565e-01
-5.10506511e-01 -8.90690863e-01 3.47461015e-01 6.39072001e-01
1.23435520e-01 -6.43389285e-01 -6.56825602e-01 8.06745946e-01
-2.10488200e+00 -1.18681526e+00 -1.40587285e-01 1.56058300e+00
7.89461911e-01 -4.52965908e-02 3.02652959e-02 -6.59020916e-02
7.59755969e-01 3.85738552e-01 -6.75727308e-01 -3.98037344e-01
-2.44967937e-01 -2.68077999e-01 3.83964963e-02 4.64401424e-01
-8.39868844e-01 1.39739823e+00 5.30222893e+00 8.70962977e-01
-1.18755472e+00 7.12366477e-02 8.02753747e-01 -3.87267709e-01
-7.40759254e-01 2.15260927e-02 -3.85700494e-01 4.77504492e-01
4.43127602e-01 -1.67682797e-01 3.72289330e-01 9.25001442e-01
4.73423541e-01 2.72454191e-02 -9.75828648e-01 1.04105330e+00
8.69767219e-02 -1.78164673e+00 5.29469550e-01 -1.60063937e-01
9.61834550e-01 -4.74035859e-01 3.50755155e-01 2.95253634e-01
5.70965946e-01 -1.22665322e+00 1.59134817e+00 6.87728226e-01
1.25685763e+00 -8.08140039e-01 1.81596324e-01 2.23259941e-01
-1.33403993e+00 1.15608051e-01 -3.45458627e-01 2.76509941e-01
5.81335545e-01 2.27028966e-01 -9.93194938e-01 4.23413485e-01
6.71969473e-01 7.52910435e-01 -4.26463425e-01 7.75413454e-01
-6.07178450e-01 3.16857934e-01 2.41897613e-01 -3.23053226e-02
2.64750898e-01 1.09439215e-03 5.99653304e-01 1.16359568e+00
3.91467422e-01 1.33292302e-01 4.80414070e-02 1.53198540e+00
-1.02693930e-01 5.00556491e-02 -4.54108745e-01 -2.70404220e-01
5.54255903e-01 1.27002728e+00 -6.91776097e-01 -5.98029733e-01
-1.33975998e-01 1.55905569e+00 1.91510037e-01 5.91673613e-01
-1.03784788e+00 -3.64673287e-01 4.41268295e-01 3.84049155e-02
5.93583822e-01 -8.63880739e-02 -4.08717513e-01 -1.04221618e+00
1.69313140e-02 -8.96157026e-01 -8.23744852e-03 -1.67405391e+00
-7.05747247e-01 7.67360389e-01 1.13164110e-03 -1.16735840e+00
-5.77197552e-01 -6.25615343e-02 -8.70365024e-01 9.04275417e-01
-1.27046084e+00 -1.76332581e+00 -4.67601597e-01 6.71867669e-01
1.01234770e+00 -4.90648188e-02 4.60204065e-01 -1.46352857e-01
-5.26860893e-01 6.98626995e-01 1.43875360e-01 2.31599301e-01
4.54210430e-01 -1.36126935e+00 8.71896088e-01 1.20072460e+00
1.73810184e-01 2.26373836e-01 8.95017862e-01 -9.40679491e-01
-1.16083407e+00 -1.29863381e+00 6.81433141e-01 -3.35400999e-01
1.63868606e-01 -6.15784407e-01 -6.34812772e-01 8.17583442e-01
7.78033137e-01 -5.50704300e-01 2.55271077e-01 -5.72431147e-01
-2.19331309e-01 4.33590591e-01 -6.58596039e-01 1.00931931e+00
7.20095873e-01 -3.58910501e-01 -2.92830765e-01 1.42839000e-01
1.05496979e+00 -7.02319741e-01 -4.01453733e-01 -1.66335568e-01
6.95631146e-01 -8.86731267e-01 8.32821190e-01 -3.70807797e-01
1.44465387e+00 -3.86850238e-01 2.72910278e-02 -1.29133332e+00
-3.09897065e-01 -7.92999148e-01 6.86411187e-02 1.36550629e+00
3.60546947e-01 8.27582553e-02 7.96082139e-01 4.47206467e-01
-3.43598753e-01 -7.29632854e-01 -5.88124454e-01 -3.18895698e-01
3.23787741e-02 -4.01870608e-01 6.81616843e-01 9.20449376e-01
-2.01860324e-01 3.19791824e-01 -9.05032694e-01 -1.12997264e-01
5.50141931e-01 2.23122522e-01 1.04200840e+00 -5.70989192e-01
-3.06376517e-01 -5.93491375e-01 1.29964679e-01 -1.18185699e+00
-1.25377432e-01 -7.39530444e-01 3.16729575e-01 -1.95274806e+00
6.17240548e-01 -3.90016921e-02 2.10081354e-01 6.07465327e-01
-4.50084925e-01 1.49460897e-01 7.06365824e-01 2.34405339e-01
-5.59895217e-01 1.08505487e+00 1.62558031e+00 -2.13466108e-01
-2.59653419e-01 -5.51766098e-01 -7.01429904e-01 2.94774950e-01
5.03050864e-01 -2.85708219e-01 -6.31939888e-01 -6.82427824e-01
1.11605078e-01 4.56026137e-01 5.48232019e-01 -8.12327087e-01
1.66631386e-01 -3.49696040e-01 8.27732265e-01 -1.08253741e+00
6.18424177e-01 -2.36882210e-01 3.47376287e-01 3.90524834e-01
-5.11672854e-01 2.87841678e-01 1.18542686e-01 2.69074738e-01
-1.59215461e-02 -1.70100015e-02 6.72452450e-01 -1.77211016e-01
-7.35700786e-01 2.96818614e-01 -4.79716897e-01 -2.03139201e-01
1.13751972e+00 -5.19411206e-01 -4.19293374e-01 -8.59106600e-01
-3.72161031e-01 5.48482895e-01 7.41384506e-01 6.93549573e-01
1.04104507e+00 -1.42248929e+00 -1.22004128e+00 -4.01892737e-02
3.91463265e-02 1.57401219e-01 6.32455230e-01 2.91793615e-01
-7.44396091e-01 2.20401868e-01 -3.89915437e-01 -6.58541083e-01
-1.36391330e+00 5.45307338e-01 7.60049000e-02 -2.12419838e-01
-8.90798688e-01 1.15668273e+00 5.75657904e-01 3.30359526e-02
1.98843688e-01 -2.00423077e-01 -2.55263209e-01 -1.14102364e-02
6.74010694e-01 1.00427210e-01 -6.22328103e-01 -6.66422844e-01
7.90164769e-02 3.59315544e-01 -2.31802344e-01 -6.57423556e-01
1.40807438e+00 -1.29754499e-01 2.24284604e-01 2.66803503e-01
8.27681720e-01 -1.76247269e-01 -1.89947796e+00 1.62760243e-02
-6.52616799e-01 -4.52535212e-01 -6.50998577e-02 -7.02819943e-01
-1.10919547e+00 1.07317340e+00 5.35552144e-01 -1.49588421e-01
1.00999212e+00 -1.46332964e-01 1.06019521e+00 -2.24684879e-01
-2.20875233e-01 -1.00061071e+00 6.26704395e-01 3.28434497e-01
1.31200349e+00 -9.42436099e-01 -1.74474239e-01 -2.55174965e-01
-1.36955392e+00 9.92086768e-01 7.39557028e-01 3.04582901e-02
-8.18473175e-02 1.77093551e-01 2.25661725e-01 -8.59457925e-02
-9.49032009e-01 -2.54625585e-02 2.52559245e-01 6.82315826e-01
1.79862738e-01 8.02545995e-02 6.33178055e-02 5.54826379e-01
-6.10178828e-01 -5.82499355e-02 7.15123951e-01 7.12701678e-01
-2.11916447e-01 -9.92161453e-01 -3.63817841e-01 1.79217085e-01
-8.12508538e-02 -3.93724710e-01 -4.58570629e-01 5.27585983e-01
5.57218418e-02 8.76504064e-01 2.34762833e-01 -5.16030967e-01
1.15321033e-01 -8.35759044e-02 2.59434104e-01 -4.12957072e-01
-6.09051347e-01 3.32720369e-01 -1.20459363e-01 -5.23129642e-01
7.05855787e-02 -7.51964331e-01 -1.24776256e+00 -4.89503711e-01
-3.80558730e-03 -9.75868627e-02 6.71116292e-01 5.83020389e-01
3.12722176e-01 9.58853066e-01 4.34793204e-01 -9.59792912e-01
-7.40230083e-02 -9.25748765e-01 -3.80258560e-01 5.78575909e-01
2.87477940e-01 -3.38752359e-01 1.68746978e-01 6.76055789e-01]
|
[11.178110122680664, 0.540488600730896]
|
a919056c-4241-43a3-8e3d-97b55b5d9719
|
part-aware-panoptic-segmentation
|
2106.06351
| null |
https://arxiv.org/abs/2106.06351v1
|
https://arxiv.org/pdf/2106.06351v1.pdf
|
Part-aware Panoptic Segmentation
|
In this work, we introduce the new scene understanding task of Part-aware Panoptic Segmentation (PPS), which aims to understand a scene at multiple levels of abstraction, and unifies the tasks of scene parsing and part parsing. For this novel task, we provide consistent annotations on two commonly used datasets: Cityscapes and Pascal VOC. Moreover, we present a single metric to evaluate PPS, called Part-aware Panoptic Quality (PartPQ). For this new task, using the metric and annotations, we set multiple baselines by merging results of existing state-of-the-art methods for panoptic segmentation and part segmentation. Finally, we conduct several experiments that evaluate the importance of the different levels of abstraction in this single task.
|
['Gijs Dubbelman', 'Xiaoxiao Wen', 'Chenyang Lu', 'Panagiotis Meletis', 'Daan de Geus']
|
2021-06-11
| null |
http://openaccess.thecvf.com//content/CVPR2021/html/de_Geus_Part-Aware_Panoptic_Segmentation_CVPR_2021_paper.html
|
http://openaccess.thecvf.com//content/CVPR2021/papers/de_Geus_Part-Aware_Panoptic_Segmentation_CVPR_2021_paper.pdf
|
cvpr-2021-1
|
['scene-parsing', 'part-level-panoptic-segmentation']
|
['computer-vision', 'computer-vision']
|
[ 3.56057882e-01 -3.58140767e-02 -1.20842382e-01 -6.21525407e-01
-9.70984280e-01 -8.93038988e-01 6.44590914e-01 2.73236156e-01
-7.88067728e-02 2.09598064e-01 2.42288277e-01 -1.38905495e-01
2.10268155e-01 -8.73194695e-01 -7.49116123e-01 -4.83494967e-01
2.10678428e-01 4.07991022e-01 8.51314187e-01 3.58436108e-02
1.39228910e-01 6.03743017e-01 -1.55925858e+00 5.32998741e-01
8.48317742e-01 1.05654025e+00 3.02935660e-01 7.96507120e-01
-4.18986112e-01 6.80712283e-01 -5.69858015e-01 -5.39386988e-01
4.96972889e-01 -2.50691324e-01 -1.23111606e+00 5.04588783e-01
9.51129615e-01 -1.09464020e-01 1.18862972e-01 1.06120574e+00
-1.72371715e-01 9.73925367e-02 3.84931386e-01 -1.05832553e+00
-1.76262274e-01 5.38102388e-01 -5.53910553e-01 2.08178401e-01
2.32372270e-03 3.23570102e-01 1.35637975e+00 -3.09387982e-01
5.57960510e-01 1.53515899e+00 5.53126156e-01 6.83684573e-02
-1.16460264e+00 -3.78528923e-01 5.77327728e-01 -1.41129091e-01
-1.01705658e+00 -1.05455212e-01 4.53321755e-01 -6.90876305e-01
9.39009130e-01 3.87062013e-01 6.05918646e-01 5.38054407e-01
-1.58590734e-01 1.10371995e+00 1.30734134e+00 -1.12069957e-01
1.57302797e-01 -1.42903522e-01 6.10655963e-01 6.05330408e-01
3.20628464e-01 -3.01616162e-01 4.99497205e-02 1.99369267e-01
5.20513773e-01 -1.76849887e-01 -1.16127186e-01 -4.06073838e-01
-1.09123695e+00 7.25669861e-01 5.38016200e-01 2.39645660e-01
-5.26392311e-02 4.80235927e-02 3.80473822e-01 -3.37810099e-01
6.66139066e-01 6.08988523e-01 -8.94254804e-01 2.36172061e-02
-9.31957066e-01 2.29232416e-01 8.77489448e-01 9.50205088e-01
9.51293349e-01 -3.48535925e-01 -4.03115630e-01 7.97183394e-01
3.33804905e-01 4.62708354e-01 -1.99495912e-01 -1.20968294e+00
6.77629232e-01 8.68568659e-01 7.10959733e-02 -4.26829666e-01
-5.35417438e-01 -2.53063023e-01 -4.24041748e-01 4.93729152e-02
3.74699652e-01 -5.33397347e-02 -1.53001678e+00 1.48030996e+00
6.49224997e-01 1.91392541e-01 7.81308785e-02 7.61605680e-01
1.05383861e+00 7.82653689e-01 5.28365016e-01 2.89745837e-01
1.77137125e+00 -1.62532866e+00 -3.51797819e-01 -5.09022415e-01
3.86085212e-01 -7.03476548e-01 1.16241944e+00 2.35233724e-01
-6.87507033e-01 -9.19108570e-01 -9.32371974e-01 -2.17258945e-01
-7.34262466e-01 6.72338381e-02 6.29506111e-01 5.08698225e-01
-7.73356497e-01 5.45954645e-01 -9.47221160e-01 -5.27077675e-01
2.34962717e-01 -1.56091973e-01 -1.94750652e-01 2.65185852e-02
-7.83939183e-01 6.14144504e-01 7.30873942e-01 -3.00475091e-01
-7.52844334e-01 -8.47481549e-01 -1.15096402e+00 2.96114892e-01
4.15645272e-01 -7.30879724e-01 1.52063906e+00 -7.45535970e-01
-1.45839357e+00 1.18650103e+00 2.00419631e-02 -4.59244668e-01
1.14855178e-01 -5.64201593e-01 -3.69698554e-01 4.92827386e-01
4.13454622e-01 1.15633214e+00 5.64790010e-01 -1.57327235e+00
-1.07200432e+00 -3.08466136e-01 3.35900247e-01 2.51666605e-01
4.62757379e-01 4.69570085e-02 -1.00662541e+00 -3.90588999e-01
1.17102429e-01 -7.92421937e-01 -4.17259604e-01 -2.62003392e-01
-7.01685309e-01 -5.34696467e-02 8.02914262e-01 -7.44880021e-01
9.63912845e-01 -2.36755133e+00 2.73799896e-03 -1.62142187e-01
1.71672940e-01 1.63241863e-01 -3.69406700e-01 8.81773606e-02
1.61057010e-01 4.54654783e-01 -1.12593007e+00 -7.02125013e-01
1.36134066e-02 4.48156357e-01 -2.26635456e-01 -7.80318677e-02
2.87786990e-01 7.42673159e-01 -7.73630977e-01 -6.44339681e-01
4.44530755e-01 9.10834298e-02 -4.18353707e-01 2.88157046e-01
-6.69890583e-01 6.36141717e-01 -4.41334039e-01 7.73112893e-01
1.05071342e+00 -1.77807704e-01 -9.19748619e-02 -6.99690729e-02
-4.64869291e-01 4.26113963e-01 -1.00789523e+00 1.88416684e+00
-4.24596757e-01 4.27625775e-01 1.68469474e-01 -6.56092703e-01
5.26084244e-01 -8.54141414e-02 3.82752210e-01 -6.43161178e-01
-2.94278581e-02 5.72374761e-02 -3.21642995e-01 -3.41421962e-01
8.49161804e-01 2.14040324e-01 -3.35956931e-01 1.71180014e-02
1.99912846e-01 -9.92768347e-01 6.56537414e-01 2.91289911e-02
8.63125563e-01 3.20598096e-01 4.50672984e-01 -4.15027440e-01
4.50725824e-01 4.82118756e-01 7.13925302e-01 7.84244776e-01
-3.85867387e-01 1.09428000e+00 6.64145291e-01 -3.57177973e-01
-8.87514710e-01 -1.09247303e+00 -4.03793454e-01 9.88555789e-01
4.35102671e-01 -5.13917625e-01 -9.34129119e-01 -9.68472123e-01
-9.65522006e-02 6.86064005e-01 -7.07225740e-01 6.54290378e-01
-5.51656961e-01 -1.04509676e+00 4.40333337e-01 6.46491170e-01
1.03295314e+00 -9.91216838e-01 -7.80453265e-01 -1.37985889e-02
-3.38321269e-01 -1.84640467e+00 -1.83341131e-01 1.44469664e-01
-7.78648317e-01 -1.36306107e+00 -5.47581255e-01 -5.09320855e-01
1.22302487e-01 5.30131519e-01 1.69472504e+00 -1.17025688e-01
-2.08344236e-01 4.04006988e-01 -6.36785269e-01 -4.08975929e-01
-2.92671412e-01 1.68118015e-01 -7.85264194e-01 -1.09288558e-01
4.77354936e-02 -2.69277722e-01 -6.67326868e-01 2.56697416e-01
-1.16625082e+00 4.23410296e-01 5.99064946e-01 1.13655012e-02
1.04113472e+00 2.44673342e-02 -4.16614860e-01 -1.20227432e+00
2.00224034e-02 -3.24637324e-01 -1.00914955e+00 4.66597021e-01
7.54158050e-02 -3.03559024e-02 3.97024006e-01 2.63627410e-01
-1.30408251e+00 2.60234267e-01 -4.74564552e-01 -3.10222842e-02
-7.49317586e-01 2.02181175e-01 -5.53367138e-01 1.89658105e-01
1.61476776e-01 -4.07316685e-02 -8.53463352e-01 -7.34695911e-01
8.60758960e-01 2.80108988e-01 7.84364581e-01 -4.63533401e-01
5.92185497e-01 6.84151828e-01 -8.46136510e-02 -9.09579992e-01
-1.52367771e+00 -1.05588233e+00 -1.07712066e+00 2.14532197e-01
1.51579320e+00 -1.11494839e+00 -7.15338252e-03 6.49174809e-01
-1.26651764e+00 -6.58111393e-01 -4.08946276e-01 9.80599821e-02
-4.28647459e-01 6.24375165e-01 -5.13154089e-01 -5.00295579e-01
-3.54200006e-01 -1.22460234e+00 1.73949039e+00 3.62924725e-01
1.51144281e-01 -7.70853877e-01 5.40620983e-01 5.00007093e-01
-6.95287809e-02 5.87299228e-01 9.75693047e-01 -5.67521036e-01
-7.99378991e-01 2.91440338e-01 -6.83594584e-01 4.67593431e-01
-8.21325779e-02 4.26402122e-01 -1.12661970e+00 2.93378979e-01
-1.38822094e-01 -1.03295028e-01 1.31639588e+00 4.22133803e-01
1.29874682e+00 -9.26045626e-02 -3.10102880e-01 9.04273152e-01
1.69735575e+00 1.87455416e-01 7.82592297e-01 5.27776837e-01
8.12268853e-01 9.07832742e-01 6.81282699e-01 1.03604808e-01
6.11456931e-01 5.69183648e-01 5.76226294e-01 -2.85547495e-01
-2.32845649e-01 -1.53369606e-01 -6.02663755e-02 5.99592626e-01
1.91292599e-01 -4.77587879e-01 -1.04422855e+00 7.54105747e-01
-1.94288361e+00 -5.58343530e-01 -5.36227763e-01 1.80015957e+00
4.35767382e-01 1.17935851e-01 5.17190844e-02 -1.51356265e-01
4.44209337e-01 6.11833334e-01 -1.69690847e-01 -4.31207299e-01
-2.33478129e-01 3.65450621e-01 6.03235424e-01 4.61384505e-01
-1.75214016e+00 1.38692331e+00 6.79716396e+00 5.95633566e-01
-8.78247201e-01 1.26845270e-01 8.01680148e-01 4.97164071e-01
-3.34030800e-02 3.38641405e-01 -9.94597197e-01 2.05144763e-01
7.80680716e-01 6.48395717e-01 2.31963135e-02 1.14837193e+00
-1.20039195e-01 -5.05087376e-01 -9.47384477e-01 5.58307052e-01
-8.35878551e-02 -1.04845953e+00 1.73239246e-01 -3.02918613e-01
9.42968607e-01 4.94961202e-01 -2.66167194e-01 3.58800769e-01
5.37580311e-01 -7.36632586e-01 7.79447794e-01 1.76088795e-01
3.33307624e-01 -3.14588249e-01 6.32881701e-01 1.72192343e-02
-1.70806730e+00 1.07624553e-01 -1.90978602e-01 3.06535214e-01
4.15598720e-01 7.13952184e-01 -3.67192686e-01 7.92498112e-01
9.97224987e-01 6.66565716e-01 -9.67615068e-01 1.38752997e+00
-6.42089963e-01 5.87147057e-01 -4.13512409e-01 7.09859133e-01
6.85864747e-01 -2.91380763e-01 4.91132349e-01 1.76932824e+00
-3.21289748e-02 2.10257441e-01 5.09251833e-01 9.26441669e-01
-1.66823119e-01 6.04600348e-02 -3.34591955e-01 1.54006228e-01
6.47331327e-02 1.56632674e+00 -1.16746342e+00 -7.48952329e-01
-4.68306899e-01 9.16406095e-01 7.73457885e-02 1.47924528e-01
-8.26682687e-01 -2.10176960e-01 9.87089097e-01 -2.19350353e-01
6.13400102e-01 -2.08240017e-01 -5.78269124e-01 -1.26824629e+00
-2.11174667e-01 -5.94003022e-01 5.98919511e-01 -8.83051455e-01
-9.16007459e-01 6.00510418e-01 5.38380861e-01 -9.46912587e-01
1.73738435e-01 -6.99742615e-01 -8.60157013e-01 3.58759314e-01
-1.86164284e+00 -1.42048109e+00 -6.22223675e-01 1.40305668e-01
1.01374125e+00 6.46568835e-01 5.25993407e-01 5.41384444e-02
-7.42501557e-01 -1.27098963e-01 -1.61786512e-01 2.07656056e-01
2.78978586e-01 -1.67241490e+00 1.08992589e+00 1.29812467e+00
3.69355232e-01 -5.75086437e-02 4.82133150e-01 -5.22583067e-01
-6.69718802e-01 -1.46010315e+00 7.26108015e-01 -6.69646025e-01
5.37943006e-01 -3.88604343e-01 -8.68314266e-01 6.89574420e-01
9.19257551e-02 -2.79692989e-02 3.46377552e-01 2.27577542e-03
-5.44322431e-01 7.20670000e-02 -1.12362146e+00 4.16085184e-01
1.06738698e+00 -2.89087296e-01 -6.33872509e-01 3.50542217e-01
1.42817700e+00 -4.06844229e-01 -7.76272535e-01 9.13506031e-01
2.76764005e-01 -1.22733557e+00 1.08870840e+00 -2.33245805e-01
4.01550680e-01 -3.52749646e-01 -4.05729979e-01 -1.16227102e+00
-4.03958708e-01 -1.54720306e-01 2.77810633e-01 1.32470620e+00
4.62825596e-01 -2.72608876e-01 5.52470148e-01 3.23445737e-01
-5.63854337e-01 -4.15727079e-01 -6.78359032e-01 -7.11546004e-01
6.39068484e-02 -5.27924001e-01 9.32801902e-01 6.75358236e-01
-6.92989647e-01 4.05295074e-01 6.40238896e-02 6.30235434e-01
4.09665436e-01 6.56571805e-01 9.69320238e-01 -1.37061334e+00
-2.35207647e-01 -5.46549678e-01 -1.27764910e-01 -1.11807144e+00
2.08692346e-02 -5.59843004e-01 3.82650435e-01 -2.04144454e+00
3.17379862e-01 -2.21849382e-01 -2.22661510e-01 4.98174787e-01
-3.08733433e-01 2.66818076e-01 5.32254338e-01 1.95306972e-01
-9.26376224e-01 3.74705613e-01 1.28916788e+00 -3.41026038e-01
-3.41221690e-01 -4.05999720e-02 -6.97017670e-01 1.09409153e+00
8.70446920e-01 -4.97843981e-01 -1.49909317e-01 -7.14855492e-01
-3.01694542e-01 -2.89993584e-01 2.47139797e-01 -1.26195312e+00
-3.54612201e-01 -2.98138231e-01 -1.01281226e-01 -1.03963864e+00
3.60097647e-01 -7.02963710e-01 -1.46460235e-01 2.58340180e-01
2.20620945e-01 3.42097133e-02 4.67463136e-01 3.75565439e-01
-4.16552216e-01 -6.65914491e-02 1.04474437e+00 -4.11483437e-01
-1.32190943e+00 2.47825071e-01 -2.14435250e-01 3.19084615e-01
8.69184494e-01 -1.89395640e-02 -4.99544919e-01 2.68439204e-01
-5.46255589e-01 4.85505939e-01 4.80266333e-01 5.21341681e-01
1.41633004e-01 -6.72909796e-01 -6.67786121e-01 -1.90377548e-01
4.16429311e-01 4.29506242e-01 2.96537995e-01 7.12369084e-01
-1.07195866e+00 6.86956584e-01 -1.35737002e-01 -8.13863516e-01
-1.18316722e+00 4.19242501e-01 3.13709795e-01 -7.45207906e-01
-6.52470469e-01 6.46684468e-01 9.84788060e-01 -8.10571969e-01
-6.35000616e-02 -1.19248080e+00 -4.28022802e-01 2.62190010e-02
3.20176601e-01 1.19406335e-01 1.03078902e-01 -6.40312433e-01
-3.54122758e-01 7.70343781e-01 2.94724792e-01 -2.71945130e-02
1.25686800e+00 -5.85556701e-02 -2.31942788e-01 2.83721834e-01
9.54301834e-01 -5.22408374e-02 -1.48534966e+00 -5.50096296e-02
1.38369560e-01 -2.91024595e-01 -2.26422176e-01 -9.57225204e-01
-1.07553840e+00 9.32878673e-01 4.79155034e-01 1.84615180e-01
1.13806450e+00 1.68727621e-01 8.97013426e-01 3.49918634e-01
9.08798799e-02 -1.03734553e+00 -2.71889508e-01 8.48545134e-01
6.12675309e-01 -1.30294609e+00 3.84273231e-02 -1.05729961e+00
-7.78020203e-01 7.28370428e-01 6.92669630e-01 7.12816566e-02
6.45215631e-01 4.97453548e-02 1.65867552e-01 -3.76106769e-01
-2.91878343e-01 -9.10074294e-01 6.39807284e-01 4.54747021e-01
1.10938206e-01 3.71338874e-01 -1.89575721e-02 4.09234583e-01
-2.15556279e-01 -5.01243055e-01 1.85442477e-01 8.62827301e-01
-8.39546442e-01 -9.72566903e-01 -3.71375561e-01 9.01382118e-02
-4.95932788e-01 -1.58028603e-01 -7.79410601e-01 1.16893756e+00
3.61458212e-01 9.08106089e-01 3.58194977e-01 -1.08614594e-01
4.76527900e-01 2.29849890e-02 2.68787205e-01 -8.50318849e-01
-6.50399685e-01 -2.06212662e-02 3.56015682e-01 -6.41496837e-01
-8.06645095e-01 -4.45634037e-01 -1.07279205e+00 2.02641100e-01
8.28450620e-02 1.45038605e-01 7.22351551e-01 9.04227674e-01
2.18884334e-01 4.57344294e-01 4.02364641e-01 -9.03856575e-01
1.67496696e-01 -9.53200281e-01 -4.60308790e-01 6.25183523e-01
-1.58893615e-02 -3.79547656e-01 -1.33810371e-01 9.15600061e-02]
|
[9.563271522521973, 0.43641841411590576]
|
3b42515b-0dd3-4ebb-a63e-c949a960c46f
|
video-smoke-detection-based-on-deep-saliency
|
1809.02802
| null |
http://arxiv.org/abs/1809.02802v2
|
http://arxiv.org/pdf/1809.02802v2.pdf
|
Video Smoke Detection Based on Deep Saliency Network
|
Video smoke detection is a promising fire detection method especially in open
or large spaces and outdoor environments. Traditional video smoke detection
methods usually consist of candidate region extraction and classification, but
lack powerful characterization for smoke. In this paper, we propose a novel
video smoke detection method based on deep saliency network. Visual saliency
detection aims to highlight the most important object regions in an image. The
pixel-level and object-level salient convolutional neural networks are combined
to extract the informative smoke saliency map. An end-to-end framework for
salient smoke detection and existence prediction of smoke is proposed for
application in video smoke detection. The deep feature map is combined with the
saliency map to predict the existence of smoke in an image. Initial and
augmented dataset are built to measure the performance of frameworks with
different design strategies. Qualitative and quantitative analysis at
frame-level and pixel-level demonstrate the excellent performance of the
ultimate framework.
|
['Zhong Wang', 'Yongming Zhang', 'Gao Xu', 'Jinjun Wang', 'Gaohua Lin', 'Yang Jia', 'Qixing Zhang']
|
2018-09-08
| null | null | null | null |
['fire-detection']
|
['time-series']
|
[ 6.78157628e-01 -7.49029160e-01 -1.34348586e-01 -8.36130381e-02
-5.48216760e-01 -3.47305089e-01 5.36521852e-01 6.09922931e-02
-3.57458621e-01 3.35252672e-01 2.34159395e-01 -6.30479380e-02
-5.73831387e-02 -9.48407352e-01 -4.38327521e-01 -7.74342895e-01
2.81460583e-01 -6.30025387e-01 9.94785547e-01 7.12932274e-02
6.57340646e-01 3.52549762e-01 -1.55252564e+00 4.87086773e-01
6.42728269e-01 1.04102731e+00 1.13759756e+00 9.00552750e-01
-5.21583915e-01 1.08057141e+00 -1.90101206e-01 3.54636461e-01
2.40398645e-01 -4.73074436e-01 -5.29915512e-01 9.36225951e-02
7.04412878e-01 -5.01674116e-01 -3.80538255e-01 1.18583429e+00
3.76755863e-01 1.42645270e-01 3.50384146e-01 -1.17122877e+00
-3.22607309e-01 4.83955503e-01 -5.51769555e-01 8.31583202e-01
1.35138899e-01 2.99186170e-01 6.55692399e-01 -1.35049522e+00
5.87030113e-01 8.55060577e-01 5.83365142e-01 5.92461348e-01
-5.76829851e-01 -6.13607466e-01 2.59727657e-01 2.13145345e-01
-1.11465096e+00 -1.21362738e-01 1.30692840e+00 -3.89644861e-01
3.75734538e-01 6.68725610e-01 8.61489117e-01 7.91093409e-01
2.50092708e-02 8.66654515e-01 9.97566462e-01 -8.14320669e-02
2.87787139e-01 1.06494531e-01 4.11340632e-02 1.04683053e+00
4.49320167e-01 4.26234663e-01 -6.38735533e-01 1.62667096e-01
7.27999210e-01 7.47801900e-01 -1.58546209e-01 -8.65467265e-02
-9.64232385e-01 7.84159958e-01 1.02461958e+00 2.95649290e-01
-4.07611609e-01 5.01327395e-01 2.17551619e-01 -4.67193604e-01
6.11851931e-01 1.50979199e-02 -8.19526762e-02 2.40943998e-01
-1.54819167e+00 3.56967658e-01 3.18848997e-01 5.54999828e-01
7.16510177e-01 4.93973017e-01 -5.85960984e-01 4.97914225e-01
4.83550608e-01 7.79455841e-01 5.75066134e-02 -1.00840187e+00
1.46868959e-01 7.15353429e-01 3.09421457e-02 -1.07402861e+00
-2.89812893e-01 -2.47073814e-01 -6.17471993e-01 8.87441039e-01
4.73672263e-02 -1.97674721e-01 -1.04600048e+00 1.01421034e+00
4.26753253e-01 7.88293839e-01 -9.19033810e-02 1.42363119e+00
1.47279000e+00 6.50540173e-01 5.33162117e-01 -1.04921117e-01
1.31986284e+00 -1.11248994e+00 -4.39736009e-01 -3.52427363e-01
-2.40239367e-01 -6.17746353e-01 9.40373957e-01 -2.06480354e-01
-9.22722578e-01 -6.47157550e-01 -7.59241521e-01 -1.46403280e-03
-4.05273855e-01 1.15977459e-01 4.85230744e-01 6.02526188e-01
-7.66763449e-01 5.90774536e-01 -3.52832228e-01 -3.72045934e-01
9.77728546e-01 -4.83485498e-02 2.99421072e-01 8.80766474e-03
-8.81553411e-01 6.37860477e-01 2.35776588e-01 2.94668049e-01
-1.46215618e+00 -9.85927701e-01 -5.75362027e-01 1.56917363e-01
3.04959923e-01 -7.40780711e-01 9.31737125e-01 -1.03754342e+00
-9.78367329e-01 5.84538937e-01 -2.78661162e-01 -2.74306774e-01
4.60493177e-01 -1.94078788e-01 -1.17947444e-01 6.57835841e-01
1.56054378e-01 7.09332705e-01 9.91550326e-01 -1.65971851e+00
-1.11688435e+00 1.84886053e-01 8.14418346e-02 2.31280208e-01
-1.67173013e-01 4.74042684e-01 2.00539544e-01 -6.80820465e-01
-1.36677846e-01 -5.08093715e-01 -5.25396347e-01 4.11772937e-01
-3.56839418e-01 -3.54581326e-02 1.67268097e+00 -6.12589955e-01
1.03822327e+00 -1.90483367e+00 -4.77844745e-01 3.61452885e-02
4.16730762e-01 4.93612289e-01 -4.74184044e-02 -5.86971529e-02
2.36966670e-01 2.57351965e-01 -5.07588983e-01 2.33341113e-01
-3.76204908e-01 -3.07614625e-01 -2.16232240e-01 2.15666130e-01
6.34033084e-01 8.61303687e-01 -1.19654500e+00 -8.14398825e-01
6.73785925e-01 5.47169566e-01 -2.34331474e-01 9.15955156e-02
-4.94285703e-01 2.40467414e-01 -9.33301747e-01 8.50271583e-01
7.29543924e-01 7.77234808e-02 -3.94425124e-01 -2.07077220e-01
-7.24747896e-01 -4.29008417e-02 -8.52779388e-01 1.26970553e+00
-2.03152359e-01 9.64453161e-01 2.26791680e-01 -1.81287646e-01
8.14435601e-01 3.07893753e-01 8.05290759e-01 -4.80839252e-01
3.13966811e-01 -1.26110852e-01 -6.03221714e-01 -9.38397050e-01
5.89590251e-01 -2.90026873e-01 2.98738301e-01 1.45342410e-01
-4.49424952e-01 -3.01760882e-01 1.34411097e-01 1.32399648e-01
1.01751995e+00 6.05952740e-02 -2.48920456e-01 -3.93381327e-01
3.13507110e-01 3.43949080e-01 3.67537946e-01 6.93113744e-01
-6.98289335e-01 9.32747185e-01 -2.43261352e-01 -4.27391261e-01
-8.31507742e-01 -1.23888230e+00 1.55452579e-01 1.10929847e+00
8.10120583e-01 -2.02898413e-01 -7.50084817e-01 -8.19370389e-01
-2.05584764e-01 8.18380117e-01 -5.19474387e-01 -1.09680615e-01
-4.97610867e-01 -4.33406800e-01 1.05287060e-01 6.58020973e-01
9.16581452e-01 -1.51384473e+00 -1.35179555e+00 -5.92958331e-02
-2.22560182e-01 -8.62169743e-01 -4.56374317e-01 1.33356228e-01
-6.87747538e-01 -1.24000108e+00 -5.13929665e-01 -1.18066657e+00
4.11391228e-01 1.20371604e+00 8.84120107e-01 5.09158492e-01
-6.42727673e-01 2.76008934e-01 -3.29093814e-01 -7.13195860e-01
-2.39869699e-01 -2.67887771e-01 -5.17549813e-01 -3.09409708e-01
8.56916085e-02 -2.48261914e-01 -1.17631543e+00 -1.73149884e-01
-1.07228959e+00 3.24425131e-01 4.85906869e-01 2.46120051e-01
2.16478676e-01 2.47098491e-01 3.64940822e-01 -6.55694544e-01
4.39397454e-01 -4.20145154e-01 -4.19415057e-01 1.81151330e-01
-3.57102960e-01 -3.18412453e-01 2.11771592e-01 1.95885990e-02
-1.13902104e+00 5.62419236e-01 8.96028578e-02 -7.29717612e-01
-3.66669893e-01 1.54637799e-01 1.49227872e-01 -1.22946575e-01
6.71836376e-01 2.98588485e-01 -3.72129083e-01 -2.53094912e-01
3.07973862e-01 4.17240977e-01 4.70483184e-01 -3.85349952e-02
1.19763553e+00 6.17089212e-01 1.45674888e-02 -1.01635289e+00
-9.93584454e-01 -7.81196117e-01 -6.72766984e-01 -8.92424524e-01
1.10453296e+00 -9.68572974e-01 -6.00721478e-01 8.64989832e-02
-9.95965779e-01 -3.10417771e-01 -3.33691090e-01 1.55447856e-01
-6.69125095e-02 3.08175683e-01 -4.58144337e-01 -1.22674930e+00
-9.59626734e-01 -9.49396610e-01 1.29319370e+00 7.99239099e-01
2.55061865e-01 -6.97646677e-01 -1.56803742e-01 2.74848670e-01
6.55888379e-01 5.27481437e-01 3.60754430e-01 1.60830572e-01
-9.73274767e-01 1.05588481e-01 -5.43734193e-01 9.81426165e-02
1.17627285e-01 5.35432220e-01 -1.25596213e+00 2.38983601e-01
-1.23581693e-01 2.86100030e-01 1.46668661e+00 6.75592244e-01
9.70818281e-01 -1.62039146e-01 -3.44381928e-01 8.39684755e-02
1.91837084e+00 9.99739841e-02 4.67573315e-01 2.85915285e-01
1.04254270e+00 2.82887429e-01 7.18078136e-01 3.19863915e-01
-1.89490154e-01 2.57950693e-01 9.85957026e-01 -3.99241716e-01
-5.84122300e-01 -8.92967172e-03 7.14391530e-01 2.46696368e-01
-1.33012578e-01 3.75970080e-02 -6.28643394e-01 5.70217133e-01
-1.55271125e+00 -1.49378347e+00 -5.37971437e-01 1.64874363e+00
4.48598117e-01 5.92235709e-03 3.91017258e-01 -1.15434676e-01
9.21943426e-01 3.82533789e-01 -2.85358161e-01 1.68224536e-02
-1.98262006e-01 1.87780857e-01 5.34726381e-01 3.97984892e-01
-1.63246918e+00 1.02156210e+00 6.34215689e+00 6.26215875e-01
-1.32772791e+00 2.25169480e-01 5.31173289e-01 -2.51773864e-01
-3.35535675e-01 1.16473474e-01 -4.90385562e-01 6.03766620e-01
5.61581850e-01 7.88609684e-03 2.58217543e-01 9.42292035e-01
7.55296171e-01 -3.88915122e-01 -2.04148054e-01 5.31189620e-01
-1.40639514e-01 -1.73617351e+00 -1.54488906e-01 -5.59788465e-01
8.70484352e-01 3.24622214e-01 -5.10626519e-03 -2.35713795e-01
1.70804203e-01 -6.93976700e-01 1.07321250e+00 7.92003334e-01
3.84990811e-01 -6.15937769e-01 3.68602514e-01 5.47124743e-02
-1.62159336e+00 -3.60196114e-01 -4.32688177e-01 -5.81112020e-02
9.06594098e-02 7.63202071e-01 -9.51737642e-01 3.04403067e-01
6.46132410e-01 7.84956872e-01 -8.58458996e-01 1.45632660e+00
-1.39905825e-01 8.45941603e-01 -1.44527033e-01 -3.56097490e-01
4.04883295e-01 -2.36860037e-01 6.22457325e-01 1.62405956e+00
3.77311170e-01 -6.58804774e-02 5.36879659e-01 1.25094879e+00
1.28724962e-01 -1.77508757e-01 -7.42737472e-01 2.88554188e-02
1.03417695e-01 1.61839974e+00 -1.15969074e+00 -4.55053538e-01
-3.00148338e-01 7.36823082e-01 -2.66967416e-01 2.03172147e-01
-9.81698394e-01 -1.04240105e-01 6.04343295e-01 4.06927109e-01
5.02480924e-01 -1.46778449e-01 -4.91312444e-01 -5.25724709e-01
-2.83068210e-01 -6.70714155e-02 2.70111561e-01 -1.12737072e+00
-9.20915365e-01 3.85814935e-01 -3.33078295e-01 -1.24202847e+00
3.41631979e-01 -4.29031372e-01 -1.32746267e+00 4.81275052e-01
-1.66666245e+00 -1.22844613e+00 -1.02595830e+00 3.07181805e-01
1.00967610e+00 1.33335784e-01 5.10806322e-01 1.74240425e-01
-5.77195704e-01 -2.86315084e-01 7.66394939e-03 2.69036025e-01
8.76455233e-02 -1.11055267e+00 2.86751799e-02 1.38950276e+00
3.15160245e-01 6.66568875e-02 7.45895445e-01 -9.39230978e-01
-1.12766397e+00 -1.82285774e+00 5.08575261e-01 -4.68937039e-01
1.91593066e-01 -2.16032088e-01 -6.01902485e-01 -1.85202375e-01
2.39541233e-01 1.75881714e-01 -1.64615246e-03 -5.66625297e-01
2.80098356e-02 -3.20386477e-02 -1.22636235e+00 7.76244164e-01
1.00491977e+00 -5.35920739e-01 -2.62437612e-01 4.53679740e-01
7.28742719e-01 -4.33488116e-02 -2.81374216e-01 3.56865048e-01
4.80685323e-01 -9.06142235e-01 1.32254267e+00 -3.73790085e-01
9.63597655e-01 -6.66688979e-01 -1.32796749e-01 -8.25900912e-01
-7.47866452e-01 -1.71720996e-01 -9.96697322e-02 1.22991097e+00
8.83003026e-02 4.98449206e-01 1.12829852e+00 3.55423898e-01
-1.99544072e-01 -4.83682543e-01 -7.45337725e-01 -1.20679848e-01
-4.14318025e-01 -3.63574356e-01 1.00814827e-01 6.50041461e-01
-1.80480748e-01 1.96313843e-01 -2.47194931e-01 3.62003446e-01
5.14343500e-01 5.76672018e-01 2.13920787e-01 -1.11055005e+00
5.36297560e-01 -7.49667287e-01 -3.06879282e-01 -1.39311746e-01
-1.21601313e-01 -9.22686160e-01 6.26670361e-01 -1.97649670e+00
7.26562679e-01 -5.87387122e-02 -8.14244032e-01 3.29500437e-01
-8.14630091e-01 5.24513602e-01 5.76033294e-01 -1.93650976e-01
-8.16912830e-01 4.82744962e-01 1.28641248e+00 -5.66566527e-01
-8.68615434e-02 1.61377266e-01 -5.73992968e-01 7.43733048e-01
1.22642589e+00 -5.64725399e-01 -2.73694664e-01 -1.43289149e-01
-4.62033004e-01 -2.07202807e-01 1.11195135e+00 -1.42526364e+00
2.08404452e-01 -7.05729723e-01 6.57217860e-01 -7.74614394e-01
2.10420057e-01 -1.05008125e+00 -8.11092556e-02 7.40573764e-01
-4.26253796e-01 -1.60256058e-01 6.98087141e-02 8.59822810e-01
1.53360739e-01 -3.72602820e-01 1.02642488e+00 -2.87154317e-01
-1.37795126e+00 3.13325286e-01 -6.25655472e-01 -2.04206139e-01
1.08641934e+00 -4.16593522e-01 -3.37244481e-01 -1.46364808e-01
-4.40229118e-01 -2.35584319e-01 2.28237212e-01 5.58486164e-01
1.11872697e+00 -1.29511511e+00 -6.21626198e-01 1.64544652e-03
-3.69237438e-02 -1.24421671e-01 4.03461218e-01 6.08554125e-01
-5.85628092e-01 -8.81205127e-02 -2.26583898e-01 -5.67017376e-01
-1.74923086e+00 6.42399251e-01 2.57557958e-01 2.98213750e-01
-9.18819726e-01 8.27994764e-01 3.15859646e-01 2.39500016e-01
-9.67534557e-02 -7.22446322e-01 -2.77418226e-01 -2.77244151e-01
6.40113950e-01 3.90985638e-01 -4.82360125e-01 -6.37036502e-01
-4.58861291e-01 4.49952722e-01 6.08145177e-01 2.75274575e-01
1.47218585e+00 -3.35148759e-02 7.11521283e-02 3.12305719e-01
8.60665441e-01 5.04642958e-03 -1.76311648e+00 8.16495493e-02
4.33146060e-02 -7.99695790e-01 4.66903210e-01 -7.28177547e-01
-1.57478154e+00 8.37381780e-01 1.16951382e+00 1.13408782e-01
1.04368925e+00 -2.36650884e-01 7.84499168e-01 -1.08067952e-01
-7.32791722e-02 -1.01697791e+00 2.32634321e-01 2.39969179e-01
7.11794019e-01 -1.34258497e+00 1.87009886e-01 -8.35292935e-01
-6.09428167e-01 1.08105218e+00 5.66376984e-01 -2.89569736e-01
8.58730257e-01 6.01320207e-01 2.40656391e-01 -5.34270883e-01
-4.65308636e-01 -7.31506169e-01 3.00434977e-01 6.33612216e-01
1.40523031e-01 2.95538604e-02 1.24330260e-01 4.44633514e-01
3.93610001e-01 5.35984039e-02 2.58271694e-01 1.04098272e+00
-1.38738489e+00 -2.83591330e-01 -4.69353855e-01 6.40121460e-01
-2.47719064e-01 -1.98435262e-01 -5.98437726e-01 3.21720719e-01
4.34421152e-01 1.33619761e+00 -1.54856071e-01 -4.26369190e-01
5.63568212e-02 -1.56909198e-01 2.56310165e-01 -4.64361608e-01
-9.29240584e-01 8.21624249e-02 -2.66861413e-02 -2.44566187e-01
-7.84910917e-01 -4.95627314e-01 -1.42732334e+00 -2.17975434e-02
-4.79122907e-01 -5.43393791e-02 7.21943140e-01 8.51895332e-01
-2.65763272e-02 8.73527765e-01 7.15714455e-01 -1.15649259e+00
1.65724918e-01 -6.63605988e-01 -6.16824090e-01 5.98172069e-01
3.51783276e-01 -5.56253254e-01 -9.78737473e-02 3.26838255e-01]
|
[9.757658958435059, -0.4506576657295227]
|
044ae7ad-4a75-435d-9096-3c11d6501956
|
a-neural-framework-for-learning-subgraph-and
|
2112.13143
| null |
https://arxiv.org/abs/2112.13143v3
|
https://arxiv.org/pdf/2112.13143v3.pdf
|
GREED: A Neural Framework for Learning Graph Distance Functions
|
Among various distance functions for graphs, graph and subgraph edit distances (GED and SED respectively) are two of the most popular and expressive measures. Unfortunately, exact computations for both are NP-hard. To overcome this computational bottleneck, neural approaches to learn and predict edit distance in polynomial time have received much interest. While considerable progress has been made, there exist limitations that need to be addressed. First, the efficacy of an approximate distance function lies not only in its approximation accuracy, but also in the preservation of its properties. To elaborate, although GED is a metric, its neural approximations do not provide such a guarantee. This prohibits their usage in higher order tasks that rely on metric distance functions, such as clustering or indexing. Second, several existing frameworks for GED do not extend to SED due to SED being asymmetric. In this work, we design a novel siamese graph neural network called GREED, which through a carefully crafted inductive bias, learns GED and SED in a property-preserving manner. Through extensive experiments across 10 real graph datasets containing up to 7 million edges, we establish that GREED is not only more accurate than the state of the art, but also up to 3 orders of magnitude faster. Even more significantly, due to preserving the triangle inequality, the generated embeddings are indexable and consequently, even in a CPU-only environment, GREED is up to 50 times faster than GPU-powered baselines for graph / subgraph retrieval.
|
['Sayan Ranu', 'Yogish Sabharwal', 'Venkatesan Chakaravarthy', 'Sourav Medya', 'Siddharth Grover', 'Rishabh Ranjan']
|
2021-12-24
| null | null | null | null |
['graph-similarity']
|
['graphs']
|
[ 6.73572272e-02 3.93418744e-02 -2.31667668e-01 -2.26919547e-01
-5.02173960e-01 -6.51870430e-01 2.76477486e-01 6.22337878e-01
-4.97100711e-01 6.58004761e-01 -2.29150075e-02 -4.00265574e-01
-4.67537820e-01 -1.13481116e+00 -8.04558516e-01 -5.59501529e-01
-2.79010028e-01 5.61032772e-01 1.56940520e-01 -1.85546771e-01
2.30400473e-01 6.29031420e-01 -1.50238121e+00 -3.27853203e-01
9.56317425e-01 1.00019813e+00 -1.61660001e-01 4.00468588e-01
-1.43932492e-01 4.48365301e-01 -2.40007475e-01 -8.03697467e-01
3.39126170e-01 -2.35771805e-01 -8.73617530e-01 -3.61054778e-01
6.23601854e-01 -4.06643867e-01 -6.88650668e-01 1.20542514e+00
5.11835515e-01 2.48535529e-01 6.14815474e-01 -1.37758982e+00
-1.01662552e+00 7.90746927e-01 -4.34252411e-01 8.24252367e-02
2.44498059e-01 -1.75775126e-01 1.48954380e+00 -6.73117936e-01
5.63192904e-01 8.29823434e-01 8.35384607e-01 3.15097123e-01
-1.32817161e+00 -4.79295224e-01 1.54614329e-01 1.00433052e-01
-1.58101952e+00 4.43715081e-02 8.06421280e-01 -8.59475732e-02
1.04194939e+00 3.86714518e-01 7.67051935e-01 8.61651659e-01
4.14443240e-02 6.50851607e-01 8.47198427e-01 -1.96450844e-01
2.10843444e-01 3.78834419e-02 1.63387433e-01 9.29274678e-01
6.46717310e-01 3.91461551e-02 -2.57378161e-01 -2.51598746e-01
4.33296084e-01 -7.41594285e-02 -4.50979769e-01 -7.32010603e-01
-1.04363334e+00 1.01079273e+00 7.01064348e-01 4.28433150e-01
-4.04372290e-02 3.22908729e-01 6.71668470e-01 5.43895543e-01
5.75434923e-01 6.21784151e-01 -1.98144168e-01 -2.60961652e-01
-8.38124931e-01 2.48235732e-01 1.05650330e+00 8.65161657e-01
7.97772825e-01 -1.63131803e-01 7.70599172e-02 5.72980881e-01
-7.58497864e-02 2.55264789e-01 1.54536203e-01 -5.07548451e-01
4.44173723e-01 7.79141903e-01 -4.00497258e-01 -1.80954838e+00
-4.04282957e-01 -6.45222783e-01 -1.12754560e+00 -8.18187743e-02
5.42159796e-01 3.52814019e-01 -5.05228400e-01 1.98324919e+00
2.88813710e-01 1.44410029e-01 -4.34665680e-01 8.28326046e-01
5.12448013e-01 4.40670580e-01 -2.17387721e-01 8.10364112e-02
1.04938853e+00 -8.09754670e-01 -3.12782913e-01 -6.21160083e-02
8.62516522e-01 -2.84569114e-01 1.13609231e+00 3.71931046e-01
-1.00714803e+00 -2.25355461e-01 -1.21446967e+00 -3.61005574e-01
-7.10585475e-01 -2.42299095e-01 9.53225970e-01 7.26360559e-01
-1.22700918e+00 9.87328649e-01 -8.47792089e-01 -4.95878667e-01
5.01867712e-01 4.59398925e-01 -5.57188988e-01 -2.43766204e-01
-1.15709817e+00 7.94138014e-01 4.53008145e-01 6.14801832e-02
-3.31787378e-01 -7.49917388e-01 -9.29228902e-01 4.35052484e-01
4.04813290e-01 -6.39082491e-01 5.86516261e-01 -6.44711614e-01
-1.21338654e+00 7.77292669e-01 5.82911074e-02 -7.16353536e-01
4.91696656e-01 -6.56931940e-03 -1.93446279e-01 1.59622610e-01
-1.59311384e-01 3.32547843e-01 4.47132379e-01 -8.21528256e-01
-3.52703243e-01 -4.08239901e-01 5.40735006e-01 -1.31716160e-02
-9.15307641e-01 -2.06180081e-01 -5.67665398e-01 -6.56140566e-01
-2.64254566e-02 -1.03827202e+00 -1.35667369e-01 3.43403608e-01
-3.32080513e-01 -3.91103506e-01 4.11266237e-01 -4.53390926e-01
1.49983072e+00 -2.08300352e+00 2.26215154e-01 4.24409360e-01
7.36087739e-01 3.20055008e-01 -2.42272153e-01 7.15742052e-01
5.51989749e-02 1.83519885e-01 -4.58121687e-01 -1.95023686e-01
3.74571979e-01 1.57691747e-01 -1.55315027e-01 7.16161251e-01
1.86885491e-01 9.59534109e-01 -9.71429288e-01 -2.89355397e-01
1.10139310e-01 6.19040906e-01 -8.21247935e-01 -3.66773531e-02
5.30887069e-03 -1.30383760e-01 -3.62663716e-01 4.65525597e-01
7.60004461e-01 -4.52271342e-01 1.93455532e-01 -3.02943677e-01
1.69791698e-01 2.12811142e-01 -1.11336386e+00 1.81019723e+00
-4.24870193e-01 5.47423005e-01 -2.27555886e-01 -1.36188138e+00
8.10909748e-01 -1.38440743e-01 5.24263024e-01 -7.40777194e-01
1.99550614e-01 3.85314971e-01 -1.41800314e-01 -1.19551204e-01
5.74516296e-01 2.20271260e-01 3.57368067e-02 7.22349167e-01
-1.25575617e-01 1.94479331e-01 5.19574642e-01 4.37673151e-01
1.45504642e+00 -5.06576747e-02 1.50268823e-01 -3.56373370e-01
3.72782975e-01 -1.97305322e-01 3.68033260e-01 6.36883378e-01
-7.26327375e-02 5.46252012e-01 6.85410202e-01 -4.08825248e-01
-9.87988114e-01 -1.07800293e+00 -3.79672907e-02 1.06321526e+00
2.49757007e-01 -5.75058460e-01 -7.47658253e-01 -7.04875708e-01
2.11457819e-01 3.41453880e-01 -6.26347601e-01 -4.86679196e-01
-6.47830009e-01 -6.15544677e-01 7.19948471e-01 5.75329423e-01
2.27217063e-01 -5.85167587e-01 -2.67263710e-01 2.87812233e-01
1.88356578e-01 -1.13561344e+00 -8.23948801e-01 1.27206251e-01
-8.09289396e-01 -1.14990878e+00 -6.32772982e-01 -7.94013739e-01
8.08082700e-01 3.64454627e-01 1.27671838e+00 4.30059105e-01
-2.82413155e-01 1.49975166e-01 -2.65616447e-01 2.67938934e-02
-1.10744528e-01 4.55819845e-01 1.54966507e-02 -1.56750485e-01
4.77523059e-01 -1.05769980e+00 -5.69685400e-01 8.13344941e-02
-9.65986073e-01 -1.65289015e-01 6.61318064e-01 8.50466490e-01
5.78191698e-01 1.99162886e-01 4.33396667e-01 -1.13625717e+00
8.10342550e-01 -4.00973648e-01 -7.78341711e-01 3.06077331e-01
-9.80375111e-01 3.78596932e-01 9.74203646e-01 -2.98705906e-01
-3.68367702e-01 -1.79637149e-01 -7.02274144e-02 -3.54465663e-01
1.18128583e-01 8.44933093e-01 -2.22639292e-02 -2.38715649e-01
5.88342965e-01 2.19138309e-01 1.55096814e-01 -2.52646178e-01
3.63038689e-01 3.07521343e-01 4.18841183e-01 -7.36458659e-01
1.06771541e+00 3.95867020e-01 2.60175496e-01 -6.07972980e-01
-7.99711883e-01 -2.70066440e-01 -4.62193727e-01 1.60227403e-01
5.50172389e-01 -5.22064447e-01 -8.79748225e-01 1.95063010e-01
-8.97427678e-01 -1.31245360e-01 -1.41339645e-01 4.27001834e-01
-4.03223127e-01 7.11614192e-01 -7.70509303e-01 -5.90426803e-01
-4.51062918e-01 -9.95737791e-01 7.96948373e-01 -6.19316101e-02
-9.95874926e-02 -1.13569224e+00 -2.26325914e-02 1.07429989e-01
6.99541986e-01 6.26065910e-01 1.25990510e+00 -7.99670041e-01
-6.73109055e-01 -3.47799897e-01 -5.94205379e-01 2.19723478e-01
6.47061914e-02 -3.39577906e-02 -5.93277454e-01 -6.05427027e-01
-3.45840544e-01 -4.21302408e-01 8.19222152e-01 9.60763767e-02
1.45317495e+00 -3.06680143e-01 -2.44458035e-01 8.70185733e-01
1.54627275e+00 -2.34924898e-01 4.09416139e-01 2.58378804e-01
7.61411369e-01 5.05582213e-01 3.87865126e-01 3.03959429e-01
5.88620424e-01 5.43655574e-01 5.87982535e-01 -9.20480415e-02
-8.76321271e-02 -3.71683002e-01 1.71179086e-01 1.17500138e+00
-1.10421389e-01 -3.97977173e-01 -6.97927296e-01 5.69520175e-01
-1.59467721e+00 -8.08543801e-01 -1.07637063e-01 2.41508746e+00
7.90698528e-01 1.81776941e-01 1.33118436e-01 3.33264440e-01
5.89565396e-01 2.57128179e-01 -6.33011818e-01 -6.50946319e-01
-2.03798994e-01 4.80330378e-01 5.78736961e-01 2.55344599e-01
-9.78828549e-01 6.64445460e-01 4.75040054e+00 8.36642742e-01
-1.04716623e+00 -1.58046558e-01 3.14947426e-01 1.12941051e-02
-5.51438689e-01 -1.08354650e-02 -3.57320040e-01 3.60271305e-01
8.27425003e-01 -4.48123157e-01 9.67808843e-01 7.55901277e-01
-2.84210563e-01 2.54906237e-01 -1.40945578e+00 1.10036361e+00
2.19523340e-01 -1.18370891e+00 -2.98875067e-02 3.98294151e-01
5.33234537e-01 1.29027918e-01 2.73455959e-02 3.59519839e-01
1.57798901e-01 -1.20887816e+00 5.59124231e-01 2.12931097e-01
7.15643167e-01 -1.15857255e+00 6.75468028e-01 1.32817268e-01
-1.25881314e+00 2.34762087e-01 -6.64205074e-01 8.98590032e-03
5.75383380e-03 9.02891695e-01 -4.51409549e-01 7.94787288e-01
5.18376887e-01 7.49191880e-01 -5.68467438e-01 9.76351798e-01
-1.04890361e-01 4.49803233e-01 -5.35597265e-01 -2.63512582e-01
4.66692895e-01 -3.46249789e-01 4.03570324e-01 1.08441937e+00
4.12214190e-01 -2.00535223e-01 3.34538706e-02 9.45407391e-01
-4.97677803e-01 1.42651618e-01 -8.48067820e-01 -4.81413692e-01
7.19608963e-01 1.17039311e+00 -6.73442304e-01 1.72069967e-02
-4.53506887e-01 1.11979496e+00 9.09614027e-01 2.13450909e-01
-1.10759449e+00 -9.34907794e-01 7.03517020e-01 -2.08911039e-02
3.81089091e-01 -4.56648320e-01 -7.47814700e-02 -1.02156150e+00
3.95709366e-01 -8.61679375e-01 4.48323458e-01 -1.16006017e-01
-1.52987206e+00 7.02529788e-01 -2.15280607e-01 -9.49595213e-01
-4.30762619e-02 -8.19900632e-01 -2.40721017e-01 5.72063386e-01
-1.63254726e+00 -9.51968908e-01 -3.66937518e-01 4.64233607e-01
-9.23338011e-02 2.03828752e-01 9.02096808e-01 7.34375417e-01
-4.99600649e-01 9.53952730e-01 1.49487659e-01 2.56386340e-01
4.86558050e-01 -1.36835492e+00 5.75008035e-01 6.93649471e-01
5.31898916e-01 7.54486740e-01 6.61815822e-01 -2.41214171e-01
-1.90206242e+00 -1.06380892e+00 8.29946518e-01 -1.71402037e-01
8.55136096e-01 -5.87689042e-01 -1.05797875e+00 5.30572057e-01
-1.00342490e-01 3.66075903e-01 5.00749469e-01 3.03199023e-01
-5.55474579e-01 -2.60171890e-01 -9.27122891e-01 6.15792394e-01
1.50917435e+00 -7.42715001e-01 -1.53989524e-01 3.12811047e-01
6.92143321e-01 -4.25947487e-01 -1.16318643e+00 3.64752680e-01
5.78016639e-01 -1.14779770e+00 8.59687984e-01 -5.65450668e-01
3.36552441e-01 -1.59209967e-01 -1.38358876e-01 -1.20441389e+00
-2.26763144e-01 -4.09385264e-01 -3.40322793e-01 1.33565152e+00
3.58204156e-01 -9.94777322e-01 9.38098133e-01 5.26402950e-01
7.55062476e-02 -1.23289013e+00 -9.40347970e-01 -1.09998536e+00
1.76676080e-01 -3.46871674e-01 8.17300856e-01 1.12169337e+00
-1.74672350e-01 2.01233238e-01 -2.93175966e-01 1.41969055e-01
5.64408004e-01 3.25872719e-01 7.31887579e-01 -1.40663064e+00
-3.41756284e-01 -7.27760553e-01 -7.17392027e-01 -1.07148087e+00
3.97279650e-01 -1.30971980e+00 -2.78095514e-01 -1.48942137e+00
1.52302817e-01 -6.57687664e-01 -4.18113530e-01 3.78220916e-01
5.83067872e-02 3.62382174e-01 -1.35925218e-01 -1.95817247e-01
-4.21025008e-01 7.59192884e-01 8.91170561e-01 -2.57031411e-01
8.04933384e-02 -2.83566058e-01 -7.71347821e-01 4.43646878e-01
8.33171487e-01 -5.23162425e-01 -4.76831198e-01 -7.28086472e-01
6.13258064e-01 -3.15099508e-01 2.78994799e-01 -9.45867658e-01
3.96047831e-01 1.80961400e-01 -2.11519524e-01 -2.22197965e-01
2.20102698e-01 -7.63763547e-01 1.82157867e-02 4.30713952e-01
-2.29807213e-01 3.16804796e-01 -1.12956464e-01 7.93809891e-01
-3.02212030e-01 -5.00417203e-02 5.60947597e-01 2.57885605e-01
-4.25316155e-01 7.76656866e-01 1.79594219e-01 3.79325122e-01
8.51714253e-01 -2.86826134e-01 -2.67489851e-01 -3.21750164e-01
-1.61073253e-01 9.19720531e-03 5.20684958e-01 3.42524022e-01
4.88030732e-01 -1.38224840e+00 -6.17736399e-01 -5.47083728e-02
2.06087247e-01 4.85588759e-02 1.76169835e-02 1.02894783e+00
-6.39243662e-01 3.97046536e-01 9.36506763e-02 -2.61320472e-01
-1.00454235e+00 8.18403184e-01 4.35341187e-02 -5.73030293e-01
-8.29019845e-01 7.36317754e-01 -4.05936204e-02 -4.61404711e-01
4.43598509e-01 -4.24874395e-01 2.32866883e-01 -1.21839710e-01
3.33376318e-01 4.97144192e-01 2.85584480e-01 -3.98589909e-01
-4.19364065e-01 5.43104410e-01 -3.99459712e-02 4.95788693e-01
1.50563955e+00 3.30286808e-02 -2.93715984e-01 8.16131160e-02
1.41837466e+00 1.95495114e-01 -8.45066309e-01 -2.29800865e-01
1.40484899e-01 -5.32452166e-01 2.16418654e-02 -3.65190178e-01
-1.33450460e+00 9.13286388e-01 2.33966202e-01 4.73847389e-01
1.18577504e+00 -2.07398266e-01 1.23154759e+00 5.85049629e-01
5.62047243e-01 -9.43959892e-01 -1.87250659e-01 3.79902482e-01
6.62515640e-01 -1.05780470e+00 8.91217068e-02 -4.58709180e-01
-7.53196850e-02 1.07506990e+00 4.52595323e-01 -3.26114118e-01
4.26603138e-01 2.27520198e-01 -4.71181780e-01 -2.45878741e-01
-5.03723681e-01 -1.69802189e-01 3.08835536e-01 5.11523068e-01
4.50126976e-01 -2.08221003e-01 -4.98745292e-01 3.27055484e-01
-3.59461695e-01 -1.82327017e-01 1.56984493e-01 6.10907376e-01
-1.07196100e-01 -1.12550569e+00 2.74180919e-01 6.18459404e-01
-3.84395808e-01 -1.65367201e-01 -3.32294106e-01 9.87319529e-01
-2.30608016e-01 7.04264700e-01 -1.11286923e-01 -3.85529548e-01
3.24506789e-01 -4.38635468e-01 4.95862037e-01 -2.83871382e-01
-4.27560836e-01 -8.11739564e-01 1.37069121e-01 -7.54324019e-01
-1.02849424e-01 -1.46866366e-01 -1.19945407e+00 -7.50725508e-01
-4.70334470e-01 1.52145594e-01 5.02871811e-01 7.02549517e-01
5.41103721e-01 2.92718649e-01 5.75886190e-01 -5.51239729e-01
-9.32127118e-01 -5.10191023e-01 -6.82472050e-01 4.48364675e-01
-4.36687376e-03 -6.12122715e-01 -5.48776627e-01 -6.91614687e-01]
|
[7.107424736022949, 6.133922576904297]
|
f9482924-e0f5-4174-926d-e8f77f13544a
|
hindi-history-note-generation-with
| null | null |
https://aclanthology.org/2020.aacl-srw.7
|
https://aclanthology.org/2020.aacl-srw.7.pdf
|
Hindi History Note Generation with Unsupervised Extractive Summarization
|
In this work, the task of extractive single document summarization applied to an education setting to generate summaries of chapters from grade 10 Hindi history textbooks is undertaken. Unsupervised approaches to extract summaries are employed and evaluated. TextRank, LexRank, Luhn and KLSum are used to extract summaries. When evaluated intrinsically, Luhn and TextRank summaries have the highest ROUGE scores. When evaluated extrinsically, the effective measure of a summary in answering exam questions, TextRank summaries performs the best.
|
['Meet Chetan Gadoya', 'Dhruv Mathew', 'Dhineshkumar Ramasubbu', 'Aayush Shah']
|
2020-12-01
| null | null | null |
asian-chapter-of-the-association-for
|
['unsupervised-extractive-summarization']
|
['natural-language-processing']
|
[ 2.56340533e-01 6.19200408e-01 -4.78866786e-01 -1.81295395e-01
-1.47246575e+00 -7.33220398e-01 9.34668362e-01 1.00427318e+00
-4.48148489e-01 1.52479804e+00 1.20926654e+00 -1.86700836e-01
-7.17141628e-01 -5.23810804e-01 -4.12031233e-01 -1.98791817e-01
2.76973397e-01 4.47759509e-01 4.86908015e-03 -2.34826952e-01
1.05996633e+00 1.83414623e-01 -1.54175663e+00 2.11634919e-01
1.60809350e+00 2.23121896e-01 1.24251582e-01 1.62269521e+00
-2.45952785e-01 1.34284806e+00 -1.46392143e+00 -7.08438933e-01
-5.34697175e-01 -9.08067405e-01 -1.18887258e+00 -2.55167991e-01
1.16774178e+00 -5.56484580e-01 -4.47526544e-01 8.98497522e-01
6.34954333e-01 6.10745907e-01 1.24741173e+00 -7.02362120e-01
-4.74818259e-01 1.49197888e+00 1.05898092e-02 4.68123943e-01
1.22099411e+00 -4.01048750e-01 1.43977702e+00 -4.30316061e-01
1.03072071e+00 1.02686477e+00 -5.63696818e-03 2.84842581e-01
-1.01378047e+00 -3.31361331e-02 -3.44513297e-01 1.12102348e-02
-6.80770457e-01 -2.51661539e-01 7.68598676e-01 -2.36759737e-01
9.35252726e-01 7.27100432e-01 7.66060829e-01 9.53763723e-01
6.95233226e-01 1.45132589e+00 8.37797761e-01 -3.61790478e-01
3.08217585e-01 6.84612012e-03 6.99143708e-01 4.36776459e-01
7.77957678e-01 -5.58184624e-01 -1.01264358e+00 -8.76346901e-02
-4.30556722e-02 -4.50492084e-01 -1.21243432e-01 3.99130583e-01
-1.17506123e+00 9.04994726e-01 -1.68439627e-01 1.27516240e-01
-4.69081759e-01 1.00356748e-03 4.45919365e-01 3.64013702e-01
1.51627377e-01 1.35943103e+00 1.16839549e-02 -4.56168920e-01
-1.62757885e+00 8.90448213e-01 1.13961947e+00 1.21407366e+00
3.20688784e-01 1.39105683e-02 -9.95834053e-01 5.74338317e-01
1.63867876e-01 5.52643478e-01 8.74945164e-01 -1.28637338e+00
6.20948732e-01 6.99954748e-01 -1.74365848e-01 -8.58951449e-01
-9.30656791e-02 -4.43468511e-01 -5.71892500e-01 -2.97339737e-01
-7.49350861e-02 -2.82923281e-01 -6.94763064e-01 9.16533172e-01
-1.47910044e-01 -4.48706090e-01 7.32413709e-01 1.38431385e-01
2.09338284e+00 9.06615555e-01 3.67426574e-02 -4.09538060e-01
1.12343073e+00 -1.05027902e+00 -1.16240501e+00 9.43616312e-03
7.15367436e-01 -1.21238625e+00 5.48630238e-01 6.20969653e-01
-1.71887231e+00 -5.20311594e-01 -1.39602649e+00 -3.79098237e-01
-4.50278640e-01 3.86615872e-01 2.05204830e-01 3.85014653e-01
-9.83402908e-01 8.90432656e-01 -3.41364950e-01 -1.16912156e-01
1.12732708e-01 1.70989171e-01 -1.20676450e-01 -6.02337644e-02
-1.09078360e+00 7.79380083e-01 7.86282718e-01 -7.82444358e-01
-6.98099315e-01 -7.57774234e-01 -1.00884056e+00 1.85742632e-01
1.80176385e-02 -7.58970976e-01 1.60905504e+00 2.55809575e-01
-1.64038169e+00 5.60435832e-01 2.25604460e-01 -6.82842970e-01
4.69227165e-01 -4.41960633e-01 2.04064623e-01 6.99636638e-01
3.30628693e-01 5.54669738e-01 1.52852461e-01 -7.98452556e-01
-8.03051353e-01 1.12134732e-01 2.70950813e-02 6.70387030e-01
-3.20488036e-01 -1.22572206e-01 -2.29083095e-02 -5.49260318e-01
3.25675346e-02 -3.61360878e-01 -1.50667354e-01 -1.38290465e+00
-7.83057213e-01 -8.38567197e-01 5.88863552e-01 -9.10829186e-01
1.70308530e+00 -1.24184322e+00 1.25280380e-01 1.16251193e-01
4.39883769e-01 1.53579488e-01 5.60684949e-02 8.23888958e-01
5.51124692e-01 1.80537045e-01 1.53122410e-01 4.34836894e-02
3.22319895e-01 -7.24329511e-05 -2.56035119e-01 -2.55882591e-01
-2.17098862e-01 8.61214697e-01 -1.43876517e+00 -1.25162780e+00
1.10957764e-01 -5.12937792e-02 -4.95199025e-01 3.97221476e-01
-1.47247970e-01 4.04399112e-02 -7.19993830e-01 3.47026885e-01
7.16998577e-02 4.78717200e-02 -1.78548530e-01 2.40084633e-01
-2.25018546e-01 7.13648498e-01 -1.03459656e+00 1.63355732e+00
-3.15500915e-01 1.13951385e+00 -6.87408745e-01 -5.58390260e-01
1.06526971e+00 3.85870993e-01 1.69569775e-01 -1.99673519e-01
-8.60056728e-02 3.06047529e-01 -2.65753508e-01 -5.35893738e-01
1.57290304e+00 6.19879186e-01 -5.52138209e-01 6.33708477e-01
7.57696867e-01 -9.85971034e-01 1.30081844e+00 1.29620194e+00
1.25552237e+00 1.24311537e-01 8.13935876e-01 -4.08764094e-01
6.56421483e-01 2.76866466e-01 -1.60164624e-01 1.16724110e+00
1.70882016e-01 8.23955417e-01 7.04957962e-01 2.26955920e-01
-7.24822402e-01 -1.29277265e+00 1.77771077e-01 1.25445676e+00
-3.80232096e-01 -1.02943158e+00 -9.02743876e-01 -7.68304348e-01
-2.31263369e-01 1.31320679e+00 -2.68593907e-01 -1.73253998e-01
-4.11301553e-01 -1.09260954e-01 6.33555233e-01 2.02266365e-01
2.66995132e-01 -1.17044806e+00 -7.71710277e-01 2.95638233e-01
-3.78574520e-01 -6.33256614e-01 -4.66564626e-01 1.53288975e-01
-9.28262532e-01 -6.97479546e-01 -1.00760353e+00 -7.25911677e-01
6.41649723e-01 -2.34910250e-01 1.44711745e+00 -8.01623613e-02
5.06699495e-02 6.59923077e-01 -3.64058852e-01 -7.87064791e-01
-7.12589741e-01 8.19071352e-01 -3.45713705e-01 -1.06533921e+00
1.28307402e-01 -2.36555696e-01 -4.42884684e-01 -7.05356121e-01
-1.06356168e+00 -5.08068390e-02 6.89051449e-01 7.12978184e-01
2.94649452e-01 -1.40888229e-01 7.57597387e-01 -1.08302784e+00
1.54447210e+00 -2.95912568e-02 -1.17131911e-01 3.68575543e-01
-6.36683881e-01 3.92645001e-01 7.09719598e-01 -2.98312038e-01
-1.20992482e+00 -3.77789319e-01 6.36846796e-02 4.08480883e-01
-1.42517537e-01 8.32828939e-01 1.25413373e-01 3.28989446e-01
8.59639525e-01 5.07325649e-01 -6.28545642e-01 -6.10170774e-02
3.21605414e-01 5.69815636e-01 7.14444876e-01 -6.78723991e-01
3.20254117e-01 -3.04738760e-01 -6.78510070e-02 -1.17982638e+00
-1.30468154e+00 -7.69389331e-01 -8.18881452e-01 -6.64330244e-01
8.05723906e-01 -5.36789596e-01 -4.93576050e-01 -1.07441247e-01
-1.27791858e+00 2.57979870e-01 -7.42042422e-01 9.19532061e-01
-6.68968260e-01 4.39818382e-01 -6.33313835e-01 -7.13945687e-01
-8.33668530e-01 -9.97219801e-01 8.83075356e-01 1.09762037e+00
-8.07102978e-01 -1.09737527e+00 4.61665958e-01 7.20834136e-01
1.42842978e-02 4.83953297e-01 8.10343623e-01 -1.34013653e+00
-1.65755168e-01 -4.52007741e-01 3.51761222e-01 2.53863722e-01
-1.16157532e-01 4.34733152e-01 -6.09638274e-01 9.06662121e-02
-5.58575213e-01 -6.21717155e-01 1.01805544e+00 6.40586972e-01
6.84799075e-01 -1.01884460e+00 6.85307533e-02 7.57437944e-02
1.10989892e+00 -3.28515172e-02 3.92096400e-01 3.81061554e-01
5.38694739e-01 6.15874410e-01 6.16106093e-01 4.38382059e-01
4.78416234e-01 -1.63504153e-01 -1.76353917e-01 6.01300716e-01
-3.93126696e-01 -5.56376457e-01 4.11606014e-01 1.92634106e+00
-9.57148597e-02 -3.44549745e-01 -8.66513908e-01 7.14802563e-01
-1.46371460e+00 -1.18202472e+00 -3.56699824e-01 1.80533016e+00
1.16083848e+00 5.40153384e-01 -7.09079877e-02 2.54471213e-01
1.75160497e-01 3.41840833e-01 -2.25523137e-04 -9.61007357e-01
-2.16905832e-01 4.80124712e-01 2.43880227e-01 5.26871622e-01
-7.47021496e-01 8.45391572e-01 6.88477135e+00 9.82141078e-01
-2.07931891e-01 -4.83121127e-01 3.53791833e-01 -7.31288046e-02
-4.77096826e-01 -1.96011942e-02 -1.13403368e+00 1.87415525e-01
1.32688355e+00 -1.06500685e+00 -2.96054542e-01 8.02899122e-01
-1.17359690e-01 -5.95217645e-01 -1.05073917e+00 4.54190969e-01
4.57032353e-01 -1.40250862e+00 6.22624397e-01 -6.12167716e-02
1.65113711e+00 -5.78624725e-01 3.04372031e-02 5.65539181e-01
9.33442056e-01 -1.00395179e+00 3.55539143e-01 8.12663198e-01
5.84554346e-03 -1.34974384e+00 1.02044761e+00 3.58403116e-01
-5.86797833e-01 3.69008392e-01 -6.60454929e-01 -4.15164232e-02
-1.31330505e-01 5.82782626e-01 -1.06971943e+00 8.60609770e-01
9.25100595e-02 5.92225671e-01 -1.07647455e+00 1.20758593e+00
-7.58582115e-01 9.41827953e-01 6.64139241e-02 -1.00309706e+00
6.35415375e-01 -2.60697454e-01 9.07177091e-01 1.57116163e+00
3.60074103e-01 3.21818888e-01 1.45629644e-01 2.78337210e-01
-3.62170994e-01 6.01139903e-01 -6.26598597e-01 -2.95494914e-01
3.95945311e-01 1.24773049e+00 -7.01370358e-01 -9.38751578e-01
3.74714315e-01 6.68303549e-01 -1.09198101e-01 2.04997554e-01
-1.44752368e-01 -1.13652837e+00 -1.96557283e-01 -4.53358710e-01
-1.19646341e-01 4.43108790e-02 -5.08082211e-01 -7.18727350e-01
-5.10000050e-01 -1.02530479e+00 6.52012646e-01 -8.54888320e-01
-6.87166810e-01 3.43027204e-01 4.26795334e-01 -1.12252784e+00
-8.31078112e-01 -1.48613915e-01 -1.10041976e+00 3.40899676e-01
-1.02304983e+00 -4.61401552e-01 -2.37771481e-01 -1.57748640e-01
9.25620437e-01 -5.96508205e-01 6.02180421e-01 -3.05756450e-01
-2.05375493e-01 5.96485078e-01 5.45713603e-01 -9.37918723e-02
6.80588961e-01 -1.96818185e+00 -8.39152560e-02 7.75977075e-01
1.98658317e-01 5.83021820e-01 1.26935387e+00 -9.26808476e-01
-1.11354268e+00 -8.42991829e-01 1.47452343e+00 -4.30168331e-01
4.76515174e-01 4.65187788e-01 -6.75159812e-01 2.44387895e-01
1.00341046e+00 -1.16915202e+00 1.00908148e+00 -2.89933443e-01
3.52928489e-01 1.40125677e-01 -9.02566850e-01 7.08607793e-01
2.94389367e-01 -1.12656936e-01 -1.36477220e+00 6.35233521e-01
6.29904866e-01 -6.56874299e-01 -1.38142836e+00 -9.05631930e-02
4.67921942e-01 -5.61602533e-01 8.56339812e-01 -6.58953786e-01
1.41304827e+00 1.09417148e-01 1.63344413e-01 -1.71292949e+00
7.56553784e-02 -7.68067241e-01 -6.37724340e-01 1.57619226e+00
5.99316299e-01 8.45827907e-02 9.12575424e-01 2.43240133e-01
-1.53512672e-01 -6.66068316e-01 -3.77126604e-01 -4.55183446e-01
2.40311146e-01 2.42767110e-01 3.44217032e-01 6.32528067e-01
3.48865628e-01 7.81397402e-01 4.88609355e-03 -5.55430591e-01
9.07653689e-01 -3.00114229e-02 8.06858301e-01 -1.43261576e+00
3.65185052e-01 -9.78128552e-01 -4.17557210e-01 -9.57087457e-01
1.76330477e-01 -9.17432129e-01 2.87390620e-01 -2.29260182e+00
3.62879276e-01 6.79464102e-01 -9.47772618e-03 -2.49744594e-01
-5.14595687e-01 -5.11795431e-02 1.42300338e-01 -1.54578850e-01
-1.11664772e+00 3.55695784e-01 1.42141426e+00 -4.82063800e-01
-2.87676036e-01 1.91634998e-01 -7.98897743e-01 6.92662656e-01
9.03363109e-01 -4.53870595e-01 -5.25319338e-01 6.49025887e-02
3.41380209e-01 2.28905171e-01 -4.58425105e-01 -1.06402147e+00
8.07719409e-01 -1.88945740e-01 6.94942892e-01 -1.44399667e+00
-1.92201778e-01 2.59935051e-01 -5.91008246e-01 3.34658921e-01
-1.38851547e+00 2.84836978e-01 -1.40761241e-01 5.89014068e-02
-4.98125166e-01 -1.03592300e+00 3.39352220e-01 -2.53077656e-01
-7.86452517e-02 -2.56428123e-01 -8.41501772e-01 6.02252722e-01
4.12474811e-01 -7.66312703e-02 -3.95478696e-01 -8.44198585e-01
-3.09532464e-01 4.16701525e-01 -1.31211177e-01 6.59592226e-02
9.72103357e-01 -1.23171425e+00 -1.40373492e+00 -6.66802585e-01
-1.28267929e-01 -1.12802431e-01 1.06426468e-02 4.80303705e-01
-8.09684753e-01 9.49051261e-01 -2.34235704e-01 -2.27858946e-01
-1.52199078e+00 -3.97185087e-01 -6.43785954e-01 -1.07185316e+00
-4.38564748e-01 6.27003849e-01 -3.60899478e-01 -4.76569653e-01
4.23377067e-01 -3.11437696e-01 -1.12917042e+00 6.57606840e-01
9.35728192e-01 1.22376525e+00 1.95967332e-01 -3.23615223e-01
4.08261061e-01 1.69159323e-01 -3.32010746e-01 -3.48953456e-01
1.27503455e+00 3.63779068e-02 1.61722507e-02 4.95950729e-01
1.07383096e+00 3.07264805e-01 -4.92614269e-01 1.94637328e-01
4.87451792e-01 1.80641070e-01 2.56509840e-01 -8.26513767e-01
-2.36361563e-01 5.04729211e-01 -2.21944824e-01 4.64191288e-01
7.42658138e-01 -1.53045416e-01 6.32503092e-01 9.78313684e-01
-4.66083169e-01 -1.66187477e+00 3.52961481e-01 8.60160589e-01
8.78298581e-01 -1.04212594e+00 7.99113870e-01 2.24362418e-01
-7.99906313e-01 1.38310218e+00 5.38571060e-01 -1.27860785e-01
-1.46902621e-01 -1.64131880e-01 -2.78337061e-01 -1.76383227e-01
-8.30891013e-01 -5.85137904e-02 1.21129751e+00 3.60339224e-01
9.75860476e-01 5.21725602e-02 -1.04434454e+00 5.22674680e-01
-1.31673133e+00 -5.25909066e-01 1.41711748e+00 9.03454781e-01
-8.48021805e-01 -5.27985215e-01 -3.53326291e-01 1.07231283e+00
-9.11392927e-01 -1.02618031e-01 -9.65936959e-01 6.51396215e-01
-7.89398551e-01 1.15210605e+00 -3.72347206e-01 9.17318761e-02
3.09136689e-01 5.42634390e-02 5.91251791e-01 -7.92840183e-01
-1.06839609e+00 -2.51385093e-01 1.93568796e-01 2.13066921e-01
-2.81904191e-01 -6.03513718e-01 -1.26977742e+00 -2.34300286e-01
-3.62058371e-01 1.01037443e+00 7.83750951e-01 8.80705178e-01
-2.95719087e-01 1.04873276e+00 3.66276950e-01 -4.17817205e-01
-8.15927088e-01 -1.34598494e+00 -3.96124750e-01 1.35217503e-01
3.21211100e-01 4.65893522e-02 -4.51486379e-01 1.76927865e-01]
|
[12.527356147766113, 9.496230125427246]
|
e1335512-eaea-4ee7-8c99-118e1f5d7ded
|
symbolic-music-generation-conditioned-on
|
2203.16165
| null |
https://arxiv.org/abs/2203.16165v2
|
https://arxiv.org/pdf/2203.16165v2.pdf
|
Symbolic music generation conditioned on continuous-valued emotions
|
In this paper we present a new approach for the generation of multi-instrument symbolic music driven by musical emotion. The principal novelty of our approach centres on conditioning a state-of-the-art transformer based on continuous-valued valence and arousal labels. In addition, we provide a new large-scale dataset of symbolic music paired with emotion labels in terms of valence and arousal. We evaluate our approach in a quantitative manner in two ways, first by measuring its note prediction accuracy, and second via a regression task in the valence-arousal plane. Our results demonstrate that our proposed approaches outperform conditioning using control tokens which is representative of the current state of the art.
|
['Paula Viana', 'Matthew E. P. Davies', 'Serkan Sulun']
|
2022-03-30
| null | null | null | null |
['music-generation', 'music-generation']
|
['audio', 'music']
|
[ 3.92186612e-01 -3.80894803e-02 -8.63262564e-02 -2.28715107e-01
-1.04899693e+00 -9.27235305e-01 5.74067295e-01 1.15374528e-01
-2.41179615e-01 7.59932339e-01 2.41106614e-01 5.47442555e-01
-3.21317792e-01 -7.50783801e-01 -3.81920040e-01 -4.71293718e-01
-2.15377629e-01 5.00995517e-01 -3.04604769e-01 -5.56381881e-01
4.54113096e-01 3.31366181e-01 -1.88641798e+00 5.52169323e-01
4.73064750e-01 1.40689158e+00 -5.42057335e-01 6.74160361e-01
1.94428295e-01 7.92948008e-01 -8.50051165e-01 -6.66260540e-01
1.59177735e-01 -6.99835300e-01 -6.28684044e-01 -4.18480992e-01
4.68526959e-01 4.32332993e-01 4.45575058e-01 1.04044652e+00
6.70280755e-01 3.41646403e-01 7.49994636e-01 -1.39770842e+00
-3.21935713e-01 9.80298042e-01 -7.76998103e-02 -4.61161256e-01
7.08531022e-01 -8.33957791e-02 1.67623305e+00 -7.12825537e-01
7.44209588e-01 9.25876796e-01 9.83248353e-01 6.23396456e-01
-1.57750225e+00 -8.19442153e-01 -1.98945031e-01 8.26577768e-02
-1.32740355e+00 -3.91369194e-01 1.32103753e+00 -4.88059759e-01
7.19289899e-01 5.10692000e-01 1.04920197e+00 1.18066907e+00
-3.37079018e-01 6.12207472e-01 1.53417253e+00 -5.74839056e-01
4.12441254e-01 2.52536014e-02 -2.39584163e-01 3.52708519e-01
-2.89011359e-01 3.03546131e-01 -9.59021807e-01 -2.70547479e-01
6.19990945e-01 -7.37751782e-01 1.38091490e-01 -1.72971085e-01
-1.47493434e+00 7.38177240e-01 1.46172941e-01 5.05817235e-01
-5.01418948e-01 5.14585674e-01 6.66505039e-01 3.76139641e-01
3.70054662e-01 1.19188511e+00 -4.22853529e-01 -6.83588386e-01
-1.51351166e+00 5.51906705e-01 8.85345280e-01 5.64132035e-01
5.02186596e-01 2.36872733e-01 -3.83011281e-01 9.09909487e-01
5.39564341e-02 3.34960818e-01 5.66657722e-01 -1.25052845e+00
-3.33185755e-02 2.55993217e-01 3.31180319e-02 -7.28483915e-01
-2.75742143e-01 -3.91623825e-01 -5.42653501e-01 5.02289832e-01
3.28589350e-01 1.66243613e-02 -3.61952007e-01 2.14993691e+00
-1.77206516e-01 4.28661376e-01 8.98727104e-02 8.24171245e-01
6.63800597e-01 2.62909740e-01 8.84839296e-02 -4.95275736e-01
1.18085492e+00 -6.19079232e-01 -8.64202857e-01 4.15359139e-01
2.03066379e-01 -7.84471333e-01 1.45066452e+00 1.12857950e+00
-1.34739995e+00 -6.26700819e-01 -1.14214885e+00 1.32612437e-01
-3.16505939e-01 2.52765924e-01 9.26617861e-01 7.48055696e-01
-9.28611338e-01 9.98565793e-01 -2.78352201e-01 8.54919106e-02
1.70427307e-01 3.19724441e-01 -6.28061742e-02 8.07659388e-01
-1.41375554e+00 6.81436658e-01 3.10415000e-01 -2.45313063e-01
-6.48508012e-01 -6.65774226e-01 -6.12590134e-01 -1.89894438e-02
-1.43199898e-02 -3.71118873e-01 1.42889929e+00 -1.14699197e+00
-1.98840749e+00 1.12363851e+00 2.68754840e-01 -2.95052469e-01
3.82610142e-01 -2.40004539e-01 -4.76337343e-01 1.19325384e-01
-5.40358871e-02 6.32411182e-01 6.84300780e-01 -1.35722387e+00
-3.44280154e-01 -1.37822554e-01 -5.44717833e-02 -6.08344972e-02
-4.57632303e-01 1.83485821e-02 -2.97346804e-02 -1.14205158e+00
-1.36512652e-01 -1.04746878e+00 7.29762986e-02 -3.69924337e-01
-4.45659250e-01 -1.66477606e-01 8.35492536e-02 -2.00567827e-01
1.43583417e+00 -2.23996425e+00 4.76562500e-01 5.04567742e-01
-1.74103037e-01 -2.42942810e-01 -1.03925653e-01 4.38306868e-01
-2.37662479e-01 1.68857768e-01 -2.20195398e-01 -4.47992951e-01
5.67954242e-01 -3.13907005e-02 -7.56541193e-01 5.31519428e-02
1.97394758e-01 8.38955760e-01 -8.93338859e-01 -7.64101088e-01
-4.84334268e-02 4.28457558e-01 -7.96671569e-01 1.35371774e-01
-4.05809611e-01 4.52724010e-01 -1.97223693e-01 8.66619110e-01
-4.51336540e-02 3.74418318e-01 1.00999884e-01 -2.11725041e-01
-2.77212143e-01 4.14355099e-01 -1.25727189e+00 2.22925949e+00
-3.19034934e-01 4.68757898e-01 -2.95284420e-01 -7.61040807e-01
1.26363945e+00 6.24592602e-01 7.23078012e-01 -3.60695153e-01
2.73551822e-01 3.07849139e-01 -2.11428568e-01 -2.06585348e-01
8.13862860e-01 -8.04046392e-01 -7.45080590e-01 5.09940743e-01
3.47500682e-01 -7.37173915e-01 4.78142887e-01 -2.71834910e-01
7.98291564e-01 5.38855433e-01 3.43642950e-01 -1.49159253e-01
4.27244186e-01 -1.24757580e-01 4.33149189e-01 4.73950446e-01
-1.20588228e-01 5.52701414e-01 8.15799952e-01 1.19272806e-01
-8.60001981e-01 -1.14666009e+00 -1.27725109e-01 1.30810630e+00
-4.44935054e-01 -7.54758775e-01 -7.17581928e-01 -2.36432597e-01
-3.17167677e-02 8.17873478e-01 -8.76814067e-01 -9.87776816e-02
-3.91093016e-01 -4.10647392e-01 1.23814392e+00 5.47205746e-01
-1.26931638e-01 -1.69922137e+00 -6.69796228e-01 1.47703245e-01
-3.31041962e-01 -8.20163608e-01 -5.83176091e-02 3.29063535e-01
-9.71397817e-01 -7.93316960e-01 -4.22022998e-01 -5.30223429e-01
-1.28405690e-01 -8.64504635e-01 1.48891151e+00 -3.23909312e-01
-3.15040767e-01 3.91425371e-01 -4.35026348e-01 -7.06448138e-01
-2.97854155e-01 1.51312292e-01 1.16269402e-01 1.24405466e-01
6.30938485e-02 -1.12722659e+00 -2.14433908e-01 -1.29127696e-01
-8.19819152e-01 -1.52766153e-01 4.88205642e-01 7.03166842e-01
1.02037084e+00 -2.78116375e-01 8.78438711e-01 -7.04241037e-01
9.71269548e-01 -8.20478275e-02 -2.95400530e-01 3.45429266e-03
-6.49349928e-01 1.03366666e-01 4.21408117e-01 -7.64214694e-01
-6.36481941e-01 2.84962356e-01 -2.54019469e-01 -5.30836642e-01
-1.49412349e-01 5.18130422e-01 8.01143143e-03 1.28201276e-01
7.09062278e-01 -1.09102950e-01 -4.60924804e-01 -3.68867815e-01
8.00111771e-01 3.32762122e-01 1.05750585e+00 -1.10955632e+00
6.41848266e-01 2.11749002e-01 3.51301968e-01 -3.58325005e-01
-7.74359465e-01 -2.21789137e-01 -8.47723544e-01 -6.23531044e-01
7.60126829e-01 -6.40641391e-01 -1.24638951e+00 1.94759458e-01
-7.88016856e-01 -2.93181866e-01 -9.35566783e-01 4.74762887e-01
-1.28533351e+00 -4.62517478e-02 -6.52009010e-01 -1.04587185e+00
-4.63976890e-01 -5.38424134e-01 1.14562845e+00 -1.33487269e-01
-8.69302571e-01 -7.37723053e-01 7.15772569e-01 6.54095113e-02
2.64002979e-01 9.28889334e-01 7.54574835e-01 -5.61616838e-01
-7.45205348e-03 -1.76597729e-01 3.20547014e-01 3.70375156e-01
-2.62493551e-01 1.93663657e-01 -1.32608569e+00 1.87195167e-01
-1.69516265e-01 -8.29567313e-01 6.55650020e-01 1.42076254e-01
1.00702357e+00 1.44649833e-01 3.64133984e-01 5.95901251e-01
1.39490199e+00 -1.44278422e-01 6.22022688e-01 4.46820080e-01
3.36903900e-01 4.51167434e-01 9.62095737e-01 7.02092350e-01
-1.00497110e-02 8.28725040e-01 3.63644600e-01 1.11692205e-01
-4.04343754e-03 -4.03178275e-01 3.75493407e-01 9.62085724e-01
-6.25221848e-01 2.87317038e-01 -5.20726264e-01 5.03814697e-01
-1.75741601e+00 -1.34759319e+00 8.28667264e-03 2.05121303e+00
1.20676816e+00 1.47594765e-01 5.52418888e-01 1.06687391e+00
2.75979996e-01 -7.67359808e-02 -1.05774574e-01 -7.70833731e-01
-3.41522723e-01 9.69062984e-01 -1.02920584e-01 2.03249037e-01
-1.08398509e+00 8.59582007e-01 7.32088518e+00 7.09719062e-01
-1.26390612e+00 -1.10832900e-01 4.84544784e-02 -5.37107766e-01
-2.26988971e-01 -2.60609806e-01 -2.19172031e-01 1.32402718e-01
8.24874341e-01 -3.81216407e-02 7.33237624e-01 8.39682400e-01
-1.40276775e-01 2.88157821e-01 -1.25791192e+00 1.21284842e+00
3.04028004e-01 -9.39265549e-01 -8.86049345e-02 -1.93290904e-01
7.02026129e-01 -4.99901444e-01 3.07256639e-01 5.05488157e-01
-1.08628511e-01 -1.07709193e+00 1.34271193e+00 8.93372297e-01
1.20974553e+00 -9.35534954e-01 3.23099196e-01 -2.04341978e-01
-1.25043643e+00 2.25104429e-02 3.56159806e-01 -4.61777866e-01
1.82062879e-01 4.75105405e-01 -4.14595932e-01 3.29357833e-01
5.51159680e-01 6.89404845e-01 -4.44643229e-01 7.13787854e-01
-4.79447454e-01 8.20008337e-01 -1.11764528e-01 -2.71715168e-02
-3.10844760e-02 -1.68141380e-01 5.38236558e-01 1.53646576e+00
4.59703028e-01 1.98570769e-02 -7.61239305e-02 1.16558242e+00
-6.34939149e-02 4.70068842e-01 -3.11124116e-01 -3.56694430e-01
3.88602078e-01 1.43411207e+00 -6.62985206e-01 -4.93422329e-01
1.70586586e-01 8.14720094e-01 1.50229439e-01 -1.26757458e-01
-8.70945275e-01 -5.41465342e-01 5.43412745e-01 -3.55346292e-01
3.41289252e-01 3.94613333e-02 -7.91937947e-01 -1.12928462e+00
-1.28731176e-01 -8.68506193e-01 3.06916147e-01 -9.97964740e-01
-1.40164185e+00 7.47476757e-01 -7.67352581e-02 -1.39606285e+00
-6.53441846e-01 -6.14767790e-01 -5.45134366e-01 6.28847301e-01
-1.24841404e+00 -1.16169584e+00 -1.67852134e-01 6.87229216e-01
-1.20247811e-01 -2.08537579e-01 1.60166025e+00 4.10313189e-01
-3.88772488e-02 7.17193484e-01 -3.68195027e-01 1.54349534e-02
1.11721778e+00 -1.57905161e+00 -3.39521328e-03 1.55427977e-01
6.88658893e-01 3.43421876e-01 8.75920117e-01 -1.71652228e-01
-1.01240754e+00 -5.93339980e-01 9.96340156e-01 -5.30308902e-01
7.52177179e-01 -4.44353223e-01 -5.14758706e-01 4.11168665e-01
-4.83186133e-02 -1.91794261e-01 1.21337438e+00 5.53328097e-01
-6.01490676e-01 -4.10768315e-02 -1.08242595e+00 3.60032797e-01
9.18563485e-01 -7.59495318e-01 -8.71513963e-01 -1.73046082e-01
3.12140673e-01 -3.09249938e-01 -1.25355995e+00 5.98045647e-01
1.05863392e+00 -1.05360806e+00 9.93576825e-01 -5.81927419e-01
7.67645299e-01 -2.13573799e-01 -4.77142900e-01 -1.33939934e+00
-2.92135656e-01 -7.40805089e-01 -4.04903032e-02 1.49010837e+00
4.35900122e-01 -2.37147454e-02 6.25196099e-01 -1.65511090e-02
-8.37205425e-02 -7.15908408e-01 -9.09669280e-01 -6.85524762e-01
8.48540887e-02 -1.04145765e+00 6.73258483e-01 1.10450447e+00
4.80487615e-01 3.93763334e-01 -4.58606660e-01 -4.73727643e-01
5.17791152e-01 4.70230669e-01 5.79158962e-01 -1.60120726e+00
-5.85078776e-01 -9.38186288e-01 -6.20187223e-01 2.33888477e-02
2.73460418e-01 -1.04182029e+00 7.96767697e-02 -9.74113882e-01
6.35453537e-02 -2.95035720e-01 -1.00996244e+00 5.53456604e-01
2.88928151e-01 1.05228555e+00 5.17103136e-01 2.36896276e-02
-7.26390302e-01 4.48873192e-01 8.58845770e-01 1.11749431e-03
-2.49371663e-01 -1.66876510e-01 -5.97052395e-01 7.41131842e-01
9.09821033e-01 -5.45021474e-01 -5.49078695e-02 4.01414573e-01
8.09749961e-01 9.95869711e-02 4.06607270e-01 -1.21937335e+00
-1.78806826e-01 -1.33780073e-02 2.18273312e-01 -4.82962132e-01
9.19056594e-01 -3.73404980e-01 2.31577948e-01 1.55864909e-01
-7.70574033e-01 -2.93092174e-03 2.27411017e-01 7.64212757e-02
-4.98014629e-01 -1.15658522e-01 6.44370019e-01 9.18947607e-02
-3.64048183e-01 -3.41552049e-01 -7.64391497e-02 2.15602323e-01
6.29197657e-01 1.90602355e-02 1.62366137e-01 -4.24725235e-01
-1.04496109e+00 -5.16083360e-01 2.99919248e-01 3.93935919e-01
3.59501839e-01 -1.89449871e+00 -8.29667151e-01 2.30848268e-02
3.79631966e-01 -9.06810582e-01 -1.60146385e-01 7.98779368e-01
2.11095493e-02 1.61610708e-01 -6.76791191e-01 -3.57352436e-01
-1.40363252e+00 4.81525749e-01 8.57195556e-02 -4.29932892e-01
-1.34297937e-01 6.98256314e-01 -4.53098446e-01 -3.89330417e-01
3.37854773e-01 -3.55497152e-01 -2.78151959e-01 5.52023292e-01
2.64067501e-01 3.32668006e-01 -1.25174135e-01 -7.52874374e-01
-2.19319955e-01 6.98147655e-01 8.75231266e-01 -9.15839732e-01
1.35615849e+00 6.13531053e-01 -4.03814495e-01 1.39879954e+00
8.16757441e-01 3.96255106e-01 -6.21558189e-01 2.34043971e-01
-3.45926322e-02 -2.45590732e-01 -1.25133395e-01 -1.25501966e+00
-7.24231780e-01 7.66990423e-01 4.77196395e-01 2.87890255e-01
1.29993010e+00 -1.66784704e-01 5.87106109e-01 4.01574612e-01
2.73928523e-01 -1.40082204e+00 2.20315367e-01 4.81661409e-01
1.11489522e+00 -8.63794446e-01 -1.08564809e-01 -5.74808344e-02
-8.55336487e-01 1.03542387e+00 1.15115188e-01 -3.45109403e-01
4.71911550e-01 2.89911658e-01 2.56897777e-01 -2.11348772e-01
-7.69349396e-01 -3.04411054e-01 5.99562764e-01 4.54366028e-01
1.09795952e+00 2.60080040e-01 -5.40326238e-01 1.32173657e+00
-9.81470525e-01 2.03630298e-01 2.77437985e-01 7.54829645e-01
-1.10471986e-01 -1.33557999e+00 -5.22466660e-01 1.33144692e-01
-7.61471272e-01 -1.21191286e-01 -8.59013140e-01 7.19319999e-01
4.63662028e-01 8.73707712e-01 -3.01179767e-01 -7.35953212e-01
5.73118269e-01 5.69014370e-01 8.28034461e-01 -4.28029031e-01
-1.10281146e+00 2.03900382e-01 2.17942700e-01 -4.45516378e-01
-7.12633133e-01 -8.41742992e-01 -1.26809192e+00 1.19445629e-01
-1.60969436e-01 2.59009153e-01 9.23324049e-01 4.73832220e-01
-1.30816489e-01 8.45575988e-01 6.73617840e-01 -1.26932430e+00
-4.60318267e-01 -1.10310733e+00 -8.93559635e-01 8.61075699e-01
-5.13253026e-02 -6.05009735e-01 -3.00569624e-01 2.56338596e-01]
|
[15.903129577636719, 5.347635746002197]
|
8ecfea2b-d4ca-4b59-92ce-56cf88876d57
|
multi-view-spatial-temporal-network-for
|
2204.08747
| null |
https://arxiv.org/abs/2204.08747v1
|
https://arxiv.org/pdf/2204.08747v1.pdf
|
Multi-View Spatial-Temporal Network for Continuous Sign Language Recognition
|
Sign language is a beautiful visual language and is also the primary language used by speaking and hearing-impaired people. However, sign language has many complex expressions, which are difficult for the public to understand and master. Sign language recognition algorithms will significantly facilitate communication between hearing-impaired people and normal people. Traditional continuous sign language recognition often uses a sequence learning method based on Convolutional Neural Network (CNN) and Long Short-Term Memory Network (LSTM). These methods can only learn spatial and temporal features separately, which cannot learn the complex spatial-temporal features of sign language. LSTM is also difficult to learn long-term dependencies. To alleviate these problems, this paper proposes a multi-view spatial-temporal continuous sign language recognition network. The network consists of three parts. The first part is a Multi-View Spatial-Temporal Feature Extractor Network (MSTN), which can directly extract the spatial-temporal features of RGB and skeleton data; the second is a sign language encoder network based on Transformer, which can learn long-term dependencies; the third is a Connectionist Temporal Classification (CTC) decoder network, which is used to predict the whole meaning of the continuous sign language. Our algorithm is tested on two public sign language datasets SLR-100 and PHOENIX-Weather 2014T (RWTH). As a result, our method achieves excellent performance on both datasets. The word error rate on the SLR-100 dataset is 1.9%, and the word error rate on the RWTHPHOENIX-Weather dataset is 22.8%.
|
['Lu Meng', 'Ronghui Li']
|
2022-04-19
| null | null | null | null |
['sign-language-recognition']
|
['computer-vision']
|
[-1.87725529e-01 -6.28208101e-01 -2.18182132e-01 -2.92761624e-01
-6.35709345e-01 -1.25367507e-01 3.18961829e-01 -1.00253999e+00
-7.19725728e-01 4.82126743e-01 5.37891865e-01 -2.31653795e-01
1.20153300e-01 -7.56726325e-01 -3.44987780e-01 -8.32564294e-01
-1.00780033e-01 9.15266797e-02 3.79378468e-01 -1.98466435e-01
2.86578480e-02 5.81971169e-01 -1.64231944e+00 6.83255255e-01
5.67177594e-01 1.12005496e+00 2.50018507e-01 7.08179533e-01
-6.27639413e-01 1.18032181e+00 -3.41808438e-01 1.35432318e-01
-8.85727406e-02 -5.31716585e-01 -7.00675845e-01 -1.56924203e-01
1.46851763e-01 -5.76755822e-01 -7.29618669e-01 6.94201589e-01
9.57039654e-01 -1.52087897e-01 4.25029278e-01 -1.12734044e+00
-6.56554341e-01 1.05507784e-01 -3.49465787e-01 -1.70853361e-01
2.52738357e-01 5.01742780e-01 7.76581526e-01 -8.49127531e-01
4.98192906e-01 1.28633845e+00 4.70071644e-01 7.93202579e-01
-2.90597737e-01 -9.69040275e-01 3.39686722e-01 6.57894969e-01
-1.44119906e+00 -2.21462816e-01 5.27056634e-01 -4.16419148e-01
1.17194307e+00 -1.04795836e-01 1.21665907e+00 1.04821205e+00
1.78956568e-01 1.19911933e+00 1.29585218e+00 -3.35824430e-01
-2.11939156e-01 -5.43450296e-01 1.21007651e-01 8.36777449e-01
-2.15464383e-01 3.52315545e-01 -7.26630688e-01 2.95356452e-01
7.84106135e-01 2.16853887e-01 -4.52278078e-01 1.54682428e-01
-1.21340036e+00 4.28405523e-01 6.85931623e-01 5.38358510e-01
-2.49569029e-01 3.44687134e-01 3.83968145e-01 4.87386853e-01
-1.13938265e-01 -5.80283105e-01 -5.55965900e-01 -5.05867302e-01
-5.12104988e-01 -3.35094303e-01 6.09479845e-01 7.11521506e-01
1.84527189e-01 3.55855725e-03 -1.69411823e-01 1.03959918e+00
8.04831207e-01 9.34210777e-01 9.05990958e-01 -2.59649187e-01
4.76159781e-01 6.35458350e-01 -2.47205675e-01 -3.24178457e-01
-4.52959508e-01 -1.33100510e-01 -9.16509867e-01 6.43276215e-01
3.71349901e-01 -8.54293704e-02 -1.63342595e+00 1.47739196e+00
-2.05065385e-01 -9.81610417e-02 3.77688254e-03 1.13917935e+00
1.29855847e+00 5.60369909e-01 1.59459636e-01 6.09526783e-02
1.43858004e+00 -8.88775647e-01 -7.89219260e-01 -1.00824349e-01
5.24913371e-01 -6.49437666e-01 1.17400861e+00 3.65779281e-01
-6.96361184e-01 -3.05920392e-01 -7.38067448e-01 -5.08411288e-01
-6.65560424e-01 3.91932368e-01 5.03348351e-01 2.78661847e-01
-9.29258168e-01 -7.30206668e-02 -8.95643055e-01 -4.30274427e-01
5.69222569e-01 3.84315342e-01 -4.69979793e-01 -2.93457150e-01
-1.05627751e+00 1.00825942e+00 2.75729075e-02 4.51453805e-01
-3.70655090e-01 -1.98869660e-01 -8.49088371e-01 -2.89701074e-01
-1.64192721e-01 -5.57393253e-01 1.16483259e+00 -9.18672085e-01
-1.82206869e+00 9.42084908e-01 -6.50219440e-01 1.19407676e-01
7.10014999e-01 -7.31042698e-02 -7.58804560e-01 -5.74839972e-02
-1.61787540e-01 5.67637563e-01 7.85665989e-01 -8.94944310e-01
-8.92657995e-01 -4.13904160e-01 -3.99870694e-01 1.15060799e-01
-1.26039624e-01 2.52289921e-01 -4.68767881e-01 -7.30446994e-01
5.61990023e-01 -6.88930571e-01 3.13760847e-01 6.13470256e-01
-2.10079089e-01 -4.45195019e-01 9.98921812e-01 -1.06385159e+00
9.17216063e-01 -2.24449492e+00 -4.29160185e-02 2.77382940e-01
-5.73324114e-02 3.57661605e-01 -2.94606090e-01 1.36561230e-01
-9.14066434e-02 -1.20735541e-01 -3.07037413e-01 5.25502786e-02
-1.80308849e-01 5.64345479e-01 -4.12000209e-01 3.26155126e-01
-7.02968687e-02 1.23068726e+00 -7.03813374e-01 -3.17257762e-01
4.15415138e-01 7.95517385e-01 -1.85373630e-02 1.24752246e-01
-8.27635527e-02 4.75477934e-01 -3.17236811e-01 9.21891928e-01
3.94152224e-01 -1.17131725e-01 -2.53721118e-01 -2.15646774e-01
-3.54248822e-01 3.84352863e-01 -8.05261135e-01 1.65962756e+00
-7.27447152e-01 7.91704834e-01 -1.64963335e-01 -6.77749097e-01
8.14122498e-01 4.81457084e-01 4.21702713e-01 -1.20427179e+00
3.15053761e-01 5.73627591e-01 -2.95372065e-02 -1.08970976e+00
-3.96598160e-01 -4.11469281e-01 1.75075442e-01 4.22711074e-01
-1.75106809e-01 1.79856598e-01 -3.93136516e-02 -3.21624368e-01
1.07854855e+00 -1.09018795e-01 -4.31716479e-02 4.35525447e-01
5.82388937e-01 -3.43643755e-01 5.38115680e-01 2.24681333e-01
-1.83254138e-01 5.31839967e-01 1.18731402e-01 -6.27671778e-01
-4.60899800e-01 -1.33887875e+00 4.89085075e-03 7.08794236e-01
-8.01021829e-02 -1.05297379e-01 1.47716612e-01 -5.49961209e-01
-3.09199002e-02 3.91261578e-01 -3.01825255e-01 1.48989493e-03
-6.93652213e-01 -2.95631915e-01 6.32913828e-01 8.16545904e-01
1.13876891e+00 -1.56878221e+00 -6.75758541e-01 -3.71110551e-02
-4.12044376e-01 -1.14930332e+00 -6.02187395e-01 -2.33892933e-01
-6.68750226e-01 -1.13713050e+00 -1.18085372e+00 -1.27500522e+00
5.33200443e-01 -4.77405498e-03 3.51698250e-01 2.52937209e-02
-3.29309732e-01 3.94961417e-01 -5.13175070e-01 -3.61003786e-01
8.31196979e-02 -2.34254345e-01 -2.45836377e-01 1.26777887e-01
7.49479532e-01 -8.46745491e-01 -3.72950166e-01 2.17399985e-01
-6.53015971e-01 2.76219636e-01 9.04687941e-01 1.15024579e+00
4.61928934e-01 -4.10726458e-01 1.36142358e-01 8.68671760e-02
3.93617034e-01 1.40791342e-01 -4.52694207e-01 5.78590572e-01
-1.86675444e-01 1.10329978e-01 3.90676051e-01 -6.06971264e-01
-8.45464528e-01 1.49644837e-01 -5.17614841e-01 -1.02078535e-01
-8.58910605e-02 5.65752566e-01 -3.06020617e-01 -3.77692997e-01
4.06632423e-02 9.04148877e-01 7.82090873e-02 -5.91158211e-01
1.86328247e-01 1.22657561e+00 6.02704465e-01 -6.02032132e-02
5.25742948e-01 6.46356642e-01 -1.78936645e-01 -1.14252710e+00
-4.87988085e-01 -3.90231103e-01 -6.50644422e-01 -3.37426454e-01
9.76359665e-01 -8.53893101e-01 -1.01218593e+00 1.30814123e+00
-1.30876446e+00 -5.58676243e-01 -1.51336387e-01 9.45087492e-01
-4.95174974e-01 3.28254163e-01 -6.01481974e-01 -6.74987197e-01
-2.60560751e-01 -9.53078330e-01 1.03996682e+00 2.58586884e-01
1.86853409e-01 -5.74574769e-01 -2.17363417e-01 2.42142677e-01
4.49825376e-01 1.75108001e-01 9.10921395e-01 2.00264767e-01
-6.71273410e-01 -1.75794959e-01 -5.82883835e-01 5.78765035e-01
3.26925486e-01 -2.72321612e-01 -8.64745438e-01 -1.45503357e-01
-3.88121307e-01 -3.95728886e-01 1.29279947e+00 5.61364651e-01
1.13726962e+00 -1.15010016e-01 -1.04445480e-01 8.74703825e-01
1.04106140e+00 4.87185955e-01 9.56901729e-01 1.42345607e-01
7.45521188e-01 3.04888576e-01 4.23262380e-02 1.02369279e-01
7.92541623e-01 5.88539124e-01 -6.16152547e-02 -2.46983349e-01
-6.93243563e-01 -3.81186068e-01 7.25089192e-01 1.08971274e+00
-3.31899464e-01 1.72100917e-01 -1.05856514e+00 6.53540671e-01
-1.64637125e+00 -9.47717726e-01 -1.00797936e-01 1.77785945e+00
6.91229820e-01 -2.12458685e-01 -1.53163895e-02 5.05129933e-01
3.20469141e-01 9.15888175e-02 -7.15014815e-01 -2.18150869e-01
-4.86387163e-01 4.94046777e-01 3.70072246e-01 4.74943727e-01
-8.97029757e-01 1.05790651e+00 5.36358452e+00 4.49316323e-01
-1.83069611e+00 6.84662089e-02 -1.05351105e-01 -2.05282271e-01
-2.01957580e-03 -3.29350114e-01 -4.57738519e-01 3.73167276e-01
4.27566439e-01 8.88002589e-02 4.30993915e-01 4.90942776e-01
3.95035118e-01 -2.04172786e-02 -7.36531615e-01 1.52691245e+00
2.05180109e-01 -8.93346786e-01 2.12599039e-01 -4.29266458e-03
1.11978143e-01 6.33208275e-01 -1.30214959e-01 3.42921168e-01
1.86120614e-01 -1.35124624e+00 5.32440841e-01 9.20078397e-01
1.35756314e+00 -4.86206681e-01 6.89584315e-01 3.75297219e-01
-1.59476995e+00 -2.50018179e-01 1.51088640e-01 -1.58608392e-01
5.32988548e-01 3.36890697e-01 -9.55550224e-02 5.99572845e-02
9.14576828e-01 1.02890074e+00 -2.02354118e-01 1.30986011e+00
-8.11878920e-01 5.64291835e-01 -5.68927824e-01 -3.48548472e-01
2.18706235e-01 1.38481349e-01 4.62301522e-01 1.20434833e+00
4.30301070e-01 3.47933948e-01 8.03324357e-02 4.13588464e-01
1.23000942e-01 6.57669157e-02 -5.65011501e-01 -4.88649420e-02
-7.64142573e-02 5.66847205e-01 -2.11125538e-01 -2.34652832e-01
-7.18363464e-01 1.33052063e+00 -7.61257187e-02 6.59227431e-01
-4.51884210e-01 -6.69585168e-01 6.63737357e-01 -1.73161134e-01
2.84648806e-01 -5.35690188e-01 -3.52550983e-01 -1.41489172e+00
6.41884446e-01 -5.70462406e-01 5.77517331e-01 -1.02678263e+00
-1.36320734e+00 3.98327470e-01 -4.42603320e-01 -1.52022064e+00
-7.81828389e-02 -1.18744695e+00 -4.86717671e-01 1.12210822e+00
-1.99281025e+00 -1.62229002e+00 -6.20425403e-01 1.12350321e+00
3.25765431e-01 -2.95740098e-01 9.73346472e-01 4.43845659e-01
-2.00953618e-01 6.98504329e-01 -1.34634733e-01 8.87493730e-01
4.12883043e-01 -6.82208240e-01 9.31401774e-02 6.35240912e-01
1.22903042e-01 2.95405269e-01 -1.90320656e-01 -4.50565338e-01
-1.20128977e+00 -1.02518725e+00 1.40212679e+00 1.19329877e-01
4.96659100e-01 -1.24649480e-01 -5.67309439e-01 6.14823520e-01
-3.05787921e-01 3.77625316e-01 4.23982054e-01 -3.01452547e-01
-6.97853923e-01 -1.71554044e-01 -1.01014006e+00 5.95348299e-01
1.42288041e+00 -1.04149389e+00 -6.47258878e-01 3.00517827e-01
3.54382813e-01 -3.80110502e-01 -4.99892086e-01 4.59907740e-01
1.38441896e+00 -7.27978826e-01 7.69182324e-01 -5.63580334e-01
1.77915305e-01 -4.28294986e-01 -3.13636899e-01 -1.09598887e+00
1.19618692e-01 -7.22882897e-02 -3.40436399e-02 6.29913211e-01
2.67172009e-01 -1.10642445e+00 4.90743905e-01 2.97623098e-01
4.87758406e-02 -7.08078265e-01 -1.52096987e+00 -1.08892965e+00
7.78350234e-02 -8.15171778e-01 3.66293818e-01 6.63020909e-01
-7.59399459e-02 1.72902375e-01 -3.24576110e-01 7.95618445e-03
4.04604256e-01 1.79690465e-01 5.84339082e-01 -1.00661051e+00
-7.38120526e-02 -6.79882586e-01 -7.34703600e-01 -1.42763853e+00
1.23141490e-01 -9.35958982e-01 1.44459894e-02 -2.03496838e+00
-1.34756371e-01 -7.90970698e-02 -3.10611933e-01 1.05813265e+00
5.44172466e-01 -4.40595485e-02 1.03834219e-01 1.16713606e-01
4.81462739e-02 8.13395143e-01 1.68808544e+00 -4.91241932e-01
-2.94318229e-01 1.50511324e-01 8.34723189e-03 7.51330078e-01
6.88652277e-01 -1.24878980e-01 -7.58721679e-03 -7.53391862e-01
-2.21042529e-01 -5.92544712e-02 7.72128522e-01 -1.00436103e+00
5.14992595e-01 -1.58722639e-01 3.31883937e-01 -9.15226161e-01
4.05041218e-01 -7.48329282e-01 -4.17421907e-01 9.54200506e-01
1.47504685e-02 -2.08005995e-01 4.87781353e-02 6.25137761e-02
-5.36171734e-01 5.05421400e-01 7.94339955e-01 -1.64944738e-01
-1.18445516e+00 5.84083319e-01 -5.38332403e-01 -1.25147775e-01
6.19738102e-01 -3.10640156e-01 -2.49742746e-01 -6.32090390e-01
-7.88760841e-01 3.24999332e-01 -2.23102346e-01 7.86276579e-01
1.16156924e+00 -1.51822448e+00 -7.64627755e-01 5.98909557e-01
3.81712139e-01 -8.88702497e-02 1.98524654e-01 8.80906463e-01
-6.53313160e-01 4.66518670e-01 -2.56896317e-01 -5.69623291e-01
-1.41309726e+00 -2.18422219e-01 7.45394409e-01 1.21438652e-01
-1.14860296e+00 1.05281031e+00 -3.50729257e-01 -5.19129276e-01
5.44251919e-01 -7.14216650e-01 -1.78944811e-01 -5.41773513e-02
6.63082421e-01 3.85786928e-02 -2.52961397e-01 -9.31342900e-01
-5.77288628e-01 1.10408616e+00 2.64211774e-01 -5.49906790e-01
1.30615866e+00 1.94213539e-01 -3.98849636e-01 6.63261950e-01
1.41041613e+00 -4.33490574e-01 -7.77686715e-01 -5.00426173e-01
-9.95712355e-02 -3.16698790e-01 2.08092496e-01 -1.20762289e+00
-1.17997456e+00 1.36766636e+00 1.11224866e+00 -6.36473477e-01
1.44608951e+00 -1.56733811e-01 1.35437131e+00 6.33005559e-01
5.30445039e-01 -1.02282381e+00 -7.36195445e-02 1.07645071e+00
1.29584157e+00 -1.16626883e+00 -5.11823297e-01 -1.11312780e-03
-2.88800806e-01 1.38873434e+00 4.25161809e-01 1.04426108e-01
1.14230800e+00 2.90405095e-01 6.10362411e-01 -9.26618576e-02
-4.38637912e-01 -7.26557791e-01 4.45101947e-01 6.01141870e-01
3.93978864e-01 2.52682716e-01 -4.16409969e-01 5.68111002e-01
-3.32300514e-01 5.42994082e-01 -3.09258234e-02 1.00778461e+00
-2.70271420e-01 -1.15834928e+00 -1.93161160e-01 4.85117763e-01
1.06505319e-01 -8.37163478e-02 -3.88738900e-01 6.98205471e-01
5.06584883e-01 8.79478335e-01 -5.95461577e-02 -6.70926690e-01
5.39630115e-01 2.99267322e-01 5.30713379e-01 -1.83872908e-01
-1.39091045e-01 -1.19731888e-01 -9.24845859e-02 -5.78822851e-01
-4.42082763e-01 -5.15007615e-01 -1.75877929e+00 1.73027664e-02
2.24997774e-01 -4.59620178e-01 7.17363775e-01 1.35185170e+00
-4.90865000e-02 4.40441072e-01 2.16825426e-01 -3.58376980e-01
-1.99025765e-01 -9.95808780e-01 -6.05893016e-01 2.39635020e-01
6.93007350e-01 -6.50756359e-01 -1.50997713e-01 -5.66659756e-02]
|
[9.207999229431152, -6.499528884887695]
|
a81e95c2-06ae-4ef0-89a4-aea98ccbe28b
|
a-persian-asr-based-ser-modification-of
|
2211.09956
| null |
https://arxiv.org/abs/2211.09956v1
|
https://arxiv.org/pdf/2211.09956v1.pdf
|
A Persian ASR-based SER: Modification of Sharif Emotional Speech Database and Investigation of Persian Text Corpora
|
Speech Emotion Recognition (SER) is one of the essential perceptual methods of humans in understanding the situation and how to interact with others, therefore, in recent years, it has been tried to add the ability to recognize emotions to human-machine communication systems. Since the SER process relies on labeled data, databases are essential for it. Incomplete, low-quality or defective data may lead to inaccurate predictions. In this paper, we fixed the inconsistencies in Sharif Emotional Speech Database (ShEMO), as a Persian database, by using an Automatic Speech Recognition (ASR) system and investigating the effect of Farsi language models obtained from accessible Persian text corpora. We also introduced a Persian/Farsi ASR-based SER system that uses linguistic features of the ASR outputs and Deep Learning-based models.
|
['Yasser Shekofteh', 'Ali Yazdani']
|
2022-11-18
| null | null | null | null |
['speech-emotion-recognition']
|
['speech']
|
[-2.16445431e-01 3.26099783e-01 4.52171415e-01 -6.70084536e-01
-5.01290858e-01 -1.93936765e-01 4.78176087e-01 -7.89893419e-02
-6.15636408e-01 6.92706466e-01 3.41872424e-01 -1.64898887e-01
2.28245690e-01 -3.47362995e-01 -1.01566479e-01 -2.15061069e-01
6.17972501e-02 4.38173324e-01 8.51436183e-02 -7.49754846e-01
2.47483719e-02 6.09715581e-01 -1.48912382e+00 5.53564429e-01
6.09990835e-01 1.15411806e+00 3.05679888e-01 9.11066890e-01
-3.25402379e-01 1.37628007e+00 -1.05254924e+00 -5.52022994e-01
-2.04967573e-01 -3.75449419e-01 -1.04225075e+00 4.14190292e-02
-6.15894198e-01 9.51200649e-02 -3.36206496e-01 7.64982641e-01
5.82173407e-01 2.47839212e-01 4.62461442e-01 -1.26884854e+00
-8.97430778e-01 7.82068968e-01 1.27884164e-01 9.60180014e-02
5.63423276e-01 -2.10219294e-01 6.37274206e-01 -9.34416890e-01
4.74367052e-01 1.57710683e+00 3.08616549e-01 7.91057229e-01
-8.51299465e-01 -4.58735943e-01 -1.29997700e-01 6.19627237e-01
-1.32311010e+00 -9.14651394e-01 9.81738865e-01 -2.21176565e-01
1.52586555e+00 1.61067680e-01 5.69787860e-01 1.43565178e+00
-2.46452987e-02 9.11811054e-01 1.18557549e+00 -6.87816501e-01
4.15490180e-01 6.25034273e-01 1.28324643e-01 2.98543334e-01
-5.81694722e-01 3.19913745e-01 -6.88120246e-01 8.19820091e-02
8.37253779e-02 -5.23625672e-01 -1.21003285e-01 3.98088217e-01
-7.29166389e-01 7.38120973e-01 6.70570061e-02 6.52869105e-01
-4.99529749e-01 -4.99108046e-01 6.38751924e-01 6.03623390e-01
5.17219424e-01 4.56264526e-01 -6.09307289e-01 -6.41008139e-01
-6.85052335e-01 -4.87666458e-01 9.25984621e-01 6.28604949e-01
4.16002363e-01 5.84388316e-01 3.63956004e-01 1.51284266e+00
3.30482960e-01 5.52770257e-01 8.61152351e-01 -7.24358857e-01
1.35164186e-01 6.96807623e-01 -6.64099976e-02 -1.00078869e+00
-4.41470325e-01 2.47968242e-01 -6.34784341e-01 1.96593866e-01
-2.42926583e-01 -6.64554462e-02 -8.50849688e-01 1.47540879e+00
-4.69863340e-02 -3.79209042e-01 7.75934577e-01 8.16216946e-01
1.04032004e+00 9.76338983e-01 3.64126503e-01 -2.82267600e-01
1.20787990e+00 -6.81863308e-01 -1.09833848e+00 -3.39940399e-01
5.33556759e-01 -6.95115268e-01 1.13291109e+00 8.84960115e-01
-9.43845153e-01 -5.85509121e-01 -1.00338399e+00 -3.28686740e-03
-8.18034768e-01 2.96811283e-01 2.73062497e-01 8.09604526e-01
-1.14328849e+00 2.81038940e-01 -6.38533890e-01 -5.77648461e-01
5.50978817e-03 2.46533483e-01 -7.66163051e-01 3.58742595e-01
-1.45907640e+00 1.34724987e+00 5.77453494e-01 2.33676612e-01
-6.03232920e-01 2.76681930e-01 -9.50856149e-01 5.77585511e-02
1.56917751e-01 2.28469715e-01 1.24285054e+00 -1.60426807e+00
-1.91372335e+00 9.59516108e-01 -1.88975587e-01 -6.12842143e-01
-9.79341194e-02 1.31049920e-02 -1.24790645e+00 2.11485684e-01
-6.07878029e-01 6.94960296e-01 6.78346276e-01 -1.11851394e+00
-2.71913201e-01 -5.34398556e-01 -3.65394086e-01 -6.35624528e-02
-1.44287437e-01 6.76972747e-01 1.57339931e-01 -5.31769514e-01
-2.84144841e-02 -8.07307959e-01 1.97767586e-01 -5.69820583e-01
-1.59647182e-01 -4.85255569e-01 8.19476128e-01 -1.19974470e+00
1.14403188e+00 -2.56574917e+00 -4.28817421e-02 -2.40168236e-02
-2.10337043e-01 7.81175733e-01 -2.24751934e-01 5.73439002e-01
-2.79279917e-01 2.41928443e-01 -1.11986451e-01 -4.38079655e-01
1.47772655e-01 6.55264378e-01 -3.45150024e-01 6.97817877e-02
5.82365274e-01 5.83199620e-01 -4.06935453e-01 -5.25401533e-01
5.26388645e-01 5.90776503e-01 -1.69904828e-01 6.26858532e-01
-2.01044083e-02 1.67091966e-01 -4.64078598e-02 5.57982087e-01
4.11555350e-01 5.04547238e-01 2.00886559e-02 7.79949315e-03
-1.09822750e-01 4.88126040e-01 -9.12179112e-01 1.29123855e+00
-6.15572214e-01 7.61287451e-01 2.57554203e-01 -1.21918094e+00
1.24333417e+00 8.18448901e-01 1.59235656e-01 -1.02664804e+00
3.97633880e-01 1.29829198e-01 2.06742167e-01 -8.15734208e-01
6.20964587e-01 -3.84306848e-01 -2.25618169e-01 5.30519076e-02
3.92513663e-01 -1.48924053e-01 -4.45230640e-02 2.52346043e-02
9.03747857e-01 -2.59692788e-01 5.22236466e-01 2.15090528e-01
8.35104167e-01 -9.46197286e-02 6.84902549e-01 1.06467135e-01
-5.64387739e-01 4.94041026e-01 3.18902761e-01 -3.80127370e-01
-8.63869250e-01 -8.93198848e-01 6.62888512e-02 9.30790484e-01
-3.73952508e-01 -3.84679258e-01 -6.79561198e-01 -4.20297921e-01
-6.54391646e-01 1.29549325e+00 -3.23864877e-01 -3.78477991e-01
-2.29334116e-01 -2.75460899e-01 9.21084881e-01 4.28689301e-01
4.24138337e-01 -1.85854805e+00 -3.78078818e-01 3.53164524e-01
-1.50579274e-01 -1.28803444e+00 3.92928086e-02 5.09151399e-01
-1.80713072e-01 -8.19160700e-01 -9.19896290e-02 -6.35460019e-01
1.01041138e-01 -1.85262427e-01 9.94415283e-01 -6.93946704e-02
-1.79722354e-01 4.39773083e-01 -9.41556334e-01 -6.27911747e-01
-1.07599711e+00 -3.78326774e-01 1.43826067e-01 -4.16511558e-02
6.73138499e-01 -3.82922858e-01 1.49518043e-01 2.60410875e-01
-8.14761460e-01 -1.76369786e-01 4.84231800e-01 8.16599905e-01
8.14350098e-02 2.06348807e-01 9.36496317e-01 -4.16323721e-01
8.15462589e-01 -2.66059041e-01 -2.04618976e-01 2.18187019e-01
-2.62291640e-01 1.11094445e-01 7.15031087e-01 -2.81435698e-01
-1.47191787e+00 -2.45517753e-02 -8.18849325e-01 -2.85607457e-01
-6.58580005e-01 4.39946413e-01 -6.76858246e-01 2.63450503e-01
5.09716749e-01 2.36465126e-01 -5.85189573e-02 -2.89670855e-01
1.61057323e-01 1.70454323e+00 4.33190376e-01 -4.27211821e-01
-3.80816124e-02 -1.37732133e-01 -6.18140221e-01 -1.47096944e+00
-4.92569327e-01 -3.46117467e-01 -5.01229465e-01 -2.82170951e-01
1.00337374e+00 -6.75772548e-01 -7.75757968e-01 5.52136242e-01
-1.42291617e+00 -2.82348901e-01 -2.16846958e-01 5.27413547e-01
-5.08932710e-01 3.80599260e-01 -7.67049432e-01 -1.52562141e+00
-3.87837350e-01 -1.09479606e+00 8.83192301e-01 2.30976241e-03
-5.94629765e-01 -7.65608549e-01 -1.57921873e-02 5.31189263e-01
3.43879461e-01 -4.41577524e-01 7.53197014e-01 -1.25752687e+00
2.44625255e-01 -4.95517850e-02 1.03465535e-01 1.14194703e+00
-2.27100924e-02 1.93870321e-01 -1.25388658e+00 2.17895135e-01
2.96765000e-01 -7.46321857e-01 3.94641221e-01 -5.43410629e-02
9.05425489e-01 -3.06484640e-01 2.94065148e-01 1.14347056e-01
8.18212330e-01 9.92625475e-01 9.63635802e-01 1.03446692e-01
1.13534503e-01 8.69385481e-01 6.32313192e-01 4.29141611e-01
4.40536857e-01 4.28973347e-01 9.94971171e-02 1.07794210e-01
-5.89689091e-02 -2.15921164e-01 8.24782670e-01 1.24850857e+00
2.82945007e-01 -4.63639051e-01 -8.82109582e-01 4.28135216e-01
-1.52448857e+00 -9.93181646e-01 1.94806457e-01 1.83881664e+00
7.53466249e-01 2.07917299e-02 -7.94757754e-02 3.86040211e-01
6.56130254e-01 6.09408394e-02 -1.95398048e-01 -1.11449492e+00
-5.22720098e-01 1.60777360e-01 -1.69493467e-01 5.16673505e-01
-8.61619532e-01 1.31429911e+00 6.40060234e+00 6.83834195e-01
-1.27866757e+00 1.19503856e-01 5.81593871e-01 2.79910475e-01
6.26207143e-02 -3.05725962e-01 -4.08974886e-01 4.54530984e-01
1.56254399e+00 5.36755919e-02 7.64886260e-01 8.85312736e-01
4.15795892e-01 -2.05961421e-01 -8.71322930e-01 1.33140755e+00
4.30185080e-01 -6.26454949e-01 -2.31562197e-01 -2.75938690e-01
-1.11479141e-01 6.86755925e-02 -2.76390731e-01 5.82389235e-01
3.04405808e-01 -1.17908061e+00 7.53722131e-01 3.26726019e-01
4.65805918e-01 -9.05768394e-01 1.01009226e+00 5.38939476e-01
-7.13060796e-01 -1.77892819e-01 -2.97615111e-01 -9.78569239e-02
1.00257874e-01 1.19054943e-01 -1.14049995e+00 2.02848911e-01
7.59644747e-01 3.60092998e-01 -4.85093504e-01 3.13856810e-01
-2.76906013e-01 8.34846258e-01 -2.27415219e-01 -2.48521701e-01
-1.36807427e-01 -2.04537138e-01 5.45728028e-01 1.25761342e+00
2.84425944e-01 2.74737418e-01 1.87619869e-02 6.72516167e-01
1.83487147e-01 4.00169611e-01 -6.73797131e-01 -5.61252058e-01
4.96734798e-01 1.16722953e+00 -4.41428900e-01 -2.06881598e-01
-3.00527036e-01 1.11847770e+00 3.74480039e-01 1.66079938e-01
-5.14021158e-01 -2.14866981e-01 5.64418912e-01 -2.82237321e-01
-1.52905464e-01 -2.63195485e-01 6.07394651e-02 -1.03117263e+00
-1.86994791e-01 -1.17426789e+00 2.74376988e-01 -1.50648558e+00
-1.40683460e+00 1.24433839e+00 -4.39135313e-01 -7.87678123e-01
-4.27169532e-01 -7.54093409e-01 -3.36617619e-01 7.71193087e-01
-1.13994312e+00 -1.17792594e+00 1.20958410e-01 5.45843303e-01
8.87917101e-01 -6.52597606e-01 1.11972415e+00 2.41201535e-01
-5.05879462e-01 2.77671576e-01 -5.32450140e-01 3.81441593e-01
7.60436118e-01 -1.11307859e+00 1.56788609e-03 6.70827031e-01
3.49635363e-01 2.73431718e-01 7.78899789e-01 -4.50814545e-01
-1.14703238e+00 -6.10351264e-01 1.15410352e+00 -1.94007844e-01
6.69414639e-01 -4.14564550e-01 -1.01872993e+00 6.55350924e-01
6.37317479e-01 -3.29202294e-01 8.98493707e-01 -4.13487069e-02
-1.99781612e-01 -1.72163054e-01 -1.31001019e+00 4.82955605e-01
6.88317239e-01 -8.34999025e-01 -1.09839296e+00 -2.20864922e-01
7.41664827e-01 1.70186356e-01 -7.43277431e-01 1.39052808e-01
6.07215986e-02 -1.02840984e+00 5.99777341e-01 -6.13924742e-01
5.45973815e-02 -1.53421074e-01 -5.24443984e-01 -1.83846819e+00
1.36323627e-02 -4.97028798e-01 2.84633696e-01 1.59603941e+00
4.07859862e-01 -6.55188978e-01 2.61514544e-01 7.14482605e-01
-3.74014348e-01 -7.81842321e-02 -1.21056390e+00 -5.08313239e-01
-2.41504833e-01 -9.98170078e-01 4.78307366e-01 8.51447165e-01
5.65007806e-01 6.24891341e-01 -4.11018431e-01 1.12369008e-01
-1.84451908e-01 -6.26818538e-01 3.75516385e-01 -1.26584542e+00
-1.19625172e-02 -2.80027300e-01 -7.05555856e-01 -2.91089147e-01
8.24019909e-01 -6.83861136e-01 1.88533083e-01 -1.31099522e+00
-4.05637234e-01 -2.37021819e-01 -2.00798124e-01 6.66802466e-01
4.03824866e-01 -2.30956391e-01 2.22277686e-01 -2.41312802e-01
-5.26275933e-01 8.95714045e-01 5.36724091e-01 -1.36503533e-01
-1.60287410e-01 -2.84234911e-01 -4.56587732e-01 9.72288907e-01
9.04531777e-01 -3.12690109e-01 -2.83902347e-01 -9.50682629e-03
2.32271999e-01 3.52310508e-01 1.37067392e-01 -1.08022976e+00
1.55174538e-01 -1.46459006e-02 2.90110409e-01 -5.29555082e-01
8.66548598e-01 -1.09126127e+00 9.67054293e-02 6.45943731e-02
-4.42218125e-01 4.25026827e-02 2.71876812e-01 -1.04010412e-02
-7.18013763e-01 -2.60151833e-01 9.15157437e-01 -1.08224653e-01
-1.03763723e+00 -2.35713914e-01 -9.00702536e-01 -1.51722595e-01
9.01200175e-01 -6.14527091e-02 -2.29103118e-01 -5.21964073e-01
-9.66614306e-01 -7.49601424e-02 2.36464113e-01 7.74113715e-01
9.11343575e-01 -9.91271794e-01 -5.15843511e-01 4.80480552e-01
2.86844492e-01 -6.19936585e-01 3.40754390e-01 5.30438066e-01
-2.91405976e-01 4.20944780e-01 -2.67914832e-01 -8.59198943e-02
-1.55109739e+00 7.08028138e-01 5.31075060e-01 2.15371758e-01
-2.21253693e-01 5.10210454e-01 -2.17583701e-01 -6.81488335e-01
2.57777542e-01 -9.44615230e-02 -5.42740703e-01 1.08434327e-01
4.98086363e-01 2.65130281e-01 2.48720109e-01 -1.16154230e+00
-3.77901584e-01 -2.33946487e-01 2.64030516e-01 -5.41232169e-01
1.43992841e+00 -3.66914660e-01 -2.07589835e-01 9.09303427e-01
9.51584697e-01 7.80726001e-02 -4.88544345e-01 8.75867233e-02
2.45938614e-01 -2.55129356e-02 2.99760520e-01 -1.14559674e+00
-7.91285276e-01 1.18076897e+00 6.53735995e-01 5.15627503e-01
1.19508648e+00 8.38039070e-02 7.46405303e-01 6.17654681e-01
3.91127557e-01 -1.75509834e+00 -1.44711882e-01 7.27914631e-01
1.13656688e+00 -1.43302810e+00 -8.30798507e-01 -1.61417574e-01
-1.35274625e+00 1.17425621e+00 4.99754041e-01 2.80865580e-01
7.14319289e-01 3.91988724e-01 5.08122504e-01 -3.17335338e-03
-1.08749950e+00 -2.27033183e-01 -6.99371425e-03 8.23573351e-01
6.64794564e-01 2.78346866e-01 -1.63878858e-01 1.19353282e+00
-4.09642071e-01 5.29531948e-02 5.48892498e-01 6.77877724e-01
-4.28778738e-01 -1.23243332e+00 -4.85421091e-01 7.16693848e-02
-4.35558796e-01 -3.93982045e-02 -9.14217949e-01 6.16454244e-01
1.29734417e-02 1.51927495e+00 -3.23363394e-02 -5.44049323e-01
4.31354046e-01 5.83399475e-01 4.30336893e-02 -4.16064352e-01
-6.04943871e-01 -4.34128232e-02 6.52795315e-01 -3.85165840e-01
-4.44436550e-01 -5.16663909e-01 -1.49925125e+00 1.70091260e-02
-1.30000606e-01 4.15113270e-01 8.55148137e-01 1.02437854e+00
4.55036789e-01 4.11950320e-01 6.99777424e-01 -5.23482740e-01
-3.56117100e-01 -1.26296461e+00 -8.91177833e-01 3.93195122e-01
-5.43704517e-02 -5.13129234e-01 -4.90460217e-01 1.10056996e-01]
|
[13.627829551696777, 5.836152076721191]
|
8719ac54-9a3e-4792-95c8-df8d9087ae6c
|
transformer-tracking-with-cyclic-shifting
|
2205.03806
| null |
https://arxiv.org/abs/2205.03806v1
|
https://arxiv.org/pdf/2205.03806v1.pdf
|
Transformer Tracking with Cyclic Shifting Window Attention
|
Transformer architecture has been showing its great strength in visual object tracking, for its effective attention mechanism. Existing transformer-based approaches adopt the pixel-to-pixel attention strategy on flattened image features and unavoidably ignore the integrity of objects. In this paper, we propose a new transformer architecture with multi-scale cyclic shifting window attention for visual object tracking, elevating the attention from pixel to window level. The cross-window multi-scale attention has the advantage of aggregating attention at different scales and generates the best fine-scale match for the target object. Furthermore, the cyclic shifting strategy brings greater accuracy by expanding the window samples with positional information, and at the same time saves huge amounts of computational power by removing redundant calculations. Extensive experiments demonstrate the superior performance of our method, which also sets the new state-of-the-art records on five challenging datasets, along with the VOT2020, UAV123, LaSOT, TrackingNet, and GOT-10k benchmarks.
|
['Wei Yang', 'Yi-Ping Phoebe Chen', 'Junqing Yu', 'Zikai Song']
|
2022-05-08
| null |
http://openaccess.thecvf.com//content/CVPR2022/html/Song_Transformer_Tracking_With_Cyclic_Shifting_Window_Attention_CVPR_2022_paper.html
|
http://openaccess.thecvf.com//content/CVPR2022/papers/Song_Transformer_Tracking_With_Cyclic_Shifting_Window_Attention_CVPR_2022_paper.pdf
|
cvpr-2022-1
|
['visual-object-tracking']
|
['computer-vision']
|
[-6.18937351e-02 -5.62800646e-01 -1.55916110e-01 -8.87851790e-02
-3.99608552e-01 -3.02419513e-01 3.62322718e-01 -2.21927956e-01
-2.97640771e-01 5.05600691e-01 2.49376241e-02 -9.20626000e-02
-3.69749926e-02 -6.73741937e-01 -6.94975555e-01 -6.44739330e-01
-5.64044192e-02 1.93407480e-02 9.68363464e-01 -2.37731442e-01
4.48760316e-02 4.13343877e-01 -1.66157770e+00 3.77931714e-01
7.63824463e-01 1.49124217e+00 3.61545414e-01 2.98705071e-01
-3.74629162e-02 7.90944457e-01 -4.86332804e-01 -4.75483567e-01
4.51626658e-01 3.93909886e-02 -4.24702555e-01 -4.40069027e-02
7.51449943e-01 -4.05442208e-01 -5.29021084e-01 1.16435313e+00
4.59128618e-01 -1.63342655e-01 2.94356905e-02 -1.48448884e+00
-1.25942504e+00 5.27077734e-01 -1.04233897e+00 6.53237879e-01
-1.22615010e-01 4.31767672e-01 9.75218654e-01 -9.56992745e-01
3.31192017e-01 1.40296710e+00 8.04782867e-01 2.74800330e-01
-9.90348995e-01 -1.04250956e+00 5.65974951e-01 5.88245749e-01
-1.39658308e+00 -9.25261900e-02 5.09257078e-01 -4.21185374e-01
9.30693507e-01 3.58150005e-01 9.22952652e-01 6.91753268e-01
2.13232756e-01 9.64262486e-01 8.87031257e-01 1.02428263e-02
-3.03746104e-01 -1.70426026e-01 1.21409431e-01 6.38576627e-01
4.15535003e-01 2.19390869e-01 -3.82809639e-01 1.64898306e-01
6.85850978e-01 4.37695682e-01 -4.83662009e-01 -4.45406348e-01
-1.39896774e+00 5.86745620e-01 9.72225606e-01 2.31815591e-01
-3.72412384e-01 4.19878334e-01 4.66220886e-01 1.59468334e-02
4.24832344e-01 5.84583059e-02 -5.21216512e-01 2.86604106e-01
-8.05567145e-01 8.30733329e-02 -9.81750190e-02 1.14717746e+00
4.62869912e-01 2.59586304e-01 -7.16050208e-01 4.26444173e-01
3.32756937e-01 7.38832533e-01 5.46424627e-01 -4.27488863e-01
4.27040368e-01 8.18504751e-01 2.25331306e-01 -1.09839761e+00
-1.88329577e-01 -6.53667808e-01 -7.46414006e-01 4.04574245e-01
2.09898844e-01 1.96965545e-01 -1.20883727e+00 1.76932871e+00
5.01080215e-01 3.68833840e-01 -1.65518090e-01 1.11725521e+00
9.57790792e-01 7.50775218e-01 3.94432068e-01 -5.51343672e-02
1.66414046e+00 -1.21821976e+00 -8.47463608e-01 -1.88835338e-01
-2.01702546e-02 -7.56953657e-01 1.11501324e+00 1.48089945e-01
-9.64778900e-01 -9.40561593e-01 -1.12252915e+00 -1.03770345e-01
-5.40253401e-01 3.43652815e-01 6.15437865e-01 3.61522496e-01
-8.85967135e-01 3.95897567e-01 -6.84766769e-01 -2.28139505e-01
8.15144897e-01 3.79846245e-01 -1.04090273e-01 2.11190730e-01
-1.16582203e+00 6.23368323e-01 4.18862820e-01 2.55985439e-01
-7.88541555e-01 -1.04232645e+00 -6.35572672e-01 4.06302571e-01
5.05868018e-01 -4.71451581e-01 1.10608566e+00 -8.86290133e-01
-1.01086259e+00 5.17930686e-01 1.30405918e-01 -7.09515929e-01
5.41934669e-01 -4.21174824e-01 -3.66864651e-01 -1.92895383e-02
1.17949925e-01 9.95511949e-01 1.04676783e+00 -8.46858263e-01
-1.01528549e+00 -4.62882042e-01 1.14246607e-01 3.54831256e-02
-6.48202956e-01 9.48540941e-02 -7.43171751e-01 -7.93773115e-01
-1.15344219e-01 -7.40085781e-01 -1.00228563e-01 3.91324401e-01
-1.76401839e-01 -3.49411041e-01 1.48014426e+00 -5.29328704e-01
1.45668745e+00 -2.22914410e+00 5.82533851e-02 -4.42230761e-01
4.23988074e-01 6.66237473e-01 -5.80549240e-02 -3.76170734e-03
-9.47807282e-02 -7.19997585e-02 9.30289850e-02 -1.43966451e-01
-1.44766629e-01 -1.76859125e-01 -3.62781465e-01 3.60608131e-01
3.89680117e-01 1.35576916e+00 -8.09130669e-01 -8.23236227e-01
3.69907856e-01 4.91677314e-01 -4.55872506e-01 -9.20477957e-02
-2.12303355e-01 2.26656031e-02 -5.31519890e-01 8.28752756e-01
7.41161644e-01 -5.42971134e-01 -1.90171167e-01 -7.02802956e-01
-3.75067770e-01 -2.66804755e-01 -9.89787936e-01 1.34333074e+00
-2.91903112e-02 8.50568593e-01 -1.95225984e-01 -4.97606218e-01
8.42611194e-01 1.95761910e-03 5.05176008e-01 -9.34254467e-01
2.71275938e-01 -2.63129026e-01 4.67116982e-02 -3.21566731e-01
6.50549054e-01 2.84281969e-01 6.01070002e-04 -5.94893359e-02
-1.75371796e-01 4.86104548e-01 2.57936209e-01 6.09219819e-02
6.44227207e-01 1.79331616e-01 2.30616391e-01 -4.40723211e-01
6.18894815e-01 2.64919754e-02 7.51504779e-01 3.53451997e-01
-4.50164169e-01 4.30880278e-01 1.89161211e-01 -7.09687054e-01
-8.94931972e-01 -8.04152429e-01 -1.01813033e-01 1.08377409e+00
5.88899910e-01 -2.84570843e-01 -5.93717933e-01 -6.76938057e-01
2.82732129e-01 2.71973133e-01 -1.05442691e+00 -1.52138144e-01
-5.98846436e-01 -6.39322639e-01 3.49123478e-01 1.08025169e+00
8.53814960e-01 -1.24581707e+00 -1.16017592e+00 1.48594752e-01
-2.91771501e-01 -1.06500578e+00 -9.97683346e-01 -1.43334717e-01
-6.69651449e-01 -1.14144289e+00 -8.31838489e-01 -7.96433032e-01
5.58109820e-01 5.88830590e-01 9.30471241e-01 1.48449138e-01
-4.93859142e-01 -2.11833064e-02 -3.13096583e-01 -5.90924561e-01
4.44171160e-01 -1.26527146e-01 -1.77213654e-01 2.49259815e-01
1.69874445e-01 1.04268556e-02 -7.34894395e-01 5.47486305e-01
-6.60426438e-01 -7.65070021e-02 6.96563005e-01 8.40348125e-01
7.50840902e-01 -6.07004352e-02 5.05947530e-01 -3.54655206e-01
1.34907514e-01 -1.35581732e-01 -1.02804852e+00 4.38162029e-01
-4.63164121e-01 -1.89500585e-01 5.71938992e-01 -7.06098378e-01
-9.03669059e-01 4.75661978e-02 3.04684788e-01 -1.03715909e+00
5.21252334e-01 -8.56938735e-02 -1.56991944e-01 -3.87736022e-01
1.95185855e-01 3.96262854e-01 -3.05520177e-01 -4.45761830e-01
2.81418860e-01 2.05715403e-01 5.76992631e-01 -1.25528485e-01
8.55956495e-01 6.49483681e-01 -1.57203227e-01 -3.85083228e-01
-7.14668751e-01 -2.57858306e-01 -3.90272349e-01 -3.57032329e-01
1.00487959e+00 -8.27392757e-01 -1.07370234e+00 6.29259646e-01
-1.01875305e+00 -2.89324261e-02 -3.59017313e-01 3.15845013e-01
-1.25680432e-01 2.65525579e-01 -3.96674156e-01 -6.52529597e-01
-7.39925086e-01 -1.31793463e+00 1.15869188e+00 6.13954782e-01
2.58964807e-01 -4.22711790e-01 -3.78722608e-01 -5.23303188e-02
6.14810109e-01 2.43797258e-01 4.84422773e-01 -2.16040894e-01
-1.10635149e+00 -4.94445898e-02 -7.22026110e-01 3.96657549e-02
-7.44826049e-02 1.02993168e-01 -8.15039635e-01 -5.19490004e-01
-4.42170531e-01 -8.35396051e-02 1.02771986e+00 5.07220447e-01
1.34711945e+00 -2.96595823e-02 -6.08555675e-01 7.86583006e-01
1.52870834e+00 3.64167690e-01 5.37237406e-01 3.96166623e-01
7.89388299e-01 6.97717443e-02 1.02696192e+00 2.20226228e-01
2.74360955e-01 8.92230630e-01 8.04531157e-01 -3.42561722e-01
-3.73677671e-01 -9.93882492e-02 1.11498483e-01 2.56390661e-01
-1.12166936e-02 -3.32402289e-01 -5.85239947e-01 8.70869398e-01
-1.88356006e+00 -1.18127298e+00 -1.21143982e-01 2.13929415e+00
2.98837602e-01 2.80365616e-01 1.56560332e-01 -9.31022987e-02
9.22096729e-01 2.91573435e-01 -6.81566417e-01 1.40559956e-01
-1.80237502e-01 2.11695638e-02 7.72890329e-01 3.80579755e-02
-1.28766286e+00 1.00898254e+00 5.87986755e+00 9.88044620e-01
-1.29284894e+00 1.63415521e-01 5.08680046e-01 -2.00714901e-01
2.15254202e-01 -2.78817743e-01 -1.04302943e+00 7.48993278e-01
2.73093343e-01 -2.25500867e-01 1.70652300e-01 9.35233116e-01
-1.58176675e-01 1.97212398e-01 -7.49552071e-01 9.09581065e-01
-3.69141363e-02 -1.35043383e+00 5.95209002e-02 -4.37942185e-02
5.95152557e-01 1.71326950e-01 3.54722768e-01 2.62976438e-01
8.23307931e-02 -6.95915341e-01 1.08311081e+00 4.53671664e-01
7.78891385e-01 -6.95950866e-01 6.69139624e-01 -1.64718091e-01
-2.05199862e+00 -3.87368023e-01 -3.88884515e-01 3.26235980e-01
3.86482403e-02 1.69591218e-01 -3.21410030e-01 5.64354241e-01
1.39355373e+00 7.50412703e-01 -8.64879489e-01 1.30369008e+00
1.20412953e-01 2.85586357e-01 -3.14692587e-01 -1.57063246e-01
3.99080396e-01 1.03412695e-01 6.01093233e-01 1.10410774e+00
2.83501565e-01 8.92551914e-02 3.13823193e-01 7.82774270e-01
-3.87155861e-02 -7.17480853e-02 -2.86720276e-01 2.21131638e-01
5.50566018e-01 1.46474802e+00 -8.92854452e-01 -5.78844309e-01
-5.16821206e-01 6.73716545e-01 2.43304119e-01 2.19444335e-02
-1.42656553e+00 -4.61970538e-01 8.26071680e-01 1.43155038e-01
1.15515149e+00 -7.17363367e-03 -1.91577263e-02 -8.04283321e-01
9.46690068e-02 -7.82253265e-01 4.87038702e-01 -9.27446961e-01
-8.59050512e-01 8.75857711e-01 -1.17384315e-01 -1.51208913e+00
4.52896684e-01 -5.79491019e-01 -6.63365662e-01 6.41943634e-01
-1.66292381e+00 -1.41351712e+00 -6.86330736e-01 6.35156631e-01
7.32475758e-01 1.23265915e-01 1.20009728e-01 6.18429005e-01
-6.20516837e-01 7.46828794e-01 -1.66134968e-01 2.36868247e-01
6.35536730e-01 -1.01468527e+00 6.25756145e-01 9.03805852e-01
-6.01590201e-02 4.81832683e-01 3.94821525e-01 -6.51651025e-01
-1.36238539e+00 -1.40623641e+00 4.26067859e-01 -4.21978712e-01
6.35227442e-01 -1.21931881e-01 -1.00315070e+00 7.20013261e-01
3.63024622e-01 4.96206015e-01 -1.24576380e-02 -2.75695980e-01
-3.27781796e-01 -5.20121992e-01 -1.02322984e+00 5.05313337e-01
1.11097693e+00 5.70406392e-02 -4.08272147e-01 1.97388485e-01
1.09383559e+00 -5.91700673e-01 -8.57700467e-01 6.32972240e-01
7.30050862e-01 -9.06720519e-01 1.33522439e+00 -6.16146624e-01
1.79922178e-01 -8.65578115e-01 -6.97563067e-02 -8.84860218e-01
-8.55577171e-01 -4.36545283e-01 -1.33426353e-01 1.39239073e+00
1.30795777e-01 -6.09313369e-01 6.31471932e-01 2.51071304e-02
-1.95329487e-01 -1.00215423e+00 -8.07559073e-01 -8.80065739e-01
-4.45901394e-01 -5.75865991e-02 1.02793944e+00 6.71108246e-01
-4.92891848e-01 1.62057668e-01 -6.35124087e-01 4.24043804e-01
8.79252970e-01 4.27056938e-01 6.35265827e-01 -1.20243800e+00
-8.10789242e-02 -4.05032277e-01 -6.59961462e-01 -1.04355288e+00
-3.80588591e-01 -4.50424194e-01 -3.38349119e-02 -1.14239836e+00
2.62555718e-01 -2.52759337e-01 -6.81080580e-01 6.15466297e-01
-5.50478399e-01 6.36517227e-01 6.58105075e-01 1.09547675e-01
-9.52284157e-01 6.44881845e-01 1.44779515e+00 -3.03658724e-01
1.13335863e-01 -3.26961040e-01 -6.80723250e-01 6.43060148e-01
5.28796136e-01 -4.60738689e-01 -3.83228660e-01 -6.41269267e-01
-3.19801986e-01 -2.93500036e-01 6.29087627e-01 -1.27838671e+00
4.14910614e-01 -1.15232207e-01 7.81158268e-01 -1.13786709e+00
3.34804833e-01 -1.06833041e+00 2.57244855e-01 8.38380218e-01
-4.16828990e-02 4.43003446e-01 6.65027082e-01 6.77711666e-01
-1.03467815e-01 2.56295294e-01 1.10094452e+00 8.30682442e-02
-1.23745930e+00 7.49573350e-01 2.22825781e-01 1.25402868e-01
1.47217715e+00 -3.39921355e-01 -3.86197448e-01 2.07993895e-01
-4.10149813e-01 6.48861170e-01 3.74614120e-01 8.34313393e-01
4.94358093e-01 -1.63935113e+00 -6.86348617e-01 2.70830482e-01
1.96089849e-01 -2.30186954e-01 5.40212333e-01 9.00700450e-01
-2.48084232e-01 6.74734592e-01 -5.73472440e-01 -9.40430582e-01
-1.60005355e+00 1.07154679e+00 2.83782601e-01 -2.87140548e-01
-8.27981472e-01 1.02818370e+00 6.00809097e-01 2.83525854e-01
1.54745653e-01 -6.70224190e-01 -2.33323544e-01 1.31717190e-01
7.27634132e-01 2.90275931e-01 -7.62372166e-02 -7.28715539e-01
-6.43498838e-01 1.11678612e+00 -2.53130555e-01 5.39365232e-01
1.02794647e+00 -8.93968344e-02 2.97431707e-01 -5.11021428e-02
8.09295952e-01 -5.48681244e-02 -1.63898075e+00 -4.25454170e-01
-3.80576968e-01 -8.32200527e-01 1.56253189e-01 -5.73166370e-01
-1.62696314e+00 7.61826873e-01 8.32729340e-01 4.10374850e-01
1.41874504e+00 -1.52501494e-01 9.01295543e-01 -1.63851380e-02
2.95146614e-01 -7.13242173e-01 1.50026038e-01 1.66752443e-01
9.41208363e-01 -1.21018171e+00 2.06463382e-01 -3.70600551e-01
-6.10096812e-01 7.40763247e-01 1.10318387e+00 -1.63354829e-01
3.61359686e-01 3.21152061e-01 -7.66635910e-02 -1.42926171e-01
-7.89506853e-01 -4.26986068e-01 4.44910944e-01 5.71209013e-01
6.15263470e-02 -9.53065753e-02 -9.75379050e-02 4.22227681e-01
1.45789281e-01 2.13769749e-02 -2.34869719e-02 7.20033288e-01
-5.36175847e-01 -5.59133351e-01 -4.85842764e-01 3.30031723e-01
-5.08041322e-01 -2.48062417e-01 -2.16077149e-01 1.14208782e+00
3.37720186e-01 5.54790556e-01 9.93150473e-02 -3.05409372e-01
5.44169903e-01 -3.32977742e-01 3.04647595e-01 -4.32945602e-03
-7.19739735e-01 -3.19218785e-02 -3.91549587e-01 -7.03791916e-01
-3.47615004e-01 -7.14888811e-01 -1.10009980e+00 -2.37563938e-01
-6.34408534e-01 4.19508293e-02 2.61580199e-01 5.53999007e-01
6.59376144e-01 1.16732645e+00 3.52006435e-01 -8.65944207e-01
-5.38875461e-01 -8.45900953e-01 -5.23694679e-02 2.76899487e-01
4.83059436e-01 -9.98177469e-01 1.31174490e-01 -7.45859221e-02]
|
[6.25180721282959, -2.123166799545288]
|
4e39a429-ed9c-4191-a63c-2d31a3956fd3
|
learnability-and-algorithm-for-continual
|
2306.12646
| null |
https://arxiv.org/abs/2306.12646v1
|
https://arxiv.org/pdf/2306.12646v1.pdf
|
Learnability and Algorithm for Continual Learning
|
This paper studies the challenging continual learning (CL) setting of Class Incremental Learning (CIL). CIL learns a sequence of tasks consisting of disjoint sets of concepts or classes. At any time, a single model is built that can be applied to predict/classify test instances of any classes learned thus far without providing any task related information for each test instance. Although many techniques have been proposed for CIL, they are mostly empirical. It has been shown recently that a strong CIL system needs a strong within-task prediction (WP) and a strong out-of-distribution (OOD) detection for each task. However, it is still not known whether CIL is actually learnable. This paper shows that CIL is learnable. Based on the theory, a new CIL algorithm is also proposed. Experimental results demonstrate its effectiveness.
|
['Bing Liu', 'Tatsuya Konishi', 'Changnan Xiao', 'Gyuhak Kim']
|
2023-06-22
| null | null | null | null |
['class-incremental-learning', 'incremental-learning']
|
['computer-vision', 'methodology']
|
[ 5.23296595e-01 1.28471285e-01 -6.74714863e-01 -5.48612177e-01
-9.60388899e-01 -6.07826889e-01 3.53770882e-01 4.26056743e-01
-1.33243039e-01 1.02322865e+00 -5.25916338e-01 -4.02230769e-01
-3.21910858e-01 -6.19600475e-01 -1.00822222e+00 -6.60756111e-01
-3.25683445e-01 9.01912630e-01 6.70375645e-01 2.16310248e-01
9.92176533e-02 1.77088857e-01 -1.82603729e+00 7.13829517e-01
9.09558415e-01 1.04242873e+00 4.11074936e-01 8.20348084e-01
-2.39616022e-01 1.01658249e+00 -7.13929951e-01 -2.12423548e-01
4.06594604e-01 -4.46847677e-01 -9.84267473e-01 2.15526447e-01
6.40175223e-01 -1.09767936e-01 2.17768982e-01 8.49029362e-01
2.63728946e-01 1.11668473e-02 7.53665686e-01 -1.77287328e+00
-4.19385791e-01 6.11924112e-01 -5.58585644e-01 1.73339605e-01
1.28632620e-01 -1.85888842e-01 8.83939564e-01 -1.18343806e+00
4.74027395e-01 9.85895932e-01 6.49017215e-01 6.03429377e-01
-1.23161840e+00 -8.82530212e-01 4.84868377e-01 3.67013127e-01
-1.05439758e+00 -3.17245722e-02 8.64153385e-01 -4.45876420e-01
6.88214242e-01 1.60236016e-01 6.31601274e-01 1.08286619e+00
2.64123350e-01 1.32088411e+00 1.44323218e+00 -8.50838363e-01
5.77286422e-01 6.93872273e-01 5.89546025e-01 4.07840371e-01
3.06290716e-01 -1.12329736e-01 -6.14641368e-01 -1.91316992e-01
1.74603298e-01 1.16648316e-01 -2.39873268e-02 -6.67817175e-01
-6.31436884e-01 5.69813669e-01 6.83088750e-02 3.52841079e-01
6.43386319e-03 -2.54260689e-01 4.50468123e-01 8.86569738e-01
9.30123568e-01 4.00049649e-02 -9.39588785e-01 5.08467667e-02
-8.19399714e-01 2.87065089e-01 1.00537610e+00 1.05125797e+00
9.98700798e-01 -2.61893302e-01 -5.64194433e-02 8.20677221e-01
-3.75349462e-01 2.81967133e-01 6.48045897e-01 -4.64877665e-01
4.23882544e-01 5.60555160e-01 -5.38682602e-02 -5.53547859e-01
-2.96137720e-01 -5.75824857e-01 -5.52142382e-01 2.77291447e-01
2.84871608e-01 -1.35080710e-01 -6.66860461e-01 1.60556805e+00
4.70299423e-01 6.86235905e-01 7.63424532e-03 1.83915943e-01
2.44365096e-01 5.28313994e-01 4.35096137e-02 -5.25284946e-01
6.27566159e-01 -8.40743124e-01 -4.14021879e-01 -4.39496636e-01
5.25574803e-01 -4.45830643e-01 1.01708925e+00 7.96534002e-01
-7.97086954e-01 -9.13902462e-01 -1.01486051e+00 2.40293249e-01
-3.38590860e-01 -2.67029017e-01 5.60484111e-01 6.01582766e-01
-9.83362377e-01 2.49961078e-01 -6.80884004e-01 -1.39649361e-01
4.72601235e-01 3.41854960e-01 -1.72874138e-01 -5.60188115e-01
-7.76099861e-01 6.37400746e-01 5.89651108e-01 -2.34961137e-01
-1.35591888e+00 -6.60387456e-01 -5.22409439e-01 -1.33140475e-01
5.82261860e-01 -4.00161743e-01 1.51578593e+00 -1.56121922e+00
-1.14679575e+00 6.29852474e-01 -3.97921085e-01 -5.70787430e-01
5.86906552e-01 -3.81138742e-01 -2.41798952e-01 -6.32856339e-02
1.19484901e-01 4.59139109e-01 1.11101031e+00 -1.54859412e+00
-1.28756690e+00 -4.81867313e-01 1.92599282e-01 2.34582126e-01
-7.72291183e-01 -2.85054594e-01 -1.96998000e-01 -3.93845439e-01
4.25870232e-02 -8.48980665e-01 1.93226039e-01 -3.50358635e-01
-4.22267497e-01 -8.75152409e-01 1.09482801e+00 -2.57055640e-01
1.16092300e+00 -2.05055237e+00 -1.66511908e-01 9.99112725e-02
2.21180007e-01 2.02468649e-01 -3.18407297e-01 2.09336713e-01
-2.85339624e-01 -1.18665351e-02 -2.41603985e-01 -3.61320078e-01
-2.64575541e-01 2.71049678e-01 -5.63796461e-01 1.79614469e-01
1.28139794e-01 7.72270858e-01 -9.26666975e-01 -3.83766115e-01
-1.01715341e-01 3.53581784e-03 -4.53853250e-01 4.03910816e-01
-5.43819308e-01 1.80031165e-01 -3.29235941e-01 6.95370197e-01
7.63311923e-01 -1.64032817e-01 3.25973690e-01 3.37663770e-01
1.88815072e-01 1.92234799e-01 -1.16796970e+00 1.18463814e+00
-4.29514736e-01 4.82994735e-01 -2.99802184e-01 -1.77806795e+00
8.94736946e-01 3.50904763e-01 4.71713334e-01 -5.54769516e-01
-5.14210641e-01 2.77524531e-01 -2.33370867e-02 -5.05503118e-01
4.99753617e-02 -2.44384646e-01 7.99807385e-02 4.95072871e-01
2.37087250e-01 2.22408518e-01 1.65222600e-01 8.13619867e-02
1.21403313e+00 1.36490352e-02 2.63628751e-01 -3.18850547e-01
3.64361763e-01 2.00369850e-01 6.66542172e-01 1.17555296e+00
-1.14359133e-01 2.63008654e-01 3.95313323e-01 -5.48683047e-01
-8.05684447e-01 -1.17584002e+00 -3.57739776e-01 1.62521255e+00
-1.25419372e-03 -1.52567402e-01 -5.68269730e-01 -1.25699317e+00
1.69155449e-01 6.40163541e-01 -7.04416394e-01 -1.81638047e-01
-4.42639351e-01 -5.99243283e-01 -4.00595106e-02 5.34452617e-01
2.67845213e-01 -9.86828804e-01 -3.39336962e-01 3.45258832e-01
1.99482590e-02 -8.70848477e-01 -2.58100539e-01 6.59044445e-01
-1.12939751e+00 -1.49154186e+00 -4.11021829e-01 -1.20208609e+00
7.13519633e-01 5.49919248e-01 1.22355962e+00 -5.34233190e-02
-2.45431572e-01 7.08193421e-01 -3.66806060e-01 -9.56389368e-01
-4.02157515e-01 1.47397876e-01 2.30342135e-01 2.16513053e-01
6.02733850e-01 -3.08905065e-01 -1.39630333e-01 2.87630439e-01
-8.28471839e-01 -1.42179012e-01 6.64893627e-01 9.37718213e-01
8.58320832e-01 4.86630350e-01 1.17002130e+00 -1.54777586e+00
5.79967737e-01 -7.10770190e-01 -4.30209875e-01 6.72375500e-01
-9.43428874e-01 -2.30081141e-01 7.63529003e-01 -8.59371960e-01
-1.15678263e+00 4.00564447e-02 1.97938025e-01 -3.17277163e-01
-4.37799424e-01 6.01546466e-01 -1.69886407e-02 1.20004147e-01
6.54041886e-01 4.86114770e-01 -1.55364543e-01 -4.76230562e-01
1.48192691e-02 7.29964912e-01 3.00256759e-01 -7.89208293e-01
7.42722929e-01 2.66080499e-01 -3.26944888e-01 -8.05681527e-01
-1.56936717e+00 -7.23575115e-01 -9.64930713e-01 -3.44381541e-01
3.80869657e-01 -1.02565265e+00 -3.25431019e-01 4.39125329e-01
-6.78107858e-01 -8.70084763e-01 -4.99709159e-01 2.77403951e-01
-5.07662535e-01 1.43178508e-01 -2.70486861e-01 -1.00118995e+00
-1.59591027e-02 -7.10235894e-01 6.00362122e-01 -4.17428836e-02
-9.17036843e-04 -1.17833793e+00 1.65368482e-01 1.55881450e-01
2.18436778e-01 -1.24777690e-01 1.08185995e+00 -1.02637136e+00
-4.12425041e-01 -3.90127718e-01 2.37386316e-01 7.02779353e-01
2.10093394e-01 -3.47196043e-01 -1.17947471e+00 -6.43984616e-01
2.78912991e-01 -9.60294604e-01 8.79224122e-01 4.21379864e-01
1.63981724e+00 -3.86190176e-01 -4.55460191e-01 1.25802517e-01
1.36301816e+00 3.57254535e-01 3.04870844e-01 3.73194307e-01
4.54646379e-01 4.88239795e-01 9.24135149e-01 1.53633043e-01
2.19306648e-01 4.34773177e-01 2.14971706e-01 1.24662563e-01
-2.57765204e-01 -2.18381584e-01 6.24965966e-01 6.12264395e-01
2.51552641e-01 -1.97312966e-01 -1.05029762e+00 5.56716800e-01
-1.84307086e+00 -1.00931644e+00 -5.80740534e-02 2.48252726e+00
1.09446204e+00 5.30028224e-01 1.27716020e-01 5.27815700e-01
5.94442487e-01 -4.30630684e-01 -8.67935002e-01 -1.25944480e-01
4.13591182e-03 4.62689817e-01 5.29811457e-02 5.56633294e-01
-1.26768756e+00 6.79596245e-01 6.78943157e+00 9.80385184e-01
-1.03805971e+00 2.03569829e-01 8.86026561e-01 9.54094082e-02
-2.02352822e-01 -5.12934178e-02 -1.11005032e+00 1.98302671e-01
7.99543560e-01 -5.96122563e-01 -1.03265502e-01 1.49236608e+00
-4.58182096e-01 -3.52911055e-01 -1.33169997e+00 6.57757282e-01
4.15829986e-01 -8.95032465e-01 5.29606827e-02 -2.08325535e-01
1.11776698e+00 -2.32176930e-02 2.49913037e-01 9.33894396e-01
3.31735969e-01 -5.64633846e-01 5.64594567e-01 3.04342896e-01
7.50454605e-01 -8.79409432e-01 6.55989289e-01 9.93892789e-01
-1.10398173e+00 -6.19172037e-01 -6.53805494e-01 -3.41287851e-02
-6.55546367e-01 8.48851025e-01 -1.23938966e+00 4.17427063e-01
6.85330510e-01 7.74893999e-01 -8.82892132e-01 1.28094542e+00
-1.82368040e-01 1.17807508e+00 -1.06062874e-01 2.22937137e-01
-3.89931654e-03 1.86369956e-01 2.56990880e-01 1.06655836e+00
2.75675803e-01 -1.86182573e-01 8.91319573e-01 3.07684869e-01
-1.72451824e-01 3.74274552e-02 -8.64188790e-01 4.68870968e-01
5.20132363e-01 9.74820971e-01 -6.52963340e-01 -5.15434325e-01
-4.91924137e-01 9.82227385e-01 5.32583833e-01 3.47836703e-01
-4.83925641e-01 1.07552066e-01 3.47381294e-01 2.56056458e-01
2.14158207e-01 1.71072595e-02 -1.79664955e-01 -9.67126667e-01
1.62739411e-01 -7.80033350e-01 7.96892643e-01 -3.02233309e-01
-1.74215126e+00 3.21028918e-01 1.92093909e-01 -1.34196532e+00
-3.75541121e-01 -4.85843420e-01 -5.16826749e-01 4.92756337e-01
-1.73854375e+00 -8.65437865e-01 -1.16756603e-01 8.70159745e-01
1.09002209e+00 -3.90829742e-01 8.40982854e-01 5.46005890e-02
-4.29788828e-01 6.85448229e-01 3.45052660e-01 -8.71804059e-02
8.88172925e-01 -1.58703792e+00 -2.34728575e-01 5.91880441e-01
3.14471394e-01 2.52776563e-01 3.58293772e-01 -6.85810328e-01
-1.27116692e+00 -1.44444382e+00 9.99086142e-01 -6.40954852e-01
4.57849801e-01 -4.98389870e-01 -1.16276932e+00 1.00807130e+00
-1.27839252e-01 2.31691644e-01 8.42454851e-01 5.62559605e-01
-4.43276733e-01 -6.25239432e-01 -9.26570952e-01 -1.11856967e-01
7.79081583e-01 -5.88204384e-01 -5.98712325e-01 7.26160407e-01
6.76540315e-01 -1.54003426e-01 -2.91840523e-01 4.75773901e-01
3.47628295e-01 -9.62300956e-01 7.85676420e-01 -6.75188422e-01
2.36302018e-01 -9.90216583e-02 9.89822447e-02 -1.21145797e+00
-1.59951642e-01 -6.49312660e-02 -4.25673515e-01 1.10903311e+00
5.19043624e-01 -5.53456068e-01 8.50064099e-01 2.70689875e-01
-1.78985223e-01 -6.82731211e-01 -9.00162101e-01 -1.46199942e+00
2.79432505e-01 -9.22262549e-01 -4.15712111e-02 1.16498172e+00
8.21585804e-02 6.01605177e-01 -2.78790772e-01 2.10543990e-01
8.84770930e-01 3.52051526e-01 8.39700699e-01 -1.77722132e+00
-4.35153097e-01 -3.33917066e-02 -1.03786677e-01 -8.76664579e-01
3.28240186e-01 -1.18294179e+00 1.68726146e-01 -1.32135201e+00
6.60302103e-01 -9.63123441e-01 -7.32192218e-01 7.66506374e-01
-3.80736351e-01 -1.55300889e-02 -7.61317834e-02 3.82200390e-01
-1.13373196e+00 2.08295643e-01 8.39049757e-01 -2.15397164e-01
-3.03709000e-01 5.01388073e-01 -5.96445858e-01 6.83852613e-01
9.78536785e-01 -7.68082380e-01 -9.07282650e-01 -1.96703255e-01
9.68330652e-02 -1.03064559e-01 2.42428795e-01 -1.20939767e+00
3.68165433e-01 -3.85670662e-02 4.83094037e-01 -7.06699729e-01
-7.08749518e-02 -8.02540779e-01 -2.16384798e-01 6.12446070e-01
-6.22356534e-01 -3.48186225e-01 2.23526478e-01 9.34506178e-01
-2.32737184e-01 -3.16548705e-01 6.81922019e-01 4.90082391e-02
-8.53897810e-01 5.24954736e-01 6.20732689e-03 2.40045086e-01
1.46042657e+00 2.00308356e-02 -1.66170485e-02 -2.20684946e-01
-8.19660902e-01 2.43248671e-01 -2.94526909e-02 3.71347606e-01
8.12426627e-01 -1.12390602e+00 -8.62935185e-01 2.57625580e-01
4.27988023e-01 2.51323193e-01 2.60736763e-01 4.61468190e-01
2.11773604e-01 2.10095838e-01 7.60771856e-02 -8.56192231e-01
-1.35609198e+00 8.69048715e-01 3.68382549e-04 -4.99687761e-01
-5.98070323e-01 1.02387822e+00 3.67059797e-01 -5.72611511e-01
5.60600579e-01 1.60528436e-01 -1.90775260e-01 1.78194419e-01
8.80964339e-01 1.04056396e-01 1.25591159e-01 1.98239267e-01
3.26154493e-02 2.83111483e-01 -3.53713542e-01 1.89166784e-01
1.44719160e+00 -1.92349683e-02 -1.45227406e-02 1.22048569e+00
9.52967346e-01 -2.44289786e-01 -1.45473981e+00 -5.55410385e-01
4.03605908e-01 -3.65940809e-01 -2.79050976e-01 -9.46455061e-01
-6.26736283e-01 8.59404624e-01 6.90654278e-01 5.54287910e-01
1.28786564e+00 1.95542261e-01 3.50943863e-01 6.84475899e-01
6.83963835e-01 -1.32665634e+00 5.44273853e-01 6.26979589e-01
7.88018107e-01 -1.42536378e+00 -3.19186002e-01 -3.69258076e-01
-4.38528866e-01 1.22888899e+00 8.51265788e-01 -3.57341184e-03
1.01483524e+00 5.45071542e-01 -3.44815463e-01 1.84217289e-01
-1.24767458e+00 -1.52048439e-01 2.63020664e-01 8.86847556e-01
3.37291390e-01 1.33382410e-01 -1.73388105e-02 5.12718081e-01
1.85857743e-01 2.48554692e-01 2.91520000e-01 1.13081467e+00
-9.42979813e-01 -1.23851097e+00 -2.71042079e-01 6.18326485e-01
-9.40052420e-02 4.46780287e-02 -3.07418078e-01 8.44304025e-01
4.24546659e-01 8.50848436e-01 1.30537510e-01 -4.40306515e-01
1.58874437e-01 6.25282943e-01 4.57638443e-01 -1.15728724e+00
-1.62145451e-01 -1.94131672e-01 -2.70197719e-01 -2.47268170e-01
-3.80999237e-01 -6.85442805e-01 -9.63903010e-01 7.84924254e-02
-2.10829467e-01 2.65546501e-01 2.46404812e-01 9.82829094e-01
-2.43796948e-02 3.86770368e-01 1.06644630e+00 -3.77108485e-01
-6.09212399e-01 -9.97799039e-01 -8.23439002e-01 2.89119899e-01
3.09660345e-01 -5.68414450e-01 -5.25214672e-01 3.67890120e-01]
|
[9.696581840515137, 3.407879590988159]
|
cdbc1fed-599b-48ca-bf5f-5c2b0cc3eecc
|
large-scale-adversarial-training-for-vision
|
2006.06195
| null |
https://arxiv.org/abs/2006.06195v2
|
https://arxiv.org/pdf/2006.06195v2.pdf
|
Large-Scale Adversarial Training for Vision-and-Language Representation Learning
|
We present VILLA, the first known effort on large-scale adversarial training for vision-and-language (V+L) representation learning. VILLA consists of two training stages: (i) task-agnostic adversarial pre-training; followed by (ii) task-specific adversarial finetuning. Instead of adding adversarial perturbations on image pixels and textual tokens, we propose to perform adversarial training in the embedding space of each modality. To enable large-scale training, we adopt the "free" adversarial training strategy, and combine it with KL-divergence-based regularization to promote higher invariance in the embedding space. We apply VILLA to current best-performing V+L models, and achieve new state of the art on a wide range of tasks, including Visual Question Answering, Visual Commonsense Reasoning, Image-Text Retrieval, Referring Expression Comprehension, Visual Entailment, and NLVR2.
|
['Yen-Chun Chen', 'Yu Cheng', 'Linjie Li', 'Jingjing Liu', 'Chen Zhu', 'Zhe Gan']
|
2020-06-11
| null |
http://proceedings.neurips.cc/paper/2020/hash/49562478de4c54fafd4ec46fdb297de5-Abstract.html
|
http://proceedings.neurips.cc/paper/2020/file/49562478de4c54fafd4ec46fdb297de5-Paper.pdf
|
neurips-2020-12
|
['visual-commonsense-reasoning', 'visual-entailment']
|
['reasoning', 'reasoning']
|
[ 5.99325836e-01 2.47225434e-01 5.29648252e-02 -2.83096731e-01
-1.01985538e+00 -8.78446877e-01 8.69895995e-01 -1.44840464e-01
-4.23449904e-01 3.56144756e-01 4.02851105e-01 -6.47067964e-01
4.45987225e-01 -6.29306078e-01 -1.02993226e+00 -2.83525050e-01
3.32357168e-01 3.75068575e-01 -5.78731745e-02 -2.63357371e-01
5.78113534e-02 5.82411826e-01 -9.30493772e-01 5.83939373e-01
6.38270676e-01 1.05431867e+00 -3.18879843e-01 9.01219964e-01
-2.08158493e-01 1.59114623e+00 -6.59632742e-01 -7.56079018e-01
1.03620790e-01 -5.28016329e-01 -1.17536628e+00 5.14120944e-02
8.03701937e-01 -4.16440666e-01 -8.12088490e-01 9.56562102e-01
5.27394235e-01 3.17415327e-01 9.09169972e-01 -1.30139434e+00
-1.58239651e+00 3.57049346e-01 -5.35705864e-01 1.50349051e-01
5.10098577e-01 6.76906645e-01 9.94640589e-01 -9.34274852e-01
7.75682569e-01 1.61261892e+00 4.18081850e-01 1.03065634e+00
-1.48744750e+00 -5.78335643e-01 1.89191297e-01 2.71171182e-01
-1.29500437e+00 -4.95239496e-01 1.00041115e+00 -5.14965534e-01
1.03003454e+00 2.20569670e-01 3.23136538e-01 1.53701973e+00
7.24664107e-02 9.38148260e-01 1.20185804e+00 -3.71573031e-01
1.71279833e-01 3.43594179e-02 -2.52629429e-01 6.76921606e-01
-4.21377212e-01 2.93427203e-02 -2.28309423e-01 -7.99782760e-03
6.20383561e-01 -2.20647305e-01 -2.30812639e-01 -2.33577520e-01
-1.15729129e+00 9.90956426e-01 9.13755357e-01 -8.12290609e-02
-3.18586886e-01 5.60424626e-01 7.55811095e-01 5.98286152e-01
3.60790700e-01 6.65306807e-01 -1.26463234e-01 3.53572279e-01
-6.00770831e-01 4.10885632e-01 6.63752437e-01 8.01769972e-01
5.36526620e-01 3.67590040e-01 -7.75902748e-01 8.45799088e-01
2.14348838e-01 7.20358074e-01 5.02683818e-01 -1.18361580e+00
4.16424483e-01 3.53230864e-01 -2.33477876e-01 -8.96385372e-01
9.03222337e-02 1.69696882e-01 -8.91720772e-01 5.01557946e-01
3.71464223e-01 -2.16522273e-02 -1.28855765e+00 1.83156347e+00
5.49960136e-02 -1.15728833e-01 3.38564873e-01 7.99624801e-01
1.17871284e+00 6.26739085e-01 4.43655312e-01 2.19712973e-01
1.33339453e+00 -1.10280240e+00 -5.67272246e-01 -4.49708164e-01
2.57302374e-01 -7.80524015e-01 1.47600543e+00 7.85464719e-02
-1.33337927e+00 -5.39016485e-01 -7.88846135e-01 -7.09589362e-01
-6.27721071e-01 -3.05980116e-01 5.58182478e-01 2.55360812e-01
-9.89439607e-01 1.29602894e-01 -3.78013611e-01 -2.85858046e-02
9.95068789e-01 -3.40439826e-02 -4.72200722e-01 -4.48045969e-01
-1.26774132e+00 1.02874613e+00 1.68930948e-01 -4.82383631e-02
-1.34889555e+00 -7.95426250e-01 -1.21426201e+00 -2.34114707e-01
3.45061719e-01 -1.04357994e+00 1.03463340e+00 -1.45847940e+00
-1.35268617e+00 1.59558523e+00 -5.43105043e-02 -5.15623152e-01
6.57055616e-01 -2.07493633e-01 -2.78958857e-01 2.90600300e-01
-7.49568343e-02 1.00010729e+00 1.46428919e+00 -1.40824699e+00
3.84170651e-01 -2.25661427e-01 5.93509734e-01 1.72329396e-01
-1.18606515e-01 -2.36571599e-02 -4.43964124e-01 -9.37701762e-01
-4.26688105e-01 -7.82639623e-01 -8.80621448e-02 5.14519453e-01
-4.60947573e-01 -2.69858778e-01 6.41039014e-01 -8.61135066e-01
7.11452246e-01 -2.25181055e+00 3.72678101e-01 -7.90401548e-02
5.24253845e-01 3.75629991e-01 -7.13612556e-01 2.71962583e-01
-4.35775340e-01 1.15111411e-01 -2.71762878e-01 -4.02156413e-01
2.38632143e-01 3.48495066e-01 -6.89568698e-01 3.01524878e-01
6.77906036e-01 1.65854108e+00 -1.00742114e+00 -7.39858270e-01
2.51635939e-01 3.70110124e-01 -6.36304438e-01 5.03538847e-01
-6.52489543e-01 4.33385491e-01 -3.54888588e-01 7.30342031e-01
6.76128685e-01 -2.59146780e-01 -2.12746471e-01 -3.95709097e-01
4.30431277e-01 -1.86244115e-01 -5.47774434e-01 1.81656790e+00
-6.47825539e-01 9.70043361e-01 5.51723130e-02 -1.05454719e+00
4.34567064e-01 4.24186103e-02 6.45153672e-02 -9.10803199e-01
3.36976983e-02 -2.95050919e-01 -2.41842434e-01 -7.29100645e-01
2.31238708e-01 -2.47079194e-01 -1.79575413e-01 2.74621129e-01
2.03905493e-01 -6.16350532e-01 -2.15524629e-01 6.44344151e-01
1.14032114e+00 2.21286371e-01 2.67602891e-01 1.84276238e-01
6.80721998e-01 -1.05363138e-01 -1.02707364e-01 6.94480956e-01
-4.70104545e-01 5.67107737e-01 6.68989122e-01 -2.81270444e-01
-1.10474575e+00 -1.35405862e+00 9.60913226e-02 1.37147939e+00
-7.37200305e-02 -7.96554461e-02 -7.70348430e-01 -9.36320841e-01
2.85191476e-01 8.73181880e-01 -1.06210852e+00 -4.68937725e-01
-4.88627732e-01 -1.52268350e-01 1.04113221e+00 7.33493090e-01
5.01660347e-01 -1.52350247e+00 -1.46573290e-01 -3.24007601e-01
9.05873068e-03 -1.28602207e+00 -6.34650707e-01 6.45458624e-02
-2.84424931e-01 -1.01923263e+00 -7.41510093e-01 -7.25617349e-01
6.89792037e-01 -1.53277248e-01 1.53120196e+00 1.32653480e-02
-3.88667077e-01 7.66222596e-01 -2.66633838e-01 -3.09915751e-01
-5.74539781e-01 -4.43815261e-01 -5.19883096e-01 -1.74597755e-01
4.03095484e-02 -4.97874558e-01 -6.10881805e-01 -1.13270871e-01
-1.21864212e+00 -2.71389097e-01 5.29870152e-01 1.08362842e+00
6.74364030e-01 -6.48798943e-01 4.61036265e-01 -9.29127812e-01
7.82952547e-01 -3.88460129e-01 -3.46320570e-01 3.94651949e-01
-2.66122699e-01 -2.39953343e-02 8.24151695e-01 -8.18207383e-01
-7.41850674e-01 -1.96395218e-01 -2.96174407e-01 -1.09572506e+00
-7.22700208e-02 2.47531265e-01 -2.34429747e-01 -3.55725884e-01
9.59891617e-01 2.51928210e-01 -1.19636446e-01 7.45245218e-02
1.18173134e+00 2.84553111e-01 9.17316020e-01 -4.79317129e-01
1.11832309e+00 4.11160618e-01 -1.03214653e-02 -6.55573308e-01
-1.01426530e+00 -5.23506589e-02 -4.35492158e-01 2.90151611e-02
1.33248949e+00 -9.06494141e-01 -8.68319869e-01 3.88928324e-01
-1.37874484e+00 -7.15176523e-01 -6.99764729e-01 -6.81810975e-02
-7.29077280e-01 4.43850100e-01 -5.95375657e-01 -5.12103319e-01
-5.62952757e-01 -8.13151419e-01 1.03374565e+00 -3.82101350e-02
-1.42614573e-01 -1.17138457e+00 4.14242707e-02 7.35981584e-01
4.62265164e-01 6.30533159e-01 1.06387782e+00 -6.00537896e-01
-5.45877457e-01 -1.22076340e-01 -6.34512782e-01 8.62378955e-01
-2.83928782e-01 -2.31200159e-01 -1.16613650e+00 -3.22919965e-01
-1.93709582e-01 -1.26078045e+00 1.02386749e+00 4.64431420e-02
1.63663697e+00 -4.32316363e-01 1.34922877e-01 9.90797400e-01
1.23493409e+00 -3.34983677e-01 8.53939652e-01 9.63370427e-02
1.17894769e+00 3.45719963e-01 2.75632650e-01 8.99044499e-02
4.17122871e-01 3.76890481e-01 8.57119441e-01 -3.58276814e-01
-6.08079791e-01 -3.76036674e-01 4.06974494e-01 1.30799755e-01
1.01693071e-01 -4.04373258e-01 -8.46677363e-01 5.91260195e-01
-1.59727943e+00 -1.07883906e+00 2.55747497e-01 1.69966340e+00
9.02172685e-01 1.49266189e-02 -4.50601690e-02 -2.99860001e-01
3.06970388e-01 6.24607980e-01 -9.86503184e-01 -7.35209823e-01
-2.76872844e-01 5.12243748e-01 2.52862930e-01 5.67796111e-01
-1.15175188e+00 1.44659746e+00 6.64702511e+00 8.29704523e-01
-1.11997664e+00 2.67813325e-01 5.23247659e-01 -1.65964998e-02
-7.42385209e-01 -1.68878987e-01 1.10897817e-01 1.29491746e-01
5.86459398e-01 3.33811110e-03 7.73369908e-01 7.79807210e-01
-3.64293069e-01 2.96232402e-01 -1.21956682e+00 1.07725227e+00
4.72780675e-01 -1.36671007e+00 5.74560344e-01 -2.88507491e-01
7.75232375e-01 6.83215708e-02 3.12535197e-01 6.12119973e-01
4.70220864e-01 -1.50588763e+00 6.28176570e-01 5.61806440e-01
1.17390490e+00 -6.08526647e-01 4.08287048e-01 -5.47133498e-02
-7.47761250e-01 1.36024177e-01 -2.30214268e-01 2.69138366e-01
5.86072430e-02 1.86455056e-01 -3.78554314e-01 2.73435742e-01
4.67367947e-01 6.52218282e-01 -6.59731507e-01 1.06084123e-01
-5.41594386e-01 4.44097519e-01 1.82319805e-01 3.25324029e-01
3.62594277e-01 1.23229019e-01 5.35061359e-01 1.27400672e+00
-4.31329697e-01 4.50666398e-02 1.68085128e-01 1.17966831e+00
-7.51352012e-01 -4.92969975e-02 -7.79499948e-01 -2.89752424e-01
1.69808403e-01 9.96806502e-01 -6.93978593e-02 -4.62679595e-01
-4.84413952e-01 1.74089086e+00 5.46405792e-01 6.67771280e-01
-1.06938088e+00 -4.40458745e-01 6.03023946e-01 -9.36959013e-02
3.65014553e-01 1.79027617e-02 -4.43110503e-02 -1.25094938e+00
-9.67162326e-02 -1.17831933e+00 3.89149606e-01 -1.22137022e+00
-1.70956123e+00 5.97073257e-01 -2.51973361e-01 -6.91609800e-01
-4.05860484e-01 -7.78856158e-01 -8.01379979e-01 1.05478179e+00
-1.72154653e+00 -1.79371095e+00 -3.68124545e-01 1.34557223e+00
5.25066137e-01 -2.27492005e-01 1.01171196e+00 4.21006866e-02
-3.49145681e-01 9.61417317e-01 -1.44540682e-01 3.52243781e-01
9.24801469e-01 -1.45477164e+00 5.00437975e-01 6.48199081e-01
4.75410651e-03 3.44667077e-01 6.00627840e-01 -2.72781610e-01
-1.48534536e+00 -1.26409566e+00 5.65147102e-01 -8.33410084e-01
1.09096742e+00 -5.15031040e-01 -8.98182988e-01 1.02687192e+00
4.87375289e-01 6.05544865e-01 5.86441875e-01 -1.43530324e-01
-1.02881444e+00 6.29150420e-02 -1.32610846e+00 8.23465049e-01
9.52347279e-01 -1.28516269e+00 -8.13862443e-01 6.85345888e-01
9.99891579e-01 -5.23053765e-01 -9.13533449e-01 1.86800495e-01
4.07905489e-01 -4.14087176e-01 1.49226439e+00 -1.19370532e+00
9.32149827e-01 -3.97618189e-02 -4.15624827e-01 -1.15801024e+00
-1.72175705e-01 -7.00813711e-01 -2.69620359e-01 1.16276813e+00
9.86305177e-02 -4.00548369e-01 4.21493620e-01 4.18444365e-01
-3.44375484e-02 -6.63314879e-01 -7.22586513e-01 -4.53860760e-01
6.67630553e-01 -5.22660553e-01 2.20180437e-01 1.12120295e+00
-3.91531527e-01 5.61106563e-01 -4.66241688e-01 -1.25449030e-02
5.70985079e-01 3.04199532e-02 9.88682508e-01 -5.73049307e-01
-4.98120338e-01 -4.32974070e-01 -2.74716407e-01 -1.00759482e+00
8.24882507e-01 -1.00065315e+00 -1.82617567e-02 -1.38760436e+00
1.71198770e-01 8.97320881e-02 -2.67871380e-01 6.83770716e-01
-3.86867076e-01 6.08229518e-01 3.11524481e-01 1.68197617e-01
-7.85888970e-01 4.92789000e-01 1.68312657e+00 -5.62643766e-01
2.69743532e-01 -4.04753953e-01 -8.10011446e-01 7.27642000e-01
3.99078399e-01 -1.75072014e-01 -4.92581546e-01 -5.71000278e-01
2.42194295e-01 -7.99076185e-02 9.48303401e-01 -4.36239779e-01
-2.29509339e-01 -2.92679608e-01 6.65599942e-01 -1.54962495e-01
4.91064280e-01 -5.99426091e-01 -5.30849516e-01 2.53570110e-01
-7.31796741e-01 -3.24618965e-02 4.70705777e-01 5.35534143e-01
-2.72017300e-01 2.84037031e-02 1.03633380e+00 -1.82821482e-01
-8.37330103e-01 3.68533313e-01 -1.29109293e-01 6.51613832e-01
9.66401637e-01 1.79336175e-01 -6.22447670e-01 -6.24474168e-01
-8.77053678e-01 3.34971249e-01 5.74245393e-01 3.97066146e-01
8.58178735e-01 -1.31649864e+00 -8.10501337e-01 -5.15889190e-02
4.63034570e-01 -1.62369177e-01 4.77664083e-01 4.09781665e-01
-3.82804066e-01 9.30606350e-02 -2.90509760e-01 -3.19387138e-01
-1.04651785e+00 1.13442183e+00 3.57714355e-01 -3.91797245e-01
-4.95242178e-01 1.09000230e+00 5.24626493e-01 -3.94721240e-01
2.25377664e-01 2.32103858e-02 -4.25085463e-02 -2.27329776e-01
3.61489236e-01 -6.48052469e-02 -4.16507512e-01 -6.10343874e-01
-4.37283456e-01 5.75257540e-01 -1.53049707e-01 -1.36116788e-01
8.63811553e-01 -1.82396080e-02 -1.17082849e-01 2.98665315e-01
1.44214320e+00 2.89259665e-02 -1.15374172e+00 -3.18214685e-01
-5.18119037e-01 -2.19103575e-01 6.00782558e-02 -1.07369590e+00
-1.00254560e+00 1.10734606e+00 4.26987052e-01 5.70910722e-02
1.34192693e+00 3.75349641e-01 7.73117006e-01 5.65857768e-01
-4.00702119e-01 -7.25178361e-01 6.51404619e-01 6.33193135e-01
1.40541124e+00 -1.42904568e+00 -8.69412720e-02 -2.81891171e-02
-1.02105665e+00 7.27580547e-01 6.23733580e-01 -5.16075790e-01
3.20341617e-01 -8.32424909e-02 2.53799140e-01 -2.34320149e-01
-6.68471992e-01 -3.44911873e-01 6.99453056e-01 9.44345653e-01
3.69550735e-01 -1.17336325e-01 1.83755636e-01 1.34881407e-01
2.48949248e-02 -1.75908551e-01 1.76560096e-02 7.17470050e-01
1.90530896e-01 -9.35412943e-01 -2.47871444e-01 2.10579291e-01
-5.01734734e-01 -5.33028007e-01 -7.04557240e-01 8.84247303e-01
4.60721739e-02 6.30657554e-01 6.16417229e-02 -1.97245896e-01
4.50414181e-01 1.38562515e-01 8.94665718e-01 -4.35110211e-01
-6.55097306e-01 -4.35387135e-01 7.73436874e-02 -9.48466778e-01
-2.81445801e-01 -3.43150944e-01 -1.09249794e+00 -3.80514443e-01
2.01923773e-01 -3.83182287e-01 4.72994208e-01 8.98435891e-01
2.41924047e-01 8.19782257e-01 4.86732244e-01 -7.03966975e-01
-7.08909392e-01 -8.16617131e-01 -1.39281571e-01 1.08129466e+00
6.22453272e-01 -2.42519155e-01 -3.66152972e-01 3.22982430e-01]
|
[10.887338638305664, 1.7763948440551758]
|
1f7cb39e-92e4-460b-b806-236eff40504a
|
decentralized-optimization-with-distributed
|
2208.11224
| null |
https://arxiv.org/abs/2208.11224v1
|
https://arxiv.org/pdf/2208.11224v1.pdf
|
Decentralized Optimization with Distributed Features and Non-Smooth Objective Functions
|
We develop a new consensus-based distributed algorithm for solving learning problems with feature partitioning and non-smooth convex objective functions. Such learning problems are not separable, i.e., the associated objective functions cannot be directly written as a summation of agent-specific objective functions. To overcome this challenge, we redefine the underlying optimization problem as a dual convex problem whose structure is suitable for distributed optimization using the alternating direction method of multipliers (ADMM). Next, we propose a new method to solve the minimization problem associated with the ADMM update step that does not rely on any conjugate function. Calculating the relevant conjugate functions may be hard or even unfeasible, especially when the objective function is non-smooth. To obviate computing any conjugate function, we solve the optimization problem associated with each ADMM iteration in the dual domain utilizing the block coordinate descent algorithm. Unlike the existing related algorithms, the proposed algorithm is fully distributed and does away with the conjugate of the objective function. We prove theoretically that the proposed algorithm attains the optimal centralized solution. We also confirm its network-wide convergence via simulations.
|
['Stefan Werner', 'Reza Arablouei', 'Naveen K. D. Venkategowda', 'Cristiano Gratton']
|
2022-08-23
| null | null | null | null |
['distributed-optimization']
|
['methodology']
|
[-3.06908280e-01 -1.53931761e-02 5.98470829e-02 -2.31601402e-01
-7.54200399e-01 -5.74641705e-01 1.16920762e-01 1.35028988e-01
-6.00056946e-01 1.03582120e+00 -1.46025628e-01 -2.46461496e-01
-5.77110589e-01 -6.33967757e-01 -6.42413914e-01 -1.19638169e+00
-1.15049705e-01 6.46281898e-01 -3.23638529e-01 -2.35650409e-02
1.35559782e-01 2.29161754e-01 -7.97426701e-01 -2.50036955e-01
1.10563493e+00 9.28375781e-01 2.77728677e-01 4.89861161e-01
8.40890557e-02 6.55767322e-01 -5.43616831e-01 -8.74095112e-02
6.22461021e-01 -5.17262876e-01 -9.08507228e-01 4.69531924e-01
4.26234961e-01 -3.81351322e-01 -1.60184145e-01 1.29906332e+00
5.03067851e-01 3.18901122e-01 4.55043167e-01 -1.68112922e+00
-4.91405666e-01 3.73620749e-01 -9.97308731e-01 -8.45134910e-03
-7.53260329e-02 -3.67703289e-01 1.01106155e+00 -9.49325383e-01
4.82139528e-01 1.13063848e+00 5.84993541e-01 2.83340961e-01
-1.23611259e+00 -4.03902054e-01 5.23087919e-01 2.46501952e-01
-1.56251609e+00 -2.03530714e-01 8.91884804e-01 -2.24925682e-01
2.88237721e-01 3.58433396e-01 6.03519440e-01 2.26678371e-01
-4.24766205e-02 8.70391428e-01 1.07597852e+00 -2.40473166e-01
5.87370396e-01 -6.47434443e-02 1.63162112e-01 7.75431693e-01
6.30524755e-01 -4.03970212e-01 -4.05324519e-01 -6.20390534e-01
4.65130717e-01 1.88580677e-01 -4.81560558e-01 -5.74766815e-01
-1.40942013e+00 9.92470264e-01 4.22189623e-01 -7.68063590e-02
-7.70134389e-01 3.45268458e-01 1.27406016e-01 5.68235278e-01
7.15925574e-01 -1.37593657e-01 -3.99052441e-01 2.47930720e-01
-1.06802833e+00 3.29882562e-01 1.07889676e+00 7.64490962e-01
1.08381307e+00 5.87339923e-02 2.23801032e-01 6.52924955e-01
7.97263741e-01 5.89011312e-01 4.41643856e-02 -1.11364710e+00
5.86147487e-01 3.41409147e-01 3.07020813e-01 -1.36896408e+00
-3.82676899e-01 -5.70425808e-01 -1.16342759e+00 3.42080563e-01
6.46299839e-01 -8.44228327e-01 -7.80749246e-02 1.71283734e+00
9.56231952e-01 3.72409284e-01 4.38155513e-03 1.42563462e+00
1.99233502e-01 8.58224750e-01 -3.87938738e-01 -6.29003644e-01
9.03319895e-01 -1.35629547e+00 -7.49788702e-01 1.13320183e-02
7.02862740e-01 -6.28737807e-01 1.40727922e-01 2.70453036e-01
-1.21983743e+00 4.79882881e-02 -9.64543045e-01 1.95089653e-01
1.40359566e-01 1.85907960e-01 5.64237475e-01 2.94530928e-01
-1.27490807e+00 3.45512539e-01 -8.47111166e-01 -1.55488178e-01
1.61734924e-01 5.71072221e-01 -3.29190463e-01 -7.76218697e-02
-7.21384585e-01 6.64928436e-01 1.05984457e-01 5.87708294e-01
-8.69812965e-01 -6.76027179e-01 -4.49470460e-01 -1.59273654e-01
4.81846571e-01 -1.02524734e+00 8.98165166e-01 -1.16338158e+00
-1.61414683e+00 3.24096441e-01 -2.19459400e-01 -1.52050361e-01
7.95278013e-01 -3.13365683e-02 2.23911434e-01 2.27101650e-02
6.02422692e-02 -8.44209194e-02 1.09103763e+00 -1.21998656e+00
-6.94963872e-01 -3.36160570e-01 3.46321493e-01 5.81325531e-01
-6.81361198e-01 -3.15998137e-01 -2.14818344e-01 -5.34934640e-01
7.15733021e-02 -1.01944900e+00 -6.10759735e-01 3.98300678e-01
-3.58797550e-01 8.20830557e-03 1.03034019e+00 -5.82579494e-01
1.07304657e+00 -1.94210351e+00 6.31981015e-01 5.64772487e-01
7.67965853e-01 2.31872946e-02 -2.42770299e-01 5.94570577e-01
3.15372139e-01 -2.58490384e-01 -3.38748664e-01 -4.49816078e-01
1.57223821e-01 2.84153432e-01 1.17743477e-01 1.28124678e+00
-4.90398735e-01 5.08601010e-01 -1.03287184e+00 -4.83439505e-01
-5.89545295e-02 2.88507730e-01 -7.97303140e-01 2.35505728e-03
1.34442002e-01 5.01410186e-01 -9.18751061e-01 2.78031081e-01
1.09835649e+00 -4.69941318e-01 6.61735833e-01 -3.09515715e-01
-2.17695639e-01 -2.93964416e-01 -1.89461493e+00 1.79524064e+00
-5.01262009e-01 3.74030799e-01 1.26189017e+00 -1.74884355e+00
6.44822240e-01 2.62939006e-01 1.02236998e+00 -5.15359342e-02
-9.49339494e-02 3.33862811e-01 -1.47620305e-01 -1.97378591e-01
2.38842249e-01 3.96355130e-02 2.05555961e-01 9.99500573e-01
-1.24171399e-01 2.68872648e-01 1.49376497e-01 3.32370341e-01
1.08628380e+00 -3.37770551e-01 1.22608328e-02 -8.15421402e-01
9.53501761e-01 -3.27861644e-02 8.04227531e-01 6.12293303e-01
-7.77716860e-02 1.94084316e-01 2.73496360e-01 -4.75715578e-01
-7.44583726e-01 -8.86158943e-01 1.55870944e-01 1.01609063e+00
5.00804245e-01 -4.37262148e-01 -7.29300201e-01 -8.74254227e-01
2.35129938e-01 -1.08108923e-01 -1.92614108e-01 1.21393017e-01
-3.77349645e-01 -1.10156322e+00 -3.52074578e-02 -9.84153599e-02
6.20402157e-01 -1.29982859e-01 -2.04536468e-01 4.36627239e-01
-3.33916873e-01 -6.41099274e-01 -9.02480543e-01 -6.96421042e-02
-7.64969707e-01 -1.17655349e+00 -1.07838273e+00 -9.78263915e-01
1.34828186e+00 7.32297838e-01 8.27488422e-01 3.47889632e-01
9.66080576e-02 7.89325595e-01 -1.79752950e-02 -1.93698741e-02
7.18454123e-02 1.31933048e-01 3.11026186e-01 6.74736559e-01
-2.06451386e-01 -6.38003707e-01 -8.56970847e-01 3.69215816e-01
-8.33350360e-01 7.28470758e-02 3.65360975e-01 8.79014969e-01
7.36192763e-01 1.30303964e-01 8.28654706e-01 -7.29184091e-01
9.64255810e-01 -7.16979444e-01 -8.72731507e-01 4.71510410e-01
-5.53225577e-01 -3.61010842e-02 7.29208529e-01 -3.29513043e-01
-8.97736251e-01 3.08114678e-01 5.27423024e-01 -3.15236270e-01
5.49254179e-01 8.90844226e-01 7.38123208e-02 -5.87034523e-01
2.44643316e-01 2.67124236e-01 4.44467962e-01 -3.73665720e-01
4.87206548e-01 6.47133291e-01 5.08779697e-02 -8.66636753e-01
1.12240493e+00 8.36173892e-01 2.26548299e-01 -7.64167130e-01
-7.23997414e-01 -4.13706154e-01 -3.46319377e-01 -3.67507309e-01
3.52283418e-01 -9.66626465e-01 -1.14927375e+00 4.08733547e-01
-1.27772164e+00 -1.40334442e-01 -2.08311006e-01 7.14601934e-01
-5.34008384e-01 6.84117258e-01 -4.84604865e-01 -7.61680126e-01
-5.08264184e-01 -9.31743562e-01 5.85213184e-01 9.42931026e-02
1.12125240e-01 -1.47343445e+00 3.98550510e-01 2.12584630e-01
6.53778851e-01 1.99128777e-01 3.98612767e-01 -4.32211399e-01
-5.81384540e-01 -3.18472460e-02 -1.33061752e-01 1.99141666e-01
1.68477088e-01 -1.75398722e-01 -2.38462135e-01 -8.61983955e-01
2.86552221e-01 -2.09532499e-01 3.45356524e-01 4.01121676e-01
7.89113998e-01 -7.16592968e-01 -1.69191107e-01 6.87857211e-01
1.56186128e+00 -3.90367091e-01 -1.00213170e-01 2.60884583e-01
7.58924067e-01 3.74666125e-01 4.60557669e-01 1.01163232e+00
7.53606319e-01 5.69835305e-01 5.34353793e-01 -2.66676337e-01
1.95063561e-01 2.77450413e-01 4.89702016e-01 1.25216317e+00
-7.20982179e-02 -9.33688357e-02 -7.13352621e-01 5.08511543e-01
-2.57432604e+00 -7.34721839e-01 -4.16780293e-01 2.09535956e+00
7.74183154e-01 -8.26201379e-01 5.65946773e-02 -2.13490501e-01
1.04389763e+00 9.72665474e-02 -7.20623314e-01 -2.17912868e-01
-3.95378377e-03 -1.91692710e-01 6.92251861e-01 6.41550899e-01
-8.96762073e-01 4.29105580e-01 5.90286636e+00 6.23218238e-01
-1.09507501e+00 4.25924778e-01 2.87491232e-01 -2.43416518e-01
-3.09377342e-01 3.69854346e-02 -4.16351110e-01 5.52970290e-01
5.50805151e-01 -6.21131897e-01 9.42549407e-01 7.24948049e-01
5.79792202e-01 -6.78873509e-02 -8.69791150e-01 1.06541371e+00
-1.20418124e-01 -1.26162922e+00 -3.12403172e-01 2.74383515e-01
1.16343784e+00 8.95342082e-02 -9.93411690e-02 -9.19567123e-02
5.02769411e-01 -4.63522911e-01 6.53795898e-01 3.10865074e-01
2.45739833e-01 -6.43829763e-01 4.52569962e-01 4.93644297e-01
-1.24887872e+00 3.00216896e-04 -5.22753239e-01 -2.92602122e-01
2.13406652e-01 9.28403080e-01 -1.28571793e-01 7.30202079e-01
4.03730094e-01 8.95275950e-01 7.36018941e-02 1.26386964e+00
-5.54992184e-02 3.27733666e-01 -8.16084862e-01 1.53700188e-01
3.55701387e-01 -8.52269828e-01 7.31158078e-01 7.51407981e-01
4.44862217e-01 -2.82135867e-02 6.98954523e-01 6.34484589e-01
-1.84370190e-01 3.84878427e-01 -2.62892485e-01 2.33694002e-01
4.99082237e-01 1.60618567e+00 -4.73607868e-01 -3.44487101e-01
-7.56561220e-01 1.12234116e+00 5.62763393e-01 8.14784884e-01
-8.32045913e-01 -2.20310822e-01 9.34143484e-01 -3.26079518e-01
1.92752838e-01 -4.03788030e-01 -7.53623098e-02 -1.50380170e+00
2.80525178e-01 -7.16538429e-01 4.49561626e-01 -8.56911689e-02
-1.57306349e+00 2.96016693e-01 -2.96119690e-01 -1.17544937e+00
-9.33079869e-02 -1.78856045e-01 -8.02994609e-01 9.73211408e-01
-1.72630560e+00 -1.04091704e+00 -2.41890028e-01 1.19479406e+00
4.13665101e-02 4.79198955e-02 4.85326320e-01 6.34177029e-01
-7.92532265e-01 3.64021033e-01 7.19953120e-01 -1.50360584e-01
6.87835872e-01 -1.17253482e+00 -6.90431118e-01 9.07431602e-01
-2.97962397e-01 3.47116381e-01 4.71502155e-01 -3.99523288e-01
-1.96898663e+00 -1.17746258e+00 7.30580807e-01 2.51974881e-01
8.88248920e-01 -1.68866917e-01 -4.22683626e-01 6.32502258e-01
3.11455876e-01 4.99653608e-01 6.97931707e-01 -8.67299885e-02
1.75710887e-01 -5.07522464e-01 -1.16300642e+00 2.23516017e-01
7.90725827e-01 -2.45390683e-01 -7.42030889e-02 8.54352832e-01
1.41788468e-01 -3.16653341e-01 -7.97224045e-01 -1.00740217e-01
1.48758546e-01 -3.98959816e-01 8.25373709e-01 -5.46991646e-01
-7.69752115e-02 -8.27059627e-01 -2.32813805e-01 -1.58049846e+00
-4.89047348e-01 -1.03424299e+00 -4.09893036e-01 9.37777102e-01
2.51261890e-01 -1.00977600e+00 8.90881538e-01 6.43542230e-01
4.75255176e-02 -5.36642551e-01 -1.18175590e+00 -7.89151251e-01
-7.77915791e-02 1.77810844e-02 2.74352998e-01 1.27600121e+00
1.69853896e-01 1.14475049e-01 -5.40139616e-01 5.14497578e-01
1.16908634e+00 5.14666259e-01 8.13061595e-01 -1.03802419e+00
-4.45003182e-01 -2.13885412e-01 -8.88227373e-02 -1.25954401e+00
4.36252922e-01 -1.17623389e+00 1.36563079e-02 -1.67356169e+00
2.40988523e-01 -6.95548356e-01 -3.48656148e-01 4.98642117e-01
-4.11634194e-03 -3.05917095e-02 4.19023246e-01 3.77064496e-01
-9.68761444e-01 6.48772061e-01 1.28460634e+00 -3.83059323e-01
-1.23682514e-01 8.62189685e-04 -6.35469437e-01 5.75247645e-01
6.36457145e-01 -5.86380482e-01 -5.69815218e-01 -9.29004014e-01
2.95758933e-01 2.62083799e-01 2.19373450e-01 -6.15540385e-01
9.05375779e-01 -5.58842957e-01 -2.83746332e-01 -1.82580337e-01
6.28687367e-02 -1.27729356e+00 2.25488201e-01 5.32123387e-01
-4.45765369e-02 7.32010230e-02 -3.55151176e-01 7.72828221e-01
-3.14756244e-01 -1.88285604e-01 5.68051755e-01 1.00860879e-01
-3.48253816e-01 6.04026914e-01 -5.05184352e-01 -2.73867864e-02
1.34260499e+00 1.86776385e-01 -1.93354532e-01 -5.80829322e-01
-8.03696156e-01 8.01479042e-01 2.81583428e-01 -1.50730535e-01
5.38954318e-01 -1.47306502e+00 -1.02515161e+00 -1.63256541e-01
-5.94772041e-01 1.12062968e-01 3.14457685e-01 1.44602001e+00
-4.86450136e-01 7.87272230e-02 7.60124698e-02 -4.18150127e-01
-9.46525455e-01 4.15385991e-01 4.34128612e-01 -3.51531804e-01
-2.83352077e-01 5.89825690e-01 1.78955764e-01 -4.48197067e-01
8.40908661e-02 1.78639919e-01 2.36948043e-01 1.80769444e-01
3.97298872e-01 8.04620206e-01 -8.08332413e-02 -4.56299156e-01
-5.10786116e-01 6.79518640e-01 5.00604473e-02 -9.91680622e-02
1.68374741e+00 -6.82427764e-01 -7.38735139e-01 -1.39901400e-01
1.54329896e+00 1.10739894e-01 -1.15995002e+00 -3.94662708e-01
-3.97280723e-01 -2.69394308e-01 5.34338832e-01 -3.29096258e-01
-1.78953278e+00 3.08798015e-01 2.25727707e-01 1.99270651e-01
1.07227826e+00 -4.37336624e-01 8.86192203e-01 5.66314399e-01
5.64246058e-01 -1.28064990e+00 -2.21988216e-01 5.53901851e-01
7.14709759e-01 -1.11409938e+00 3.45973432e-01 -3.67594749e-01
-9.07315388e-02 1.24547064e+00 3.85808468e-01 -4.36107785e-01
9.97473419e-01 2.01688394e-01 -3.31980065e-02 -1.60346195e-01
-7.26991832e-01 8.51960182e-02 4.43414375e-02 2.63528347e-01
1.46208048e-01 4.32744287e-02 -9.06869233e-01 2.21119970e-01
3.19069117e-01 -7.00618979e-03 5.32507539e-01 1.09394395e+00
-2.19817713e-01 -1.18396556e+00 -5.30925810e-01 3.02081645e-01
-2.77974099e-01 2.03384668e-01 -1.76893204e-01 2.98598230e-01
-2.47227430e-01 1.05302536e+00 -5.24640158e-02 2.20104363e-02
-7.82556161e-02 -3.62718761e-01 2.53510565e-01 -3.42117250e-01
-4.06525135e-01 8.01692829e-02 -2.27688417e-01 -4.87874329e-01
-6.55517519e-01 -4.82774347e-01 -1.26082671e+00 -7.93590426e-01
-4.10107225e-01 5.67718923e-01 6.06424868e-01 8.62910151e-01
4.91819322e-01 1.94915265e-01 1.07206917e+00 -9.09085810e-01
-8.96869242e-01 -3.75714034e-01 -8.06606174e-01 2.00211406e-01
4.26523358e-01 -3.29719543e-01 -4.20331627e-01 -1.34974003e-01]
|
[6.253385066986084, 4.9205756187438965]
|
9cbe0494-d4c7-494e-bbee-4da32c34cd93
|
geometry-preserving-lie-group-integrators-for
|
2210.08842
| null |
https://arxiv.org/abs/2210.08842v4
|
https://arxiv.org/pdf/2210.08842v4.pdf
|
Geometry-preserving Lie Group Integrators For Differential Equations On The Manifold Of Symmetric Positive Definite Matrices
|
In many applications, one encounters signals that lie on manifolds rather than a Euclidean space. In particular, covariance matrices are examples of ubiquitous mathematical objects that have a non Euclidean structure. The application of Euclidean methods to integrate differential equations lying on such objects does not respect the geometry of the manifold, which can cause many numerical issues. In this paper, we propose to use Lie group methods to define geometry-preserving numerical integration schemes on the manifold of symmetric positive definite matrices. These can be applied to a number of differential equations on covariance matrices of practical interest. We show that they are more stable and robust than other classical or naive integration schemes on an example.
|
['Franck Vermet', 'Naoufal El Bekri', 'Alexandre Reiffers-Masson', 'Lucas Drumetz']
|
2022-10-17
| null | null | null | null |
['numerical-integration']
|
['miscellaneous']
|
[-1.72571987e-01 -6.11095652e-02 3.97547096e-01 -3.56609188e-02
-1.46052867e-01 -6.48838699e-01 3.56251955e-01 -6.37413442e-01
-3.64116579e-01 8.34263861e-01 -3.04234505e-01 -2.58276433e-01
-5.77365577e-01 -5.91237545e-01 -1.15435459e-01 -8.56344879e-01
-2.21259087e-01 2.89928522e-02 5.87238371e-02 -4.07503754e-01
2.42963478e-01 8.99780869e-01 -1.06999099e+00 -5.30134261e-01
1.04775131e+00 7.95483649e-01 -5.47339022e-01 7.07193494e-01
1.82679087e-01 5.00935316e-01 -2.87904471e-01 -3.96391630e-01
5.31776607e-01 -5.32716334e-01 -4.26751167e-01 2.70175040e-01
3.07019949e-01 1.21287219e-01 -4.74948406e-01 1.69010639e+00
2.25169852e-01 2.99244821e-01 1.34311056e+00 -1.30071902e+00
-5.83104968e-01 4.09534335e-01 -5.75702250e-01 -4.75286953e-02
1.32139370e-01 -3.48661065e-01 6.69355810e-01 -1.24729860e+00
6.13222301e-01 1.33702028e+00 1.01942801e+00 3.27149332e-01
-1.45860672e+00 -2.44874939e-01 -4.64825690e-01 -6.33254498e-02
-1.98592138e+00 -2.74687797e-01 9.69813585e-01 -7.80466974e-01
1.42710328e-01 5.84869266e-01 4.75746334e-01 2.38108680e-01
5.51315665e-01 1.85533255e-01 8.94250274e-01 -3.29612494e-01
4.48794574e-01 9.68367010e-02 2.98750550e-01 6.62371814e-01
4.18456793e-01 -1.70405358e-02 2.05298200e-01 -2.89922565e-01
9.66733634e-01 -1.29297853e-01 -2.49211580e-01 -9.26129937e-01
-1.53583539e+00 1.24087453e+00 1.83569238e-01 7.50814795e-01
-2.54383713e-01 -5.29409796e-02 8.51913616e-02 2.58808762e-01
2.98188239e-01 4.66131270e-01 1.14093520e-01 3.86891551e-02
-5.94788611e-01 1.74017921e-01 1.04882371e+00 9.38449800e-01
7.03615487e-01 2.46265262e-01 3.69022220e-01 4.17851925e-01
5.22786975e-01 8.30241561e-01 -2.05079541e-02 -1.33214092e+00
1.15412287e-01 4.37524289e-01 9.17513762e-03 -1.52499998e+00
-7.10420012e-01 -5.95021062e-02 -1.34655476e+00 5.28883636e-01
7.45408773e-01 -4.04092759e-01 -4.02604900e-02 1.45256531e+00
4.89323497e-01 -2.24708603e-03 2.70850569e-01 9.64965403e-01
2.44828254e-01 4.88082528e-01 -6.89084589e-01 -2.67430156e-01
1.01386893e+00 -2.28114948e-01 -9.87601817e-01 5.36390960e-01
8.68168950e-01 -9.64101851e-01 3.16878736e-01 3.60240221e-01
-8.77804697e-01 -1.85221463e-01 -1.39696765e+00 2.00756401e-01
-3.16084534e-01 3.50134760e-01 4.80020583e-01 9.55906749e-01
-1.15571368e+00 1.06716919e+00 -5.87613404e-01 -2.60698438e-01
-1.33648112e-01 3.31943899e-01 -3.45637023e-01 5.83411455e-01
-1.11837673e+00 9.72093642e-01 4.31145951e-02 4.69080210e-01
2.73190558e-01 -7.09167182e-01 -8.64929676e-01 -3.81121814e-01
-3.25845409e-04 -4.00440395e-01 7.78979897e-01 -6.82524741e-01
-1.60321283e+00 7.34521508e-01 2.15097040e-01 -1.13726772e-01
7.69717157e-01 7.80380145e-02 -5.41473150e-01 3.40344645e-02
-1.98730290e-01 2.60637552e-02 1.18490577e+00 -5.99836111e-01
-1.05328122e-02 -4.38250899e-01 -8.73847157e-02 -3.61233801e-02
-2.12351084e-02 7.16366917e-02 3.35545629e-01 -5.53448200e-01
6.14485145e-01 -1.13782763e+00 -5.18066168e-01 3.28891695e-01
-5.21098852e-01 -1.58282500e-02 8.85973513e-01 -4.98917311e-01
1.01704788e+00 -2.37226081e+00 6.89861715e-01 6.56706333e-01
1.68019250e-01 1.06998652e-01 2.28188306e-01 2.71821588e-01
-2.03785241e-01 9.10102855e-03 -4.28508192e-01 2.22534806e-01
9.25643295e-02 -1.42552853e-01 -2.59454936e-01 1.29901612e+00
2.32117295e-01 5.54531276e-01 -6.56485736e-01 -6.38248026e-01
1.31799385e-01 7.27642894e-01 -5.85259974e-01 -4.26219881e-01
3.01335990e-01 5.93306899e-01 -6.20860577e-01 1.14024840e-01
1.02773631e+00 8.12364817e-02 -5.74431866e-02 -5.42148411e-01
-3.76625746e-01 -2.82902330e-01 -2.11265159e+00 1.44853055e+00
7.07387226e-03 6.66229665e-01 6.55228436e-01 -1.16150415e+00
9.38035071e-01 1.20176934e-01 6.48810208e-01 7.41926804e-02
5.05580187e-01 3.40379506e-01 5.23279607e-01 -1.72164977e-01
4.34253573e-01 -2.57927030e-01 6.98271692e-02 4.51890439e-01
-1.31186038e-01 -5.86648345e-01 2.90515006e-01 2.59226620e-01
7.81394184e-01 -3.10091734e-01 3.94146621e-01 -1.01202834e+00
1.17987359e+00 -3.83102417e-01 4.80799377e-01 1.35389909e-01
-1.62899941e-01 6.82201564e-01 4.24313664e-01 8.21420550e-02
-1.03591490e+00 -1.15314996e+00 -7.56993532e-01 1.13056928e-01
1.48067936e-01 -1.32632017e-01 -9.88718510e-01 -1.43746033e-01
6.89311400e-02 2.89484203e-01 -4.55651462e-01 -2.29944602e-01
-4.80366498e-01 -7.43428469e-01 4.95653450e-01 9.52273235e-02
4.76583898e-01 -2.70268559e-01 -1.63356230e-01 2.17467353e-01
2.43875727e-01 -9.77264166e-01 -9.62005019e-01 -2.77375728e-01
-1.19546020e+00 -1.12159932e+00 -8.93665314e-01 -4.91905034e-01
8.24208379e-01 2.39497721e-01 5.95533550e-01 -1.04143828e-01
-3.18425536e-01 8.09217513e-01 1.36523172e-01 -2.06370294e-01
-5.48440099e-01 -2.28379309e-01 6.19954348e-01 7.01275945e-01
2.87114501e-01 -6.51616096e-01 -2.74158031e-01 7.39198387e-01
-8.35568249e-01 -4.71672595e-01 6.25146925e-02 6.48430288e-01
3.74897152e-01 1.27972722e-01 4.99661565e-01 -4.43107396e-01
7.29327142e-01 -3.23065013e-01 -9.64829206e-01 6.54750094e-02
-2.47914270e-01 4.00223762e-01 3.33339125e-01 -5.69751918e-01
-7.12799788e-01 3.03526193e-01 3.11867803e-01 -3.56020033e-01
4.14352149e-01 5.20797491e-01 -2.64629394e-01 -7.51175165e-01
8.58796358e-01 -1.11870587e-01 4.80629593e-01 -3.98402005e-01
4.92768317e-01 5.32143176e-01 3.79909784e-01 -4.65640396e-01
1.14880610e+00 5.58684170e-01 8.16617906e-01 -1.54347122e+00
-3.68201345e-01 -3.54923069e-01 -1.24456608e+00 -1.79667398e-01
8.45344365e-01 -4.90384519e-01 -4.90868241e-01 5.10833025e-01
-1.23563027e+00 2.36590698e-01 -4.21461135e-01 9.73655999e-01
-8.97388577e-01 5.80632865e-01 -6.35589004e-01 -1.00209117e+00
1.02936462e-01 -8.49947631e-01 6.53004885e-01 6.99245781e-02
-3.58909607e-01 -1.52386594e+00 3.23467225e-01 -3.97700191e-01
4.65165913e-01 1.09281585e-01 4.74924088e-01 -4.26491588e-01
-5.11852264e-01 -2.56783158e-01 5.00232214e-03 5.44540107e-01
3.24645013e-01 5.33117652e-01 -6.19651198e-01 -5.66109084e-02
7.84067690e-01 5.62937796e-01 1.04847305e-01 3.08470517e-01
3.83413970e-01 -3.09587433e-03 -3.05567324e-01 5.69680393e-01
1.10359108e+00 3.81195128e-01 6.63849354e-01 -2.28075713e-01
6.68521106e-01 6.90189064e-01 5.32107592e-01 2.24930361e-01
-2.31352389e-01 7.67940581e-01 -7.53169358e-02 3.76659296e-02
4.53382581e-01 3.86649907e-01 3.63276184e-01 1.33917522e+00
-1.20458372e-01 7.57678747e-01 -5.56269825e-01 2.26711884e-01
-1.81281781e+00 -1.00016749e+00 -6.17646754e-01 2.31112933e+00
7.88332880e-01 -3.50110799e-01 1.33878365e-02 4.57320869e-01
9.95337188e-01 -2.50304967e-01 -4.77705806e-01 -3.23412836e-01
-5.23616672e-01 -1.37513191e-01 5.30890286e-01 7.17490971e-01
-1.21985137e+00 2.64834642e-01 7.01340628e+00 6.09485745e-01
-1.09793675e+00 7.00074211e-02 -1.00824453e-01 4.64996755e-01
1.58227030e-02 -1.11637220e-01 -7.28232980e-01 2.93358147e-01
7.44810939e-01 -7.78947473e-01 1.80640757e-01 7.34982967e-01
4.88066673e-02 -2.12609358e-02 -1.29110229e+00 1.37554955e+00
-2.01194026e-02 -9.19587135e-01 -3.36226374e-01 4.08149183e-01
9.44233477e-01 -5.50581694e-01 2.06004992e-01 -2.65729219e-01
-1.15833662e-01 -8.94132793e-01 3.11620325e-01 8.78823757e-01
3.18754762e-01 -8.13831031e-01 4.47712094e-01 2.18116507e-01
-1.12153685e+00 5.26716888e-01 -7.25718856e-01 -8.15844238e-02
3.46248478e-01 9.92588878e-01 -2.69028991e-01 5.37714005e-01
2.08901182e-01 1.00361860e+00 -3.26397330e-01 1.36781013e+00
2.57733017e-01 1.30568789e-02 -5.88350654e-01 -1.35589600e-01
7.19461888e-02 -1.50970340e+00 9.14517939e-01 8.08920264e-01
8.28689337e-01 4.30083513e-01 -3.45231861e-01 1.19226551e+00
4.06108320e-01 6.14901960e-01 -1.26446533e+00 -3.39009392e-04
-2.58286446e-02 1.40939987e+00 -7.65603423e-01 -1.84642330e-01
-6.96322322e-01 7.21096694e-01 -3.68406236e-01 3.78871202e-01
-6.44364059e-01 -7.89985001e-01 1.21377313e+00 -8.35592207e-03
2.04116523e-01 -7.59345710e-01 -7.83825219e-02 -1.47662008e+00
-6.37033433e-02 -6.03247464e-01 7.32942224e-02 -4.73461181e-01
-1.36986196e+00 3.32908124e-01 -8.27947631e-02 -1.69892883e+00
-3.07154655e-01 -1.02650726e+00 -6.34824693e-01 8.35097075e-01
-3.99381012e-01 -2.14409277e-01 2.82186836e-01 8.42219710e-01
-3.25791150e-01 -2.40823284e-01 5.39051056e-01 4.07893151e-01
-2.83214957e-01 1.38941705e-01 6.34315848e-01 2.64773935e-01
4.05984551e-01 -1.35242629e+00 1.27802333e-02 8.70154560e-01
2.41179049e-01 8.32799077e-01 7.77216077e-01 -3.41601551e-01
-1.84931612e+00 -8.69051397e-01 4.51130629e-01 -5.20520568e-01
1.15588069e+00 -2.58341819e-01 -1.04123855e+00 8.02180648e-01
-1.27873998e-02 -1.51816215e-02 5.51614761e-01 -2.59240955e-01
-8.23739544e-02 2.41172072e-02 -1.34037697e+00 8.53389025e-01
9.79571879e-01 -5.24371207e-01 -5.92653573e-01 3.42157185e-01
1.63729101e-01 -2.10327208e-01 -1.22276199e+00 3.20061892e-01
3.01478207e-01 -8.19938481e-01 1.03825128e+00 -4.63263601e-01
-5.66790104e-01 -7.76250601e-01 -3.89056236e-01 -1.40183258e+00
-4.51201648e-02 -1.17615151e+00 2.34463111e-01 1.09859860e+00
1.92357063e-01 -1.11668801e+00 3.99435610e-01 7.18573272e-01
3.22002113e-01 -1.00315444e-01 -1.23076451e+00 -1.29720819e+00
6.46512926e-01 -1.85705513e-01 3.06195378e-01 1.01440310e+00
5.49334347e-01 3.11547577e-01 -1.43711582e-01 1.52807549e-01
1.09937942e+00 -6.08585700e-02 6.09446406e-01 -1.73068404e+00
1.87096372e-01 -5.16970217e-01 -1.18572628e+00 -9.26598012e-01
4.34531540e-01 -9.96940315e-01 -3.39380391e-02 -7.91232288e-01
-3.77000302e-01 -5.27835965e-01 2.78948009e-01 -6.53904676e-01
3.55189264e-01 1.69005394e-01 5.25417514e-02 -1.28335422e-02
-2.91909784e-01 6.89276636e-01 1.26339805e+00 -1.28638610e-01
-2.10030779e-01 2.24393070e-01 -2.25531042e-01 1.04598248e+00
6.64217055e-01 -1.50161713e-01 -3.61269265e-01 2.61843324e-01
2.99776763e-01 -1.52247369e-01 2.16963105e-02 -1.19089532e+00
4.32986766e-01 3.75602543e-02 -5.40170856e-02 -4.89096850e-01
3.47146809e-01 -9.66849446e-01 7.03061879e-01 3.77726883e-01
1.81204647e-01 1.34042017e-02 -1.48768976e-01 2.58497238e-01
-3.81277084e-01 -6.92866027e-01 1.11970341e+00 2.96931565e-01
-1.98739931e-01 1.62622258e-01 -6.56365097e-01 1.81162059e-01
1.18635499e+00 -8.29274580e-02 5.97065091e-02 -5.38915694e-01
-8.64710271e-01 -1.75618619e-01 6.06492341e-01 5.13241440e-02
4.30561930e-01 -1.89467406e+00 -4.94215995e-01 3.31382930e-01
-3.46346259e-01 -3.39769274e-01 6.91231564e-02 1.46188343e+00
-7.13551521e-01 5.62987924e-01 -1.42647251e-01 -8.31253946e-01
-1.14259756e+00 8.37688506e-01 8.67561519e-01 2.95628250e-01
-5.03032744e-01 2.75887251e-01 3.25384259e-01 -7.54621565e-01
-2.37718038e-02 -5.92580914e-01 -7.11120591e-02 2.97005415e-01
6.83262527e-01 8.39873433e-01 -1.82693809e-01 -1.22569454e+00
-3.59591812e-01 1.37543213e+00 5.37108064e-01 -4.19796616e-01
7.40869403e-01 -3.31858933e-01 -5.00501573e-01 7.74961889e-01
1.57656515e+00 2.46775702e-01 -9.29958642e-01 -3.66721839e-01
2.00709179e-01 -2.82931387e-01 -2.93297350e-01 3.34348649e-01
-1.34973812e+00 8.94000113e-01 4.42161977e-01 5.61479628e-01
6.22225702e-01 -2.17046753e-01 5.85686937e-02 5.67519069e-01
5.12448967e-01 -1.08044386e+00 -2.98143595e-01 6.43119991e-01
1.24823678e+00 -9.59234774e-01 1.84429009e-02 -9.96771872e-01
-1.76889002e-01 1.31103241e+00 6.87029352e-03 -5.41157961e-01
1.48567486e+00 2.33246252e-01 1.46299556e-01 7.19249994e-02
-1.95792139e-01 -3.91605906e-02 5.11538029e-01 7.41672218e-01
4.93164420e-01 1.34723350e-01 -8.17286491e-01 2.63745010e-01
-3.06452304e-01 -1.97649315e-01 9.65772331e-01 7.01626062e-01
-2.40266994e-01 -9.69975352e-01 -8.61734867e-01 9.48416740e-02
-3.33010554e-01 2.84605056e-01 -4.72075671e-01 7.33055890e-01
-3.91671509e-01 8.81865740e-01 1.28061369e-01 -1.35044739e-01
4.67017204e-01 6.59162998e-02 6.53808951e-01 -7.25079551e-02
1.10260129e-01 -4.88507859e-02 -1.48269266e-01 -5.54218173e-01
-5.93222260e-01 -1.37833321e+00 -1.24594879e+00 -4.56409961e-01
-2.62344003e-01 4.66569424e-01 6.93737984e-01 6.31438613e-01
2.32401431e-01 1.16861816e-02 7.79325724e-01 -8.41649950e-01
-1.00676167e+00 -8.98643851e-01 -1.63243067e+00 1.21023655e-01
3.38356704e-01 -9.81324613e-01 -9.14751053e-01 -9.16505083e-02]
|
[7.44998025894165, 4.174729347229004]
|
02bf97bc-7ba3-41a7-a4e1-16045556207e
|
audio-classification-using-ml-methods
|
2305.19304
| null |
https://arxiv.org/abs/2305.19304v1
|
https://arxiv.org/pdf/2305.19304v1.pdf
|
Audio classification using ML methods
|
Machine Learning systems have achieved outstanding performance in different domains. In this paper machine learning methods have been applied to classification task to classify music genre. The code shows how to extract features from audio files and classify them using supervised learning into 2 genres namely classical and metal. Algorithms used are LogisticRegression, SVC using different kernals (linear, sigmoid, rbf and poly), KNeighborsClassifier , RandomForestClassifier, DecisionTreeClassifier and GaussianNB.
|
['Krishna Kumar']
|
2023-05-30
| null | null | null | null |
['audio-classification']
|
['audio']
|
[ 1.73706245e-02 -2.92974323e-01 -2.26746142e-01 -4.22060817e-01
-5.82189083e-01 -9.30897236e-01 4.17357862e-01 2.38243341e-01
-4.18638825e-01 1.13227069e+00 2.47275069e-01 -3.03123027e-01
-5.84048212e-01 -5.33088565e-01 5.43236136e-02 -5.99002004e-01
-1.72630861e-01 5.21603048e-01 2.93208390e-01 -7.44744837e-02
7.58167148e-01 4.27890360e-01 -1.75307596e+00 9.45604444e-01
5.20124197e-01 1.13883150e+00 -2.71657854e-01 1.17108119e+00
-1.27724772e-02 8.33105206e-01 -9.17848229e-01 1.42794639e-01
6.62001744e-02 -2.80857921e-01 -8.44992816e-01 -3.37386012e-01
-3.85955572e-01 4.76626158e-01 -4.09422517e-02 5.26502311e-01
6.14661932e-01 4.20697063e-01 1.50215042e+00 -1.33662343e+00
-1.96324050e-01 7.58923233e-01 -4.08236712e-01 5.93259990e-01
7.78864920e-01 -4.04803872e-01 4.99282837e-01 -6.70961499e-01
2.19050974e-01 1.16143656e+00 9.79581535e-01 -2.12088786e-02
-1.13359034e+00 -1.12415469e+00 -5.97133160e-01 5.51089346e-01
-1.49805474e+00 -2.38800738e-02 6.18210256e-01 -8.20864916e-01
1.19678354e+00 5.81466615e-01 5.79306483e-01 6.62602723e-01
3.95171940e-01 4.01943952e-01 1.74026072e+00 -7.53514171e-01
2.39543676e-01 6.61025703e-01 5.32407522e-01 -1.17601328e-01
-1.95348144e-01 7.25915059e-02 -5.61095953e-01 -5.71781456e-01
4.05412972e-01 -2.78323174e-01 -6.95994217e-03 6.10520124e-01
-7.54404306e-01 8.22393537e-01 -1.66883841e-01 6.36359870e-01
-4.08194989e-01 -2.56277341e-02 6.89472914e-01 8.29450071e-01
3.65544647e-01 5.79409957e-01 -8.44834328e-01 -5.31638622e-01
-1.06924772e+00 3.95251960e-01 9.01890159e-01 7.54151881e-01
2.20667005e-01 3.36129993e-01 2.30585858e-02 1.16478539e+00
1.31012604e-01 3.86275202e-01 1.08541226e+00 -4.96566683e-01
2.49693133e-02 3.84871930e-01 -2.05699325e-01 -1.01245952e+00
-5.84049642e-01 -2.36399561e-01 -6.70038104e-01 3.75509709e-01
2.45947033e-01 -2.13469386e-01 -6.50114417e-01 5.96368730e-01
5.05779646e-02 2.60078043e-01 9.82723832e-02 3.57956082e-01
1.11576200e+00 8.47267866e-01 -9.72961485e-02 -2.88541526e-01
1.05645013e+00 -5.65342903e-01 -6.28120482e-01 7.78773785e-01
3.30001190e-02 -1.41672862e+00 8.51184368e-01 1.63390744e+00
-6.68736637e-01 -9.85278189e-01 -8.88629198e-01 4.57093179e-01
-6.16274178e-01 3.04343134e-01 5.31938493e-01 7.51038194e-01
-6.63299799e-01 9.86498415e-01 -5.88132322e-01 7.34884962e-02
3.39060336e-01 7.19818294e-01 -2.17676118e-01 7.29806662e-01
-7.89654255e-01 4.02770847e-01 9.79521215e-01 -4.20166165e-01
-4.46973711e-01 -1.37500972e-01 -3.23722452e-01 -1.70094073e-01
-2.41235137e-01 -1.24203628e-02 1.22431016e+00 -1.10333550e+00
-1.77363229e+00 7.12815285e-01 1.99387759e-01 -5.02618670e-01
-7.88487494e-03 -1.00404501e-01 -7.23165810e-01 -5.11304401e-02
-5.65705538e-01 -8.13756362e-02 1.12235796e+00 -8.14452052e-01
-8.90033901e-01 -5.17547727e-01 -3.90262216e-01 4.03212160e-02
-1.68105111e-01 5.01346529e-01 7.02769399e-01 -1.22198713e+00
5.22412777e-01 -6.61245167e-01 2.51487255e-01 -9.59433436e-01
-6.35511458e-01 -7.88194954e-01 8.59395742e-01 -7.38776445e-01
1.53567088e+00 -2.17794681e+00 -2.36619241e-03 4.58982974e-01
-4.25331235e-01 5.35042882e-02 5.66952229e-01 5.08420646e-01
-3.28514069e-01 -2.48854086e-02 2.80716866e-01 5.51626086e-01
-1.79282486e-01 1.93915097e-03 -4.13553298e-01 1.87715665e-01
-5.55653036e-01 6.07585125e-02 -3.79492790e-01 -5.88518679e-01
6.53574169e-02 3.83696169e-01 -3.02522779e-01 9.27690268e-02
1.56199276e-01 -1.96029320e-02 -1.98391885e-01 7.07797348e-01
2.36493751e-01 4.94590282e-01 -6.21831492e-02 -3.29732858e-02
-2.16937348e-01 4.69430715e-01 -1.49536502e+00 9.60322201e-01
-3.98400575e-01 8.57755423e-01 -2.22107738e-01 -1.36766839e+00
1.46351457e+00 6.82619035e-01 1.85825631e-01 4.50809717e-01
3.33739042e-01 4.74468112e-01 4.77949642e-02 -7.43807077e-01
9.93931592e-02 -2.87770391e-01 -6.34025335e-02 1.63874224e-01
4.98722047e-01 -2.64044166e-01 -3.36679555e-02 -5.12307703e-01
7.65549004e-01 2.88708091e-01 5.42810380e-01 -3.15577894e-01
6.02435768e-01 -9.75823775e-03 2.38388970e-01 6.81555569e-01
1.14972048e-01 4.58495617e-01 3.40662479e-01 -4.58426386e-01
-6.39609635e-01 -1.00451803e+00 -3.64607215e-01 1.46040285e+00
-5.40400088e-01 -3.98363292e-01 -4.66003895e-01 -5.44407189e-01
1.61655694e-01 6.10780835e-01 -2.33797342e-01 3.86058679e-03
-3.63504648e-01 -5.98809302e-01 8.40719461e-01 4.27443355e-01
1.23160012e-01 -1.35618520e+00 -4.61487293e-01 2.78341860e-01
3.32463890e-01 -3.90449226e-01 1.25804201e-01 5.80967546e-01
-1.06584501e+00 -8.94088209e-01 -4.16623563e-01 -9.48408544e-01
-1.57944262e-01 -3.33349735e-01 8.94942284e-01 -5.42056799e-01
-5.55274725e-01 5.66976294e-02 -7.21182466e-01 -9.22153533e-01
-3.44592482e-01 1.33046806e-01 3.36060792e-01 -2.84029603e-01
4.48469073e-01 -1.17340338e+00 -4.09198582e-01 1.34343341e-01
-3.41794014e-01 -7.62045979e-01 2.68600941e-01 7.72106290e-01
4.71736014e-01 4.60834861e-01 6.43061876e-01 -7.45684326e-01
9.72298801e-01 -4.53516960e-01 -2.66216040e-01 -1.07712127e-01
-7.35665977e-01 -3.82357061e-01 7.96887159e-01 -8.50385547e-01
-5.39419472e-01 1.53263003e-01 -2.59411365e-01 -4.42536473e-02
-6.21864021e-01 4.01574016e-01 2.02375695e-01 3.16084847e-02
1.04432154e+00 4.10217762e-01 -3.68062139e-01 -9.24030602e-01
-1.89558625e-01 1.69068551e+00 4.91361916e-01 -3.32311988e-01
4.83118623e-01 5.86604364e-02 -2.40919396e-01 -1.17274690e+00
-4.10077721e-01 -7.46074200e-01 -6.02699161e-01 -4.95905012e-01
7.35187113e-01 -5.42932391e-01 -7.82813668e-01 4.63351578e-01
-8.54963422e-01 1.26488999e-01 -4.07085896e-01 1.03192019e+00
-7.94829547e-01 -1.94578648e-01 -6.49268925e-01 -1.41083753e+00
-7.14369774e-01 -3.98192137e-01 5.35281599e-01 4.77793157e-01
-6.82011068e-01 -6.96392953e-01 2.07435414e-01 2.64485311e-02
-1.05205737e-02 3.67540449e-01 9.80628431e-01 -1.27203548e+00
5.02589583e-01 -4.22451437e-01 3.66671622e-01 6.67552650e-01
1.70042485e-01 2.45041862e-01 -1.23549724e+00 2.07036108e-01
-1.21852756e-01 -4.05454040e-01 8.06194425e-01 5.13889730e-01
1.53323531e+00 -3.38843375e-01 -1.97306395e-01 7.30152011e-01
1.22671700e+00 8.34035218e-01 5.44487059e-01 5.83134830e-01
-1.07399255e-01 1.89665481e-01 6.53484404e-01 5.40188313e-01
-1.04906023e-01 5.21398008e-01 -2.55491436e-01 4.29591715e-01
1.78757370e-01 -1.15591489e-01 2.77833879e-01 8.15656245e-01
-6.51728332e-01 3.94220985e-02 -9.08674777e-01 -1.10705934e-01
-1.61707687e+00 -1.17833948e+00 -4.04783309e-01 2.13691139e+00
9.17162061e-01 5.57151020e-01 7.83923388e-01 1.41269481e+00
5.27306557e-01 -4.27339733e-01 8.41782894e-03 -1.06697834e+00
-1.52418435e-01 5.80261469e-01 1.83339462e-01 1.64459068e-02
-1.38373756e+00 7.46722519e-01 7.36650181e+00 1.15183091e+00
-1.05532575e+00 2.76625901e-02 3.39858979e-01 1.19852401e-01
6.13032997e-01 -4.86852415e-02 -8.05368602e-01 4.96803850e-01
1.15043652e+00 -1.03133626e-01 5.62406003e-01 1.01708233e+00
1.00980230e-01 -1.77715227e-01 -4.40548360e-01 1.25349271e+00
6.25626221e-02 -7.48944938e-01 -2.00740799e-01 -3.19497079e-01
3.82365912e-01 -2.38058046e-01 -6.92195669e-02 6.34134650e-01
1.62851870e-01 -1.07415342e+00 3.08408052e-01 7.38860786e-01
5.16727924e-01 -1.01748204e+00 6.95242226e-01 2.80387819e-01
-9.55308259e-01 -4.78300840e-01 -3.43072683e-01 -4.57061619e-01
-2.66950130e-01 4.03574795e-01 -1.13588834e+00 4.69289750e-01
1.04307985e+00 6.42477334e-01 -3.43352228e-01 1.45176339e+00
1.61204398e-01 1.29869413e+00 -5.41095018e-01 -2.82159567e-01
-1.29996985e-01 -1.40375748e-01 5.47918916e-01 1.44331610e+00
4.96123582e-01 5.32089397e-02 2.31273383e-01 4.11966965e-02
6.69849098e-01 7.14363039e-01 -5.75810522e-02 -1.28272595e-02
4.16259080e-01 1.05670106e+00 -1.02291596e+00 -3.95075440e-01
2.08920494e-01 5.33401847e-01 -4.59658533e-01 1.71848819e-01
-3.58938575e-01 -1.12986076e+00 1.92991167e-01 5.46035767e-01
2.11209312e-01 -6.18505478e-02 -5.95696867e-01 -7.43097484e-01
-6.81539297e-01 -9.41146553e-01 9.11512077e-01 -7.56967247e-01
-1.11435199e+00 8.05870116e-01 2.67350674e-01 -1.20226204e+00
-4.56543773e-01 -9.61670756e-01 -6.53584480e-01 5.64493895e-01
-5.33454418e-01 -8.35642338e-01 3.16994041e-02 7.76173234e-01
7.08390236e-01 -1.09816933e+00 1.07198429e+00 1.52353749e-01
7.93498084e-02 3.85371357e-01 5.52728474e-01 7.51662180e-02
7.76568770e-01 -1.60179341e+00 -6.09355271e-01 -4.77063298e-01
4.58659559e-01 2.28750363e-01 9.28692162e-01 -3.16211224e-01
-6.90480232e-01 -5.25668263e-01 7.79154301e-01 1.00717716e-01
4.86643523e-01 -2.07721949e-01 -6.60233498e-01 2.44785696e-01
1.13044225e-01 -3.13304752e-01 1.14144576e+00 5.10155261e-01
-3.83537292e-01 -4.73821849e-01 -1.09790051e+00 -2.53411084e-01
2.60020673e-01 -3.25496316e-01 -8.77614856e-01 6.01939857e-01
-5.55154346e-02 -3.89409065e-02 -1.32171321e+00 3.83792996e-01
1.15532625e+00 -9.47458267e-01 8.30237508e-01 -9.33614433e-01
2.18728825e-01 -1.85587704e-01 -1.95764929e-01 -1.22504413e+00
-1.39276600e-02 -7.44236469e-01 -1.98264927e-01 1.17017257e+00
6.65746987e-01 -6.03585362e-01 6.55356646e-01 -3.63385975e-01
8.77506286e-02 -8.48761380e-01 -9.13600743e-01 -9.03234482e-01
1.59971699e-01 -5.97378433e-01 2.94973522e-01 9.46783841e-01
5.16355872e-01 4.28107709e-01 -1.43043324e-01 -4.84114259e-01
1.56350389e-01 2.96442956e-01 8.16377103e-01 -1.81770253e+00
-7.67533123e-01 -4.18932199e-01 -1.25292814e+00 -1.54543564e-01
-7.21953437e-02 -7.89790809e-01 -2.56689668e-01 -1.07495475e+00
-1.68028846e-01 -4.38728869e-01 -6.19889617e-01 4.17457819e-01
3.42881799e-01 6.27688229e-01 -8.09879899e-02 2.45925292e-01
-6.34329990e-02 -1.93460256e-01 6.21833742e-01 8.41240212e-03
-7.59332061e-01 8.50050330e-01 -4.62861396e-02 1.08324778e+00
1.15286458e+00 -7.42445052e-01 -2.87279427e-01 4.61068898e-01
1.78126186e-01 -1.46629706e-01 -5.47875203e-02 -1.06731927e+00
-1.48384608e-02 -6.21541180e-02 5.53432286e-01 -6.85246468e-01
4.36885476e-01 -3.99786353e-01 2.50267833e-01 4.21987981e-01
-4.80238825e-01 3.96528281e-02 4.33073230e-02 3.01087111e-01
-3.91950130e-01 -7.25174427e-01 6.92473292e-01 1.26672491e-01
-2.36070320e-01 -4.24482793e-01 -8.16703260e-01 -1.81184262e-01
8.79900277e-01 -5.09981453e-01 3.56670558e-01 -6.59227788e-01
-1.72147250e+00 -4.72060412e-01 -1.70342609e-01 2.80448139e-01
4.97894585e-01 -1.14532781e+00 -8.10009301e-01 8.75715464e-02
-2.15256587e-01 -8.64895463e-01 -1.43730029e-01 7.37953901e-01
-6.17309868e-01 2.44651183e-01 -3.63789946e-01 -5.30966401e-01
-1.99150312e+00 2.70325392e-01 4.22497950e-02 -1.23399563e-01
-3.41989309e-01 9.73778486e-01 -6.91495001e-01 -3.94188702e-01
4.47447658e-01 -2.52470851e-01 -9.72710788e-01 1.57492414e-01
5.41720152e-01 9.27724421e-01 -4.17342782e-02 -5.71573019e-01
-1.08774625e-01 5.96139252e-01 1.01556510e-01 -2.28287280e-01
1.40297472e+00 5.13863623e-01 -1.31905705e-01 1.23133755e+00
1.00784779e+00 -4.35187705e-02 -3.10966402e-01 3.24237645e-01
3.88344973e-01 -3.02505910e-01 5.30483909e-02 -9.77495968e-01
-3.01721245e-01 8.38391781e-01 8.87363315e-01 8.81930828e-01
1.39858854e+00 -2.90747195e-01 1.56607732e-01 3.58658671e-01
2.28311166e-01 -1.58230746e+00 -5.10446250e-01 3.78179133e-01
8.99005830e-01 -7.77292728e-01 3.52488071e-01 -4.21540469e-01
-5.00367701e-01 1.76147103e+00 8.14765766e-02 -7.08165109e-01
1.44741082e+00 5.38303256e-01 -6.03110269e-02 3.18087846e-01
-8.98990154e-01 9.29775909e-02 6.34350657e-01 8.26452672e-01
9.20586646e-01 1.85429692e-01 -8.73306632e-01 1.35507202e+00
-9.99778032e-01 2.23884910e-01 3.20217133e-01 9.00103569e-01
-5.88923812e-01 -8.68499935e-01 -7.52497435e-01 9.43702638e-01
-9.59920585e-01 1.68120280e-01 -5.46319604e-01 7.91097224e-01
4.98743534e-01 1.28127789e+00 -6.75125569e-02 -8.79199684e-01
3.83925170e-01 2.76889890e-01 6.18652046e-01 -4.84943390e-01
-1.10763872e+00 4.56915200e-01 1.84997484e-01 1.15867242e-01
-2.15126619e-01 -7.05738425e-01 -1.14882863e+00 1.06711924e-01
-5.67532241e-01 7.72856414e-01 1.02262521e+00 7.53728569e-01
-1.77220091e-01 3.45184952e-01 7.81910121e-01 -7.65710831e-01
-6.62204802e-01 -1.66220915e+00 -9.47068512e-01 2.64650527e-02
-5.60337268e-02 -6.89470589e-01 -4.83041316e-01 4.69024569e-01]
|
[15.865103721618652, 5.214579105377197]
|
5077c425-8b01-45f2-9dfb-0f0bf08ec9b9
|
karsl-arabic-sign-language-database
| null | null |
https://dl.acm.org/doi/10.1145/3423420#:~:text=Signs%20in%20KArSL%20database%20are,language%20recognition%20using%20this%20database
|
https://dl.acm.org/doi/10.1145/3423420#:~:text=Signs%20in%20KArSL%20database%20are,language%20recognition%20using%20this%20database
|
KArSL: Arabic Sign Language Database
|
Sign language is the major means of communication for the deaf community. It uses body language and gestures such as hand shapes, lib patterns, and facial expressions to convey a message. Sign language is geography-specific, as it differs from one country to another. Arabic Sign language is used in all Arab countries. The availability of a comprehensive benchmarking database for ArSL is one of the challenges of the automatic recognition of Arabic Sign language. This article introduces KArSL database for ArSL, consisting of 502 signs that cover 11 chapters of ArSL dictionary. Signs in KArSL database are performed by three professional signers, and each sign is repeated 50 times by each signer. The database is recorded using state-of-art multi-modal Microsoft Kinect V2. We also propose three approaches for sign language recognition using this database. The proposed systems are Hidden Markov Models, deep learning images’ classification model applied on an image composed of shots of the video of the sign, and attention-based deep learning captioning system. Recognition accuracies of these systems indicate their suitability for such a large number of Arabic signs. The techniques are also tested on a publicly available database. KArSL database will be made freely available for interested researchers.
|
['Mohamed Mohandes', 'Sabri Mahmoud', 'Hamzah Luqman', 'Ala Addin I. Sidig']
|
2021-01-01
| null | null | null |
acm-transactions-on-asian-and-low-resource-1
|
['sign-language-recognition']
|
['computer-vision']
|
[-2.34924749e-01 -5.58295071e-01 -3.72478396e-01 -4.04752105e-01
-6.50760591e-01 -5.01121223e-01 5.85810006e-01 -9.42502558e-01
-7.73628592e-01 3.37257802e-01 7.23212361e-01 6.13413788e-02
5.91562279e-02 -3.40052068e-01 -1.30731046e-01 -1.04598296e+00
7.45507851e-02 4.26256865e-01 8.90851542e-02 -2.32498720e-01
5.12238324e-01 8.95644784e-01 -1.73133707e+00 2.06339404e-01
2.11615682e-01 7.74555326e-01 -4.56168875e-03 8.92882109e-01
-3.51457030e-01 9.78326082e-01 -4.73852456e-01 -1.77028283e-01
1.98490262e-01 -5.46023071e-01 -5.43200135e-01 -8.41859356e-02
6.00187361e-01 -7.28563190e-01 -5.53741217e-01 7.08626807e-01
1.23521090e+00 -7.21242055e-02 8.88513029e-01 -1.14303041e+00
-8.08190882e-01 2.76269406e-01 -1.79949790e-01 -1.26851916e-01
5.73040783e-01 2.93786228e-01 8.06241810e-01 -9.55239952e-01
7.72447169e-01 1.11645973e+00 4.37847644e-01 9.37111497e-01
-6.87699020e-02 -8.11055005e-01 -2.25036010e-01 5.76729476e-01
-1.30067623e+00 -4.68011558e-01 6.83212876e-01 -6.20510280e-01
8.93604279e-01 1.87204644e-01 9.33251500e-01 9.25525069e-01
-2.01891229e-01 1.16003764e+00 1.29459977e+00 -7.55126595e-01
1.16089374e-01 -6.41192913e-01 2.81506121e-01 8.07863414e-01
1.14406988e-01 6.46359697e-02 -8.05490196e-01 5.07998988e-02
8.95446062e-01 -7.47111738e-02 -3.40695977e-01 -1.93456903e-01
-1.03569257e+00 5.74138880e-01 2.13878259e-01 4.32183444e-01
-5.46575367e-01 3.09520692e-01 1.86995655e-01 3.04182321e-01
-4.89115626e-01 -3.85004699e-01 -2.79483855e-01 -7.73293376e-01
-7.06059217e-01 4.14402075e-02 8.31916511e-01 8.05298448e-01
-8.53334591e-02 1.80733278e-01 -1.51079586e-02 8.75189841e-01
1.37960255e+00 1.08969188e+00 8.74985218e-01 -5.54392278e-01
2.89006889e-01 6.45788670e-01 1.23277664e-01 -5.05028009e-01
-3.27123404e-01 5.98291218e-01 -4.82116550e-01 7.91944325e-01
6.95404887e-01 -1.03260793e-01 -1.82169020e+00 1.09048533e+00
1.09839067e-01 -1.78046629e-01 2.36544117e-01 1.22773445e+00
1.48281705e+00 4.21843231e-01 1.58725455e-01 4.10174459e-01
1.32319474e+00 -8.11150908e-01 -8.93987954e-01 1.95142314e-01
4.15287077e-01 -9.29685533e-01 9.04064119e-01 5.45775533e-01
-7.55269706e-01 5.92295565e-02 -6.85801804e-01 -2.59132385e-01
-4.97772008e-01 4.94310677e-01 5.79338849e-01 9.49580789e-01
-8.77823114e-01 -1.75488845e-01 -9.18173552e-01 -8.13011050e-01
4.10414517e-01 4.25723970e-01 -5.48502743e-01 -1.33342236e-01
-6.41598582e-01 1.16491723e+00 -1.15682065e-01 6.73073947e-01
-4.33295250e-01 2.23304898e-01 -9.25957620e-01 -5.23564637e-01
-5.45947373e-01 4.78133224e-02 1.26485336e+00 -9.00970519e-01
-2.07491255e+00 1.34054101e+00 -2.49343947e-01 2.32798368e-01
6.70236528e-01 -3.62854898e-01 -6.72157109e-01 2.58489493e-02
-3.66286844e-01 3.62670064e-01 7.94371545e-01 -1.09741282e+00
-4.46734190e-01 -4.28648591e-01 -4.77958560e-01 2.88167655e-01
3.03966939e-01 7.12586939e-01 -6.00061774e-01 -4.33684140e-01
5.12408614e-01 -8.69671106e-01 2.04850867e-01 2.37205699e-01
-3.85841392e-02 -2.10040629e-01 9.93456125e-01 -1.18361664e+00
8.91161323e-01 -1.96659601e+00 -4.62949276e-02 4.56287831e-01
-2.83346385e-01 6.05954647e-01 -2.71610647e-01 6.57617927e-01
3.19040686e-01 -2.48541981e-01 -2.36688435e-01 1.23462170e-01
2.01057434e-01 6.79657042e-01 -6.20942973e-02 8.52033436e-01
-3.42607766e-01 9.45477009e-01 -6.47290707e-01 -5.68407238e-01
5.02836883e-01 8.03407669e-01 -7.90265948e-02 5.58539070e-02
2.62875706e-01 6.22322738e-01 -1.54458538e-01 1.43212521e+00
5.82477093e-01 3.83902460e-01 -9.32507813e-02 -1.70743644e-01
-2.51081824e-01 -2.10681871e-01 -1.31303477e+00 1.46574938e+00
-2.68860787e-01 8.42764080e-01 -6.80724904e-02 -5.42044759e-01
9.01751041e-01 4.80391532e-01 2.75021464e-01 -5.72751224e-01
4.69152093e-01 7.13216484e-01 2.03766569e-01 -1.24287629e+00
-1.61390051e-01 -1.33183509e-01 2.52500117e-01 6.87705457e-01
1.46678826e-02 -9.66789499e-02 2.15110213e-01 -4.93833482e-01
7.99298048e-01 1.80336311e-01 2.74068296e-01 5.71029484e-01
6.25740230e-01 -1.91943660e-01 2.35269457e-01 5.07009983e-01
-5.46673417e-01 6.23712957e-01 -2.17194527e-01 -5.18144488e-01
-7.40217447e-01 -9.67395723e-01 -5.16025312e-02 1.03777170e+00
-3.49910706e-02 3.67941260e-01 -2.54341215e-01 -2.91110009e-01
-4.62341458e-02 2.38974884e-01 -4.38727647e-01 5.11266112e-01
-7.38977134e-01 -4.51943815e-01 9.65102613e-01 7.26200819e-01
9.93457377e-01 -1.95124328e+00 -1.20373106e+00 -1.11162476e-01
1.24644479e-02 -9.10526752e-01 -3.09211850e-01 -3.28190058e-01
-5.37971735e-01 -1.47322822e+00 -1.59973049e+00 -1.42059648e+00
7.52921820e-01 -3.78201932e-01 4.47333276e-01 1.60800561e-01
-2.50199735e-01 8.10149908e-01 -8.80051017e-01 -5.23315310e-01
-3.11167061e-01 -4.21239287e-01 -5.36337607e-02 2.16998443e-01
9.68585253e-01 -2.12795570e-01 -5.16167462e-01 2.78481215e-01
-6.51420593e-01 -3.34698886e-01 9.52997446e-01 7.63371527e-01
1.92269981e-01 -8.82983565e-01 -2.55230963e-01 1.82325259e-01
5.47668695e-01 1.97901204e-01 -5.84792256e-01 4.50608641e-01
-4.95300768e-03 1.01131581e-01 -4.63392437e-01 -4.34837908e-01
-7.88557291e-01 4.42244798e-01 -3.42859566e-01 1.54111981e-01
-3.89464915e-01 4.53352600e-01 -7.13656470e-02 -6.48550510e-01
2.83058286e-01 6.81471884e-01 2.22918168e-01 -6.38839006e-01
3.42875570e-01 1.40559566e+00 6.99240983e-01 -7.44505078e-02
5.28149366e-01 6.04097247e-01 -2.28970841e-01 -1.17313826e+00
8.90116394e-02 -8.09262097e-01 -8.99568379e-01 -7.63258517e-01
7.63865113e-01 -6.20190680e-01 -1.05019403e+00 1.62069285e+00
-1.13664758e+00 -4.12442178e-01 -2.66094785e-02 9.19602692e-01
-5.15568852e-01 3.71765941e-01 -4.83335704e-01 -1.04920435e+00
-3.62084985e-01 -1.00154364e+00 1.09743989e+00 3.24153095e-01
-1.12957723e-01 -5.45265019e-01 5.33593059e-01 4.16564763e-01
3.14866453e-01 2.35428348e-01 3.84766430e-01 -3.76748055e-01
-3.68220627e-01 -7.35608459e-01 -2.21071854e-01 4.79309618e-01
4.03886169e-01 2.79289126e-01 -9.21935678e-01 5.71328253e-02
-5.95278680e-01 -4.71395224e-01 8.12784314e-01 5.23534119e-01
3.25119525e-01 -2.68047154e-01 3.84013802e-02 3.98854315e-01
1.39490187e+00 7.66121328e-01 8.65824163e-01 3.76582533e-01
7.45466411e-01 3.41746658e-01 1.41472235e-01 3.42949063e-01
5.39538860e-01 5.74346006e-01 8.79752114e-02 5.12496717e-02
-4.35745865e-01 -4.21602018e-02 5.85064054e-01 1.09462047e+00
-9.08618927e-01 1.25125349e-01 -1.43130350e+00 6.60574317e-01
-1.53415465e+00 -1.15893161e+00 -2.41836756e-01 1.80812097e+00
6.30054712e-01 -7.88803935e-01 2.29604140e-01 5.47611654e-01
4.46320862e-01 1.54890551e-03 -4.05506343e-01 -5.62724710e-01
-4.52394336e-01 3.78143340e-01 6.29239202e-01 6.78019166e-01
-1.07514143e+00 1.21209252e+00 6.55517292e+00 1.89033039e-02
-1.65329444e+00 -5.93855884e-03 -3.88174385e-01 2.38397211e-01
3.10752988e-01 -3.68436813e-01 -7.27774084e-01 3.85849684e-01
3.75112981e-01 2.52292663e-01 2.67371237e-01 6.97672904e-01
3.14999729e-01 -3.32968563e-01 -8.50572884e-01 1.56557703e+00
5.08859158e-01 -1.04668748e+00 2.37040728e-01 2.33366806e-02
5.35077155e-01 4.94760752e-01 -2.20864519e-01 2.32329927e-02
4.04741049e-01 -1.25963748e+00 8.30110013e-01 7.89667368e-01
9.53650296e-01 -2.33454496e-01 9.17198360e-01 5.13181612e-02
-1.15783167e+00 -1.08492143e-01 2.26721555e-01 -1.80938572e-01
4.31704849e-01 -5.11821806e-01 -5.47963202e-01 -2.13301226e-01
7.30535209e-01 5.31125128e-01 -2.55324394e-01 1.62413323e+00
-6.43666983e-01 5.65808415e-01 -6.24366939e-01 -6.35600507e-01
3.68771344e-01 -1.37456313e-01 3.65577430e-01 1.41411519e+00
4.34079200e-01 2.89629966e-01 -3.18311006e-01 5.01547493e-02
3.48087430e-01 4.86462384e-01 -8.03765535e-01 -2.19967499e-01
8.33878815e-02 5.53756833e-01 -4.25052941e-01 -1.97392166e-01
-4.87701595e-01 1.21237218e+00 -5.18840849e-01 6.00198209e-01
-2.51442134e-01 -5.03781617e-01 5.79409182e-01 -3.30582619e-01
3.51323426e-01 -4.73892629e-01 -2.05984399e-01 -1.05425501e+00
1.92573622e-01 -8.33061099e-01 4.16184217e-01 -9.98744965e-01
-1.03404748e+00 3.48137081e-01 -2.01299652e-01 -1.44506383e+00
-2.87192374e-01 -1.13884866e+00 -5.15903592e-01 8.99669707e-01
-1.46557367e+00 -1.52602160e+00 -7.70037234e-01 8.78383636e-01
4.31323349e-01 -5.60545743e-01 1.08270144e+00 5.11411726e-01
-5.45760393e-02 4.19696987e-01 1.43018663e-01 1.03937042e+00
4.33588624e-01 -9.00299489e-01 5.12263477e-02 6.78590715e-01
4.59289163e-01 2.59033501e-01 2.74935991e-01 -5.68908274e-01
-1.25893962e+00 -2.61219025e-01 1.24656224e+00 -4.44170505e-01
3.26830834e-01 2.71114409e-01 -2.83440143e-01 7.84559011e-01
-8.42096936e-03 -4.51235324e-02 7.13841498e-01 -6.07183397e-01
-3.32089275e-01 2.34970987e-01 -1.22105551e+00 6.69741392e-01
9.91400480e-01 -4.85097826e-01 -7.96895266e-01 2.28210509e-01
-5.94965756e-01 -6.06323242e-01 -3.23735327e-01 2.07405284e-01
1.43543017e+00 -5.55055499e-01 7.62735188e-01 -6.58897936e-01
6.54179556e-03 -3.31622154e-01 -4.57846105e-01 -8.70017886e-01
3.47754717e-01 -3.94422591e-01 -1.51041836e-01 8.60996366e-01
5.51361777e-02 -6.64788902e-01 8.43935907e-01 7.24104226e-01
3.54377419e-01 -1.59407139e-01 -1.27038348e+00 -8.34839761e-01
-7.01907948e-02 -5.91532528e-01 4.54792082e-01 6.76237464e-01
-6.80468455e-02 -3.20029080e-01 -5.39548695e-01 9.07036290e-02
6.13811612e-01 -1.03300981e-01 9.89879012e-01 -1.10856760e+00
2.28796750e-01 -7.08160162e-01 -1.12762237e+00 -9.47205782e-01
-2.25472420e-01 -6.87724054e-01 1.95812911e-01 -2.08245897e+00
-3.05678934e-01 -1.10534377e-01 6.70649633e-02 8.54972363e-01
6.44091725e-01 4.78063881e-01 4.22412187e-01 3.47142190e-01
7.80103635e-03 3.98055911e-01 1.23390794e+00 -3.06410611e-01
-3.57050449e-01 8.98788720e-02 1.77259073e-01 9.99936879e-01
8.24262977e-01 -1.36361212e-01 2.91674703e-01 -5.09145916e-01
-2.66305119e-01 -4.06682283e-01 5.04995942e-01 -1.00942266e+00
4.03577387e-01 -1.76815808e-01 1.65665016e-01 -8.74126792e-01
5.33759654e-01 -9.16536152e-01 -1.74567595e-01 8.93236101e-01
1.19148076e-01 -3.56887244e-02 -1.19768992e-01 -5.49541675e-02
-4.16110784e-01 -1.78281650e-01 8.33169281e-01 -1.67166293e-01
-1.54610610e+00 1.25276178e-01 -7.17234433e-01 -2.19539210e-01
8.55433106e-01 -7.17169583e-01 1.33921012e-01 -7.24291742e-01
-9.11901772e-01 -6.72141612e-02 9.61201452e-03 5.10737598e-01
1.00398731e+00 -1.62537503e+00 -8.56822729e-01 5.23190796e-01
4.34455931e-01 -3.82180542e-01 -6.37738630e-02 7.83259809e-01
-1.44998312e+00 5.12429237e-01 -7.23255336e-01 -4.67181325e-01
-1.87569809e+00 -5.01080036e-01 4.89356995e-01 3.64429474e-01
-8.45373571e-01 9.48433459e-01 -7.94740438e-01 -5.65764546e-01
7.82694936e-01 -6.17645442e-01 -5.24570704e-01 8.36785808e-02
6.28965557e-01 2.81486660e-01 -3.04957390e-01 -1.47249937e+00
-7.46013522e-01 1.37967014e+00 4.59734976e-01 -7.64366984e-01
1.34333241e+00 2.59378821e-01 3.21047418e-02 5.24567246e-01
9.65371430e-01 -2.99304966e-02 -7.35563934e-01 -1.81722090e-01
4.50724736e-02 -5.23889720e-01 -1.31134123e-01 -1.32700515e+00
-1.13692474e+00 9.05755460e-01 1.24427497e+00 -8.60105455e-01
1.00078988e+00 -2.17926726e-02 7.68596888e-01 6.48989975e-01
4.46973741e-01 -1.28451014e+00 -1.85102597e-01 7.14381337e-01
1.50644720e+00 -1.38280070e+00 -3.14389765e-01 2.31099918e-01
-8.39640081e-01 1.21160150e+00 2.15953767e-01 -3.78247611e-02
1.01687348e+00 1.97173089e-01 1.21344900e+00 -2.95552641e-01
2.40395993e-01 -6.36391521e-01 4.56825167e-01 9.43802714e-01
4.37378556e-01 1.22045964e-01 -7.27127731e-01 2.54245967e-01
-2.74002075e-01 5.26085079e-01 3.32133681e-01 1.39045548e+00
-3.15932423e-01 -1.06259918e+00 -7.86522210e-01 9.40111205e-02
-2.11284794e-02 -3.76902558e-02 -6.82988346e-01 9.06321228e-01
8.71062651e-02 6.90272570e-01 -2.75179416e-01 -1.67610601e-01
3.57620984e-01 4.94181305e-01 8.12699497e-01 -1.19955666e-01
-1.90302789e-01 -1.76072955e-01 -9.08788741e-02 -3.77090394e-01
-8.82879376e-01 -9.38268006e-01 -1.46778238e+00 6.46118121e-03
2.11536676e-01 -5.41059315e-01 1.12844169e+00 1.01607180e+00
-2.77399778e-01 -3.48630369e-01 3.08985338e-02 -1.03308129e+00
-3.49861801e-01 -1.18175626e+00 -7.43458390e-01 4.24322307e-01
3.56165379e-01 -7.38619089e-01 -1.91485122e-01 2.84298956e-01]
|
[9.102933883666992, -6.416418075561523]
|
4449ce3c-2ed0-42df-aca7-433dc5b3be95
|
ask-me-anything-a-simple-strategy-for
|
2210.02441
| null |
https://arxiv.org/abs/2210.02441v3
|
https://arxiv.org/pdf/2210.02441v3.pdf
|
Ask Me Anything: A simple strategy for prompting language models
|
Large language models (LLMs) transfer well to new tasks out-of-the-box simply given a natural language prompt that demonstrates how to perform the task and no additional training. Prompting is a brittle process wherein small modifications to the prompt can cause large variations in the model predictions, and therefore significant effort is dedicated towards designing a painstakingly "perfect prompt" for a task. To mitigate the high degree of effort involved in prompt-design, we instead ask whether producing multiple effective, yet imperfect, prompts and aggregating them can lead to a high quality prompting strategy. Our observations motivate our proposed prompting method, ASK ME ANYTHING (AMA). We first develop an understanding of the effective prompt formats, finding that question-answering (QA) prompts, which encourage open-ended generation ("Who went to the park?") tend to outperform those that restrict the model outputs ("John went to the park. Output True or False."). Our approach recursively uses the LLM itself to transform task inputs to the effective QA format. We apply the collected prompts to obtain several noisy votes for the input's true label. We find that the prompts can have very different accuracies and complex dependencies and thus propose to use weak supervision, a procedure for combining the noisy predictions, to produce the final predictions for the inputs. We evaluate AMA across open-source model families (e.g., EleutherAI, BLOOM, OPT, and T0) and model sizes (125M-175B parameters), demonstrating an average performance lift of 10.2% over the few-shot baseline. This simple strategy enables the open-source GPT-J-6B model to match and exceed the performance of few-shot GPT3-175B on 15 of 20 popular benchmarks. Averaged across these tasks, the GPT-J-6B model outperforms few-shot GPT3-175B. We release our code here: https://github.com/HazyResearch/ama_prompting
|
['Laurel Orr', 'Christopher Ré', 'Frederic Sala', 'Ines Chami', 'Kush Bhatia', 'Neel Guha', 'Mayee F. Chen', 'Avanika Narayan', 'Simran Arora']
|
2022-10-05
| null | null | null | null |
['coreference-resolution']
|
['natural-language-processing']
|
[ 4.57115412e-01 3.34370434e-01 4.48651724e-02 -5.93641400e-01
-1.47837782e+00 -7.50486135e-01 7.57209480e-01 -3.03720329e-02
-4.16686803e-01 8.17008078e-01 2.76666075e-01 -6.74022257e-01
7.09773377e-02 -5.06138206e-01 -7.59204865e-01 -4.44856822e-01
5.27086794e-01 7.46672273e-01 2.13937998e-01 -5.50231874e-01
3.82001474e-02 -1.58535451e-01 -1.32830894e+00 5.88765740e-01
9.62744474e-01 6.73896313e-01 5.31018555e-01 9.54640627e-01
-3.07435274e-01 9.41531897e-01 -9.56000447e-01 -6.55163169e-01
1.68145120e-01 -4.48891521e-01 -1.10679471e+00 -2.89462805e-01
5.94462454e-01 -3.63078922e-01 7.04698414e-02 5.74302554e-01
5.35536289e-01 3.83507013e-01 6.04325294e-01 -1.29903734e+00
-7.79623568e-01 8.15063477e-01 -7.39913657e-02 1.91055536e-01
6.61537349e-01 7.25633502e-01 1.24154937e+00 -9.75281417e-01
6.02408111e-01 1.24775136e+00 6.87745869e-01 9.56943870e-01
-1.38420916e+00 -5.80315113e-01 1.34812519e-01 8.21223930e-02
-9.52098310e-01 -5.13454676e-01 3.55381966e-01 -4.38280821e-01
1.19916534e+00 6.14331126e-01 3.63099240e-02 1.72799301e+00
2.08231807e-01 6.29051685e-01 1.23903966e+00 -5.85574508e-01
1.62799835e-01 1.23109557e-01 5.60665071e-01 4.62383300e-01
-1.48473615e-02 1.05369419e-01 -6.58447385e-01 -3.91822040e-01
1.28575668e-01 -2.41778597e-01 -3.62661660e-01 5.64988136e-01
-1.15109396e+00 7.25725532e-01 1.85635522e-01 9.96395499e-02
-3.01027209e-01 1.25342727e-01 1.10993676e-01 5.60324490e-01
3.80140275e-01 9.79412377e-01 -7.77951300e-01 -5.01179874e-01
-8.15043271e-01 6.65246248e-01 1.00466859e+00 1.00665712e+00
8.73998523e-01 -1.98918089e-01 -9.70115185e-01 1.00690615e+00
1.34239644e-02 4.05281126e-01 4.59984601e-01 -1.13410139e+00
6.75780118e-01 3.28580260e-01 4.61809218e-01 -4.00190085e-01
-3.20476145e-01 -9.38275680e-02 -4.94306177e-01 -1.65263802e-01
6.41493082e-01 -5.85754752e-01 -1.08729112e+00 1.86344254e+00
1.43309817e-01 -9.08702686e-02 1.41101599e-01 6.69114947e-01
1.06705153e+00 9.57846224e-01 4.03556466e-01 3.62904593e-02
1.40005147e+00 -1.22981393e+00 -5.30205250e-01 -7.38158226e-01
9.14368093e-01 -9.17165399e-01 1.81770945e+00 1.41825706e-01
-9.16830122e-01 -6.99049830e-01 -5.92756808e-01 -3.05102140e-01
-1.70260787e-01 -5.81177175e-02 3.45647663e-01 3.01743507e-01
-1.18446076e+00 5.74232101e-01 -4.88817483e-01 -4.30267185e-01
1.65201575e-02 6.32230043e-02 -1.74860999e-01 -3.38145018e-01
-1.36738384e+00 1.11127138e+00 1.27896726e-01 -2.26239637e-01
-9.12958443e-01 -9.44028139e-01 -7.25953460e-01 6.89641014e-02
5.59981406e-01 -8.42091680e-01 2.19537473e+00 -7.16607451e-01
-1.44601381e+00 6.37286246e-01 -4.72857952e-01 -4.25279498e-01
3.60864967e-01 -4.09610152e-01 -1.22775085e-01 -2.38673553e-01
2.91465521e-01 1.03913331e+00 5.32943010e-01 -1.22137511e+00
-6.58573627e-01 1.89049453e-01 3.58583957e-01 1.78335533e-01
2.80709919e-02 2.92439640e-01 -3.98832373e-02 -4.48365837e-01
-3.27628076e-01 -9.50387895e-01 -2.56592184e-01 -5.53668916e-01
-6.13926232e-01 -7.18883693e-01 3.61040741e-01 -5.80280840e-01
1.16254568e+00 -1.79861319e+00 -2.48915896e-01 -3.42607737e-01
1.29738245e-02 2.30221704e-01 -5.36899030e-01 6.47574484e-01
7.60745853e-02 3.40271533e-01 -2.94908941e-01 -5.36638021e-01
6.26858473e-02 4.56325680e-01 -6.15702868e-01 -3.67194921e-01
4.04255241e-01 1.19475245e+00 -1.06633234e+00 -3.57204854e-01
-1.34548083e-01 2.15196721e-02 -6.00656986e-01 6.80406392e-01
-6.79347873e-01 3.05938303e-01 -2.29892746e-01 4.84211057e-01
2.70191938e-01 -5.54036617e-01 -1.30269587e-01 2.70456731e-01
1.46974459e-01 9.14289176e-01 -7.11216033e-01 1.59558940e+00
-5.42259872e-01 4.63267118e-01 -1.17519110e-01 -4.53453660e-01
9.60641444e-01 4.69317168e-01 -1.14867859e-01 -6.36094928e-01
-2.18166970e-02 2.45156080e-01 1.38275117e-01 -5.77702284e-01
8.19604456e-01 -2.98010588e-01 -2.66972810e-01 5.52482128e-01
3.54257733e-01 -3.69509131e-01 3.35497588e-01 5.09432137e-01
1.44289029e+00 1.32881269e-01 2.91011781e-01 -7.84594864e-02
7.45243877e-02 2.71522909e-01 4.93507415e-01 1.28842902e+00
-2.00390205e-01 6.76329792e-01 4.77679759e-01 -2.49319047e-01
-8.91408026e-01 -8.94437551e-01 2.15505615e-01 1.65950143e+00
-2.07558423e-01 -4.99568820e-01 -7.60369956e-01 -8.20074081e-01
-1.62418365e-01 1.41316748e+00 -4.52256441e-01 -1.69062734e-01
-5.68998873e-01 -3.72653395e-01 6.16598487e-01 3.20957035e-01
1.59017593e-01 -1.41778874e+00 -6.16555393e-01 2.64210492e-01
-6.61544144e-01 -1.00154495e+00 -6.33839071e-01 5.86313546e-01
-6.33536041e-01 -8.04206312e-01 -5.66385448e-01 -5.84980130e-01
4.85255420e-01 1.78495094e-01 1.62898636e+00 1.22422077e-01
2.70496607e-01 3.58414620e-01 -6.93441749e-01 -4.42696303e-01
-8.63296390e-01 2.65865117e-01 -2.39090070e-01 -4.22070771e-01
4.05061752e-01 -3.85861665e-01 -3.41919899e-01 2.66311139e-01
-7.17347026e-01 3.52559060e-01 5.91183424e-01 1.10675895e+00
2.10751846e-01 -6.02209032e-01 6.93956017e-01 -1.28992474e+00
1.05662799e+00 -6.38045490e-01 -1.43354088e-01 4.13275629e-01
-5.61400175e-01 1.92838475e-01 8.66050184e-01 -5.35942733e-01
-1.33024430e+00 -2.44976938e-01 -3.52271378e-01 -1.50638565e-01
-4.11870539e-01 3.47416401e-01 1.42657459e-02 4.38954860e-01
1.18598938e+00 5.35556450e-02 -4.10595089e-01 -5.89930952e-01
6.80294752e-01 7.42678285e-01 6.07278407e-01 -9.73904252e-01
6.67957127e-01 -1.69868901e-01 -6.79669261e-01 -3.44180673e-01
-1.17842770e+00 -4.19451743e-01 -1.58056840e-02 6.30671112e-03
5.45796514e-01 -7.81878114e-01 -3.34748179e-01 4.51317936e-01
-1.68406880e+00 -9.14186418e-01 -3.81543845e-01 -5.50402701e-02
-4.30121154e-01 2.22580060e-02 -7.27147400e-01 -8.16415131e-01
-6.38982058e-01 -1.09585500e+00 1.08744848e+00 3.95355701e-01
-8.13341618e-01 -7.50048697e-01 2.57126689e-01 5.60840607e-01
3.70716870e-01 -2.70286530e-01 1.04560959e+00 -1.06877232e+00
-3.24297220e-01 1.58364370e-01 2.42201202e-02 3.93208534e-01
2.85718758e-02 -9.87957790e-02 -1.21768677e+00 -1.32814236e-02
9.31721926e-02 -6.88609242e-01 6.74715579e-01 9.26433802e-02
8.74940395e-01 -6.88624978e-01 -1.98532611e-01 2.83956647e-01
1.05190480e+00 1.61940753e-01 4.23006296e-01 1.80711403e-01
4.59068596e-01 5.47706366e-01 8.04362178e-01 1.85110494e-01
5.42257547e-01 6.23041868e-01 7.16556013e-02 1.72640786e-01
-2.24308610e-01 -5.65772951e-01 6.98540270e-01 8.16503167e-01
2.76165575e-01 -4.70124513e-01 -1.02287459e+00 6.68387353e-01
-1.67631686e+00 -7.67964900e-01 -1.84994087e-01 1.93487203e+00
1.46302617e+00 1.74332976e-01 -2.77086616e-01 -2.84555852e-01
3.52966130e-01 2.87485480e-01 -3.69369864e-01 -7.52739966e-01
5.97048029e-02 5.16048431e-01 6.33618459e-02 7.67911553e-01
-6.19322538e-01 1.13485193e+00 6.33668041e+00 9.39421177e-01
-8.71009171e-01 3.69733661e-01 7.42368340e-01 -8.31028074e-02
-6.49429977e-01 3.15363526e-01 -1.10497916e+00 6.44765437e-01
1.40869582e+00 -2.94788450e-01 5.10292947e-01 8.30318689e-01
2.50495613e-01 -1.90393075e-01 -1.48472893e+00 5.99657595e-01
-1.09392494e-01 -1.14843452e+00 5.50051145e-02 -4.20855820e-01
8.56498539e-01 1.41635194e-01 -1.41049325e-01 9.69907463e-01
8.36999178e-01 -1.07040548e+00 7.90505052e-01 5.01679540e-01
6.77526832e-01 -1.08449481e-01 5.56200862e-01 8.23407233e-01
-6.47478700e-01 -1.48319036e-01 -3.47930759e-01 -4.43112284e-01
3.63149017e-01 5.10052383e-01 -1.35764706e+00 4.04168963e-01
6.35913849e-01 1.28045514e-01 -6.89163625e-01 7.17373133e-01
-6.09736741e-01 9.91655946e-01 -2.80156493e-01 -2.38619447e-01
2.42984161e-01 1.99615091e-01 5.46549678e-01 1.36990345e+00
3.38091522e-01 2.96027094e-01 1.52509123e-01 1.05861652e+00
-3.04276794e-01 -1.32278189e-01 -4.40026015e-01 3.29589024e-02
8.23307395e-01 1.33661342e+00 -1.01112783e-01 -6.65450573e-01
-4.46923912e-01 8.52269828e-01 5.30441165e-01 4.93925929e-01
-7.77718127e-01 -9.90401432e-02 5.39341509e-01 9.40174237e-02
-5.04084975e-02 1.59158006e-01 -3.57181609e-01 -9.25411344e-01
6.74079508e-02 -1.39269364e+00 2.94415265e-01 -1.21658802e+00
-1.51550210e+00 7.02309072e-01 4.88742143e-02 -7.34990537e-01
-6.76484585e-01 -4.10143584e-01 -9.70193267e-01 1.15068913e+00
-1.39692426e+00 -8.30836356e-01 -3.73083562e-01 2.65434206e-01
9.72174466e-01 2.23070562e-01 9.63848948e-01 6.44946694e-02
-3.45225006e-01 6.41236067e-01 -3.08893591e-01 -1.29964173e-01
1.14904511e+00 -1.41443419e+00 9.07201707e-01 8.55278611e-01
2.75459766e-01 7.40355432e-01 9.38236892e-01 -6.22040570e-01
-9.47034597e-01 -1.12773323e+00 1.43056381e+00 -1.03456140e+00
5.95937252e-01 -3.49339515e-01 -1.15851820e+00 9.05782878e-01
4.76010233e-01 -2.61192918e-01 5.33993781e-01 2.36754835e-01
-3.39044243e-01 2.43844502e-02 -8.55524361e-01 7.02969253e-01
8.70980322e-01 -4.73081499e-01 -1.10355186e+00 4.58027691e-01
1.35798669e+00 -5.25790632e-01 -4.98939365e-01 2.09399909e-01
2.14778841e-01 -9.46076155e-01 5.81627548e-01 -8.75070274e-01
6.71345770e-01 1.10979773e-01 -4.72292639e-02 -1.57037210e+00
-2.70654738e-01 -8.96104395e-01 -3.35572027e-02 1.34587157e+00
8.10000718e-01 -6.29184902e-01 4.55564678e-01 1.05524802e+00
-3.61407608e-01 -9.74686742e-01 -7.32907236e-01 -6.16177440e-01
2.15965524e-01 -5.52138150e-01 6.66725576e-01 6.76969886e-01
-1.69488192e-02 7.33816683e-01 -3.47533047e-01 -1.75332621e-01
1.97599977e-01 -5.72016351e-02 9.90240097e-01 -9.84326363e-01
-6.51807308e-01 -1.33301705e-01 5.87327242e-01 -1.37487066e+00
-1.47574134e-02 -9.16026354e-01 4.92164642e-01 -1.64540064e+00
2.69557178e-01 -6.09204948e-01 -1.77626416e-01 1.06101418e+00
-7.78978288e-01 -2.23969370e-01 5.00680387e-01 2.58015722e-01
-6.28389478e-01 3.36766213e-01 1.18158364e+00 -6.66545257e-02
-2.66160220e-01 -1.42486393e-02 -1.04042602e+00 5.02377748e-01
6.77680671e-01 -6.82784677e-01 -3.09400171e-01 -6.74633682e-01
2.54853606e-01 1.60144538e-01 4.32786465e-01 -7.66279638e-01
2.36975104e-01 -2.79585063e-01 -1.26961485e-01 -2.85646379e-01
3.76672417e-01 -2.77998358e-01 1.59505615e-03 1.62762761e-01
-7.40326703e-01 8.06447864e-02 1.69774771e-01 2.27304026e-01
2.43813414e-02 -5.72867155e-01 5.08290589e-01 -2.65932679e-01
-6.07559144e-01 1.48456004e-02 -2.72786260e-01 5.16521931e-01
5.22383213e-01 1.16703577e-01 -8.52979362e-01 -4.98043358e-01
-3.71504515e-01 3.72340143e-01 2.66924948e-01 6.07321203e-01
3.06594193e-01 -1.02364552e+00 -8.44820142e-01 -1.14039190e-01
2.06227854e-01 2.68057376e-01 9.89615545e-02 6.80451453e-01
1.40364990e-02 5.63858509e-01 1.07710429e-01 -4.72792208e-01
-1.07932103e+00 1.39961392e-01 2.58245349e-01 -6.68964386e-01
-3.77096683e-01 1.20870554e+00 1.71982706e-01 -7.86369383e-01
1.21241547e-01 -6.38399243e-01 8.46835300e-02 -6.09451532e-02
3.08220506e-01 2.11901411e-01 1.39923230e-01 -1.34455726e-01
-1.01106651e-01 -4.19124588e-02 -2.62735188e-01 -2.88255155e-01
1.04974580e+00 6.09324612e-02 1.71308115e-01 6.31901979e-01
8.35667372e-01 -1.96081713e-01 -1.27676690e+00 -3.23013097e-01
1.36039034e-01 -2.84432709e-01 -3.96314800e-01 -1.40080702e+00
-2.86707580e-01 8.97687316e-01 6.81100413e-02 2.75311291e-01
8.19106698e-01 1.67622775e-01 9.94814873e-01 7.19111681e-01
3.59976619e-01 -9.32634830e-01 1.97481111e-01 8.66352916e-01
9.86814559e-01 -1.33515811e+00 -4.58494842e-01 2.44537275e-02
-9.62048233e-01 6.24877334e-01 9.84412193e-01 1.53162047e-01
-1.12886481e-01 1.96860731e-01 4.31847274e-01 -2.59460807e-02
-1.54325247e+00 3.07814013e-02 1.33267149e-01 5.66640317e-01
5.63945413e-01 2.29054376e-01 -2.44877890e-01 1.10915697e+00
-5.50674498e-01 -3.31932120e-02 5.68356097e-01 7.79098749e-01
-6.17692947e-01 -1.27645648e+00 -4.46841329e-01 6.59177959e-01
-2.87958831e-01 -5.74502945e-01 -5.36787868e-01 5.71099460e-01
1.23271700e-02 1.31550014e+00 -2.28816509e-01 -3.73277575e-01
4.41416830e-01 6.77206337e-01 1.86280683e-01 -1.05967915e+00
-9.42512333e-01 -3.28117549e-01 4.31061298e-01 -4.99596149e-01
3.73482965e-02 -4.44018900e-01 -1.17659116e+00 -2.05374137e-01
-2.39569828e-01 2.30666488e-01 3.49518389e-01 1.05831969e+00
4.71453965e-01 4.16242003e-01 4.08484519e-01 -5.99161685e-01
-1.06683445e+00 -1.42175817e+00 2.04657987e-02 5.84746540e-01
2.25057572e-01 -5.14016509e-01 -4.85630900e-01 1.48749709e-01]
|
[11.117913246154785, 8.491462707519531]
|
37791ff1-f4a6-4e1d-85c0-bc515ced1162
|
model-aggregation-for-risk-evaluation-and
|
2201.0637
| null |
https://arxiv.org/abs/2201.06370v2
|
https://arxiv.org/pdf/2201.06370v2.pdf
|
Model Aggregation for Risk Evaluation and Robust Optimization
|
We introduce a new approach for prudent risk evaluation based on stochastic dominance, which will be called the model aggregation (MA) approach. In contrast to the classic worst-case risk (WR) approach, the MA approach produces not only a robust value of risk evaluation but also a robust distributional model which is useful for modeling, analysis and simulation, independent of any specific risk measure. The MA approach is easy to implement even if the uncertainty set is non-convex or the risk measure is computationally complicated, and it provides great tractability in distributionally robust optimization. Via an equivalence property between the MA and the WR approaches, new axiomatic characterizations are obtained for a few classes of popular risk measures. In particular, the Expected Shortfall (ES, also known as CVaR) is the unique risk measure satisfying the equivalence property for convex uncertainty sets among a very large class. The MA approach for Wasserstein and mean-variance uncertainty sets admits explicit formulas for the obtained robust models, and the new approach is illustrated with various risk measures and examples from portfolio optimization.
|
['Qinyu Wu', 'Ruodu Wang', 'Tiantian Mao']
|
2022-01-17
| null | null | null | null |
['portfolio-optimization']
|
['time-series']
|
[-8.86558220e-02 2.45343998e-01 1.26842305e-01 -2.61814117e-01
-9.48533654e-01 -8.03816438e-01 4.10253435e-01 4.65145648e-01
-3.11958730e-01 8.42050433e-01 7.25677460e-02 -4.34298337e-01
-1.02335620e+00 -9.71946657e-01 -3.76822710e-01 -9.18308556e-01
-2.51413137e-01 3.71556938e-01 -5.36529757e-02 -9.59181264e-02
3.57145756e-01 7.66792536e-01 -1.44535220e+00 -6.40957415e-01
1.30708778e+00 1.57526147e+00 -3.55040669e-01 2.31200516e-01
-1.11774029e-02 3.68070304e-01 -5.77472627e-01 -7.29046464e-01
4.52270508e-01 -1.56011865e-01 -5.23845732e-01 -3.16176653e-01
-2.30404690e-01 -2.49652378e-02 5.20250320e-01 1.49641120e+00
2.89995342e-01 4.07867581e-01 1.17384839e+00 -1.28036845e+00
-4.70882833e-01 9.23897445e-01 -5.27214348e-01 2.49751974e-02
1.24441378e-01 -4.77597177e-01 1.11064100e+00 -7.06620574e-01
1.44294307e-01 1.22382843e+00 6.00067019e-01 2.36252069e-01
-1.13169658e+00 -1.84584960e-01 5.93124032e-02 -3.56876016e-01
-1.32124996e+00 1.61121264e-01 4.13169175e-01 -6.08021677e-01
1.66970372e-01 7.44025409e-01 3.07587326e-01 4.04301852e-01
6.71298921e-01 3.48692566e-01 1.26821089e+00 -3.83241892e-01
6.95911884e-01 1.23071820e-01 2.84711987e-01 4.04074728e-01
8.56660247e-01 3.16056162e-01 1.56749114e-01 -6.35251582e-01
4.66607720e-01 3.89464051e-02 -3.86666685e-01 -5.58887362e-01
-1.04488206e+00 8.95583749e-01 -1.61439136e-01 3.43234658e-01
-2.97883213e-01 5.27518429e-02 1.72711983e-01 4.28708106e-01
6.87232494e-01 5.14075458e-01 -7.20681548e-02 1.85022727e-01
-6.82219744e-01 4.73914117e-01 1.14491975e+00 6.55529857e-01
1.40122488e-01 1.84181109e-01 -4.58142102e-01 3.59347194e-01
5.78091383e-01 7.65391290e-01 7.52117252e-03 -9.95480001e-01
5.51305950e-01 1.44396544e-01 6.63178802e-01 -8.85810614e-01
-4.82306927e-01 -4.82121110e-01 -6.37060583e-01 6.72601759e-01
7.47044086e-01 -2.32378736e-01 -2.67417550e-01 1.88694263e+00
9.94527936e-02 -1.56756893e-01 2.13177338e-01 3.51341099e-01
9.52374004e-03 5.23046851e-01 -7.72476643e-02 -9.41762745e-01
8.12442541e-01 -1.56665087e-01 -9.79650080e-01 5.25158167e-01
2.99082667e-01 -5.37779391e-01 7.57766843e-01 6.95350707e-01
-1.27605748e+00 1.56233847e-01 -1.19313490e+00 6.25229597e-01
-2.84100533e-01 -6.48853481e-01 1.39661089e-01 1.13707542e+00
-8.37628305e-01 1.08651531e+00 -3.57960284e-01 1.01361774e-01
8.20511952e-02 -4.07986157e-03 -9.43203717e-02 3.12935293e-01
-1.14548230e+00 1.12982345e+00 3.57265085e-01 2.76562810e-01
-8.80534291e-01 -7.42138028e-01 -8.31947923e-01 1.64781049e-01
4.31192487e-01 -1.98523685e-01 1.05292058e+00 -4.45721775e-01
-1.54571080e+00 3.08458924e-01 4.13563073e-01 -5.34959972e-01
1.07699955e+00 -2.57645935e-01 -3.66890192e-01 3.07913292e-02
-1.25663847e-01 -5.55444360e-01 9.11051452e-01 -1.05869102e+00
-2.43052796e-01 -2.40276799e-01 1.56611383e-01 6.37470782e-02
3.04316799e-03 5.49137950e-01 7.79076993e-01 -9.98904228e-01
8.12149048e-02 -5.35745442e-01 -5.80984354e-01 -1.14513122e-01
-4.26885098e-01 -1.71620309e-01 -5.89809706e-03 -3.64358634e-01
1.47783709e+00 -2.00283265e+00 3.03997070e-01 8.60397637e-01
-1.57252118e-01 -2.29096010e-01 4.07460511e-01 5.49716651e-01
-2.91429996e-01 4.52048331e-01 -8.33869696e-01 9.27884504e-02
6.19743943e-01 -1.54859349e-01 -5.34782946e-01 7.54328847e-01
7.34812766e-02 4.29279566e-01 -8.56857598e-01 -2.74886489e-01
1.78680271e-01 2.23677941e-02 -1.33102164e-01 2.23565698e-01
-1.14679284e-01 1.49376048e-02 -4.61697310e-01 4.48287994e-01
8.05675447e-01 2.49374717e-01 -4.48303893e-02 1.63500890e-01
-2.37291813e-01 -4.12702322e-01 -1.72851586e+00 8.26155841e-01
-3.88455480e-01 -1.38291925e-01 3.72170471e-02 -8.77666295e-01
1.08405423e+00 3.90605062e-01 4.87895846e-01 2.45995224e-01
3.71833384e-01 4.72835690e-01 -3.11396658e-01 -8.41118842e-02
2.67978907e-01 -7.95777798e-01 -4.18351352e-01 8.68899465e-01
-2.34775543e-01 -2.75325596e-01 3.02586295e-02 -1.02565072e-01
6.14601374e-01 -1.97488159e-01 5.96178532e-01 -1.10370517e+00
5.80700219e-01 -5.87166131e-01 7.22613513e-01 5.36825120e-01
-3.70256573e-01 4.97107804e-01 9.19050574e-01 -1.06287181e-01
-5.38010061e-01 -1.75738358e+00 -6.40006781e-01 5.95083594e-01
2.18273625e-01 -7.74988160e-02 -7.52041757e-01 -5.02063274e-01
3.43791008e-01 1.02081895e+00 -6.73333943e-01 -2.65479982e-01
-1.16634905e-01 -1.05700088e+00 2.76137590e-01 3.86944562e-01
1.92519784e-01 -5.69191694e-01 -3.75770032e-01 1.73038125e-01
2.05715552e-01 -4.15804416e-01 -4.92581159e-01 1.43315941e-01
-8.40189576e-01 -1.20553851e+00 -9.24324274e-01 8.17238390e-02
3.35573345e-01 -3.51311386e-01 9.86431777e-01 -5.51735699e-01
4.23439196e-04 5.90378821e-01 -1.15175441e-01 -9.10913765e-01
-3.88637602e-01 -9.62631643e-01 4.40017402e-01 4.75904882e-01
-1.81885377e-01 -3.51475239e-01 -2.85363704e-01 4.20778692e-01
-9.63615716e-01 -9.88667488e-01 2.23118439e-02 7.23948658e-01
9.29197073e-01 5.90495646e-01 1.04134476e+00 -6.31349385e-01
9.74268675e-01 -6.27817869e-01 -1.01891732e+00 6.13523543e-01
-8.84164035e-01 2.73073316e-01 3.35917741e-01 -2.82489490e-02
-1.38097465e+00 -4.64612454e-01 2.96167552e-01 1.13607459e-02
3.60899597e-01 6.71813369e-01 -4.60295767e-01 -1.06261976e-01
4.24924344e-01 -3.09435189e-01 -7.34494627e-02 -5.73503911e-01
2.62973636e-01 3.39701295e-01 3.26673090e-01 -8.65426898e-01
7.18535066e-01 4.67317373e-01 5.21378636e-01 -5.70537865e-01
-9.31293249e-01 6.10826798e-02 -5.00861585e-01 -2.31603086e-01
9.20443356e-01 -2.27509126e-01 -5.50422072e-01 4.04737413e-01
-7.60297954e-01 1.02161013e-01 -8.62138271e-01 6.62628293e-01
-1.02199280e+00 4.95820135e-01 -3.61299902e-01 -1.64835930e+00
-2.44924381e-01 -1.11644876e+00 5.12391627e-01 -1.46958763e-02
1.05664715e-01 -1.40493405e+00 9.01139602e-02 -1.30313873e-01
2.96179861e-01 1.01243901e+00 1.06760597e+00 -8.41298103e-01
-7.93662444e-02 -5.40874302e-01 -7.14680925e-02 8.21705759e-01
-1.84620425e-01 2.35711798e-01 -6.13301694e-01 -1.11711793e-01
6.52449131e-01 1.43274233e-01 6.37995660e-01 5.85138977e-01
1.01674533e+00 -3.24189812e-01 1.74180284e-01 3.91922653e-01
1.57989669e+00 2.02635378e-01 3.53941530e-01 2.29853407e-01
1.68058053e-01 9.64306653e-01 8.93934965e-01 6.22756362e-01
-5.22604659e-02 3.94292325e-01 5.92073321e-01 6.05601251e-01
8.61387014e-01 2.78780282e-01 5.43335080e-01 6.06381774e-01
-3.38539332e-01 3.86634097e-02 -7.73309290e-01 1.81721270e-01
-1.84105587e+00 -1.18977237e+00 -1.03773601e-01 2.94517612e+00
7.04293787e-01 1.25584587e-01 2.82744110e-01 1.80429012e-01
9.76205111e-01 1.07363120e-01 -2.56404817e-01 -9.16609704e-01
-2.52227753e-01 1.30784824e-01 7.00127661e-01 8.36789727e-01
-1.05648673e+00 -7.64275119e-02 7.55455303e+00 1.00111628e+00
-2.01439112e-01 1.51837945e-01 5.51694095e-01 -5.15175536e-02
-7.05247700e-01 -1.89593658e-01 -6.20995343e-01 5.81239939e-01
9.66307878e-01 -1.12183619e+00 -1.73438445e-01 9.54586565e-01
1.56023487e-01 -5.96867017e-02 -1.06831586e+00 6.46277905e-01
-3.04506212e-01 -9.05240953e-01 -2.44448006e-01 2.21633986e-01
8.04241121e-01 -5.13888180e-01 2.18649566e-01 -1.98212236e-01
5.36979377e-01 -1.07181859e+00 8.75418842e-01 1.06140625e+00
6.68352902e-01 -1.54632485e+00 1.17179358e+00 2.32944027e-01
-1.05954647e+00 -3.15507412e-01 -5.41718364e-01 1.70627892e-01
5.18893361e-01 1.13211143e+00 2.96564877e-01 9.67749834e-01
4.41725463e-01 2.76999712e-01 -1.42111540e-01 1.22679293e+00
-2.75313482e-02 1.59617007e-01 -5.33527851e-01 1.04080662e-01
1.62934795e-01 -6.10041261e-01 9.38304663e-01 8.55056345e-01
7.45721638e-01 -8.34161788e-02 -1.91627041e-01 9.31301057e-01
2.79316783e-01 4.07125086e-01 -7.27349460e-01 2.23852649e-01
3.92002225e-01 8.67339790e-01 -5.61687708e-01 6.24586083e-02
-1.04585618e-01 2.51341403e-01 -1.21544488e-01 2.39620537e-01
-7.38532543e-01 -7.14736819e-01 5.78805149e-01 -2.28822455e-01
-7.62383491e-02 5.06197400e-02 -4.17673558e-01 -1.14333713e+00
2.34938636e-01 -4.45090115e-01 8.26410890e-01 -1.74577817e-01
-1.69455123e+00 6.08575821e-01 6.72031105e-01 -1.43756473e+00
-2.52811015e-01 -8.28350723e-01 -9.60543096e-01 1.12272358e+00
-1.12234914e+00 -3.59198600e-01 3.16684872e-01 4.84909058e-01
-2.13969573e-01 -2.29884669e-01 7.54996300e-01 -3.28071177e-01
-6.95845127e-01 5.16534030e-01 7.32741833e-01 -5.91731131e-01
3.83219719e-01 -1.85615444e+00 -1.83932692e-01 9.86001372e-01
-3.75134826e-01 4.29187089e-01 1.03549612e+00 -6.09462261e-01
-9.33288693e-01 -9.95232105e-01 6.51280344e-01 -5.91181934e-01
1.16091239e+00 2.95596570e-02 -7.12009847e-01 4.07228351e-01
-1.45285144e-01 -1.69696450e-01 9.37991619e-01 -5.76897562e-02
-3.33973587e-01 -2.06006646e-01 -1.67069590e+00 5.07534683e-01
6.06102467e-01 -1.24456614e-01 -1.02158833e+00 3.90429765e-01
6.93294048e-01 1.76671401e-01 -1.46087670e+00 5.48828602e-01
5.29293478e-01 -1.21855807e+00 8.87802184e-01 -5.17437041e-01
-9.86909792e-02 -2.39052802e-01 -5.71679354e-01 -1.19319284e+00
-1.25473112e-01 -1.09852958e+00 -2.18858078e-01 1.25887859e+00
4.02812839e-01 -1.18935800e+00 2.81243455e-02 8.12646329e-01
1.38774678e-01 -9.93739903e-01 -1.27054918e+00 -1.53194273e+00
6.63579047e-01 -6.44180417e-01 9.47424054e-01 6.71133637e-01
2.74515629e-01 -5.76137364e-01 -2.55531609e-01 1.02710389e-01
1.30450797e+00 1.93123043e-01 -1.49761125e-01 -1.56059921e+00
-4.24033731e-01 -8.38846028e-01 -4.01329994e-01 -4.54232208e-02
1.41646042e-01 -7.77442575e-01 5.99685544e-03 -1.17155468e+00
-1.94142357e-01 -4.25602615e-01 -8.17182779e-01 -2.24389508e-01
-4.96178493e-02 -2.39317670e-01 2.12219208e-01 7.00649396e-02
-1.90500021e-01 7.21268296e-01 9.33267653e-01 2.02423885e-01
-6.11085817e-02 6.85887992e-01 -9.02251184e-01 1.22294545e+00
6.96914792e-01 -2.79394656e-01 -5.85620046e-01 2.19560921e-01
5.72262347e-01 2.34577388e-01 2.11549655e-01 -7.61740267e-01
-2.77489036e-01 -5.30598819e-01 -8.88977051e-02 -5.06698251e-01
-7.82390311e-02 -8.26728880e-01 3.80047590e-01 4.30902719e-01
-4.93345022e-01 1.82858080e-01 -1.96677744e-01 8.88588071e-01
-2.32983693e-01 -9.08125460e-01 1.07452512e+00 1.42921835e-01
-7.17054009e-02 3.65169197e-01 -1.71260208e-01 1.90520212e-01
1.50739527e+00 -3.00041977e-02 -3.47404063e-01 -4.54524457e-01
-1.03640819e+00 2.79174745e-01 3.60721260e-01 -6.73274547e-02
4.88528848e-01 -1.60178316e+00 -8.40649307e-01 -3.03560853e-01
-3.49529646e-02 -9.42481458e-02 2.35423684e-01 8.80997539e-01
-5.16891360e-01 1.28751203e-01 -3.34958509e-02 -8.80149230e-02
-5.69927454e-01 8.34763706e-01 5.30074418e-01 -4.55352664e-01
-3.42808932e-01 7.11550772e-01 3.16255450e-01 1.65284388e-02
3.24778944e-01 -3.96893382e-01 -2.72588760e-01 2.57088691e-01
1.00048447e+00 9.75320637e-01 -2.29950964e-01 -6.25869513e-01
-2.32959419e-01 6.23154640e-01 4.26771015e-01 -4.39910322e-01
1.13800025e+00 -3.07120204e-01 -5.60989678e-01 8.21757078e-01
8.46342504e-01 1.08569421e-01 -1.03528178e+00 -2.27876119e-02
4.58713174e-01 -3.85078639e-01 -2.12243378e-01 -6.11979544e-01
-7.07063615e-01 8.36314380e-01 3.07027310e-01 8.13115060e-01
1.10032630e+00 -2.03925103e-01 2.08860919e-01 3.91258180e-01
6.89400852e-01 -1.23091424e+00 -3.63500148e-01 2.98347503e-01
1.53346586e+00 -7.82023251e-01 7.62699544e-02 -4.82548654e-01
-7.53667533e-01 1.12778711e+00 -5.61625138e-02 -3.71651202e-01
1.39028788e+00 3.20641905e-01 -1.85324728e-01 2.30121225e-01
-3.48004460e-01 -9.75833982e-02 4.91050839e-01 7.25699008e-01
8.88891146e-02 4.38936353e-01 -7.00809538e-01 1.17712140e+00
-2.71939069e-01 -5.19077241e-01 7.89220989e-01 7.72087038e-01
-7.09535241e-01 -8.44157398e-01 -4.60003763e-01 5.45083046e-01
-8.54126990e-01 5.26938289e-02 7.22713768e-02 5.66998005e-01
-3.42419714e-01 9.90594685e-01 -1.14497602e-01 -1.40721664e-01
4.55671877e-01 8.14558566e-02 2.87744015e-01 -5.89047432e-01
-2.99780369e-01 -8.50331709e-02 8.68129805e-02 -7.22991824e-01
-2.52272397e-01 -8.69390488e-01 -8.71621549e-01 -3.54811043e-01
-4.46042955e-01 5.22872865e-01 3.66433412e-01 1.00595510e+00
-1.97362348e-01 5.61805367e-02 9.70060766e-01 -6.05593801e-01
-1.61762500e+00 -6.62330687e-01 -1.57725084e+00 -4.55412455e-02
1.80947423e-01 -9.67246532e-01 -9.35898483e-01 -4.88696098e-01]
|
[5.06982421875, 3.933800220489502]
|
778a0c6c-a428-4531-8935-77bfe1f6cdbd
|
less-is-more-removing-text-regions-improves
|
2305.05095
| null |
https://arxiv.org/abs/2305.05095v1
|
https://arxiv.org/pdf/2305.05095v1.pdf
|
Less is More: Removing Text-regions Improves CLIP Training Efficiency and Robustness
|
The CLIP (Contrastive Language-Image Pre-training) model and its variants are becoming the de facto backbone in many applications. However, training a CLIP model from hundreds of millions of image-text pairs can be prohibitively expensive. Furthermore, the conventional CLIP model doesn't differentiate between the visual semantics and meaning of text regions embedded in images. This can lead to non-robustness when the text in the embedded region doesn't match the image's visual appearance. In this paper, we discuss two effective approaches to improve the efficiency and robustness of CLIP training: (1) augmenting the training dataset while maintaining the same number of optimization steps, and (2) filtering out samples that contain text regions in the image. By doing so, we significantly improve the classification and retrieval accuracy on public benchmarks like ImageNet and CoCo. Filtering out images with text regions also protects the model from typographic attacks. To verify this, we build a new dataset named ImageNet with Adversarial Text Regions (ImageNet-Attr). Our filter-based CLIP model demonstrates a top-1 accuracy of 68.78\%, outperforming previous models whose accuracy was all below 50\%.
|
['Yantao Zheng', 'Zhiyun Lu', 'Wencong Zhang', 'Xianzhi Du', 'Yinfei Yang', 'Chen Chen', 'BoWen Zhang', 'Liangliang Cao']
|
2023-05-08
| null | null | null | null |
['adversarial-text']
|
['adversarial']
|
[ 3.05816084e-01 -3.66043538e-01 -2.92046536e-02 -1.43298909e-01
-7.94822097e-01 -7.84438372e-01 5.47395170e-01 -1.97248552e-02
-4.53109175e-01 3.85643125e-01 -1.05462102e-02 -2.28124589e-01
5.40662766e-01 -5.96566319e-01 -1.13539362e+00 -5.13801277e-01
1.52400434e-01 -4.48707156e-02 5.01142085e-01 -1.01725966e-01
2.68863589e-01 1.87453672e-01 -1.40182245e+00 7.31711984e-01
8.85610521e-01 1.21940446e+00 8.68136063e-02 6.24430120e-01
-1.11035839e-01 8.78272414e-01 -7.81003654e-01 -7.97140718e-01
6.33897662e-01 -3.22071701e-01 -5.10930061e-01 -7.19146281e-02
1.04929101e+00 -4.31010783e-01 -5.36142170e-01 1.46809268e+00
3.59943151e-01 -1.78533092e-01 4.34261441e-01 -1.29094982e+00
-9.62982059e-01 6.02216244e-01 -7.64330506e-01 1.88158322e-02
2.25001082e-01 2.10768014e-01 8.22938979e-01 -1.13797879e+00
8.18584383e-01 1.10777223e+00 6.66115165e-01 3.57896030e-01
-1.08944929e+00 -1.01292598e+00 1.37923926e-01 2.18810305e-01
-1.56095362e+00 -2.99008191e-01 6.29899323e-01 -4.37800765e-01
6.19859993e-01 4.43144411e-01 4.74552870e-01 1.34182489e+00
2.56518483e-01 8.68735254e-01 1.14106011e+00 -3.15488189e-01
3.41312066e-02 3.81133407e-01 -1.38810098e-01 4.88381386e-01
4.57033627e-02 1.12513386e-01 -4.59664315e-01 -7.82763511e-02
3.87795955e-01 5.78217022e-02 -1.61509052e-01 -1.79110646e-01
-1.04895937e+00 7.11640716e-01 5.58841944e-01 1.76378191e-01
4.20809835e-02 6.62470311e-02 6.09989762e-01 4.11877930e-01
4.38914657e-01 3.93913209e-01 -4.18569110e-02 2.19207391e-01
-1.14668989e+00 2.93441117e-01 5.46954334e-01 1.19164872e+00
5.63970983e-01 7.72844926e-02 -1.24576434e-01 7.93093443e-01
-1.05938099e-01 7.98029840e-01 2.41062075e-01 -6.62791312e-01
6.25169873e-01 5.07495821e-01 -4.09284532e-02 -1.31992412e+00
2.63631970e-01 -3.48651946e-01 -9.53273237e-01 3.18468451e-01
4.04485226e-01 6.79914420e-03 -1.17712247e+00 1.47443557e+00
-6.96403235e-02 5.72190247e-02 -1.98780179e-01 8.41003597e-01
7.08243012e-01 7.91163743e-01 1.73464209e-01 2.63812453e-01
1.42636311e+00 -1.06853545e+00 -6.56664550e-01 -3.34496498e-01
2.36539453e-01 -1.17887628e+00 1.30208063e+00 3.83592814e-01
-8.77983272e-01 -6.32640004e-01 -1.16306829e+00 -7.60448202e-02
-7.47135520e-01 2.00638305e-02 2.47583747e-01 6.20051742e-01
-8.40454161e-01 2.96514392e-01 -3.79676431e-01 -1.35773584e-01
6.37352049e-01 2.10487291e-01 -4.81673300e-01 -2.60360599e-01
-1.19541109e+00 5.26539862e-01 2.58378893e-01 -2.45397776e-01
-8.40861678e-01 -8.49822283e-01 -7.27416873e-01 1.16746433e-01
4.10229325e-01 -1.15096100e-01 8.03544939e-01 -1.60308182e+00
-9.89895225e-01 1.02705896e+00 -1.75483916e-02 -5.45261204e-01
8.40333283e-01 -3.99278134e-01 -6.24128342e-01 3.82500768e-01
5.12084477e-02 8.63721669e-01 1.44564688e+00 -1.40484691e+00
-5.73379159e-01 -1.04062527e-01 -8.49395692e-02 -8.77846107e-02
-5.63888073e-01 2.36469015e-01 -9.47386920e-01 -1.19831884e+00
-2.59070963e-01 -1.03407598e+00 3.76191363e-02 3.43644321e-01
-6.37711823e-01 3.54241878e-01 1.19181538e+00 -7.04159081e-01
1.16165495e+00 -2.56995845e+00 -3.71540785e-01 3.43474895e-01
2.64601439e-01 5.97122669e-01 -3.91340584e-01 3.89529169e-01
-7.06470907e-02 4.81042027e-01 -1.19192250e-01 -3.11791807e-01
-1.09344155e-01 -7.51092359e-02 -7.89571345e-01 3.88232648e-01
1.55958265e-01 8.91201794e-01 -5.61986625e-01 -7.01632679e-01
2.67008811e-01 5.28209925e-01 -7.50464201e-01 2.22721472e-02
-3.24947804e-01 1.67971805e-01 -2.08840892e-01 7.26517022e-01
9.48255062e-01 -1.50278240e-01 -8.66165608e-02 -2.38913640e-01
1.42759174e-01 -1.17045157e-01 -9.78665054e-01 1.34177470e+00
-1.22830682e-01 9.94959772e-01 4.57139947e-02 -6.84786022e-01
6.59604907e-01 1.46266565e-01 4.60926890e-01 -1.05753422e+00
5.98638467e-02 6.25064038e-03 -2.44964838e-01 -5.76850593e-01
6.61011517e-01 2.24156693e-01 -2.12655991e-01 2.18885124e-01
-2.27997676e-01 4.79224361e-02 9.42708626e-02 4.37664360e-01
9.06164229e-01 -2.17686683e-01 -2.80425668e-01 -2.18414679e-01
4.80756283e-01 5.62576801e-02 4.79234189e-01 9.48183954e-01
-2.42365181e-01 9.10076201e-01 5.14715195e-01 -4.47068214e-01
-1.30800915e+00 -1.17104542e+00 -1.30673259e-01 1.06005859e+00
3.52592975e-01 -7.04854488e-01 -9.46080208e-01 -7.27145374e-01
4.99652028e-02 3.15930635e-01 -7.05129385e-01 -7.96428099e-02
-5.96586227e-01 -3.81755620e-01 8.79835248e-01 4.44870561e-01
8.81356716e-01 -8.73006642e-01 -2.14912862e-01 -2.64313728e-01
-3.82080466e-01 -1.43924749e+00 -9.00297165e-01 -2.34958217e-01
-3.54641110e-01 -1.11122096e+00 -6.45694673e-01 -1.00947714e+00
9.77521122e-01 3.65427524e-01 1.08081257e+00 4.08862472e-01
-3.76185268e-01 3.77571471e-02 -5.81329584e-01 -3.95459056e-01
-4.08697784e-01 -6.68124035e-02 -2.91379035e-01 6.39887825e-02
2.22999185e-01 -2.77981430e-01 -6.47532701e-01 4.49016154e-01
-1.31045556e+00 1.50421202e-01 5.77491343e-01 9.34159100e-01
6.09485269e-01 1.59116641e-01 2.12320089e-01 -1.03296518e+00
2.85041660e-01 -3.01972061e-01 -6.15438640e-01 3.31188530e-01
-4.43774223e-01 -3.21241766e-01 9.52692449e-01 -8.39226484e-01
-7.56212115e-01 2.90105730e-01 -8.39091986e-02 -6.98669434e-01
-5.43024112e-03 2.84532487e-01 -1.61964476e-01 -2.38510892e-01
6.47174776e-01 2.21283570e-01 -5.18197194e-02 -3.69751185e-01
3.37400526e-01 6.22869432e-01 8.07175934e-01 -4.13755447e-01
1.15161562e+00 5.89010298e-01 -4.55179065e-01 -7.93726683e-01
-7.40010619e-01 -3.19598675e-01 -2.42537796e-01 -1.65544525e-01
8.67403030e-01 -1.02836001e+00 -3.48111182e-01 5.82684577e-01
-9.08009052e-01 -2.42476881e-01 -6.64097536e-03 2.12221995e-01
-1.72311634e-01 5.25842071e-01 -7.36303151e-01 -3.18612009e-01
-3.65200400e-01 -1.22531104e+00 8.93028677e-01 7.50658917e-04
1.02734402e-01 -5.63180864e-01 -3.55174363e-01 3.28468502e-01
5.82925320e-01 3.56460243e-01 8.86159241e-01 -5.19926131e-01
-6.39961243e-01 -5.30228317e-01 -4.70461756e-01 5.35828531e-01
-2.87060887e-01 1.71166569e-01 -9.88015056e-01 -4.41333622e-01
-7.33209476e-02 -4.10040617e-01 8.77624273e-01 7.62786642e-02
1.44207263e+00 -5.44885218e-01 -1.67287678e-01 8.32995713e-01
1.71859896e+00 -2.86297370e-02 1.06263721e+00 3.18432719e-01
6.78646982e-01 3.32036823e-01 5.43927431e-01 9.93971005e-02
-1.05006903e-01 5.40043592e-01 5.29863238e-01 -3.44804257e-01
-1.74637571e-01 -6.16680384e-01 5.72990835e-01 4.82558817e-01
4.31265384e-01 -4.48994040e-01 -7.52772331e-01 5.88375747e-01
-1.61636424e+00 -1.12163925e+00 -3.85914035e-02 2.09756541e+00
8.18068564e-01 3.66247386e-01 -6.87368214e-02 -6.38485476e-02
8.92207980e-01 3.02738279e-01 -4.97910619e-01 -2.22274199e-01
-4.05598134e-01 2.24145174e-01 8.35447967e-01 3.10618222e-01
-1.36547101e+00 1.13950074e+00 6.74681091e+00 1.18482482e+00
-1.33646917e+00 2.71697668e-03 7.74966002e-01 -1.00711897e-01
-1.53244674e-01 -4.61380295e-02 -6.83256686e-01 7.48182058e-01
5.66709220e-01 -5.82161546e-02 4.37360406e-01 8.93723369e-01
-1.06674284e-01 -5.60931750e-02 -8.34890068e-01 1.05630362e+00
3.88465554e-01 -1.33506000e+00 2.82525092e-01 1.58027969e-02
8.98790658e-01 7.34163746e-02 3.98026347e-01 3.50061893e-01
1.80619881e-01 -1.22886717e+00 1.05202448e+00 1.83192164e-01
1.15350962e+00 -6.56142533e-01 6.92926168e-01 8.98884162e-02
-1.06414759e+00 -2.01782975e-02 -4.13846910e-01 3.62363368e-01
-1.36318013e-01 6.40697896e-01 -3.21088135e-01 1.69000298e-01
9.73916173e-01 6.15603745e-01 -1.08615041e+00 1.08021808e+00
-9.89840627e-02 6.64453447e-01 -3.72068316e-01 2.30616108e-01
2.04322055e-01 2.77445223e-02 5.17322361e-01 1.34801817e+00
9.01764557e-02 -2.88489163e-01 2.32226253e-01 9.66727376e-01
-5.64376116e-01 2.40416855e-01 -6.37426972e-01 -6.95176050e-02
4.95811075e-01 9.83355999e-01 -7.09630907e-01 -4.29011792e-01
-6.81872904e-01 1.26367629e+00 -1.86211932e-02 4.74938750e-01
-1.16041577e+00 -6.22462034e-01 5.44788301e-01 9.89848599e-02
4.45397824e-01 -5.63199446e-02 -2.73063451e-01 -1.18562901e+00
3.39992136e-01 -1.27617753e+00 2.28844479e-01 -7.36661494e-01
-1.25286341e+00 5.88959694e-01 -2.85497606e-01 -1.33967710e+00
2.22335592e-01 -4.72651601e-01 -4.94231641e-01 5.97397983e-01
-1.31885982e+00 -1.31289816e+00 -3.70698482e-01 8.01207304e-01
5.42184174e-01 -6.56015351e-02 6.58234000e-01 6.27027273e-01
-4.79181439e-01 1.11489463e+00 2.53187835e-01 6.92911565e-01
1.09625804e+00 -9.83557284e-01 4.94123906e-01 1.08980083e+00
8.80615860e-02 4.81018573e-01 6.48239613e-01 -6.23727560e-01
-1.41990781e+00 -1.22281885e+00 5.77651203e-01 -3.57898414e-01
6.60070777e-01 -6.84437275e-01 -9.81899559e-01 6.25843942e-01
3.02909940e-01 7.65589401e-02 4.60490793e-01 -4.51116681e-01
-9.30275857e-01 -1.69688299e-01 -1.08812487e+00 8.26643646e-01
8.08916032e-01 -8.46178412e-01 -2.98447758e-01 3.96749705e-01
7.93462217e-01 -3.86159122e-01 -7.34896064e-01 2.66417861e-01
6.77469313e-01 -8.89901519e-01 1.09638476e+00 -4.66468424e-01
7.08752871e-01 -1.72828585e-01 -2.44661123e-01 -8.06993484e-01
-3.57560813e-02 -6.30158782e-01 1.89887211e-01 1.29271877e+00
2.90695816e-01 -4.45270181e-01 7.55826235e-01 4.33258444e-01
1.55579463e-01 -4.68794882e-01 -5.95352948e-01 -9.40118968e-01
3.14436853e-01 -4.44637895e-01 4.39581990e-01 9.66719329e-01
-2.32956633e-01 -5.35291582e-02 -6.11495376e-01 1.39614135e-01
5.45802951e-01 -1.30816817e-01 8.55280101e-01 -6.80629194e-01
-5.16877696e-02 -4.78871822e-01 -3.90059859e-01 -9.76586521e-01
1.03941448e-01 -8.13114643e-01 3.62080894e-02 -8.85999084e-01
4.01574612e-01 -4.06019032e-01 -2.29827359e-01 4.27632123e-01
-1.13829076e-01 7.71460652e-01 5.41210055e-01 4.75214958e-01
-7.26555169e-01 1.40173838e-01 1.13787889e+00 -3.96081954e-01
1.78131819e-01 -3.74953210e-01 -6.12025619e-01 6.29153788e-01
8.48678172e-01 -5.91319501e-01 -2.81204760e-01 -5.82050443e-01
2.86820889e-01 -3.80489677e-01 5.13378799e-01 -1.01060092e+00
2.54682243e-01 -8.33669603e-02 5.02390623e-01 -6.86288059e-01
1.35209709e-01 -1.03535151e+00 1.03465587e-01 3.88764054e-01
-5.48932254e-01 1.46973759e-01 3.22968751e-01 5.93909740e-01
-3.84073496e-01 -1.28050044e-01 1.03905582e+00 -7.28155226e-02
-6.90952122e-01 2.89930016e-01 -4.00436282e-01 2.51985878e-01
1.04163885e+00 -1.10530853e-01 -6.64526701e-01 -3.44889760e-01
-3.34473044e-01 2.22442821e-01 9.19372797e-01 6.58944368e-01
6.39564037e-01 -1.12088883e+00 -5.52111149e-01 4.49057877e-01
2.53632188e-01 -3.15971136e-01 2.52034783e-01 6.35551631e-01
-8.71836662e-01 1.86436191e-01 -2.42114350e-01 -5.50733209e-01
-1.43057585e+00 9.06299412e-01 2.46461838e-01 -1.31514117e-01
-8.61893356e-01 6.02178514e-01 2.96946496e-01 -1.47009222e-02
6.03481770e-01 -6.11543730e-02 2.10506856e-01 -1.95767924e-01
7.99779713e-01 2.74373088e-02 -1.05317384e-01 -7.87260711e-01
-1.13658145e-01 5.86782038e-01 -4.15008008e-01 -9.92628857e-02
1.04590416e+00 -1.15962001e-02 -1.00414515e-01 -9.22104865e-02
1.50470448e+00 3.79920393e-01 -1.12123930e+00 -2.28440881e-01
-2.21481964e-01 -7.79352844e-01 -7.72230923e-02 -8.27419221e-01
-1.38103378e+00 8.15355361e-01 7.60191619e-01 1.40732750e-01
1.14576054e+00 -2.64934480e-01 1.14494693e+00 2.55726755e-01
1.69576645e-01 -1.29559577e+00 2.23036826e-01 3.21799368e-01
8.53998542e-01 -1.40148675e+00 7.62387589e-02 -5.63852370e-01
-6.17218494e-01 8.16188276e-01 5.91902137e-01 -3.44741225e-01
7.87095964e-01 4.21165466e-01 1.66224658e-01 -1.05400205e-01
-6.30500615e-01 1.94616467e-02 3.36635619e-01 5.86233377e-01
3.27606983e-02 -6.41907677e-02 -9.79618728e-02 3.32868516e-01
-2.71796763e-01 -2.54923791e-01 3.27379137e-01 8.86231720e-01
-2.97550410e-01 -9.86622632e-01 -5.51379919e-01 4.46535170e-01
-9.84789312e-01 -2.45158300e-01 -6.65850222e-01 8.43230188e-01
2.09932134e-01 8.12357068e-01 2.63563003e-02 -6.73274577e-01
1.56526104e-01 -6.92492574e-02 9.57666114e-02 -3.30659389e-01
-7.99254358e-01 1.02390751e-01 -2.42222354e-01 -6.46086633e-01
-8.11201036e-02 -3.00726086e-01 -9.16442096e-01 -6.25978112e-01
-3.25161219e-01 -1.13183066e-01 7.20105350e-01 5.33614933e-01
3.53238076e-01 1.59771055e-01 5.90412855e-01 -5.54113507e-01
-4.67583597e-01 -6.15740180e-01 -4.10078347e-01 1.00166440e+00
2.56771028e-01 -2.60705888e-01 -4.76156443e-01 5.01884818e-01]
|
[11.471902847290039, 0.93805330991745]
|
6e3064a7-5586-4826-ab28-be30448f73b5
|
multi-site-clinical-federated-learning-using
|
2306.16367
| null |
https://arxiv.org/abs/2306.16367v1
|
https://arxiv.org/pdf/2306.16367v1.pdf
|
Multi-Site Clinical Federated Learning using Recursive and Attentive Models and NVFlare
|
The prodigious growth of digital health data has precipitated a mounting interest in harnessing machine learning methodologies, such as natural language processing (NLP), to scrutinize medical records, clinical notes, and other text-based health information. Although NLP techniques have exhibited substantial potential in augmenting patient care and informing clinical decision-making, data privacy and adherence to regulations persist as critical concerns. Federated learning (FL) emerges as a viable solution, empowering multiple organizations to train machine learning models collaboratively without disseminating raw data. This paper proffers a pragmatic approach to medical NLP by amalgamating FL, NLP models, and the NVFlare framework, developed by NVIDIA. We introduce two exemplary NLP models, the Long-Short Term Memory (LSTM)-based model and Bidirectional Encoder Representations from Transformers (BERT), which have demonstrated exceptional performance in comprehending context and semantics within medical data. This paper encompasses the development of an integrated framework that addresses data privacy and regulatory compliance challenges while maintaining elevated accuracy and performance, incorporating BERT pretraining, and comprehensively substantiating the efficacy of the proposed approach.
|
['Joongheon Kim', 'Samuel Kim', 'Won Joon Yun']
|
2023-06-28
| null | null | null | null |
['decision-making']
|
['reasoning']
|
[ 2.86919087e-01 4.76573467e-01 -2.68949538e-01 -5.48653603e-01
-9.02161300e-01 -3.94112229e-01 2.43149742e-01 8.38453352e-01
-5.18429458e-01 7.97289491e-01 6.90091610e-01 -7.20369518e-01
-4.11246449e-01 -6.28614426e-01 -4.87392545e-01 -2.85742372e-01
-2.41643786e-01 3.90006453e-01 -7.22881377e-01 1.61410600e-01
-1.81581810e-01 4.36840326e-01 -1.03045321e+00 8.62845778e-01
8.65113020e-01 1.09406579e+00 -2.77125686e-01 5.17783225e-01
-4.21791285e-01 1.40639687e+00 -4.69663084e-01 -8.02360952e-01
2.50821076e-02 9.85908285e-02 -1.02221155e+00 -4.48713958e-01
9.14946645e-02 -3.96868259e-01 -1.14021286e-01 9.21323895e-01
6.26889765e-01 -3.73224556e-01 -3.58634554e-02 -8.95740807e-01
-1.21050346e+00 4.82065588e-01 7.48285204e-02 -4.59029302e-02
5.34284234e-01 1.87578931e-01 7.92754591e-01 -4.10963118e-01
8.31354797e-01 8.29810202e-01 1.03274691e+00 6.35866880e-01
-8.52488518e-01 -4.38464314e-01 -1.45053670e-01 -1.52999327e-01
-1.09360647e+00 -3.96443158e-01 3.27221870e-01 -4.35256004e-01
1.37490582e+00 5.66376805e-01 6.02005482e-01 1.39577591e+00
8.58591497e-01 8.12142789e-01 9.65061665e-01 -4.01238859e-01
2.09855869e-01 5.20136237e-01 3.42648178e-01 6.10073388e-01
4.56867903e-01 2.41716519e-01 -7.31713831e-01 -8.81607950e-01
3.58338326e-01 3.95233274e-01 -4.17568624e-01 -1.19198032e-01
-1.19834387e+00 6.60189450e-01 3.34448725e-01 4.00106251e-01
-7.28939176e-01 -3.14281106e-01 9.60362017e-01 1.84012875e-01
5.31254888e-01 5.54855287e-01 -6.71599030e-01 -2.12874293e-01
-8.00232351e-01 2.12054551e-01 9.96609330e-01 8.35240185e-01
1.33574501e-01 -3.11401755e-01 -2.36701965e-01 4.57055867e-01
4.20938671e-01 3.73376250e-01 7.48043537e-01 -6.89714372e-01
3.75218451e-01 8.07174146e-01 1.17264807e-01 -1.13038886e+00
-4.24321026e-01 -2.90665299e-01 -1.01692557e+00 -3.45081955e-01
-1.38027519e-01 -3.43617767e-01 -6.43139124e-01 1.54520607e+00
2.36536443e-01 -1.09397799e-01 5.35638988e-01 5.11016250e-01
8.27214956e-01 3.07873785e-01 5.22653401e-01 -1.98144704e-01
1.45448911e+00 -3.66967678e-01 -1.37603033e+00 1.61842853e-01
7.67445028e-01 -3.61538172e-01 6.15103960e-01 5.29799581e-01
-9.98650551e-01 -9.33381617e-02 -8.24655116e-01 -2.81028003e-01
-6.12147033e-01 -3.28870207e-01 6.38182759e-01 7.78877735e-01
-1.01856363e+00 7.05415070e-01 -1.08052278e+00 -3.24084938e-01
9.28913414e-01 4.30209875e-01 -6.17414951e-01 7.72076100e-03
-1.35935044e+00 9.36431110e-01 3.13850850e-01 2.66985357e-01
-5.05619466e-01 -9.40212786e-01 -9.37007129e-01 4.98231091e-02
1.14760257e-01 -1.06714833e+00 1.21991694e+00 -5.99512458e-01
-1.04083717e+00 8.96013558e-01 1.71668708e-01 -1.03127420e+00
5.35716474e-01 -5.18642783e-01 -8.40297818e-01 6.29234985e-02
-6.26498088e-02 4.74965096e-01 3.07547271e-01 -6.18037403e-01
-4.94603395e-01 -6.50827169e-01 -5.05145848e-01 -8.09384510e-02
-6.48278236e-01 6.69187531e-02 2.34099120e-01 -3.89896095e-01
-4.95129973e-01 -3.14100385e-01 -4.84066963e-01 1.70298353e-01
-6.55916274e-01 -1.17568476e-02 6.59215510e-01 -9.63081419e-01
1.42090702e+00 -2.14359713e+00 -5.23843050e-01 8.65748003e-02
5.46968877e-01 6.00507259e-01 2.43641227e-01 6.32806897e-01
1.27776563e-01 3.66938710e-01 -2.42764547e-01 -2.99779803e-01
-7.66758323e-02 4.24782336e-01 -4.26973403e-01 3.19075376e-01
2.49587253e-01 1.21499276e+00 -7.85011828e-01 -5.23547769e-01
2.55156577e-01 9.18063819e-01 -4.34041560e-01 2.69539297e-01
-2.62913823e-01 4.22479868e-01 -5.18141270e-01 7.80799448e-01
3.88268173e-01 -7.18706071e-01 5.30761063e-01 -1.70750380e-01
-9.18275714e-02 3.59318584e-01 -6.41079962e-01 1.75348425e+00
-2.19240993e-01 2.44977832e-01 2.97171533e-01 -5.77877223e-01
9.23482597e-01 8.11911285e-01 7.87309408e-01 -8.15499723e-01
1.58261657e-01 7.83785284e-02 -3.72401267e-01 -1.28965449e+00
3.63543004e-01 -2.03252941e-01 -2.70657931e-02 2.22091705e-01
-5.56120975e-03 6.38813376e-01 -6.20244324e-01 1.27582729e-01
1.27482700e+00 5.73567040e-02 6.99160099e-01 -2.27283478e-01
4.76155996e-01 2.50672698e-01 5.56851268e-01 5.52445412e-01
-5.59696853e-01 1.36406094e-01 1.96084470e-01 -8.26551318e-01
-6.40405118e-01 -7.83018172e-01 -4.01930839e-01 6.51823103e-01
-5.34863353e-01 -3.67840558e-01 -4.97603059e-01 -6.18298233e-01
2.41238192e-01 6.77246988e-01 -5.64020395e-01 -1.16714373e-01
-3.70534718e-01 -6.50748372e-01 9.46014524e-01 4.91070598e-01
3.52160722e-01 -1.18407834e+00 -1.26489758e+00 4.52496380e-01
-2.35637337e-01 -1.00453758e+00 -2.02612489e-01 3.10005963e-01
-8.57771754e-01 -1.28405380e+00 -2.53899008e-01 -6.15769863e-01
3.45574677e-01 -4.84523416e-01 1.04401803e+00 -3.89316559e-01
-6.92920268e-01 4.40793365e-01 -9.60635468e-02 -7.37948358e-01
-5.90838253e-01 -2.28460327e-01 -1.06737413e-01 -1.30751386e-01
8.27630460e-01 -1.24754265e-01 -6.98366404e-01 -3.97218436e-01
-1.03051484e+00 -7.09943548e-02 5.50725281e-01 8.92188728e-01
6.00082695e-01 -4.98487562e-01 8.94882977e-01 -1.57226670e+00
9.05101895e-01 -7.99214542e-01 -8.35010037e-02 5.44064999e-01
-9.97009277e-01 -1.67209327e-01 7.82634914e-01 6.92456290e-02
-1.08657706e+00 -9.78620606e-04 -5.88911414e-01 -4.54894125e-01
-3.53080153e-01 8.37925792e-01 -9.99291390e-02 2.63612032e-01
7.10763156e-01 1.29592746e-01 5.49149871e-01 -4.95761126e-01
3.97907943e-01 1.24138844e+00 6.63303435e-01 -1.16797358e-01
-1.31982684e-01 5.62393725e-01 -6.21215641e-01 -5.43012440e-01
-6.64806008e-01 -3.98101419e-01 -2.05938116e-01 3.25147450e-01
9.61933553e-01 -9.50570583e-01 -1.06618631e+00 -9.99966171e-03
-1.09478378e+00 3.18157822e-01 -6.10134304e-01 3.51091355e-01
-3.30637097e-01 2.61733860e-01 -8.67343783e-01 -7.93011963e-01
-1.09165108e+00 -9.01569724e-01 8.62992764e-01 -2.02797726e-01
-7.61899710e-01 -1.05657005e+00 3.44793022e-01 4.73037928e-01
6.56424880e-01 7.24950492e-01 1.27857363e+00 -1.28345871e+00
-2.85176665e-01 -5.09288907e-01 -4.39284444e-02 3.12951028e-01
4.15417790e-01 -3.53429645e-01 -1.22992468e+00 -3.33394110e-01
4.59177256e-01 -4.20735389e-01 1.79073304e-01 2.37341821e-01
1.22548509e+00 -8.33734512e-01 -4.73061919e-01 6.78558469e-01
1.43793821e+00 3.52609813e-01 3.57878327e-01 2.05259830e-01
4.52274352e-01 6.31681800e-01 1.47979468e-01 6.59116030e-01
4.71102923e-01 -1.15778737e-01 4.15479928e-01 -9.03628394e-02
2.83187598e-01 -3.74063224e-01 -9.00897235e-02 7.73552656e-01
5.55977702e-01 -1.09408848e-01 -1.18542922e+00 4.07954037e-01
-1.78749132e+00 -6.23563886e-01 5.75986169e-02 1.96959639e+00
9.57230985e-01 -3.40709120e-01 -3.90653700e-01 -1.99854180e-01
3.64046752e-01 -6.12029955e-02 -7.95173347e-01 -7.11937666e-01
2.66983900e-02 2.15063274e-01 4.59653884e-01 1.31510377e-01
-1.00946701e+00 2.27366388e-01 6.84070253e+00 6.30394369e-02
-1.26656890e+00 2.41303861e-01 8.66599739e-01 -9.94284600e-02
-4.41549867e-01 -7.01105416e-01 -4.52743977e-01 3.99518073e-01
1.47298682e+00 -3.10518980e-01 -5.14422879e-02 9.69011724e-01
3.32638770e-01 4.29933816e-01 -1.35279799e+00 1.09122169e+00
-4.26524226e-03 -2.01197577e+00 2.97942132e-01 1.44346014e-01
3.38559866e-01 3.10311407e-01 2.83351570e-01 1.16905190e-01
3.53178471e-01 -1.32027280e+00 1.64421022e-01 5.81795156e-01
8.35420191e-01 -4.69664454e-01 1.01675630e+00 2.97266930e-01
-4.99761969e-01 -4.13097292e-01 3.66984606e-02 1.57506347e-01
1.33536607e-01 4.31560814e-01 -1.10952914e+00 8.71237874e-01
8.20174754e-01 7.87363052e-01 -2.44357675e-01 6.74948275e-01
4.48407620e-01 2.07113639e-01 7.25321546e-02 1.40843064e-01
1.41086563e-01 1.25359997e-01 2.67733753e-01 1.25238049e+00
3.01077757e-02 4.59623747e-02 5.49633764e-02 8.45943213e-01
-1.37454972e-01 3.20074409e-01 -9.64806020e-01 -4.87167567e-01
4.60559160e-01 1.08831692e+00 -5.57527132e-02 -3.39367986e-01
-6.43652081e-01 6.36464477e-01 2.17035502e-01 1.10468321e-01
-6.25194788e-01 3.12820054e-03 6.89334273e-01 1.52456000e-01
-1.01347491e-01 3.56851161e-01 -5.16700506e-01 -8.94391596e-01
4.17038836e-02 -1.37275970e+00 7.87212908e-01 -3.31962705e-01
-1.57500017e+00 1.04246509e+00 -3.45885128e-01 -1.13401389e+00
-4.76071417e-01 -5.14487922e-01 1.89394262e-02 9.02774334e-01
-1.55564582e+00 -1.27369726e+00 -1.99966487e-02 7.12435842e-01
1.49772773e-02 -1.60603195e-01 1.58157718e+00 5.23328424e-01
-5.74364007e-01 5.30411482e-01 9.77811068e-02 2.54463136e-01
6.06382787e-01 -9.20407414e-01 1.66413993e-01 4.36004579e-01
-1.23192273e-01 1.14535105e+00 2.82471806e-01 -8.78874063e-01
-1.69828320e+00 -1.46618116e+00 1.20879138e+00 -4.38858807e-01
3.52185696e-01 -3.19378853e-01 -1.02252507e+00 9.87300396e-01
4.49745387e-01 1.06201485e-01 1.58182919e+00 4.55221348e-02
-2.13236690e-01 9.63556394e-03 -1.78383744e+00 2.02864334e-01
5.06940186e-01 -8.43351185e-01 -7.84568071e-01 4.37866926e-01
1.00184441e+00 -3.79700482e-01 -1.28743076e+00 3.89106125e-01
5.04623234e-01 -8.80833268e-01 7.22551644e-01 -1.05159414e+00
3.83924574e-01 2.37272799e-01 -1.69971168e-01 -7.92089105e-01
-1.75372157e-02 -9.74168360e-01 -3.65709484e-01 9.63683784e-01
3.87761325e-01 -1.01275802e+00 1.01468146e+00 1.24488890e+00
-7.30130598e-02 -1.16287899e+00 -9.16579425e-01 -2.48432860e-01
-6.04263693e-02 -5.26481390e-01 8.64875495e-01 1.29391527e+00
5.67090511e-01 -3.16251665e-02 -2.90172428e-01 2.56800234e-01
3.47202539e-01 -1.46850973e-01 1.99851424e-01 -1.00634241e+00
-2.53260255e-01 9.32657272e-02 -3.32248837e-01 -4.13898200e-01
-2.58871526e-01 -1.09854388e+00 -3.88411283e-01 -1.58584023e+00
7.32750818e-02 -3.94372851e-01 -5.49813986e-01 8.45658064e-01
1.70714319e-01 -2.75615096e-01 -6.59692213e-02 1.65942177e-01
-4.43035603e-01 1.96198449e-01 9.04770136e-01 -1.52234495e-01
-2.00647503e-01 -3.07406247e-01 -1.10229540e+00 5.42416692e-01
6.59822643e-01 -3.49366039e-01 -3.49741548e-01 -7.30503678e-01
2.94707716e-01 3.52953672e-01 3.99621934e-01 -6.80914581e-01
5.78288198e-01 3.06792259e-01 4.86898541e-01 -4.60551322e-01
-5.49857505e-02 -1.26111829e+00 4.26715434e-01 8.18325520e-01
-8.11403692e-01 2.25565299e-01 4.61125463e-01 6.40930116e-01
-4.77335244e-01 3.36128891e-01 3.34169924e-01 -3.11594516e-01
-4.11819935e-01 3.29480648e-01 -3.75365764e-01 -1.91124361e-02
1.07161450e+00 -1.53613627e-01 -5.22259414e-01 1.11650780e-01
-1.06781256e+00 5.13682961e-01 6.89458922e-02 3.58558804e-01
8.71258438e-01 -8.58111143e-01 -5.56069672e-01 7.82935798e-01
1.25427604e-01 4.35904227e-02 5.52158833e-01 7.09904730e-01
-6.31071746e-01 8.60837340e-01 -1.91535145e-01 -4.42178339e-01
-1.13238263e+00 8.69026363e-01 2.93119699e-01 -4.46464479e-01
-1.01249754e+00 7.04303145e-01 -3.44887614e-01 -6.17096663e-01
4.88082826e-01 -5.29400587e-01 -4.19864953e-02 -1.22533768e-01
8.36935639e-01 1.78747118e-01 4.34999377e-01 -9.70763043e-02
-5.88813066e-01 -3.40449154e-01 -5.33291936e-01 3.64181846e-01
1.47549951e+00 -6.10744841e-02 -2.74475038e-01 3.17560762e-01
1.31569517e+00 -3.83389592e-01 -6.34782195e-01 -1.00037016e-01
3.92230093e-01 -1.98171243e-01 1.06570525e-02 -1.43027246e+00
-6.23442531e-01 6.88305318e-01 7.75186539e-01 4.15723175e-02
1.09198463e+00 -4.03420597e-01 1.11201119e+00 3.74785274e-01
4.11714464e-01 -7.80722678e-01 -3.75661135e-01 8.77095014e-02
5.21316051e-01 -1.08717430e+00 -1.74016237e-01 -2.02716649e-01
-7.76715636e-01 9.79117870e-01 2.09003836e-01 2.41152599e-01
8.09142292e-01 8.31796348e-01 5.37531793e-01 -4.45899487e-01
-9.61915612e-01 6.12964749e-01 -1.54866338e-01 5.99880815e-01
6.88021958e-01 1.09036863e-01 8.02595243e-02 8.96162450e-01
-4.55889329e-02 8.63096952e-01 1.09159403e-01 1.20406926e+00
1.57927290e-01 -1.10466433e+00 -2.67441213e-01 8.74442458e-01
-1.07579076e+00 -3.07854325e-01 -1.78264409e-01 5.17895937e-01
2.66171694e-01 7.64211178e-01 1.35789171e-01 -1.33885130e-01
3.57150018e-01 4.71582383e-01 -1.17952719e-01 -5.92528462e-01
-1.28522098e+00 -2.68547267e-01 1.35132343e-01 -8.62209797e-01
-2.66258717e-01 -4.60005671e-01 -1.10888422e+00 -6.11574166e-02
1.20138265e-01 2.37760305e-01 6.20039046e-01 6.05954885e-01
1.16848290e+00 6.75372899e-01 1.31881550e-01 1.70779854e-01
-9.33830321e-01 -5.07945597e-01 -3.80210996e-01 4.36298907e-01
7.07980812e-01 3.00980061e-01 2.63353318e-01 3.05526704e-02]
|
[6.4742841720581055, 6.6603569984436035]
|
0a6cad48-4e5d-4399-baad-5615193a669d
|
controlling-learned-effects-to-reduce
|
2305.16863
| null |
https://arxiv.org/abs/2305.16863v2
|
https://arxiv.org/pdf/2305.16863v2.pdf
|
Controlling Learned Effects to Reduce Spurious Correlations in Text Classifiers
|
To address the problem of NLP classifiers learning spurious correlations between training features and target labels, a common approach is to make the model's predictions invariant to these features. However, this can be counter-productive when the features have a non-zero causal effect on the target label and thus are important for prediction. Therefore, using methods from the causal inference literature, we propose an algorithm to regularize the learnt effect of the features on the model's prediction to the estimated effect of feature on label. This results in an automated augmentation method that leverages the estimated effect of a feature to appropriately change the labels for new augmented inputs. On toxicity and IMDB review datasets, the proposed algorithm minimises spurious correlations and improves the minority group (i.e., samples breaking spurious correlations) accuracy, while also improving the total accuracy compared to standard training.
|
['Amit Sharma', 'Parikshit Bansal']
|
2023-05-26
| null | null | null | null |
['causal-inference', 'causal-inference']
|
['knowledge-base', 'miscellaneous']
|
[ 8.62815797e-01 4.07399595e-01 -7.24777460e-01 -5.33601046e-01
-6.14897251e-01 -6.66396558e-01 7.60766566e-01 6.01918757e-01
-3.87127846e-01 1.20598400e+00 3.08765650e-01 -4.96811450e-01
-4.24102306e-01 -8.52345109e-01 -1.04188359e+00 -8.86010170e-01
2.28332020e-02 1.64339051e-01 8.18937644e-03 4.29935932e-01
4.45925891e-01 3.32356155e-01 -1.52661133e+00 4.89015251e-01
8.42132747e-01 7.05890715e-01 -3.13127577e-01 2.35774294e-01
6.04390129e-02 1.03756070e+00 -4.65927422e-01 -4.88101214e-01
5.52862100e-02 -5.82780302e-01 -8.30936372e-01 -1.32750496e-01
3.28450829e-01 3.01962942e-02 1.88258946e-01 9.29664552e-01
1.17276281e-01 -3.88591960e-02 1.18632364e+00 -1.32048345e+00
-3.98164988e-01 8.27730179e-01 -4.22574311e-01 5.90335317e-02
2.09045514e-01 1.18067548e-01 1.36604023e+00 -6.67692244e-01
4.91333932e-01 1.16682887e+00 9.55189645e-01 1.81771815e-01
-1.60468900e+00 -1.08435357e+00 3.62589180e-01 1.26409516e-01
-1.16674888e+00 -4.25005168e-01 4.00974482e-01 -6.39754295e-01
7.64951348e-01 3.13106179e-01 3.64093482e-01 1.15617168e+00
3.87616217e-01 1.37988389e-01 1.23439968e+00 -5.97637951e-01
4.01370227e-01 3.64076555e-01 1.23903595e-01 5.40383816e-01
3.78701717e-01 5.37319005e-01 -5.65013468e-01 -3.94393176e-01
2.44100958e-01 -3.70100476e-02 -1.64726585e-01 -1.31851152e-01
-8.66077662e-01 1.10173738e+00 3.41037363e-01 8.16239193e-02
-4.83601272e-01 2.41318181e-01 1.38533235e-01 8.46280605e-02
6.96678460e-01 7.82046974e-01 -1.03651130e+00 2.20672086e-01
-7.46198833e-01 3.71061087e-01 5.47636688e-01 3.23749304e-01
8.16279113e-01 -5.92826247e-01 -4.32793379e-01 7.63584554e-01
3.60276848e-01 1.03241995e-01 2.52702206e-01 -5.88964939e-01
2.40024135e-01 1.05224049e+00 9.50841010e-02 -8.56117666e-01
-4.84259337e-01 -4.18793112e-01 -4.39056486e-01 1.50733486e-01
5.86537421e-01 -1.85125440e-01 -1.07760215e+00 1.98658907e+00
5.55854082e-01 4.29785728e-01 5.64034516e-03 4.93229479e-01
5.10236263e-01 3.17512810e-01 7.65959144e-01 -5.37819922e-01
1.02808261e+00 -2.74672240e-01 -6.86444879e-01 -2.37300694e-01
1.15385878e+00 -5.97546041e-01 8.61829221e-01 2.69936889e-01
-3.97746176e-01 -1.83951274e-01 -1.01401353e+00 2.56061822e-01
-3.66128206e-01 -6.03636429e-02 9.92284417e-01 6.52920008e-01
-2.27198750e-01 1.04060578e+00 -4.14218813e-01 -3.23449485e-02
5.90222001e-01 7.02742279e-01 -3.27077985e-01 -1.21745408e-01
-1.39144123e+00 1.08863914e+00 6.77489102e-01 -1.82696790e-01
-5.30606091e-01 -1.35864568e+00 -3.75116497e-01 1.19278878e-01
6.21576667e-01 -5.26535928e-01 9.61172342e-01 -1.04665327e+00
-1.03531790e+00 4.41914052e-01 -5.12738377e-02 -2.92077750e-01
3.74657601e-01 -8.34943280e-02 -1.26205489e-01 -3.67475659e-01
2.24627614e-01 5.74055970e-01 8.13897669e-01 -1.11938131e+00
-8.02028894e-01 -3.42618436e-01 -1.00217938e-01 3.73664796e-02
-2.08274946e-01 -2.66877681e-01 2.31543317e-01 -3.69336665e-01
-1.85681768e-02 -9.92602706e-01 -3.30403090e-01 -1.85946807e-01
-5.17702222e-01 -3.00328821e-01 5.63627481e-01 -3.28368694e-01
1.03480232e+00 -1.96729183e+00 -2.01535329e-01 4.22605067e-01
7.22857565e-02 2.17761267e-02 2.82938294e-02 3.22503000e-01
-6.30659461e-01 3.24315310e-01 -4.62060332e-01 1.35031715e-01
-3.85692984e-01 3.45025033e-01 -3.80097926e-01 4.63716954e-01
6.19350195e-01 6.28681540e-01 -9.22407508e-01 -3.56839269e-01
2.49251872e-02 2.60508120e-01 -7.88561225e-01 4.61740941e-02
-3.28453243e-01 5.52262545e-01 -3.41481388e-01 1.99261919e-01
4.15009022e-01 -1.17973216e-01 4.72355098e-01 -6.13130704e-02
-1.26704395e-01 6.88309312e-01 -1.15676856e+00 7.21217275e-01
-4.84412313e-01 1.23798393e-01 -7.77954340e-01 -1.02019870e+00
8.21484029e-01 2.91375905e-01 4.87095177e-01 -3.09967995e-01
-8.04060698e-02 1.71310470e-01 3.21359009e-01 -3.44306260e-01
-1.12451918e-01 -6.45003498e-01 1.18140757e-01 3.14657360e-01
7.32334703e-02 2.21538529e-01 -1.36690274e-01 -4.49254513e-02
1.30119014e+00 -4.06643003e-02 7.13488936e-01 -1.57733291e-01
3.58503014e-01 -2.83872406e-03 9.28459704e-01 7.29236484e-01
4.73936200e-01 4.56985049e-02 9.68324780e-01 -2.64611155e-01
-6.91179276e-01 -7.25930989e-01 -4.94910181e-01 7.21898675e-01
-5.42661309e-01 -2.73828089e-01 -3.59675825e-01 -1.19561982e+00
3.13721687e-01 1.11419129e+00 -1.06329751e+00 -6.74709618e-01
-6.37991400e-03 -1.17456865e+00 4.23971891e-01 4.41304088e-01
3.07353549e-02 -8.53584051e-01 -2.24562570e-01 1.69580981e-01
1.61828883e-02 -7.47018635e-01 -7.39150345e-02 7.28856742e-01
-8.73002768e-01 -1.47483313e+00 1.67066634e-01 -2.27309987e-01
9.06569719e-01 -2.65880555e-01 8.16506922e-01 1.86954722e-01
-9.60398559e-03 -1.75894469e-01 -3.48387808e-01 -6.70440733e-01
-6.11429334e-01 -2.12459043e-01 -1.37682492e-02 7.38826320e-02
4.98127908e-01 -3.95068020e-01 -1.63456693e-01 4.63054255e-02
-7.85295010e-01 -7.36036077e-02 6.37195528e-01 1.10923207e+00
6.37134969e-01 2.89886117e-01 1.06430459e+00 -1.59273660e+00
4.09577370e-01 -7.22039342e-01 -5.11173546e-01 2.19705939e-01
-1.29311192e+00 3.45754296e-01 6.81949914e-01 -7.69458473e-01
-1.04575467e+00 3.97503376e-01 1.22345954e-01 -2.14576814e-02
-1.31900415e-01 8.40707362e-01 -3.35315704e-01 1.06413797e-01
8.04993987e-01 -3.42505366e-01 -3.20213467e-01 -4.45648283e-01
2.53316432e-01 4.39614922e-01 4.36784551e-02 -3.96389723e-01
5.70267141e-01 7.05527127e-05 5.78950405e-01 -3.35079908e-01
-1.35639453e+00 -2.95517534e-01 -7.42807388e-01 -5.85168181e-03
5.46815157e-01 -6.33206725e-01 -6.29842758e-01 7.81211406e-02
-9.51481283e-01 -2.55002558e-01 -4.77533281e-01 7.74107456e-01
-2.25212663e-01 -2.91132908e-02 5.33690415e-02 -7.37874269e-01
1.57722294e-01 -9.75682914e-01 6.17322922e-01 -7.27152685e-03
-6.43503726e-01 -1.13308501e+00 1.72658786e-01 3.68114188e-02
-2.16281056e-01 2.95056790e-01 1.55874848e+00 -1.12564754e+00
-1.41430795e-01 -3.83530110e-01 -1.06745653e-01 3.10104907e-01
3.57665569e-01 3.46442536e-02 -1.14504957e+00 1.58112586e-01
-3.70615087e-02 -1.41124859e-01 8.68630528e-01 5.03449917e-01
9.73635435e-01 -7.66710937e-01 -5.81048131e-01 1.13514803e-01
1.22318482e+00 2.30409771e-01 3.60947132e-01 5.86385354e-02
5.58481872e-01 7.37508833e-01 7.08413422e-01 3.77299875e-01
-1.34444267e-01 6.99495852e-01 2.29945242e-01 2.44940855e-02
4.69871052e-02 -5.54728687e-01 2.37148046e-01 -9.42133069e-02
2.35212371e-01 -1.86119825e-02 -7.19505012e-01 3.57311070e-01
-1.75420439e+00 -7.00911701e-01 -6.65936768e-01 2.45302701e+00
1.14244270e+00 2.80259192e-01 -5.10448031e-02 2.61750281e-01
5.91097653e-01 -3.45010966e-01 -6.07420385e-01 -4.67562109e-01
1.10754006e-01 3.50783348e-01 6.58970475e-01 4.48323309e-01
-8.47769260e-01 7.50293970e-01 6.71846437e+00 5.52207530e-01
-8.33648145e-01 1.16457865e-01 6.19952440e-01 5.54890521e-02
-3.93998146e-01 4.02856231e-01 -8.84372056e-01 4.18243408e-01
1.20379484e+00 -1.10768855e-01 2.18443945e-01 5.19456863e-01
3.99317592e-01 -3.16741347e-01 -1.53996432e+00 1.06141515e-01
-2.35446692e-01 -1.20383608e+00 1.27124935e-01 3.99010986e-01
8.31252158e-01 -4.94601816e-01 -6.10656850e-02 2.76721567e-01
4.40996587e-01 -1.17517602e+00 3.41228604e-01 5.38228989e-01
7.26467073e-01 -8.99150252e-01 9.64969575e-01 4.40376997e-01
-2.76416570e-01 -3.37568641e-01 -1.67593658e-01 -4.54248011e-01
-3.81940126e-01 1.04411006e+00 -1.52823544e+00 1.96349531e-01
3.13789934e-01 6.95208669e-01 -6.73006058e-01 8.10124040e-01
-6.89258814e-01 1.05869901e+00 -9.75757465e-02 -5.28511293e-02
-2.09176987e-02 -4.22607772e-02 1.95008993e-01 9.27674592e-01
1.38955498e-02 1.41058326e-01 -7.55319297e-02 9.21969891e-01
-1.63869679e-01 2.70873487e-01 -7.82197356e-01 -1.23799480e-01
4.51117158e-01 1.00525796e+00 -6.38761163e-01 -2.71046460e-01
-4.41569000e-01 4.08361197e-01 2.96606600e-01 1.77262202e-01
-6.93583846e-01 3.91048146e-03 5.27554989e-01 4.40273285e-02
4.46897224e-02 5.91678798e-01 -6.30828261e-01 -4.98030841e-01
-2.19980925e-01 -7.79719234e-01 6.33018136e-01 -3.77048284e-01
-1.33020234e+00 -1.87606126e-01 5.27935028e-02 -9.52443242e-01
-2.35941216e-01 -4.90138441e-01 -4.80580211e-01 1.03428841e+00
-1.18656635e+00 -9.95414019e-01 3.09887558e-01 9.40727592e-02
9.99753270e-03 1.67177930e-01 1.04136527e+00 9.21691880e-02
-3.78965169e-01 5.86998343e-01 -8.39379430e-02 -2.53764570e-01
8.88933122e-01 -1.22694814e+00 -1.85698286e-01 6.17578864e-01
-7.95851275e-02 6.20369375e-01 8.62613976e-01 -1.04452026e+00
-7.29572356e-01 -1.25722075e+00 1.29331088e+00 -6.22194827e-01
7.01275945e-01 -1.75039977e-01 -9.96265113e-01 7.64248788e-01
-4.90316838e-01 -4.47188504e-02 1.08069623e+00 6.64269030e-01
-5.23568273e-01 5.53110391e-02 -1.29559612e+00 5.08594096e-01
7.86577165e-01 -3.23007822e-01 -6.10884726e-01 4.92211848e-01
5.28463244e-01 -2.41951942e-02 -9.62585092e-01 4.58300799e-01
5.74068964e-01 -4.60730791e-01 8.11603427e-01 -1.21308053e+00
7.08409905e-01 -2.37955794e-01 -1.08274415e-01 -1.61713612e+00
-4.18425113e-01 -9.24560577e-02 1.71223298e-01 1.27884543e+00
9.04791653e-01 -5.83856761e-01 5.57033539e-01 1.01734626e+00
4.28598076e-02 -6.62449002e-01 -1.02527344e+00 -5.25185764e-01
1.18484244e-01 -5.29477179e-01 4.16572958e-01 1.22186816e+00
1.70238540e-01 6.86294615e-01 -3.57216716e-01 3.92090797e-01
5.10074317e-01 -1.17882907e-01 5.10223567e-01 -1.49775052e+00
-3.40862393e-01 -3.52027975e-02 -2.81165302e-01 -1.76091596e-01
5.18513143e-01 -1.20059943e+00 5.12326881e-02 -1.06247878e+00
5.74803650e-01 -5.31161845e-01 -3.46252829e-01 9.84814703e-01
-6.18253589e-01 2.72084936e-03 -4.93483879e-02 -6.28333613e-02
-4.26306650e-02 1.42789245e-01 9.55321133e-01 6.85168877e-02
-3.36402923e-01 3.34282666e-01 -9.52702820e-01 7.27464616e-01
6.49530292e-01 -1.32746935e+00 -3.67074251e-01 3.46628785e-01
5.33015430e-01 -2.30787456e-01 4.83888566e-01 -3.90648633e-01
-7.50139430e-02 -4.24213886e-01 6.67283654e-01 -1.75905213e-01
6.21892279e-04 -9.43801522e-01 4.25557047e-01 7.67078817e-01
-1.04786193e+00 -3.21000189e-01 1.89883485e-01 7.38667548e-01
1.45946458e-01 -6.71989739e-01 6.62721395e-01 4.97428589e-02
-6.29672036e-02 -1.69425130e-01 -3.95721138e-01 -1.10613219e-01
8.54050279e-01 1.56627685e-01 -3.18274409e-01 5.21316146e-03
-6.50259197e-01 -3.38789634e-02 8.54895189e-02 3.53723556e-01
3.01271141e-01 -1.09291363e+00 -6.77628815e-01 2.11869944e-02
1.08413048e-01 -3.53363127e-01 1.69472098e-02 8.95961523e-01
3.27633232e-01 5.95430732e-01 1.84934661e-01 -2.04315364e-01
-1.33625019e+00 5.43218553e-01 1.80710465e-01 -4.56005543e-01
-2.97485173e-01 6.70238972e-01 4.79738772e-01 -4.95705694e-01
-4.11209017e-02 -3.08476120e-01 -3.51293892e-01 3.70296866e-01
4.07539994e-01 3.35630834e-01 3.59979421e-01 -2.23886162e-01
-3.53712380e-01 1.23522311e-01 -2.53997743e-01 3.93162258e-02
1.58761382e+00 3.14628333e-01 -1.21198051e-01 7.56051779e-01
1.19006598e+00 1.53945073e-01 -1.17716801e+00 -1.38619691e-01
3.62913847e-01 -5.42422175e-01 4.04211730e-01 -1.32170117e+00
-6.71782732e-01 4.48771060e-01 6.53596282e-01 -9.98106971e-02
8.02191198e-01 -1.31957069e-01 -9.92408395e-02 2.03838542e-01
-9.32347625e-02 -9.35453236e-01 -1.66370451e-01 2.27181405e-01
7.61622190e-01 -1.16940784e+00 3.20903689e-01 -7.48451293e-01
-5.56614995e-01 9.68722105e-01 4.48048174e-01 1.44295409e-01
5.22386730e-01 5.82974516e-02 -1.96686819e-01 -2.74367899e-01
-1.09575498e+00 -7.62441289e-03 4.14223254e-01 4.74817395e-01
5.84116161e-01 1.94494426e-01 -6.41707599e-01 7.16411948e-01
9.07572359e-02 2.40574643e-01 5.22190571e-01 4.51202005e-01
-6.12219274e-02 -1.10970163e+00 -3.62316489e-01 1.15488935e+00
-6.55060768e-01 -3.51761907e-01 -7.82572091e-01 8.52374732e-01
3.76712322e-01 9.97505605e-01 -1.02710024e-01 -3.19285303e-01
3.58499974e-01 5.82762003e-01 3.21826786e-01 -8.73000741e-01
-7.07455933e-01 -1.99106544e-01 3.67795706e-01 -3.69570315e-01
-5.33745766e-01 -1.05091846e+00 -1.23554468e+00 7.61531200e-03
-7.81018674e-01 1.51456445e-01 4.97069001e-01 1.40475082e+00
1.36035025e-01 5.64966261e-01 7.55189538e-01 -6.78414404e-02
-7.13599324e-01 -9.89848793e-01 -3.14551085e-01 4.75954086e-01
1.35674953e-01 -1.08074331e+00 -6.22331977e-01 5.42033091e-02]
|
[8.66940975189209, 5.503654956817627]
|
295a2766-dbe7-49b5-983f-95d62fc6f5af
|
explanations-for-automatic-speech-recognition
|
2302.14062
| null |
https://arxiv.org/abs/2302.14062v1
|
https://arxiv.org/pdf/2302.14062v1.pdf
|
Explanations for Automatic Speech Recognition
|
We address quality assessment for neural network based ASR by providing explanations that help increase our understanding of the system and ultimately help build trust in the system. Compared to simple classification labels, explaining transcriptions is more challenging as judging their correctness is not straightforward and transcriptions as a variable-length sequence is not handled by existing interpretable machine learning models. We provide an explanation for an ASR transcription as a subset of audio frames that is both a minimal and sufficient cause of the transcription. To do this, we adapt existing explainable AI (XAI) techniques from image classification-Statistical Fault Localisation(SFL) and Causal. Additionally, we use an adapted version of Local Interpretable Model-Agnostic Explanations (LIME) for ASR as a baseline in our experiments. We evaluate the quality of the explanations generated by the proposed techniques over three different ASR ,Google API, the baseline model of Sphinx, Deepspeech and 100 audio samples from the Commonvoice dataset.
|
['Ajitha Rajan', 'Peter Bell', 'Xiaoliang Wu']
|
2023-02-27
| null | null | null | null |
['interpretable-machine-learning']
|
['methodology']
|
[ 2.99872518e-01 8.21117043e-01 -1.41088488e-02 -7.20346391e-01
-1.11213505e+00 -5.17757654e-01 3.92759502e-01 6.21386524e-03
5.10340154e-01 7.74780333e-01 6.89606190e-01 -4.34748799e-01
-3.00342679e-01 -6.44863024e-02 -1.07609761e+00 -2.70628273e-01
-7.89655093e-03 4.89487767e-01 -1.32436574e-01 -3.43953334e-02
3.15327466e-01 3.63536656e-01 -1.64189577e+00 1.02274287e+00
5.45879841e-01 1.25425684e+00 -7.74995163e-02 1.16807330e+00
7.69586191e-02 1.54045141e+00 -1.24666250e+00 -1.69604227e-01
-1.47451445e-01 -4.99385059e-01 -1.12442732e+00 3.29437673e-01
5.11634648e-01 -2.13875338e-01 -3.97310667e-02 6.77735150e-01
1.75105855e-01 -2.18290329e-01 7.47142196e-01 -1.78390932e+00
-8.05967271e-01 1.13299847e+00 1.26800284e-01 2.52616197e-01
4.54194725e-01 7.62671009e-02 1.22066259e+00 -9.63845193e-01
2.06329271e-01 1.57737613e+00 4.69709635e-01 7.75538385e-01
-9.20736492e-01 -6.13630652e-01 2.61397511e-01 5.68912685e-01
-1.05142939e+00 -7.09882557e-01 4.52766448e-01 -2.84217745e-01
1.30643010e+00 7.50569701e-01 3.26041043e-01 1.47346067e+00
2.22152457e-01 7.14688361e-01 8.46355617e-01 -3.62356842e-01
3.93912613e-01 6.21931180e-02 4.74898994e-01 4.69230056e-01
-7.96538740e-02 -1.69848293e-01 -1.32257879e+00 -1.94297224e-01
4.63941962e-01 -1.30997017e-01 -4.74615246e-01 4.09122705e-01
-1.13453221e+00 5.75420558e-01 1.38584122e-01 3.82365286e-02
-6.30456507e-01 8.86017203e-01 2.86484569e-01 6.15817249e-01
4.81879592e-01 4.86571729e-01 -7.76399255e-01 -3.94287884e-01
-9.60951209e-01 2.29811855e-02 7.29288638e-01 8.51476729e-01
5.89572132e-01 7.96376824e-01 -2.19094500e-01 4.97215033e-01
6.65017605e-01 3.01059008e-01 5.61038554e-01 -1.34592319e+00
2.90091246e-01 5.30751467e-01 1.52276218e-01 -7.11787641e-01
9.42165181e-02 -6.43601656e-01 -6.44125521e-01 1.35569453e-01
-1.43119827e-01 8.10379386e-02 -1.14699328e+00 1.39481640e+00
-2.55504966e-01 6.40733004e-01 2.83629209e-01 1.01513541e+00
8.02920461e-01 6.42111421e-01 -3.70375067e-01 -8.52344483e-02
1.16190314e+00 -1.15439379e+00 -1.02475095e+00 -3.08146924e-01
4.56153572e-01 -5.80308318e-01 1.32344759e+00 9.65036750e-01
-8.89378369e-01 -5.68677902e-01 -1.28964436e+00 8.07904750e-02
1.34857163e-01 1.72268480e-01 4.08331186e-01 4.22760397e-01
-1.25371289e+00 5.40087223e-01 -7.43197024e-01 -8.00512731e-02
1.74515262e-01 4.84027207e-01 -2.60292023e-01 3.07647198e-01
-1.17397070e+00 6.03549659e-01 1.29334271e-01 9.88688841e-02
-1.64339280e+00 -4.10210520e-01 -6.68735683e-01 2.86931038e-01
3.77016991e-01 -3.93949509e-01 1.62170804e+00 -1.35262632e+00
-1.26790297e+00 2.42882177e-01 -5.08795559e-01 -9.41651940e-01
2.71856427e-01 -7.33917475e-01 -3.65356445e-01 1.58186525e-01
-8.45067725e-02 6.18366838e-01 1.05144286e+00 -1.46233153e+00
-5.00464380e-01 -2.66857427e-02 2.07085729e-01 -4.05266741e-03
-6.54353052e-02 6.79034442e-02 4.05325275e-03 -7.44115233e-01
2.33276039e-01 -9.50426161e-01 1.16105728e-01 -2.40928710e-01
-1.02667308e+00 -1.77934051e-01 9.49071825e-01 -8.99509668e-01
1.18813956e+00 -2.02798772e+00 -5.21003716e-02 3.34592402e-01
3.76710117e-01 -1.23055324e-01 -2.32180879e-02 2.94907689e-01
-5.66488266e-01 6.76575184e-01 -1.81486215e-02 -4.14608657e-01
6.31864518e-02 5.70339799e-01 -8.22269619e-01 2.58256607e-02
5.44431925e-01 3.95734698e-01 -6.04531884e-01 -2.22478658e-01
1.28998235e-01 4.66053098e-01 -4.46903080e-01 5.19525349e-01
-3.17963839e-01 3.72431517e-01 -2.05765501e-01 6.54246986e-01
-1.62417129e-01 -3.03061038e-01 -2.35001341e-01 -8.31987858e-02
3.93042326e-01 7.16130376e-01 -1.33654237e+00 1.38436580e+00
-3.31425101e-01 1.03452015e+00 -4.06624526e-01 -9.22314584e-01
1.19290447e+00 1.17080998e+00 -8.34919810e-02 -1.31610006e-01
-5.96060194e-02 2.18108833e-01 2.07091242e-01 -5.36884487e-01
3.83433849e-01 9.14708525e-02 2.38314345e-01 6.81851327e-01
1.76898837e-01 2.13120319e-02 -4.08105671e-01 5.81699789e-01
1.61003268e+00 -6.21194541e-02 3.89713407e-01 1.10444009e-01
3.69947821e-01 -1.71747744e-01 4.84136879e-01 9.96524990e-01
-5.79226064e-03 1.15657353e+00 5.42445898e-01 -4.87419575e-01
-1.09567225e+00 -7.54340947e-01 3.23897749e-01 7.75732100e-01
-2.81140625e-01 -7.33004689e-01 -9.72014308e-01 -8.51641834e-01
-4.38185960e-01 1.21110845e+00 -6.03973448e-01 -1.47348180e-01
-1.14813849e-01 6.66038617e-02 7.70231307e-01 7.18064904e-01
1.69612497e-01 -1.25896978e+00 -4.81921017e-01 1.17688797e-01
-5.76941252e-01 -1.09107387e+00 -2.27395326e-01 7.26825118e-01
-8.00709963e-01 -9.18490589e-01 1.35170221e-01 -2.05985740e-01
4.21682477e-01 4.73503247e-02 1.35529327e+00 7.23705888e-01
1.30484894e-01 3.84342968e-01 -5.82550704e-01 -7.23269641e-01
-9.75941300e-01 -2.44498551e-01 3.87116790e-01 -1.44616757e-02
1.59112573e-01 -3.97052050e-01 -3.05399746e-01 5.63813508e-01
-8.29385221e-01 7.36674592e-02 4.62336957e-01 6.89303875e-01
5.11347294e-01 -3.84262279e-02 5.90532243e-01 -8.52415383e-01
7.05233753e-01 -5.84014714e-01 1.83664680e-01 3.07541490e-01
-7.52772212e-01 3.81370276e-01 6.75613225e-01 -3.70698810e-01
-9.56834733e-01 -1.39850732e-02 -1.35032564e-01 -6.47699118e-01
-5.52225649e-01 5.12485921e-01 -1.48103297e-01 4.97217864e-01
1.02773976e+00 3.25700343e-02 -4.55483347e-01 -3.30492556e-01
2.89315522e-01 1.21694171e+00 7.33452618e-01 -4.63017732e-01
8.09281170e-01 1.53260276e-01 -3.94905239e-01 -4.68872190e-01
-9.03639078e-01 -1.92435756e-01 -4.80871737e-01 -3.60931933e-01
6.17912829e-01 -8.00080717e-01 -6.07055366e-01 -2.76988745e-01
-1.72076964e+00 -7.67568871e-02 -2.23327681e-01 4.33378458e-01
-8.33849013e-01 -3.22406180e-02 -6.20675266e-01 -1.19884992e+00
-4.86981034e-01 -1.41090488e+00 1.21076798e+00 -3.60462964e-01
-1.10503078e+00 -6.60570860e-01 -3.17921698e-01 6.51742756e-01
1.56658188e-01 2.29123309e-01 9.46996093e-01 -1.26076627e+00
-7.25086808e-01 -9.39014629e-02 8.50622579e-02 6.23943865e-01
7.90522546e-02 2.79316485e-01 -1.52103817e+00 2.86852479e-01
2.92179763e-01 -5.34878969e-01 5.78553140e-01 2.65077084e-01
1.25677776e+00 -9.56235945e-01 5.86610027e-02 2.52993125e-02
8.26283395e-01 2.88750142e-01 6.70535684e-01 4.50240403e-01
5.64362347e-01 5.23874402e-01 8.59487236e-01 4.35371399e-01
8.45959689e-03 6.27851307e-01 9.20049906e-01 -1.86050013e-02
-2.57108927e-01 -2.96357661e-01 7.79920638e-01 9.83952463e-01
-8.24763700e-02 -6.79213941e-01 -9.17200327e-01 4.14536119e-01
-2.19359446e+00 -8.83492351e-01 -5.27655780e-01 1.77444601e+00
6.15494788e-01 1.80181116e-01 -3.92156631e-01 1.07680023e+00
5.99299848e-01 -2.69129574e-01 -5.27189195e-01 -8.46597135e-01
-1.33470237e-01 -1.86461918e-02 3.81955802e-02 6.54457688e-01
-5.60023427e-01 6.60075068e-01 6.40861797e+00 7.54180193e-01
-9.27476048e-01 2.98918843e-01 7.63499439e-01 4.56473604e-02
-5.90156138e-01 1.71024278e-01 -4.26983327e-01 1.47175327e-01
1.60654271e+00 2.41461560e-01 5.18723309e-01 9.79811430e-01
7.08967984e-01 3.55355382e-01 -1.66247988e+00 8.85859251e-01
1.42710641e-01 -1.36803126e+00 3.48181367e-01 -1.78449094e-01
3.10872376e-01 -1.62936017e-01 9.96757671e-02 -1.04701407e-01
2.92393386e-01 -1.43264818e+00 1.47202134e+00 4.36811715e-01
5.10046721e-01 -6.02515042e-01 9.51757073e-01 2.92912900e-01
-7.91057110e-01 -1.42482907e-01 -1.52589872e-01 -3.69766921e-01
-9.98333916e-02 3.37742388e-01 -1.48339045e+00 2.39309922e-01
8.08815777e-01 5.71776807e-01 -6.62639141e-01 5.93037844e-01
-7.68205106e-01 1.46559632e+00 -1.06046153e-02 2.79073566e-01
8.08788016e-02 7.54201829e-01 7.53720105e-01 9.67684984e-01
3.53413761e-01 -8.59975517e-02 -2.72862196e-01 9.03474569e-01
-3.67838629e-02 -3.30374360e-01 -5.13158202e-01 -3.04257702e-02
6.84114158e-01 9.44427311e-01 -6.03242993e-01 -5.50599694e-01
1.93140525e-02 1.01491785e+00 6.46389052e-02 2.75018781e-01
-1.06854892e+00 1.72836497e-01 4.18396294e-01 8.26207325e-02
-2.55995572e-01 1.62979931e-01 -3.10500979e-01 -1.04304945e+00
2.19714642e-02 -1.54127657e+00 2.81714350e-01 -1.55279720e+00
-7.91773558e-01 1.11286259e+00 -1.54563963e-01 -1.30057800e+00
-7.10754037e-01 -2.49352500e-01 -4.09053624e-01 4.96544808e-01
-1.23296070e+00 -9.68963146e-01 -4.24059898e-01 3.74918282e-01
1.22627020e+00 -3.18330258e-01 9.34486091e-01 3.87876034e-02
-2.22111359e-01 4.21961516e-01 -5.60965121e-01 -2.22994149e-01
7.04060912e-01 -1.51676214e+00 5.02281189e-01 9.91176605e-01
9.02124226e-01 8.38191211e-01 1.31337130e+00 -6.63397551e-01
-1.04720485e+00 -1.13998139e+00 8.31711590e-01 -8.20296645e-01
5.60497403e-01 -2.05028534e-01 -1.08449376e+00 1.09981120e+00
4.19942617e-01 -2.42407262e-01 5.07056296e-01 6.26413971e-02
-3.29462439e-01 -8.34246054e-02 -8.83526027e-01 1.96521938e-01
8.83800983e-01 -6.93448484e-01 -6.90364301e-01 5.28440118e-01
1.21264899e+00 -2.94639170e-01 -5.56567430e-01 9.36448500e-02
3.44001770e-01 -1.11350977e+00 6.42445385e-01 -7.73226678e-01
5.94753206e-01 -6.45635366e-01 -5.72005868e-01 -1.35624897e+00
1.15266226e-01 -8.54109943e-01 -1.55805886e-01 1.43147933e+00
7.26473868e-01 -2.03856543e-01 8.02810729e-01 7.33292937e-01
-7.51672983e-01 -3.68981510e-01 -8.57138813e-01 -6.72926307e-01
-8.43880475e-01 -1.22601366e+00 7.52843082e-01 7.94249117e-01
-1.23496473e-01 4.52830136e-01 -8.48095417e-01 6.84262514e-01
4.45224464e-01 -5.97957611e-01 7.23991871e-01 -1.11920691e+00
-5.29608488e-01 3.25389862e-01 -6.77218795e-01 -6.86218858e-01
2.63483226e-01 -6.30788386e-01 3.98919135e-01 -1.47090161e+00
1.69394761e-01 -9.57718864e-02 -2.46604189e-01 1.01947355e+00
1.12603039e-01 2.44411245e-01 1.58406600e-01 4.83183354e-01
-6.10523522e-01 2.78624743e-01 4.65895802e-01 -2.42476508e-01
1.89031377e-01 -2.02523500e-01 -6.32865012e-01 9.63505864e-01
5.81708491e-01 -9.34932232e-01 -6.13617837e-01 -5.23371875e-01
3.52501631e-01 4.26390916e-01 5.37500024e-01 -1.11864352e+00
7.14421198e-02 -8.18903185e-03 1.91741344e-02 -4.42210495e-01
3.43995899e-01 -1.00326085e+00 5.65191627e-01 2.80547947e-01
-1.01799166e+00 2.84998327e-01 -5.63683324e-02 6.48335397e-01
-4.97405201e-01 -2.85773337e-01 1.54398561e-01 8.57150257e-02
-3.83803517e-01 -2.03014523e-01 -6.15084827e-01 -3.20792615e-01
4.87059325e-01 -4.01886970e-01 -4.39281225e-01 -1.14269078e+00
-8.84033382e-01 -2.56225973e-01 6.26361743e-02 6.09623611e-01
1.16504085e+00 -1.19736040e+00 -5.36258519e-01 -1.72017828e-01
1.70464039e-01 -1.51386306e-01 -2.22347274e-01 4.69155192e-01
-3.28251958e-01 6.58504844e-01 1.42195210e-01 -7.79773533e-01
-1.63896251e+00 1.00802280e-01 3.57901961e-01 2.82557040e-01
-3.58153760e-01 7.90120542e-01 2.41896331e-01 -1.61367804e-01
5.87407827e-01 -7.00345159e-01 -2.99345136e-01 -5.51732242e-01
5.89409769e-01 2.35046297e-01 2.87197590e-01 -4.97372597e-01
-2.75083572e-01 -2.02120729e-02 2.49022260e-01 -3.70200306e-01
1.30684555e+00 -2.80283391e-01 1.52690187e-01 1.03397059e+00
5.16929269e-01 -2.18975827e-01 -1.03045869e+00 2.02658504e-01
1.00872606e-01 -2.31065825e-01 4.25745472e-02 -1.08190656e+00
-8.84919405e-01 1.19917226e+00 7.17981339e-01 4.22435254e-01
8.66467357e-01 1.08423255e-01 4.05820727e-01 6.02718830e-01
1.32149175e-01 -6.46111727e-01 5.18666148e-01 3.00441712e-01
1.39953029e+00 -1.00401223e+00 -3.53588134e-01 -2.56071508e-01
-9.85892475e-01 1.42487288e+00 6.32970035e-01 1.96641967e-01
1.07540652e-01 3.72036576e-01 4.87655729e-01 -3.41646284e-01
-1.44890881e+00 2.54435569e-01 4.04454768e-01 4.95382011e-01
6.18193150e-01 -8.38482752e-02 3.09663236e-01 9.36912537e-01
-4.37632650e-01 2.35635471e-02 1.00961268e+00 4.56363529e-01
-4.76005316e-01 -6.84697568e-01 -6.37461722e-01 3.78851235e-01
-7.35571086e-01 -2.59898096e-01 -8.96563709e-01 5.13350248e-01
-4.53547761e-02 1.56219268e+00 -2.04163939e-01 -7.62018442e-01
9.83225852e-02 2.79956162e-01 -2.32610971e-01 -8.91876161e-01
-6.96956158e-01 -3.42141017e-02 5.29026866e-01 -6.88957393e-01
-4.03418541e-01 -2.65934825e-01 -1.77936327e+00 2.93352362e-03
-6.71169102e-01 3.01051974e-01 9.15001988e-01 1.43572056e+00
3.90165567e-01 8.74859452e-01 4.70420986e-01 -4.43813175e-01
-3.45621347e-01 -1.09278870e+00 -2.88590431e-01 4.56554949e-01
5.76746345e-01 -5.18909216e-01 -7.58916497e-01 5.72184861e-01]
|
[8.944230079650879, 5.694268703460693]
|
6849e30f-a920-44ae-8f97-53e6bd4ed761
|
dynamic-measurement-scheduling-for-event
|
1901.09699
| null |
https://arxiv.org/abs/1901.09699v3
|
https://arxiv.org/pdf/1901.09699v3.pdf
|
Dynamic Measurement Scheduling for Event Forecasting using Deep RL
|
Imagine a patient in critical condition. What and when should be measured to forecast detrimental events, especially under the budget constraints? We answer this question by deep reinforcement learning (RL) that jointly minimizes the measurement cost and maximizes predictive gain, by scheduling strategically-timed measurements. We learn our policy to be dynamically dependent on the patient's health history. To scale our framework to exponentially large action space, we distribute our reward in a sequential setting that makes the learning easier. In our simulation, our policy outperforms heuristic-based scheduling with higher predictive gain and lower cost. In a real-world ICU mortality prediction task (MIMIC3), our policies reduce the total number of measurements by $31\%$ or improve predictive gain by a factor of $3$ as compared to physicians, under the off-policy policy evaluation.
|
['Chun-Hao Chang', 'Mingjie Mai', 'Anna Goldenberg']
|
2019-01-24
| null | null | null | null |
['icu-mortality']
|
['medical']
|
[ 1.48575589e-01 4.97398436e-01 -4.30501342e-01 -1.30132452e-01
-9.55486000e-01 -1.29422531e-01 -2.86748439e-01 5.39132595e-01
-7.99046576e-01 1.36123967e+00 1.87222466e-01 -7.43876755e-01
-4.48048025e-01 -6.07390165e-01 -5.42142034e-01 -5.98173738e-01
-4.70201284e-01 9.11673546e-01 -1.33871555e-01 1.97873399e-01
4.27169167e-02 1.41010165e-01 -6.62300467e-01 -1.42581284e-01
6.80714607e-01 1.06974268e+00 2.17819467e-01 8.39175403e-01
5.94845355e-01 1.18483472e+00 -5.27504265e-01 9.92399752e-02
5.88648796e-01 -6.94845140e-01 -6.96221530e-01 2.71798551e-01
-3.08047682e-01 -8.40672493e-01 -2.53646016e-01 5.47571719e-01
7.78846979e-01 2.57629335e-01 3.47350508e-01 -8.90322566e-01
9.52731892e-02 7.02918291e-01 -2.45544568e-01 4.17559057e-01
7.03556538e-02 6.73687339e-01 6.96072340e-01 3.73524636e-01
1.80703834e-01 1.03169429e+00 3.78927469e-01 8.47112954e-01
-1.35208976e+00 -2.11776778e-01 3.25915396e-01 1.86513383e-02
-7.67921925e-01 -2.47718260e-01 2.61291951e-01 -2.19063669e-01
1.00340140e+00 1.75704896e-01 8.57599497e-01 9.24063325e-01
5.42130589e-01 4.16205406e-01 1.11322379e+00 -2.25183889e-01
5.83271205e-01 -2.11105213e-01 -2.34151915e-01 6.72544837e-01
1.38123587e-01 4.46004093e-01 -4.64717522e-02 -4.51020896e-01
1.01354063e+00 5.87745190e-01 -4.27551091e-01 -1.81405902e-01
-1.29245579e+00 7.51737118e-01 -9.87857878e-02 -5.23640275e-01
-1.02813423e+00 4.76639032e-01 2.98087507e-01 5.45619786e-01
-3.05302292e-02 1.04198527e+00 -9.07397866e-01 -3.84627074e-01
-5.36879838e-01 1.97608531e-01 8.56464207e-01 8.88481259e-01
7.66417012e-02 1.87226385e-01 -7.28665352e-01 4.72833335e-01
-2.68417358e-01 7.99675107e-01 4.04136807e-01 -1.77500188e+00
4.76718962e-01 8.18332881e-02 9.80034471e-01 -1.84889600e-01
-7.55973518e-01 -4.23885584e-01 -7.00053751e-01 -1.68007053e-02
3.80485982e-01 -1.11985838e+00 -7.58647919e-01 1.91179180e+00
5.17872795e-02 1.27139673e-01 -1.78578019e-01 9.33154106e-01
-3.86757135e-01 3.01707476e-01 1.24407731e-01 -9.32813883e-01
1.17253065e+00 -6.75409377e-01 -5.35555065e-01 -1.99022368e-01
7.61349261e-01 -3.19945067e-01 1.06909931e+00 5.26276171e-01
-1.35760474e+00 9.36449617e-02 -5.16865909e-01 8.54666352e-01
6.55112147e-01 -1.09285548e-01 2.77763903e-01 2.90344059e-01
-8.90190065e-01 9.80976760e-01 -1.20846522e+00 -1.18169181e-01
4.13961619e-01 4.31981355e-01 3.51339787e-01 6.98563084e-02
-1.03704941e+00 1.03113639e+00 2.20900878e-01 -3.34889978e-01
-1.55202734e+00 -8.07593226e-01 -4.82137114e-01 5.30119598e-01
1.11869943e+00 -1.12624848e+00 1.83301628e+00 -5.14261127e-01
-1.69407094e+00 3.41313928e-01 1.59190819e-01 -8.86154473e-01
7.25391567e-01 -3.15280825e-01 5.86009212e-02 3.58891517e-01
-1.38790593e-01 3.07518393e-01 6.42504752e-01 -8.17463577e-01
-6.73930526e-01 -1.28925592e-01 4.27814543e-01 5.00508904e-01
-8.01888406e-02 -2.67840356e-01 2.19315946e-01 -3.61353964e-01
-4.24006194e-01 -1.28208828e+00 -1.04951227e+00 -3.76307338e-01
-1.64812297e-01 2.87572443e-01 1.03815272e-01 -3.68147254e-01
1.26741064e+00 -1.87522459e+00 -1.03441454e-01 8.66969302e-02
2.53778607e-01 -1.62396312e-01 4.35120091e-02 4.28738773e-01
2.06007451e-01 4.39597815e-02 -2.52447426e-01 -6.37825057e-02
-1.58164665e-01 5.44039369e-01 -9.65999886e-02 2.97835171e-01
-2.26014525e-01 7.79410720e-01 -1.02726293e+00 -4.19848770e-01
1.77590474e-01 -3.66670221e-01 -8.88142824e-01 7.43469059e-01
-3.67357373e-01 7.73863494e-01 -7.15400457e-01 4.26117897e-01
-1.52432248e-02 -6.96185470e-01 7.40304887e-01 6.62559211e-01
4.08686578e-01 9.55637395e-02 -8.86393130e-01 1.06414151e+00
-6.66946948e-01 -1.53671488e-01 9.94838774e-02 -1.27352881e+00
4.59612399e-01 4.87302870e-01 1.15463650e+00 -4.91121680e-01
1.93255618e-01 -8.00780877e-02 2.52078354e-01 -7.55775690e-01
-2.21432626e-01 -6.69651985e-01 -2.56793704e-02 6.52092874e-01
-3.84949178e-01 5.01533337e-02 -1.89409226e-01 5.20823374e-02
1.67480099e+00 -3.47457528e-01 7.24474132e-01 -4.46971953e-01
-1.51800960e-01 -1.32259682e-01 9.76010263e-01 1.05313766e+00
-4.34501827e-01 7.83025250e-02 9.64531422e-01 -4.84632015e-01
-8.20150614e-01 -9.00845110e-01 2.40056723e-01 7.98931897e-01
-6.78474903e-02 2.33900785e-01 -5.98207116e-01 -8.02677810e-01
4.42401141e-01 8.48823428e-01 -6.20863199e-01 -4.22097296e-01
-8.67619932e-01 -9.17250574e-01 -1.93563625e-01 6.45314455e-01
-9.00602788e-02 -1.24937522e+00 -1.53475821e+00 6.89010322e-01
-1.18425176e-01 -9.40333724e-01 -7.24907994e-01 4.16155696e-01
-1.08952630e+00 -1.15969682e+00 -8.92597854e-01 3.41561250e-02
7.27815986e-01 -9.67033505e-02 1.32001281e+00 -1.15239970e-01
-2.93230116e-01 7.22192168e-01 -5.78396581e-02 -5.41023016e-01
-1.58046275e-01 -8.59536156e-02 2.24055082e-01 -4.33497906e-01
-1.45928606e-01 -4.36407954e-01 -1.26749349e+00 2.30203912e-01
-4.91967201e-01 -2.61574984e-01 5.41254163e-01 1.18344569e+00
5.95838428e-01 -2.71721750e-01 8.64321887e-01 -1.08705091e+00
9.17060256e-01 -5.35270929e-01 -8.58000219e-01 3.23916584e-01
-1.16962075e+00 2.69615442e-01 8.75031412e-01 -5.22667646e-01
-6.94425642e-01 -3.51375751e-02 4.32385117e-01 -6.45174623e-01
1.44628301e-01 2.73381650e-01 3.58307481e-01 4.81243461e-01
4.12061691e-01 8.38861912e-02 1.31194010e-01 -2.41200790e-01
-1.00304611e-01 2.67188072e-01 3.89371216e-02 -9.32362080e-01
-1.68119359e-03 1.67129077e-02 4.27513897e-01 4.69404086e-02
-1.05999446e+00 -1.55580878e-01 -1.45657044e-02 -5.97423054e-02
6.86214805e-01 -8.20964217e-01 -1.66876876e+00 -3.67861986e-01
-6.16175115e-01 -1.03520572e+00 -7.58788228e-01 8.45295072e-01
-1.22606850e+00 5.71395010e-02 -6.29886866e-01 -9.94392693e-01
-3.56042862e-01 -1.18959439e+00 7.18150556e-01 -8.60403851e-02
-1.30647242e-01 -8.38810921e-01 3.88132669e-02 -1.70988068e-02
4.13708985e-01 3.63340735e-01 9.63962018e-01 -3.77032012e-01
-5.80352128e-01 1.91741765e-01 7.71234632e-02 3.29086035e-01
1.64273426e-01 -6.82016253e-01 -1.96697235e-01 -6.87791824e-01
3.98829058e-02 -3.31574798e-01 4.73802477e-01 7.80593216e-01
1.76124358e+00 -7.92225838e-01 -4.05740976e-01 5.20950019e-01
1.34390783e+00 7.91801333e-01 4.97214422e-02 2.49567151e-01
1.81541249e-01 1.30516320e-01 7.72799075e-01 1.33860230e+00
2.50409514e-01 3.21729422e-01 5.88092864e-01 9.17238817e-02
5.32609701e-01 -1.49432689e-01 1.85588911e-01 4.16400105e-01
-2.77025521e-01 -3.40624392e-01 -1.10185862e+00 2.83115327e-01
-2.12876368e+00 -7.25440264e-01 7.49831975e-01 2.56658864e+00
9.49183822e-01 3.51387203e-01 3.56512278e-01 -3.87751549e-01
1.79019973e-01 -2.39694923e-01 -1.04973876e+00 -5.98141313e-01
5.17627835e-01 2.32045844e-01 1.13195455e+00 5.39496779e-01
-7.11502075e-01 4.61091459e-01 7.44979334e+00 1.30010530e-01
-1.12500513e+00 7.24719763e-02 1.21949947e+00 -7.49095142e-01
1.52926683e-01 -1.81942120e-01 -4.02295560e-01 5.96421242e-01
1.70402896e+00 -3.43677014e-01 7.94352651e-01 9.11491334e-01
8.24252903e-01 -1.47374943e-01 -1.34517789e+00 7.83105910e-01
-5.04100621e-01 -1.21602869e+00 -5.41802108e-01 8.82070884e-02
6.56515002e-01 -4.03225683e-02 -2.23701268e-01 5.69172800e-01
1.02094364e+00 -9.77333844e-01 1.40758201e-01 7.42044747e-01
8.13382387e-01 -9.07075465e-01 8.38591278e-01 6.13789201e-01
-4.82988268e-01 -6.93871617e-01 -2.96837389e-01 -2.06245899e-01
5.63259363e-01 5.58682740e-01 -1.15882802e+00 -6.18332028e-02
5.00407040e-01 1.22426145e-01 2.09858149e-01 8.92403066e-01
-1.31159171e-01 9.55232561e-01 -2.77787119e-01 5.97156584e-02
2.74071097e-01 -9.15213972e-02 3.25453490e-01 8.49686563e-01
3.47375870e-01 7.95649529e-01 8.70909393e-01 3.43700200e-01
-1.22471474e-01 -9.94477645e-02 -2.85026789e-01 4.95620333e-02
4.95221794e-01 7.79654562e-01 -3.93744648e-01 -6.97170675e-01
1.19157262e-01 7.62375236e-01 1.95938796e-01 2.42510259e-01
-8.53212833e-01 8.65205973e-02 4.61069167e-01 4.09456730e-01
4.41512130e-02 8.53426829e-02 -2.45920241e-01 -8.74938428e-01
-1.71868026e-01 -9.12585020e-01 6.01789355e-01 -3.53286117e-01
-1.02644265e+00 2.82625437e-01 -5.77465305e-03 -1.24673295e+00
-7.37407506e-01 -2.86872774e-01 -3.47905129e-01 6.87831342e-01
-1.35992360e+00 1.27395587e-02 1.70723006e-01 3.49966824e-01
3.89773786e-01 1.81542248e-01 9.35716927e-01 1.14035301e-01
-7.60706544e-01 3.40839118e-01 1.48705974e-01 -2.66300082e-01
5.97282708e-01 -1.38172889e+00 -1.43211573e-01 2.52687156e-01
-8.81445229e-01 2.84824699e-01 6.89615250e-01 -4.48380798e-01
-1.14966226e+00 -1.01864493e+00 2.84678370e-01 -3.45218122e-01
5.61789393e-01 1.67616099e-01 -6.11335635e-01 6.76997781e-01
6.86373189e-02 1.44465327e-01 5.00462532e-01 -6.68725520e-02
2.96514302e-01 -2.38703474e-01 -1.30629575e+00 5.62219977e-01
1.09733176e+00 7.64642730e-02 -3.37069780e-01 6.72446012e-01
8.58490884e-01 -4.14125532e-01 -1.33434033e+00 4.62383479e-01
3.12594622e-01 -4.85306323e-01 6.56364322e-01 -1.27061450e+00
3.51340562e-01 4.20217603e-01 -1.67819206e-02 -1.70622432e+00
-4.34692472e-01 -1.08145142e+00 -3.57206374e-01 1.03575058e-01
4.26610023e-01 -7.99851835e-01 7.61433840e-01 8.46893847e-01
-7.13645667e-02 -1.37719870e+00 -8.96681011e-01 -8.10269654e-01
7.00777844e-02 1.69301495e-01 5.55117667e-01 6.84886277e-01
2.88389891e-01 1.28262475e-01 -7.98640788e-01 1.88759621e-02
5.92643082e-01 2.62193620e-01 2.27618694e-01 -7.05036342e-01
-1.00147688e+00 -2.85537869e-01 2.31036916e-01 -9.91932929e-01
-1.55607462e-01 -3.19775343e-01 1.97575808e-01 -1.35978448e+00
4.55203772e-01 -5.57769597e-01 -9.60541487e-01 7.20334172e-01
-5.20057738e-01 -6.97292745e-01 3.87243658e-01 4.55575343e-03
-8.62169445e-01 4.88925725e-01 1.42059481e+00 1.78257167e-01
-2.10404128e-01 2.68353075e-01 -6.80873036e-01 4.69628423e-01
9.86285388e-01 -6.74428046e-01 -6.07049584e-01 -3.36720675e-01
-1.75366879e-01 1.44145072e+00 -1.21188991e-01 -6.61374509e-01
-3.41274828e-01 -1.08599949e+00 1.86363488e-01 -1.99828781e-02
1.56528741e-01 -8.00020874e-01 -1.05755009e-01 1.18918991e+00
-7.75536239e-01 4.83278066e-01 9.35446750e-03 6.81966960e-01
2.86016077e-01 1.06675379e-01 1.04337478e+00 -6.72773898e-01
5.65737039e-02 7.25871861e-01 -4.52379227e-01 4.73912418e-01
1.24447870e+00 3.65444303e-01 -1.79902524e-01 -7.62250900e-01
-1.15115380e+00 8.04873526e-01 1.47999033e-01 -1.13186568e-01
4.89658207e-01 -7.79271007e-01 -6.70578301e-01 -3.33651125e-01
-2.39274129e-01 -1.22127488e-01 3.01273912e-01 1.06274748e+00
-2.87277550e-01 4.61552650e-01 -3.87766995e-02 -2.83338934e-01
-8.23217988e-01 9.06770468e-01 6.85372710e-01 -7.88595796e-01
-6.64557874e-01 4.20251817e-01 7.35993385e-02 -2.84187458e-02
2.63974577e-01 -3.25818926e-01 3.27349663e-01 -3.85190576e-01
3.66335601e-01 7.13511050e-01 -1.95221499e-01 4.70103562e-01
-2.44929895e-01 -2.04451177e-02 1.42972291e-01 -1.09954938e-01
1.31534111e+00 -1.05570510e-01 3.99297744e-01 2.26721913e-01
7.03694403e-01 -5.06716311e-01 -1.73256183e+00 -1.80899516e-01
3.80573496e-02 -3.69246811e-01 2.04028226e-02 -1.16001225e+00
-1.05586982e+00 5.99123478e-01 7.26042211e-01 2.10029900e-01
1.16358721e+00 -3.02627146e-01 8.10305297e-01 7.24669397e-01
5.97796082e-01 -1.08152103e+00 1.47684366e-02 1.06816731e-01
5.82545042e-01 -1.36636925e+00 9.14670601e-02 2.18916655e-01
-1.18611670e+00 6.37105048e-01 6.53817773e-01 -3.98534894e-01
7.88491905e-01 4.00874525e-01 1.18006654e-01 5.60733899e-02
-1.49493027e+00 5.05391322e-02 -4.09511507e-01 2.56230056e-01
3.46478969e-02 8.94095063e-01 -4.05188382e-01 3.44726235e-01
1.38312563e-01 3.53815079e-01 7.29274631e-01 1.09825742e+00
-6.86897159e-01 -1.02149367e+00 -2.49706224e-01 1.07393420e+00
-7.18539417e-01 2.53144391e-02 2.94307590e-01 4.37808961e-01
-2.23040253e-01 7.96731114e-01 5.80484420e-02 8.04475397e-02
5.45519829e-01 1.95696335e-02 4.61648226e-01 -7.07225382e-01
-3.41754675e-01 3.56065452e-01 1.03952512e-01 -9.27502632e-01
-4.55198623e-02 -6.05875254e-01 -1.53702271e+00 -2.03666478e-01
2.47723177e-01 1.64628029e-01 7.50859380e-02 8.83495569e-01
5.46723962e-01 1.03221035e+00 1.11372209e+00 -3.00447941e-01
-1.58080196e+00 -8.41039121e-01 -5.66571534e-01 2.40446672e-01
5.72237730e-01 -6.30117357e-01 -2.39390433e-01 -2.63117611e-01]
|
[3.9733691215515137, 2.7591826915740967]
|
9134cd84-e180-45d1-a3a3-e668bdf008bb
|
sok-privacy-preserving-data-synthesis
|
2307.02106
| null |
https://arxiv.org/abs/2307.02106v1
|
https://arxiv.org/pdf/2307.02106v1.pdf
|
SoK: Privacy-Preserving Data Synthesis
|
As the prevalence of data analysis grows, safeguarding data privacy has become a paramount concern. Consequently, there has been an upsurge in the development of mechanisms aimed at privacy-preserving data analyses. However, these approaches are task-specific; designing algorithms for new tasks is a cumbersome process. As an alternative, one can create synthetic data that is (ideally) devoid of private information. This paper focuses on privacy-preserving data synthesis (PPDS) by providing a comprehensive overview, analysis, and discussion of the field. Specifically, we put forth a master recipe that unifies two prominent strands of research in PPDS: statistical methods and deep learning (DL)-based methods. Under the master recipe, we further dissect the statistical methods into choices of modeling and representation, and investigate the DL-based methods by different generative modeling principles. To consolidate our findings, we provide comprehensive reference tables, distill key takeaways, and identify open problems in the existing literature. In doing so, we aim to answer the following questions: What are the design principles behind different PPDS methods? How can we categorize these methods, and what are the advantages and disadvantages associated with each category? Can we provide guidelines for method selection in different real-world scenarios? We proceed to benchmark several prominent DL-based methods on the task of private image synthesis and conclude that DP-MERF is an all-purpose approach. Finally, upon systematizing the work over the past decade, we identify future directions and call for actions from researchers.
|
['Dawn Song', 'Bo Li', 'David Forsyth', 'Bolin Ding', 'Chang Ge', 'Gonzalo Munilla Garrido', 'Yunhui Long', 'Qinbin Li', 'Fan Wu', 'Yuzheng Hu']
|
2023-07-05
| null | null | null | null |
['image-generation']
|
['computer-vision']
|
[ 5.18867493e-01 1.42711893e-01 -2.89740533e-01 -3.70322913e-01
-8.35088611e-01 -8.15547705e-01 6.74538195e-01 1.41085103e-01
-3.50687355e-01 6.86887503e-01 3.41117412e-01 -4.94452506e-01
-9.81214717e-02 -7.68303394e-01 -5.76352596e-01 -7.80076325e-01
2.18787059e-01 -1.99878603e-01 -2.89790064e-01 6.49927929e-02
2.09472105e-01 6.19370222e-01 -1.47127938e+00 3.04994315e-01
7.00267911e-01 1.01313853e+00 -5.26302516e-01 3.17826748e-01
8.75659436e-02 6.64393842e-01 -6.27342820e-01 -1.07298529e+00
6.14458680e-01 -5.17858446e-01 -7.73822486e-01 1.12716168e-01
4.29080993e-01 -3.20797980e-01 -4.79839325e-01 1.16893029e+00
6.11596048e-01 -8.80955607e-02 4.45360541e-01 -1.61315572e+00
-7.59244442e-01 3.44266683e-01 -4.47521985e-01 -7.69157037e-02
4.33639362e-02 2.59159505e-01 9.67270792e-01 -4.96237040e-01
6.57949090e-01 1.05584538e+00 6.03311837e-01 7.75644004e-01
-1.44678986e+00 -6.76982403e-01 1.86851159e-01 -1.49747012e-02
-1.32248628e+00 -8.03087711e-01 6.39857054e-01 -4.36163634e-01
4.59570140e-01 7.49591529e-01 4.66788203e-01 1.38962269e+00
6.10252395e-02 9.96154964e-01 1.21079791e+00 -3.29499692e-01
3.04864258e-01 4.99588817e-01 -6.26722202e-02 3.28326970e-01
5.86973965e-01 2.39632711e-01 -5.92109144e-01 -5.84153354e-01
4.28340167e-01 5.29508013e-03 -3.19327265e-01 -9.35705543e-01
-9.94857550e-01 9.91916955e-01 -9.14824232e-02 9.62919593e-02
2.57343873e-02 -1.07098237e-01 5.40936708e-01 2.96240538e-01
4.02162820e-01 5.75486004e-01 -2.52910227e-01 1.02106608e-01
-1.05972564e+00 5.88486075e-01 8.95338178e-01 1.00337255e+00
4.98214900e-01 -1.65184457e-02 -2.24312693e-01 4.40265894e-01
2.01297730e-01 2.02934429e-01 1.20699190e-01 -1.11248314e+00
3.47870559e-01 4.74811532e-02 1.14764273e-01 -1.24828207e+00
-1.38371736e-02 -2.36202493e-01 -1.04186785e+00 7.56482109e-02
4.12673086e-01 -3.21496904e-01 -4.69352990e-01 1.91868901e+00
1.83948055e-01 -3.58439118e-01 8.41175541e-02 4.58740234e-01
6.32535696e-01 3.86779845e-01 1.87485456e-01 -3.68576080e-01
1.43297207e+00 -5.41005969e-01 -8.01067770e-01 -5.40225022e-02
4.60358143e-01 -4.66202319e-01 8.98477674e-01 3.23234051e-01
-9.07624722e-01 -1.98301390e-01 -9.28390086e-01 -3.01750511e-01
-5.51575840e-01 -2.27499977e-02 6.93884790e-01 1.11369002e+00
-1.01460433e+00 4.84492868e-01 -8.50111723e-01 -5.14060736e-01
6.95381165e-01 1.72020674e-01 -4.90740895e-01 1.83401689e-01
-1.21707428e+00 6.93059444e-01 1.18453354e-01 -8.26987773e-02
-6.47738993e-01 -8.06945145e-01 -7.34942675e-01 -3.02170888e-02
4.15050715e-01 -9.45578218e-01 9.56857383e-01 -8.14827442e-01
-1.19450891e+00 1.24815392e+00 -1.76168129e-01 -6.82091773e-01
8.52612734e-01 5.61458282e-02 -4.10863549e-01 -1.12651259e-01
-1.25762463e-01 4.12743986e-01 7.63945460e-01 -1.48619628e+00
-7.13649511e-01 -4.98867542e-01 -1.57811016e-01 -7.16285259e-02
-3.44332010e-01 2.77915508e-01 -3.65403026e-01 -9.59453225e-01
-3.23290616e-01 -9.36383665e-01 -3.72797906e-01 3.24460924e-01
-6.22789800e-01 2.03035519e-01 7.13054419e-01 -4.25009400e-01
1.53143334e+00 -2.74723458e+00 -2.36883655e-01 8.26909766e-02
5.99476397e-01 2.94129103e-01 2.08600745e-01 6.38189852e-01
-5.20060621e-02 5.32336771e-01 -6.70614064e-01 -6.01710975e-01
3.14391583e-01 -6.30445480e-02 -6.68092728e-01 6.38897598e-01
-7.57682398e-02 9.27268922e-01 -6.27410293e-01 -2.61349171e-01
9.98880640e-02 4.56313163e-01 -6.44135654e-01 5.61281033e-02
-4.15022559e-02 3.55912030e-01 -3.41804445e-01 6.75823450e-01
9.20903265e-01 -1.08837016e-01 3.59747589e-01 -6.93259165e-02
-1.79342583e-01 1.17732801e-01 -9.86704946e-01 1.32408226e+00
-7.23005785e-03 7.53410161e-01 3.42793286e-01 -9.36782956e-01
7.82149971e-01 9.64010134e-02 4.99515295e-01 -3.13655496e-01
-1.34282606e-02 1.84878334e-01 -2.35767812e-01 -3.54773402e-01
4.65423495e-01 -1.28784373e-01 -2.52083898e-01 4.56740022e-01
-3.93660069e-01 -6.16609585e-03 -2.67120659e-01 -1.11098457e-02
1.15054798e+00 -1.38234451e-01 5.28777659e-01 -1.21565439e-01
3.23015004e-01 -1.78972930e-01 5.55151105e-01 9.14826214e-01
-6.40521824e-01 7.96666205e-01 6.94433808e-01 -5.15356719e-01
-8.76748562e-01 -1.03468871e+00 -3.07011694e-01 8.22142780e-01
-1.74863487e-01 -7.05387831e-01 -8.28113616e-01 -8.52168202e-01
2.88061082e-01 7.68752635e-01 -7.45064437e-01 -9.97594744e-02
-2.52050191e-01 -1.31285250e+00 7.64487982e-01 1.24147736e-01
5.83554268e-01 -8.49203110e-01 -7.15852976e-01 -2.37597808e-01
-2.31272414e-01 -9.36383069e-01 -3.22124541e-01 -1.65320095e-02
-7.20614433e-01 -8.40864778e-01 -5.30861795e-01 -3.42587143e-01
6.54025197e-01 3.65518451e-01 9.29112196e-01 -1.12958491e-01
-1.90575033e-01 4.69682276e-01 -1.90521330e-01 -5.58791578e-01
-4.49191213e-01 1.28406152e-01 -1.38038233e-01 3.16005081e-01
6.69295371e-01 -5.66204786e-01 -6.33095026e-01 5.98879419e-02
-1.21928585e+00 -1.88540131e-01 5.53474605e-01 5.41465759e-01
6.17910087e-01 1.29152820e-01 2.95799553e-01 -1.26684165e+00
8.61315012e-01 -6.04235888e-01 -6.65908456e-01 3.15070897e-01
-1.04428077e+00 5.98436594e-02 4.11036432e-01 3.48330364e-02
-9.25254464e-01 2.02501357e-01 -2.17064574e-01 -3.77418876e-01
-1.32440776e-01 1.67813733e-01 -5.52210450e-01 -2.50759814e-02
6.20975256e-01 4.32625741e-01 3.76797765e-01 -5.40577412e-01
5.24658918e-01 6.93762362e-01 3.54314625e-01 -6.21787786e-01
6.09181702e-01 8.21903765e-01 -1.33083984e-01 -8.14332128e-01
-4.64883983e-01 4.51239571e-02 -3.02276909e-01 2.35223964e-01
7.78713942e-01 -7.22759247e-01 -7.48521268e-01 5.64992130e-01
-9.32396531e-01 -1.14929184e-01 -6.06912017e-01 5.92654124e-02
-7.11961210e-01 6.03081763e-01 -2.87936091e-01 -8.17403913e-01
-2.56545275e-01 -1.31547785e+00 8.87173235e-01 -1.53632969e-01
-4.37944382e-01 -8.88077497e-01 -3.59182917e-02 4.64346707e-01
4.71779913e-01 5.80882430e-01 8.90600383e-01 -7.56599128e-01
-7.99606085e-01 -3.02574188e-01 -2.20170856e-01 4.39060837e-01
1.03234135e-01 1.80937752e-01 -1.27783167e+00 -4.06454384e-01
4.28088218e-01 -3.39351483e-02 7.51898825e-01 3.87651861e-01
1.66637993e+00 -8.31356466e-01 -3.09060216e-01 9.28547442e-01
1.37280154e+00 9.08221453e-02 7.97521234e-01 3.39352101e-01
4.43725049e-01 1.02376068e+00 3.23977023e-01 6.20683372e-01
6.02318227e-01 5.35922229e-01 2.50982106e-01 -1.69353768e-01
7.04964995e-02 -5.08426130e-01 3.24276030e-01 2.19180495e-01
2.55273402e-01 -4.02720898e-01 -5.32164156e-01 5.04083037e-01
-1.86975563e+00 -1.03962493e+00 -1.77900456e-02 2.46823049e+00
7.77116835e-01 -2.52178192e-01 3.78420562e-01 3.27686295e-02
5.34179389e-01 5.70943475e-01 -5.42518735e-01 -3.55010927e-01
-2.97897398e-01 -1.33282036e-01 7.84122229e-01 2.29534015e-01
-1.23116052e+00 6.92007065e-01 7.29419947e+00 7.04492450e-01
-1.01958871e+00 -5.35343848e-02 9.87155080e-01 -5.59688471e-02
-6.70117438e-01 1.67591527e-01 -7.16456890e-01 5.80562890e-01
8.50548387e-01 -5.01842439e-01 4.01796669e-01 8.40321660e-01
1.34163931e-01 4.29809466e-02 -1.25592828e+00 1.13904500e+00
7.23768920e-02 -1.49678075e+00 9.29954275e-02 4.88516480e-01
4.51920956e-01 -1.61480844e-01 4.75422800e-01 -8.16039667e-02
4.27652925e-01 -1.05147946e+00 5.91340482e-01 4.75894004e-01
7.49687135e-01 -5.52583933e-01 3.75056624e-01 1.72306865e-01
-5.99029064e-01 -5.35313822e-02 -1.08203277e-01 8.53533670e-02
-1.00905076e-01 6.34449661e-01 -2.47875735e-01 6.98546231e-01
8.53998721e-01 6.19437814e-01 -4.29748237e-01 8.55614960e-01
-6.66844696e-02 3.98999393e-01 -1.25998974e-01 1.54075623e-01
-1.99542776e-01 -2.10703135e-01 5.23102939e-01 1.18173492e+00
2.64366716e-01 6.12901933e-02 -2.34452143e-01 1.00823748e+00
-1.86073139e-01 9.16112438e-02 -8.41509044e-01 -2.77719408e-01
7.12701142e-01 1.03078151e+00 -3.87150675e-01 3.27793211e-02
-5.90504229e-01 8.87489855e-01 -1.07800215e-02 3.33645076e-01
-4.70673472e-01 -3.25068086e-01 1.07840705e+00 3.09305519e-01
-7.56263062e-02 -1.05046585e-01 -7.40882635e-01 -1.33488286e+00
1.04005612e-01 -1.44016612e+00 7.40849435e-01 -3.39163959e-01
-1.41901195e+00 3.09137434e-01 9.37937796e-02 -1.13225138e+00
-4.29509766e-02 -3.31837118e-01 -1.08931638e-01 7.61040032e-01
-1.30522215e+00 -9.50144768e-01 -4.27069049e-03 4.25731480e-01
4.42838706e-02 -1.44892722e-01 9.34995830e-01 2.57976085e-01
-6.76467478e-01 9.09475803e-01 4.63828832e-01 1.70209378e-01
7.69749403e-01 -9.02238309e-01 6.19838417e-01 1.04344380e+00
4.44643870e-02 9.99210417e-01 7.16933012e-01 -4.53595191e-01
-1.51830089e+00 -1.03692830e+00 9.92046177e-01 -8.06473374e-01
4.81541008e-01 -7.68051028e-01 -7.76286364e-01 8.64909351e-01
1.43999338e-01 -2.94616651e-02 1.07263064e+00 -1.27243266e-01
-3.43611747e-01 -2.56985128e-01 -1.65137398e+00 7.43683338e-01
9.31661606e-01 -5.25087953e-01 -8.20263848e-02 1.24704719e-01
5.99614739e-01 -2.53898382e-01 -6.95041835e-01 2.15528175e-01
6.75764680e-01 -1.28295219e+00 9.15601373e-01 -5.87140262e-01
1.67363688e-01 -1.37793481e-01 -3.44928086e-01 -7.59663343e-01
-2.39768088e-01 -1.16887355e+00 4.77220351e-03 1.35714519e+00
1.92382231e-01 -9.40057039e-01 1.09950101e+00 1.36582875e+00
3.59679997e-01 -6.82478964e-01 -7.70724058e-01 -6.31914079e-01
1.81638658e-01 -4.68880832e-01 8.34535539e-01 1.18342257e+00
-1.78808957e-01 -1.36739433e-01 -5.93715847e-01 2.04992760e-02
7.65496969e-01 -1.87586751e-02 1.14879715e+00 -1.10714698e+00
-7.19603896e-02 -4.25734937e-01 -2.98056304e-01 -8.60613346e-01
-2.83111483e-02 -8.43756139e-01 -4.12939847e-01 -1.15823722e+00
2.33327746e-01 -4.18019623e-01 -1.32771343e-01 5.36156535e-01
4.81527075e-02 3.95873236e-03 3.90528619e-01 3.50788593e-01
-1.52900919e-01 3.80108684e-01 9.17518318e-01 7.05992281e-02
-2.81432960e-02 3.73496354e-01 -1.41623163e+00 3.97501200e-01
6.70150518e-01 -5.32035291e-01 -6.13926709e-01 -1.80800691e-01
3.11296821e-01 -1.68399349e-01 5.37841439e-01 -6.59392357e-01
1.17542341e-01 -4.56725389e-01 1.09243251e-01 -3.40614498e-01
1.31180525e-01 -1.02591705e+00 4.41850066e-01 2.89344728e-01
-5.32374322e-01 -3.48643251e-02 1.11979634e-01 6.00265622e-01
-1.39322102e-01 4.02565971e-02 9.39141572e-01 -2.30562776e-01
-3.50188971e-01 4.78681505e-01 -3.40284944e-01 1.49263859e-01
1.15776241e+00 -2.58201480e-01 -2.00579062e-01 -6.11093879e-01
-5.53062081e-01 1.62023440e-01 8.31712425e-01 3.65634233e-01
3.26980114e-01 -1.08153164e+00 -5.61519682e-01 3.94589841e-01
1.00722030e-01 -2.16820985e-01 3.04813892e-01 6.97480142e-01
-2.39115044e-01 4.26926374e-01 9.17922631e-02 -3.76591533e-02
-1.21609128e+00 8.50761056e-01 3.18571299e-01 -6.41094372e-02
-4.85580057e-01 8.06208551e-01 4.80723768e-01 -3.12503040e-01
4.44412917e-01 -8.37397277e-02 2.22184762e-01 1.02493815e-01
4.55559045e-01 3.68699074e-01 3.82007174e-02 -5.23054719e-01
-4.16030407e-01 -9.75796059e-02 -1.46628320e-01 -9.97407585e-02
1.31406772e+00 -2.88052350e-01 -1.95059359e-01 1.37964457e-01
1.31766713e+00 3.61360580e-01 -1.19355118e+00 -3.26256678e-02
-9.04539898e-02 -6.84976697e-01 -2.08754629e-01 -7.16838777e-01
-1.13963425e+00 8.72827172e-01 4.12194043e-01 4.39670324e-01
1.36621237e+00 -1.88074037e-01 5.81016958e-01 -1.17555670e-01
3.42669904e-01 -9.09853935e-01 -4.85770524e-01 -4.42441888e-02
8.45195711e-01 -1.08131135e+00 2.44174808e-01 -4.39570278e-01
-8.63855481e-01 7.58456051e-01 1.87945709e-01 2.71728426e-01
8.09200466e-01 1.44556850e-01 -1.09224737e-01 -1.14158645e-01
-5.30557990e-01 2.46539190e-01 2.44814251e-02 8.57606351e-01
2.13886991e-01 -3.41161042e-02 -4.35348421e-01 7.87823677e-01
-3.92472237e-01 8.21898058e-02 4.94186461e-01 1.09394217e+00
1.37608409e-01 -1.50270462e+00 -4.01007921e-01 5.28223097e-01
-7.01829851e-01 -9.28144604e-02 -7.47823596e-01 7.48322606e-01
1.36582866e-01 8.89596283e-01 -1.29945144e-01 -4.25374955e-01
2.26065964e-01 2.23953873e-02 1.08096078e-01 -3.80709708e-01
-5.56050360e-01 -2.92049199e-01 -1.29931103e-02 -4.29294050e-01
-2.84307331e-01 -1.09513807e+00 -5.22428274e-01 -6.94103777e-01
3.71790305e-02 2.64557619e-02 5.13957918e-01 5.99042952e-01
8.70063543e-01 -1.61440447e-02 6.28123283e-01 -1.62642971e-01
-5.86942732e-01 -1.85567424e-01 -5.26340604e-01 2.01697499e-01
5.55191875e-01 -2.56788969e-01 -5.04290462e-01 4.08900855e-03]
|
[6.2056660652160645, 6.764781951904297]
|
2305ac34-81a8-487b-b3b6-1986dab0dc1d
|
pseudo-lidar-for-visual-odometry
|
2209.01567
| null |
https://arxiv.org/abs/2209.01567v1
|
https://arxiv.org/pdf/2209.01567v1.pdf
|
Pseudo-LiDAR for Visual Odometry
|
In the existing methods, LiDAR odometry shows superior performance, but visual odometry is still widely used for its price advantage. Conventionally, the task of visual odometry mainly rely on the input of continuous images. However, it is very complicated for the odometry network to learn the epipolar geometry information provided by the images. In this paper, the concept of pseudo-LiDAR is introduced into the odometry to solve this problem. The pseudo-LiDAR point cloud back-projects the depth map generated by the image into the 3D point cloud, which changes the way of image representation. Compared with the stereo images, the pseudo-LiDAR point cloud generated by the stereo matching network can get the explicit 3D coordinates. Since the 6 Degrees of Freedom (DoF) pose transformation occurs in 3D space, the 3D structure information provided by the pseudo-LiDAR point cloud is more direct than the image. Compared with sparse LiDAR, the pseudo-LiDAR has a denser point cloud. In order to make full use of the rich point cloud information provided by the pseudo-LiDAR, a projection-aware dense odometry pipeline is adopted. Most previous LiDAR-based algorithms sampled 8192 points from the point cloud as input to the odometry network. The projection-aware dense odometry pipeline takes all the pseudo-LiDAR point clouds generated from the images except for the error points as the input to the network. While making full use of the 3D geometric information in the images, the semantic information in the images is also used in the odometry task. The fusion of 2D-3D is achieved in an image-only based odometry. Experiments on the KITTI dataset prove the effectiveness of our method. To the best of our knowledge, this is the first visual odometry method using pseudo-LiDAR.
|
['Hesheng Wang', 'Yanzi Miao', 'Xinrui Wu', 'Chaokang Jiang', 'Zhiheng Feng', 'Guangming Wang', 'Huiying Deng']
|
2022-09-04
| null | null | null | null |
['stereo-matching-1']
|
['computer-vision']
|
[-3.68279815e-01 -7.38909394e-02 -2.33996049e-01 -3.62729073e-01
-3.04799415e-02 -1.96006790e-01 3.03706050e-01 -2.22365394e-01
-4.69931096e-01 6.07459307e-01 -2.72544801e-01 -1.59269735e-01
1.47210717e-01 -1.24372411e+00 -7.70960271e-01 -5.61282933e-01
2.85973251e-01 1.13123918e+00 4.99514073e-01 -2.89912462e-01
1.62147149e-01 8.00910056e-01 -1.72862625e+00 -3.47039044e-01
9.18526649e-01 8.53433132e-01 6.55553043e-01 3.62866074e-01
-5.72206438e-01 8.91370848e-02 -1.49459735e-01 8.96248296e-02
6.94922805e-01 5.97177781e-02 -2.15353593e-01 2.18179956e-01
7.08383024e-01 -6.45883441e-01 -4.39454734e-01 1.34385920e+00
4.13365513e-01 -1.70903057e-02 2.93250293e-01 -1.45843351e+00
-4.42899942e-01 -1.67887628e-01 -5.36007643e-01 -4.35397804e-01
4.96713847e-01 1.56544253e-01 5.92524350e-01 -1.15941405e+00
9.79664207e-01 1.36950684e+00 4.63771641e-01 7.05939084e-02
-1.04574406e+00 -8.76651645e-01 -1.74133331e-01 1.32903188e-01
-1.84503055e+00 -2.02119667e-02 9.15387869e-01 -4.57852542e-01
7.82572269e-01 -3.25200766e-01 1.03954446e+00 5.17077744e-01
-3.05524524e-02 2.23960862e-01 9.88940895e-01 -2.86420971e-01
1.15638815e-01 2.87821621e-01 -1.28923595e-01 8.63769531e-01
5.74771821e-01 4.20025617e-01 -5.61054289e-01 1.82448283e-01
1.20505714e+00 7.16730118e-01 -3.68254185e-01 -1.13407135e+00
-8.90434980e-01 8.59105885e-01 1.06562006e+00 -1.67509571e-01
-1.59830600e-01 -5.40040508e-02 -1.82938457e-01 9.25668776e-02
3.05550575e-01 7.18306899e-02 -1.27979293e-01 -2.89745852e-02
-8.38521600e-01 2.33412515e-02 7.00117052e-01 1.33360577e+00
1.79409266e+00 1.53714955e-01 7.06185102e-01 5.20513713e-01
6.51620865e-01 1.04246879e+00 3.10876608e-01 -1.04717922e+00
7.17857838e-01 1.01023912e+00 -6.16684444e-02 -1.17550993e+00
-3.60983402e-01 -2.84274787e-01 -9.57746267e-01 6.77231133e-01
2.82097071e-01 1.32281899e-01 -1.09394181e+00 1.16977513e+00
5.04268825e-01 2.12848246e-01 -4.52627279e-02 1.18677342e+00
7.09932506e-01 4.39274907e-01 -7.20021307e-01 6.04833439e-02
1.02025235e+00 -5.41593313e-01 -5.43109715e-01 -3.93675655e-01
4.29980665e-01 -7.13736534e-01 6.95846021e-01 1.73281897e-02
-5.70356548e-01 -7.66585529e-01 -1.50327182e+00 -3.83468956e-01
-3.96726996e-01 -2.20105365e-01 2.37514004e-01 3.17084223e-01
-1.04414988e+00 2.36916065e-01 -7.28245199e-01 -4.18208778e-01
9.49523374e-02 3.97070467e-01 -5.95351696e-01 -6.15445197e-01
-1.12526286e+00 1.08996403e+00 5.63244104e-01 3.74430493e-02
-5.18149495e-01 -4.91208792e-01 -1.14929235e+00 -1.88791484e-01
4.44950610e-01 -9.24257815e-01 7.45612860e-01 -5.55203974e-01
-1.21562624e+00 8.23414862e-01 -3.63337785e-01 -2.20385164e-01
6.17048502e-01 8.60814527e-02 -1.14705870e-02 2.89236724e-01
3.13134104e-01 9.22398686e-01 4.46993142e-01 -1.47882450e+00
-7.11190820e-01 -7.35518694e-01 2.25037485e-01 4.72242832e-01
2.73179859e-01 -7.70795822e-01 -7.34694958e-01 2.09653288e-01
9.32613492e-01 -1.06970644e+00 -1.29823714e-01 5.09288847e-01
-2.29949027e-01 4.02367473e-01 1.08165860e+00 -1.61046505e-01
5.28983653e-01 -2.21431875e+00 -9.16237757e-02 2.13907376e-01
3.30416769e-01 9.56358090e-02 1.24895751e-01 2.46169925e-01
1.72228754e-01 -2.92716503e-01 -3.69518995e-02 -6.05230808e-01
-2.42107719e-01 7.17221081e-01 -2.11467624e-01 5.65987647e-01
-7.75656942e-03 6.39232635e-01 -9.73269820e-01 -6.72975421e-01
6.13766432e-01 7.09567070e-01 -4.43608463e-01 3.19212042e-02
-2.61317659e-02 4.46975827e-01 -2.66631961e-01 5.91894388e-01
1.23180711e+00 1.15153939e-01 -1.70353025e-01 -2.15242580e-01
-3.98027241e-01 3.34631294e-01 -1.40365362e+00 1.97835720e+00
-3.46300304e-01 4.77013737e-01 -1.73168201e-02 -4.65058714e-01
1.27322340e+00 1.75707296e-01 3.94942284e-01 -7.78766334e-01
-4.70618391e-03 4.64759886e-01 -1.70867160e-01 -1.03120640e-01
6.47481620e-01 -4.29295152e-01 2.17290878e-01 1.54217018e-03
-4.12549153e-02 -8.50790799e-01 1.37637276e-03 1.53083339e-01
4.14885938e-01 4.29240763e-01 3.95313829e-01 2.32241422e-01
4.53901440e-01 4.73003119e-01 7.82791018e-01 3.50195616e-01
1.33882448e-01 8.87360036e-01 -5.00689521e-02 -4.92855400e-01
-1.20613444e+00 -1.27830648e+00 -2.30492577e-01 -1.60445899e-01
6.46126688e-01 -1.99662179e-01 -2.10637391e-01 -2.54758686e-01
3.09908032e-01 2.24830970e-01 -1.21998981e-01 2.78293993e-03
-5.12241602e-01 -1.79405332e-01 2.04890475e-01 4.12494361e-01
9.26131010e-01 -5.86107254e-01 -4.07419771e-01 1.44747058e-02
-1.27744526e-01 -1.24923575e+00 -3.01745236e-01 1.08146876e-01
-1.32224500e+00 -1.04133165e+00 -3.38894993e-01 -6.30978465e-01
7.89420903e-01 8.61586034e-01 5.58034956e-01 1.21207675e-03
3.51548679e-02 -8.00798535e-02 -1.23677989e-02 -4.00000542e-01
-9.07090157e-02 -1.73292488e-01 2.36744270e-01 -3.17103267e-01
6.77832425e-01 -8.60888541e-01 -2.88867384e-01 4.07763600e-01
-6.83607221e-01 2.37150833e-01 4.88133699e-01 6.64855957e-01
7.51115024e-01 -5.35860807e-02 -2.97166348e-01 -4.45417643e-01
-1.47664905e-01 -2.36508206e-01 -9.84294116e-01 -5.96039891e-01
-6.69212759e-01 5.39740138e-02 2.49150485e-01 -1.24795802e-01
-7.75408626e-01 5.93542278e-01 -8.25361116e-04 -9.23125744e-01
-9.13720280e-02 4.22688603e-01 -3.54480833e-01 -1.00958012e-01
5.65439999e-01 1.43227488e-01 5.94330490e-01 -6.19147718e-01
2.49225438e-01 6.90415978e-01 4.68313247e-01 -4.46309820e-02
1.29215896e+00 1.03630662e+00 4.79966015e-01 -8.72320592e-01
-7.45577395e-01 -8.05852056e-01 -1.09625435e+00 3.28690112e-02
7.87308931e-01 -1.30323851e+00 -5.00345886e-01 2.32718542e-01
-1.40019464e+00 3.05760980e-01 -3.59721631e-01 8.46376836e-01
-4.33716208e-01 5.76248884e-01 -2.85118431e-01 -6.79157794e-01
2.98518818e-02 -1.34219515e+00 1.07656717e+00 1.31377608e-01
2.87612319e-01 -8.75862539e-01 2.42170289e-01 1.30801246e-01
-2.28204519e-01 1.59081314e-02 4.41205204e-01 6.30292892e-02
-1.33159447e+00 -2.80932963e-01 -4.39761341e-01 2.31543928e-01
1.13238394e-01 -1.59240738e-01 -9.45182502e-01 -7.00818822e-02
1.08558185e-01 1.70481801e-01 5.45183063e-01 1.88817546e-01
2.55193859e-01 2.72736847e-01 -2.63147831e-01 1.10897803e+00
1.83894217e+00 1.18475750e-01 6.24809206e-01 6.49944961e-01
1.12663496e+00 5.21988392e-01 9.24329162e-01 1.35111094e-01
7.11175740e-01 9.08689499e-01 1.07267189e+00 -1.80171445e-01
-2.57486291e-02 -8.16737175e-01 3.08019519e-01 8.22247624e-01
-1.32889926e-01 4.00342017e-01 -1.04829013e+00 3.10603231e-01
-1.87248433e+00 -8.35672498e-01 -4.30530012e-01 2.48686981e+00
5.04639268e-01 1.61916643e-01 -1.81660935e-01 1.51419342e-01
7.16792345e-01 1.59246609e-01 -3.89822543e-01 -1.95755154e-01
-6.28711889e-03 -6.97011948e-02 8.32394004e-01 8.85188520e-01
-5.06054342e-01 8.41418922e-01 4.77678013e+00 3.34569246e-01
-1.31132364e+00 -4.65390459e-03 -4.91471142e-01 -9.49626639e-02
-7.81630874e-02 3.85936230e-01 -1.32603335e+00 4.05917704e-01
4.69941348e-01 -1.30261276e-02 2.40420297e-01 1.11307061e+00
2.86352456e-01 -3.30489069e-01 -8.87942672e-01 1.31460476e+00
2.02881530e-01 -1.14121222e+00 1.40699416e-01 6.63794875e-01
6.46956086e-01 4.97490525e-01 -1.68629885e-01 4.31233197e-02
3.36215824e-01 -7.18559980e-01 7.81352639e-01 3.76725793e-01
1.04913652e+00 -6.61362827e-01 8.69583368e-01 9.61508691e-01
-1.30007339e+00 1.09537750e-01 -8.69376779e-01 -6.23747766e-01
5.50116718e-01 9.89109337e-01 -1.33094001e+00 9.31316793e-01
6.19163930e-01 9.48278010e-01 -3.80159229e-01 1.10714626e+00
-5.53671241e-01 -2.12505311e-01 -6.71144426e-01 4.44175065e-01
2.48884887e-01 -6.58601522e-01 5.68622530e-01 5.66458941e-01
6.47907794e-01 -3.63212913e-01 5.32024145e-01 9.12050664e-01
9.85954255e-02 -1.88517377e-01 -1.24923158e+00 4.81361419e-01
6.32805347e-01 1.06680977e+00 -2.89277852e-01 -5.24790049e-01
-3.31407517e-01 6.00349486e-01 1.55330598e-01 2.17147365e-01
-3.33198339e-01 -2.77445823e-01 8.70021343e-01 3.69054824e-01
2.00807914e-01 -8.24004054e-01 -3.42616379e-01 -1.16303420e+00
4.78605069e-02 -3.14048439e-01 -1.62222877e-01 -1.23111796e+00
-9.02767718e-01 3.66120845e-01 1.39726877e-01 -1.75423610e+00
-4.27403510e-01 -7.06451833e-01 -2.46584132e-01 1.43171120e+00
-1.79685688e+00 -9.21763539e-01 -9.85704184e-01 4.79628384e-01
1.49046436e-01 2.87375659e-01 5.03448963e-01 3.46252412e-01
-3.51156369e-02 -4.12298441e-01 -2.05927249e-02 1.96167365e-01
8.05454910e-01 -1.13507748e+00 3.51319969e-01 6.65145218e-01
1.63605675e-01 7.04501212e-01 5.44348657e-01 -9.04442430e-01
-1.33939016e+00 -9.46341038e-01 8.69435787e-01 -4.06272411e-01
2.91374743e-01 -3.68299007e-01 -9.30194616e-01 7.32184768e-01
-3.42723668e-01 1.97771057e-01 -3.14372662e-03 -3.00963223e-01
-2.99662143e-01 -3.28033149e-01 -1.12484694e+00 4.48825419e-01
1.26510465e+00 -7.91362405e-01 -6.32432938e-01 -4.75979336e-02
8.45183969e-01 -8.27012897e-01 -5.47893167e-01 2.69138694e-01
5.24453461e-01 -1.07085800e+00 1.27762759e+00 2.16033101e-01
2.45212615e-01 -9.64472175e-01 -1.68024868e-01 -1.37452471e+00
-3.12702768e-02 3.57565358e-02 2.14516237e-01 8.70838702e-01
-1.12380028e-01 -9.33597803e-01 1.00279975e+00 3.08673419e-02
-1.49819508e-01 -2.59849548e-01 -1.08835185e+00 -9.41370368e-01
-3.06973666e-01 -5.06130993e-01 8.48037243e-01 7.87397742e-01
-2.99145520e-01 4.35089439e-01 -1.95571840e-01 4.77355957e-01
7.94039667e-01 1.96264893e-01 1.27880406e+00 -1.66220486e+00
3.97646241e-02 1.32191047e-01 -9.88388956e-01 -1.60862756e+00
-2.71805376e-02 -9.71148014e-01 1.91808909e-01 -1.59886289e+00
-2.53778130e-01 -4.80339289e-01 3.85117918e-01 1.58571690e-01
2.21263006e-01 3.98750842e-01 4.61489171e-01 5.07572472e-01
1.63268685e-01 5.24532855e-01 1.53654909e+00 -6.35891967e-03
-3.10254842e-01 -2.57699043e-01 -6.79667145e-02 8.75487089e-01
5.65289378e-01 -6.37413502e-01 -5.43571353e-01 -6.15410388e-01
2.08149686e-01 1.61262929e-01 4.69865710e-01 -1.01915836e+00
4.99273509e-01 -1.55063495e-01 3.34188640e-01 -1.30048728e+00
9.20845807e-01 -1.24248242e+00 4.07107890e-01 5.73827267e-01
7.30183601e-01 6.92867488e-02 -1.79189220e-01 4.50358063e-01
-4.65650290e-01 -3.06438178e-01 7.18298197e-01 -3.63161922e-01
-7.07909644e-01 6.97758913e-01 9.63189378e-02 -3.11616331e-01
7.07504392e-01 -9.73337233e-01 -3.16078216e-01 -3.47031176e-01
-3.31111193e-01 2.34967977e-01 1.23053217e+00 2.27197826e-01
9.23310399e-01 -1.43274307e+00 -1.47852227e-01 6.93021059e-01
2.89227784e-01 9.87130165e-01 -4.37195823e-02 1.01208913e+00
-8.67626369e-01 5.14456034e-01 -4.52471852e-01 -1.14427614e+00
-1.16614163e+00 6.53487265e-01 3.75535816e-01 1.44988850e-01
-8.48852575e-01 9.91746858e-02 3.55250835e-01 -6.96572483e-01
-1.52092084e-01 -4.33604062e-01 1.59075744e-02 -6.89783767e-02
2.34760717e-01 3.66466671e-01 -5.57151027e-02 -1.02932811e+00
-2.42007881e-01 1.13510144e+00 3.10453475e-01 -4.56333548e-01
1.04359949e+00 -3.57543528e-01 -1.23055995e-01 6.32480681e-01
1.43204212e+00 4.00489084e-02 -1.29541004e+00 -3.98538202e-01
-2.41447046e-01 -9.16368902e-01 9.44064632e-02 -4.78655137e-02
-9.50875759e-01 1.40443766e+00 4.16017234e-01 -2.26365551e-01
6.23915315e-01 -2.94304281e-01 5.70982814e-01 4.99894142e-01
1.02325964e+00 -8.65682900e-01 -2.21117005e-01 8.74249578e-01
6.68288291e-01 -1.29303694e+00 2.43909448e-01 -7.17948139e-01
-3.48715454e-01 1.12392414e+00 7.12660611e-01 -1.86329007e-01
5.37545621e-01 1.03259712e-01 3.36105168e-01 -1.83383405e-01
-1.19138397e-01 -4.62701261e-01 -5.12721017e-02 9.25236642e-01
-2.17681423e-01 -1.60215423e-01 1.98126212e-01 -4.56597894e-01
-3.53778541e-01 2.27405131e-01 6.27761960e-01 1.01393318e+00
-8.15750182e-01 -1.22168648e+00 -7.50749767e-01 -4.66307327e-02
4.91068929e-01 1.51376262e-01 -1.93365201e-01 9.32406664e-01
4.13377017e-01 5.40588975e-01 3.68334472e-01 -3.56558710e-01
4.95620549e-01 3.34324650e-02 3.81393760e-01 -9.38255966e-01
9.03293937e-02 -1.15528598e-01 -2.20474094e-01 -6.72180116e-01
-3.71279210e-01 -3.21513504e-01 -1.73611224e+00 -4.52202201e-01
-2.94134229e-01 6.70182481e-02 1.21393728e+00 7.23494589e-01
2.99726635e-01 2.43397728e-02 4.67821032e-01 -1.23253131e+00
-3.27017128e-01 -7.75752962e-01 -8.35060835e-01 2.67112762e-01
4.25549507e-01 -1.12341213e+00 -5.61125457e-01 -3.07018071e-01]
|
[7.543024063110352, -2.332383871078491]
|
068c88db-e491-45e4-8165-e92451e4c87a
|
proportional-aggregation-of-preferences-for
|
2306.14858
| null |
https://arxiv.org/abs/2306.14858v1
|
https://arxiv.org/pdf/2306.14858v1.pdf
|
Proportional Aggregation of Preferences for Sequential Decision Making
|
We study the problem of fair sequential decision making given voter preferences. In each round, a decision rule must choose a decision from a set of alternatives where each voter reports which of these alternatives they approve. Instead of going with the most popular choice in each round, we aim for proportional representation. We formalize this aim using axioms based on Proportional Justified Representation (PJR), which were proposed in the literature on multi-winner voting and were recently adapted to multi-issue decision making. The axioms require that every group of $\alpha\%$ of the voters, if it agrees in every round (i.e., approves a common alternative), then those voters must approve at least $\alpha\%$ of the decisions. A stronger version of the axioms requires that every group of $\alpha\%$ of the voters that agrees in a $\beta$ fraction of rounds must approve $\beta\cdot\alpha\%$ of the decisions. We show that three attractive voting rules satisfy axioms of this style. One of them (Sequential Phragm\'en) makes its decisions online, and the other two satisfy strengthened versions of the axioms but make decisions semi-online (Method of Equal Shares) or fully offline (Proportional Approval Voting). The first two are polynomial-time computable, and the latter is based on an NP-hard optimization, but it admits a polynomial-time local search algorithm that satisfies the same axiomatic properties. We present empirical results about the performance of these rules based on synthetic data and U.S. political elections. We also run experiments where votes are cast by preference models trained on user responses from the moral machine dataset about ethical dilemmas.
|
['Dominik Peters', 'Shashwat Goel', 'Nikhil Chandak']
|
2023-06-26
| null | null | null | null |
['decision-making']
|
['reasoning']
|
[ 2.65125781e-01 6.90001667e-01 -4.54697758e-01 -6.19032204e-01
-6.59496903e-01 -8.94303381e-01 4.39998716e-01 2.10821316e-01
-1.04421461e+00 9.95259047e-01 1.45602167e-01 -8.62059653e-01
-5.12343466e-01 -9.99357104e-01 -3.49561244e-01 -7.05638111e-01
2.26520792e-01 1.11465478e+00 -1.14818193e-01 -4.87981945e-01
4.20789629e-01 -5.83677478e-02 -1.30562687e+00 2.81606734e-01
7.51541853e-01 1.19106102e+00 -5.47851205e-01 5.29252350e-01
2.44490832e-01 7.56066442e-01 -1.34885293e-02 -7.89224088e-01
9.91396844e-01 -2.90997863e-01 -1.19064903e+00 -3.81363213e-01
1.93808049e-01 -5.30750096e-01 3.38181965e-02 1.15095615e+00
2.98184007e-01 2.27054536e-01 9.19979036e-01 -1.27826834e+00
-5.73189259e-01 1.12231040e+00 -6.61756396e-01 -2.49338999e-01
3.61919224e-01 -8.02497938e-03 1.65302479e+00 -4.55084652e-01
6.68623269e-01 1.08157134e+00 1.65261775e-01 8.97587001e-01
-1.45353770e+00 -6.33218229e-01 1.59896612e-01 -1.14527024e-01
-1.04247570e+00 -3.76503289e-01 2.44482085e-01 -2.60215551e-01
5.02547324e-01 6.79286480e-01 3.84225845e-01 5.03730595e-01
1.06514767e-01 4.02394742e-01 1.92376924e+00 -6.38570428e-01
7.04729378e-01 1.79944895e-02 5.96744061e-01 1.93312988e-01
5.63664973e-01 3.51074547e-01 -3.03364486e-01 -7.50093102e-01
1.59795180e-01 1.58607110e-03 -1.76624507e-01 -2.77396441e-01
-8.23988259e-01 1.19878483e+00 2.18102053e-01 1.53496772e-01
-3.47314447e-01 1.03373937e-01 9.01824906e-02 7.50248849e-01
1.66473612e-01 3.73376191e-01 -4.95878607e-01 2.78441101e-01
-7.76839375e-01 5.92880607e-01 1.14576769e+00 5.08036077e-01
9.69626546e-01 -7.19963253e-01 -1.94691464e-01 3.59552205e-01
3.88158023e-01 5.05594492e-01 1.80832073e-01 -1.56627929e+00
4.78086144e-01 5.92403352e-01 7.85237730e-01 -7.16761708e-01
-3.97464037e-01 2.10790291e-01 -9.52753901e-01 7.50849426e-01
9.30559814e-01 -2.60154843e-01 -4.95652348e-01 2.13209057e+00
3.50893617e-01 -9.34229791e-01 -1.27583191e-01 9.00351644e-01
3.58458012e-02 5.69204152e-01 2.57815551e-02 -6.73088372e-01
1.60215759e+00 -2.04140171e-01 -2.90545225e-01 -9.14231688e-03
4.61003333e-01 -7.37061918e-01 7.53167689e-01 5.38660169e-01
-1.31187272e+00 4.72046435e-02 -7.22398579e-01 1.17731147e-01
2.11115882e-01 -1.84973016e-01 7.17589438e-01 8.45147312e-01
-1.25337839e+00 5.27703762e-01 7.27399252e-03 -1.83799073e-01
1.79458022e-01 6.88685834e-01 -4.00086015e-01 -3.52733186e-03
-1.16390288e+00 9.53954995e-01 -2.54003346e-01 -2.15020016e-01
-4.68356997e-01 -1.13837987e-01 -5.00195265e-01 1.29658625e-01
5.81092894e-01 -8.31337988e-01 1.66708517e+00 -1.43777430e+00
-1.58004665e+00 1.40673971e+00 -1.05864607e-01 -4.52627540e-01
1.03240252e+00 2.63330102e-01 -2.29624063e-02 -2.70836741e-01
4.84532565e-01 4.60679024e-01 3.61953557e-01 -8.34327340e-01
-1.00488329e+00 -5.95618308e-01 7.72555172e-01 3.04781467e-01
7.79144019e-02 3.51193964e-01 4.23860818e-01 9.12099984e-03
5.84669076e-02 -1.26267159e+00 -6.89856648e-01 2.23914087e-02
-2.50524402e-01 -5.10214925e-01 -1.70674205e-01 -9.01276097e-02
1.09652960e+00 -1.79728532e+00 -1.76087096e-01 7.53834546e-01
1.93532690e-01 -2.63948739e-01 -9.66327488e-02 3.86028558e-01
-4.70528826e-02 5.24234951e-01 -2.75997907e-01 -6.90487176e-02
6.01777375e-01 1.62660152e-01 -5.28427660e-01 7.02657640e-01
-6.75738335e-01 4.81045425e-01 -5.92660725e-01 -1.36942402e-01
-2.05415562e-01 -5.81112862e-01 -1.02164495e+00 -2.06775084e-01
-5.48462085e-02 -3.34055535e-02 -4.07188952e-01 -3.34388502e-02
8.64186525e-01 -1.91707447e-01 8.28236043e-01 3.70367199e-01
-2.08323896e-01 5.77380717e-01 -1.64867747e+00 8.60791206e-01
-1.91519558e-01 -1.42012490e-03 4.21300232e-01 -7.49112129e-01
3.27891558e-01 3.96401733e-01 2.18945056e-01 -7.47628272e-01
5.32803178e-01 4.87285674e-01 1.27183378e-01 2.14896277e-01
5.88432014e-01 -6.38058662e-01 -6.84861600e-01 1.37078297e+00
-5.37711918e-01 -1.15327172e-01 2.52667278e-01 4.64526147e-01
1.03174829e+00 -3.98116380e-01 8.53237212e-01 -7.38635480e-01
4.98232335e-01 5.16062742e-03 1.07615685e+00 1.07956147e+00
-2.67251760e-01 1.79318741e-01 6.98493481e-01 -8.83758187e-01
-1.03746629e+00 -7.94982195e-01 1.48397580e-01 1.47505045e+00
3.54875803e-01 -2.20252186e-01 -4.90389884e-01 -6.69154108e-01
1.21930622e-01 9.77840662e-01 -7.89774120e-01 3.84920567e-01
-3.56944829e-01 -6.12499356e-01 7.68883079e-02 2.29451433e-01
4.14582998e-01 -7.42255032e-01 -9.30259645e-01 5.41137122e-02
-4.30314898e-01 -4.32704538e-01 -7.99483776e-01 3.40860635e-01
-5.24562895e-01 -1.42353082e+00 -2.97633350e-01 -1.97273314e-01
9.95646119e-01 2.41962582e-01 1.01291537e+00 9.51393768e-02
1.82846829e-01 1.35426437e-02 -2.13467121e-01 -2.16015130e-01
-4.73592013e-01 -2.93968767e-01 6.15027070e-01 9.18800309e-02
4.60392386e-01 -4.01488066e-01 -8.20008457e-01 5.92496753e-01
-6.96204960e-01 -1.26157954e-01 2.52077132e-01 6.76802337e-01
5.33596814e-01 -3.24478507e-01 1.03955857e-01 -1.22071004e+00
7.85565197e-01 -2.91630536e-01 -9.44847226e-01 2.23078817e-01
-8.01183999e-01 3.47257555e-01 8.83488417e-01 -1.07667297e-01
-7.89052904e-01 -6.64678514e-02 1.67253420e-01 4.77495253e-01
1.40232921e-01 1.99557289e-01 -7.65414312e-02 1.76315725e-01
9.26358044e-01 -2.77370751e-01 -2.44247392e-01 -1.57387897e-01
5.86834311e-01 5.72282851e-01 2.01623514e-01 -9.86390114e-01
6.63850605e-01 5.31139612e-01 -3.18530644e-03 9.65560414e-03
-6.15110397e-01 -1.02418065e-01 3.86083312e-02 5.22635542e-02
6.43061280e-01 -5.87340057e-01 -1.58497918e+00 4.63369675e-02
-9.81830060e-01 -2.95017838e-01 -6.00075126e-01 3.64136517e-01
-6.60663068e-01 5.39610684e-01 -4.22900438e-01 -1.22235906e+00
-3.11760277e-01 -1.01168907e+00 2.40328252e-01 2.95054227e-01
-6.93086326e-01 -1.98671833e-01 3.17998797e-01 5.82559049e-01
3.77585679e-01 -1.67723596e-01 9.57459807e-01 -7.70491898e-01
-3.23866457e-01 -2.06955791e-01 -8.31141397e-02 1.53638557e-01
-1.81434020e-01 -1.78947113e-02 -5.81907988e-01 -4.06144887e-01
-6.39462546e-02 -4.96498644e-01 8.15428972e-01 5.33171117e-01
7.41409540e-01 -9.18653429e-01 -1.44004181e-01 -1.41995892e-01
1.30782139e+00 2.03639537e-01 5.11525571e-01 1.94797158e-01
-4.03602540e-01 6.87796354e-01 3.77233654e-01 6.10198140e-01
8.11783254e-01 6.88032150e-01 4.11214471e-01 3.80409122e-01
7.70191371e-01 -6.05904870e-02 3.63076687e-01 1.81168899e-01
-6.04814172e-01 -2.36243159e-02 -5.08545518e-01 4.33918238e-01
-2.09076571e+00 -1.11758709e+00 -5.54052964e-02 2.63144064e+00
1.04007053e+00 8.31204802e-02 2.99939185e-01 1.60307750e-01
6.98643088e-01 -1.17223896e-01 -3.74458730e-01 -1.26308763e+00
-1.68524116e-01 6.64620757e-01 7.88450837e-01 8.60656142e-01
-8.43268871e-01 5.54949164e-01 5.85450506e+00 6.38089418e-01
-5.71748018e-01 2.00609401e-01 1.25324285e+00 -3.45674753e-01
-8.96501243e-01 6.69735909e-01 -6.04119122e-01 4.82502282e-01
5.51838577e-01 -5.04935145e-01 6.31513536e-01 9.35154080e-01
2.19885707e-02 -6.59280062e-01 -1.37317157e+00 6.57072067e-01
-4.70908433e-01 -1.23057783e+00 -2.38467023e-01 3.71584922e-01
1.07900727e+00 -3.41774434e-01 1.62003711e-01 1.80952057e-01
1.61758661e+00 -1.05228782e+00 1.14465439e+00 8.34654644e-02
9.21329677e-01 -8.01248550e-01 6.95581853e-01 8.60287309e-01
-7.84074247e-01 -5.32814860e-01 -4.55147028e-01 -7.13496447e-01
3.53891924e-02 6.25971377e-01 -1.27344295e-01 5.91160119e-01
5.54516912e-01 -3.75466406e-01 2.45784789e-01 5.44370174e-01
-5.50410867e-01 1.99766099e-01 -6.33771420e-01 -3.71312685e-02
2.65433669e-01 -4.46267635e-01 1.75428048e-01 4.92713392e-01
3.23174089e-01 8.77420843e-01 1.04253531e-01 6.26081645e-01
-4.77250069e-01 1.53892651e-01 -4.19308543e-01 6.42522395e-01
6.95246339e-01 1.21976519e+00 -3.66393656e-01 -1.25146776e-01
-8.09922069e-02 6.59209847e-01 2.63565749e-01 6.67391792e-02
-4.53134537e-01 6.57399818e-02 7.37252593e-01 2.04037517e-01
-1.02217592e-01 2.44566247e-01 -3.85797203e-01 -1.14300239e+00
-4.81883548e-02 -1.07011402e+00 1.00749803e+00 -3.29816520e-01
-1.44231880e+00 2.51619458e-01 -1.11057967e-01 -8.66355538e-01
-2.78347611e-01 -5.68958342e-01 -9.17898536e-01 8.63188922e-01
-1.25492382e+00 -5.26722491e-01 3.46755505e-01 6.34033144e-01
-6.95609748e-02 1.91656560e-01 8.56891274e-01 -2.33247399e-01
-1.32177845e-01 5.22748947e-01 3.30403075e-02 -2.13367313e-01
6.66877627e-01 -1.20914137e+00 -2.50848066e-02 6.77312911e-01
-1.26236141e-01 7.71407664e-01 7.05804050e-01 -4.15227599e-02
-1.01003373e+00 -5.38717866e-01 1.65104079e+00 -3.75401109e-01
2.11602554e-01 -1.55486129e-02 -3.18946876e-02 7.10108042e-01
4.04027104e-01 -2.68436760e-01 1.04224360e+00 2.03339189e-01
-5.08905947e-01 -3.38363260e-01 -1.44196320e+00 7.91547656e-01
8.46400559e-01 -3.48347425e-01 -7.69186974e-01 -3.93787473e-02
-1.57421947e-01 4.64758240e-02 -1.56415597e-01 2.91711688e-01
8.55720758e-01 -1.16887867e+00 4.14726883e-01 -7.84125268e-01
3.61438066e-01 -4.99319166e-01 -5.08686781e-01 -1.11212599e+00
-5.85039735e-01 -9.30927336e-01 4.73745704e-01 7.04676688e-01
7.06041932e-01 -1.03907943e+00 5.72134376e-01 1.41667390e+00
5.47818661e-01 -6.35330260e-01 -1.62389469e+00 -2.64172435e-01
6.49902105e-01 -1.26918197e-01 5.62248707e-01 8.16879690e-01
5.66147447e-01 1.71251297e-01 -4.85931784e-01 -3.62262070e-01
6.97185516e-01 5.60909271e-01 7.55176783e-01 -1.33790326e+00
-5.76302290e-01 -6.85358346e-01 2.82508880e-01 -1.04950237e+00
-3.69140729e-02 -7.88926780e-01 -1.34651691e-01 -1.44010103e+00
5.77669263e-01 -7.31333315e-01 -2.61325210e-01 7.70547330e-01
-3.26681733e-02 5.84998541e-02 2.71048993e-01 3.24305356e-01
-6.65921986e-01 7.70031810e-02 7.42710769e-01 -2.24442974e-01
6.04590401e-02 3.20973188e-01 -1.45167255e+00 9.16633666e-01
6.89666927e-01 -5.75101554e-01 -5.99618303e-03 -1.27297059e-01
1.19652271e+00 3.29612941e-01 1.32583678e-01 -2.65116274e-01
5.23663163e-01 -7.58514583e-01 3.70078199e-02 -2.07658082e-01
-1.15567103e-01 -7.39072502e-01 3.35179806e-01 8.05445969e-01
-7.24101841e-01 2.19526008e-01 -5.40022731e-01 3.53823990e-01
2.63633102e-01 -1.01893395e-01 1.13647735e+00 -2.46436968e-01
-1.05483703e-01 1.71326712e-01 -7.07311213e-01 -3.30462083e-02
1.14654493e+00 -1.22302681e-01 -4.31257367e-01 -7.02828944e-01
-5.72430253e-01 4.84819114e-01 8.19257617e-01 -1.44308999e-01
2.03230232e-01 -1.23193896e+00 -9.35617030e-01 -1.64387286e-01
1.32781640e-01 -3.56064618e-01 1.66014373e-01 9.66735840e-01
-2.24070668e-01 2.51341909e-01 -2.14940310e-01 1.96027309e-01
-1.23022568e+00 3.79711211e-01 3.18051487e-01 -7.45248795e-01
2.53210783e-01 8.59481096e-01 5.58453858e-01 -6.94280088e-01
-2.58671552e-01 -2.05175430e-01 -2.25454010e-02 3.26334685e-02
3.87569398e-01 4.44278359e-01 -3.38387907e-01 -5.71960628e-01
-6.38992727e-01 4.63798642e-01 -1.23205975e-01 -6.20989799e-01
1.09558499e+00 -1.37271866e-01 -5.35867929e-01 -1.13471955e-01
3.28302324e-01 4.04639572e-01 -7.10915625e-01 -4.30895865e-01
-2.45860040e-01 -5.03308594e-01 -4.84178692e-01 -8.98636699e-01
-7.48171210e-01 4.37253892e-01 1.27791896e-01 2.53354371e-01
9.55022156e-01 -1.98939458e-01 1.97104469e-01 5.04014432e-01
1.17331731e+00 -1.57215357e+00 -5.27632475e-01 4.95787710e-01
5.34963548e-01 -9.12229717e-01 1.32680327e-01 7.35929757e-02
-8.34809840e-01 9.44828153e-01 1.67314366e-01 -2.56511211e-01
6.16116524e-01 -1.52480289e-01 -3.28246839e-02 1.99475169e-01
-1.10609043e+00 6.84032738e-02 -7.46773258e-02 1.45192733e-02
2.22530112e-01 7.72659361e-01 -1.29011726e+00 1.08758605e+00
-2.66359866e-01 -1.27537651e-02 6.26105428e-01 6.48669600e-01
-5.53688586e-01 -1.62756610e+00 -3.22761655e-01 5.17332256e-01
-6.20325327e-01 -1.87070832e-01 -7.37810612e-01 3.95401388e-01
3.02262127e-01 1.34216321e+00 -2.83808172e-01 -2.33191475e-01
1.94726348e-01 -9.91986096e-02 4.02885467e-01 -5.01744926e-01
-9.59599376e-01 -3.70922327e-01 2.22231939e-01 -4.78421420e-01
-3.40751231e-01 -6.85993791e-01 -1.02092028e+00 -1.26039743e+00
-3.54029536e-01 5.30782700e-01 8.47776979e-02 8.40739250e-01
1.23786896e-01 -5.36157668e-01 1.13463449e+00 -3.19426835e-01
-1.42281711e+00 -6.34896100e-01 -7.98395693e-01 3.86890829e-01
-1.51063219e-01 -1.62399754e-01 -6.67397499e-01 -3.48353446e-01]
|
[8.921228408813477, 5.417226314544678]
|
81c4b2f9-90b9-45b4-8a4e-0b51f609e18d
|
context-aware-polyunet-for-liver-and-lesion
|
2106.1133
| null |
https://arxiv.org/abs/2106.11330v1
|
https://arxiv.org/pdf/2106.11330v1.pdf
|
Context-aware PolyUNet for Liver and Lesion Segmentation from Abdominal CT Images
|
Accurate liver and lesion segmentation from computed tomography (CT) images are highly demanded in clinical practice for assisting the diagnosis and assessment of hepatic tumor disease. However, automatic liver and lesion segmentation from contrast-enhanced CT volumes is extremely challenging due to the diversity in contrast, resolution, and quality of images. Previous methods based on UNet for 2D slice-by-slice or 3D volume-by-volume segmentation either lack sufficient spatial contexts or suffer from high GPU computational cost, which limits the performance. To tackle these issues, we propose a novel context-aware PolyUNet for accurate liver and lesion segmentation. It jointly explores structural diversity and consecutive t-adjacent slices to enrich feature expressive power and spatial contextual information while avoiding the overload of GPU memory consumption. In addition, we utilize zoom out/in and two-stage refinement strategy to exclude the irrelevant contexts and focus on the specific region for the fine-grained segmentation. Our method achieved very competitive performance at the MICCAI 2017 Liver Tumor Segmentation (LiTS) Challenge among all tasks with a single model and ranked the $3^{rd}$, $12^{th}$, $2^{nd}$, and $5^{th}$ places in the liver segmentation, lesion segmentation, lesion detection, and tumor burden estimation, respectively.
|
['Simon Chun-Ho Yu', 'Liping Zhang']
|
2021-06-21
| null | null | null | null |
['liver-segmentation']
|
['medical']
|
[ 5.96989170e-02 -2.46972367e-01 -2.01871529e-01 -2.34579831e-01
-9.40716088e-01 -4.45281327e-01 2.53503025e-01 4.38174039e-01
-3.38377774e-01 4.02716458e-01 1.86633021e-01 -6.00654781e-01
-3.11789513e-01 -5.18116713e-01 -6.38059974e-02 -9.47404861e-01
-3.18519086e-01 5.91030121e-01 5.34352839e-01 2.30320886e-01
-8.49227756e-02 5.96118867e-01 -6.50547564e-01 1.30422354e-01
1.10586715e+00 9.97319341e-01 4.13427055e-01 4.54521626e-01
-2.16100618e-01 4.78419781e-01 -1.01023503e-01 -2.77124513e-02
3.09289008e-01 -8.01205814e-01 -6.60413742e-01 4.14753884e-01
2.03054592e-01 -1.51941150e-01 -1.47192091e-01 1.17105174e+00
6.16586924e-01 -1.04855254e-01 6.71816230e-01 -7.08808243e-01
9.01313499e-02 5.31300604e-01 -9.03385520e-01 6.20626509e-01
-2.58165151e-01 3.95583540e-01 3.89036119e-01 -1.01871300e+00
4.31426555e-01 4.09661889e-01 6.03569388e-01 2.46644765e-01
-9.30108547e-01 -4.85008299e-01 1.25270084e-01 -4.84610721e-02
-1.36898148e+00 1.09033585e-01 4.79499012e-01 -5.22679090e-01
6.34225011e-01 6.37294888e-01 9.97610450e-01 1.65648833e-01
2.09803715e-01 6.74480617e-01 1.10305381e+00 -2.49785081e-01
-4.13390100e-02 -3.28161925e-01 3.74950911e-03 1.12974131e+00
2.42524490e-01 2.23573763e-02 1.25119239e-01 -1.57819614e-01
1.03610146e+00 1.46330670e-01 -5.77393055e-01 -6.41139388e-01
-1.61239445e+00 7.48640299e-01 8.51575732e-01 3.82339925e-01
-4.98034656e-01 7.03400895e-02 6.15749300e-01 -3.16446930e-01
3.03581685e-01 3.54702294e-01 -2.45198518e-01 1.85607091e-01
-1.18662155e+00 -1.55194998e-01 4.92313236e-01 7.58729279e-01
1.50791183e-01 -8.00313726e-02 -5.17856717e-01 5.03196657e-01
4.89575267e-02 2.37030506e-01 6.43107951e-01 -3.61806601e-01
7.78045580e-02 5.80626011e-01 -1.34241357e-01 -3.89273345e-01
-9.55016077e-01 -7.96257555e-01 -1.31493258e+00 -9.02362391e-02
5.15782058e-01 -3.31824832e-02 -1.39393771e+00 1.12604523e+00
7.81598687e-01 2.25406513e-01 -5.91796279e-01 1.34750545e+00
1.01542282e+00 2.16302127e-01 4.43017453e-01 -5.43045700e-01
1.88327873e+00 -1.27716506e+00 -2.98905373e-01 1.67054348e-02
8.69170189e-01 -8.67799759e-01 9.92018878e-01 -1.71652455e-02
-1.09604025e+00 2.18921527e-02 -7.75310576e-01 2.26812512e-01
1.57352135e-01 2.37450778e-01 8.30206215e-01 6.08963490e-01
-7.20197797e-01 3.13572884e-01 -1.19777083e+00 -6.71305731e-02
6.44960165e-01 4.96909142e-01 -4.08419184e-02 -2.17724442e-01
-7.14170873e-01 7.85160124e-01 2.39297688e-01 -1.39990486e-02
-9.05021608e-01 -1.19534683e+00 -7.52975464e-01 -3.16535980e-02
7.42403507e-01 -7.83705592e-01 1.01701725e+00 -4.19560015e-01
-1.26134050e+00 7.83798218e-01 -4.72081415e-02 -4.03492838e-01
7.47581244e-01 5.19612968e-01 1.63579432e-04 2.43681386e-01
1.15413284e-02 5.07082224e-01 5.55826247e-01 -7.48400509e-01
-6.39873028e-01 -5.59553742e-01 -4.09305215e-01 4.97049570e-01
1.96624860e-01 -1.68219656e-02 -6.35176122e-01 -7.84774363e-01
6.24544382e-01 -8.91761601e-01 -7.36942291e-01 4.14142571e-02
-3.16717863e-01 7.75143802e-02 6.52825177e-01 -8.01719964e-01
1.18413448e+00 -1.87641776e+00 5.74154332e-02 4.04510379e-01
5.01612127e-01 5.27676679e-02 3.93097073e-01 -5.06264567e-01
9.10619199e-02 1.83695391e-01 -3.39414686e-01 7.25179911e-02
-1.68773323e-01 -1.96368862e-02 4.75889534e-01 7.83948183e-01
-2.49333933e-01 1.03735209e+00 -9.55920398e-01 -1.00678933e+00
3.02902788e-01 3.51578027e-01 -5.56473076e-01 -7.27047175e-02
-1.34619087e-01 1.05667877e+00 -6.01280510e-01 7.22280204e-01
6.57247901e-01 -5.70943356e-01 3.02810401e-01 -4.66995478e-01
-3.10729980e-01 1.77217305e-01 -8.81808162e-01 1.64206004e+00
-4.57179725e-01 5.16611040e-02 4.05348718e-01 -6.26893699e-01
4.12781984e-01 4.77676839e-01 1.05229557e+00 -8.07435691e-01
2.26198345e-01 4.51754361e-01 4.03561562e-01 -4.19873953e-01
8.66009109e-03 -3.35063785e-01 4.21775766e-02 4.11869913e-01
-4.57493037e-01 -5.76466739e-01 3.30379903e-01 -9.29927528e-02
8.93902540e-01 -1.44941077e-01 5.18621325e-01 -6.50981188e-01
5.45044482e-01 2.87027895e-01 6.69632375e-01 4.68961447e-01
-5.00716209e-01 5.83678067e-01 4.96825337e-01 -6.39859140e-01
-9.11776364e-01 -8.89403164e-01 -4.27429229e-01 8.39512050e-01
2.82756060e-01 -1.05365142e-01 -7.52502680e-01 -1.03771245e+00
-3.60204160e-01 2.18380138e-01 -5.23305833e-01 3.50877225e-01
-9.17198002e-01 -1.25399458e+00 2.27084875e-01 4.77518827e-01
5.59732914e-01 -7.63690174e-01 -1.06423664e+00 3.71286392e-01
-2.85672605e-01 -9.48763788e-01 -9.02657092e-01 3.23923975e-01
-1.10579264e+00 -1.15514660e+00 -1.10408258e+00 -8.72912347e-01
1.00363231e+00 2.35193893e-01 1.31344748e+00 3.67149264e-01
-7.33354390e-01 -3.06294292e-01 -1.05266586e-01 -1.35873139e-01
-6.20589443e-02 1.32171974e-01 -3.72834891e-01 -4.34599400e-01
-2.06104536e-02 -2.39038423e-01 -1.08142269e+00 4.79970008e-01
-6.90761149e-01 4.06837910e-01 6.98415816e-01 1.07633531e+00
1.05307364e+00 9.87354964e-02 7.30758831e-02 -9.01362538e-01
1.99376792e-01 -2.41033345e-01 -6.87020004e-01 2.83776939e-01
-3.74943793e-01 -1.69273898e-01 4.64850277e-01 -3.04664046e-01
-7.36932099e-01 3.48118305e-01 -1.09466381e-01 -3.03069770e-01
-2.85427589e-02 5.76429129e-01 2.35934585e-01 -3.96260172e-01
5.43969333e-01 5.17365456e-01 -2.32366458e-01 -2.17673317e-01
-1.24037310e-01 -5.00534549e-02 3.84039998e-01 -3.40074837e-01
4.04672354e-01 5.15744090e-01 3.74135971e-01 -4.50924337e-01
-5.92208862e-01 -5.10517359e-01 -5.89465916e-01 4.46596704e-02
8.13519299e-01 -6.93140984e-01 -4.08609301e-01 1.77080527e-01
-5.66669166e-01 -3.05484802e-01 -4.22249228e-01 8.28927517e-01
-1.85494527e-01 4.73751992e-01 -7.76053965e-01 -2.33491257e-01
-7.48768687e-01 -1.96562684e+00 8.21742058e-01 2.80272067e-01
3.66110168e-02 -8.59830320e-01 -4.13276404e-01 1.04940742e-01
7.93323576e-01 4.60031539e-01 1.28881729e+00 -5.71884274e-01
-8.22683692e-01 -4.52055223e-02 -5.68964839e-01 -1.85658321e-01
9.00082439e-02 -3.44861567e-01 -1.86480522e-01 -4.03129339e-01
3.33140716e-02 8.84272307e-02 7.72909880e-01 9.72119391e-01
1.34873390e+00 -1.50551289e-01 -3.94351810e-01 1.05262208e+00
1.34319317e+00 1.31984204e-01 1.27625763e-01 -7.58680105e-02
5.96044540e-01 1.72644645e-01 3.05880517e-01 5.33944547e-01
2.22884282e-01 6.47563398e-01 6.18133426e-01 -5.10058701e-01
-3.69560897e-01 1.90985277e-01 -4.22995269e-01 6.36633277e-01
1.56057114e-02 1.52572662e-01 -1.23004079e+00 7.64592350e-01
-1.26491237e+00 -2.88091362e-01 -3.18033546e-01 2.14409184e+00
8.08049381e-01 -4.71689068e-02 1.70225710e-01 -1.77394331e-01
6.30495787e-01 -9.10194889e-02 -4.84300971e-01 2.23444447e-01
2.31686518e-01 3.26526463e-01 6.26158059e-01 4.51356560e-01
-1.18469489e+00 5.68196774e-01 4.98954582e+00 1.05123627e+00
-1.36921000e+00 3.10022473e-01 9.32044148e-01 -1.98590145e-01
-1.92938641e-01 -5.97508810e-02 -5.01389027e-01 3.91112596e-01
8.74567851e-02 -1.09233428e-02 5.51052317e-02 6.93703890e-01
1.55543700e-01 -5.05126953e-01 -9.89897072e-01 9.76496518e-01
-2.37207368e-01 -1.46525228e+00 -1.31985322e-01 1.13167174e-01
8.26835752e-01 1.65794685e-01 9.90068391e-02 1.34833395e-01
5.11323512e-02 -1.12779427e+00 3.45167249e-01 3.01365461e-02
9.85629380e-01 -5.74813306e-01 7.52160966e-01 4.32616115e-01
-1.38204420e+00 2.35818610e-01 5.72730694e-03 5.34368873e-01
1.07996300e-01 6.79593503e-01 -1.08923662e+00 5.61087966e-01
4.71787721e-01 1.35739475e-01 -5.42546272e-01 1.43403280e+00
1.34217158e-01 3.45482320e-01 -6.02908075e-01 5.53381816e-02
4.29368764e-01 -1.42975867e-01 5.67932487e-01 1.32635820e+00
5.12034595e-01 5.51800847e-01 6.96566105e-01 4.56942528e-01
-6.27491847e-02 2.93798327e-01 1.59517646e-01 5.18509150e-01
1.95323408e-01 1.49488199e+00 -1.46063840e+00 -4.79991555e-01
-2.22522900e-01 7.25343645e-01 -2.35985637e-01 -6.39548302e-02
-1.00248873e+00 9.66524780e-02 -7.17335641e-02 5.09760737e-01
2.27779046e-01 -9.01195556e-02 -5.52338064e-01 -1.20672107e+00
-2.23167419e-01 -6.83605492e-01 7.18086660e-01 -6.11287504e-02
-9.31596339e-01 8.77743959e-01 -6.31763041e-02 -1.24001992e+00
-9.22288448e-02 -2.16642484e-01 -5.07049143e-01 8.71465981e-01
-1.62577474e+00 -1.08005822e+00 -7.50912607e-01 5.72425425e-01
5.60300469e-01 3.30039829e-01 5.68124950e-01 3.60229284e-01
-3.04225355e-01 5.74741483e-01 -1.79883555e-01 3.45206529e-01
3.64176929e-01 -1.20473778e+00 -1.28935412e-01 6.37047589e-01
-3.03293914e-01 1.04670152e-01 3.57711166e-01 -4.72008646e-01
-1.24743831e+00 -1.14721143e+00 5.74474216e-01 -4.85771075e-02
2.81778306e-01 9.33063626e-02 -7.18558729e-01 3.31382841e-01
-2.77218688e-02 7.96650112e-01 5.54153800e-01 -5.31062961e-01
1.63676724e-01 1.67016238e-01 -1.55622447e+00 6.17026210e-01
6.93939269e-01 9.97017920e-02 -3.45966145e-02 4.91436422e-01
3.68284732e-01 -1.19437432e+00 -1.12835085e+00 6.94947779e-01
2.98729390e-01 -9.07573700e-01 1.07606745e+00 -2.80050933e-01
7.86658525e-02 -3.46389532e-01 2.17986867e-01 -1.05941176e+00
-4.90578711e-01 -3.88648033e-01 1.51815116e-01 2.52208680e-01
4.09517705e-01 -3.10807645e-01 1.04000247e+00 5.72480917e-01
-5.32438517e-01 -1.27195537e+00 -9.96962726e-01 -2.82396734e-01
2.31920272e-01 -1.08970456e-01 6.02106392e-01 9.38514054e-01
-1.76011994e-01 -1.40233070e-01 1.61714196e-01 9.75486338e-02
8.62571597e-01 4.64638889e-01 1.69989720e-01 -8.92556548e-01
-2.33263582e-01 -8.92173707e-01 -1.64403781e-01 -1.03243589e+00
-5.42736948e-01 -9.89193022e-01 -9.78806391e-02 -1.55048251e+00
5.69398761e-01 -7.39994407e-01 -2.14100197e-01 4.19006288e-01
-3.74672830e-01 5.30714035e-01 1.96713120e-01 2.76619166e-01
-5.17887235e-01 1.24071613e-01 1.90589285e+00 -5.52292168e-02
-4.23946343e-02 1.24966376e-01 -2.54951924e-01 9.07741964e-01
5.22720695e-01 -3.01215768e-01 -1.05621547e-01 -3.64558905e-01
-2.24629328e-01 7.36711621e-01 4.16451335e-01 -8.73461664e-01
4.22072232e-01 -2.86489576e-01 6.77503467e-01 -7.39537299e-01
2.24115476e-02 -7.97965944e-01 -7.62080923e-02 1.02473760e+00
-6.00943528e-02 -6.18295511e-03 1.16443008e-01 1.52277768e-01
-1.99074522e-01 -1.25520349e-01 1.27584803e+00 -6.49972141e-01
-4.36622053e-01 8.86623442e-01 -2.26913333e-01 1.62528053e-01
1.23466480e+00 -1.40131876e-01 9.90900397e-02 3.62565438e-03
-1.09664595e+00 5.28684914e-01 2.28944957e-01 -2.46728912e-01
5.04205942e-01 -1.06504476e+00 -1.06617737e+00 3.24349254e-01
-1.30382374e-01 6.39841914e-01 6.01038814e-01 1.76378345e+00
-9.13466632e-01 5.69804788e-01 1.57781020e-01 -8.66054654e-01
-1.15514791e+00 3.98747355e-01 8.29295874e-01 -1.11022842e+00
-8.64215136e-01 1.07484019e+00 5.15501559e-01 -7.74172246e-02
1.08231269e-01 -5.94048738e-01 1.34144276e-01 -2.87264973e-01
2.99538255e-01 1.65181965e-01 4.22993332e-01 -5.40377736e-01
-5.62216520e-01 5.55253744e-01 -1.78713426e-01 2.10780039e-01
9.27626848e-01 -8.89445469e-02 -5.00721224e-02 -3.52259785e-01
8.50433111e-01 -1.29722536e-01 -1.33419693e+00 -4.53653067e-01
6.00657724e-02 -3.87345314e-01 4.36728001e-01 -1.05741131e+00
-1.38133645e+00 8.05812597e-01 6.45411432e-01 -6.24723509e-02
1.28606510e+00 -6.26354516e-02 9.94846582e-01 -4.46470350e-01
3.04302543e-01 -5.42595744e-01 -1.03208132e-01 3.90999526e-01
5.47840059e-01 -1.22473979e+00 4.28815544e-01 -7.77230740e-01
-6.33566737e-01 1.06777072e+00 5.80197453e-01 3.47799286e-02
6.37276351e-01 5.72072625e-01 -4.14363369e-02 -2.10699856e-01
-4.19087797e-01 -1.45766169e-01 4.13353890e-01 2.25277513e-01
5.63941360e-01 3.84867549e-01 -4.32764918e-01 5.35288692e-01
1.11657731e-01 -1.05836391e-01 2.19186887e-01 8.79986584e-01
-4.82263058e-01 -7.29952872e-01 -2.93969512e-01 5.93778491e-01
-7.40410924e-01 -3.68342698e-01 2.73771316e-01 9.75478768e-01
1.83697790e-01 2.20500544e-01 -1.24818742e-01 2.92730033e-01
2.11305350e-01 -9.46458578e-02 6.27141535e-01 -5.47417998e-01
-1.02700460e+00 6.56341255e-01 -3.23158205e-01 -3.54219764e-01
4.03135717e-02 -5.82466722e-01 -1.53567791e+00 -5.10014892e-02
-3.05344999e-01 2.77170002e-01 6.92978024e-01 7.87245810e-01
9.19780731e-02 6.70765877e-01 6.12029016e-01 -7.37578690e-01
-7.03941762e-01 -8.77418220e-01 -3.61543119e-01 2.25619659e-01
1.75292000e-01 -3.75618100e-01 -1.01881512e-01 -1.85745865e-01]
|
[14.566838264465332, -2.6458566188812256]
|
b850fb42-e2b1-4d92-a728-1ff1885a51af
|
alo-vc-any-to-any-low-latency-one-shot-voice
|
2306.011
| null |
https://arxiv.org/abs/2306.01100v1
|
https://arxiv.org/pdf/2306.01100v1.pdf
|
ALO-VC: Any-to-any Low-latency One-shot Voice Conversion
|
This paper presents ALO-VC, a non-parallel low-latency one-shot phonetic posteriorgrams (PPGs) based voice conversion method. ALO-VC enables any-to-any voice conversion using only one utterance from the target speaker, with only 47.5 ms future look-ahead. The proposed hybrid signal processing and machine learning pipeline combines a pre-trained speaker encoder, a pitch predictor to predict the converted speech's prosody, and positional encoding to convey the phoneme's location information. We introduce two system versions: ALO-VC-R, which uses a pre-trained d-vector speaker encoder, and ALO-VC-E, which improves performance using the ECAPA-TDNN speaker encoder. The experimental results demonstrate both ALO-VC-R and ALO-VC-E can achieve comparable performance to non-causal baseline systems on the VCTK dataset and two out-of-domain datasets. Furthermore, both proposed systems can be deployed on a single CPU core with 55 ms latency and 0.78 real-time factor. Our demo is available online.
|
['Milos Cernak', 'Damien Ronssin', 'Bohan Wang']
|
2023-06-01
| null | null | null | null |
['voice-conversion', 'voice-conversion']
|
['audio', 'speech']
|
[ 8.14982578e-02 1.62890956e-01 -1.70839056e-01 -2.85055816e-01
-1.38363945e+00 -5.16577601e-01 4.33346331e-01 -1.62117288e-01
-3.31626892e-01 4.59975541e-01 5.32664061e-01 -4.84850109e-01
4.40810442e-01 -4.77480501e-01 -6.91886127e-01 -3.46656442e-01
9.22340453e-02 4.56007361e-01 5.25116622e-01 2.93355696e-02
-6.32795990e-02 3.20174724e-01 -1.29181707e+00 4.55018491e-01
1.55662909e-01 1.17857075e+00 5.30651093e-01 1.39833152e+00
-6.16583508e-03 6.79945827e-01 -6.07162356e-01 -1.15246624e-01
-8.05778801e-02 -3.91982287e-01 -5.96483350e-01 -4.93196666e-01
1.06867887e-02 -2.98286110e-01 -2.48060867e-01 6.11318469e-01
1.16196239e+00 6.25358149e-02 3.64900619e-01 -9.74862099e-01
-4.73041534e-01 9.78112400e-01 -9.47044119e-02 3.78477037e-01
5.46525359e-01 1.13070987e-01 9.74046886e-01 -1.13360918e+00
3.77319455e-01 1.43877912e+00 6.76873207e-01 7.58992970e-01
-9.44008410e-01 -8.32249820e-01 -1.66331649e-01 3.04493994e-01
-1.36031485e+00 -1.18271780e+00 8.09549510e-01 -2.65277863e-01
1.62615061e+00 3.28644067e-01 4.71813411e-01 1.12442005e+00
3.00964445e-01 5.24957299e-01 8.54030132e-01 -6.54524684e-01
3.26820105e-01 6.65405318e-02 -1.95571050e-01 3.64304125e-01
-6.92393661e-01 5.21633208e-01 -1.17810237e+00 -4.72810492e-02
4.51329410e-01 -6.70294404e-01 -1.46341205e-01 5.37742198e-01
-1.03907621e+00 5.35760701e-01 -1.22058548e-01 8.93179551e-02
-4.30020034e-01 2.40699470e-01 6.80357933e-01 4.32015538e-01
3.49810481e-01 2.34438423e-02 -6.41605794e-01 -8.19335163e-01
-9.67395186e-01 -1.55327410e-01 8.58912289e-01 1.19562125e+00
7.44471103e-02 6.55141056e-01 -1.18312337e-01 1.06627011e+00
3.43623579e-01 4.69536871e-01 1.03068566e+00 -7.16801286e-01
4.97377276e-01 -5.24369001e-01 -3.13018024e-01 -2.10940450e-01
-1.50767222e-01 -2.79390126e-01 -3.06749642e-01 -1.34726986e-01
-3.00919205e-01 -4.57712531e-01 -7.25476801e-01 1.73085558e+00
2.44089514e-01 4.02606428e-01 4.09258276e-01 7.90767312e-01
7.84312129e-01 1.24070621e+00 -3.13291311e-01 -5.53960741e-01
1.37602544e+00 -1.37075162e+00 -1.01903594e+00 -1.83195949e-01
9.88795906e-02 -1.15407217e+00 1.36638641e+00 3.18670779e-01
-1.25312161e+00 -8.04538131e-01 -1.04245579e+00 -1.95843145e-01
-1.56087475e-02 2.76236594e-01 2.58866191e-01 6.23069763e-01
-1.12332559e+00 2.95694470e-01 -1.00977635e+00 -9.41317230e-02
-7.05372095e-02 2.97362536e-01 6.82827272e-03 5.82019866e-01
-1.29465032e+00 3.95349711e-01 2.65726149e-01 -1.30046040e-01
-1.14856493e+00 -9.07143176e-01 -4.40666556e-01 1.79014981e-01
1.73629031e-01 -2.13401198e-01 1.88337719e+00 -5.89272857e-01
-2.65806341e+00 2.07225397e-01 -4.74118680e-01 -5.45163333e-01
2.74998188e-01 -4.60418135e-01 -1.12709451e+00 9.01832357e-02
-1.11005967e-02 6.49617493e-01 1.07918394e+00 -6.49642527e-01
-8.54002595e-01 -4.08876501e-02 -8.33562732e-01 3.19492429e-01
-1.14099696e-01 3.83365721e-01 -6.23240232e-01 -5.66061437e-01
-1.87533110e-01 -8.42525125e-01 3.52323920e-01 -3.72193485e-01
-5.69063067e-01 -2.70528823e-01 9.20272946e-01 -1.00502563e+00
1.40283394e+00 -2.40095305e+00 -1.45522431e-01 -3.41169864e-01
-4.48010027e-01 3.75493556e-01 -5.04257791e-02 5.43031514e-01
5.02674021e-02 -1.93206519e-01 4.83597107e-02 -4.98252988e-01
-4.68503870e-02 -1.06793903e-01 -7.68195391e-01 1.31058827e-01
2.02024966e-01 6.45775795e-01 -5.31483293e-01 -3.14075261e-01
2.30405003e-01 7.01313376e-01 -6.75032318e-01 4.89460349e-01
-1.48343205e-01 3.08558494e-01 2.65524566e-01 3.96484584e-01
2.58176208e-01 5.90241551e-01 2.58111447e-01 1.00416787e-01
-4.14349258e-01 1.23836148e+00 -8.92540991e-01 1.70283878e+00
-9.15816724e-01 9.57282543e-01 1.72101215e-01 -2.72710294e-01
9.20137763e-01 1.03300500e+00 -3.76159102e-02 -5.51296413e-01
-1.16130672e-02 3.94652516e-01 1.64383575e-01 -1.67243257e-01
5.18046916e-01 -2.51006484e-01 7.08327303e-03 2.97082812e-01
3.37681353e-01 -2.31163144e-01 -4.97796118e-01 -1.68451503e-01
1.01045382e+00 9.90460217e-02 4.52465117e-01 4.98827472e-02
5.11134028e-01 -2.90207654e-01 8.21564257e-01 3.08953971e-01
-3.48118186e-01 5.56021988e-01 2.76170582e-01 3.56035113e-01
-8.48736286e-01 -1.21372318e+00 -1.11762799e-01 1.34922338e+00
-2.87672400e-01 -5.89049101e-01 -8.17961037e-01 -2.18555890e-02
-3.24248642e-01 1.12335610e+00 1.38278425e-01 1.33018503e-02
-7.92588770e-01 -2.62748990e-02 1.06441677e+00 6.59271598e-01
8.12142342e-02 -1.24600267e+00 -3.33726257e-01 7.29139447e-01
-7.84936100e-02 -1.25845945e+00 -1.02355170e+00 3.65316093e-01
-6.31876767e-01 -1.64791495e-01 -2.58994848e-01 -9.81124341e-01
-2.57332593e-01 -6.38722777e-02 6.60407364e-01 -7.95437574e-01
-6.88569769e-02 7.45643303e-02 -2.05847859e-01 -5.60833216e-01
-7.33015776e-01 1.44785762e-01 5.31480789e-01 -8.52437913e-02
5.01765788e-01 -8.06758165e-01 -2.52519935e-01 3.91626321e-02
-8.40053037e-02 5.82972988e-02 4.25362110e-01 6.97420776e-01
9.00837958e-01 -1.59857258e-01 1.09349132e+00 -5.21676540e-01
7.42080688e-01 -2.96450198e-01 -5.05266011e-01 -1.03958353e-01
-7.85829246e-01 -1.66859388e-01 8.83394957e-01 -6.65615976e-01
-1.15959024e+00 3.50775272e-01 -8.31539333e-01 -6.96260154e-01
1.87419988e-02 2.08847553e-01 -4.94270712e-01 7.42344975e-01
3.45040381e-01 4.28594708e-01 -1.99831784e-01 -9.06553924e-01
7.88284659e-01 1.63765800e+00 1.19234133e+00 -2.16457367e-01
3.78403991e-01 -2.34141171e-01 -7.35457838e-01 -1.25838995e+00
-1.98681653e-01 -4.53211576e-01 -5.71575344e-01 -8.24738741e-02
7.99982786e-01 -1.33608902e+00 -7.42374241e-01 5.08729875e-01
-1.41951323e+00 -3.79065812e-01 -3.92321646e-01 7.88224220e-01
-7.43436217e-01 -1.54512838e-01 -9.00175393e-01 -8.65088224e-01
-7.60357082e-01 -1.02895319e+00 1.03446746e+00 2.06710421e-03
-1.90413967e-01 -5.71599841e-01 2.13345900e-01 2.86498636e-01
3.91661704e-01 -4.88489270e-01 6.85763717e-01 -6.88293338e-01
-2.92929322e-01 9.40766558e-02 3.40844750e-01 5.52589297e-01
2.35430762e-01 -3.61945257e-02 -1.55955398e+00 -2.84755260e-01
-3.26944957e-03 -1.03068806e-01 5.02687395e-01 3.27577561e-01
9.61003840e-01 -6.25166237e-01 -1.53052017e-01 5.31177938e-01
9.63876307e-01 7.55639851e-01 2.25865006e-01 -4.89453644e-01
5.33869505e-01 1.66765854e-01 5.48537850e-01 3.21764112e-01
3.46162558e-01 1.02929997e+00 -5.03616445e-02 2.69334972e-01
-8.56735647e-01 -6.86967969e-01 1.06112790e+00 2.10523891e+00
2.36018404e-01 -3.39146286e-01 -6.52505219e-01 5.73743999e-01
-1.34254110e+00 -8.28354836e-01 1.03849694e-01 2.10262465e+00
1.34666336e+00 2.38055348e-01 1.27005264e-01 3.68985057e-01
8.22909653e-01 1.94384992e-01 -5.90735078e-01 -1.05585277e+00
3.54667187e-01 6.77530646e-01 2.26593420e-01 1.06302190e+00
-7.84175277e-01 1.35967863e+00 5.91621685e+00 9.45217669e-01
-1.77578199e+00 7.76962757e-01 -8.04101229e-02 -5.20571768e-01
2.16465406e-02 -2.54770309e-01 -1.04120803e+00 4.55305636e-01
2.03247142e+00 -3.19835752e-01 7.61586487e-01 1.05883205e+00
2.89487034e-01 4.98625785e-01 -1.27563155e+00 9.86424506e-01
4.37845774e-02 -1.21689415e+00 -2.86631286e-01 -6.26336560e-02
7.92136639e-02 2.66717196e-01 2.53213085e-02 4.97241974e-01
5.00627346e-02 -7.18340039e-01 1.24174976e+00 2.78843462e-01
1.61715055e+00 -1.01658833e+00 2.44703621e-01 3.26236188e-01
-1.64478636e+00 1.03735320e-01 -2.74118543e-01 -3.51331681e-02
3.58027875e-01 2.99645901e-01 -1.46500516e+00 3.57065707e-01
5.44624686e-01 3.36418897e-01 1.04667127e-01 3.95799190e-01
-3.85819227e-01 1.43033779e+00 -3.31007302e-01 -5.06090634e-02
-1.44798160e-01 5.09563625e-01 7.76942790e-01 1.48330641e+00
5.53564131e-01 1.87446147e-01 -1.36662573e-01 3.57285798e-01
-1.44641995e-01 1.55874193e-01 -1.35109007e-01 -2.09221870e-01
1.22196674e+00 8.68624747e-01 -1.65428594e-01 -3.04654777e-01
-2.75112718e-01 1.21938336e+00 8.10342431e-02 1.73433460e-02
-9.06541049e-01 -8.53028476e-01 9.72934663e-01 -2.11626980e-02
4.84747112e-01 -1.88212812e-01 1.02204848e-02 -7.21878052e-01
-2.69063115e-01 -8.68029296e-01 -2.09325597e-01 -8.13602924e-01
-8.17021012e-01 1.11636817e+00 -5.04763842e-01 -1.17837620e+00
-9.57536101e-01 -2.37506092e-01 -6.07532978e-01 1.04688096e+00
-1.38102615e+00 -9.68521714e-01 2.67110169e-01 5.53771079e-01
1.03429317e+00 -5.59116244e-01 1.22037053e+00 3.74232918e-01
-5.87859094e-01 7.21946180e-01 6.67805746e-02 -4.89610173e-02
8.75086129e-01 -1.12328482e+00 9.63660657e-01 7.71707654e-01
2.73591071e-01 2.73080349e-01 3.91088843e-01 -5.53639710e-01
-1.45156777e+00 -1.36201560e+00 1.28762734e+00 -1.80304542e-01
3.49025160e-01 -7.02303290e-01 -7.72484243e-01 7.73558438e-01
4.12673622e-01 5.73793799e-02 8.67045045e-01 -2.32494120e-02
-2.41418645e-01 -4.35534298e-01 -7.82384753e-01 3.54948282e-01
8.42564762e-01 -1.11434317e+00 -9.46368635e-01 -4.16609086e-02
1.59272146e+00 -4.87585723e-01 -8.18551004e-01 -1.42785132e-01
5.03914773e-01 -6.58938885e-01 7.81260967e-01 -2.91243106e-01
1.45272985e-01 -1.74736798e-01 -6.62974060e-01 -1.36814320e+00
-1.25510722e-01 -1.16853118e+00 -2.51620561e-01 1.64216459e+00
5.77306271e-01 -5.10837913e-01 8.31346810e-02 -2.41142988e-01
-7.03224778e-01 -4.77226585e-01 -1.51960981e+00 -8.72677028e-01
-5.95392771e-02 -1.05188823e+00 7.72290945e-01 5.55114865e-01
1.25181168e-01 8.21056545e-01 -5.96294582e-01 4.97946352e-01
2.30533227e-01 -1.84137896e-01 5.13018668e-01 -8.67137551e-01
-7.02498317e-01 -4.97543328e-02 -1.56081012e-02 -1.10553825e+00
9.23478156e-02 -1.02507758e+00 4.50175732e-01 -1.02366590e+00
-4.87947434e-01 -1.61002979e-01 -3.49561393e-01 5.47346354e-01
2.39246964e-01 -3.56479622e-02 2.55624473e-01 1.58982635e-01
1.35763988e-01 6.89161599e-01 7.67399549e-01 1.11272611e-01
-7.62631595e-01 7.97738507e-02 -2.52424836e-01 5.98748744e-01
7.54493356e-01 -6.59085214e-01 -5.27758956e-01 -2.34894618e-01
-4.87460077e-01 7.87747681e-01 -1.73363596e-01 -1.20229495e+00
4.52022761e-01 8.74126405e-02 8.21901038e-02 -8.40794325e-01
9.47654963e-01 -3.75564247e-01 1.83091789e-01 5.09111524e-01
-4.44291979e-01 -1.92695074e-02 3.07353854e-01 5.19753754e-01
-3.18091154e-01 7.61179328e-02 8.61182809e-01 3.34168613e-01
-5.50960422e-01 -6.11005053e-02 -7.54847229e-01 3.63403484e-02
7.55151689e-01 3.11458051e-01 -1.46493673e-01 -3.43525112e-01
-7.25263655e-01 -4.18587536e-01 -3.37744415e-01 7.21823394e-01
6.56438053e-01 -1.41558385e+00 -5.31158924e-01 5.57654142e-01
-1.33600801e-01 -3.35839093e-01 5.01221120e-02 3.58387589e-01
-3.78082007e-01 9.29200411e-01 -6.38664216e-02 -6.14972770e-01
-1.42531931e+00 3.32887888e-01 2.89214224e-01 3.11199427e-01
-8.52361560e-01 1.06734681e+00 -2.66739815e-01 -5.79648733e-01
4.72608805e-01 -4.41436172e-01 1.42187208e-01 -1.58825498e-02
6.42459154e-01 4.83158857e-01 1.36222273e-01 -7.01148868e-01
-7.62003422e-01 2.01390013e-01 2.51532737e-02 -1.01543617e+00
1.09992146e+00 -3.14492971e-01 2.48877943e-01 9.78085101e-01
1.15303695e+00 4.79057521e-01 -1.19346809e+00 -2.96853483e-01
-1.28796622e-01 -2.00400352e-02 5.27929008e-01 -1.09967029e+00
-7.55955458e-01 1.41576636e+00 8.07036221e-01 -3.21923047e-01
1.15355921e+00 -1.14665538e-01 1.22556269e+00 1.10899627e-01
2.87516266e-01 -1.19077420e+00 -4.19892855e-02 6.12815738e-01
1.08515310e+00 -6.98462903e-01 -4.96853143e-01 -3.11341256e-01
-9.04139459e-01 1.12942374e+00 4.32911932e-01 2.96863437e-01
8.16009641e-01 7.12460816e-01 3.50532770e-01 5.32184660e-01
-1.51638138e+00 1.93217918e-02 7.13250041e-02 5.77882707e-01
5.38377881e-01 5.58946192e-01 -4.86999340e-02 1.11555219e+00
-9.56825078e-01 -5.46362484e-03 4.40703809e-01 4.90176976e-01
-3.99310738e-01 -1.03143072e+00 -2.73894697e-01 -9.16349795e-03
-5.32963872e-01 -4.14837360e-01 -2.83412606e-01 3.25499296e-01
-1.64655045e-01 1.22686577e+00 5.02152622e-01 -8.61744404e-01
2.70201236e-01 6.30717874e-01 1.25412375e-01 -7.58747995e-01
-7.56088793e-01 7.74117708e-01 2.61764824e-01 -4.88492608e-01
3.07491094e-01 -7.66334772e-01 -1.60034704e+00 -2.52703816e-01
-3.39885950e-01 1.15509354e-01 1.05711830e+00 5.58166027e-01
6.56702518e-01 1.05655158e+00 9.69935477e-01 -6.22056365e-01
-7.68896818e-01 -1.28717506e+00 -4.63287950e-01 -6.08797491e-01
5.23857713e-01 -9.39229801e-02 -2.17754513e-01 1.45949900e-01]
|
[14.846231460571289, 6.631141662597656]
|
e905649f-d12a-4f37-9b9d-30f46c9d3cfd
|
cross-lingual-complex-word-identification
| null | null |
https://aclanthology.org/W18-0518
|
https://aclanthology.org/W18-0518.pdf
|
Cross-lingual complex word identification with multitask learning
|
We approach the 2018 Shared Task on Complex Word Identification by leveraging a cross-lingual multitask learning approach. Our method is highly language agnostic, as evidenced by the ability of our system to generalize across languages, including languages for which we have no training data. In the shared task, this is the case for French, for which our system achieves the best performance. We further provide a qualitative and quantitative analysis of which words pose problems for our system.
|
['Johannes Bjerva', 'Joachim Bingel']
|
2018-06-01
| null | null | null |
ws-2018-6
|
['complex-word-identification']
|
['natural-language-processing']
|
[-2.54655480e-01 -4.65201914e-01 -1.77065790e-01 -1.22003414e-01
-1.23940635e+00 -1.07806385e+00 8.00581932e-01 9.54631343e-02
-9.23348665e-01 5.64558744e-01 4.58796114e-01 -6.37548327e-01
1.89084321e-01 -3.04313064e-01 -3.82856399e-01 -1.60200149e-01
1.10144503e-01 2.97619224e-01 -4.29211646e-01 -5.24417102e-01
-7.15676248e-02 -5.77568188e-02 -9.03589606e-01 4.15468872e-01
7.93468714e-01 5.30044794e-01 7.79373720e-02 4.37661350e-01
-1.63767025e-01 2.56964266e-01 -4.78122115e-01 -6.35259271e-01
2.68999428e-01 1.75448716e-01 -1.01829469e+00 -3.12192798e-01
6.76319301e-01 -1.17719099e-01 -5.11560440e-02 9.39152896e-01
5.77264965e-01 3.01777069e-02 7.65729070e-01 -9.44029212e-01
-7.78849661e-01 8.11976314e-01 -3.04469854e-01 3.62343431e-01
4.75114673e-01 2.85528988e-01 1.49586439e+00 -1.25389385e+00
4.49741513e-01 1.52527475e+00 7.92340875e-01 5.23050189e-01
-1.13081002e+00 -9.46684599e-01 4.90149975e-01 -1.81632675e-02
-1.53500068e+00 -7.64705479e-01 3.91875327e-01 -5.76061189e-01
1.53668022e+00 -2.43900701e-01 2.07189888e-01 1.46715593e+00
1.15978122e-01 7.51809001e-01 1.23908854e+00 -5.61721802e-01
-1.10739961e-01 1.83030237e-02 1.29263923e-01 6.12749696e-01
2.41325483e-01 -4.78223804e-03 -6.52358532e-01 -2.36532334e-02
1.25935316e-01 -3.97414744e-01 -1.13480054e-01 1.56316772e-01
-1.33454883e+00 9.09032941e-01 4.65275832e-02 5.22124767e-01
1.08468220e-01 2.85079125e-02 7.85581410e-01 5.20489573e-01
9.62261915e-01 6.75926864e-01 -1.08062398e+00 -1.09178625e-01
-7.51852453e-01 1.72547158e-02 9.13309813e-01 8.25811923e-01
7.43564546e-01 1.41175181e-01 -8.55287760e-02 1.04777908e+00
2.90566385e-01 5.30058146e-01 8.20245862e-01 -2.38397226e-01
5.29674113e-01 3.44441056e-01 -1.53742060e-01 -3.21274608e-01
-5.52114069e-01 -4.55965400e-01 -2.89189398e-01 -9.76046771e-02
5.31788290e-01 -5.24526238e-01 -6.99070573e-01 1.97425127e+00
-3.00494671e-01 -2.31672212e-01 2.03552812e-01 4.69129771e-01
9.68300700e-01 3.89674097e-01 5.28004050e-01 4.63551544e-02
1.58276856e+00 -1.00239825e+00 -7.39925802e-01 -7.36755371e-01
1.21119654e+00 -1.03344417e+00 1.49622536e+00 3.38517219e-01
-5.70138872e-01 -4.22521025e-01 -1.01120901e+00 -1.38523385e-01
-8.05088103e-01 2.98003346e-01 8.68739665e-01 8.08819771e-01
-8.83200765e-01 -2.10588928e-02 -5.55181801e-01 -6.69680595e-01
1.62969694e-01 -2.85596270e-02 -5.03274977e-01 -3.01561002e-02
-1.37968993e+00 1.41406643e+00 6.80484772e-01 -2.58369923e-01
-7.96517491e-01 -7.59634852e-01 -1.17404008e+00 -1.87171653e-01
3.82882267e-01 -4.00715917e-01 1.32369697e+00 -7.44677126e-01
-1.20246875e+00 1.18542707e+00 -3.60108197e-01 -2.65075117e-01
2.76771426e-01 -4.41131741e-01 -7.44023263e-01 -5.17704725e-01
2.80181944e-01 4.29336339e-01 5.10600805e-01 -1.02328706e+00
-5.86847901e-01 -5.66649996e-02 8.10341462e-02 3.42222601e-02
-8.35252941e-01 4.83879745e-01 -3.63219738e-01 -8.74808967e-01
-4.63191509e-01 -9.51774776e-01 4.13935855e-02 -4.59101707e-01
-4.10593241e-01 -8.20086062e-01 5.29305816e-01 -7.12173462e-01
1.40820479e+00 -2.09143138e+00 -1.15748875e-01 -3.14812362e-02
8.79898369e-02 1.91400096e-01 -4.01390314e-01 5.32431483e-01
-5.68902455e-02 4.88268912e-01 -7.29815755e-03 -6.77337110e-01
1.93613768e-01 -9.85468626e-02 -3.06001157e-01 3.80403668e-01
3.27479422e-01 1.18064487e+00 -7.40538597e-01 -3.07106614e-01
-1.78384826e-01 1.01581007e-01 -3.97212505e-01 -5.61235547e-02
-1.57150850e-01 1.84485510e-01 -3.06124091e-01 5.99325240e-01
3.27617079e-01 -1.77602842e-01 5.97337745e-02 -8.89261812e-02
-3.44969690e-01 6.00732744e-01 -7.48511672e-01 1.92136598e+00
-1.00580657e+00 3.99397016e-01 2.72672959e-02 -6.05003655e-01
6.42586946e-01 4.56815034e-01 2.00881794e-01 -5.99394381e-01
-8.93991962e-02 4.80396688e-01 2.75214791e-01 -3.35715801e-01
3.88122648e-01 -2.10018769e-01 -6.29409611e-01 8.40065598e-01
4.33882564e-01 -2.61568189e-01 1.27553612e-01 2.97254652e-01
9.75394547e-01 -4.40304130e-02 5.64601362e-01 -8.51582289e-01
4.90307868e-01 3.44065838e-02 2.79696316e-01 6.60923183e-01
-3.03073883e-01 1.56730980e-01 -4.52757850e-02 -3.19273889e-01
-8.50040019e-01 -9.84148383e-01 -2.85784185e-01 1.64058316e+00
-3.66616070e-01 -6.91008389e-01 -3.79010379e-01 -7.85535753e-01
1.89023316e-01 5.26922405e-01 -5.45450628e-01 -1.02011621e-01
-4.15286332e-01 -6.82878673e-01 8.65785420e-01 3.15188199e-01
1.07385106e-01 -1.08535194e+00 1.72493994e-01 1.38985634e-01
-7.83764049e-02 -1.63386011e+00 -9.08477783e-01 2.31705487e-01
-1.70125246e-01 -9.51439738e-01 -4.79378551e-01 -1.11033452e+00
1.65756971e-01 -2.32146005e-03 1.36186993e+00 3.69322896e-02
-1.10698260e-01 5.31597197e-01 -2.16699272e-01 -6.00818813e-01
-4.63947415e-01 6.74932778e-01 3.85771811e-01 -2.51777649e-01
6.03330016e-01 -6.88199177e-02 -7.39768222e-02 -7.31795207e-02
-4.59737092e-01 -2.37630442e-01 3.90518159e-01 7.24264205e-01
5.71286678e-02 -4.00119931e-01 9.20747161e-01 -9.54030275e-01
1.12454748e+00 -6.55630052e-01 -3.72194320e-01 5.06901741e-01
-6.64217412e-01 3.86895128e-02 6.22746468e-01 -3.86139333e-01
-7.36456037e-01 -5.02211675e-02 -1.64609984e-01 1.06558442e-01
-6.72781989e-02 7.23851085e-01 -4.83869649e-02 -1.60737485e-01
6.32128835e-01 8.10083281e-03 -3.43462914e-01 -7.50258327e-01
4.23270404e-01 7.23383307e-01 2.99967676e-01 -6.95740163e-01
8.38395655e-01 6.33926615e-02 -5.98147750e-01 -7.10774958e-01
-1.15438044e+00 -5.16021907e-01 -7.43318856e-01 9.50728580e-02
1.02012503e+00 -1.46384895e+00 -7.17540026e-01 6.05663717e-01
-1.26268733e+00 -3.99842978e-01 2.09235251e-01 5.92327416e-01
-2.27889076e-01 6.63441792e-02 -6.31696582e-01 -5.80789268e-01
-4.70000654e-01 -1.08710349e+00 1.08359671e+00 -3.17774683e-01
-5.53107142e-01 -1.64483452e+00 6.71407357e-02 8.54841173e-02
7.26482093e-01 -1.59978211e-01 1.12424052e+00 -1.20383847e+00
7.31363446e-02 1.51637629e-01 -1.40004784e-01 2.00238347e-01
4.00994331e-01 -1.79543063e-01 -9.92926657e-01 -6.86850905e-01
-2.93705732e-01 -8.10158610e-01 1.09973383e+00 3.54618840e-02
8.78266454e-01 -1.43145666e-01 -3.32964718e-01 5.82887232e-01
1.29811168e+00 -4.45784956e-01 -2.35834286e-01 4.35330003e-01
8.90435398e-01 6.21404231e-01 2.36740321e-01 9.84307751e-02
8.87927115e-01 7.69077122e-01 -1.40515000e-01 -2.29225442e-01
-2.73172021e-01 -2.25477487e-01 5.87140203e-01 1.12218821e+00
3.83219451e-01 -3.45288068e-01 -1.49354947e+00 7.97659636e-01
-1.52045584e+00 -4.15736020e-01 2.12520007e-02 1.92121005e+00
9.82276976e-01 1.95166662e-01 1.32909790e-01 -2.32594416e-01
5.05933702e-01 2.70302296e-01 -3.38742971e-01 -6.32090211e-01
-4.19116318e-01 4.96541530e-01 5.35519540e-01 7.41906047e-01
-1.47093904e+00 1.62912107e+00 7.68977499e+00 9.73605871e-01
-1.13734221e+00 5.10990560e-01 2.85691559e-01 -1.09779358e-01
-4.51416492e-01 -1.60229892e-01 -8.39179575e-01 2.92462498e-01
1.01083601e+00 -6.33228481e-01 5.53355575e-01 4.14260954e-01
-1.14065900e-01 1.37288526e-01 -1.17795253e+00 8.50136757e-01
2.84493983e-01 -8.46419632e-01 1.25321791e-01 3.23170354e-03
7.16264427e-01 5.20198047e-01 1.33441463e-01 5.12254000e-01
6.99696302e-01 -1.38127446e+00 6.71490252e-01 3.89978625e-02
1.05919635e+00 -6.73835278e-01 3.58251154e-01 3.24348122e-01
-1.10815835e+00 5.29909879e-02 7.54994303e-02 -8.54057074e-02
1.36845738e-01 4.11577106e-01 -6.79243565e-01 1.94089055e-01
5.33310235e-01 7.72823274e-01 -8.34961534e-01 5.67101002e-01
-3.95996571e-01 7.89789915e-01 -2.18905181e-01 4.95466180e-02
2.33527616e-01 7.45415688e-02 4.46939796e-01 1.75198817e+00
1.46622390e-01 -3.95540833e-01 6.75657451e-01 4.28372890e-01
-3.33071560e-01 5.46636283e-01 -8.69997740e-01 -2.70958602e-01
4.85597491e-01 1.38644242e+00 -2.47682840e-01 -4.00810480e-01
-8.52490842e-01 8.42234194e-01 7.59777248e-01 3.86259675e-01
-1.54188052e-01 -1.70878842e-01 7.32401431e-01 -4.86249238e-01
1.31567657e-01 -8.42067778e-01 -4.74426568e-01 -1.38520765e+00
-1.75644815e-01 -1.04466867e+00 5.74220240e-01 -1.74224913e-01
-1.77637613e+00 5.72876871e-01 -1.44981146e-01 -8.82762313e-01
-4.95258242e-01 -9.16231275e-01 -5.79126358e-01 1.19802082e+00
-1.62695408e+00 -1.62291992e+00 2.50986308e-01 7.83443034e-01
5.92265189e-01 -8.25108945e-01 1.25267518e+00 4.38880593e-01
-5.28107285e-01 1.10032189e+00 -1.66320369e-01 1.79576099e-01
1.21757746e+00 -1.27424300e+00 9.73524690e-01 8.27911854e-01
5.15960932e-01 7.25478470e-01 4.40365076e-01 -5.25698721e-01
-1.51218557e+00 -1.04920435e+00 1.33377206e+00 -7.25645661e-01
1.35590982e+00 -9.27838445e-01 -7.50986814e-01 8.87405694e-01
2.50771493e-01 7.37319216e-02 1.03707862e+00 9.50974166e-01
-8.17114949e-01 3.55569154e-01 -8.03550243e-01 4.39711392e-01
1.12570965e+00 -1.24111164e+00 -5.02092183e-01 7.22106755e-01
9.01252568e-01 -6.33154437e-02 -8.51564109e-01 2.81716645e-01
6.59670115e-01 -2.36475363e-01 7.31266916e-01 -8.97094905e-01
2.85668224e-01 1.57889009e-01 -2.25911468e-01 -1.82136369e+00
-4.28462744e-01 -3.16047430e-01 1.56204909e-01 1.32836998e+00
1.04327011e+00 -1.00665319e+00 8.30523819e-02 1.99723914e-01
-1.13738894e-01 -4.86594468e-01 -9.62702572e-01 -1.08115423e+00
8.33899975e-01 -7.14830756e-01 3.89064312e-01 1.25426400e+00
3.01875155e-02 6.15292072e-01 -1.62376285e-01 -3.93108092e-02
3.01783055e-01 4.17281874e-02 3.80643427e-01 -1.07670343e+00
-2.45320827e-01 -5.73122740e-01 -6.39086515e-02 -6.80017352e-01
1.02118742e+00 -1.54696381e+00 -1.13179907e-01 -1.24145830e+00
2.88802117e-01 -4.47922975e-01 -4.23815638e-01 7.38480330e-01
-4.38481063e-01 4.52048033e-01 2.60899007e-01 2.86437362e-01
-5.68200707e-01 4.01470512e-01 9.42716956e-01 -1.66204572e-01
8.85776952e-02 -2.64230639e-01 -1.08313072e+00 6.59376860e-01
7.05577374e-01 -3.29814702e-01 9.13289785e-02 -1.04706359e+00
2.78724253e-01 -6.51553631e-01 -1.09391361e-01 -6.14918411e-01
1.21300191e-01 4.74715419e-02 1.61141053e-01 2.25120652e-02
1.69961303e-01 -4.38617349e-01 -4.28096861e-01 4.04294997e-01
-2.73391724e-01 5.19002557e-01 6.26888454e-01 9.66546461e-02
-1.05095886e-01 1.71603829e-01 4.91097629e-01 -1.31149307e-01
-7.82680094e-01 4.57215726e-01 -4.22473818e-01 4.70719934e-01
7.27295518e-01 3.04604709e-01 -4.24942911e-01 -2.31256932e-01
-5.41619658e-01 5.02144039e-01 3.17722499e-01 9.55892503e-01
2.05296054e-01 -1.39404714e+00 -1.18925560e+00 2.03054488e-01
7.73523211e-01 -6.85873628e-01 -2.34582737e-01 5.14341891e-01
1.08382761e-01 5.49947321e-01 1.78565323e-01 -2.79365748e-01
-9.65791404e-01 5.23620188e-01 3.07406187e-01 -2.95660615e-01
-2.81137913e-01 7.00180054e-01 1.55514985e-01 -7.87160337e-01
-1.92339881e-03 -7.39737973e-02 -2.94601768e-01 3.63863468e-01
3.57162446e-01 -6.34100661e-03 3.74097228e-01 -8.35350096e-01
-6.47390246e-01 5.99135518e-01 -3.81861985e-01 -3.59692037e-01
1.17353523e+00 -1.15530320e-01 -1.53178051e-01 8.42284977e-01
1.42528713e+00 4.00181055e-01 -6.36916637e-01 -6.09918773e-01
2.67149270e-01 -1.22695737e-01 2.26272404e-01 -1.17888474e+00
-6.94082201e-01 6.64614260e-01 4.85131413e-01 3.57202329e-02
6.35004640e-01 2.98756808e-01 7.38678455e-01 4.03044909e-01
4.81998682e-01 -1.18834007e+00 -3.33041102e-01 1.32296550e+00
8.69289637e-01 -1.66946316e+00 -8.68378282e-02 -1.08837917e-01
-4.87094879e-01 9.08718348e-01 6.88573718e-01 1.89875159e-02
8.97461772e-01 4.34530109e-01 3.06310385e-01 -2.39426956e-01
-9.10820007e-01 -3.91952366e-01 6.20239735e-01 3.15076053e-01
9.14316714e-01 3.73367190e-01 -5.58083415e-01 7.73773849e-01
-6.46074355e-01 -5.76611638e-01 1.07817225e-01 5.75547457e-01
-1.56915650e-01 -1.15937364e+00 3.41532975e-02 4.43948865e-01
-6.00217819e-01 -6.53178036e-01 -5.14152706e-01 5.86015344e-01
-2.45057419e-02 1.01863921e+00 8.12218636e-02 -2.21679866e-01
3.12498182e-01 4.50684577e-01 4.34872001e-01 -1.06007981e+00
-7.61468589e-01 -2.84902126e-01 3.32587540e-01 -3.68628234e-01
-7.41487741e-02 -6.83959246e-01 -9.64629650e-01 -1.40452862e-01
-7.05769435e-02 -1.23248607e-01 8.08390737e-01 1.32547307e+00
1.26290366e-01 2.24262342e-01 6.12966776e-01 -5.13871074e-01
-5.22985458e-01 -1.28769898e+00 -2.92649806e-01 4.23442692e-01
5.48192263e-01 -4.83336091e-01 -4.81261730e-01 -2.74794757e-01]
|
[10.735148429870605, 10.223640441894531]
|
bd34bb5e-355a-4585-8dcd-1e005c82abcf
|
cats-complementary-cnn-and-transformer
|
2208.11572
| null |
https://arxiv.org/abs/2208.11572v1
|
https://arxiv.org/pdf/2208.11572v1.pdf
|
Cats: Complementary CNN and Transformer Encoders for Segmentation
|
Recently, deep learning methods have achieved state-of-the-art performance in many medical image segmentation tasks. Many of these are based on convolutional neural networks (CNNs). For such methods, the encoder is the key part for global and local information extraction from input images; the extracted features are then passed to the decoder for predicting the segmentations. In contrast, several recent works show a superior performance with the use of transformers, which can better model long-range spatial dependencies and capture low-level details. However, transformer as sole encoder underperforms for some tasks where it cannot efficiently replace the convolution based encoder. In this paper, we propose a model with double encoders for 3D biomedical image segmentation. Our model is a U-shaped CNN augmented with an independent transformer encoder. We fuse the information from the convolutional encoder and the transformer, and pass it to the decoder to obtain the results. We evaluate our methods on three public datasets from three different challenges: BTCV, MoDA and Decathlon. Compared to the state-of-the-art models with and without transformers on each task, our proposed method obtains higher Dice scores across the board.
|
['Ipek Oguz', 'Jiacheng Wang', 'Han Liu', 'Dewei Hu', 'Hao Li']
|
2022-08-24
| null | null | null | null |
['3d-medical-imaging-segmentation']
|
['medical']
|
[ 1.53829336e-01 2.64049019e-03 -7.33325034e-02 -5.17876446e-01
-7.31739104e-01 -2.14598328e-01 1.72744900e-01 -7.95704573e-02
-6.70510948e-01 4.52829242e-01 -2.88648363e-02 -3.33647966e-01
2.44294778e-01 -7.19234169e-01 -7.87452877e-01 -7.52768815e-01
1.27189621e-01 2.63438791e-01 5.49350560e-01 7.68239498e-02
-1.16016634e-01 3.24286461e-01 -9.57695723e-01 6.58563137e-01
9.62516963e-01 1.43600953e+00 3.00613999e-01 4.83298838e-01
-1.53166071e-01 1.09384763e+00 -4.40655738e-01 -4.74745274e-01
8.56053457e-02 -3.45793158e-01 -8.85960340e-01 -1.70753032e-01
8.94774944e-02 -4.52370584e-01 -5.22243738e-01 9.93374348e-01
6.44511163e-01 -4.60880876e-01 4.94636923e-01 -7.65147507e-01
-5.21373808e-01 5.52601397e-01 -6.42974496e-01 2.80815035e-01
-1.37343571e-01 8.13238472e-02 7.74854243e-01 -5.84600389e-01
4.60887998e-01 1.10083318e+00 8.94234180e-01 4.02838349e-01
-9.35366988e-01 -8.24620068e-01 -2.66770478e-02 3.65819603e-01
-1.29688752e+00 -2.18099251e-01 6.61729813e-01 -4.22947943e-01
8.71990681e-01 -5.17840199e-02 6.78398609e-01 8.63171756e-01
5.47580361e-01 1.20558739e+00 9.72215533e-01 5.50389402e-02
-4.71152877e-03 -1.82651192e-01 7.10271597e-02 9.95689094e-01
-4.27896641e-02 -2.67590676e-02 -2.62386739e-01 3.23659509e-01
8.22432160e-01 1.56923890e-01 -1.97665989e-01 2.53414698e-02
-1.23717928e+00 7.68202722e-01 9.54110563e-01 3.71114820e-01
-5.55550158e-01 3.80622223e-02 5.75311482e-01 9.00731832e-02
6.44233525e-01 3.79441679e-02 -6.11353517e-01 -3.48195694e-02
-1.10963452e+00 -3.66263539e-02 6.10534430e-01 7.77711689e-01
5.84064901e-01 -3.30073953e-01 -5.52911580e-01 8.78611624e-01
2.42912576e-01 1.66522622e-01 5.67607343e-01 -4.82918888e-01
4.27084506e-01 8.58920395e-01 -5.28155625e-01 -7.67075121e-01
-5.64691305e-01 -8.58203530e-01 -1.21232438e+00 2.23806426e-02
3.43766451e-01 -1.71185851e-01 -1.38459873e+00 1.43282425e+00
2.27079004e-01 4.43081856e-01 -1.41533017e-01 9.13011670e-01
1.10955334e+00 5.71852744e-01 -8.44815746e-02 3.31906796e-01
1.45334196e+00 -1.35002708e+00 -8.71176600e-01 -2.25739792e-01
7.58232594e-01 -7.23650575e-01 6.43732429e-01 4.29163843e-01
-1.02376103e+00 -6.75514519e-01 -1.15390015e+00 -4.55297977e-01
-3.80866349e-01 4.53123868e-01 3.56697857e-01 4.39030677e-01
-1.01928353e+00 7.05530226e-01 -1.25482357e+00 -3.80642489e-02
1.04505563e+00 5.33556104e-01 -2.66999632e-01 -2.60859787e-01
-1.09137428e+00 9.14118826e-01 2.41965637e-01 3.75901163e-01
-9.09025788e-01 -6.65614665e-01 -8.21532845e-01 -1.27296057e-02
3.32411528e-02 -6.62003696e-01 1.17683375e+00 -8.55112195e-01
-1.43554449e+00 7.96407759e-01 2.59687267e-02 -6.19733155e-01
5.75257123e-01 -2.30265170e-01 8.39313120e-02 1.63147360e-01
7.44534582e-02 9.08443391e-01 5.66566527e-01 -7.70927608e-01
-6.49271667e-01 -3.46346229e-01 5.39195612e-02 -2.97623035e-02
-2.91881204e-01 -4.03317511e-02 -8.93879294e-01 -6.19686186e-01
7.70434141e-02 -8.00749540e-01 -2.82817155e-01 2.05169871e-01
-6.59487963e-01 -1.24245137e-01 1.07607925e+00 -9.22445476e-01
9.88539279e-01 -2.21157551e+00 1.65873975e-01 -8.27358440e-02
4.08524096e-01 4.62218016e-01 -5.69286272e-02 -8.13683495e-02
-3.97006795e-02 -2.33799834e-02 -3.46387327e-01 -5.90373218e-01
-2.25075826e-01 3.05982381e-01 2.70300210e-01 5.06865025e-01
2.35081360e-01 1.26701140e+00 -6.36065245e-01 -7.72707999e-01
2.61203080e-01 8.13644230e-01 -5.91336668e-01 2.62507588e-01
1.11250445e-01 5.25930583e-01 -5.59947908e-01 5.69383919e-01
8.12783778e-01 -6.05052888e-01 4.97434624e-02 -4.74010080e-01
6.55050352e-02 4.45924610e-01 -6.16743684e-01 1.90819693e+00
-3.90078068e-01 6.41046405e-01 1.11458361e-01 -1.39860725e+00
7.64792740e-01 4.58978415e-01 5.72821081e-01 -9.30280805e-01
4.02362406e-01 2.89027363e-01 1.42217264e-01 -5.08303344e-01
-9.04528499e-02 -3.88634168e-02 1.26389742e-01 4.90643121e-02
2.41070151e-01 1.60533383e-01 4.42142822e-02 6.86703324e-02
1.27357090e+00 1.56181917e-01 1.67151138e-01 -1.27935141e-01
6.24360085e-01 -2.26860330e-01 7.59005249e-01 4.74236190e-01
-1.75168082e-01 8.33493769e-01 6.84502244e-01 -6.53127074e-01
-8.77141774e-01 -7.91499019e-01 -1.80416614e-01 6.14540279e-01
6.17880747e-02 -3.48647296e-01 -1.09231544e+00 -9.84895229e-01
-2.34798789e-01 1.09775126e-01 -8.99629235e-01 -1.19545512e-01
-5.93773246e-01 -8.22965562e-01 8.02804768e-01 7.12881625e-01
9.74429905e-01 -1.01825345e+00 -7.32896566e-01 3.99817437e-01
-3.31425637e-01 -1.51895070e+00 -5.16801536e-01 2.73200989e-01
-9.52359140e-01 -1.10026598e+00 -9.20233607e-01 -9.24332321e-01
6.75469935e-01 -1.35064840e-01 9.78501916e-01 9.68361273e-02
-3.58660132e-01 -3.66661310e-01 -4.22797829e-01 -4.91103739e-01
-1.65738761e-01 4.21019703e-01 -7.17411816e-01 5.98185323e-02
1.98597461e-01 -3.12097698e-01 -9.50136721e-01 2.89497346e-01
-9.03660238e-01 5.24496257e-01 9.77215111e-01 1.05836284e+00
7.41295576e-01 -6.76572472e-02 3.25655073e-01 -1.08045697e+00
3.12380135e-01 -3.24521691e-01 -3.38776290e-01 1.35592908e-01
-1.90276369e-01 1.08362786e-01 7.37919807e-01 -9.58533585e-02
-8.32376242e-01 2.80715495e-01 -6.07149124e-01 -4.61934000e-01
-1.02965154e-01 5.97052038e-01 -3.83702181e-02 -1.81374345e-02
2.40176693e-01 2.42983416e-01 1.15487844e-01 -6.70861244e-01
-7.32911378e-02 7.35700607e-01 3.59239340e-01 -2.53547221e-01
3.78317446e-01 4.67672855e-01 -1.26408637e-01 -4.31005538e-01
-1.06834531e+00 -2.40029618e-01 -8.35090458e-01 -2.44685598e-02
1.39423454e+00 -9.35674131e-01 -5.21064818e-01 8.64121020e-01
-1.35781014e+00 -3.84258181e-01 -7.42078424e-02 4.60780948e-01
-2.97704935e-01 7.68360496e-02 -9.68235612e-01 -2.04628378e-01
-6.45004630e-01 -1.87975895e+00 1.20750082e+00 2.37766653e-01
2.01649636e-01 -1.04333282e+00 -2.52779871e-01 4.60835099e-01
5.53589225e-01 3.61040145e-01 8.65669250e-01 -6.07255578e-01
-5.01526594e-01 -1.78469613e-01 -4.81272668e-01 6.52460933e-01
2.35948145e-01 -3.52519125e-01 -1.07125247e+00 -1.78453013e-01
1.19982380e-02 -2.97744423e-01 1.13466418e+00 6.87527597e-01
1.69128418e+00 -7.76010975e-02 -5.38045168e-01 1.11515009e+00
1.30640209e+00 2.55402863e-01 9.14197445e-01 2.29379907e-01
7.99011707e-01 1.79979846e-01 2.92205572e-01 1.13276884e-01
5.32020926e-01 5.43749869e-01 6.22273266e-01 -7.94647753e-01
-2.81855583e-01 9.38099809e-03 1.70298696e-01 1.05420887e+00
7.17002749e-02 -1.55736938e-01 -9.91279662e-01 4.70615029e-01
-1.77889514e+00 -4.37522918e-01 -5.76489232e-02 1.59242749e+00
9.40666735e-01 1.87616929e-01 -2.18183175e-01 1.21266216e-01
4.49675947e-01 1.14794351e-01 -6.74593568e-01 -2.19091639e-01
1.78703174e-01 5.76656461e-01 6.48304641e-01 1.38137996e-01
-1.38576066e+00 8.55528951e-01 6.11137724e+00 9.22408044e-01
-1.31030631e+00 2.88578629e-01 1.15256560e+00 6.50424734e-02
1.61114916e-01 -3.33880514e-01 -5.34920454e-01 5.20319641e-01
7.99488604e-01 4.68809575e-01 1.75615586e-02 5.83472550e-01
1.45629540e-01 5.27506508e-02 -1.05088818e+00 9.25725102e-01
-3.67319807e-02 -1.46589541e+00 -1.60290357e-02 9.32126567e-02
6.66516960e-01 4.04182106e-01 1.37350753e-01 1.20521434e-01
1.06328137e-01 -1.32269549e+00 5.39884567e-01 3.53076696e-01
9.49986994e-01 -7.17044711e-01 1.23612499e+00 3.36361527e-01
-1.19348812e+00 5.32187968e-02 -3.51237893e-01 2.55264550e-01
6.67692944e-02 8.67030621e-01 -6.59312963e-01 7.40922928e-01
8.95156384e-01 1.09279728e+00 -5.50647795e-01 1.13542473e+00
-2.23828137e-01 7.38620639e-01 -1.40464187e-01 2.05478922e-01
5.81546307e-01 -8.83744359e-02 -4.05155867e-03 1.36201334e+00
3.43184918e-01 1.75795443e-02 1.04807898e-01 8.71826291e-01
-3.90106857e-01 2.59281173e-02 -2.88567543e-01 1.17583185e-01
1.56471096e-02 1.35297692e+00 -9.01991010e-01 -4.89876300e-01
-7.27772713e-01 1.02743113e+00 3.05224448e-01 1.32837445e-01
-1.09893823e+00 -5.89124382e-01 5.64545035e-01 1.96747296e-02
5.56871951e-01 3.46688814e-02 -4.93014693e-01 -1.01711118e+00
-6.02411851e-02 -1.01203954e+00 3.66979122e-01 -5.22505343e-01
-1.08315027e+00 8.26117516e-01 -4.54625905e-01 -1.12376595e+00
1.88746125e-01 -9.04026985e-01 -4.57317591e-01 9.02296245e-01
-1.85539246e+00 -1.29151881e+00 -2.84706503e-01 7.13860214e-01
5.62189937e-01 -5.24013862e-02 6.70113385e-01 7.04883337e-01
-6.37764573e-01 5.89402020e-01 9.58352908e-02 7.36216366e-01
5.88050127e-01 -1.15417528e+00 4.63491201e-01 6.99236751e-01
-1.62695378e-01 3.91222984e-01 6.69279844e-02 -4.43518430e-01
-1.13538802e+00 -1.31937671e+00 6.73149168e-01 -5.62649183e-02
2.54251093e-01 -4.21421289e-01 -8.58563960e-01 6.93830848e-01
4.30002064e-01 5.14549315e-01 5.29145122e-01 -1.90730363e-01
-1.70333445e-01 -2.35990047e-01 -1.08388793e+00 1.30075648e-01
8.11841190e-01 -3.44861060e-01 -2.67111063e-01 1.74996004e-01
6.60049200e-01 -7.69326150e-01 -8.77604008e-01 4.12856400e-01
6.20309770e-01 -1.02621138e+00 9.67014134e-01 -2.21118256e-01
9.19693112e-01 -1.66841686e-01 1.13613173e-01 -1.29861557e+00
-2.40899518e-01 -2.92221099e-01 1.94646761e-01 7.51411617e-01
4.31605875e-01 -4.85986620e-01 7.63171136e-01 -6.88260933e-03
-4.95252728e-01 -1.25719249e+00 -1.03169715e+00 -4.16027606e-01
2.90768862e-01 -3.50664169e-01 6.53052568e-01 6.96930647e-01
-3.46447676e-01 3.42936903e-01 -2.02056676e-01 -1.21921249e-01
5.26795805e-01 -6.61300495e-02 3.98628056e-01 -1.00103641e+00
-1.66174099e-02 -3.52887630e-01 -5.16172111e-01 -1.30854464e+00
-1.23270620e-02 -1.09921575e+00 1.83113962e-01 -1.88012433e+00
4.34944421e-01 -3.13770890e-01 -4.94990975e-01 7.66377389e-01
-1.52178153e-01 5.01034141e-01 8.34892839e-02 -5.43262810e-02
-6.04913533e-01 4.43232834e-01 1.77285147e+00 -4.05538350e-01
1.78182386e-02 -1.49967000e-01 -7.14169204e-01 7.51033485e-01
7.29545116e-01 -5.06889880e-01 -2.93419391e-01 -9.00490761e-01
-1.17013618e-01 -1.19136600e-02 3.62467855e-01 -1.19394922e+00
4.09234047e-01 3.47818702e-01 7.30783820e-01 -7.53971636e-01
2.04797193e-01 -7.79927433e-01 -5.20246178e-02 7.20161796e-01
-3.28767270e-01 3.69779915e-02 2.20012799e-01 1.12573698e-01
-5.39154172e-01 1.72875598e-01 1.00358045e+00 -1.88371390e-01
-3.45748514e-01 7.15184987e-01 -1.80622339e-01 -1.37298107e-02
7.88675427e-01 -7.48015195e-02 -2.33981922e-01 -1.04863048e-01
-7.05807626e-01 2.18342334e-01 3.85540910e-02 2.42548451e-01
7.96803176e-01 -1.14004385e+00 -8.11260641e-01 3.12163293e-01
-1.96490988e-01 4.04521346e-01 2.59831607e-01 1.26609790e+00
-7.77008712e-01 4.69727755e-01 -2.49998510e-01 -8.70627999e-01
-1.13422072e+00 2.49331340e-01 6.94370329e-01 -6.50621414e-01
-8.60157430e-01 9.46216106e-01 5.53479195e-01 -3.90407205e-01
1.48454532e-01 -6.80686235e-01 -2.71277547e-01 -9.58479196e-02
4.73689258e-01 -5.51893078e-02 2.66911685e-01 -5.28490484e-01
-5.26246488e-01 6.17943406e-01 -3.55914265e-01 1.92956582e-01
1.55635083e+00 1.58167034e-01 -2.58744895e-01 4.79851961e-02
1.64043856e+00 -6.10717535e-01 -1.36903262e+00 -3.06378514e-01
-2.78567433e-01 -1.51992708e-01 3.37387592e-01 -8.98040950e-01
-1.86920917e+00 1.42466867e+00 7.07413554e-01 -2.30389819e-01
1.39936292e+00 -5.25825806e-02 1.22303867e+00 2.32817084e-02
7.54337162e-02 -9.42662060e-01 8.79721995e-03 6.27771437e-01
5.82061708e-01 -1.12813962e+00 -1.41819894e-01 -4.54709470e-01
-6.03442669e-01 1.11735272e+00 5.11845708e-01 -1.55184194e-01
8.60432088e-01 7.09762454e-01 1.55919760e-01 -2.37973124e-01
-6.82824790e-01 -1.44389346e-01 4.11180228e-01 4.72478658e-01
6.92830741e-01 -4.67900112e-02 -1.18699610e-01 6.79946959e-01
-1.98135502e-03 3.64805013e-01 6.88839853e-02 7.29970455e-01
-1.30725473e-01 -1.08045173e+00 -6.04515523e-03 6.70816839e-01
-9.13202047e-01 -1.84824362e-01 -1.27781272e-01 6.19404376e-01
4.07389164e-01 6.83897376e-01 1.27671733e-01 -5.19006193e-01
3.12304258e-01 -1.38824835e-01 3.87032807e-01 -4.45860595e-01
-8.59507143e-01 1.66937813e-01 -1.95573568e-01 -7.90556490e-01
-3.81176472e-01 -5.09075403e-01 -1.34065437e+00 -1.11307569e-01
-1.71976298e-01 1.06673846e-02 5.70042133e-01 1.10556829e+00
2.92666942e-01 1.22264171e+00 3.08827668e-01 -5.82218587e-01
-2.96498477e-01 -9.86817181e-01 -4.03286904e-01 3.19753468e-01
4.55755472e-01 -4.73925799e-01 8.77554268e-02 5.63565418e-02]
|
[14.595026969909668, -2.5979437828063965]
|
dd052412-9bd6-43f4-adfc-9856a1b0d3e2
|
cmu-net-a-strong-convmixer-based-medical
|
2210.13012
| null |
https://arxiv.org/abs/2210.13012v4
|
https://arxiv.org/pdf/2210.13012v4.pdf
|
CMU-Net: A Strong ConvMixer-based Medical Ultrasound Image Segmentation Network
|
U-Net and its extensions have achieved great success in medical image segmentation. However, due to the inherent local characteristics of ordinary convolution operations, U-Net encoder cannot effectively extract global context information. In addition, simple skip connections cannot capture salient features. In this work, we propose a fully convolutional segmentation network (CMU-Net) which incorporates hybrid convolutions and multi-scale attention gate. The ConvMixer module extracts global context information by mixing features at distant spatial locations. Moreover, the multi-scale attention gate emphasizes valuable features and achieves efficient skip connections. We evaluate the proposed method using both breast ultrasound datasets and a thyroid ultrasound image dataset; and CMU-Net achieves average Intersection over Union (IoU) values of 73.27% and 84.75%, and F1 scores of 84.81% and 91.71%. The code is available at https://github.com/FengheTan9/CMU-Net.
|
['Jianrui Ding', 'Min Xian', 'Chunping Ning', 'Lingtao Wang', 'Fenghe Tang']
|
2022-10-24
| null | null | null | null |
['tumor-segmentation']
|
['computer-vision']
|
[ 8.74938965e-02 3.46760690e-01 -3.37415844e-01 -4.99727279e-01
-8.59571218e-01 -1.59688175e-01 1.50451586e-01 1.82304844e-01
-4.01645690e-01 5.61217189e-01 3.19140673e-01 -3.82275254e-01
-4.14275266e-02 -6.92163765e-01 -6.33578062e-01 -6.38741553e-01
-4.74011377e-02 -1.07903883e-01 2.19670743e-01 4.66518216e-02
1.46333380e-02 3.04332912e-01 -1.00389290e+00 4.22487766e-01
9.14038360e-01 1.15989780e+00 4.10994262e-01 8.50637794e-01
-2.93572545e-01 8.64359200e-01 -4.85767663e-01 -9.32014436e-02
-1.26641588e-02 -5.00144839e-01 -8.99643719e-01 -1.61940619e-01
2.25058466e-01 -5.13588965e-01 -5.13160825e-01 1.00749481e+00
6.64358437e-01 -2.13286467e-02 3.22663516e-01 -7.28029191e-01
-8.09987843e-01 8.00083280e-01 -8.00713420e-01 6.03494406e-01
-2.04762936e-01 -1.59838554e-02 9.81961429e-01 -5.67403793e-01
5.82469225e-01 8.86154950e-01 5.99505067e-01 5.45218289e-01
-9.18586493e-01 -6.45242929e-01 -2.41022594e-02 -5.41666821e-02
-1.29334092e+00 -5.44702150e-02 4.83229429e-01 -8.59742761e-02
7.81759322e-01 2.96075225e-01 4.47260201e-01 5.88409901e-01
4.51745093e-01 1.28924978e+00 7.98615098e-01 -2.09084943e-01
-1.00155361e-01 -2.62240410e-01 3.43535811e-01 9.12338078e-01
1.12320021e-01 -2.56432921e-01 -7.99879506e-02 2.13808194e-01
1.21854913e+00 4.49609458e-01 -3.14559281e-01 1.89060658e-01
-1.22554243e+00 7.12337911e-01 1.09435451e+00 7.09766448e-01
-3.93635988e-01 4.40702200e-01 3.72480720e-01 -5.12687601e-02
2.64417440e-01 3.49005193e-01 -3.41062307e-01 -9.34102535e-02
-8.00926805e-01 -2.07203642e-01 2.61587232e-01 1.15533543e+00
6.56841636e-01 -1.14178404e-01 -5.77162564e-01 7.28814483e-01
-3.30401212e-02 2.36079514e-01 5.66813171e-01 -8.86533856e-01
9.35718343e-02 5.33524156e-01 -2.84818888e-01 -5.72273493e-01
-6.82624578e-01 -8.91042531e-01 -1.01337636e+00 -3.05021971e-01
2.97008485e-01 -3.97059172e-01 -1.29763174e+00 1.52372468e+00
9.51990560e-02 4.00679201e-01 -3.65723521e-02 9.64548588e-01
1.34411097e+00 4.49980855e-01 2.80583233e-01 1.69685315e-02
1.47727811e+00 -1.14832449e+00 -9.89118874e-01 -2.47026738e-02
8.70331407e-01 -8.90417159e-01 8.85380268e-01 -1.77288130e-01
-1.32924986e+00 -6.57240331e-01 -8.95670593e-01 -3.20838690e-01
-2.65692994e-02 4.16805148e-01 6.58682287e-01 2.52300620e-01
-1.16657293e+00 5.47820985e-01 -1.05097234e+00 -2.08218932e-01
8.92541647e-01 6.11318648e-01 -5.98626137e-02 -1.66016981e-01
-1.07036209e+00 3.44391555e-01 2.77338177e-01 2.18703508e-01
-6.54033959e-01 -7.54860938e-01 -8.50363433e-01 3.78578156e-01
3.61817211e-01 -5.52508354e-01 1.68540478e+00 -8.16745460e-01
-1.15956855e+00 5.43006003e-01 -6.44703880e-02 -3.85697901e-01
4.04222965e-01 -1.83309183e-01 -1.16130792e-01 4.68904525e-01
1.53096378e-01 9.16782975e-01 1.51497692e-01 -7.86710560e-01
-7.74370492e-01 -1.71149790e-01 2.17488378e-01 1.11797474e-01
-1.72270954e-01 5.45778908e-02 -7.41643250e-01 -8.42226326e-01
2.75290877e-01 -6.31929874e-01 -4.15682584e-01 -1.74196698e-02
-3.90781909e-01 -2.89480034e-02 7.97283351e-01 -7.70657420e-01
1.37191153e+00 -2.26298666e+00 -2.54207492e-01 6.12414740e-02
4.29667592e-01 4.47598547e-01 -1.91631168e-01 -1.87550038e-02
1.19370958e-02 1.41520545e-01 -3.80338699e-01 -1.53842285e-01
-3.20827961e-01 2.12339073e-01 3.12595606e-01 1.89244777e-01
4.91745472e-01 1.50767958e+00 -1.12085652e+00 -6.14636958e-01
2.53020078e-01 5.53036034e-01 -6.62629545e-01 -8.89308229e-02
5.24637587e-02 5.34303129e-01 -6.99329972e-01 1.01377308e+00
6.15240335e-01 -4.67068136e-01 1.32227883e-01 -2.71867603e-01
2.92058326e-02 3.57463248e-02 -6.66246414e-01 1.88351715e+00
-4.90714788e-01 6.87586844e-01 1.54027730e-01 -1.06209767e+00
7.47574329e-01 4.12374645e-01 7.27295220e-01 -8.24679136e-01
5.80414057e-01 2.73467839e-01 2.61726737e-01 -4.94371116e-01
4.62704360e-01 -8.70635509e-02 -3.90396938e-02 1.61694452e-01
3.33325505e-01 3.72176647e-01 3.63420770e-02 2.70077676e-01
1.24996150e+00 4.44379672e-02 3.64337146e-01 -5.87943196e-01
3.49244207e-01 -2.79987425e-01 7.28623390e-01 7.91120052e-01
-4.82462496e-01 8.75187874e-01 6.55492246e-01 -4.55748498e-01
-6.83714271e-01 -8.69051218e-01 -4.03320283e-01 7.94128895e-01
2.94032961e-01 -3.36407363e-01 -7.11112916e-01 -6.20439470e-01
-1.02541566e-01 3.46799761e-01 -9.15385544e-01 -1.97948322e-01
-7.73981750e-01 -5.81819713e-01 5.08924007e-01 1.17732048e+00
7.14684844e-01 -1.09522963e+00 -6.00459993e-01 3.94784778e-01
-2.90873408e-01 -1.03394282e+00 -6.57317519e-01 1.22643113e-01
-1.10256803e+00 -9.74200845e-01 -9.93615508e-01 -1.06277180e+00
9.12892342e-01 3.43535990e-01 1.07738876e+00 2.60521948e-01
-6.87813044e-01 -1.69136390e-01 -3.35984677e-01 -2.86640048e-01
-7.71106631e-02 2.46205226e-01 -6.63382173e-01 -2.17978969e-01
4.18698639e-01 -3.12491804e-01 -9.54289377e-01 2.28007987e-01
-9.37054038e-01 1.78387061e-01 9.93534446e-01 1.07858229e+00
8.54596972e-01 -4.48830336e-01 6.03247881e-01 -9.78716671e-01
2.44589910e-01 -5.08478880e-01 -2.72220999e-01 1.25431582e-01
5.09683508e-04 -2.02246919e-01 3.39135975e-01 -1.88984200e-01
-9.93120313e-01 -6.53081089e-02 -3.83772194e-01 -5.47600746e-01
-1.00598641e-01 5.27172685e-01 1.16754860e-01 1.90489918e-01
3.86461943e-01 1.65820092e-01 -1.06312513e-01 -4.22839195e-01
7.56631941e-02 6.76146388e-01 5.82002521e-01 -2.59376913e-01
6.33186474e-02 4.66640741e-01 -2.76457667e-01 -8.02420259e-01
-8.67598116e-01 -5.26659250e-01 -5.64356446e-01 1.93937905e-02
1.15881789e+00 -8.35248053e-01 -4.27323073e-01 2.52135664e-01
-8.32867682e-01 -4.75553960e-01 -3.74423474e-01 6.47232354e-01
-3.13069522e-01 1.53817564e-01 -9.81949389e-01 -3.79624337e-01
-5.41055262e-01 -1.38592303e+00 1.05663157e+00 7.33673930e-01
-4.62396815e-03 -8.30693245e-01 -6.56986654e-01 2.39708722e-01
5.60161769e-01 3.56045067e-01 4.92184430e-01 -5.00996649e-01
-7.09896743e-01 -2.68914312e-01 -6.90781713e-01 1.80269286e-01
4.31631327e-01 -5.52227125e-02 -9.40893233e-01 -2.82493085e-01
-3.01368564e-01 -1.22795507e-01 1.14089561e+00 1.16372466e+00
1.66419399e+00 4.29334072e-03 -4.33256447e-01 8.40056002e-01
1.33791745e+00 3.84065688e-01 5.93377948e-01 -5.20254970e-02
5.83634973e-01 2.82949489e-02 4.95112360e-01 2.93829799e-01
1.83542013e-01 2.08741412e-01 4.23437178e-01 -6.18375778e-01
-3.71803194e-01 1.08955130e-01 -2.37561107e-01 7.68009067e-01
-9.32090953e-02 1.96528118e-02 -9.90564048e-01 8.79138231e-01
-1.81913960e+00 -6.02104187e-01 -2.76982058e-02 1.64206672e+00
6.32354379e-01 2.19073325e-01 -1.72718316e-01 -2.36188799e-01
7.27832139e-01 4.41253074e-02 -4.65643734e-01 -3.60872567e-01
4.98009585e-02 4.24846828e-01 6.48591161e-01 4.46416914e-01
-1.35046065e+00 7.45820940e-01 5.22366858e+00 8.56277943e-01
-1.13728082e+00 3.14361095e-01 1.06200469e+00 -2.33840555e-01
-1.63663417e-01 -2.96423554e-01 -5.36968887e-01 2.59322673e-01
5.81633627e-01 1.60670772e-01 -2.49277130e-01 6.09490931e-01
-1.04977027e-01 -6.88352138e-02 -5.06869912e-01 6.91757560e-01
-1.97783962e-01 -1.51156199e+00 -2.99763054e-01 5.04838973e-02
1.00888181e+00 2.83666939e-01 4.55151871e-02 2.95220643e-01
2.30821818e-01 -9.98471916e-01 1.99942335e-01 3.34741086e-01
9.94499266e-01 -8.80963087e-01 1.20330846e+00 2.21401770e-02
-1.24138832e+00 -3.88412438e-02 -3.00573677e-01 1.60498127e-01
5.20499125e-02 5.18557787e-01 -6.52592421e-01 6.32131994e-01
8.37110937e-01 6.84219360e-01 -3.03067416e-01 1.10319006e+00
-2.26469100e-01 6.34447217e-01 -3.41714412e-01 -1.64287910e-02
8.57460380e-01 6.84621781e-02 2.64457226e-01 1.52342081e+00
2.49769136e-01 3.62583011e-01 7.77943283e-02 6.56651914e-01
-3.55640382e-01 1.73608974e-01 -1.87518761e-01 5.48685454e-02
2.40134224e-01 1.45259404e+00 -1.05813634e+00 -3.73843640e-01
-6.27880991e-01 7.88932443e-01 9.20019224e-02 3.79595935e-01
-1.06055307e+00 -6.98267281e-01 6.14443362e-01 -1.71397105e-01
4.53590274e-01 1.62200853e-01 -4.56770360e-01 -1.04140353e+00
-1.17058381e-01 -3.95300418e-01 4.29826438e-01 -5.36451221e-01
-7.61221886e-01 7.90196657e-01 -3.03900540e-01 -1.09044409e+00
2.64325559e-01 -4.25695658e-01 -7.69667029e-01 7.97641039e-01
-1.76928806e+00 -1.36930680e+00 -5.07091939e-01 2.92846292e-01
5.89571416e-01 5.22628762e-02 5.49024463e-01 3.78820777e-01
-6.31880939e-01 8.01717103e-01 2.59246200e-01 5.69552004e-01
5.12024283e-01 -1.35676897e+00 3.80955398e-01 6.90109789e-01
-3.04925650e-01 6.66083276e-01 1.36139989e-01 -5.27893841e-01
-9.12600100e-01 -1.31997275e+00 7.88679361e-01 4.33840454e-02
4.45587516e-01 -1.57745797e-02 -9.55637455e-01 7.88257480e-01
2.93865263e-01 5.11831343e-01 7.38668740e-01 -1.46008149e-01
1.10462099e-01 8.18981528e-02 -1.12504506e+00 5.25035381e-01
8.69147301e-01 -1.45986289e-01 -2.05468059e-01 1.47067025e-01
8.22727621e-01 -7.87501097e-01 -1.14909804e+00 5.59573114e-01
6.16131544e-01 -8.23323190e-01 8.79013062e-01 -4.26300436e-01
9.02703345e-01 -9.00643095e-02 -1.15274563e-01 -8.61458302e-01
-5.20645738e-01 -4.31316406e-01 -8.96942317e-02 8.81154239e-01
4.61169958e-01 -5.77534974e-01 7.25230753e-01 1.80787489e-01
-6.78860962e-01 -1.39766920e+00 -7.58544207e-01 -2.99782336e-01
1.71897307e-01 -1.42850384e-01 5.35335004e-01 8.26715648e-01
-1.64746180e-01 7.77961835e-02 -4.22651805e-02 1.10802852e-01
4.48603094e-01 3.04936916e-01 1.25879869e-01 -7.37332582e-01
-3.23686630e-01 -6.42210007e-01 -3.50069761e-01 -1.08555400e+00
-3.30271393e-01 -9.62359667e-01 -1.48235843e-01 -1.75562727e+00
3.34100276e-01 -4.07237828e-01 -8.60325754e-01 7.17430890e-01
-4.33818460e-01 5.43352544e-01 1.18615195e-01 8.50937739e-02
-4.74430501e-01 2.80518085e-01 1.86092722e+00 -1.38135508e-01
-2.25876585e-01 2.58321855e-02 -8.95380020e-01 5.85816264e-01
1.44467890e+00 -7.02651069e-02 -2.16805354e-01 -6.02571070e-01
-4.23618823e-01 2.08261773e-01 1.06953956e-01 -9.03073370e-01
1.39646813e-01 1.11079916e-01 6.52084708e-01 -6.63181424e-01
-3.18499096e-02 -5.04704475e-01 -2.88470089e-01 8.39968622e-01
-3.73257488e-01 -2.26978540e-01 3.93983930e-01 2.98417330e-01
-4.44151759e-01 -1.13514483e-01 7.79051781e-01 -3.27397823e-01
-8.22262824e-01 4.26416904e-01 -1.82032436e-01 -1.21685565e-01
1.04145634e+00 -1.68656230e-01 -2.06784293e-01 -1.96516454e-01
-8.60536814e-01 4.99327451e-01 2.65773851e-02 2.83512741e-01
6.13326192e-01 -1.25407767e+00 -7.30477691e-01 3.14801574e-01
5.49391843e-02 3.92058432e-01 6.50441647e-01 1.22545719e+00
-9.17021453e-01 5.56092799e-01 -2.36259833e-01 -7.06350625e-01
-1.20117342e+00 1.95116252e-01 4.09893543e-01 -2.61024296e-01
-8.56481314e-01 1.16799808e+00 5.74766695e-01 -2.95184702e-01
1.56701744e-01 -8.43836546e-01 -9.28359926e-02 -1.66661352e-01
6.29463911e-01 1.90163165e-01 7.81534687e-02 -4.68527913e-01
-2.12352991e-01 3.99546802e-01 -3.43312860e-01 4.65752423e-01
1.24626493e+00 5.86862043e-02 3.72665115e-02 -1.16170324e-01
1.39700544e+00 -3.81405443e-01 -1.26716697e+00 -5.29355466e-01
-2.11115405e-01 -2.80925632e-01 3.75874132e-01 -7.87612677e-01
-1.57393181e+00 1.16021955e+00 5.34138203e-01 -6.76468909e-02
1.41728616e+00 1.65865287e-01 1.07652533e+00 4.72640507e-02
-1.26983628e-01 -7.77886033e-01 -7.22958660e-03 3.97327125e-01
6.29935920e-01 -1.25776136e+00 -1.33316219e-01 -4.67862755e-01
-5.94111443e-01 9.14654195e-01 8.04085910e-01 -2.15705723e-01
8.03226590e-01 4.43304747e-01 2.97828317e-01 -2.08555758e-01
-4.65266943e-01 -4.19164181e-01 1.91643149e-01 1.92399964e-01
8.37787151e-01 1.60535812e-01 -5.15824378e-01 7.74449348e-01
1.55067429e-01 1.32455826e-01 4.16353464e-01 1.22772002e+00
-4.02704090e-01 -7.34937012e-01 -1.39699370e-01 8.07126701e-01
-9.92441356e-01 -5.79076335e-02 1.18385546e-01 7.55303264e-01
1.30356371e-01 5.84479690e-01 2.72989780e-01 -3.54451716e-01
1.61054730e-01 -2.13325649e-01 3.15407902e-01 -4.54211950e-01
-7.06928372e-01 7.14184999e-01 -1.81523800e-01 -5.93645811e-01
-3.50210249e-01 -4.73132163e-01 -1.69441605e+00 -7.66794682e-02
-4.87917423e-01 2.33082194e-02 3.74660760e-01 6.04284942e-01
3.69360715e-01 1.35830808e+00 3.96859795e-01 -5.77223122e-01
-2.45726839e-01 -1.07719541e+00 -5.41773438e-01 2.13465020e-01
4.80340898e-01 -4.45608824e-01 2.24866420e-01 -7.33068064e-02]
|
[14.649928092956543, -2.6139607429504395]
|
c778c875-96b6-472e-9159-d8856fa44b8f
|
multi-view-partial-mvp-point-cloud-challenge
|
2112.12053
| null |
https://arxiv.org/abs/2112.12053v1
|
https://arxiv.org/pdf/2112.12053v1.pdf
|
Multi-View Partial (MVP) Point Cloud Challenge 2021 on Completion and Registration: Methods and Results
|
As real-scanned point clouds are mostly partial due to occlusions and viewpoints, reconstructing complete 3D shapes based on incomplete observations becomes a fundamental problem for computer vision. With a single incomplete point cloud, it becomes the partial point cloud completion problem. Given multiple different observations, 3D reconstruction can be addressed by performing partial-to-partial point cloud registration. Recently, a large-scale Multi-View Partial (MVP) point cloud dataset has been released, which consists of over 100,000 high-quality virtual-scanned partial point clouds. Based on the MVP dataset, this paper reports methods and results in the Multi-View Partial Point Cloud Challenge 2021 on Completion and Registration. In total, 128 participants registered for the competition, and 31 teams made valid submissions. The top-ranked solutions will be analyzed, and then we will discuss future research directions.
|
['Yang Yang', 'Qinlong Wang', 'Francisco Gómez-Fernández', 'Xin Li', 'Dongrui Liu', 'Changwei Lin', 'Lifa Zhu', 'Junyi An', 'Yuanjie Yan', 'Zhizhong Han', 'Yu-Shen Liu', 'Peng Xiang', 'Xin Wen', 'Junsheng Zhou', 'Yu Qiao', 'Yali Wang', 'Manning Wang', 'Peng Gao', 'Kexue Fu', 'Xiaoyuan Luo', 'Mingye Xu', 'Jie zhou', 'Jiwen Lu', 'Yongming Rao', 'Xumin Yu', 'Ziwei Liu', 'Zhongang Cai', 'Tong Wu', 'Liang Pan']
|
2021-12-22
| null | null | null | null |
['point-cloud-completion']
|
['computer-vision']
|
[-1.47693098e-01 -2.00072423e-01 7.98946396e-02 -5.01107097e-01
-1.05841517e+00 -7.54924715e-01 5.36943316e-01 -1.94771111e-01
4.63611260e-02 9.05245077e-03 -1.15709022e-01 1.20249346e-01
1.91818967e-01 -4.37331110e-01 -9.26727474e-01 -2.13234812e-01
2.81388313e-01 1.32194197e+00 4.27445412e-01 -1.38745263e-01
5.29553771e-01 8.54214489e-01 -1.75064826e+00 4.52736884e-01
5.99326193e-01 7.11936533e-01 5.22529483e-01 4.02312279e-01
-2.82427639e-01 -4.54425097e-01 -1.83071673e-01 -3.38721544e-01
6.81248963e-01 4.85904843e-01 -6.11856163e-01 2.80375987e-01
1.09278810e+00 -3.33527207e-01 9.30736810e-02 9.59799469e-01
3.08101118e-01 -2.52941579e-01 3.34249467e-01 -1.64671302e+00
-2.77324945e-01 -2.72895575e-01 -8.52231145e-01 -3.75918210e-01
6.78095996e-01 -4.69488092e-03 1.03898191e+00 -1.70409024e+00
1.01335406e+00 1.24415410e+00 8.72248411e-01 3.52873951e-01
-1.21198022e+00 -6.61198854e-01 -2.62852851e-02 2.44287565e-01
-1.50140119e+00 -3.03185672e-01 8.67591798e-01 -5.31030536e-01
1.05034459e+00 3.76162589e-01 8.51890981e-01 7.15087950e-01
-8.77408162e-02 5.39776564e-01 1.03898025e+00 1.74584817e-02
9.79130641e-02 -6.83964416e-02 -8.42955261e-02 1.88281298e-01
3.94678891e-01 1.08184844e-01 -5.53555131e-01 -6.79320157e-01
1.03300130e+00 2.35601172e-01 -3.09580922e-01 -1.04555786e+00
-1.49499667e+00 4.95274872e-01 2.64410168e-01 -1.03343919e-01
-4.74927694e-01 -2.19494969e-01 -1.01149799e-02 2.54085422e-01
6.03844881e-01 1.35634944e-01 -7.02947319e-01 -1.27703890e-01
-9.54932630e-01 5.23731530e-01 5.99821806e-01 1.39355671e+00
1.06185985e+00 -3.58920217e-01 7.81309307e-01 6.70986056e-01
4.93847430e-01 9.66180682e-01 -3.90405029e-01 -1.48599899e+00
7.88086772e-01 8.01436424e-01 2.37288460e-01 -9.57174122e-01
-3.69708300e-01 -7.27035478e-02 -7.42372870e-01 5.44488490e-01
-2.94601098e-02 4.73025531e-01 -6.91195965e-01 9.48311746e-01
6.50337636e-01 3.07794780e-01 -3.20519209e-01 1.15425050e+00
1.22898209e+00 4.20062125e-01 -7.20952690e-01 -1.27921030e-01
1.28783023e+00 -6.85520053e-01 -3.27645153e-01 -2.29412809e-01
3.29782367e-02 -1.05686474e+00 8.28723907e-01 7.12622404e-01
-1.30279315e+00 -6.41590595e-01 -1.07923555e+00 -3.30963224e-01
2.63965249e-01 -3.48557085e-01 1.24174923e-01 1.82693973e-01
-1.19949162e+00 5.82085192e-01 -1.07692814e+00 -1.80061623e-01
4.60355341e-01 4.19797421e-01 -8.39151680e-01 -5.63376665e-01
-2.10589886e-01 6.14753306e-01 -2.43882224e-01 1.78092476e-02
-4.90613520e-01 -1.21383560e+00 -5.04945695e-01 -2.83702254e-01
2.96775013e-01 -1.20370865e+00 1.05239320e+00 -1.78296044e-01
-8.95646334e-01 1.31430471e+00 -5.40463507e-01 1.17258310e-01
5.92881382e-01 -2.30149776e-01 -1.68104276e-01 3.09678733e-01
3.53710383e-01 6.77636206e-01 6.08751655e-01 -1.84844458e+00
-5.96274674e-01 -1.02376354e+00 -3.19956064e-01 4.08630013e-01
3.98429215e-01 -6.10550381e-02 -8.24359953e-01 -1.83967426e-02
1.18142712e+00 -1.14604032e+00 -3.11591834e-01 2.38387376e-01
-1.95885241e-01 -2.00833127e-01 1.01915026e+00 -4.57124352e-01
1.75120421e-02 -2.05887866e+00 2.44227871e-01 1.33215889e-01
4.39515412e-01 -2.16465801e-01 -2.07915306e-01 5.21224380e-01
-2.00497046e-01 4.61718179e-02 -1.05021462e-01 -1.08992934e+00
-1.53736100e-01 2.98695326e-01 -5.05665004e-01 6.87036514e-01
-3.44014585e-01 8.40919435e-01 -6.51051581e-01 -3.45001757e-01
4.88387495e-01 5.05100310e-01 -3.97460282e-01 -5.69174662e-02
-2.29741842e-01 7.66356409e-01 -4.16036904e-01 9.44020748e-01
1.53451109e+00 -4.26990300e-01 -2.93471158e-01 -3.81296873e-01
-2.91524768e-01 -1.21237837e-01 -1.39149773e+00 2.34720898e+00
-2.37644389e-01 3.27569962e-01 5.02458394e-01 -3.34172726e-01
9.97693956e-01 4.61184829e-01 1.10058534e+00 -2.24008888e-01
-2.25998506e-01 4.33567643e-01 -6.50644481e-01 -3.06324498e-03
7.68564522e-01 -1.13734491e-01 2.23735213e-01 3.12671065e-01
-2.66603023e-01 -9.79324400e-01 -3.62273812e-01 1.98457107e-01
9.49914992e-01 3.17792952e-01 1.00082397e-01 1.30173698e-01
2.67672181e-01 5.85389555e-01 5.62467337e-01 2.04201177e-01
1.78857613e-02 1.64028466e+00 -7.32584372e-02 -7.35314310e-01
-1.25091505e+00 -1.30123198e+00 -3.02407444e-01 1.92218795e-01
3.65292609e-01 -5.75358391e-01 -1.30361050e-01 -2.61302114e-01
1.88231036e-01 2.55814195e-01 -9.62691233e-02 5.22261024e-01
-6.42878830e-01 -1.50679767e-01 -1.69827208e-01 3.00694317e-01
2.32600406e-01 -7.96572864e-01 -4.23878849e-01 -8.43477920e-02
-5.26025832e-01 -1.51723051e+00 -3.61230373e-01 -5.59853673e-01
-1.58010852e+00 -1.34483945e+00 -7.03783274e-01 -4.99940693e-01
6.71769977e-01 1.20497704e+00 1.50146401e+00 2.26757199e-01
4.75294404e-02 8.28282773e-01 -3.45384240e-01 -4.69523847e-01
-9.15621072e-02 -2.04323813e-01 3.59423965e-01 -1.67778596e-01
3.68961096e-01 -9.64378238e-01 -4.90826458e-01 7.78621793e-01
-6.25192702e-01 1.45126596e-01 3.00823957e-01 3.20996523e-01
1.29821134e+00 -5.48953950e-01 -2.20976099e-01 -5.73155940e-01
-1.37516260e-01 -1.15188472e-01 -1.02910292e+00 -5.52588739e-02
-3.19761336e-01 -5.80494404e-01 6.33652732e-02 3.53456363e-02
-6.79065943e-01 5.19492090e-01 -1.42621577e-01 -1.35547030e+00
-1.85321793e-01 1.87950820e-01 -2.73100257e-01 -4.07973945e-01
4.89720225e-01 1.71850458e-01 9.56379101e-02 -9.53643441e-01
2.60467857e-01 3.22661757e-01 5.10295570e-01 -3.36657107e-01
1.23546016e+00 1.17007899e+00 2.58790672e-01 -9.66029644e-01
-4.63898271e-01 -9.92633343e-01 -1.07706606e+00 -3.03510278e-02
5.71800768e-01 -1.25058460e+00 -6.91991925e-01 3.54128271e-01
-1.74980032e+00 1.74870729e-01 -3.08005244e-01 4.76718873e-01
-7.11938322e-01 7.48324633e-01 -6.45171329e-02 -4.60631371e-01
-1.97548255e-01 -1.25788987e+00 1.68292165e+00 -2.89948851e-01
3.36219296e-02 -5.62094629e-01 4.72353578e-01 7.21381247e-01
-8.82492587e-02 1.52044073e-01 2.39129603e-01 -1.98541373e-01
-1.34297597e+00 -3.98870349e-01 -1.38268337e-01 3.67431454e-02
-1.25402302e-01 4.19147536e-02 -9.11167979e-01 -6.41630709e-01
4.54077691e-01 -4.05977555e-02 4.34409976e-01 5.82726538e-01
7.97041059e-01 4.16608304e-01 -5.04405558e-01 1.06783903e+00
1.55830717e+00 -4.27426517e-01 5.49119949e-01 2.59360254e-01
9.42720890e-01 5.68415225e-01 8.67167771e-01 5.50626338e-01
7.77856708e-01 1.06910300e+00 1.15019166e+00 2.73079187e-01
8.90382938e-03 -1.64643958e-01 -8.71196687e-02 1.12361574e+00
-5.81109941e-01 3.64064872e-01 -1.12034190e+00 3.67241174e-01
-1.38942945e+00 -9.14963365e-01 -1.06343651e+00 2.38059711e+00
6.69746473e-02 -2.25082457e-01 -9.60085019e-02 -1.13073863e-01
5.92516780e-01 1.67346194e-01 -5.31816423e-01 3.68878216e-01
-2.68332452e-01 7.84726217e-02 2.92321801e-01 4.71099108e-01
-7.25512564e-01 5.86963356e-01 6.37136269e+00 4.43355411e-01
-7.59672761e-01 2.83108503e-01 -1.49261564e-01 -1.45292342e-01
-5.12030542e-01 4.26240653e-01 -8.39437127e-01 -1.71624944e-02
2.81433672e-01 1.89695098e-02 2.26317212e-01 9.05660331e-01
-5.93200289e-02 -2.95842439e-03 -1.08129954e+00 1.64545608e+00
8.35690424e-02 -1.50767386e+00 3.27439606e-02 6.65884316e-01
9.54701245e-01 1.10221803e+00 -8.42646360e-02 -1.80191621e-01
4.19739597e-02 -6.16627336e-01 6.15238965e-01 5.01070499e-01
9.64100540e-01 -4.12750840e-01 4.91516978e-01 8.37228358e-01
-1.43071437e+00 5.04193127e-01 -8.56054604e-01 -5.09250618e-04
5.41592419e-01 6.29048347e-01 -6.70630872e-01 1.19793987e+00
9.77715790e-01 1.03027141e+00 -5.14661849e-01 1.32128072e+00
9.06760991e-02 -2.57120840e-02 -6.24059737e-01 6.57598853e-01
-3.68712783e-01 -5.67515552e-01 1.00742507e+00 2.07630202e-01
5.93541145e-01 4.34326679e-01 3.26130837e-01 9.21556652e-01
1.73424363e-01 -2.22211808e-01 -6.95025206e-01 6.79976285e-01
4.97375220e-01 1.43661594e+00 -5.31601787e-01 -1.09001167e-01
-6.38979435e-01 9.12620425e-01 1.01562217e-01 2.10924521e-01
-3.10850233e-01 4.55772221e-01 7.82439649e-01 2.87563562e-01
1.12303697e-01 -7.00470150e-01 -6.81712627e-01 -1.46857238e+00
4.95288402e-01 -5.41303158e-01 2.93701887e-02 -1.31215584e+00
-1.40813100e+00 7.26527333e-01 2.11165428e-01 -2.10787988e+00
-4.84679389e-04 -3.86164516e-01 -2.98912257e-01 1.16433060e+00
-1.39256251e+00 -1.28860772e+00 -7.74372697e-01 6.58073366e-01
6.86476350e-01 -2.86492463e-02 7.69177973e-01 1.41276628e-01
3.37520033e-01 -1.82943746e-01 -6.91417977e-02 -3.92675966e-01
5.47520757e-01 -9.23882604e-01 1.02630210e+00 3.79980952e-01
5.03746390e-01 5.42439222e-01 6.66618109e-01 -8.72684538e-01
-1.94056129e+00 -8.09668303e-01 7.93786287e-01 -1.17940962e+00
2.83826143e-02 -3.48946422e-01 -1.18394995e+00 8.82370591e-01
-8.07028711e-02 5.61294615e-01 2.04107076e-01 -8.35589692e-02
-3.13976705e-01 1.09455720e-01 -1.23891234e+00 1.39701724e-01
1.30780196e+00 -3.25059772e-01 -7.61402547e-01 4.26291138e-01
7.57020175e-01 -1.07193041e+00 -8.33172023e-01 5.32728136e-01
3.63014787e-01 -1.18729627e+00 1.41614413e+00 -1.59885138e-01
2.53179878e-01 -3.42314035e-01 -4.05014008e-01 -1.22922266e+00
-2.82145977e-01 -4.85630453e-01 1.18286736e-01 7.57927239e-01
-6.00487441e-02 -5.26660442e-01 1.33912241e+00 4.91169989e-01
-5.82663178e-01 -4.28696752e-01 -1.26438844e+00 -8.60682249e-01
-2.44762599e-01 -7.77554691e-01 8.28131437e-01 9.10824537e-01
-6.28404558e-01 1.68750435e-01 -1.19404398e-01 6.44413650e-01
1.14347398e+00 6.33562446e-01 1.47681141e+00 -1.91380012e+00
-8.35809782e-02 1.40934121e-02 -4.73422021e-01 -1.35911071e+00
-1.04551822e-01 -9.54169333e-01 -2.48598084e-01 -1.73899364e+00
1.37276232e-01 -5.90428412e-01 4.76575077e-01 9.56353247e-02
2.28312999e-01 5.38123965e-01 4.45058107e-01 9.05477345e-01
-4.14642930e-01 5.29413760e-01 1.34213769e+00 4.82838899e-02
-1.23746820e-01 4.43529367e-01 -9.46247652e-02 8.42105508e-01
3.37435871e-01 -4.00425971e-01 -1.09491535e-01 -8.44978452e-01
4.23721194e-01 4.99646664e-01 6.84523880e-01 -8.70624542e-01
3.32025081e-01 -2.78020471e-01 1.32455528e-01 -1.84912848e+00
1.32327282e+00 -1.31648827e+00 7.97419310e-01 8.92737135e-02
5.73713064e-01 4.82039422e-01 2.65886933e-02 5.29145002e-01
-2.63572127e-01 -3.46637913e-03 4.05052483e-01 -4.84258145e-01
-4.95899737e-01 9.40162778e-01 6.01617515e-01 -8.13249499e-03
9.35082376e-01 -6.28404081e-01 -2.85581108e-02 -1.99584201e-01
-8.35074842e-01 3.94577205e-01 1.20345616e+00 6.14093184e-01
1.23494148e+00 -1.56870902e+00 -1.00095832e+00 3.99197221e-01
3.94317240e-01 7.81712592e-01 6.89830124e-01 6.57064080e-01
-6.88206494e-01 4.06898797e-01 -2.74666607e-01 -1.51457119e+00
-1.64905822e+00 4.56945032e-01 -9.18711722e-02 1.53776303e-01
-1.05348158e+00 6.13256335e-01 1.98534504e-01 -9.42989111e-01
-2.50887811e-01 -2.01240495e-01 7.80308321e-02 -4.22120601e-01
4.84946042e-01 5.48412919e-01 6.04418814e-01 -9.63505089e-01
-4.50030357e-01 1.38719165e+00 2.93153256e-01 -3.84674668e-01
1.72712040e+00 -1.85291231e-01 -3.37165236e-01 4.45755988e-01
1.22618377e+00 1.41864866e-01 -1.24466491e+00 -4.46310699e-01
-2.62453765e-01 -9.79425311e-01 -1.16594985e-01 -2.08980128e-01
-1.11687934e+00 1.06576157e+00 3.23353559e-01 -7.34650642e-02
7.62451589e-01 3.81682843e-01 7.11348355e-01 2.17371151e-01
1.12533092e+00 -5.07245958e-01 -2.00458735e-01 6.22724891e-01
1.46089196e+00 -1.29203904e+00 4.86036777e-01 -9.46645975e-01
-4.24014509e-01 1.07603765e+00 5.77012718e-01 -3.79548103e-01
6.69649541e-01 -5.56379184e-02 -6.53785765e-02 -6.77515864e-01
-7.64886558e-01 4.47867095e-01 4.57532644e-01 8.95721078e-01
-3.32994610e-01 -1.99713111e-02 1.95073664e-01 2.73693919e-01
-4.38040048e-01 -5.03648780e-02 5.08015692e-01 8.31432700e-01
-2.75981933e-01 -1.33004713e+00 -9.94194090e-01 3.31072152e-01
1.96885362e-01 2.37927631e-01 -4.26090986e-01 6.83593214e-01
-1.57663748e-01 7.59982169e-01 2.51484960e-01 -3.68704617e-01
8.81536424e-01 -3.19533467e-01 6.24434590e-01 -8.20749760e-01
-3.34691554e-01 4.38517034e-02 -4.05503184e-01 -1.08371234e+00
-5.14763594e-01 -1.16237545e+00 -1.14350212e+00 -4.27590936e-01
-2.02244967e-01 -7.70670027e-02 1.07945538e+00 5.44722378e-01
7.17778683e-01 -2.45623186e-01 7.70537615e-01 -1.59564042e+00
-7.07242131e-01 -6.78197503e-01 -5.83758831e-01 5.06970465e-01
3.13283980e-01 -7.60715723e-01 -3.28589529e-01 -1.15310773e-01]
|
[8.295578956604004, -3.1789424419403076]
|
d72f408e-241e-453a-89fd-a81103e9de85
|
transforming-model-prediction-for-tracking
|
2203.11192
| null |
https://arxiv.org/abs/2203.11192v1
|
https://arxiv.org/pdf/2203.11192v1.pdf
|
Transforming Model Prediction for Tracking
|
Optimization based tracking methods have been widely successful by integrating a target model prediction module, providing effective global reasoning by minimizing an objective function. While this inductive bias integrates valuable domain knowledge, it limits the expressivity of the tracking network. In this work, we therefore propose a tracker architecture employing a Transformer-based model prediction module. Transformers capture global relations with little inductive bias, allowing it to learn the prediction of more powerful target models. We further extend the model predictor to estimate a second set of weights that are applied for accurate bounding box regression. The resulting tracker relies on training and on test frame information in order to predict all weights transductively. We train the proposed tracker end-to-end and validate its performance by conducting comprehensive experiments on multiple tracking datasets. Our tracker sets a new state of the art on three benchmarks, achieving an AUC of 68.5% on the challenging LaSOT dataset.
|
['Luc van Gool', 'Fisher Yu', 'Danda Pani Paudel', 'Matthieu Paul', 'Goutam Bhat', 'Martin Danelljan', 'Christoph Mayer']
|
2022-03-21
| null |
http://openaccess.thecvf.com//content/CVPR2022/html/Mayer_Transforming_Model_Prediction_for_Tracking_CVPR_2022_paper.html
|
http://openaccess.thecvf.com//content/CVPR2022/papers/Mayer_Transforming_Model_Prediction_for_Tracking_CVPR_2022_paper.pdf
|
cvpr-2022-1
|
['visual-object-tracking']
|
['computer-vision']
|
[-2.89426625e-01 2.80346900e-01 -8.76707196e-01 -3.48610640e-01
-6.88985467e-01 -6.26699984e-01 7.85924852e-01 -2.19204843e-01
-2.33755246e-01 6.63507938e-01 2.64516138e-02 1.19369933e-02
-4.77733985e-02 -4.65546340e-01 -8.38912606e-01 -3.52943689e-01
-2.12332413e-01 6.95954740e-01 7.03833580e-01 1.45304576e-01
-1.95644125e-01 4.39651579e-01 -1.27051592e+00 -6.34294562e-03
7.55871892e-01 1.56789589e+00 -1.84716359e-01 3.71390313e-01
2.77334392e-01 1.05845523e+00 -4.89167184e-01 -6.58535421e-01
5.03785849e-01 9.24221054e-02 -1.37846753e-01 -3.70830804e-01
9.13397312e-01 -2.50761002e-01 -2.14290962e-01 9.21716154e-01
3.36159885e-01 -1.16042621e-01 3.95333111e-01 -1.31872535e+00
-2.55145133e-01 7.65723407e-01 -2.31021225e-01 2.96412706e-01
1.06060728e-01 3.51789981e-01 1.30277491e+00 -7.90746570e-01
6.20459616e-01 1.32331479e+00 1.24731255e+00 6.85652852e-01
-1.28635418e+00 -8.87677848e-01 3.90699178e-01 1.49501950e-01
-1.04951966e+00 -5.96922159e-01 7.39410818e-01 -5.07743895e-01
8.21456075e-01 1.04081675e-01 8.16382945e-01 1.12279630e+00
4.85774159e-01 7.26094544e-01 9.37784672e-01 -7.07640350e-02
2.06749663e-01 6.00593239e-02 4.20622304e-02 7.39398360e-01
4.04622912e-01 9.34903562e-01 -7.90395617e-01 -5.12707196e-02
4.43388462e-01 -2.59536356e-01 1.21072799e-01 -7.62332678e-01
-1.17670882e+00 6.43067479e-01 9.67185616e-01 -2.96899416e-02
-1.15363173e-01 6.96836233e-01 2.58799672e-01 1.22542299e-01
5.23846090e-01 4.23243284e-01 -4.38643396e-01 1.30881704e-02
-9.96158540e-01 4.16944802e-01 7.51041830e-01 9.20428395e-01
4.66362923e-01 1.91022873e-01 -6.53873622e-01 4.46459293e-01
9.65152800e-01 8.99454951e-01 2.50154864e-02 -7.94555128e-01
3.41883928e-01 6.93870187e-01 2.42067650e-01 -7.29893684e-01
-5.68783522e-01 -9.92012680e-01 -1.98325694e-01 4.65712994e-01
5.64976990e-01 -1.76056728e-01 -9.09474850e-01 2.01586342e+00
7.45671093e-01 4.76610661e-01 -3.29347998e-01 1.07587838e+00
5.76616466e-01 1.20647959e-01 5.00148892e-01 1.70272104e-02
1.32521713e+00 -1.12946451e+00 -6.29856050e-01 -3.46374422e-01
6.18404031e-01 -5.68602026e-01 4.49589670e-01 3.17197978e-01
-7.75266111e-01 -6.90435946e-01 -9.71024752e-01 1.00270241e-01
-8.70002955e-02 5.93000174e-01 5.86971939e-01 7.03816414e-01
-9.03146684e-01 6.32014871e-01 -1.28607094e+00 -4.54545856e-01
6.60036922e-01 5.73686361e-01 -3.46927978e-02 4.69062477e-01
-9.34255838e-01 1.32042122e+00 4.92141396e-01 1.58414438e-01
-1.12634349e+00 -1.12121689e+00 -5.80899596e-01 -1.11869887e-01
4.80894804e-01 -8.27229321e-01 1.33007145e+00 -4.58871037e-01
-1.58102727e+00 5.36505461e-01 7.32909702e-03 -1.13408828e+00
7.42458403e-01 -5.14215291e-01 -4.77510065e-01 -1.40285850e-01
4.87524495e-02 8.54086995e-01 9.35171366e-01 -1.02193058e+00
-8.28909755e-01 -2.63951838e-01 1.92459358e-03 -1.96172625e-01
-2.56806403e-01 1.53204957e-02 -4.68412697e-01 -6.75417662e-01
-2.94592738e-01 -1.15269136e+00 -1.26325432e-02 4.50780571e-01
-1.57699898e-01 -3.94551158e-01 9.99461174e-01 -5.93182206e-01
1.17793918e+00 -1.77009857e+00 -4.63417498e-03 1.77045822e-01
4.82097954e-01 2.11797252e-01 3.63257714e-02 -5.36903180e-02
2.95803607e-01 -3.81763756e-01 2.53684044e-01 -4.35500473e-01
4.44564134e-01 1.68629512e-01 -5.00305772e-01 4.80672538e-01
3.69483948e-01 1.38308501e+00 -8.65755439e-01 -6.65662408e-01
3.49600315e-01 3.29664230e-01 -7.41176844e-01 2.36794204e-01
-6.86945498e-01 4.18916672e-01 -5.36088586e-01 8.30895483e-01
4.15093362e-01 -2.35434294e-01 2.42299989e-01 -4.92427230e-01
-8.33496526e-02 2.25111544e-01 -8.77116084e-01 1.60627306e+00
-6.86023608e-02 5.87495267e-01 -3.23286593e-01 -6.20889068e-01
1.16761696e+00 1.92560535e-02 7.66164482e-01 -6.34833455e-01
2.95063347e-01 -8.14731270e-02 1.47525325e-01 1.23761203e-02
2.61995912e-01 -1.19466253e-01 -1.42104238e-01 4.70387004e-03
4.01308447e-01 3.11823100e-01 7.89615288e-02 1.19150609e-01
1.13160491e+00 8.17893147e-01 5.29972324e-03 -4.27892476e-01
4.88247454e-01 2.88548827e-01 8.87814164e-01 7.48019755e-01
-4.27648842e-01 7.18529075e-02 2.23732352e-01 -5.82667053e-01
-8.05760384e-01 -1.19303346e+00 -1.28906488e-01 1.15572608e+00
3.02582920e-01 -6.51417375e-01 -5.72276533e-01 -9.72291291e-01
3.26556623e-01 5.95367074e-01 -8.53854120e-01 -3.98080498e-01
-5.90601385e-01 -3.20374131e-01 8.04387391e-01 9.59279060e-01
2.36620083e-01 -7.11293519e-01 -7.41201937e-01 3.42993766e-01
4.11066459e-03 -1.41080725e+00 -4.22498375e-01 4.10590321e-02
-1.04469168e+00 -1.10364854e+00 -9.02884677e-02 -1.35662600e-01
3.51475686e-01 -3.24206948e-01 1.15848494e+00 -1.94005761e-02
-1.30370915e-01 4.50244606e-01 -6.15831576e-02 -5.81896305e-01
-1.83053002e-01 2.49274850e-01 3.73900086e-01 -6.72746226e-02
4.10352111e-01 -2.50414431e-01 -3.96778375e-01 6.33427083e-01
-1.89380541e-01 -1.65280104e-01 5.68362713e-01 7.60601044e-01
5.55585563e-01 -6.64820015e-01 3.94961625e-01 -4.40191478e-01
2.30988059e-02 -1.57604203e-01 -1.28978837e+00 3.75489801e-01
-8.26561034e-01 3.49483848e-01 4.39623415e-01 -6.90978467e-01
-1.13577700e+00 3.16787511e-01 2.44908899e-01 -9.38436449e-01
3.96766692e-01 1.98022202e-01 -2.87050717e-02 -5.40271103e-01
8.62335563e-01 -2.72622138e-01 -1.30286086e-02 -5.71647882e-01
4.93133128e-01 -1.82606831e-01 7.30989754e-01 -8.34916711e-01
1.41190827e+00 4.19100434e-01 2.40615159e-01 -2.09533483e-01
-1.27739954e+00 -3.09629977e-01 -6.27278626e-01 -7.25029230e-01
7.25004733e-01 -1.06264770e+00 -1.13735509e+00 -8.28077793e-02
-7.61015475e-01 -3.46872628e-01 -4.54358578e-01 6.66426420e-01
-4.86882031e-01 2.35178210e-02 -2.24843845e-01 -8.19615185e-01
-3.96692991e-01 -8.13497901e-01 1.22752368e+00 2.82872319e-01
-2.04832897e-01 -1.01228821e+00 4.30634171e-01 2.23371714e-01
4.88155484e-01 3.91406178e-01 -8.28878433e-02 -6.43217802e-01
-1.02536428e+00 -2.11460769e-01 -1.44531548e-01 -3.41718383e-02
-3.78868103e-01 -9.48716477e-02 -1.02917957e+00 -4.54946488e-01
-4.26598608e-01 -3.35250467e-01 9.64950562e-01 3.46257091e-01
6.59425557e-01 -1.09951422e-01 -7.98023283e-01 6.84254587e-01
1.22538459e+00 -3.66842151e-01 3.47625643e-01 3.25761914e-01
6.55943811e-01 1.09777406e-01 1.07400525e+00 2.63677418e-01
3.92808944e-01 1.30171859e+00 6.15852296e-01 2.57149756e-01
-4.67638403e-01 -5.58474362e-01 6.97689593e-01 2.15055153e-01
-1.67695031e-01 2.31174707e-01 -8.91899288e-01 3.12337250e-01
-2.16220188e+00 -1.05063438e+00 5.93510224e-03 2.06684732e+00
7.34269023e-01 4.88209397e-01 4.06472772e-01 -4.83470917e-01
4.45712537e-01 -7.84102827e-03 -6.65092409e-01 2.32955247e-01
2.52337486e-01 2.73511205e-02 7.77795851e-01 4.38592583e-01
-1.36900282e+00 1.34394634e+00 6.47575140e+00 7.57852793e-01
-1.14668810e+00 2.17350006e-01 -8.67203251e-02 -4.30917859e-01
1.28433809e-01 2.57404029e-01 -1.41306961e+00 4.04125035e-01
1.02434731e+00 -1.42278180e-01 2.13689968e-01 1.14715517e+00
3.48251648e-02 1.47453547e-01 -1.30330002e+00 6.21454358e-01
-2.63550073e-01 -1.33142424e+00 -3.31269950e-01 9.52888280e-02
5.13385892e-01 4.58735794e-01 -1.28125288e-02 7.66087592e-01
6.05098665e-01 -7.38392293e-01 1.10547924e+00 9.32072878e-01
3.79035681e-01 -1.64237514e-01 4.13293302e-01 2.57072002e-01
-1.48402274e+00 -2.16374710e-01 -4.17021781e-01 1.42794356e-01
1.81740820e-01 4.81450319e-01 -9.72995818e-01 5.89489579e-01
7.67412841e-01 9.38091338e-01 -9.00688529e-01 1.45293796e+00
-3.22486758e-01 7.17216492e-01 -8.28537345e-01 2.31717210e-02
2.89750770e-02 2.09378511e-01 8.58815908e-01 1.05236936e+00
7.77265429e-02 -3.33968610e-01 3.89326990e-01 1.07686436e+00
-9.89736691e-02 -3.40885341e-01 -3.80526066e-01 1.95571184e-01
5.51764846e-01 1.52745295e+00 -4.65343297e-01 -1.68997228e-01
-1.95576310e-01 1.68508217e-02 5.50715685e-01 3.88453132e-03
-1.51378345e+00 3.66245180e-01 7.03669548e-01 -1.02597186e-02
5.67544699e-01 -1.21963911e-01 -3.09595138e-01 -1.32755840e+00
7.27949962e-02 -6.99695408e-01 4.75515991e-01 -4.32143748e-01
-1.20241976e+00 3.78167719e-01 8.73755142e-02 -1.69935155e+00
-3.32635790e-01 -6.55796409e-01 -4.30752635e-01 4.62923765e-01
-1.52197218e+00 -1.80111051e+00 -3.80064100e-01 3.71579111e-01
8.19394141e-02 -2.08822355e-01 3.14863473e-01 3.70837331e-01
-4.48146433e-01 8.90808463e-01 -4.10244375e-01 2.46637091e-01
8.79745722e-01 -1.21760952e+00 2.44382203e-01 8.87094855e-01
2.63320088e-01 6.71195805e-01 8.32950413e-01 -8.66045952e-01
-1.56195390e+00 -1.37688899e+00 6.74073994e-01 -1.11669230e+00
9.83985901e-01 -3.43235523e-01 -4.66906279e-01 9.62221861e-01
-1.38480335e-01 5.00647426e-01 2.83292353e-01 4.93612319e-01
-5.70196867e-01 -4.77960467e-01 -9.14770246e-01 2.72312552e-01
1.26804757e+00 -1.16537303e-01 -6.79192841e-01 1.39703244e-01
5.48916757e-01 -7.57088244e-01 -1.27745128e+00 7.39355326e-01
9.12655294e-01 -6.70543492e-01 1.13717294e+00 -6.25107765e-01
-1.64852723e-01 -7.75746644e-01 1.25036225e-01 -9.49825883e-01
-4.29372817e-01 -4.60292220e-01 -9.70683455e-01 1.20719218e+00
4.41623867e-01 -5.49624026e-01 9.03192163e-01 6.07784212e-01
-1.05603538e-01 -7.60792792e-01 -1.09326649e+00 -1.18104351e+00
-2.67505735e-01 -5.20527840e-01 4.46549386e-01 5.46427727e-01
-2.17744693e-01 3.65541488e-01 -6.28194392e-01 2.10271969e-01
1.32219064e+00 1.70726836e-01 1.11418927e+00 -1.63507688e+00
-3.87612492e-01 -3.78446192e-01 -6.83854461e-01 -1.23642898e+00
1.96945041e-01 -9.36646342e-01 7.50391111e-02 -9.76068974e-01
-1.89187322e-02 -5.95059335e-01 -3.95549268e-01 6.94581151e-01
-1.01919733e-02 3.90719146e-01 5.68550646e-01 3.49794537e-01
-1.21071732e+00 7.34152377e-01 1.05785120e+00 -2.02490240e-01
2.32992619e-02 1.63342893e-01 -4.16725844e-01 6.24923468e-01
5.72338283e-01 -7.37952769e-01 -3.59299220e-02 -2.72508979e-01
5.68537861e-02 -2.84123451e-01 8.04027379e-01 -1.42441130e+00
5.63396811e-01 1.24117598e-01 8.39852571e-01 -7.57097781e-01
4.12535638e-01 -1.12990689e+00 2.45341033e-01 6.31943762e-01
-3.43696773e-01 -4.57989067e-01 3.98761332e-01 8.30233693e-01
7.44601861e-02 3.69181305e-01 7.75253296e-01 6.14726782e-01
-8.93876731e-01 4.77846920e-01 3.40493947e-01 -4.38512675e-03
1.12301576e+00 -1.57924190e-01 -3.47312808e-01 -1.86696816e-02
-8.78255785e-01 6.31191850e-01 3.36851180e-01 7.89557219e-01
2.03564107e-01 -1.69365740e+00 -4.38906908e-01 7.60040013e-03
2.35959455e-01 -5.89595437e-01 -2.80880958e-01 1.14255083e+00
3.30289565e-02 6.24592483e-01 -2.07117021e-01 -1.15798724e+00
-1.07467568e+00 5.64223588e-01 6.88561738e-01 -5.53846538e-01
-7.26567924e-01 4.01182592e-01 -2.13427916e-02 -4.77709323e-01
3.79059523e-01 -6.18539512e-01 -1.74630865e-01 -1.75500005e-01
3.34096998e-01 1.99165702e-01 -2.15505391e-01 -5.70460916e-01
-6.60426140e-01 7.32896686e-01 1.80939436e-01 1.08486652e-01
1.05156124e+00 1.92146510e-01 2.74963498e-01 1.00909859e-01
5.46180308e-01 -1.20851420e-01 -1.76941955e+00 -4.32917535e-01
5.24576128e-01 -4.90268081e-01 1.48815021e-01 -1.11224771e+00
-9.92509305e-01 4.35612053e-01 8.59612525e-01 -3.14656608e-02
8.87378931e-01 1.73104748e-01 5.67248762e-01 4.24446613e-01
6.38381362e-01 -1.07055759e+00 -8.75652395e-03 6.60227954e-01
6.50186598e-01 -1.25267160e+00 2.33505458e-01 -3.29712421e-01
-4.73332226e-01 9.50156331e-01 8.87207985e-01 -3.43987405e-01
5.87347925e-01 4.30157512e-01 -5.71534084e-03 -2.40241244e-01
-1.04095268e+00 -4.47400510e-01 8.27608943e-01 6.51467144e-01
2.13090599e-01 -9.84448940e-02 -1.92137450e-01 6.44128919e-01
-1.65866017e-01 3.97908241e-01 -5.77959061e-01 4.86175805e-01
-5.60316205e-01 -1.20759416e+00 -4.40548480e-01 2.93134779e-01
-2.96686292e-01 3.05928141e-01 -3.60337615e-01 8.31832826e-01
5.99413328e-02 7.11630940e-01 -2.71060109e-01 -4.90543276e-01
3.82498860e-01 -2.39632074e-02 7.04839349e-01 -2.18751609e-01
-8.04307044e-01 -2.67133024e-02 2.88349181e-01 -1.05291641e+00
-4.77529645e-01 -7.64647245e-01 -1.03398407e+00 -1.16086051e-01
-5.97050786e-01 9.94844246e-04 4.91437823e-01 1.00859964e+00
3.66910756e-01 5.96266210e-01 2.65023470e-01 -8.61490786e-01
-9.62375581e-01 -9.92224395e-01 2.22445466e-02 2.33722478e-01
2.35295996e-01 -1.32143438e+00 1.41899377e-01 5.11762723e-02]
|
[6.290066242218018, -2.1356513500213623]
|
38a16368-e304-4d31-8c5c-a4619344b1f0
|
text-sketch-image-compression-at-ultra-low
|
2307.01944
| null |
https://arxiv.org/abs/2307.01944v1
|
https://arxiv.org/pdf/2307.01944v1.pdf
|
Text + Sketch: Image Compression at Ultra Low Rates
|
Recent advances in text-to-image generative models provide the ability to generate high-quality images from short text descriptions. These foundation models, when pre-trained on billion-scale datasets, are effective for various downstream tasks with little or no further training. A natural question to ask is how such models may be adapted for image compression. We investigate several techniques in which the pre-trained models can be directly used to implement compression schemes targeting novel low rate regimes. We show how text descriptions can be used in conjunction with side information to generate high-fidelity reconstructions that preserve both semantics and spatial structure of the original. We demonstrate that at very low bit-rates, our method can significantly improve upon learned compressors in terms of perceptual and semantic fidelity, despite no end-to-end training.
|
['Shirin Saeedi Bidokhti', 'Hamed Hassani', 'Yiğit Berkay Uslu', 'Eric Lei']
|
2023-07-04
| null | null | null | null |
['image-compression']
|
['computer-vision']
|
[ 8.22096050e-01 1.77231267e-01 -1.81009054e-01 -3.51881891e-01
-9.94053483e-01 -3.65745604e-01 8.06549072e-01 -7.40038678e-02
-1.05240680e-01 7.41449058e-01 5.69806397e-01 -2.06998616e-01
1.47564232e-01 -8.37114334e-01 -9.84575391e-01 -6.27708077e-01
7.78216273e-02 4.31363106e-01 5.43829910e-02 -2.45960429e-01
2.02802330e-01 3.36564898e-01 -1.48956895e+00 6.74882054e-01
6.53520942e-01 8.37399483e-01 6.57011867e-01 1.02927125e+00
8.85363817e-02 8.74482334e-01 -4.01735872e-01 -5.64512074e-01
2.88377136e-01 -8.01249862e-01 -6.28731370e-01 2.31205031e-01
3.82665634e-01 -8.47900748e-01 -8.39576542e-01 8.15422893e-01
5.38524151e-01 -7.62718990e-02 7.43675888e-01 -4.81855482e-01
-6.92440867e-01 5.17023861e-01 -2.63048232e-01 2.53871344e-02
2.77306050e-01 3.70909005e-01 9.56497312e-01 -8.92182946e-01
7.70273030e-01 1.27548766e+00 6.17532074e-01 5.65627873e-01
-1.62861061e+00 -3.15267801e-01 -5.35585165e-01 1.55952692e-01
-1.11862564e+00 -9.16593909e-01 4.89556253e-01 6.64527640e-02
9.25247312e-01 1.76748008e-01 5.15351117e-01 1.03297961e+00
3.31297219e-01 5.42350054e-01 9.73945260e-01 -5.00446975e-01
8.23364407e-02 -1.94039661e-02 -7.70480335e-01 6.42999172e-01
2.67143905e-01 2.67501533e-01 -8.13439965e-01 4.16457467e-02
9.04058337e-01 -1.76454335e-01 -4.02809054e-01 -1.11861348e-01
-1.12942314e+00 9.40326571e-01 4.72313821e-01 1.36552423e-01
-3.16129655e-01 6.97968841e-01 1.60261199e-01 2.48560593e-01
4.37637240e-01 4.75695074e-01 -1.90703664e-02 -2.09124893e-01
-1.33321726e+00 3.48263532e-01 6.10760868e-01 1.16494048e+00
6.57409310e-01 3.18498850e-01 -1.80839136e-01 8.19324493e-01
2.97330320e-02 6.44798934e-01 4.30213630e-01 -1.17615914e+00
4.40099239e-01 -2.52814472e-01 8.22945498e-03 -7.93417454e-01
2.65685380e-01 -5.81507862e-01 -9.15101230e-01 3.57300267e-02
-1.38892576e-01 6.26124367e-02 -1.19529343e+00 1.50706029e+00
-1.52996093e-01 8.05535689e-02 1.05961800e-01 6.25503838e-01
3.83176178e-01 1.02629256e+00 -2.62210164e-02 -2.04363540e-01
1.13945520e+00 -8.31363201e-01 -7.47699201e-01 -3.84454429e-01
4.45998996e-01 -9.26584899e-01 9.88852799e-01 2.16116309e-01
-1.81208384e+00 -5.93110681e-01 -1.14580584e+00 -3.28339994e-01
5.68070784e-02 -2.64078051e-01 2.92474061e-01 5.26020646e-01
-1.45541131e+00 1.00582111e+00 -7.81440377e-01 -9.69277248e-02
6.35879695e-01 1.57855779e-01 -1.02211975e-01 -4.96692032e-01
-8.56957555e-01 7.03121543e-01 2.66019553e-01 -3.58995557e-01
-1.32768929e+00 -6.34795785e-01 -7.62768269e-01 2.51989901e-01
4.92902286e-02 -1.20796835e+00 1.22813213e+00 -6.87732577e-01
-1.47188354e+00 6.48482442e-01 -4.36352640e-01 -7.52713263e-01
5.08411348e-01 1.20495155e-01 -1.16603851e-01 7.28291988e-01
-1.37372557e-02 1.02451849e+00 1.18839097e+00 -1.48596883e+00
-4.31614578e-01 -3.24536674e-02 -2.35502362e-01 1.93178385e-01
-4.14716661e-01 -1.60111517e-01 -3.56166780e-01 -9.34309006e-01
6.78836554e-02 -8.14419746e-01 -2.36100644e-01 4.24606979e-01
-2.37403587e-01 4.51833338e-01 8.56485188e-01 -7.99147248e-01
8.12007308e-01 -1.86482847e+00 2.49248266e-01 -1.18458077e-01
1.10270835e-01 3.36142451e-01 -3.45652759e-01 7.34844089e-01
2.75240839e-01 4.00249153e-01 -4.89923090e-01 -7.23179996e-01
-1.03806876e-01 2.88820356e-01 -6.26049936e-01 1.71652332e-01
3.84949535e-01 1.22768939e+00 -7.30608642e-01 -3.91428381e-01
2.56756932e-01 7.80914664e-01 -9.61855531e-01 5.43902397e-01
-4.25341517e-01 5.57234406e-01 -1.36380836e-01 2.99866855e-01
4.93357092e-01 -4.03182358e-01 1.13056161e-01 -2.50998110e-01
4.54078674e-01 5.06415844e-01 -4.89140123e-01 1.96331239e+00
-6.94603801e-01 8.70473683e-01 5.71012571e-02 -1.04996586e+00
5.76310158e-01 2.71162450e-01 2.53611207e-01 -7.91410565e-01
-7.54304975e-02 2.56861687e-01 -2.83251971e-01 -4.66039568e-01
5.68247795e-01 -6.14392459e-01 1.23254895e-01 4.86636728e-01
2.95732945e-01 -5.56470096e-01 9.00557712e-02 3.59043032e-01
1.08245981e+00 9.43997130e-02 -4.57966700e-04 7.36086145e-02
2.60932148e-02 -5.64785488e-02 -3.70907299e-02 7.71561205e-01
4.01834965e-01 1.15090561e+00 1.98126987e-01 -1.23737022e-01
-1.88539577e+00 -1.15090311e+00 -3.29701573e-01 8.14268470e-01
1.38770547e-02 -5.76253235e-01 -6.40584290e-01 -9.73372012e-02
-3.18160474e-01 6.93588853e-01 -2.54681587e-01 -3.71436030e-01
-5.29663742e-01 -5.84184229e-01 7.68690825e-01 3.81938994e-01
5.82373619e-01 -8.68019104e-01 -6.00740790e-01 2.75429487e-01
-3.62981021e-01 -1.42733729e+00 -4.80602831e-01 2.43783712e-01
-1.23465967e+00 -3.62143666e-01 -1.06936812e+00 -7.48591304e-01
6.71487510e-01 4.29944605e-01 1.21919203e+00 2.96040237e-01
-2.16372013e-01 1.77321598e-01 -3.75770539e-01 -6.25044629e-02
-1.02346134e+00 -8.09764713e-02 -4.30937469e-01 -2.35265419e-01
-4.21877086e-01 -1.05111969e+00 -8.54050994e-01 1.04313627e-01
-1.29414582e+00 4.07324404e-01 7.40368009e-01 1.08030128e+00
5.97946286e-01 7.17363581e-02 4.76897806e-01 -8.05568814e-01
4.70676363e-01 -3.59695256e-01 -1.07148550e-01 -4.06066962e-02
-7.11983800e-01 5.26368201e-01 9.19513881e-01 -2.64978975e-01
-8.68382096e-01 -1.85381576e-01 -4.63607639e-01 -6.21609569e-01
1.55151282e-02 4.12699223e-01 2.44206917e-02 -1.70519516e-01
6.86707795e-01 8.28747272e-01 -2.77120955e-02 -3.93979788e-01
6.52611434e-01 5.08902669e-01 8.29098344e-01 -7.52235532e-01
8.37697387e-01 6.39605403e-01 3.09014171e-01 -7.05041528e-01
-6.55480564e-01 9.23720002e-02 -4.17570412e-01 2.39347607e-01
7.71712303e-01 -1.18483996e+00 -1.26048163e-01 1.49042800e-01
-1.01293206e+00 -4.28931862e-01 -5.44153690e-01 2.80675385e-02
-1.12320542e+00 4.38316524e-01 -8.66099715e-01 -5.60296714e-01
-3.72533619e-01 -1.00506949e+00 1.39009738e+00 -2.41913237e-02
1.56901553e-01 -8.76934230e-01 -2.52753198e-01 2.69887447e-01
7.51340866e-01 1.61821656e-02 1.09156537e+00 1.39735220e-02
-9.25960422e-01 -6.09332807e-02 -2.64280945e-01 4.60474819e-01
-8.63157883e-02 -4.86606896e-01 -9.81056094e-01 -5.74111819e-01
8.58406350e-02 -6.60536647e-01 1.10659027e+00 3.57526541e-01
1.31777692e+00 -7.74608731e-01 -7.31325299e-02 1.21238029e+00
1.67118812e+00 -1.07058667e-01 1.09308875e+00 -1.25850216e-01
5.74479759e-01 3.27838480e-01 -5.48955426e-02 4.38899755e-01
2.22274020e-01 7.29900360e-01 1.99189574e-01 3.50351073e-02
-7.75557876e-01 -8.79558384e-01 3.00676107e-01 7.55981505e-01
-2.39804238e-02 -4.81568903e-01 -5.30112803e-01 5.63730240e-01
-1.45868838e+00 -1.20814025e+00 3.43273103e-01 2.13933897e+00
1.01399171e+00 1.16457105e-01 -1.14573345e-01 2.74029020e-02
4.54794794e-01 3.79057288e-01 -5.63126445e-01 -1.95112273e-01
-3.18401188e-01 5.56787670e-01 5.70412040e-01 4.17815834e-01
-6.91881835e-01 7.23022342e-01 7.79485464e+00 9.48366940e-01
-1.05971408e+00 3.05176914e-01 9.84441638e-01 -1.52462080e-01
-7.10185289e-01 2.92343497e-01 -4.21506703e-01 5.12188673e-01
1.63743186e+00 -2.48392016e-01 7.61251807e-01 5.01215339e-01
2.81854689e-01 8.81966501e-02 -1.07515454e+00 9.91864383e-01
1.80316672e-01 -1.91828060e+00 5.77247739e-01 2.20256418e-01
1.06009448e+00 -5.28436080e-02 1.88360512e-01 -3.78356203e-02
2.15390161e-01 -1.34154487e+00 8.57713521e-01 2.94481814e-01
1.40263534e+00 -6.74529195e-01 3.41684580e-01 3.13994706e-01
-9.60082173e-01 -1.89083386e-02 -7.58012652e-01 2.30307937e-01
5.73539734e-01 4.58648115e-01 -7.45107830e-01 4.49355900e-01
3.42999041e-01 7.68929362e-01 -3.58574450e-01 8.40562046e-01
-1.77878812e-01 6.49641395e-01 -2.46547654e-01 3.48745257e-01
1.12083152e-01 7.19749108e-02 4.83275324e-01 1.23028505e+00
7.68597901e-01 1.06695652e-01 -1.65905684e-01 1.09901607e+00
-3.22131097e-01 -2.80525029e-01 -8.97789121e-01 -1.42924428e-01
3.02445889e-01 6.83160603e-01 -5.07513642e-01 -4.87683237e-01
-4.59989831e-02 1.39270246e+00 1.34663746e-01 5.51217854e-01
-7.14165807e-01 -1.92278653e-01 3.71784359e-01 6.32447660e-01
6.85181260e-01 -4.58365113e-01 -4.84403759e-01 -1.27417672e+00
-8.17239881e-02 -8.43320072e-01 -8.05642530e-02 -1.16404831e+00
-9.14906323e-01 5.79695106e-01 9.45997387e-02 -1.08525741e+00
-6.30535424e-01 -3.19905311e-01 -4.11167443e-01 7.63221622e-01
-1.65390086e+00 -1.25559604e+00 -1.29300192e-01 5.71632504e-01
7.26897418e-01 -1.13854788e-01 8.07480991e-01 1.57881618e-01
2.69860774e-02 6.58542931e-01 3.62642199e-01 -1.84960648e-01
4.03533936e-01 -9.05020535e-01 5.80144703e-01 9.68858540e-01
3.58239502e-01 3.48542869e-01 9.59206760e-01 -5.42363942e-01
-1.64180255e+00 -1.17898464e+00 6.91109657e-01 -2.08963320e-01
1.54107153e-01 -4.29404199e-01 -7.23485351e-01 6.51850820e-01
4.34472948e-01 1.86254859e-01 3.09595406e-01 -4.21912104e-01
-4.22924817e-01 7.97870755e-02 -1.28874969e+00 5.52088320e-01
1.25381291e+00 -7.63453603e-01 -2.14902163e-01 4.05161083e-01
1.04241157e+00 -3.18742096e-01 -8.11741650e-01 1.32622689e-01
3.00160766e-01 -1.09625161e+00 1.24746168e+00 -3.92502367e-01
1.19253778e+00 -6.81179529e-03 -4.99658287e-01 -1.29212534e+00
-2.73382276e-01 -9.67313766e-01 -2.94316173e-01 8.61713529e-01
3.06702614e-01 -2.94647276e-01 6.39196217e-01 2.72897273e-01
-1.92139342e-01 -6.32161260e-01 -9.40788150e-01 -7.45663762e-01
1.80590123e-01 -3.06862026e-01 4.84537840e-01 3.37251782e-01
-1.10239528e-01 4.36969668e-01 -8.06357741e-01 -2.24244207e-01
7.66569495e-01 1.57898635e-01 5.63988209e-01 -5.64174294e-01
-6.11733615e-01 -3.58918488e-01 -4.60219145e-01 -1.58915269e+00
-4.30899039e-02 -1.01093388e+00 2.54967123e-01 -1.41348302e+00
4.11168009e-01 -4.18557346e-01 1.25008270e-01 1.18263289e-01
-1.03437435e-02 6.52305603e-01 4.56322819e-01 4.99521017e-01
-2.85297126e-01 9.75195885e-01 1.43093097e+00 -1.52042761e-01
2.89457142e-01 -3.27662408e-01 -8.22220206e-01 1.89853817e-01
5.65533280e-01 -5.52000165e-01 -5.86395621e-01 -7.49430180e-01
1.79137662e-01 2.97173858e-01 4.04953480e-01 -1.30089664e+00
1.12160541e-01 -1.26284227e-01 4.75500017e-01 -2.02203423e-01
5.91024935e-01 -6.17578506e-01 2.55906492e-01 3.45522106e-01
-6.85583293e-01 2.97945384e-02 1.14333905e-01 7.92245686e-01
-2.26182178e-01 -2.65489817e-01 1.14171851e+00 -3.16581488e-01
-9.85565260e-02 2.75822788e-01 -2.53280640e-01 -2.75131166e-02
6.83361888e-01 -1.39875218e-01 -2.92712063e-01 -1.10001910e+00
-4.16078746e-01 -3.46260637e-01 7.96818614e-01 6.25364408e-02
9.62258160e-01 -1.42189598e+00 -1.06074524e+00 2.87271887e-01
-1.85811535e-01 5.46639133e-03 1.26501471e-01 3.11580658e-01
-7.88772881e-01 6.46238565e-01 -2.34550759e-01 -6.35842681e-01
-7.15115190e-01 6.54359102e-01 2.65941530e-01 -3.36069286e-01
-9.56087589e-01 5.39933383e-01 2.75876641e-01 2.07460731e-01
-7.68236220e-02 -1.36030316e-01 5.78575134e-01 -6.16372824e-01
6.93589091e-01 -1.98526248e-01 -5.13890013e-02 -5.27022064e-01
2.05476910e-01 5.56456089e-01 1.06527768e-01 -5.27443588e-01
1.57379651e+00 -4.06823546e-01 2.79737622e-01 3.16919871e-02
1.38110125e+00 -2.33477622e-01 -1.53684783e+00 -1.20237753e-01
-5.36890209e-01 -8.77508342e-01 2.04232886e-01 -6.20638371e-01
-1.15629375e+00 1.17989826e+00 4.41907883e-01 2.71583796e-02
1.31106782e+00 6.57733828e-02 1.28807747e+00 1.83924809e-01
5.67794919e-01 -7.22143352e-01 4.01276410e-01 1.28592536e-01
1.01422560e+00 -9.92007256e-01 1.44589990e-01 -2.92144626e-01
-4.84424829e-01 1.07573652e+00 -7.69157186e-02 -3.33143145e-01
4.70003039e-01 4.66874599e-01 -4.84590054e-01 1.05016157e-02
-1.17821884e+00 -2.63168868e-02 1.16340473e-01 8.75086010e-01
3.39481771e-01 -1.39266878e-01 -1.61114603e-01 -8.28433409e-02
-3.71876180e-01 7.90722817e-02 7.08286881e-01 8.15486908e-01
-5.82823932e-01 -1.15776122e+00 -2.21212298e-01 6.35759950e-01
-5.26561916e-01 -3.82951081e-01 1.91097736e-01 1.69687241e-01
-1.02496266e-01 9.32004571e-01 1.07335299e-02 -4.69368547e-01
-1.34022176e-01 -8.30311924e-02 7.41503477e-01 -5.44631839e-01
-1.32564619e-01 2.70918071e-01 1.56370685e-01 -5.66175818e-01
-2.73181021e-01 -4.87820297e-01 -9.29991543e-01 -6.90764129e-01
5.30699920e-03 -2.57542640e-01 6.29725099e-01 8.04788709e-01
6.60637736e-01 3.71730268e-01 6.22715116e-01 -1.11413407e+00
-6.51345015e-01 -7.55523980e-01 -3.16231698e-01 6.21537030e-01
5.89761078e-01 -6.80079907e-02 -2.71923125e-01 5.99609971e-01]
|
[11.358951568603516, -0.8296070098876953]
|
9e8142f5-a85a-4f85-b99b-6febcfa4dcbb
|
faster-maximum-inner-product-search-in-high
|
2212.07551
| null |
https://arxiv.org/abs/2212.07551v3
|
https://arxiv.org/pdf/2212.07551v3.pdf
|
Faster Maximum Inner Product Search in High Dimensions
|
Maximum Inner Product Search (MIPS) is a ubiquitous task in machine learning applications such as recommendation systems. Given a query vector and $n$ atom vectors in $d$-dimensional space, the goal of MIPS is to find the atom that has the highest inner product with the query vector. Existing MIPS algorithms scale at least as $O(\sqrt{d})$, which becomes computationally prohibitive in high-dimensional settings. In this work, we present BanditMIPS, a novel randomized MIPS algorithm whose complexity is independent of $d$. BanditMIPS estimates the inner product for each atom by subsampling coordinates and adaptively evaluates more coordinates for more promising atoms. The specific adaptive sampling strategy is motivated by multi-armed bandits. We provide theoretical guarantees that BanditMIPS returns the correct answer with high probability, while improving the complexity in $d$ from $O(\sqrt{d})$ to $O(1)$. We also perform experiments on four synthetic and real-world datasets and demonstrate that BanditMIPS outperforms prior state-of-the-art algorithms. For example, in the Movie Lens dataset ($n$=4,000, $d$=6,000), BanditMIPS is 20$\times$ faster than the next best algorithm while returning the same answer. BanditMIPS requires no preprocessing of the data and includes a hyperparameter that practitioners may use to trade off accuracy and runtime. We also propose a variant of our algorithm, named BanditMIPS-$\alpha$, which achieves further speedups by employing non-uniform sampling across coordinates. Finally, we demonstrate how known preprocessing techniques can be used to further accelerate BanditMIPS, and discuss applications to Matching Pursuit and Fourier analysis.
|
['DongHyun Lee', 'Martin Jinye Zhang', 'Ilan Shomorony', 'Sebastian Thrun', 'Chris Piech', 'Je-Yong Lee', 'Ryan Kang', 'Mo Tiwari']
|
2022-12-14
| null | null | null | null |
['multi-armed-bandits']
|
['miscellaneous']
|
[-7.67357722e-02 -2.87560940e-01 -5.73877156e-01 -6.20552786e-02
-1.33499479e+00 -7.11414874e-01 -1.21908233e-01 1.49374545e-01
-4.74333078e-01 6.56457961e-01 -1.27455547e-01 -5.14240146e-01
-5.24925411e-01 -9.69696701e-01 -1.13248670e+00 -6.51372313e-01
-5.84142923e-01 8.02723169e-01 9.99601185e-02 -6.40552416e-02
5.97310483e-01 3.17015141e-01 -1.31849861e+00 -4.74231653e-02
7.93852925e-01 1.44366813e+00 -1.89098995e-02 5.69316626e-01
-1.63508505e-01 5.66005409e-02 -3.44802856e-01 -6.16774440e-01
8.22809756e-01 -2.65742660e-01 -7.22994268e-01 1.30107133e-02
5.70890129e-01 -3.54483783e-01 -1.42590389e-01 1.24714971e+00
3.09920937e-01 3.84040177e-01 4.07331198e-01 -8.58691514e-01
-5.74252009e-01 5.57180703e-01 -1.24306536e+00 2.53280312e-01
3.58210623e-01 -1.34205163e-01 1.61968100e+00 -1.05010462e+00
2.30306864e-01 1.13260913e+00 5.69815457e-01 -5.44642620e-02
-1.60376108e+00 -1.02341974e+00 2.13163957e-01 1.77476913e-01
-1.63155448e+00 -2.22744584e-01 7.16056228e-01 -1.79393440e-01
6.41365767e-01 5.36824405e-01 7.00120568e-01 4.36906442e-02
-5.40342152e-01 9.71714258e-01 7.89421260e-01 -3.75346750e-01
3.74506027e-01 -6.80086687e-02 4.21580642e-01 8.28951061e-01
5.14658570e-01 4.03530449e-02 -8.74241531e-01 -8.09919775e-01
8.49795878e-01 9.87464711e-02 -3.31863195e-01 -5.98510563e-01
-1.09638774e+00 1.23886025e+00 2.95465410e-01 -1.00641437e-01
-4.59598511e-01 4.68706608e-01 1.09925261e-02 2.75963336e-01
5.38012803e-01 5.32072902e-01 -5.63201606e-01 -2.55985707e-01
-1.13866556e+00 6.98821664e-01 6.63129210e-01 1.04208148e+00
1.03920197e+00 -4.35703188e-01 -5.01017608e-02 1.01685476e+00
1.42614350e-01 9.20100868e-01 1.86467730e-02 -1.16173708e+00
8.08578253e-01 3.28857780e-01 6.63095415e-01 -1.21163154e+00
-1.43721074e-01 -5.61845601e-01 -6.42697513e-01 -1.56606019e-01
4.89115775e-01 -1.08966529e-01 -4.89953488e-01 1.58731341e+00
5.80188632e-01 3.24948579e-01 -4.39503640e-01 1.02068949e+00
1.94158271e-01 8.56186986e-01 -5.77453256e-01 -4.87866372e-01
1.37574375e+00 -9.24105823e-01 -2.81683773e-01 -2.33534619e-01
6.38909936e-01 -8.90831292e-01 1.13440883e+00 6.63504004e-01
-1.53987014e+00 -6.96756765e-02 -1.11847949e+00 3.40811223e-01
6.46235049e-02 -3.17569822e-02 1.00009894e+00 8.20785165e-01
-7.34685302e-01 3.97362620e-01 -6.55817747e-01 3.59896600e-01
4.45707440e-01 6.10313594e-01 3.95369669e-03 -2.06597701e-01
-8.38521600e-01 1.96269229e-01 -3.58035900e-02 -2.39605814e-01
-4.08513486e-01 -1.01903808e+00 -5.79930305e-01 1.05089262e-01
6.91612244e-01 -5.29718339e-01 1.19003582e+00 -4.11563575e-01
-1.10711789e+00 4.63542134e-01 -4.92681623e-01 -7.90406406e-01
1.16071343e-01 -2.16799766e-01 2.36198585e-02 3.38289499e-01
2.71391630e-01 2.50130802e-01 9.58886862e-01 -7.53812313e-01
-8.15716743e-01 -5.40236652e-01 1.21411502e-01 5.47282249e-02
-4.43120420e-01 -1.51690841e-01 -8.38367522e-01 -6.49464905e-01
5.66363454e-01 -1.17893481e+00 -4.82644975e-01 -1.39122233e-01
-3.18918854e-01 -2.57192880e-01 7.25447461e-02 -1.28791690e-01
1.55835342e+00 -2.05476809e+00 4.23529074e-02 8.04698706e-01
1.59850255e-01 4.59431298e-02 -6.41308427e-02 2.82418936e-01
-2.64469860e-03 1.37714725e-02 -2.01562658e-01 -2.82133400e-01
1.84786126e-01 -6.73234314e-02 -4.54917908e-01 8.31160009e-01
-5.13123810e-01 4.23876613e-01 -6.54480636e-01 -2.75029063e-01
-2.18751281e-01 -4.25399952e-02 -1.22913766e+00 -2.03618601e-01
-3.16455454e-01 -2.28313446e-01 -3.88942629e-01 6.32390261e-01
1.04131639e+00 -7.76389778e-01 4.33102734e-02 -2.75849015e-01
-3.57830785e-02 2.18947127e-01 -1.73983943e+00 1.62662864e+00
-3.30188036e-01 3.26071113e-01 1.72265381e-01 -1.42152011e+00
7.45798826e-01 -1.24481693e-01 9.16285694e-01 -6.57579720e-01
-1.99204504e-01 5.00460804e-01 -3.52999419e-01 -1.32365540e-01
5.71510911e-01 -1.48382038e-01 -1.05781868e-01 7.92441964e-01
-5.14161229e-01 -1.88530609e-02 4.83406693e-01 2.44672149e-01
1.14263105e+00 -5.49501896e-01 8.84634182e-02 -2.90047646e-01
2.90174693e-01 1.66428655e-01 4.81192172e-01 1.07470608e+00
6.25952333e-02 4.45036590e-01 3.93861979e-01 -5.41976213e-01
-7.68619001e-01 -1.01574433e+00 -2.57433861e-01 1.34707332e+00
4.51299787e-01 -4.44649845e-01 -6.37678921e-01 -4.96220767e-01
4.85924006e-01 3.86408329e-01 -5.33880055e-01 3.26315939e-01
-6.01221442e-01 -1.06578898e+00 2.20265463e-02 2.24212438e-01
4.39673513e-01 -3.65689248e-01 -4.72268760e-01 2.85712600e-01
-2.02538610e-01 -7.73526967e-01 -1.05300164e+00 -6.65867999e-02
-8.70227635e-01 -1.19485998e+00 -9.93999422e-01 -4.95582938e-01
6.79880083e-01 7.54424572e-01 1.09400845e+00 -8.80130902e-02
-1.89338431e-01 3.13007385e-01 -3.41167152e-01 -2.66321003e-01
4.34806615e-01 5.16858660e-02 5.59811778e-02 2.36740103e-03
4.97077584e-01 -5.54363251e-01 -1.14709258e+00 3.69631141e-01
-7.59385884e-01 -2.13604197e-01 4.96046901e-01 9.65105474e-01
1.08540845e+00 6.59139380e-02 2.62205243e-01 -9.01968956e-01
6.45221531e-01 -5.62988937e-01 -1.17121100e+00 7.82857239e-02
-7.59640276e-01 2.28878453e-01 4.04170543e-01 -5.02705097e-01
-2.78792024e-01 1.36818681e-02 1.91855561e-02 -5.01464248e-01
6.70936227e-01 7.37712324e-01 5.16691625e-01 -2.58592874e-01
7.67252803e-01 3.14150751e-01 -1.04038425e-01 -6.13761187e-01
3.04845601e-01 4.01185811e-01 3.90969098e-01 -9.28105354e-01
6.86222076e-01 7.12567031e-01 1.06752940e-01 -6.45487070e-01
-9.85580027e-01 -7.48965859e-01 2.31747925e-01 4.16030645e-01
8.95740837e-02 -8.12203705e-01 -1.36833656e+00 -7.98087120e-02
-8.15861940e-01 -1.15062959e-01 -1.52374789e-01 6.31624937e-01
-6.53965056e-01 3.52334708e-01 -3.27237606e-01 -9.95571017e-01
-5.47757566e-01 -1.09262311e+00 8.95697355e-01 -1.07320631e-02
-4.36214507e-02 -3.22010487e-01 8.03120509e-02 5.39406836e-01
1.88005939e-01 -1.73434153e-01 8.57906520e-01 -4.49134409e-01
-9.61542726e-01 -2.71133393e-01 -3.88893157e-01 -2.08225823e-03
-1.36793628e-01 -5.52761257e-01 -1.97307974e-01 -4.05796528e-01
-1.46745414e-01 -5.43915816e-02 8.03938150e-01 7.47957051e-01
1.31263828e+00 -6.74336791e-01 -2.68480510e-01 7.61841536e-01
1.32825172e+00 2.72297353e-01 3.32768172e-01 3.38769972e-01
9.14222226e-02 1.88769385e-01 1.04506421e+00 9.18506026e-01
1.22172415e-01 7.48002768e-01 3.60570222e-01 6.10794015e-02
4.11478609e-01 -5.41850179e-02 -4.26031649e-02 3.13467383e-01
1.24368221e-01 1.45943314e-01 -6.74284160e-01 7.07684577e-01
-1.97111189e+00 -8.47704887e-01 1.28781214e-01 2.51628375e+00
9.75371659e-01 2.43457198e-01 4.84968603e-01 1.17903031e-01
5.46935618e-01 3.75989303e-02 -9.80816305e-01 -1.98467612e-01
-3.40696401e-03 7.32300222e-01 1.05679643e+00 5.85024297e-01
-1.13590777e+00 6.83677375e-01 6.16216564e+00 1.15186107e+00
-7.77308941e-01 1.30730192e-03 7.71095574e-01 -7.91602790e-01
-3.16794038e-01 -1.19749039e-01 -1.00809658e+00 6.07055485e-01
6.97805882e-01 -3.28560680e-01 9.09159899e-01 1.12075388e+00
6.62402734e-02 -2.49543786e-01 -9.69150305e-01 1.63299358e+00
-3.11288759e-02 -1.95504022e+00 -2.32446432e-01 3.37133378e-01
8.64355564e-01 1.22588120e-01 4.55644608e-01 9.70688928e-03
3.93656462e-01 -7.80718982e-01 4.82078910e-01 -4.25484329e-02
6.81955397e-01 -1.06057799e+00 1.56687230e-01 3.51596355e-01
-1.30925202e+00 -1.80327341e-01 -5.36044478e-01 1.33334383e-01
1.56219631e-01 7.57637978e-01 -6.84824824e-01 8.45537484e-02
9.11202073e-01 2.54526496e-01 1.08386025e-01 1.15697849e+00
3.57600272e-01 5.97447634e-01 -9.02129471e-01 -4.32831109e-01
3.98685634e-01 -5.33452213e-01 4.44204628e-01 1.02478147e+00
6.24935150e-01 4.42005754e-01 3.05604756e-01 5.40495992e-01
-1.90705091e-01 4.77936208e-01 -1.30285710e-01 -9.25254915e-03
8.65896344e-01 6.52653635e-01 -5.23100317e-01 -4.74702030e-01
-4.10334319e-01 7.50194669e-01 2.19238907e-01 3.49535793e-01
-9.27217364e-01 -5.78427017e-01 1.04585969e+00 3.31025004e-01
9.14769769e-01 -2.59409487e-01 -4.39466983e-01 -1.00773871e+00
7.37013444e-02 -9.99975681e-01 6.18469417e-01 -2.86509931e-01
-1.12973058e+00 2.32443482e-01 6.19726032e-02 -1.09322333e+00
-1.94216445e-01 -4.96622473e-01 1.44553870e-01 7.93478310e-01
-1.21717501e+00 -4.13326293e-01 2.71688581e-01 6.72966957e-01
4.17194605e-01 7.03635672e-03 6.14486098e-01 5.08549809e-01
-3.91337305e-01 8.99113774e-01 3.80628258e-01 -1.45284504e-01
4.82008636e-01 -9.95520949e-01 2.94634253e-01 5.89972496e-01
4.55684185e-01 9.19034898e-01 1.03084123e+00 -2.87211388e-01
-1.75332928e+00 -6.46517992e-01 4.06401485e-01 2.10456148e-01
7.22369075e-01 -9.37655009e-03 -4.93426472e-01 4.10560697e-01
-3.69968891e-01 2.41110802e-01 9.98102486e-01 4.54819381e-01
-6.22440875e-01 -5.26237011e-01 -1.26307869e+00 6.87422752e-01
1.19062388e+00 -3.33038837e-01 -1.07113317e-01 6.66144669e-01
7.33582854e-01 -5.19372642e-01 -9.04491365e-01 3.62770617e-01
6.98427498e-01 -1.06331706e+00 1.46483719e+00 -4.69884515e-01
-1.58773698e-02 -1.61487505e-01 -5.77557385e-01 -8.72916639e-01
-3.67599487e-01 -1.01788366e+00 -5.92808723e-01 4.03015167e-01
7.40889430e-01 -7.57646859e-01 1.35916960e+00 7.75024414e-01
1.86810076e-01 -1.17932725e+00 -1.03471482e+00 -7.38494217e-01
3.61496508e-02 -7.87644148e-01 8.84548068e-01 5.85137129e-01
1.27762824e-01 7.21892789e-02 -4.31968898e-01 3.44455987e-01
8.65807712e-01 7.91274130e-01 1.03353703e+00 -8.56265903e-01
-9.65058625e-01 -4.52959627e-01 1.33025448e-03 -2.04860616e+00
-2.76857585e-01 -5.17326236e-01 -3.29464227e-01 -1.01513696e+00
3.21065873e-01 -8.91455054e-01 -3.36001545e-01 1.51275113e-01
-3.58042419e-02 5.57608962e-01 8.67918506e-02 2.60028750e-01
-7.94832945e-01 2.94794858e-01 1.07188380e+00 -1.50929078e-01
-3.04677576e-01 2.24709213e-01 -1.02801144e+00 6.27337992e-01
4.85301405e-01 -3.32067609e-01 -3.63653064e-01 -4.07738686e-01
6.27593875e-01 3.41682017e-01 -1.71657979e-01 -6.94938838e-01
3.53691667e-01 -3.64117026e-01 -2.47941557e-02 -1.05719137e+00
6.19732261e-01 -6.56119585e-01 1.64002813e-02 4.74541694e-01
-1.98324963e-01 -1.69766769e-02 4.47474904e-02 6.79708183e-01
3.26957479e-02 -3.43724310e-01 7.29810059e-01 -1.01379186e-01
-8.33687708e-02 6.57790482e-01 1.13181189e-01 1.66867793e-01
9.70183194e-01 -1.76295335e-03 -1.76731646e-01 -4.56961811e-01
-3.82289618e-01 1.96567908e-01 1.45721108e-01 -3.38058144e-01
6.20616972e-01 -1.33123529e+00 -4.71102834e-01 2.53689378e-01
-3.71498950e-02 1.37366101e-01 2.64032423e-01 7.27371693e-01
-5.83127558e-01 6.48570597e-01 7.30089426e-01 -6.54705405e-01
-1.28296983e+00 7.71112740e-01 -6.06903434e-02 -3.54064912e-01
-2.53884405e-01 1.31014705e+00 1.11740351e-01 3.83208245e-02
3.90913278e-01 -2.44609892e-01 4.50945377e-01 -7.09983483e-02
9.33199525e-01 5.22841752e-01 -2.25097034e-02 -1.42718554e-01
-1.90189242e-01 6.78392053e-01 -3.81799638e-01 -3.66852552e-01
1.37341666e+00 -6.89428300e-02 -6.31288961e-02 -1.75363317e-01
1.47180140e+00 3.46496671e-01 -1.16636586e+00 -5.92405796e-01
-2.41200365e-02 -9.05160964e-01 6.64510718e-03 -2.66943425e-01
-1.15813208e+00 3.19679797e-01 4.80738968e-01 5.48504829e-01
1.17463565e+00 1.76759705e-01 1.27620411e+00 6.25381768e-01
6.85241640e-01 -1.11275530e+00 1.80019319e-01 2.11689427e-01
6.65251434e-01 -1.12813807e+00 2.83878833e-01 -4.15488839e-01
-1.96944878e-01 7.32652187e-01 2.06822827e-01 -5.85431337e-01
1.00845516e+00 -5.03129698e-03 -5.50342739e-01 -1.84078321e-01
-5.80668688e-01 -1.05400130e-01 2.58094341e-01 -2.25380994e-02
1.63033485e-01 1.44361079e-01 -5.80986381e-01 7.17952013e-01
-4.01088417e-01 -8.85770544e-02 -1.15580671e-01 7.97026753e-01
-7.49026239e-01 -1.19233358e+00 -8.03362608e-01 8.17548573e-01
-6.36203587e-01 -3.38211983e-01 2.25058109e-01 3.16591084e-01
-1.60735562e-01 9.51830566e-01 1.48336262e-01 -2.52665490e-01
1.97279856e-01 -3.49022329e-01 5.08483410e-01 -3.02044392e-01
-1.21108226e-01 3.95004153e-01 -4.82308827e-02 -7.66494930e-01
-3.26833844e-01 -7.04132915e-01 -1.25034761e+00 -7.35973537e-01
-2.98056841e-01 5.95986545e-01 7.21313894e-01 6.57687366e-01
5.93395889e-01 -1.97781235e-01 8.94443274e-01 -5.53309500e-01
-7.88820505e-01 -5.73768973e-01 -6.56326532e-01 2.39593595e-01
3.23157102e-01 -6.83300793e-01 -3.31922084e-01 -2.74131179e-01]
|
[6.603999137878418, 4.63886833190918]
|
212241de-3f9a-4036-b5b6-999b6bec1be9
|
omni-gan-on-the-secrets-of-cgans-and-beyond
|
2011.13074
| null |
https://arxiv.org/abs/2011.13074v3
|
https://arxiv.org/pdf/2011.13074v3.pdf
|
Omni-GAN: On the Secrets of cGANs and Beyond
|
The conditional generative adversarial network (cGAN) is a powerful tool of generating high-quality images, but existing approaches mostly suffer unsatisfying performance or the risk of mode collapse. This paper presents Omni-GAN, a variant of cGAN that reveals the devil in designing a proper discriminator for training the model. The key is to ensure that the discriminator receives strong supervision to perceive the concepts and moderate regularization to avoid collapse. Omni-GAN is easily implemented and freely integrated with off-the-shelf encoding methods (e.g., implicit neural representation, INR). Experiments validate the superior performance of Omni-GAN and Omni-INR-GAN in a wide range of image generation and restoration tasks. In particular, Omni-INR-GAN sets new records on the ImageNet dataset with impressive Inception scores of 262.85 and 343.22 for the image sizes of 128 and 256, respectively, surpassing the previous records by 100+ points. Moreover, leveraging the generator prior, Omni-INR-GAN can extrapolate low-resolution images to arbitrary resolution, even up to x60+ higher resolution. Code is available.
|
['Cong Geng', 'Qi Tian', 'Bingbing Ni', 'Lingxi Xie', 'Peng Zhou']
|
2020-11-26
| null |
http://openaccess.thecvf.com//content/ICCV2021/html/Zhou_Omni-GAN_On_the_Secrets_of_cGANs_and_Beyond_ICCV_2021_paper.html
|
http://openaccess.thecvf.com//content/ICCV2021/papers/Zhou_Omni-GAN_On_the_Secrets_of_cGANs_and_Beyond_ICCV_2021_paper.pdf
|
iccv-2021-1
|
['conditional-image-generation']
|
['computer-vision']
|
[ 4.39357549e-01 1.42941222e-01 2.32563347e-01 -1.59596786e-01
-8.30321908e-01 -3.78172427e-01 5.57093680e-01 -9.16020155e-01
-4.82727475e-02 1.00172019e+00 6.49990514e-02 -1.92148536e-01
2.08533764e-01 -1.11871564e+00 -7.85914898e-01 -9.80881870e-01
1.37065604e-01 1.60130560e-01 -1.87116459e-01 -3.84043604e-01
-1.20089315e-01 2.95332164e-01 -1.38937163e+00 2.28296667e-01
1.04619217e+00 9.91593182e-01 2.48411208e-01 8.03966284e-01
3.34783822e-01 9.59072769e-01 -8.77114654e-01 -5.46806395e-01
4.63481128e-01 -7.81344593e-01 -4.21701133e-01 -5.69204986e-02
4.20753539e-01 -6.30794346e-01 -4.95728374e-01 1.07146060e+00
7.51750112e-01 -2.07117468e-01 6.23553216e-01 -1.31516612e+00
-1.23685527e+00 3.71378511e-01 -5.67521513e-01 5.51862232e-02
2.22552434e-01 4.91703808e-01 6.40363216e-01 -7.46101081e-01
5.40601492e-01 1.10555446e+00 6.45716012e-01 8.53958547e-01
-1.24846125e+00 -9.34408367e-01 -2.85523951e-01 -2.19197616e-01
-1.40823007e+00 -3.76245469e-01 6.78641319e-01 -3.36080313e-01
5.56041539e-01 3.13099802e-01 4.68461096e-01 1.44678450e+00
2.10942760e-01 4.66477841e-01 1.39262879e+00 -2.18564376e-01
1.51337907e-01 -6.27496317e-02 -6.27609253e-01 5.07205188e-01
1.68660492e-01 4.47809517e-01 -4.41272974e-01 1.32743359e-01
1.29069185e+00 -8.23288560e-02 -3.60141963e-01 2.77381033e-01
-9.79353964e-01 8.10086727e-01 6.96286559e-01 1.70878187e-01
-4.16399062e-01 3.02279204e-01 1.01820581e-01 4.65160519e-01
3.04262429e-01 5.05677879e-01 1.20978788e-01 5.67043647e-02
-9.58635569e-01 1.56942919e-01 1.52005687e-01 1.18806601e+00
6.24639213e-01 7.60288060e-01 -2.99116790e-01 7.18269944e-01
3.78003856e-03 7.16922879e-01 6.21698320e-01 -9.94381607e-01
2.40545020e-01 2.09082767e-01 -1.49721578e-01 -7.79478312e-01
1.57058418e-01 -6.52012467e-01 -1.52449775e+00 6.92756355e-01
1.33793384e-01 -2.06184655e-01 -1.17546368e+00 1.78373575e+00
-1.05292030e-01 1.82522073e-01 1.74919352e-01 9.04659510e-01
9.93861735e-01 8.13470900e-01 8.75194147e-02 1.06399566e-01
1.11557782e+00 -8.49602699e-01 -5.61274886e-01 -3.65679413e-01
-2.68091336e-02 -6.48906946e-01 1.19297135e+00 4.72534120e-01
-1.31057692e+00 -9.53747690e-01 -1.22213471e+00 1.34977192e-01
-1.03497878e-01 2.25563645e-01 7.26063728e-01 7.00684488e-01
-1.26568747e+00 5.16893566e-01 -4.82054919e-01 5.70611469e-02
6.98473334e-01 1.08932495e-01 -4.71952528e-01 -3.41441125e-01
-1.09851325e+00 4.76549029e-01 3.85768801e-01 -2.52308440e-03
-1.28610289e+00 -7.50267088e-01 -7.34555304e-01 -4.18368652e-02
-4.81749773e-02 -9.28825498e-01 7.78899968e-01 -1.21017063e+00
-1.60783029e+00 7.57875681e-01 3.23184818e-01 -5.76187134e-01
7.30972171e-01 -7.15089589e-02 -6.00342751e-01 1.95352703e-01
1.64027244e-01 9.18538094e-01 1.22861600e+00 -1.43645132e+00
-3.45122755e-01 -1.76643997e-01 8.48898143e-02 1.05242223e-01
-1.83123723e-01 -1.40967265e-01 -2.80065358e-01 -9.93858933e-01
-1.22992016e-01 -7.59460211e-01 -1.50135621e-01 -7.70968124e-02
-6.03227615e-01 2.14104712e-01 9.16059554e-01 -7.71517456e-01
8.84289265e-01 -2.16493869e+00 -1.23019449e-01 -4.22229953e-02
2.31711537e-01 4.39075500e-01 -3.45443100e-01 1.84689149e-01
-2.18599856e-01 1.46786183e-01 -3.88142735e-01 -2.45026261e-01
-2.34922364e-01 3.29093188e-01 -4.85786736e-01 3.32492471e-01
3.29192549e-01 1.18691373e+00 -8.06347549e-01 -2.38883629e-01
2.32334495e-01 7.96223044e-01 -4.46727395e-01 5.66162944e-01
3.56787741e-02 7.87906110e-01 -1.04867019e-01 6.97801471e-01
8.23663116e-01 -3.33054572e-01 -1.21228687e-01 -1.08411551e-01
2.67215133e-01 -9.55180302e-02 -8.11997175e-01 1.61703491e+00
-4.94513899e-01 5.72514236e-01 4.93073603e-03 -7.82549381e-01
1.06557846e+00 3.82273406e-01 1.93964750e-01 -9.31668520e-01
-8.61782283e-02 1.72523350e-01 -1.47429824e-01 -1.85161963e-01
3.73006046e-01 -3.79871577e-01 -5.00299372e-02 2.26994231e-01
2.64515549e-01 -2.24730879e-01 5.70310950e-02 1.90345734e-01
9.61688101e-01 1.74695417e-01 1.03399977e-01 -4.64991294e-02
3.87337983e-01 -4.45713967e-01 4.81033087e-01 7.08002031e-01
1.48267940e-01 1.19671929e+00 2.44123623e-01 -2.83706605e-01
-1.32160985e+00 -1.28238976e+00 1.39728775e-02 6.90827429e-01
-4.74204645e-02 -1.54072255e-01 -8.57347906e-01 -5.36095858e-01
-3.80308062e-01 7.06271112e-01 -7.17196763e-01 -3.18250179e-01
-5.29841781e-01 -8.75893235e-01 7.55277038e-01 7.08328247e-01
9.35627759e-01 -1.30295002e+00 -4.16706741e-01 5.92102408e-02
-2.23879978e-01 -1.05489993e+00 -2.78058439e-01 -1.45095915e-01
-7.25461841e-01 -7.84889638e-01 -9.54128683e-01 -7.08134174e-01
8.67190301e-01 1.01575553e-01 1.41347837e+00 1.15624748e-01
-3.10637593e-01 9.10768006e-03 -2.07037330e-01 -1.01239748e-01
-7.67558455e-01 -8.34520161e-02 -3.19595426e-01 -5.63595183e-02
-2.45324910e-01 -8.75151634e-01 -8.61001492e-01 3.56989264e-01
-1.11818123e+00 3.27703387e-01 8.15627575e-01 1.04608893e+00
6.99891925e-01 3.02653670e-01 8.17063272e-01 -8.72767270e-01
5.67963898e-01 -4.02712941e-01 -4.27571446e-01 -7.47959688e-02
-6.82388783e-01 -2.88825244e-01 1.03022802e+00 -4.12264764e-01
-1.19523036e+00 -2.70575047e-01 -5.05850971e-01 -4.33736116e-01
-2.01038942e-01 1.47664800e-01 -4.10579562e-01 -8.60883016e-03
7.97300994e-01 4.14330721e-01 -1.49258628e-01 -3.06662053e-01
4.44498450e-01 3.90611440e-01 1.05116653e+00 -4.23583537e-01
1.13148868e+00 5.66903234e-01 -2.03569487e-01 -6.28788114e-01
-6.51574135e-01 2.56898016e-01 -1.06305867e-01 -1.19915284e-01
8.98438573e-01 -1.27540064e+00 -3.28100562e-01 8.36799026e-01
-8.39334011e-01 -5.97306550e-01 -6.05064094e-01 1.00661315e-01
-6.74288511e-01 9.14461762e-02 -7.96335042e-01 -6.25357628e-01
-6.30298913e-01 -9.79794741e-01 9.62920785e-01 5.02045393e-01
1.18482210e-01 -7.87665129e-01 -2.01826930e-01 4.88886505e-01
8.16936553e-01 6.73391163e-01 7.06040800e-01 -3.97198386e-02
-6.07920229e-01 -1.48340836e-01 -3.03302854e-01 7.93355167e-01
4.16604914e-02 -1.07933663e-01 -1.09575582e+00 -5.59002340e-01
7.08170459e-02 -4.78328735e-01 7.63277352e-01 3.23292613e-01
1.35732448e+00 -5.37063241e-01 1.40568852e-01 1.10915995e+00
1.67844415e+00 2.93471903e-01 1.44606102e+00 2.46350601e-01
7.04518497e-01 8.34110230e-02 2.62687683e-01 3.39222401e-01
2.90022548e-02 5.30759871e-01 6.96527660e-01 -5.04786968e-01
-6.34833872e-01 -6.33876383e-01 3.95910442e-01 6.78189754e-01
-3.27876568e-01 -2.91812867e-01 -4.91142422e-01 3.19728792e-01
-1.27151823e+00 -1.09318566e+00 1.24231592e-01 2.08333611e+00
9.94246840e-01 1.10143781e-01 -6.68407008e-02 1.53335750e-01
6.83479548e-01 1.49974763e-01 -6.08563304e-01 -1.50086761e-01
-5.41973650e-01 5.22549689e-01 4.03707892e-01 2.59126723e-01
-8.83948684e-01 8.06725979e-01 6.59762287e+00 9.89195824e-01
-1.22651637e+00 1.39873624e-01 9.94921148e-01 1.50494277e-01
-3.22084993e-01 -1.32770360e-01 -4.79082376e-01 7.51281381e-01
8.96015763e-01 -4.02895920e-02 7.19697773e-01 8.42005432e-01
-1.60036176e-01 1.86153412e-01 -7.21027911e-01 9.74865973e-01
1.80758700e-01 -1.33734131e+00 1.88006595e-01 1.49285913e-01
1.07331133e+00 -2.07467064e-01 4.82385188e-01 3.41655642e-01
4.76553202e-01 -1.58427227e+00 7.11131454e-01 4.93134946e-01
1.57204282e+00 -7.88564503e-01 7.34713852e-01 1.99756846e-01
-8.12330484e-01 8.92746001e-02 -5.29740036e-01 1.65113300e-01
2.53526360e-01 6.66067839e-01 -5.37071466e-01 6.45470738e-01
7.92154968e-01 5.66992521e-01 -6.66331887e-01 5.62566042e-01
-6.11874640e-01 7.95392334e-01 -2.41104458e-02 8.03642988e-01
1.12966366e-01 -3.39518368e-01 3.82166654e-01 1.00999212e+00
6.62603140e-01 -2.16510911e-02 -1.91792637e-01 1.19454455e+00
-3.14480245e-01 -3.95128310e-01 -6.57583177e-01 7.12468848e-02
2.87101090e-01 1.19140792e+00 -3.97808164e-01 -4.11230505e-01
-9.39577445e-02 1.18051410e+00 8.61332715e-02 5.56778848e-01
-1.07079291e+00 -4.03246284e-01 5.43908477e-01 2.22291619e-01
3.78114671e-01 -8.84386152e-03 -3.70874494e-01 -1.04050338e+00
-7.55575374e-02 -1.22908521e+00 1.37799248e-01 -1.20955467e+00
-1.36224592e+00 1.02857685e+00 -2.67160147e-01 -1.26133955e+00
-4.95735168e-01 -4.06618148e-01 -8.05970371e-01 1.03884399e+00
-1.58489275e+00 -1.44269717e+00 -7.21287906e-01 8.03911328e-01
2.68414944e-01 -3.91623437e-01 9.08648431e-01 4.20380414e-01
-4.89424706e-01 8.19783866e-01 3.02531710e-03 3.01017374e-01
5.00057340e-01 -1.04718983e+00 5.66105843e-01 1.03088796e+00
6.87502325e-02 4.75400478e-01 5.59803605e-01 -4.14405078e-01
-1.17481852e+00 -1.29846072e+00 3.07490826e-01 -1.04756474e-01
1.54666573e-01 -2.79459596e-01 -8.35929275e-01 6.94246829e-01
5.16586125e-01 1.42868608e-01 5.46069920e-01 -5.78086913e-01
-4.13099766e-01 -1.68882355e-01 -1.49206114e+00 6.40888453e-01
1.00166631e+00 -4.52453315e-01 -9.44932029e-02 1.26381546e-01
6.29219472e-01 -4.99839962e-01 -9.95913923e-01 4.12320137e-01
3.15389007e-01 -1.45720077e+00 1.31291819e+00 -1.64463148e-01
9.74562645e-01 -3.15787733e-01 -1.97055429e-01 -1.34450805e+00
-4.25351441e-01 -7.76865959e-01 -6.19585812e-02 1.32082760e+00
1.04729012e-01 -8.19867492e-01 7.05034375e-01 1.21299982e-01
-1.13761202e-01 -7.26573408e-01 -8.40927482e-01 -8.02229822e-01
2.07887277e-01 -1.70584753e-01 9.05146837e-01 7.36272693e-01
-7.66399860e-01 2.59909809e-01 -7.34836698e-01 5.43401875e-02
8.29985499e-01 4.46809120e-02 9.00217116e-01 -8.08461845e-01
-5.31217813e-01 -2.37962395e-01 -4.68712687e-01 -1.04445422e+00
-6.24027178e-02 -8.55927944e-01 -1.64940819e-01 -1.46761453e+00
7.12784678e-02 -6.42006218e-01 -1.87442809e-01 4.33576614e-01
-2.16260448e-01 9.28851426e-01 2.57961899e-01 2.91043043e-01
-6.39353618e-02 7.03991115e-01 1.57795918e+00 -2.09083781e-01
2.10207000e-01 -1.70587748e-01 -1.05279732e+00 5.41914165e-01
1.00463116e+00 -2.03912154e-01 -6.40075028e-01 -4.39645648e-01
5.43904603e-02 1.39575638e-02 6.72831953e-01 -1.27591169e+00
-2.48078331e-01 4.94653285e-02 7.57581949e-01 -3.13072026e-01
3.89808714e-01 -5.51212072e-01 7.01506078e-01 2.91011989e-01
-2.32371286e-01 1.21553719e-01 4.30794694e-02 4.14503574e-01
-3.78662229e-01 -8.28325227e-02 1.05558395e+00 -3.15225512e-01
-6.30565822e-01 4.31387722e-01 -3.98652367e-02 3.71671200e-01
9.44455206e-01 -2.84922928e-01 -5.22021413e-01 -6.40218258e-01
-4.64194536e-01 -3.43187124e-01 6.98844433e-01 2.83995628e-01
9.08628881e-01 -1.64674759e+00 -1.03315985e+00 6.45888150e-01
-1.15478814e-01 2.00598970e-01 4.19887573e-01 3.66084814e-01
-4.90826309e-01 -4.41835299e-02 -5.84057271e-01 -4.43132371e-01
-8.22904587e-01 4.98564720e-01 2.78548717e-01 -4.08425689e-01
-9.06024158e-01 8.27713788e-01 4.72057879e-01 -1.14267409e-01
-1.68694526e-01 2.28448123e-01 4.70772721e-02 -4.70433861e-01
6.30908668e-01 2.56291121e-01 -6.94521070e-02 -5.91853619e-01
5.94389811e-02 3.03336352e-01 1.73017830e-01 -8.85899812e-02
1.41120172e+00 6.14559688e-02 -8.58754516e-02 -1.75807834e-01
1.07217908e+00 -6.77269101e-02 -1.57247567e+00 5.21579050e-02
-7.85165370e-01 -7.09554493e-01 -8.18310231e-02 -9.73205745e-01
-1.59221959e+00 7.70316124e-01 6.37498736e-01 7.27871284e-02
1.55108845e+00 -1.97079122e-01 9.26481366e-01 -4.19856250e-01
4.50617880e-01 -5.91796577e-01 2.36221030e-01 1.13840200e-01
1.19759142e+00 -1.08828259e+00 -4.26142663e-02 -1.49068162e-01
-7.59938300e-01 7.39585936e-01 7.44068086e-01 -3.68084103e-01
2.68236130e-01 4.55647379e-01 2.04679877e-01 -4.64111976e-02
-5.17251015e-01 1.88453838e-01 1.58839613e-01 1.00024080e+00
3.50844175e-01 1.17355727e-01 1.04332395e-01 4.14819449e-01
-5.69889307e-01 -1.49445294e-03 5.30322015e-01 6.78425670e-01
1.12946436e-01 -8.77866983e-01 -4.15290654e-01 3.67597699e-01
-5.37117422e-01 -2.87542611e-01 1.53177157e-01 5.98136723e-01
2.90264279e-01 8.70350361e-01 -8.89395922e-03 -3.90318185e-01
1.06351838e-01 -2.61068434e-01 3.91454279e-01 -1.86907247e-01
-4.92278725e-01 -1.14308663e-01 -3.02189410e-01 -5.18972933e-01
-3.56446385e-01 -2.68872201e-01 -9.28638697e-01 -4.29750115e-01
-1.48401886e-01 -9.35883224e-02 5.31424046e-01 4.03863490e-01
4.06584352e-01 7.78687537e-01 6.47724271e-01 -8.20893824e-01
-6.12018526e-01 -9.27006900e-01 -6.02968693e-01 6.35040760e-01
1.63531914e-01 -3.49224776e-01 -4.56058502e-01 2.77365237e-01]
|
[11.632065773010254, -0.6071309447288513]
|
b45cd4be-9d0a-4165-8ba3-3e3fd244caf5
|
ordnet-capturing-omni-range-dependencies-for
|
2101.03929
| null |
https://arxiv.org/abs/2101.03929v1
|
https://arxiv.org/pdf/2101.03929v1.pdf
|
ORDNet: Capturing Omni-Range Dependencies for Scene Parsing
|
Learning to capture dependencies between spatial positions is essential to many visual tasks, especially the dense labeling problems like scene parsing. Existing methods can effectively capture long-range dependencies with self-attention mechanism while short ones by local convolution. However, there is still much gap between long-range and short-range dependencies, which largely reduces the models' flexibility in application to diverse spatial scales and relationships in complicated natural scene images. To fill such a gap, we develop a Middle-Range (MR) branch to capture middle-range dependencies by restricting self-attention into local patches. Also, we observe that the spatial regions which have large correlations with others can be emphasized to exploit long-range dependencies more accurately, and thus propose a Reweighed Long-Range (RLR) branch. Based on the proposed MR and RLR branches, we build an Omni-Range Dependencies Network (ORDNet) which can effectively capture short-, middle- and long-range dependencies. Our ORDNet is able to extract more comprehensive context information and well adapt to complex spatial variance in scene images. Extensive experiments show that our proposed ORDNet outperforms previous state-of-the-art methods on three scene parsing benchmarks including PASCAL Context, COCO Stuff and ADE20K, demonstrating the superiority of capturing omni-range dependencies in deep models for scene parsing task.
|
['Shuicheng Yan', 'Jiashi Feng', 'Bo Li', 'Jizhong Han', 'Tianrui Hui', 'Si Liu', 'Shaofei Huang']
|
2021-01-11
| null | null | null | null |
['scene-parsing']
|
['computer-vision']
|
[ 8.99151936e-02 -2.85712332e-01 -8.05332363e-02 -7.49434352e-01
-1.61085159e-01 -4.55181807e-01 5.33926904e-01 -5.26024736e-02
-4.98381108e-01 5.29297829e-01 4.67537433e-01 -3.05152327e-01
-2.10467532e-01 -9.99037743e-01 -9.55741704e-01 -5.96311927e-01
-5.19244783e-02 1.94992349e-02 8.24845970e-01 -3.42758507e-01
2.16327116e-01 5.65935194e-01 -1.33551371e+00 5.86577356e-01
8.39732409e-01 7.91547477e-01 8.76420557e-01 5.00357568e-01
-6.51658475e-01 9.28216159e-01 -6.65036380e-01 -1.42542616e-01
-1.00020275e-01 -1.60580173e-01 -8.03559124e-01 -2.77126282e-01
4.80381221e-01 -1.67214587e-01 -2.79877394e-01 1.06692421e+00
3.65190506e-01 6.36887699e-02 3.03891540e-01 -7.93753684e-01
-8.68510187e-01 6.69560075e-01 -9.78351772e-01 5.61753273e-01
5.93529940e-02 1.68930024e-01 9.97872829e-01 -6.35556936e-01
4.56297874e-01 1.65299225e+00 5.26926100e-01 3.10972393e-01
-7.18149066e-01 -7.50529528e-01 9.24568474e-01 2.05093235e-01
-1.03782523e+00 5.39574400e-02 9.19424891e-01 -2.33562097e-01
1.06207991e+00 2.12334603e-01 3.92891526e-01 9.87592995e-01
1.52463809e-01 8.69946241e-01 1.20995271e+00 -1.07529923e-01
-9.10675004e-02 -4.13862616e-01 3.02080542e-01 5.34299195e-01
1.20952033e-01 -2.60139778e-02 -2.09895238e-01 3.36670697e-01
1.05544996e+00 1.88385874e-01 -7.54475594e-02 -8.73513371e-02
-1.12703431e+00 6.57079875e-01 1.07918382e+00 5.48746884e-01
-1.34661496e-01 1.78949371e-01 3.52958202e-01 -9.48211551e-02
2.95285523e-01 9.33763161e-02 -6.20839179e-01 1.32189453e-01
-3.33193839e-01 -1.08797900e-01 2.30861589e-01 1.08264208e+00
9.79429007e-01 -2.37722486e-01 -3.85319084e-01 1.17170954e+00
2.00491875e-01 2.58276194e-01 4.05204117e-01 -5.52720606e-01
7.75931835e-01 8.26893687e-01 -3.09644699e-01 -1.00176215e+00
-7.28336215e-01 -6.44377172e-01 -1.08061934e+00 -3.67569318e-03
3.21451455e-01 3.31745632e-02 -1.15321147e+00 1.91268373e+00
3.81423861e-01 4.72956330e-01 4.47574817e-03 1.01856971e+00
1.21623719e+00 8.20778847e-01 5.10755062e-01 3.28238457e-02
1.43876994e+00 -1.30205369e+00 -4.68055129e-01 -8.37295949e-01
4.39069569e-01 -7.45881498e-01 1.40641415e+00 -1.49309725e-01
-5.66916645e-01 -9.55448031e-01 -8.63982916e-01 -4.03839916e-01
-5.26918292e-01 -3.23725119e-02 1.06049109e+00 1.30497664e-01
-7.65369236e-01 2.92600632e-01 -4.43932444e-01 -1.85848072e-01
7.59026289e-01 1.99732795e-01 -4.09568757e-01 -5.19976258e-01
-1.24803996e+00 4.58928078e-01 3.91846091e-01 3.87324333e-01
-4.88307953e-01 -5.99263012e-01 -9.95902479e-01 7.56387711e-02
4.17127728e-01 -4.78858352e-01 7.42886961e-01 -9.00179982e-01
-1.16325188e+00 7.52245784e-01 -2.18567416e-01 5.83608039e-02
2.23157451e-01 -3.42411101e-01 -3.74068618e-01 -3.33921760e-02
3.77726674e-01 8.82288992e-01 4.46873277e-01 -1.23708761e+00
-6.96227908e-01 -4.15220946e-01 4.73998457e-01 3.48103672e-01
-1.53056338e-01 -6.77491948e-02 -8.67967546e-01 -7.07877636e-01
2.59243131e-01 -6.46403134e-01 -3.42865705e-01 -1.87261224e-01
-5.16653895e-01 -3.88610333e-01 1.14546418e+00 -2.67602742e-01
1.18710220e+00 -2.17300773e+00 -1.48967654e-01 -3.10511708e-01
1.07731171e-01 5.80393434e-01 -5.07633626e-01 1.17350526e-01
-2.15920731e-01 1.36885524e-01 -3.25034022e-01 -1.69951826e-01
-3.64643365e-01 7.21890092e-01 -2.14651912e-01 2.63462905e-02
3.93522263e-01 1.09161723e+00 -9.91424680e-01 -5.44035971e-01
5.24687707e-01 4.53201115e-01 -4.29382980e-01 3.32661301e-01
-3.25331807e-01 6.27256930e-01 -6.82894766e-01 5.29856145e-01
1.06816363e+00 -3.09616357e-01 -1.13055058e-01 -3.19439381e-01
-2.44102120e-01 2.79363662e-01 -9.59375203e-01 1.87432861e+00
-7.26128459e-01 5.29309869e-01 7.79932216e-02 -9.87235129e-01
9.92692888e-01 -4.06140000e-01 8.75122547e-02 -1.06469989e+00
-7.05811009e-02 -8.64083692e-02 2.29350589e-02 -6.58019900e-01
1.67457432e-01 4.99652289e-02 -1.16224796e-01 -6.25626445e-02
-8.44566151e-02 2.81021565e-01 1.65436104e-01 1.92278236e-01
9.79140520e-01 2.02527314e-01 1.70545280e-01 -4.25182045e-01
6.97970331e-01 -3.70265484e-01 9.03987467e-01 6.92262530e-01
-3.11296105e-01 7.21565127e-01 7.09312260e-01 -7.32352018e-01
-5.08125663e-01 -8.17413807e-01 -2.15059876e-01 1.47513366e+00
6.47356987e-01 -3.02310258e-01 -5.91587126e-01 -9.68954146e-01
-1.49204820e-01 4.19591606e-01 -8.36025000e-01 5.93708530e-02
-1.00325918e+00 -8.43085468e-01 3.45898181e-01 1.04113722e+00
9.76061225e-01 -1.67964280e+00 -6.11080766e-01 2.71701902e-01
-1.08371541e-01 -1.44420278e+00 -4.85952288e-01 4.38135713e-01
-6.44298315e-01 -9.95891392e-01 -5.69113016e-01 -1.03906989e+00
5.42424917e-01 5.51387846e-01 1.40245831e+00 1.29568532e-01
-3.13597381e-01 -9.29604918e-02 -4.33537900e-01 -2.92767495e-01
2.31685698e-01 3.26945297e-02 -6.82229340e-01 -1.02089085e-01
4.78412628e-01 -7.54190981e-01 -7.15274215e-01 4.84169930e-01
-9.25412774e-01 2.55885720e-01 8.65557373e-01 9.55031872e-01
8.76414478e-01 -1.88802965e-02 5.79943180e-01 -1.16615069e+00
3.04931819e-01 -4.75265235e-01 -6.01384997e-01 3.35879982e-01
-7.65476227e-02 8.14692825e-02 8.11797440e-01 -3.68172944e-01
-1.32043076e+00 1.21835798e-01 -4.30788636e-01 -2.02203289e-01
-5.78690886e-01 3.28080118e-01 -6.65463150e-01 8.41277391e-02
1.53771937e-01 1.59212835e-02 -7.32254982e-01 -4.98715460e-01
5.19968212e-01 3.91148776e-01 6.10396624e-01 -8.67161036e-01
3.58919650e-01 4.60802853e-01 1.09371245e-01 -8.64193559e-01
-1.16424012e+00 -4.00291055e-01 -8.62596571e-01 2.83798426e-01
1.32642472e+00 -9.01889443e-01 -4.12045091e-01 6.23486400e-01
-1.38373721e+00 -5.65084338e-01 -1.33505061e-01 1.38294891e-01
-1.04030810e-01 2.62179434e-01 -6.29824460e-01 -5.15976429e-01
-9.43444520e-02 -1.27525175e+00 1.37248945e+00 6.80865169e-01
3.67317289e-01 -1.00543165e+00 -1.49184272e-01 1.96555451e-01
3.43135238e-01 2.98701644e-01 1.06028330e+00 -2.16905400e-01
-7.43246675e-01 3.55597973e-01 -8.79541814e-01 1.53848395e-01
1.06606372e-01 -2.07409874e-01 -1.07479131e+00 4.33556810e-02
-2.76802272e-01 -2.20415637e-01 1.34034169e+00 5.76544881e-01
1.66569901e+00 7.35697001e-02 -3.05081725e-01 1.12085307e+00
1.43128252e+00 1.17193229e-01 8.57716560e-01 2.92399317e-01
1.21027887e+00 6.85139537e-01 7.00122178e-01 5.36965504e-02
6.03084624e-01 4.79765505e-01 7.16207564e-01 -5.60457766e-01
-3.85747582e-01 -3.69272590e-01 -9.21772495e-02 5.86067379e-01
6.73279837e-02 -2.26813436e-01 -6.88223124e-01 4.88954991e-01
-1.87112892e+00 -6.94269180e-01 -1.69277847e-01 1.73619556e+00
4.95772243e-01 4.11133528e-01 -2.21934378e-01 -3.34356636e-01
8.98973107e-01 6.51758492e-01 -6.91475868e-01 -3.81942123e-01
-4.12313223e-01 2.23479614e-01 5.01699209e-01 3.38406175e-01
-1.28557968e+00 1.23275149e+00 5.39395666e+00 9.54408646e-01
-1.34284568e+00 2.31942963e-02 9.04198825e-01 3.40109676e-01
-3.09127957e-01 7.80817680e-03 -8.45962107e-01 4.90597337e-01
3.53747964e-01 5.75047791e-01 1.20766915e-01 8.75699878e-01
-1.41892776e-01 -5.27905561e-02 -7.61069179e-01 1.00772250e+00
-2.17630237e-01 -1.18849874e+00 9.97134577e-03 -6.76733777e-02
6.54272020e-01 2.86352873e-01 -1.43925875e-01 4.72045124e-01
3.07295799e-01 -1.18599284e+00 4.54162300e-01 1.78304985e-01
8.30573201e-01 -7.20534801e-01 7.09533870e-01 4.08301175e-01
-1.92302012e+00 -2.37138689e-01 -8.01939785e-01 -1.51299953e-01
1.20390669e-01 7.90343642e-01 -6.84785843e-03 6.19068205e-01
9.24725771e-01 9.18551624e-01 -8.36495697e-01 7.25896835e-01
-5.16136885e-01 3.88884962e-01 -2.66464680e-01 -6.10317104e-02
5.50035179e-01 -3.11054319e-01 1.57063335e-01 1.67489028e+00
-3.29962634e-02 2.69286215e-01 2.12919101e-01 9.18363392e-01
3.06071551e-03 -5.89342713e-02 -5.03596425e-01 3.61735374e-01
4.57191378e-01 1.36106658e+00 -1.06980920e+00 -3.37117165e-01
-6.13372505e-01 9.56434309e-01 6.56900227e-01 4.15912628e-01
-9.56404567e-01 -4.39043105e-01 6.05520785e-01 5.19276857e-02
5.10785580e-01 -2.32747644e-01 -3.67589206e-01 -1.13507485e+00
1.61212489e-01 -5.92163444e-01 6.02320313e-01 -8.50648284e-01
-1.38969851e+00 9.17515039e-01 1.55169025e-01 -8.57194841e-01
3.84205490e-01 -7.56040931e-01 -9.23596561e-01 8.59402180e-01
-1.97509634e+00 -1.60238683e+00 -6.92966104e-01 7.74040341e-01
8.03007126e-01 2.52291054e-01 5.65488398e-01 3.04579347e-01
-6.81673825e-01 4.10533369e-01 -2.85635501e-01 2.48741403e-01
5.39191306e-01 -1.41697955e+00 5.07138789e-01 8.84584904e-01
2.44851246e-01 7.45433867e-01 1.75360367e-02 -4.89772528e-01
-9.73732769e-01 -1.33629680e+00 4.52946752e-01 -1.77317664e-01
2.81406611e-01 -3.82872909e-01 -1.03695440e+00 6.52825892e-01
1.21995054e-01 7.25431085e-01 2.72942603e-01 3.37995410e-01
-6.59754157e-01 -3.13809216e-01 -8.35180402e-01 4.96653855e-01
1.53816152e+00 -5.33029974e-01 -5.27154446e-01 2.16627389e-01
1.01592946e+00 -3.98839593e-01 -3.69800866e-01 7.99802780e-01
2.96196520e-01 -1.37703085e+00 1.17671061e+00 -2.56097376e-01
6.35905862e-01 -3.42070132e-01 -1.98609158e-01 -1.02565849e+00
-5.45054317e-01 -3.06525886e-01 2.45137706e-01 1.44907653e+00
1.55085251e-01 -6.78154707e-01 6.08125567e-01 1.55372694e-01
-3.82628709e-01 -9.84353781e-01 -8.07736814e-01 -5.65494180e-01
3.91316205e-01 -4.07219470e-01 8.50917518e-01 9.94593561e-01
-5.73667586e-01 6.72428370e-01 -2.09410444e-01 3.15248609e-01
3.19791675e-01 5.16041756e-01 7.29953170e-01 -1.05115354e+00
-3.22152108e-01 -5.91110706e-01 -3.89963537e-01 -1.75213110e+00
1.69574171e-01 -4.93643522e-01 6.39115227e-03 -1.75626838e+00
3.67920727e-01 -8.65177453e-01 -6.33455396e-01 4.13503915e-01
-7.12572813e-01 1.61188230e-01 1.26934722e-01 -1.14879590e-02
-8.22960019e-01 4.86528218e-01 1.63052094e+00 1.19439485e-02
-1.36347055e-01 -2.25826368e-01 -8.37520957e-01 1.05082488e+00
4.84826177e-01 -2.86228538e-01 -5.87938428e-01 -8.45686018e-01
-8.30486268e-02 -1.18895635e-01 3.32376480e-01 -1.03441560e+00
9.18572098e-02 -4.57949907e-01 5.74384868e-01 -8.34649146e-01
1.73079446e-02 -6.46532357e-01 -2.65901059e-01 1.02804877e-01
-2.73852050e-01 1.22979499e-01 3.20912451e-01 5.98046362e-01
-4.44725454e-01 -4.33195569e-02 8.11073422e-01 -3.30243409e-01
-1.26497817e+00 4.25369591e-01 2.56167144e-01 2.38278598e-01
9.33293283e-01 -4.58832756e-02 -5.74402511e-01 -3.69421579e-02
-5.49690008e-01 3.60782892e-01 1.23144016e-01 6.71383560e-01
5.75677156e-01 -1.16566324e+00 -4.99485791e-01 3.48890185e-01
1.67601973e-01 8.32902730e-01 6.23380899e-01 4.84755069e-01
-5.38891494e-01 2.71756560e-01 -2.55995691e-01 -9.28010404e-01
-1.11475682e+00 6.59263909e-01 2.20639944e-01 -5.72064519e-01
-7.79748142e-01 1.25712061e+00 1.22429574e+00 -5.47073066e-01
3.82624492e-02 -6.54687583e-01 -4.71978813e-01 -3.22133482e-01
5.84257007e-01 -1.85873151e-01 -2.53579617e-01 -8.82155657e-01
-4.98698413e-01 1.12554002e+00 1.16535807e-02 5.55517972e-01
1.32859087e+00 -2.43540391e-01 -2.17186362e-01 1.17532291e-01
1.24979615e+00 1.30888134e-01 -1.67410374e+00 -2.00247481e-01
-2.41045415e-01 -5.40497303e-01 -1.07695535e-01 -8.06633592e-01
-1.30539691e+00 1.32749212e+00 5.46069443e-01 -2.24021766e-02
1.48326170e+00 1.85603052e-01 8.03263187e-01 1.56594560e-01
2.65188247e-01 -7.86882341e-01 3.50512266e-01 8.23682189e-01
8.44694197e-01 -1.32302332e+00 -6.45742118e-02 -8.48758459e-01
-7.04992712e-01 1.07610893e+00 1.12727058e+00 -2.50678539e-01
6.67982280e-01 4.58888292e-01 1.21203288e-01 -2.17839271e-01
-4.27583456e-01 -4.97307181e-01 3.09208363e-01 8.08932722e-01
4.95991141e-01 1.38198227e-01 -7.82428607e-02 5.92536569e-01
7.22272694e-02 -4.41500932e-01 7.55593330e-02 7.35228479e-01
-4.47850198e-01 -1.07945180e+00 -1.35520354e-01 1.47377893e-01
-3.71217549e-01 -2.84968734e-01 -1.70561299e-01 9.55846548e-01
6.88326955e-01 8.73059154e-01 1.90474436e-01 -2.76832044e-01
3.94304305e-01 -4.07866269e-01 2.96836764e-01 -5.38556814e-01
-4.18350726e-01 3.26753557e-01 -1.64567038e-01 -7.19741881e-01
-4.37927246e-01 -1.76671296e-01 -1.46163023e+00 7.00862035e-02
-3.04803520e-01 -2.44498000e-01 4.48661506e-01 9.49378133e-01
3.55985433e-01 1.01485944e+00 3.93420994e-01 -6.49865448e-01
-4.95122671e-02 -9.77308095e-01 -3.79212767e-01 7.38616288e-01
2.09601402e-01 -6.57332301e-01 -4.22351249e-02 -2.66231984e-01]
|
[9.585773468017578, 0.36203017830848694]
|
d77bb34c-47aa-423e-ac2c-1db1e5d2468b
|
are-neural-operators-really-neural-operators
|
2305.19913
| null |
https://arxiv.org/abs/2305.19913v1
|
https://arxiv.org/pdf/2305.19913v1.pdf
|
Are Neural Operators Really Neural Operators? Frame Theory Meets Operator Learning
|
Recently, there has been significant interest in operator learning, i.e. learning mappings between infinite-dimensional function spaces. This has been particularly relevant in the context of learning partial differential equations from data. However, it has been observed that proposed models may not behave as operators when implemented on a computer, questioning the very essence of what operator learning should be. We contend that in addition to defining the operator at the continuous level, some form of continuous-discrete equivalence is necessary for an architecture to genuinely learn the underlying operator, rather than just discretizations of it. To this end, we propose to employ frames, a concept in applied harmonic analysis and signal processing that gives rise to exact and stable discrete representations of continuous signals. Extending these concepts to operators, we introduce a unifying mathematical framework of Representation equivalent Neural Operator (ReNO) to ensure operations at the continuous and discrete level are equivalent. Lack of this equivalence is quantified in terms of aliasing errors. We analyze various existing operator learning architectures to determine whether they fall within this framework, and highlight implications when they fail to do so.
|
['Rima Alaifari', 'Siddhartha Mishra', 'Roberto Molinaro', 'Bogdan Raonić', 'Emmanuel de Bézenac', 'Francesca Bartolucci']
|
2023-05-31
| null | null | null | null |
['operator-learning']
|
['miscellaneous']
|
[ 5.81163287e-01 3.35106343e-01 9.04379264e-02 -2.57315785e-01
-2.28684247e-01 -6.74616516e-01 3.14919353e-01 2.25942001e-01
-2.79501110e-01 5.74002743e-01 -1.34781178e-03 -5.61439037e-01
-5.66635013e-01 -8.17857623e-01 -7.18225300e-01 -5.74461877e-01
-2.90473044e-01 6.65395707e-02 7.89084136e-02 -2.76937038e-01
1.06024347e-01 7.06492901e-01 -1.63932955e+00 1.97257757e-01
9.19295371e-01 1.03609991e+00 -3.18489611e-01 7.38106489e-01
-1.12940073e-01 8.66420209e-01 -4.37742203e-01 -2.17324436e-01
4.18613225e-01 -8.06264877e-01 -1.02750373e+00 -5.27298152e-02
3.71211499e-01 5.60068078e-02 -1.48918703e-01 1.23968971e+00
1.71333209e-01 1.68884307e-01 9.59640861e-01 -1.33614445e+00
-8.35110784e-01 4.67038751e-01 9.84031036e-02 2.36321732e-01
3.72023225e-01 -2.39710689e-01 1.17344189e+00 -7.23503768e-01
4.53966230e-01 1.03807414e+00 1.03963459e+00 3.39251906e-01
-1.47620034e+00 -2.06159994e-01 -2.75786757e-01 -9.42560360e-02
-1.41860437e+00 -1.76198214e-01 8.96127343e-01 -5.71002245e-01
6.80039406e-01 3.34668994e-01 7.16728270e-01 7.32581556e-01
6.17398024e-01 6.32969856e-01 9.43915129e-01 -8.82764757e-01
4.14995313e-01 2.50474811e-01 9.95733663e-02 7.20901966e-01
1.10752918e-01 1.69387653e-01 -4.25062031e-01 -4.63161571e-03
1.06826985e+00 -2.79555529e-01 -6.47001028e-01 -7.92401433e-01
-1.03551817e+00 1.10769796e+00 2.67937094e-01 6.04155540e-01
-4.35272872e-01 7.88831934e-02 5.79797149e-01 8.19170475e-01
2.56884158e-01 7.08713710e-01 -2.00135514e-01 -1.51845649e-01
-7.09044933e-01 2.19118133e-01 9.85859096e-01 5.92402518e-01
6.85265481e-01 2.65871733e-01 1.33587629e-01 3.11110318e-01
3.56508195e-02 -2.81557851e-02 4.16576356e-01 -1.23046029e+00
-1.53537586e-01 3.95826817e-01 -1.47768958e-02 -1.10082448e+00
-3.95636052e-01 -3.72969240e-01 -8.95773113e-01 5.41468263e-01
5.51792800e-01 2.91416962e-02 -2.89276868e-01 2.07731009e+00
-3.38889211e-02 3.96715969e-01 1.97367862e-01 6.60955787e-01
-6.10493217e-03 5.32089591e-01 -2.36167014e-01 -5.55673182e-01
7.68477976e-01 -3.12191784e-01 -9.63689029e-01 4.78253901e-01
6.36933744e-01 -4.31617945e-01 1.28914285e+00 6.27878070e-01
-1.31535852e+00 -8.04956913e-01 -1.40045738e+00 -9.06517059e-02
-5.61861157e-01 -3.03988755e-01 7.53111243e-01 6.95754409e-01
-1.16140187e+00 9.85612929e-01 -7.58913457e-01 -3.34692508e-01
7.51708597e-02 2.94589877e-01 -1.01154499e-01 6.44112527e-01
-1.48863244e+00 1.08559752e+00 4.38302338e-01 2.00661197e-01
-3.02092075e-01 -9.85334158e-01 -8.82321298e-01 1.45592630e-01
-1.90707281e-01 -4.86143380e-01 1.32871044e+00 -1.25941217e+00
-1.45785785e+00 8.18432808e-01 1.02598459e-01 -9.28494513e-01
5.56886911e-01 -9.12204608e-02 -6.38279676e-01 2.26878047e-01
-1.50658354e-01 3.17159116e-01 9.39267755e-01 -1.03403914e+00
-5.39756179e-01 -1.28290489e-01 3.44842255e-01 -7.27214664e-02
-4.98742133e-01 -3.00454438e-01 6.02071643e-01 -7.18594074e-01
2.76800245e-01 -6.88552499e-01 1.17496334e-01 4.35479581e-01
2.68833667e-01 -4.54961240e-01 9.11911666e-01 -3.12893510e-01
1.32751524e+00 -2.21214509e+00 2.89138168e-01 2.92212069e-01
1.25026673e-01 1.43688560e-01 3.47707838e-01 5.48468053e-01
-5.22731483e-01 -6.33478537e-02 -5.51406384e-01 6.72862306e-02
3.45850646e-01 3.93248439e-01 -8.56165469e-01 8.35026503e-01
4.79068130e-01 7.24788129e-01 -8.48562658e-01 -3.14313978e-01
3.04894894e-01 6.35780394e-01 -7.38688529e-01 -3.11260913e-02
-1.69281930e-01 5.14352798e-01 -1.89592317e-01 1.67716742e-01
4.30133641e-01 5.53284101e-02 -8.60399157e-02 -2.69520938e-01
-2.49218315e-01 2.10085824e-01 -1.65824366e+00 1.73348248e+00
-4.67290163e-01 8.13795626e-01 1.13230698e-01 -1.69299257e+00
8.43239129e-01 6.28088415e-01 8.29032540e-01 -4.84542787e-01
5.42533845e-02 5.54998338e-01 1.59621298e-01 -4.52005714e-01
2.46902600e-01 -5.56945443e-01 -1.79076269e-02 5.38527548e-01
1.79551169e-01 -3.37067097e-01 1.72146574e-01 -2.10486293e-01
8.73934507e-01 1.71981111e-01 4.48713869e-01 -8.20068538e-01
8.43539894e-01 -2.43802518e-01 1.92232966e-01 6.37125611e-01
-1.90607682e-01 5.64796090e-01 4.82191384e-01 -5.65137863e-01
-8.18657160e-01 -1.46953714e+00 -6.37209654e-01 8.03074360e-01
-8.56169686e-02 -9.95453298e-02 -7.55919337e-01 -2.19529837e-01
-5.69267310e-02 6.47890091e-01 -8.18803430e-01 -4.51033652e-01
-6.21057868e-01 -3.06842297e-01 8.02311122e-01 5.13677716e-01
2.98464954e-01 -1.12045729e+00 -1.21083212e+00 2.82961309e-01
2.48300627e-01 -7.72428572e-01 -3.43104005e-01 5.93271732e-01
-8.70496273e-01 -1.08488870e+00 -6.44869685e-01 -8.83458972e-01
4.01091427e-01 -2.83123761e-01 1.10699463e+00 -1.02551118e-01
-1.43962160e-01 7.91667879e-01 -2.41004273e-01 -5.10871053e-01
-6.06832623e-01 5.96303446e-03 3.85801882e-01 2.64045179e-01
3.30833733e-01 -9.57183540e-01 -9.39912722e-02 -4.19759080e-02
-1.37654531e+00 -3.59882772e-01 2.60193467e-01 7.92794466e-01
2.96472549e-01 4.24283355e-01 7.07441568e-01 -6.73272789e-01
8.41407537e-01 -1.26916155e-01 -3.51380497e-01 2.56672084e-01
-4.67350781e-01 2.91986793e-01 8.20180357e-01 -4.74125355e-01
-7.12452292e-01 1.15338909e-02 -1.61252558e-01 -5.02715588e-01
2.71699149e-02 5.86868823e-01 -2.32351236e-02 -3.59000295e-01
1.04660106e+00 1.75919384e-01 2.72562623e-01 -1.08661056e-01
3.23799521e-01 4.91673380e-01 7.05112755e-01 -7.09273517e-01
6.52815223e-01 4.65722114e-01 3.38102460e-01 -1.11283422e+00
-6.72758281e-01 -2.11817533e-01 -8.90747964e-01 3.20597850e-02
8.49816799e-01 -2.66325474e-01 -7.57861376e-01 2.20121056e-01
-1.17845917e+00 -2.60811716e-01 -1.03031433e+00 4.78127033e-01
-1.09629977e+00 3.53180617e-01 -5.54539502e-01 -1.08141422e+00
1.30101934e-01 -9.29525077e-01 5.52708387e-01 -7.25008398e-02
-4.90367979e-01 -1.42867422e+00 3.02057825e-02 -5.96657395e-01
4.62957054e-01 4.79358554e-01 1.02084911e+00 -4.33535874e-01
-1.91124100e-02 -1.05747968e-01 1.42269418e-01 6.98865056e-01
2.47814894e-01 -1.39883295e-01 -9.67220306e-01 -2.83477515e-01
6.95304632e-01 -3.30452889e-01 6.25009060e-01 2.62836754e-01
1.07228303e+00 -2.74599399e-02 1.30055472e-01 5.59831679e-01
1.33518028e+00 2.19709948e-01 4.32717621e-01 -1.90960780e-01
2.69890577e-01 6.92722678e-01 -1.87287778e-02 1.86334759e-01
-9.29992571e-02 5.82333684e-01 1.35916606e-01 -3.56474798e-03
6.74813092e-02 -2.19546974e-01 4.50274825e-01 8.77973318e-01
1.20821700e-01 1.94580674e-01 -7.12511897e-01 3.44262123e-01
-1.59627652e+00 -1.05414653e+00 -1.24934174e-01 2.17269087e+00
7.85671651e-01 4.39732373e-01 1.70637727e-01 8.19516659e-01
4.48989660e-01 1.07545540e-01 -4.88575608e-01 -8.18993390e-01
-2.22629428e-01 6.29260719e-01 1.52417302e-01 7.72186100e-01
-9.84694362e-01 3.37595701e-01 6.85143185e+00 4.57941502e-01
-1.31571162e+00 -1.22924216e-01 1.24673285e-01 7.30597615e-01
-3.14926893e-01 -8.42580572e-02 -7.83820450e-02 1.87261194e-01
9.45093453e-01 -3.86944830e-01 4.90927309e-01 6.04015052e-01
5.47335185e-02 2.88783610e-01 -1.61715758e+00 8.21358979e-01
1.47322714e-02 -1.16394043e+00 1.39055982e-01 -2.65056547e-02
5.59907079e-01 -7.11625218e-01 4.75345701e-02 3.59504610e-01
-2.22229436e-01 -1.31021953e+00 7.54985988e-01 7.40230978e-01
5.05315065e-01 -7.26651609e-01 4.95189309e-01 3.89387369e-01
-1.28834510e+00 3.65818702e-02 -2.32472137e-01 -5.73882461e-01
2.21030146e-01 3.37515265e-01 -3.64518315e-01 6.58397019e-01
2.65332490e-01 6.53659701e-01 -2.40096390e-01 7.83282280e-01
1.68798849e-01 4.29081798e-01 -2.63344646e-01 2.27759764e-01
1.55882254e-01 -3.37900251e-01 2.87704647e-01 1.18450594e+00
3.15077305e-01 8.17885101e-02 7.39637241e-02 1.17387033e+00
3.32120985e-01 5.49077354e-02 -7.77798831e-01 -1.78769693e-01
1.50198191e-01 6.95503950e-01 -5.95859945e-01 -4.40375991e-02
-6.87826872e-01 1.06767344e+00 8.72796327e-02 4.23778653e-01
-8.35939884e-01 -5.70632219e-01 5.26614547e-01 2.31447127e-02
3.28409448e-02 -3.80976379e-01 -4.26773101e-01 -9.53967512e-01
1.26865357e-01 -7.37966478e-01 3.89841616e-01 -4.28510517e-01
-1.28359616e+00 3.22680086e-01 2.78519429e-02 -1.33271575e+00
-4.26716000e-01 -9.98049080e-01 -5.35867393e-01 9.17293131e-01
-1.13038993e+00 -7.20999122e-01 4.90100496e-02 6.44379973e-01
1.98755816e-01 1.97438672e-01 1.15782249e+00 2.29788929e-01
1.14829175e-01 5.23284972e-01 -2.35805884e-01 9.23171639e-02
2.41853312e-01 -1.60151589e+00 -2.73694266e-02 7.07441092e-01
6.90843046e-01 6.35661364e-01 1.00017655e+00 -9.91870537e-02
-1.31989479e+00 -7.26280808e-01 8.82863283e-01 -5.08050740e-01
7.46118784e-01 -3.40673745e-01 -1.28935111e+00 6.84695303e-01
9.37567279e-02 2.24765122e-01 4.52542782e-01 -2.32516110e-01
-1.15778565e-01 -7.12666363e-02 -1.00061691e+00 4.46119994e-01
9.85311627e-01 -9.79157031e-01 -8.70367467e-01 2.31123134e-01
7.30751157e-01 -3.33688408e-01 -1.03445673e+00 5.56885421e-01
5.23937166e-01 -1.20392656e+00 9.64025497e-01 -7.67213821e-01
1.45644516e-01 -3.42382580e-01 -3.82437021e-01 -1.07956707e+00
-1.73638821e-01 -7.22455204e-01 -4.26009178e-01 8.00348520e-01
2.37337992e-01 -7.88016856e-01 4.77547944e-01 3.66531998e-01
-3.90922785e-01 -8.61874342e-01 -1.09845841e+00 -1.03214622e+00
7.31660724e-01 -7.18081653e-01 1.68476373e-01 1.08550215e+00
3.45146954e-01 1.59789309e-01 -1.47617251e-01 2.33835071e-01
2.89884686e-01 6.36261376e-03 3.10288996e-01 -1.51834512e+00
-4.08364564e-01 -7.85366237e-01 -7.13060439e-01 -1.01181734e+00
5.16395330e-01 -1.07791054e+00 -1.88187107e-01 -1.13482594e+00
-5.98333001e-01 -2.09029645e-01 -5.45754492e-01 -9.33400691e-02
3.59417498e-01 1.88377827e-01 -1.01246156e-01 1.31956026e-01
-1.84643447e-01 5.82452953e-01 1.02389419e+00 1.94673851e-01
-1.30021721e-01 1.49604201e-01 -4.73452181e-01 9.30218220e-01
7.14774430e-01 2.18132380e-02 -6.75681710e-01 -2.23099366e-01
2.52549112e-01 -7.41251633e-02 6.09261155e-01 -1.44972730e+00
4.21451181e-01 9.31273475e-02 1.77716210e-01 -9.58519503e-02
2.36554950e-01 -9.69335020e-01 9.54443440e-02 6.70569539e-01
-7.18676925e-01 3.28856438e-01 6.24174625e-02 4.42194998e-01
-3.59966367e-01 -4.88644958e-01 9.54428792e-01 -6.58538267e-02
-6.64805889e-01 -1.32544888e-02 -2.65166819e-01 2.70515323e-01
8.62348437e-01 -4.03632402e-01 4.43866581e-01 -4.20909524e-01
-7.95870245e-01 -2.65591651e-01 3.87440443e-01 2.64862448e-01
3.19876701e-01 -1.38285029e+00 -4.67179179e-01 6.12196684e-01
-1.22666411e-01 -2.03821078e-01 -5.90272024e-02 9.20336008e-01
-5.62188625e-01 5.56894302e-01 -1.97581515e-01 -5.81371248e-01
-5.36264420e-01 6.47734642e-01 7.11008847e-01 5.45712449e-02
-7.67333567e-01 6.21479809e-01 -2.31510159e-02 -3.90471458e-01
4.89089280e-01 -8.60495090e-01 1.71267733e-01 1.54880602e-02
3.00220698e-01 2.41299853e-01 1.00023210e-01 -5.55174947e-01
-1.52832463e-01 5.81457555e-01 5.64033806e-01 -3.48902941e-01
9.68410492e-01 1.88879654e-01 -2.07447439e-01 1.15411747e+00
1.37645030e+00 -1.72013845e-02 -1.05140638e+00 -1.45240530e-01
3.68490696e-01 -1.17003143e-01 -1.06203981e-01 -2.90783376e-01
-6.09129488e-01 1.05192304e+00 6.26757026e-01 1.14132547e+00
1.20251191e+00 -2.58010775e-01 5.86103976e-01 3.64379942e-01
3.37303787e-01 -1.12993765e+00 5.64346202e-02 5.57959199e-01
1.09483302e+00 -8.08435440e-01 -2.79711425e-01 -3.57389688e-01
-1.47711992e-01 1.43571460e+00 1.35088235e-01 -5.96698165e-01
9.78600860e-01 3.11204702e-01 -1.87982991e-01 -1.31723064e-04
-2.87482679e-01 -6.06212094e-02 3.21249843e-01 7.33890533e-01
8.01745772e-01 -2.33837739e-01 -4.47575808e-01 3.69532317e-01
-4.47975993e-01 1.93903804e-01 4.60418165e-01 9.67853785e-01
-5.08145511e-01 -9.38283443e-01 -2.67441094e-01 1.49104193e-01
-3.76811713e-01 2.36192316e-01 -1.28248796e-01 9.53972220e-01
2.51134455e-01 5.95541596e-01 2.29556367e-01 -5.58392294e-02
3.19110811e-01 4.76988107e-01 5.86849868e-01 -4.36913788e-01
-2.34871358e-01 -4.33897346e-01 -3.07919502e-01 -3.09832871e-01
-6.87613606e-01 -6.27240658e-01 -1.30985999e+00 1.38195353e-02
2.50006430e-02 4.10193950e-01 1.86703026e-01 9.67158794e-01
-2.90242378e-02 6.04794145e-01 4.66172904e-01 -6.16113901e-01
-1.10785341e+00 -6.91708207e-01 -7.28088677e-01 4.70005184e-01
6.42813087e-01 -7.16399789e-01 -5.91665626e-01 1.43613458e-01]
|
[7.58817195892334, 3.6802151203155518]
|
d98b5758-bc76-4b8f-b213-032cb9527350
|
deep-structural-causal-models-for-tractable
|
2006.06485
| null |
https://arxiv.org/abs/2006.06485v2
|
https://arxiv.org/pdf/2006.06485v2.pdf
|
Deep Structural Causal Models for Tractable Counterfactual Inference
|
We formulate a general framework for building structural causal models (SCMs) with deep learning components. The proposed approach employs normalising flows and variational inference to enable tractable inference of exogenous noise variables - a crucial step for counterfactual inference that is missing from existing deep causal learning methods. Our framework is validated on a synthetic dataset built on MNIST as well as on a real-world medical dataset of brain MRI scans. Our experimental results indicate that we can successfully train deep SCMs that are capable of all three levels of Pearl's ladder of causation: association, intervention, and counterfactuals, giving rise to a powerful new approach for answering causal questions in imaging applications and beyond. The code for all our experiments is available at https://github.com/biomedia-mira/deepscm.
|
['Daniel C. Castro', 'Nick Pawlowski', 'Ben Glocker']
|
2020-06-11
| null |
http://proceedings.neurips.cc/paper/2020/hash/0987b8b338d6c90bbedd8631bc499221-Abstract.html
|
http://proceedings.neurips.cc/paper/2020/file/0987b8b338d6c90bbedd8631bc499221-Paper.pdf
|
neurips-2020-12
|
['normalising-flows', 'counterfactual-inference']
|
['methodology', 'miscellaneous']
|
[ 4.13593620e-01 4.23136026e-01 -3.67068172e-01 -4.22821045e-01
-5.94442010e-01 -1.15501493e-01 1.06366503e+00 -6.37920992e-03
-2.78985471e-01 1.25466323e+00 6.37679875e-01 -8.32515538e-01
-5.79023480e-01 -8.33707094e-01 -1.06480789e+00 -7.97375977e-01
-5.62938392e-01 3.48303318e-01 -1.97871830e-02 8.91379938e-02
1.04951508e-01 2.32071057e-01 -9.59560394e-01 3.51967007e-01
8.74160945e-01 5.35544455e-01 4.79635224e-02 5.69652438e-01
1.10350221e-01 1.19003141e+00 -1.41127944e-01 -6.28680110e-01
-1.87691867e-01 -4.52895671e-01 -1.03888071e+00 -5.65118313e-01
1.15876734e-01 -4.36651766e-01 -6.13352716e-01 1.02999783e+00
4.37338233e-01 -1.90459803e-01 8.39960396e-01 -1.46725261e+00
-7.48466790e-01 1.28520894e+00 -5.65855265e-01 6.02991819e-01
1.20018326e-01 2.57891119e-01 1.01352835e+00 -5.01399517e-01
6.23715341e-01 1.51368904e+00 5.61800122e-01 5.76199353e-01
-1.28398538e+00 -8.39170575e-01 2.22172290e-02 3.80176872e-01
-6.39892399e-01 -3.91883940e-01 6.42450511e-01 -6.08084917e-01
5.17612755e-01 2.18444556e-01 5.10601401e-01 1.87953770e+00
6.92989707e-01 7.88313568e-01 1.32578802e+00 -8.80751982e-02
3.77978265e-01 -5.07033050e-01 2.38538817e-01 5.56603909e-01
3.89668137e-01 6.66412830e-01 -5.62684178e-01 -4.63735640e-01
1.15484309e+00 -1.22693062e-01 -2.31819585e-01 -2.38395765e-01
-1.36898744e+00 1.13741648e+00 6.74490809e-01 7.29950368e-02
-6.98636889e-01 7.69653678e-01 4.51367229e-01 -2.20835041e-02
3.56832772e-01 1.84544176e-01 -4.58562434e-01 3.46428156e-01
-8.42867434e-01 4.85027462e-01 5.17861605e-01 4.83674109e-01
-7.04288622e-03 -7.26808012e-02 -4.69744354e-01 4.42132622e-01
3.55856091e-01 4.26556438e-01 1.86738178e-01 -1.04109371e+00
2.27054507e-01 9.22278985e-02 1.25151426e-01 -6.87589765e-01
-6.80819809e-01 -4.78709131e-01 -1.06892633e+00 -4.29029614e-02
3.00169468e-01 -5.34176707e-01 -7.71185219e-01 1.99115932e+00
4.39457536e-01 9.78928328e-01 -7.13045225e-02 6.84758961e-01
9.91160631e-01 1.62766919e-01 4.55917805e-01 -3.85319650e-01
1.54940474e+00 -4.35268313e-01 -1.02274644e+00 1.43789966e-02
3.35136324e-01 -3.68924201e-01 8.56282890e-01 2.51590461e-01
-9.38426614e-01 -1.49726674e-01 -8.72882187e-01 2.64138039e-02
3.08492128e-02 -3.93140882e-01 1.20300543e+00 5.85581541e-01
-1.04211175e+00 5.92130780e-01 -1.03196955e+00 4.04704362e-02
1.06632614e+00 1.40537784e-01 -2.40539461e-01 -5.92746623e-02
-1.75505769e+00 6.83802843e-01 4.11314368e-01 2.57895499e-01
-1.70810413e+00 -1.29103220e+00 -7.21636295e-01 2.59283278e-02
3.73803079e-01 -1.35017145e+00 1.32040310e+00 -5.50430179e-01
-9.93318737e-01 4.83403057e-01 -6.94477707e-02 -9.04969275e-01
8.94167066e-01 -2.36806706e-01 -2.90761709e-01 5.51441163e-02
1.82160780e-01 5.11095583e-01 6.43920660e-01 -1.12951493e+00
-4.20368463e-01 -4.67291117e-01 1.12320855e-01 -1.83000043e-01
2.85526693e-01 -3.18754725e-02 4.30205405e-01 -6.95127308e-01
-4.29131955e-01 -5.10725319e-01 -4.47471708e-01 -2.76520878e-01
-7.69613087e-01 -3.11125278e-01 3.39479387e-01 -5.47451675e-01
9.32544529e-01 -1.55493784e+00 6.79019764e-02 -1.01528920e-01
4.81918693e-01 -1.71374515e-01 -2.42480692e-02 2.62802631e-01
-6.99210942e-01 3.15984786e-01 -6.29226863e-01 -5.81481867e-02
-1.25146121e-01 1.27592772e-01 -4.05374259e-01 5.27106345e-01
2.84528315e-01 1.24285007e+00 -1.26876986e+00 -5.54630399e-01
3.06226730e-01 4.49811041e-01 -6.30412519e-01 3.56516875e-02
-3.33381921e-01 6.79821610e-01 -4.19856697e-01 1.68870792e-01
6.56614661e-01 -1.87721699e-01 4.27081704e-01 7.85933882e-02
3.82834971e-02 4.59651232e-01 -1.00293958e+00 1.58036661e+00
-3.21301520e-01 2.65771091e-01 -3.09330136e-01 -1.32455063e+00
2.84687877e-01 5.99181473e-01 4.85658288e-01 -3.94244045e-01
2.13330448e-01 -7.56447688e-02 3.11787665e-01 -7.40503609e-01
-2.70851910e-01 -5.79497457e-01 4.74037603e-02 4.19973105e-01
3.68200719e-01 1.63011789e-01 4.89383601e-02 3.64240378e-01
1.27906656e+00 1.21717453e-01 4.14961547e-01 -6.59694314e-01
3.45753580e-01 -1.14182197e-01 5.47975838e-01 1.05610240e+00
-1.51103243e-01 1.71448454e-01 1.15354848e+00 -3.58746648e-01
-8.65156054e-01 -1.57014608e+00 -5.30833483e-01 6.58178568e-01
-3.52474153e-01 1.23984724e-01 -7.81953990e-01 -6.22024059e-01
-4.88233007e-02 1.20996857e+00 -1.12346482e+00 -8.52427483e-02
-4.08795536e-01 -1.30646932e+00 6.28723979e-01 4.50892389e-01
2.87484854e-01 -1.34776282e+00 -6.76595807e-01 1.98345616e-01
-5.34520626e-01 -7.77461827e-01 2.65692417e-02 4.38105781e-03
-9.74123359e-01 -1.50259614e+00 -3.53690147e-01 -1.78555131e-01
1.79948762e-01 -4.10007566e-01 1.14869964e+00 -9.11512971e-02
-4.29161966e-01 -7.07468390e-02 1.02831446e-01 -7.14235604e-01
-5.51222444e-01 -3.91612679e-01 1.24808652e-02 -1.47766232e-01
8.28017592e-02 -9.50345993e-01 -1.00746632e+00 -2.99239606e-01
-8.08784127e-01 1.66013837e-01 4.86934930e-01 1.05026901e+00
2.80019522e-01 -1.03177547e-01 1.03447819e+00 -1.32152426e+00
7.20343232e-01 -1.08591616e+00 -6.33292258e-01 4.92857769e-02
-5.42839229e-01 1.54692382e-01 2.84000576e-01 -1.86694920e-01
-1.58321738e+00 -3.00630033e-01 -4.54254657e-01 -1.00904666e-01
-3.42586339e-01 6.62230074e-01 -4.39087562e-05 8.24161887e-01
7.20661938e-01 -1.18050836e-01 -6.63056523e-02 -1.97556362e-01
7.17117608e-01 1.87433347e-01 7.48030961e-01 -6.00602627e-01
3.86414796e-01 8.42793047e-01 1.97859496e-01 -4.26215768e-01
-8.47626209e-01 7.09216893e-02 -6.29561543e-01 -2.53960520e-01
9.91190434e-01 -7.93367267e-01 -1.12853324e+00 2.31617063e-01
-1.14961481e+00 -6.24576092e-01 -7.62630329e-02 5.97829700e-01
-7.31498599e-01 5.66092320e-02 -7.79333115e-01 -6.22574747e-01
-4.09960598e-01 -1.19595397e+00 8.04789066e-01 4.08154726e-02
-2.86301553e-01 -1.49125111e+00 2.34475225e-01 2.39595726e-01
1.36510715e-01 7.42359519e-01 1.26724589e+00 -2.63547182e-01
-3.83708507e-01 2.85484642e-01 -1.80854157e-01 -1.27221629e-01
1.10914428e-02 4.89533953e-02 -9.47075784e-01 1.54119596e-01
1.16339341e-01 -1.87036797e-01 1.19268477e+00 1.18252885e+00
1.21570182e+00 -4.70336199e-01 -5.70950747e-01 3.33581686e-01
1.37194252e+00 1.39673859e-01 8.12811792e-01 9.70667452e-02
5.86902499e-01 5.76973498e-01 2.74776638e-01 4.79485571e-01
5.26128113e-01 1.78848431e-01 7.12082446e-01 -1.51619270e-01
-1.21072359e-01 -3.92659962e-01 7.14613944e-02 1.54861107e-01
-6.17092066e-02 -5.76430000e-02 -1.11561871e+00 9.06108141e-01
-2.07112622e+00 -1.26718533e+00 -7.37755001e-01 1.72626770e+00
1.09941208e+00 1.39766783e-01 5.13203070e-02 -1.62519887e-01
5.33303916e-01 8.72413963e-02 -5.78899682e-01 -3.75515342e-01
1.86521605e-01 4.75704730e-01 4.29054409e-01 6.44369781e-01
-1.20507050e+00 7.21702874e-01 6.58005571e+00 5.93223453e-01
-7.13504910e-01 5.85670769e-01 8.26225758e-01 -8.18826109e-02
-6.52307212e-01 -6.86067194e-02 -2.04790816e-01 3.82968456e-01
1.22959912e+00 -3.42219710e-01 1.72159538e-01 3.96026790e-01
7.41883934e-01 -4.69147600e-02 -1.27677631e+00 4.54111725e-01
-6.47756934e-01 -1.86722088e+00 9.83414203e-02 -6.14905246e-02
7.61575460e-01 2.48636380e-01 4.07610051e-02 -4.53085685e-03
1.10238492e+00 -1.30683923e+00 6.73214853e-01 6.82129741e-01
7.10659683e-01 -6.23974502e-01 8.34968269e-01 2.71167308e-01
-4.13072228e-01 -7.73460418e-02 -2.45961323e-01 -1.59042701e-01
3.11192840e-01 9.41564083e-01 -8.32596719e-01 8.43912244e-01
6.73747063e-01 6.35271907e-01 -6.63937181e-02 7.45667934e-01
-8.24603558e-01 1.03003716e+00 6.60843998e-02 2.50622332e-01
7.62296021e-02 2.14362442e-01 3.18982214e-01 1.14714682e+00
-4.89758924e-02 2.43104815e-01 -4.25415397e-01 1.23992407e+00
-1.65405095e-01 -3.02008688e-01 -6.95242584e-01 2.33606949e-01
4.75696117e-01 1.10378587e+00 -6.80225849e-01 -2.91023970e-01
-1.65828884e-01 4.26422209e-01 3.15604031e-01 2.99878746e-01
-1.28522968e+00 3.90250266e-01 6.88836277e-01 -1.43096551e-01
-1.11257054e-01 2.54037231e-01 -6.97623312e-01 -1.13877463e+00
-4.78613347e-01 -6.30561769e-01 7.30557024e-01 -6.57328010e-01
-1.42433202e+00 1.94086641e-01 3.95752281e-01 -4.40028787e-01
-2.77134895e-01 -4.47929442e-01 -8.90203595e-01 7.49858320e-01
-1.41883850e+00 -1.13770604e+00 1.85030296e-01 7.86626697e-01
3.22039932e-01 2.72613555e-01 7.60330975e-01 1.88447312e-01
-6.15951419e-01 2.09298134e-01 -1.91785365e-01 1.47211745e-01
4.16978687e-01 -1.38195801e+00 4.72102404e-01 1.01102126e+00
3.04245036e-02 8.78287971e-01 9.45703328e-01 -7.98497856e-01
-8.88516009e-01 -1.02976716e+00 5.79907179e-01 -5.91211140e-01
1.12680864e+00 -4.04722661e-01 -6.48483038e-01 1.00769532e+00
5.50089896e-01 -2.35504493e-01 6.78870261e-01 3.73878241e-01
-1.78785443e-01 1.69182852e-01 -1.00621867e+00 6.32830739e-01
1.28726470e+00 -1.15102969e-01 -9.47005510e-01 6.06534421e-01
9.21883702e-01 -2.18853086e-01 -1.02475452e+00 4.72896069e-01
4.94391382e-01 -9.44106162e-01 1.09700012e+00 -9.33647692e-01
1.29185092e+00 -3.84352024e-04 2.34717727e-01 -1.46044624e+00
-3.77804607e-01 -4.14756864e-01 -1.05602898e-01 9.33824420e-01
4.10018384e-01 -7.01603293e-01 3.18720698e-01 4.17162389e-01
4.69026305e-02 -7.30439067e-01 -1.16155195e+00 -2.06910253e-01
5.22687256e-01 -8.70560944e-01 7.35825360e-01 1.31940782e+00
-1.96189523e-01 3.16165030e-01 -2.47876480e-01 3.60866398e-01
1.23545837e+00 -1.19926430e-01 2.48551548e-01 -1.14967275e+00
-1.57309234e-01 -5.43596864e-01 1.15571171e-01 3.36523168e-02
4.50717300e-01 -1.01961005e+00 -3.55107278e-01 -1.77962434e+00
6.66830063e-01 -2.07561970e-01 -5.09098887e-01 5.39202154e-01
-3.56326789e-01 -3.57088335e-02 -9.57635045e-02 -1.43342093e-01
4.13543824e-03 4.73392636e-01 1.24459136e+00 2.33872924e-02
2.88861424e-01 -5.72967529e-03 -8.52385759e-01 8.13240170e-01
8.68613660e-01 -8.32929909e-01 -6.94861710e-01 -2.43108347e-01
1.53380960e-01 5.56101859e-01 1.06376719e+00 -3.69949877e-01
2.08527949e-02 -3.25699717e-01 4.28753436e-01 -3.49143326e-01
-2.76642978e-01 -3.85995954e-01 3.63904387e-01 8.78248751e-01
-6.47751570e-01 -1.89521402e-01 2.39531875e-01 6.14429355e-01
1.51287511e-01 1.48937721e-02 6.17162406e-01 -1.77952215e-01
-5.14021456e-01 3.40110868e-01 -2.32526779e-01 9.15761143e-02
8.41033220e-01 6.84071243e-01 -6.82089508e-01 -2.42268264e-01
-9.71494317e-01 5.12860596e-01 -3.33718151e-01 4.64395851e-01
6.78514540e-01 -1.28044331e+00 -1.17924690e+00 -1.58992872e-01
-2.44566277e-01 -2.43861467e-01 6.34357512e-01 1.16238427e+00
-3.93995553e-01 5.17417669e-01 -3.45615566e-01 -4.30891573e-01
-7.53495753e-01 8.17161381e-01 4.87037212e-01 -4.84840214e-01
-6.35046184e-01 8.09595585e-01 6.92570031e-01 -4.45035309e-01
-1.10630356e-01 -3.24935198e-01 -2.32129455e-01 -5.05977869e-02
5.49527764e-01 3.06050658e-01 -2.85606235e-01 -8.97457004e-02
-4.68493700e-01 -4.50049549e-01 7.95445517e-02 -2.72375584e-01
1.61837971e+00 3.73124741e-02 -4.33876574e-01 5.14470756e-01
7.34746635e-01 -5.55417299e-01 -1.14885700e+00 7.66927153e-02
2.02474803e-01 -6.37744144e-02 2.35967249e-01 -1.05777144e+00
-9.03877079e-01 9.41477656e-01 5.74274182e-01 1.54568125e-02
9.55135465e-01 2.11662650e-01 2.52215952e-01 -1.36203453e-01
2.77587056e-01 -5.05217195e-01 -2.64219999e-01 -5.68780899e-02
1.16356599e+00 -1.16585159e+00 -3.83722000e-02 -1.82458952e-01
-4.44626361e-01 7.09006608e-01 2.07668215e-01 -3.67766619e-01
9.79074657e-01 5.54161072e-01 -2.25796312e-01 -5.75121224e-01
-1.11667788e+00 3.16324160e-02 8.98270402e-03 4.91223872e-01
6.93278015e-01 6.12516582e-01 -6.32754564e-01 7.47872472e-01
-2.94071764e-01 4.07296449e-01 7.15868831e-01 2.72267580e-01
3.64643514e-01 -7.43870258e-01 -2.27707788e-01 6.54404759e-01
-7.87385464e-01 -4.37136889e-01 1.16124749e-02 1.08141971e+00
-1.24391941e-02 8.51311803e-01 6.80106059e-02 7.08753541e-02
6.74889162e-02 5.23406565e-02 5.61826766e-01 -3.12304229e-01
-1.63075402e-01 1.33502871e-01 1.50794312e-01 -7.78004944e-01
-7.77065635e-01 -1.05305266e+00 -1.35748577e+00 -3.91381681e-01
6.01093993e-02 -1.91268012e-01 3.12637299e-01 9.85428929e-01
2.06671506e-02 1.10766983e+00 2.58163989e-01 -4.11422879e-01
-3.57883036e-01 -1.14784455e+00 -3.13487560e-01 3.60161871e-01
5.15266061e-01 -9.46634769e-01 -1.32656053e-01 1.62873313e-01]
|
[8.09609603881836, 5.404637336730957]
|
e696039e-bb52-4a69-94d9-262fc7254e78
|
evaluating-word-embeddings-in-extremely-under
| null | null |
https://aclanthology.org/2022.coling-1.393
|
https://aclanthology.org/2022.coling-1.393.pdf
|
Evaluating Word Embeddings in Extremely Under-Resourced Languages: A Case Study in Bribri
|
Word embeddings are critical for numerous NLP tasks but their evaluation in actual under-resourced settings needs further examination. This paper presents a case study in Bribri, a Chibchan language from Costa Rica. Four experiments were adapted from English: Word similarities, WordSim353 correlations, odd-one-out tasks and analogies. Here we discuss their adaptation to an under-resourced Indigenous language and we use them to measure semantic and morphological learning. We trained 96 word2vec models with different hyperparameter combinations. The best models for this under-resourced scenario were Skip-grams with an intermediate size (100 dimensions) and large window sizes (10). These had an average correlation of r=0.28 with WordSim353, a 76% accuracy in semantic odd-one-out and 70% accuracy in structural/morphological odd-one-out. The performance was lower for the analogies: The best models could find the appropriate semantic target amongst the first 25 results approximately 60% of the times, but could only find the morphological/structural target 11% of the times. Future research needs to further explore the patterns of morphological/structural learning, to examine the behavior of deep learning embeddings, and to establish a human baseline. This project seeks to improve Bribri NLP and ultimately help in its maintenance and revitalization.
|
['Rolando Coto-Solano']
| null | null | null | null |
coling-2022-10
|
['odd-one-out']
|
['reasoning']
|
[-2.23317280e-01 -1.69268072e-01 -1.99829862e-01 -1.54918686e-01
-8.98800910e-01 -9.00570214e-01 7.24734008e-01 3.31637174e-01
-1.18645239e+00 2.72452265e-01 6.12532675e-01 -8.04533601e-01
-2.51714379e-01 -7.38621473e-01 -4.95320074e-02 -5.19336283e-01
-2.95905560e-01 5.77117503e-01 8.57558772e-02 -5.43963075e-01
5.46173155e-01 5.71543217e-01 -1.18327796e+00 1.65325537e-01
6.60027385e-01 4.94305074e-01 3.61373574e-01 8.16547155e-01
-3.47708046e-01 1.88433647e-01 -7.06786990e-01 -3.76455128e-01
3.76821756e-01 -1.70809142e-02 -8.38457227e-01 -3.60838562e-01
7.02972472e-01 -3.10580730e-02 -2.50768751e-01 8.09090614e-01
9.11333263e-01 5.19379318e-01 8.03623259e-01 -6.70287907e-01
-1.23883414e+00 4.89237159e-01 -2.90541619e-01 7.59889722e-01
4.32438195e-01 1.73570037e-01 1.49546850e+00 -1.20694149e+00
5.93780577e-01 1.58242345e+00 7.93171585e-01 3.44963312e-01
-1.01411736e+00 -8.49296451e-01 -1.01789072e-01 2.80958474e-01
-1.45229757e+00 -3.83525878e-01 2.55041220e-03 -2.21404552e-01
1.67067981e+00 -3.91332246e-02 5.36112428e-01 1.01025355e+00
1.22019961e-01 4.09627855e-01 1.14733946e+00 -5.82615733e-01
-6.45368248e-02 3.88061255e-01 2.06336588e-01 4.03651744e-01
5.17166018e-01 2.10333671e-02 -3.43962908e-01 -1.96125537e-01
5.95982194e-01 -3.63297284e-01 -1.01575926e-01 2.60731876e-01
-1.03921413e+00 1.24550438e+00 3.66572946e-01 9.95490193e-01
-8.06916654e-02 -9.53482911e-02 4.04704660e-01 5.09875834e-01
5.48361063e-01 1.01890624e+00 -8.55630100e-01 -2.73940295e-01
-7.35863805e-01 3.36008847e-01 8.33057463e-01 7.53454447e-01
5.92841566e-01 3.00206751e-01 1.60011396e-01 1.26190472e+00
-4.64074500e-02 4.37947392e-01 1.01541412e+00 -5.68660557e-01
7.08641589e-01 2.48481974e-01 -1.33676335e-01 -1.15072083e+00
-7.03092456e-01 -2.91537493e-01 -1.41807556e-01 -7.64887109e-02
7.32003450e-01 -2.37504497e-01 -1.05272245e+00 1.48789501e+00
-1.27071008e-01 -3.51703107e-01 1.56450167e-01 6.05548084e-01
6.20317817e-01 9.34176683e-01 4.26749259e-01 3.08873206e-01
1.66478705e+00 -6.09305203e-01 -4.12100315e-01 -5.81168652e-01
1.09107208e+00 -1.09563482e+00 1.55491805e+00 3.08216602e-01
-8.63348067e-01 -6.61877811e-01 -1.09540987e+00 -2.87575662e-01
-1.13366866e+00 -1.08139262e-01 4.88351434e-01 1.00762355e+00
-1.14165437e+00 5.05536437e-01 -5.92321992e-01 -9.68504727e-01
1.40034631e-01 1.97790310e-01 -5.32279491e-01 -3.10262263e-01
-1.33838665e+00 1.60543513e+00 6.53783977e-01 -4.69153285e-01
-2.93530822e-01 -8.26959431e-01 -1.04696476e+00 -2.52696406e-02
-1.90205604e-01 -9.38591510e-02 8.32177877e-01 -6.88917994e-01
-8.48992229e-01 1.07238567e+00 -5.46447113e-02 -2.49325112e-01
-2.60639563e-02 -2.01102287e-01 -7.94152260e-01 -2.96131819e-02
2.73995340e-01 7.03173220e-01 1.46844074e-01 -8.95299435e-01
-6.65534973e-01 -4.23204035e-01 -6.21333942e-02 3.52877766e-01
-8.24983716e-01 3.81246120e-01 -1.53330892e-01 -8.82555842e-01
-4.12419215e-02 -8.54333818e-01 -1.59675106e-02 -4.96156722e-01
3.04702491e-01 -4.12658334e-01 3.13319504e-01 -1.02650881e+00
1.39165390e+00 -2.10935760e+00 -2.49334380e-01 1.32269427e-01
-2.18063429e-01 4.55803126e-01 -6.33861959e-01 8.75814617e-01
-1.74358636e-01 6.11858189e-01 1.66048601e-01 1.41830882e-02
2.63571497e-02 2.77539909e-01 -4.16969135e-02 5.68996131e-01
2.95017779e-01 7.12072074e-01 -1.03369606e+00 -2.96602845e-01
9.24823955e-02 5.45148790e-01 -4.62555826e-01 -1.28499597e-01
3.64576578e-01 -4.26924974e-01 1.45379499e-01 6.97049558e-01
4.72470134e-01 2.96644807e-01 3.27358186e-01 -7.56797800e-03
-2.44317263e-01 4.65886354e-01 -8.85922730e-01 1.40932202e+00
-8.70441318e-01 1.06820357e+00 -1.64214000e-01 -6.71807885e-01
1.00548315e+00 1.20989479e-01 9.53518376e-02 -1.00019932e+00
5.44704087e-02 6.60223007e-01 5.80944121e-01 -6.37470484e-01
9.12417114e-01 -3.93851936e-01 -4.48607244e-02 3.49872261e-01
5.27297147e-02 -2.91695267e-01 2.91737795e-01 5.27700856e-02
1.13749778e+00 -2.40815565e-01 4.83445764e-01 -6.74315035e-01
1.55169621e-01 2.71014839e-01 1.60112679e-01 5.51157534e-01
-1.99558407e-01 6.64176047e-01 3.25126797e-01 -2.94457108e-01
-1.21407616e+00 -1.05514693e+00 -3.99412304e-01 1.57926357e+00
-2.71127254e-01 -2.34969378e-01 -3.81859064e-01 -5.94243169e-01
3.34221572e-02 1.34534800e+00 -7.29520738e-01 -1.77459419e-01
-6.80808604e-01 -8.83418083e-01 8.03691506e-01 6.18253171e-01
-2.71088984e-02 -1.17009950e+00 -4.91169035e-01 2.64490366e-01
1.59732968e-01 -8.81942570e-01 -3.86936963e-01 4.18735772e-01
-6.70439303e-01 -9.50676620e-01 -8.80570710e-01 -1.14628375e+00
3.62014562e-01 1.58912912e-01 9.96973097e-01 3.39705534e-02
-3.26922834e-01 1.40344322e-01 -6.58820808e-01 -3.26348394e-01
-1.61057875e-01 2.82399029e-01 2.16031417e-01 -8.07133019e-01
8.26156020e-01 -2.67457873e-01 -3.46773922e-01 4.81418073e-02
-1.02994359e+00 -7.09852457e-01 6.02285922e-01 9.80594456e-01
2.66697928e-02 -7.88672715e-02 6.37553692e-01 -7.97787607e-01
9.90201950e-01 -7.87385523e-01 -1.90877374e-02 1.78398639e-01
-6.95916951e-01 -9.74914804e-02 6.60425663e-01 -6.99691713e-01
-7.88117051e-01 -6.67897522e-01 -3.93517405e-01 1.50748584e-02
-2.33842671e-04 5.30060828e-01 7.39032403e-02 1.80924997e-01
9.04850721e-01 -1.02046803e-01 -1.78580135e-01 -5.57277977e-01
5.03827929e-01 8.94257903e-01 8.13917071e-02 -6.28949106e-01
7.17826128e-01 5.34003042e-02 -6.57497942e-01 -1.33683586e+00
-5.52366674e-01 -7.21585870e-01 -4.89469439e-01 4.17593718e-01
7.96638489e-01 -8.26521873e-01 -8.54468867e-02 1.28188506e-01
-1.15508258e+00 -4.26977813e-01 -1.01856828e-01 7.55963385e-01
-2.63342224e-02 1.44189268e-01 -5.03247857e-01 -4.97799695e-01
-3.34970593e-01 -1.01152706e+00 7.49995768e-01 -1.26672024e-02
-8.05016100e-01 -1.46361744e+00 2.76014566e-01 1.56952351e-01
6.87229455e-01 -2.50789851e-01 1.46938896e+00 -1.58803594e+00
5.23716807e-01 -1.77529573e-01 -4.32165682e-01 4.87331539e-01
4.13799375e-01 -1.57618776e-01 -8.15230608e-01 -5.55126607e-01
-3.10985237e-01 -3.74415845e-01 6.38527572e-01 2.85787061e-02
5.82740307e-01 -1.79351524e-01 -5.75423762e-02 3.84599566e-01
1.62445462e+00 5.32907426e-01 5.88810921e-01 7.97280610e-01
5.00650287e-01 7.11240470e-01 4.19566631e-01 8.35518837e-02
2.46809661e-01 3.86031061e-01 -1.23392530e-01 1.40039891e-01
-3.57877910e-01 -1.65841684e-01 5.79201758e-01 1.05514085e+00
2.52745986e-01 -3.96116763e-01 -1.34432864e+00 1.09503973e+00
-9.83961761e-01 -7.17717290e-01 -3.95355970e-02 1.97582293e+00
7.67282009e-01 1.34691328e-01 -1.86493210e-02 -3.02359872e-02
4.36059028e-01 3.03397655e-01 -1.70630157e-01 -1.06619048e+00
-2.59231776e-01 9.53463495e-01 6.83450162e-01 6.64004207e-01
-7.66907692e-01 1.31983995e+00 6.57715845e+00 1.14037383e+00
-8.70616674e-01 2.30859578e-01 3.94453883e-01 -9.40148607e-02
-4.37589884e-01 -6.59836307e-02 -9.30112004e-01 2.82613128e-01
1.26890373e+00 -1.44755304e-01 3.31029385e-01 5.78422785e-01
6.82160705e-02 -9.70352292e-02 -8.59324992e-01 6.87814355e-01
4.45795894e-01 -1.00858402e+00 1.85063615e-01 9.63490158e-02
6.93365037e-01 3.54491085e-01 8.72414485e-02 5.77435076e-01
3.93018007e-01 -1.34004998e+00 4.24905956e-01 -1.75517574e-01
9.30337608e-01 -9.20993209e-01 8.60076308e-01 -5.43660969e-02
-1.10427225e+00 -1.03220657e-01 -7.10575163e-01 -2.13245809e-01
-7.30653405e-02 8.05119574e-02 -9.10387337e-01 2.14084700e-01
7.18378365e-01 4.18227881e-01 -8.12233746e-01 7.82221198e-01
-3.76994982e-02 5.37994206e-01 -3.77742618e-01 -5.55331826e-01
8.12788188e-01 -2.98391581e-02 2.81304926e-01 1.81847680e+00
5.03523052e-01 -5.56545518e-02 -6.90693408e-02 3.22118521e-01
2.01196074e-01 4.92470711e-01 -7.44514823e-01 -2.95968741e-01
7.95516491e-01 1.02433729e+00 -6.71183765e-01 -3.13012898e-01
-5.47378719e-01 8.02024305e-01 4.97134745e-01 3.56077880e-01
-6.18326306e-01 -1.03438973e+00 9.83772218e-01 1.00548990e-01
2.94466734e-01 -4.02233303e-01 -3.96779060e-01 -7.90414929e-01
-3.68335426e-01 -1.04461050e+00 4.39470202e-01 -5.95562518e-01
-1.64175487e+00 6.25116825e-01 2.30290562e-01 -5.88537931e-01
1.47996033e-02 -1.23775053e+00 -5.68285823e-01 1.12312055e+00
-1.24889266e+00 -9.34283793e-01 1.88733011e-01 2.74139136e-01
8.45072985e-01 -3.31953585e-01 1.03643548e+00 2.81900138e-01
-3.39243829e-01 8.81097317e-01 2.05433711e-01 3.26646179e-01
7.62985289e-01 -1.31879127e+00 6.35633528e-01 5.73510170e-01
4.09683943e-01 9.81131732e-01 5.50239682e-01 -5.76468349e-01
-1.08161449e+00 -7.82109141e-01 1.44203591e+00 -3.44177425e-01
1.28825200e+00 -1.69019580e-01 -9.50748503e-01 6.34866059e-01
5.71174145e-01 -5.44764459e-01 7.78331101e-01 3.11526835e-01
-5.90069115e-01 2.02285379e-01 -1.12064147e+00 8.49785507e-01
9.55001533e-01 -6.93428516e-01 -1.03298962e+00 4.52546716e-01
7.39627421e-01 1.32538145e-02 -9.64273334e-01 -4.90476703e-03
6.89456940e-01 -5.94930112e-01 9.79892373e-01 -7.25430667e-01
2.99286097e-01 2.60497659e-01 -5.08419991e-01 -1.62330770e+00
-5.03876388e-01 -1.43830135e-01 5.68296492e-01 1.31246758e+00
8.43705177e-01 -8.30077589e-01 5.24276197e-01 3.44383031e-01
-1.35079131e-01 -8.95198643e-01 -8.46096218e-01 -1.14644563e+00
8.74470353e-01 -7.45752037e-01 2.12712482e-01 1.20433474e+00
-7.41776973e-02 3.06578606e-01 2.54382193e-01 -9.74877477e-02
3.57146151e-02 -5.87099552e-01 2.16286406e-01 -7.44017720e-01
-1.36560192e-02 -7.31546581e-01 -4.83956069e-01 -6.01145566e-01
2.51749635e-01 -1.09821796e+00 -2.00123027e-01 -1.60862482e+00
-1.47911608e-01 -5.60486615e-01 -1.44527361e-01 5.31531096e-01
-1.48253754e-01 2.36465842e-01 2.32394516e-01 -9.88757461e-02
1.98027000e-01 1.26579106e-01 7.24067092e-01 -7.31502175e-02
-2.54903167e-01 -5.77521026e-01 -9.65048075e-01 6.42088056e-01
1.14860487e+00 -4.67328608e-01 -3.41638744e-01 -8.20361674e-01
9.93325114e-02 -6.81658804e-01 -1.27744675e-01 -9.00924861e-01
2.79391836e-02 -1.98987588e-01 5.00598133e-01 -2.15993404e-01
4.04657871e-01 -6.67771876e-01 -3.00710291e-01 4.61796165e-01
-3.42269868e-01 9.19175804e-01 5.36193311e-01 1.04513355e-01
4.77109104e-02 -6.81206584e-01 7.07168639e-01 -3.65318090e-01
-9.95139182e-01 -1.04794152e-01 -6.53012276e-01 5.83137333e-01
8.75359535e-01 -3.69732708e-01 -3.51908416e-01 -3.51507738e-02
-4.89316851e-01 6.98726922e-02 4.67160702e-01 7.20132649e-01
6.90294504e-01 -1.22649574e+00 -7.40657091e-01 2.10021749e-01
2.52637535e-01 -6.51840687e-01 -2.03011140e-01 4.80816990e-01
-9.19748485e-01 4.74948734e-01 -1.43924460e-01 1.56712368e-01
-9.50373113e-01 2.67154813e-01 1.13895424e-01 -2.31323883e-01
-3.31374109e-01 1.07501996e+00 -1.95516527e-01 -5.79625845e-01
1.55021474e-01 -1.74602136e-01 -2.82130986e-01 7.61319160e-01
5.77430546e-01 6.08780861e-01 1.40508562e-01 -7.25506902e-01
-5.43953419e-01 6.37945890e-01 -1.17021315e-01 -5.16773343e-01
1.23469877e+00 6.32368177e-02 -3.75041552e-02 6.23960257e-01
1.67791212e+00 2.25176066e-01 -3.08727592e-01 1.39936162e-02
3.02992284e-01 -5.18656373e-01 -3.12062688e-02 -6.71111166e-01
-7.27997184e-01 9.36475277e-01 6.07754052e-01 1.16708830e-01
5.01955807e-01 -1.86617345e-01 9.50658321e-01 4.67134565e-01
1.32987395e-01 -1.39672387e+00 8.65286868e-03 7.99021423e-01
7.75982022e-01 -8.81124496e-01 1.22172825e-01 2.96786875e-01
-6.99090838e-01 1.18439817e+00 4.84382153e-01 -3.11067253e-01
7.01204538e-01 -4.24394347e-02 1.65306568e-01 -2.36125484e-01
-5.62075317e-01 -3.69668663e-01 7.99967572e-02 8.15942824e-01
8.21660280e-01 1.65358275e-01 -6.05851948e-01 3.62385869e-01
-5.72654486e-01 -7.18125463e-01 2.83374786e-01 6.46444738e-01
-5.96410394e-01 -1.05488312e+00 -3.45623165e-01 6.04741931e-01
-6.25901341e-01 -7.01723695e-01 -4.63463336e-01 1.42428184e+00
1.49645105e-01 1.06043792e+00 4.95514482e-01 -4.35525566e-01
4.46906030e-01 4.77823377e-01 3.64906162e-01 -1.00434101e+00
-1.00167346e+00 -1.63253039e-01 5.36474824e-01 -2.38074422e-01
9.84594971e-03 -7.83008456e-01 -1.26583767e+00 -6.33253634e-01
-2.05910146e-01 8.69565457e-03 7.05637693e-01 7.28486419e-01
6.08015172e-02 1.16424836e-01 1.94459260e-01 -7.13652730e-01
-6.88722253e-01 -1.37843716e+00 -6.83136523e-01 4.57368046e-01
-1.17604755e-01 -4.86535192e-01 -6.31782174e-01 -4.55498278e-01]
|
[10.755245208740234, 9.598169326782227]
|
32da43d1-8b7b-432a-b62c-9bd906374635
|
high-resolution-gan-inversion-for-degraded
|
2302.03406
| null |
https://arxiv.org/abs/2302.03406v1
|
https://arxiv.org/pdf/2302.03406v1.pdf
|
High-Resolution GAN Inversion for Degraded Images in Large Diverse Datasets
|
The last decades are marked by massive and diverse image data, which shows increasingly high resolution and quality. However, some images we obtained may be corrupted, affecting the perception and the application of downstream tasks. A generic method for generating a high-quality image from the degraded one is in demand. In this paper, we present a novel GAN inversion framework that utilizes the powerful generative ability of StyleGAN-XL for this problem. To ease the inversion challenge with StyleGAN-XL, Clustering \& Regularize Inversion (CRI) is proposed. Specifically, the latent space is firstly divided into finer-grained sub-spaces by clustering. Instead of initializing the inversion with the average latent vector, we approximate a centroid latent vector from the clusters, which generates an image close to the input image. Then, an offset with a regularization term is introduced to keep the inverted latent vector within a certain range. We validate our CRI scheme on multiple restoration tasks (i.e., inpainting, colorization, and super-resolution) of complex natural images, and show preferable quantitative and qualitative results. We further demonstrate our technique is robust in terms of data and different GAN models. To our best knowledge, we are the first to adopt StyleGAN-XL for generating high-quality natural images from diverse degraded inputs. Code is available at https://github.com/Booooooooooo/CRI.
|
['Yuan Xie', 'Zhizhong Zhang', 'Ying Tai', 'Donghao Luo', 'Chuming Lin', 'Yanbo Wang']
|
2023-02-07
| null | null | null | null |
['colorization']
|
['computer-vision']
|
[ 5.79363465e-01 -1.36945531e-01 1.71165302e-01 -7.68406466e-02
-8.30475092e-01 -4.66847956e-01 3.92826289e-01 -6.97514534e-01
-8.01861808e-02 8.77051115e-01 3.08343828e-01 6.98676407e-02
1.04080759e-01 -7.54017770e-01 -7.49262810e-01 -9.96407628e-01
4.17567760e-01 2.05833297e-02 -3.07656407e-01 -1.47828609e-01
4.63047363e-02 2.36519292e-01 -1.40331995e+00 2.85027474e-01
1.18649411e+00 7.52340913e-01 4.67259467e-01 5.30470848e-01
6.32286221e-02 6.55185401e-01 -6.79212272e-01 -2.71972507e-01
4.06520844e-01 -9.45730686e-01 -4.94410932e-01 4.79924172e-01
3.51966083e-01 -4.07936543e-01 -2.09557176e-01 1.13274395e+00
4.21898514e-01 9.65995565e-02 5.13598442e-01 -1.17055130e+00
-9.46305037e-01 3.15033793e-01 -8.59213293e-01 -1.85236931e-01
2.52529442e-01 2.57837534e-01 6.94495201e-01 -8.30648124e-01
7.39560246e-01 1.14716911e+00 2.87990123e-01 6.04862452e-01
-1.40668380e+00 -8.07666540e-01 -2.01209467e-02 1.40332475e-01
-1.41247582e+00 -6.48640275e-01 1.01997328e+00 -2.78137535e-01
1.35258853e-01 4.01579410e-01 5.43928564e-01 1.14065504e+00
1.26920836e-02 6.55526102e-01 1.39899516e+00 -4.40197438e-01
2.25489393e-01 -1.22998908e-01 -4.87289369e-01 3.85956675e-01
-8.58598296e-03 3.43975388e-02 -5.11950016e-01 1.74607813e-01
9.09817576e-01 6.68822005e-02 -6.13188446e-01 -1.55619815e-01
-1.21775997e+00 6.68617010e-01 5.38192391e-01 2.20494390e-01
-5.63165605e-01 1.93851218e-01 -5.90791479e-02 1.33793309e-01
4.96612281e-01 4.30080324e-01 1.10205896e-01 -5.88261336e-02
-1.12594211e+00 -1.38565224e-06 2.82955945e-01 9.38572049e-01
8.60885739e-01 2.56608158e-01 -1.67179018e-01 1.06836772e+00
8.86708945e-02 5.97003162e-01 3.14572811e-01 -1.36098504e+00
3.30967516e-01 2.44928345e-01 1.95945442e-01 -1.02037823e+00
2.92666286e-01 -4.38190222e-01 -1.36164129e+00 4.12583768e-01
1.68751433e-01 -7.76958913e-02 -1.08428800e+00 1.72280335e+00
1.77119970e-01 1.69001982e-01 7.94515833e-02 1.08437765e+00
5.16947806e-01 1.02369201e+00 -2.64674872e-01 -3.01887691e-01
1.15346313e+00 -9.69249547e-01 -9.62643921e-01 -3.33624870e-01
-4.72333170e-02 -9.14510608e-01 1.28576362e+00 6.05112791e-01
-1.26336205e+00 -6.63431525e-01 -9.92753744e-01 -1.70649111e-01
1.51002482e-01 2.56393313e-01 2.88814932e-01 3.60548794e-01
-1.14940059e+00 3.49019855e-01 -7.03841567e-01 -1.75745055e-01
5.56886077e-01 -1.57477558e-01 -3.76650035e-01 -6.57302082e-01
-9.01612401e-01 4.40820158e-01 2.89282084e-01 2.89696187e-01
-8.04744482e-01 -5.61652839e-01 -7.26441681e-01 -2.24151030e-01
3.04129750e-01 -7.07582891e-01 7.70244002e-01 -1.24864793e+00
-1.66627002e+00 6.46029294e-01 -2.23588899e-01 -1.81404520e-02
5.77198088e-01 -1.88458294e-01 -2.79123545e-01 1.51568994e-01
2.83977151e-01 6.62740886e-01 1.25398648e+00 -1.75050139e+00
-3.35350692e-01 -7.39977509e-02 -1.48080945e-01 3.07343990e-01
-2.79329240e-01 -1.19935304e-01 -8.38027000e-01 -8.82319748e-01
2.03028798e-01 -8.71169984e-01 -9.37275738e-02 -5.49449511e-02
-5.58791459e-01 4.16112393e-01 8.25250447e-01 -9.73285556e-01
1.04251516e+00 -2.43503952e+00 5.31827211e-01 1.28255099e-01
3.44958544e-01 9.93267521e-02 -1.64659768e-01 3.59756351e-01
-8.37800801e-02 1.81650832e-01 -7.74870753e-01 -5.36698163e-01
-1.82249442e-01 3.42177927e-01 -4.03596818e-01 4.16948736e-01
1.80898622e-01 8.06136727e-01 -9.05882537e-01 -3.63890707e-01
2.50383288e-01 6.98638022e-01 -5.26671290e-01 4.96273458e-01
-8.97922590e-02 7.97102869e-01 -1.72528595e-01 7.52847493e-01
9.73428667e-01 -1.83617726e-01 8.11109245e-02 -5.17009318e-01
-1.07941441e-01 -3.28269333e-01 -1.08310688e+00 1.86929011e+00
-5.47856152e-01 6.26501799e-01 2.56144911e-01 -7.01581240e-01
8.72213662e-01 9.59150046e-02 3.92554790e-01 -7.82764733e-01
-4.34365943e-02 1.59634560e-01 -1.81271225e-01 -2.79022694e-01
4.88325924e-01 -2.57173121e-01 2.82654781e-02 4.04777706e-01
-2.22923681e-01 -3.15294921e-01 2.78674752e-01 2.71174729e-01
7.07821310e-01 3.67767215e-01 -5.87140061e-02 8.85281414e-02
3.63743484e-01 -1.67293459e-01 5.70354939e-01 4.96889651e-01
1.72386199e-01 1.26383972e+00 4.07130688e-01 1.80135258e-02
-1.34445262e+00 -1.29178166e+00 1.15693338e-01 5.92138469e-01
2.30217859e-01 -2.90281683e-01 -8.76647472e-01 -2.73969859e-01
-3.50907147e-01 6.91212833e-01 -5.30831277e-01 -1.44692481e-01
-5.58416367e-01 -7.67597258e-01 2.90887713e-01 2.44949877e-01
7.44466364e-01 -1.18756914e+00 -2.79296517e-01 1.52227402e-01
-6.69149637e-01 -1.02304423e+00 -7.10310996e-01 -2.50798017e-01
-6.33939922e-01 -8.17938268e-01 -9.82622445e-01 -6.17478549e-01
9.96012092e-01 3.97471458e-01 1.03949928e+00 9.29681957e-03
-1.82266444e-01 1.07471772e-01 -5.40073156e-01 1.04212165e-01
-4.35372710e-01 -1.75711215e-01 -1.71122938e-01 3.59031320e-01
-3.51694882e-01 -6.98986650e-01 -8.48844528e-01 2.62311995e-01
-1.38479578e+00 5.72268844e-01 8.36607754e-01 1.03562510e+00
8.25371087e-01 2.89748520e-01 2.23684728e-01 -7.63317704e-01
5.98271251e-01 -3.54915768e-01 -4.12088096e-01 3.73135880e-03
-4.38989341e-01 -1.90775946e-03 7.34683454e-01 -4.01019096e-01
-1.17396617e+00 -6.68408796e-02 -1.08973058e-02 -6.04882658e-01
-1.13901556e-01 3.78392249e-01 -4.74621654e-01 9.51470658e-02
4.78136450e-01 4.84688908e-01 1.75577581e-01 -4.74808365e-01
6.77136481e-01 6.57022476e-01 7.99035907e-01 -4.92217869e-01
1.07065213e+00 6.52030945e-01 -3.21425676e-01 -7.65002072e-01
-6.64432108e-01 -5.48384450e-02 -3.41063172e-01 -3.46884578e-01
8.82852793e-01 -9.65775847e-01 -1.92803413e-01 8.02359700e-01
-9.23825145e-01 -5.67947268e-01 -3.92259240e-01 2.33716667e-01
-4.95302886e-01 3.54842037e-01 -7.03558147e-01 -5.69578290e-01
-3.26021075e-01 -1.12617898e+00 1.01620233e+00 2.70980209e-01
1.21462539e-01 -6.17348611e-01 -3.66783217e-02 5.18401980e-01
4.71491307e-01 4.82227176e-01 5.39446056e-01 5.12346327e-01
-6.67105556e-01 4.61685285e-02 -3.61493319e-01 6.71831429e-01
4.40166920e-01 4.99462038e-02 -8.28211784e-01 -4.96048987e-01
4.59383056e-02 -2.27211133e-01 8.69703650e-01 4.46410537e-01
1.28550291e+00 -3.13923329e-01 1.30259413e-02 9.68962908e-01
1.55151451e+00 2.17689082e-01 1.07604074e+00 3.55670214e-01
9.53105330e-01 4.01185453e-01 5.58137596e-01 3.40553969e-01
2.02634051e-01 6.85656428e-01 4.06983435e-01 -5.57738960e-01
-4.20116931e-01 -2.46702135e-01 4.85856950e-01 7.86257923e-01
-2.37079307e-01 -4.60040838e-01 -6.52559042e-01 5.41354060e-01
-1.52357125e+00 -9.48372960e-01 1.61180809e-01 2.06683230e+00
8.78796875e-01 -1.65907085e-01 -1.45842344e-01 1.65390223e-01
8.44240785e-01 2.87005633e-01 -5.43312073e-01 -1.05869360e-01
-3.38713318e-01 1.99449331e-01 2.99947351e-01 5.62195241e-01
-7.11699784e-01 8.82981420e-01 5.25299263e+00 9.64643419e-01
-1.22527421e+00 1.34805068e-01 8.68326247e-01 -1.26487374e-01
-5.78281939e-01 4.78123836e-02 -1.94808200e-01 7.47285187e-01
4.69409645e-01 -9.64163058e-03 8.12266469e-01 3.43365192e-01
3.43264580e-01 -1.18666157e-01 -5.72329402e-01 1.05062127e+00
1.75655678e-01 -1.22064054e+00 1.85121760e-01 1.69361502e-01
9.83043611e-01 -3.08522910e-01 2.72623569e-01 -1.47435606e-01
3.09652776e-01 -1.09412861e+00 7.75388420e-01 6.40039980e-01
1.23945343e+00 -8.33096623e-01 5.04283011e-01 1.07564464e-01
-9.72545326e-01 1.08381040e-01 -2.52147347e-01 2.92501271e-01
4.25283074e-01 8.67533803e-01 -2.72652477e-01 8.32371414e-01
7.70588100e-01 9.41762447e-01 -4.34426963e-01 7.03376353e-01
-5.13877392e-01 5.44241548e-01 -6.20006733e-02 7.46708333e-01
-1.14354650e-02 -7.36293018e-01 4.47044879e-01 8.95519078e-01
6.76426530e-01 1.69552982e-01 -6.09149598e-02 1.08452296e+00
-1.24238402e-01 -1.51005924e-01 -4.97478783e-01 6.63600340e-02
2.73435682e-01 1.38028038e+00 -5.28418779e-01 -2.88418114e-01
-1.91747651e-01 1.46768069e+00 8.38150233e-02 6.50603056e-01
-9.23758030e-01 -2.39308536e-01 5.63249946e-01 1.32467151e-01
3.02286178e-01 -2.55193025e-01 -2.60841191e-01 -1.32348979e+00
1.66406035e-01 -1.17483974e+00 9.57010016e-02 -1.20687044e+00
-1.24106145e+00 9.05082464e-01 -1.28900051e-01 -1.52525496e+00
-1.81132436e-01 -1.00522682e-01 -5.39434373e-01 9.56044674e-01
-1.56181192e+00 -1.26319444e+00 -7.37444162e-01 5.95469654e-01
3.78585607e-01 3.77215743e-02 4.99724925e-01 3.50196213e-01
-7.48367369e-01 3.70332509e-01 4.09754723e-01 -3.45032774e-02
8.83600235e-01 -1.10511768e+00 2.46881470e-01 1.26561964e+00
-1.26618207e-01 4.18687224e-01 6.70478404e-01 -5.41994870e-01
-1.35870564e+00 -1.21647620e+00 3.10073406e-01 3.56620364e-03
2.68713355e-01 -3.71332109e-01 -9.09756303e-01 5.22129238e-01
4.39752281e-01 -5.76074980e-02 4.62944120e-01 -4.45238173e-01
-2.22235858e-01 -3.40772152e-01 -1.15614343e+00 7.97629356e-01
1.01726460e+00 -5.12723386e-01 -3.73875424e-02 1.21432021e-01
6.40846550e-01 -4.26905692e-01 -8.40722799e-01 2.36121997e-01
4.25278544e-01 -1.17750382e+00 1.01005125e+00 1.10351458e-01
8.48369241e-01 -7.50966966e-01 -1.48850739e-01 -1.60898364e+00
-3.44865113e-01 -6.31421149e-01 1.82535462e-02 1.38108134e+00
7.69281909e-02 -5.43790400e-01 6.32475555e-01 3.28466535e-01
-1.67760000e-01 -4.97754723e-01 -6.67860985e-01 -5.64943254e-01
-9.24830288e-02 -1.81421548e-01 5.76739967e-01 7.93933809e-01
-5.65122724e-01 7.03345686e-02 -9.07913923e-01 2.15842038e-01
1.01212287e+00 4.09378976e-01 9.01228964e-01 -5.99402905e-01
-3.61809283e-01 -2.34871969e-01 -1.00031853e-01 -1.05866754e+00
-9.59957615e-02 -5.79975605e-01 1.91776872e-01 -1.53031290e+00
1.61800399e-01 -3.62420857e-01 -1.27205908e-01 3.69637668e-01
-2.64471531e-01 8.55252147e-01 3.57433349e-01 5.11422276e-01
-2.34764174e-01 7.83259213e-01 1.63888991e+00 -1.54948309e-01
-1.38427943e-01 -3.12680840e-01 -9.22932565e-01 3.82458359e-01
1.05714786e+00 -3.76206547e-01 -4.55003649e-01 -6.34342074e-01
7.06183687e-02 8.66560712e-02 4.35216665e-01 -9.81025100e-01
-4.24570739e-02 -2.42804199e-01 4.93484974e-01 -3.20104092e-01
4.63420421e-01 -6.95612013e-01 7.35395610e-01 3.44572067e-01
-7.04410598e-02 -6.29462004e-02 -2.47526821e-02 4.46253121e-01
-4.75956291e-01 1.01035923e-01 9.64592397e-01 -1.59138799e-01
-5.78607023e-01 3.52772772e-01 -1.30682364e-01 -8.17675591e-02
8.05664182e-01 -2.68704563e-01 -2.17596173e-01 -6.47040844e-01
-5.49259663e-01 -2.51355977e-03 9.88774657e-01 2.62822002e-01
8.25346529e-01 -1.51612902e+00 -1.00466979e+00 4.66829509e-01
1.06077895e-01 2.78318077e-01 5.28808475e-01 6.83215499e-01
-6.11590326e-01 -2.41704077e-01 -4.11667138e-01 -5.75520575e-01
-1.04670310e+00 4.93650645e-01 9.52003151e-02 -9.95492861e-02
-7.28581786e-01 5.66829801e-01 4.08918530e-01 -8.08888674e-02
-1.94825038e-01 -8.53386223e-02 4.36487608e-02 -2.31701997e-03
5.08847296e-01 3.10046196e-01 -2.46735096e-01 -7.24905193e-01
-5.26709668e-02 6.40536308e-01 7.94195682e-02 -2.93680340e-01
1.43107760e+00 -4.39549237e-01 -3.85614097e-01 2.20370814e-01
1.10408139e+00 2.00369403e-01 -1.61688054e+00 -1.20456241e-01
-5.69454014e-01 -9.13088679e-01 1.05800644e-01 -6.31722033e-01
-1.54812551e+00 6.37452245e-01 6.43421650e-01 5.20173758e-02
1.69846952e+00 -1.37347758e-01 8.80966485e-01 -2.90473104e-01
1.93628266e-01 -8.23542237e-01 2.48407125e-01 8.60893875e-02
1.18380177e+00 -9.90061462e-01 5.13620228e-02 -3.37744206e-01
-7.53520906e-01 8.09034109e-01 5.30173838e-01 -1.46711320e-01
1.05188586e-01 1.85084596e-01 3.53147328e-01 4.77207219e-03
-3.77603680e-01 -6.17317855e-02 1.50740415e-01 6.24313176e-01
2.51493245e-01 2.78239008e-02 -1.72316477e-01 1.43262565e-01
-1.88242465e-01 2.23876443e-02 7.36452401e-01 7.49573708e-01
-5.96262440e-02 -1.25176942e+00 -7.04531491e-01 1.76625028e-01
-2.53952175e-01 -2.43986279e-01 -2.04312298e-02 3.86613280e-01
2.34246552e-01 1.08347094e+00 -2.89514530e-02 -3.84496599e-01
1.18278161e-01 -2.85917580e-01 3.23825359e-01 -4.27798122e-01
-6.09347597e-02 3.55254024e-01 -2.49627754e-01 -6.17937386e-01
-5.70107758e-01 -5.23769140e-01 -8.90554368e-01 -3.94981056e-01
-9.25233141e-02 1.03097774e-01 5.21679103e-01 5.33845425e-01
4.28969741e-01 6.59534633e-01 8.92320275e-01 -1.07715249e+00
-4.32854006e-03 -8.32828403e-01 -7.17954278e-01 6.17546797e-01
3.81466836e-01 -4.35633689e-01 -5.93281031e-01 4.46537018e-01]
|
[11.38636302947998, -1.3780770301818848]
|
4b9a6a3a-5452-4306-9cf2-7c9561014526
|
vita-video-instance-segmentation-via-object
|
2206.04403
| null |
https://arxiv.org/abs/2206.04403v2
|
https://arxiv.org/pdf/2206.04403v2.pdf
|
VITA: Video Instance Segmentation via Object Token Association
|
We introduce a novel paradigm for offline Video Instance Segmentation (VIS), based on the hypothesis that explicit object-oriented information can be a strong clue for understanding the context of the entire sequence. To this end, we propose VITA, a simple structure built on top of an off-the-shelf Transformer-based image instance segmentation model. Specifically, we use an image object detector as a means of distilling object-specific contexts into object tokens. VITA accomplishes video-level understanding by associating frame-level object tokens without using spatio-temporal backbone features. By effectively building relationships between objects using the condensed information, VITA achieves the state-of-the-art on VIS benchmarks with a ResNet-50 backbone: 49.8 AP, 45.7 AP on YouTube-VIS 2019 & 2021, and 19.6 AP on OVIS. Moreover, thanks to its object token-based structure that is disjoint from the backbone features, VITA shows several practical advantages that previous offline VIS methods have not explored - handling long and high-resolution videos with a common GPU, and freezing a frame-level detector trained on image domain. Code is available at https://github.com/sukjunhwang/VITA.
|
['Seon Joo Kim', 'Joon-Young Lee', 'Seoung Wug Oh', 'Sukjun Hwang', 'Miran Heo']
|
2022-06-09
| null | null | null | null |
['video-instance-segmentation']
|
['computer-vision']
|
[ 3.34609300e-02 -1.06270211e-02 -3.88944358e-01 -1.02698959e-01
-6.85198605e-01 -5.52493274e-01 6.19553149e-01 8.07049721e-02
-4.92488325e-01 4.19021726e-01 7.35747954e-03 -2.14443401e-01
2.29876384e-01 -8.36870372e-01 -1.09875000e+00 -4.95915890e-01
-1.63816854e-01 2.61671066e-01 8.09209406e-01 -7.63163045e-02
2.74085909e-01 1.91705793e-01 -1.65131152e+00 7.01293826e-01
6.33585572e-01 1.19660962e+00 5.18020928e-01 7.14896441e-01
-3.37695509e-01 1.11174572e+00 -3.20131361e-01 -3.69680434e-01
4.20908451e-01 -4.40972418e-01 -1.21899438e+00 3.48920614e-01
7.11881101e-01 -6.35708153e-01 -5.39278924e-01 7.58282423e-01
4.66628559e-02 -1.53694246e-02 2.20496058e-01 -1.25380898e+00
-1.57511830e-01 6.05968118e-01 -5.60019732e-01 4.92670268e-01
2.98908591e-01 4.02366966e-01 1.08588707e+00 -8.60563517e-01
9.60238516e-01 9.71210003e-01 5.16108215e-01 3.31489593e-01
-9.69040036e-01 -3.93174499e-01 4.63583052e-01 6.84794307e-01
-1.28139651e+00 -2.81232834e-01 4.58963692e-01 -4.80690688e-01
1.05905235e+00 2.69790828e-01 1.05501342e+00 8.09476495e-01
-9.91141573e-02 1.18739676e+00 9.40254211e-01 -1.88726336e-01
6.40465170e-02 -2.04461887e-01 2.93387920e-01 9.28834498e-01
-8.37323666e-02 -1.05840184e-01 -6.72583282e-01 2.35805228e-01
8.71187091e-01 1.99172851e-02 -3.24585319e-01 -2.61002660e-01
-1.32989419e+00 4.24887240e-01 5.00453651e-01 2.72665143e-01
-2.44739816e-01 4.95296001e-01 7.10123777e-01 9.98383313e-02
4.00225550e-01 2.40806583e-02 -5.06975055e-01 -2.96237946e-01
-1.25565767e+00 8.95698741e-02 5.88322639e-01 1.15574777e+00
1.01050186e+00 -2.33563371e-02 -3.20449233e-01 4.33025956e-01
1.63934827e-01 1.77989811e-01 2.07565978e-01 -1.31484842e+00
2.42019922e-01 4.40237999e-01 -1.39461637e-01 -6.10738516e-01
6.38345107e-02 -4.18578148e-01 -4.00759906e-01 1.04027323e-01
5.67685962e-01 2.00638607e-01 -1.03658903e+00 1.44383299e+00
5.44121861e-01 8.92562091e-01 -1.36044860e-01 8.95745397e-01
9.53182995e-01 8.51921260e-01 3.97606790e-02 -9.10800323e-02
1.63976336e+00 -1.32618344e+00 -3.13367784e-01 -1.87252257e-02
6.34873748e-01 -5.21241069e-01 9.55014706e-01 4.88604784e-01
-1.33575332e+00 -7.65085220e-01 -9.18754101e-01 -4.05023456e-01
-2.96640933e-01 -1.08891398e-01 5.15068948e-01 4.08058137e-01
-1.20755053e+00 5.85334599e-01 -9.82855678e-01 -2.34580114e-01
7.11958110e-01 2.25929484e-01 -2.22846612e-01 -2.02675045e-01
-8.43363523e-01 3.91505897e-01 6.55380666e-01 -6.65494874e-02
-1.15797091e+00 -1.04402065e+00 -9.12673175e-01 6.34889305e-02
7.35657990e-01 -7.12481856e-01 1.24682927e+00 -1.15898514e+00
-1.38661718e+00 9.07999873e-01 -3.66135925e-01 -7.53403723e-01
7.33896375e-01 -3.36416632e-01 -6.42623694e-04 8.30773532e-01
1.22263975e-01 9.21778500e-01 9.55056787e-01 -1.17359889e+00
-1.07038701e+00 -1.16701826e-01 4.25133437e-01 1.23078324e-01
-1.03119180e-01 2.55071707e-02 -1.21857262e+00 -5.48781812e-01
-1.91257760e-01 -6.59766853e-01 -1.09175056e-01 1.24830812e-01
-3.09039503e-01 -2.57032424e-01 1.06291986e+00 -7.67262638e-01
1.21376884e+00 -2.20374584e+00 7.13890698e-03 5.36167547e-02
4.65487748e-01 5.25351107e-01 -1.99825302e-01 2.71320194e-01
9.95273516e-02 6.95925876e-02 -2.44333073e-01 -3.22696149e-01
-2.10851580e-01 2.78847128e-01 -2.65086800e-01 4.15480614e-01
1.91929802e-01 1.00738692e+00 -1.09442472e+00 -7.06319034e-01
4.81778234e-01 4.67040420e-01 -8.34431767e-01 -1.39595460e-04
-4.72317129e-01 3.73980671e-01 -3.60075891e-01 6.94781423e-01
6.88836992e-01 -3.04084808e-01 2.40756586e-01 -2.90917009e-01
-3.75927478e-01 3.17070127e-01 -1.08096206e+00 1.92545557e+00
-2.79323339e-01 8.66524518e-01 9.11318660e-02 -1.20922863e+00
3.33151191e-01 2.17831120e-01 7.66432583e-01 -7.74657130e-01
8.57510194e-02 8.88386518e-02 -3.48134756e-01 -4.05484885e-01
6.62217796e-01 3.77555221e-01 1.66591108e-01 1.53673440e-01
2.21841902e-01 2.59121880e-02 6.75172567e-01 5.85529566e-01
1.14025104e+00 4.86282110e-01 3.96405794e-02 -3.32755953e-01
5.39919555e-01 8.25088397e-02 7.01060832e-01 7.93049932e-01
-1.89620078e-01 7.57494152e-01 6.26861453e-01 -4.25565124e-01
-1.12398946e+00 -1.01472020e+00 -2.07375407e-01 9.42527533e-01
4.25242752e-01 -9.77516234e-01 -1.05487978e+00 -7.18493521e-01
-1.60059750e-01 3.37051034e-01 -5.50834715e-01 3.23763102e-01
-7.98958600e-01 -7.69943744e-02 2.87268728e-01 6.07536674e-01
7.62462020e-01 -9.21826661e-01 -7.21063435e-01 3.55928987e-01
-4.28345174e-01 -1.66823244e+00 -4.89892632e-01 -1.09330982e-01
-9.17814910e-01 -1.26327360e+00 -6.56974196e-01 -8.49008322e-01
4.94841814e-01 5.11645913e-01 1.35395038e+00 4.13125694e-01
-4.02442276e-01 5.87514520e-01 -4.24950868e-01 9.38038109e-04
-1.87844560e-01 4.75782305e-02 -3.20029587e-01 -4.34177322e-03
4.80710268e-02 -4.69316125e-01 -8.65406156e-01 4.23704118e-01
-1.00154328e+00 4.31353241e-01 2.78699607e-01 5.30083597e-01
8.53559554e-01 -1.89542532e-01 1.80693939e-01 -8.08906257e-01
-3.54311585e-01 -3.64496946e-01 -6.77383363e-01 1.40390560e-01
-1.09493464e-01 -2.24809691e-01 5.64493060e-01 -2.43417114e-01
-8.13188195e-01 1.31019607e-01 -2.45134458e-01 -6.19595826e-01
-1.13834195e-01 1.99446663e-01 4.07040119e-02 1.08871512e-01
2.54074365e-01 4.44888204e-01 -8.94342214e-02 -3.18487793e-01
4.76047009e-01 3.64035130e-01 6.93012536e-01 -7.60210335e-01
6.48211896e-01 7.81987786e-01 -3.15383762e-01 -9.68250632e-01
-8.78436863e-01 -8.14295590e-01 -7.11047947e-01 -4.76137102e-01
1.14643443e+00 -1.07275999e+00 -8.11665595e-01 5.23899376e-01
-1.15740585e+00 -8.27851176e-01 -4.54808265e-01 3.01073819e-01
-8.41001511e-01 6.36545122e-01 -8.91055763e-01 -2.30858162e-01
-3.06520730e-01 -1.21880567e+00 1.05654609e+00 9.53997895e-02
3.45290406e-03 -7.09235013e-01 -3.37969482e-01 5.27052999e-01
-2.00655609e-02 2.29012713e-01 4.95754510e-01 -2.71581173e-01
-1.21244872e+00 2.88576961e-01 -4.25541669e-01 4.31519836e-01
-2.48010114e-01 2.11038038e-01 -8.07865858e-01 -2.82387108e-01
-8.10234174e-02 -1.28109664e-01 1.04898715e+00 5.09914935e-01
1.50140476e+00 -2.81356841e-01 -2.76109904e-01 8.33370984e-01
1.63977408e+00 3.82737927e-02 9.38961089e-01 3.98205191e-01
8.74268055e-01 1.98905855e-01 7.39628673e-01 5.63431919e-01
6.00244582e-01 6.75787389e-01 5.75373828e-01 -1.45983011e-01
-4.32635009e-01 -1.69682562e-01 5.16552448e-01 7.64210105e-01
-3.10699910e-01 -7.71974623e-02 -8.01747441e-01 7.73661256e-01
-1.90311038e+00 -1.13826036e+00 -4.32455689e-01 1.95470679e+00
7.55000591e-01 3.00102979e-01 4.34765249e-01 1.64536200e-02
6.16639972e-01 3.21764648e-01 -2.32310995e-01 -1.32004201e-01
5.73612563e-02 2.38881528e-01 6.06725156e-01 4.32030350e-01
-1.08882165e+00 1.25225902e+00 5.33407354e+00 1.03695822e+00
-9.75492716e-01 2.37589151e-01 7.20820963e-01 -6.16295971e-02
-3.37889083e-02 2.01179802e-01 -8.91646862e-01 5.29984891e-01
7.43056238e-01 -5.69564067e-02 3.17993313e-01 8.50662649e-01
2.52997130e-01 -3.65850449e-01 -1.20579123e+00 9.70360100e-01
-7.21895769e-02 -1.88658869e+00 8.69620517e-02 6.04664795e-02
6.65906787e-01 2.84600437e-01 -1.30135819e-01 2.41926804e-01
-7.40866587e-02 -5.99128425e-01 1.18999362e+00 2.77822495e-01
8.23691010e-01 -5.66097975e-01 3.66951883e-01 1.59292623e-01
-1.67917001e+00 1.41927615e-01 -2.23809555e-01 6.50165379e-02
2.28119358e-01 4.79134291e-01 -5.38356483e-01 6.48187280e-01
9.05509531e-01 1.18814123e+00 -3.35203141e-01 1.10875535e+00
-2.44586140e-01 8.58131826e-01 -4.59763825e-01 5.18457830e-01
5.87115765e-01 -1.33140758e-01 4.64551747e-01 1.49708343e+00
2.65906826e-02 4.25779730e-01 5.02400637e-01 6.33176565e-01
-2.02863127e-01 -7.40021989e-02 -2.78786391e-01 1.90715566e-01
1.52917862e-01 1.17082405e+00 -1.18279874e+00 -8.46306682e-01
-5.47936380e-01 1.02796686e+00 2.32341029e-02 3.64537477e-01
-1.28577244e+00 -8.20442662e-02 7.41456509e-01 2.74201214e-01
9.24251854e-01 -4.51864004e-01 -7.68410787e-02 -1.30968809e+00
1.25654384e-01 -9.39180195e-01 3.82460326e-01 -6.41134858e-01
-6.87434554e-01 4.90051806e-01 3.52167040e-02 -1.28313744e+00
8.03071409e-02 -6.41647458e-01 -5.69994330e-01 3.59543025e-01
-1.84477818e+00 -1.03793633e+00 -4.43768114e-01 8.97468686e-01
1.13243699e+00 3.27938378e-01 3.29308540e-01 3.53871405e-01
-6.16805613e-01 4.18095231e-01 -4.12559286e-02 4.67741281e-01
3.60360056e-01 -1.08406997e+00 5.06730020e-01 9.29532886e-01
3.68350387e-01 2.96160072e-01 4.03200388e-01 -5.43430507e-01
-1.48331738e+00 -1.08399880e+00 4.68124807e-01 -3.98136556e-01
7.12297022e-01 -4.71945286e-01 -9.15848613e-01 8.05404007e-01
3.37243766e-01 4.38868493e-01 2.23284319e-01 -1.95161477e-01
-4.03638572e-01 -8.33355859e-02 -9.03858066e-01 5.38659692e-01
1.36439073e+00 -5.13279974e-01 -3.88997614e-01 3.79045457e-01
7.75053382e-01 -6.55866921e-01 -9.42250371e-01 2.54844993e-01
4.10679996e-01 -1.13620317e+00 1.20227897e+00 -3.32732797e-01
7.63410389e-01 -4.53225344e-01 3.26806903e-02 -6.50206625e-01
-1.37302369e-01 -7.60336280e-01 -3.55809331e-01 1.04808021e+00
6.74482286e-02 -3.79495591e-01 9.17310655e-01 2.33813614e-01
-3.63467544e-01 -8.81356537e-01 -8.18347216e-01 -8.85298550e-01
-2.26894185e-01 -8.70446503e-01 3.30572009e-01 8.21343362e-01
-1.40109971e-01 3.77418920e-02 -9.78574604e-02 1.49619639e-01
6.51590347e-01 1.04951896e-01 7.79914439e-01 -7.08034933e-01
-3.82539064e-01 -4.24203128e-01 -6.69216514e-01 -1.48443675e+00
9.36234146e-02 -8.61739814e-01 -9.34823528e-02 -1.56204295e+00
2.08751738e-01 -5.63653052e-01 -3.24233800e-01 5.69130480e-01
-2.44746655e-02 5.73894739e-01 5.98058462e-01 2.80181229e-01
-1.02483642e+00 3.07168365e-01 1.24072289e+00 -1.14063218e-01
-2.15773564e-02 -2.39206031e-01 -1.88442037e-01 8.36869121e-01
7.02494264e-01 -4.67534244e-01 -3.65466565e-01 -5.00694931e-01
-1.83316022e-01 6.88951090e-02 5.01281500e-01 -1.07755971e+00
2.21844047e-01 -5.22596240e-02 3.02453861e-02 -8.09807837e-01
3.76447111e-01 -6.82826579e-01 3.75544056e-02 6.10154688e-01
-7.77737051e-02 -1.65995434e-01 2.64683843e-01 4.79753584e-01
-3.45788032e-01 -2.41549700e-01 6.89546227e-01 -3.27069998e-01
-1.43504047e+00 5.52603364e-01 -3.11259687e-01 3.35950851e-01
1.22437787e+00 -5.99965215e-01 -2.54040807e-01 5.50815687e-02
-6.92314327e-01 3.15928936e-01 5.53104937e-01 3.18401039e-01
5.83093464e-01 -9.52285290e-01 -7.09530652e-01 2.76836380e-02
2.05552317e-02 1.59713298e-01 4.11376297e-01 9.27880943e-01
-8.77137423e-01 2.42724419e-01 -2.10222840e-01 -1.06854177e+00
-1.40972400e+00 5.54654837e-01 1.45272389e-01 -2.67222881e-01
-1.08837438e+00 8.17041814e-01 5.46559453e-01 1.51010081e-01
1.46884084e-01 -6.67980909e-01 7.44793862e-02 1.65634245e-01
6.14485264e-01 3.19627196e-01 3.48263048e-02 -5.67543030e-01
-2.74605125e-01 6.28121555e-01 -1.06620260e-01 -8.95369127e-02
1.26262701e+00 -1.05712727e-01 -6.57772049e-02 1.11393817e-01
1.24582767e+00 -1.34210765e-01 -1.69507396e+00 -3.52641582e-01
-2.74112225e-02 -6.33847654e-01 8.40741023e-02 -4.43934143e-01
-1.25635970e+00 7.40160167e-01 2.97692001e-01 1.14955984e-01
1.21258307e+00 1.36426345e-01 1.10063207e+00 1.73624948e-01
4.73664075e-01 -1.09088433e+00 3.00157815e-01 6.38597369e-01
5.22942901e-01 -1.07332420e+00 -4.86515649e-03 -8.24749827e-01
-4.69594479e-01 1.11250842e+00 4.89188015e-01 -2.43312016e-01
5.17406583e-01 3.20161790e-01 -2.54784107e-01 -8.62550884e-02
-6.68536425e-01 -5.54743350e-01 1.65923178e-01 4.21015203e-01
1.74047053e-01 -1.80814356e-01 -2.67335922e-01 1.58015341e-01
1.11972243e-01 2.27782845e-01 6.10034943e-01 9.79098380e-01
-4.24414873e-01 -1.06267023e+00 -1.65646374e-01 2.36124933e-01
-7.13146865e-01 -3.26708347e-01 2.76486337e-01 8.57480347e-01
2.64092356e-01 6.05048120e-01 2.93237269e-01 -1.85231209e-01
5.66543862e-02 -2.10768208e-01 4.43853915e-01 -6.44088507e-01
-6.87072992e-01 1.19632088e-01 1.29545882e-01 -1.01481974e+00
-6.16946280e-01 -7.54143417e-01 -1.44777572e+00 -2.99173564e-01
-1.59714445e-02 -5.95995076e-02 6.35491252e-01 9.78182197e-01
4.06030059e-01 5.13798177e-01 4.29850280e-01 -1.02746010e+00
1.53093725e-01 -4.10544664e-01 -2.21959546e-01 3.63481611e-01
2.50006437e-01 -3.45473170e-01 8.27690121e-04 4.61547047e-01]
|
[9.190641403198242, -0.046237919479608536]
|
0f01020d-e905-4dfc-ac54-cc2d56ef751b
|
double-topic-shifts-in-open-domain
| null | null |
https://aclanthology.org/W16-4408
|
https://aclanthology.org/W16-4408.pdf
|
Double Topic Shifts in Open Domain Conversations: Natural Language Interface for a Wikipedia-based Robot Application
|
The paper describes topic shifting in dialogues with a robot that provides information from Wiki-pedia. The work focuses on a double topical construction of dialogue coherence which refers to discourse coherence on two levels: the evolution of dialogue topics via the interaction between the user and the robot system, and the creation of discourse topics via the content of the Wiki-pedia article itself. The user selects topics that are of interest to her, and the system builds a list of potential topics, anticipated to be the next topic, by the links in the article and by the keywords extracted from the article. The described system deals with Wikipedia articles, but could easily be adapted to other digital information providing systems.
|
['Kristiina Jokinen', 'Graham Wilcock']
|
2016-12-01
| null | null | null |
ws-2016-12
|
['goal-oriented-dialogue-systems']
|
['natural-language-processing']
|
[-3.05369139e-01 1.52345252e+00 -1.80117860e-01 -2.02559270e-02
-2.73231506e-01 -5.76915741e-01 1.26498818e+00 5.55377185e-01
-1.73990905e-01 1.17815185e+00 1.07364976e+00 2.53599495e-01
-2.57542431e-01 -8.12864184e-01 -7.45219067e-02 -3.26117605e-01
-2.20890582e-01 9.76477027e-01 5.51010728e-01 -9.79069054e-01
3.45418692e-01 -2.11102232e-01 -1.64751077e+00 9.97399762e-02
7.11034477e-01 2.79815853e-01 7.89163053e-01 3.72422129e-01
-6.48988545e-01 9.45612073e-01 -9.15372729e-01 2.17248842e-01
-2.72271395e-01 -4.46842462e-01 -1.46692932e+00 1.65888205e-01
-9.84366238e-02 -2.98383176e-01 -1.65110916e-01 1.03442013e+00
1.17854722e-01 1.56142795e-02 5.87447405e-01 -1.28038204e+00
7.31105283e-02 1.31215501e+00 2.29386076e-01 -4.40536439e-01
1.09017575e+00 -3.36236715e-01 7.96436071e-01 -6.52901947e-01
1.71693397e+00 1.67152703e+00 5.12453854e-01 4.44649369e-01
-7.32206523e-01 1.15957007e-01 -1.46577358e-01 1.05638593e-01
-8.13048184e-01 -2.12940201e-01 5.70306659e-01 -7.26774752e-01
8.08850348e-01 1.76490158e-01 7.83992231e-01 7.50703394e-01
2.92265296e-01 4.00608063e-01 7.31700540e-01 -7.70688355e-01
1.21763624e-01 9.11832750e-01 6.01120353e-01 5.59858501e-01
1.71048313e-01 -5.36412716e-01 -7.44500458e-01 -3.53430122e-01
3.43736470e-01 -9.72011030e-01 -2.46191502e-01 -4.51200664e-01
-1.49035645e+00 9.23229873e-01 -1.24710843e-01 7.32735991e-01
-6.38246417e-01 -2.64203638e-01 8.26410472e-01 5.17618895e-01
5.39506197e-01 1.05996370e+00 -2.77856767e-01 -6.42822802e-01
-8.29764679e-02 6.94195569e-01 1.79673898e+00 1.17400217e+00
6.91568673e-01 -6.12130523e-01 -1.03588309e-02 7.05107749e-01
7.76951909e-01 6.94050491e-02 4.75620568e-01 -1.11170077e+00
3.00375491e-01 6.46773398e-01 7.47157872e-01 -1.21640003e+00
-6.31336629e-01 5.27839124e-01 1.30369648e-01 1.28178924e-01
3.66009772e-01 -7.51923680e-01 -2.97945440e-02 1.27901137e+00
7.95328021e-01 -1.01617861e+00 6.69670820e-01 4.65810776e-01
1.41533732e+00 1.08225763e+00 8.44813883e-02 -5.67018211e-01
1.69395602e+00 -9.32278454e-01 -1.48758459e+00 1.66475415e-01
7.76405275e-01 -6.85040116e-01 5.23460865e-01 1.34000480e-01
-9.94517207e-01 -3.16167086e-01 -1.10081434e+00 -2.27647826e-01
-7.26438463e-01 8.73091538e-03 2.29665071e-01 -1.63572058e-01
-1.19212055e+00 4.12220746e-01 -2.68839985e-01 -1.42131758e+00
-6.24610722e-01 1.47429854e-01 -1.63920105e-01 6.54438555e-01
-1.65375364e+00 1.48878729e+00 9.64634359e-01 -4.05583739e-01
-5.15498042e-01 -1.33759350e-01 -8.24504435e-01 -1.67889267e-01
4.88723159e-01 -5.43000877e-01 1.62825191e+00 -6.81671023e-01
-2.26649046e+00 6.39234066e-01 4.25187886e-01 -4.68482643e-01
4.57685322e-01 -6.19915605e-01 2.07666457e-02 2.28150576e-01
5.07006526e-01 8.01071048e-01 3.60675007e-01 -1.47125161e+00
-1.26847780e+00 -1.10786542e-01 4.25852120e-01 9.34923530e-01
-2.18737394e-01 7.65919909e-02 -5.01853585e-01 -1.12431929e-01
-3.13742340e-01 -7.15841293e-01 -1.07528932e-01 -4.30295140e-01
-6.80485189e-01 -8.06254387e-01 9.58536327e-01 -8.66689920e-01
1.26073933e+00 -1.96215558e+00 4.58655059e-01 -2.50478033e-02
3.83101255e-01 -2.16300964e-01 3.14087629e-01 1.29136682e+00
4.18244928e-01 -9.94663313e-02 3.37106347e-01 2.06953093e-01
2.74524897e-01 -7.89207816e-02 -1.30790487e-01 3.95752192e-02
-5.00687003e-01 3.73279080e-02 -1.25170493e+00 -8.16200316e-01
1.80470571e-01 -8.84910487e-03 1.05011165e-01 3.00274432e-01
-8.74540925e-01 3.06143519e-02 -7.94927299e-01 -1.14167452e-01
-1.15566850e-01 1.45166188e-01 5.62930822e-01 1.55036986e-01
-7.64366627e-01 8.28092039e-01 -7.40798891e-01 1.52668583e+00
-1.74176186e-01 9.21114028e-01 4.61896896e-01 -4.04748142e-01
1.16272652e+00 7.72020996e-01 5.72318852e-01 -1.31266356e-01
1.58164933e-01 -3.90427262e-02 -2.79859960e-01 -1.07752466e+00
1.27519190e+00 6.12626314e-01 -5.36290348e-01 7.31971562e-01
2.43725359e-01 -5.92204988e-01 5.68758190e-01 4.66856509e-01
7.52249300e-01 5.63175917e-01 8.99252415e-01 -6.48228824e-01
2.76454777e-01 7.51745880e-01 -1.26044974e-01 7.24561095e-01
3.49998958e-02 -3.68181705e-01 8.06469381e-01 -1.89686134e-01
-1.16997063e+00 -5.45106471e-01 -2.64866769e-01 1.23038161e+00
4.11374599e-01 -8.27576041e-01 -1.02783513e+00 -5.12290478e-01
-2.34405950e-01 9.73575652e-01 -6.77548230e-01 4.41233367e-01
-4.02174056e-01 1.84315313e-02 2.01584324e-01 -6.00743413e-01
4.57646161e-01 -1.41838396e+00 -1.15817928e+00 4.33917195e-01
-5.02204418e-01 -6.11030817e-01 9.73968878e-02 9.74445194e-02
-4.11130875e-01 -8.78644705e-01 -4.88718182e-01 -1.09300542e+00
4.32340920e-01 -7.86045045e-02 6.83104157e-01 -9.94149894e-02
1.29420653e-01 8.23907614e-01 -7.37463713e-01 -7.15681255e-01
-1.21302354e+00 5.80769539e-01 -1.46115705e-01 -6.65749729e-01
-8.95051584e-02 -7.90166259e-02 1.02850758e-01 6.18206598e-02
-3.99830014e-01 4.68548506e-01 -6.85254633e-02 5.76821268e-01
-3.54791045e-01 -7.21729221e-03 7.47570038e-01 -1.20588696e+00
1.27211654e+00 -9.93363380e-01 -2.45921254e-01 2.18767986e-01
-4.48351711e-01 -2.57235044e-03 -5.44114336e-02 1.14478404e-02
-1.58976901e+00 -7.91519657e-02 3.53706270e-01 8.62025142e-01
-2.64206678e-01 9.41483498e-01 -5.94496205e-02 4.60026234e-01
7.60045052e-01 -1.68522254e-01 4.46514875e-01 -2.15008602e-01
8.72382760e-01 9.17675614e-01 3.86547863e-01 -4.60691184e-01
2.02929452e-01 -5.76862656e-02 -6.08493805e-01 -1.35749102e+00
-3.18482146e-02 -5.75595260e-01 -7.07887352e-01 -1.03007269e+00
7.99890637e-01 -6.87162280e-01 -6.74727440e-01 2.75910050e-01
-1.66102183e+00 -4.82919365e-01 -6.14806831e-01 3.70749325e-01
-7.70199180e-01 3.73349249e-01 -5.44357419e-01 -7.21592247e-01
-3.43746513e-01 -7.89763510e-01 6.11105204e-01 4.43677038e-01
-1.23017359e+00 -1.19654620e+00 4.01454836e-01 -2.58235902e-01
3.36356133e-01 3.31062108e-01 1.07316685e+00 -1.15080345e+00
-6.77123740e-02 8.14453661e-02 9.94057059e-02 -5.51981449e-01
3.65773559e-01 1.76807553e-01 -5.09293556e-01 1.89458475e-01
-5.93693294e-02 -3.16828310e-01 -1.90282408e-02 -2.41892738e-03
-1.41741335e-01 -7.76792586e-01 -9.04251039e-01 -7.81178117e-01
9.12747204e-01 6.90798163e-01 5.95471442e-01 1.06453109e+00
1.69164389e-01 1.55626392e+00 1.05819416e+00 6.58486307e-01
8.39326620e-01 8.84382367e-01 2.88090765e-01 9.45652723e-02
-4.85970303e-02 -4.31180038e-02 3.74906272e-01 9.45230901e-01
1.95373908e-01 -1.73392415e-01 -9.74217057e-01 7.76854873e-01
-2.28767347e+00 -8.50960791e-01 -2.95589656e-01 1.65066636e+00
1.16105533e+00 -1.52639806e-01 1.95106834e-01 -4.02627051e-01
8.68332803e-01 2.36274019e-01 -1.34709671e-01 -5.02259195e-01
3.05454016e-01 -6.42188966e-01 5.99188693e-02 1.02351367e+00
-9.33148324e-01 1.09939730e+00 6.64214373e+00 3.03821802e-01
-5.65612733e-01 6.87940270e-02 -3.15987021e-02 3.80396813e-01
-1.74753368e-01 4.07311320e-02 -8.60948503e-01 2.98814535e-01
6.66217148e-01 -1.07744038e+00 -3.30582671e-02 1.04912937e+00
4.25799549e-01 -7.59250402e-01 -1.00707710e+00 9.65156481e-02
4.12058309e-02 -1.28127384e+00 -1.51970506e-01 -4.20211367e-02
3.15515876e-01 -2.01700374e-01 -4.49865133e-01 2.23592177e-01
8.75265956e-01 -2.41613835e-01 9.00127411e-01 5.89873791e-01
3.60327542e-01 -2.25108474e-01 9.12928462e-01 3.36725950e-01
-6.58766806e-01 2.34699979e-01 -1.80301934e-01 3.53459232e-02
3.98459673e-01 1.47369012e-01 -1.65514171e+00 6.50993407e-01
5.79233289e-01 6.77845001e-01 -5.27310297e-02 8.45421672e-01
-2.03308463e-01 8.48219544e-02 -2.76424259e-01 -8.26763153e-01
1.05288975e-01 -2.95631289e-01 1.31792319e+00 1.45460451e+00
1.59312904e-01 1.17989533e-01 4.26242203e-01 3.81579310e-01
2.02527747e-01 5.09298623e-01 -9.00339246e-01 1.89797193e-01
8.78187060e-01 1.11307716e+00 -7.09869921e-01 -5.12427688e-01
6.37606382e-02 6.08798742e-01 -2.35928548e-03 4.01870646e-02
-1.84891552e-01 -8.92328441e-01 2.53012329e-01 1.04346432e-01
-6.86634257e-02 -6.88170046e-02 7.60407820e-02 -3.05309415e-01
-4.65715170e-01 -6.21984124e-01 9.59864929e-02 -8.97611320e-01
-6.33817077e-01 8.46223116e-01 6.46167278e-01 -8.39107931e-01
-8.45019817e-01 -8.47230479e-03 -5.70236206e-01 4.01140749e-01
-8.24689090e-01 -8.88625324e-01 -2.22327188e-01 -4.23831679e-02
8.02290261e-01 -1.20615318e-01 1.12749875e+00 -2.70636797e-01
-2.96421386e-02 -4.68591183e-01 3.00026983e-01 -3.59201252e-01
9.10594761e-01 -1.33154094e+00 8.90922099e-02 7.02194795e-02
-7.36750960e-01 4.63286698e-01 1.44495714e+00 -1.11570728e+00
-9.87762153e-01 -4.93583024e-01 1.33043909e+00 -1.24546617e-01
8.60658884e-01 -8.56475681e-02 -6.57073200e-01 5.38358331e-01
1.33157289e+00 -1.39459538e+00 3.17124873e-01 1.30912393e-01
4.68944997e-01 5.04358709e-01 -1.06900179e+00 6.80499911e-01
4.44865435e-01 -1.60394877e-01 -1.31957400e+00 6.03274345e-01
1.11588216e+00 -7.55997717e-01 -1.01252472e+00 -2.45307788e-01
5.07772446e-01 -3.12542260e-01 3.67719054e-01 -5.16490405e-03
3.77476513e-01 -2.64633924e-01 3.28995407e-01 -1.65597939e+00
-6.34788442e-03 -9.54694629e-01 2.85893500e-01 1.66442072e+00
6.74463391e-01 -6.56028807e-01 2.44248688e-01 5.47373056e-01
-2.79622018e-01 -7.70693049e-02 -8.07750404e-01 1.81491971e-02
-1.71977773e-01 2.80691713e-01 2.40950838e-01 1.00506878e+00
1.59342420e+00 6.60851419e-01 -1.30284548e-01 -2.68641591e-01
1.36477694e-01 -4.83926117e-01 9.71292794e-01 -1.74395120e+00
3.70458245e-01 -3.04762989e-01 4.40335833e-02 -1.00121796e+00
-1.95323870e-01 -4.13205504e-01 8.73192549e-01 -2.07349896e+00
-2.26047471e-01 -3.76667768e-01 1.06032741e+00 1.88428938e-01
3.73693109e-01 -9.41609502e-01 3.84935737e-01 6.70964122e-01
-9.51118290e-01 6.51469886e-01 8.81886482e-01 1.57149002e-01
-9.58962500e-01 -3.39832962e-01 -4.90468770e-01 1.01697218e+00
7.19387949e-01 -5.29751420e-01 -9.78654921e-02 3.04699093e-01
4.08514380e-01 4.39590067e-01 -3.36688221e-01 -7.80643642e-01
5.63082576e-01 -2.62880046e-02 -4.77028280e-01 -6.57293737e-01
2.08696634e-01 -7.65254974e-01 2.07949400e-01 5.83125472e-01
-9.61195588e-01 -2.64219552e-01 1.37081876e-01 3.50467980e-01
-2.24362284e-01 -7.98483729e-01 1.25401437e-01 -3.08388829e-01
-7.60869861e-01 -6.99149787e-01 -1.41671050e+00 -3.21284175e-01
1.26922989e+00 -2.67657816e-01 -6.23736680e-01 -7.43762910e-01
-8.65233243e-01 6.66702569e-01 3.99362832e-01 6.54503465e-01
8.19395781e-02 -1.04046237e+00 -5.57383776e-01 -6.33166790e-01
6.96111768e-02 9.46928486e-02 -3.20900083e-01 2.07019299e-01
-5.77280998e-01 8.09203565e-01 -4.05409843e-01 -3.16845477e-01
-1.38562429e+00 1.00887820e-01 9.36737508e-02 -3.73691857e-01
-8.08299184e-01 2.18121707e-01 1.43848971e-01 -4.97809559e-01
6.23722374e-01 -9.10870731e-02 -1.35182071e+00 7.25710392e-01
5.06636143e-01 6.10820711e-01 -5.75243235e-01 -7.30986357e-01
1.22084342e-01 2.50019163e-01 -1.01785474e-01 -6.43710077e-01
1.21526718e+00 -1.05725062e+00 -5.78139842e-01 1.11467314e+00
5.28381348e-01 -1.00306273e-01 -8.57576549e-01 -2.19913691e-01
5.92396736e-01 3.16190362e-01 -5.44957817e-01 -9.22508180e-01
1.34001300e-01 -6.96922615e-02 1.54844657e-01 1.25574195e+00
1.72477990e-01 4.55040962e-01 2.26334110e-01 6.54242218e-01
4.78080601e-01 -1.58300638e+00 1.35222375e-01 1.08133197e+00
1.57967615e+00 -5.97052813e-01 1.47570565e-01 -7.87896156e-01
-8.80272031e-01 1.67152464e+00 7.03209400e-01 2.72956520e-01
7.39369631e-01 1.04552455e-01 2.14645281e-01 -8.43546569e-01
-8.18674743e-01 -1.68453176e-02 -3.30373138e-01 8.58370960e-01
5.08775175e-01 8.14135559e-03 -1.12086499e+00 3.37983519e-01
-5.49717128e-01 -1.51489183e-01 1.15904653e+00 1.01419628e+00
-1.03083742e+00 -9.76930678e-01 -4.86870527e-01 2.46811882e-01
-1.09341644e-01 4.45642769e-01 -7.84017026e-01 1.19330907e+00
-1.79568350e-01 1.05226707e+00 1.57520518e-01 -1.18567176e-01
3.64045352e-01 1.45532265e-01 -1.72034606e-01 -9.86622691e-01
-9.75023091e-01 -4.35011461e-02 1.32169926e+00 -2.34879196e-01
-6.37511909e-01 -1.07573318e+00 -1.58199227e+00 1.26349345e-01
-4.02386039e-01 8.33567500e-01 1.02836084e+00 8.84181678e-01
3.25872242e-01 4.21613246e-01 4.53531653e-01 -8.26437175e-01
-4.56594154e-02 -1.51396203e+00 -3.05980951e-01 -8.41429830e-02
1.11308791e-01 -6.07667327e-01 3.84068377e-02 2.29050979e-01]
|
[12.752128601074219, 7.942568302154541]
|
126d8fb1-c39a-4c8d-bcbe-db7f787ef12f
|
off-policy-evaluation-in-doubly-inhomogeneous
|
2306.08719
| null |
https://arxiv.org/abs/2306.08719v1
|
https://arxiv.org/pdf/2306.08719v1.pdf
|
Off-policy Evaluation in Doubly Inhomogeneous Environments
|
This work aims to study off-policy evaluation (OPE) under scenarios where two key reinforcement learning (RL) assumptions -- temporal stationarity and individual homogeneity are both violated. To handle the ``double inhomogeneities", we propose a class of latent factor models for the reward and observation transition functions, under which we develop a general OPE framework that consists of both model-based and model-free approaches. To our knowledge, this is the first paper that develops statistically sound OPE methods in offline RL with double inhomogeneities. It contributes to a deeper understanding of OPE in environments, where standard RL assumptions are not met, and provides several practical approaches in these settings. We establish the theoretical properties of the proposed value estimators and empirically show that our approach outperforms competing methods that ignore either temporal nonstationarity or individual heterogeneity. Finally, we illustrate our method on a data set from the Medical Information Mart for Intensive Care.
|
['Lan Wang', 'Zhengling Qi', 'Chengchun Shi', 'Zeyu Bian']
|
2023-06-14
| null | null | null | null |
['offline-rl']
|
['playing-games']
|
[ 1.35563836e-01 1.97776537e-02 -7.53633499e-01 -6.85967058e-02
-7.80670822e-01 -2.41023719e-01 3.19830090e-01 1.58324152e-01
-6.75966680e-01 1.23252952e+00 -1.58368591e-02 -5.26779711e-01
-6.48584783e-01 -2.46684268e-01 -6.65478647e-01 -1.00239289e+00
-5.07542729e-01 7.13659286e-01 1.47020400e-01 3.12468372e-02
1.52176082e-01 3.63825262e-01 -1.31313014e+00 -2.30039939e-01
9.60906208e-01 6.80749834e-01 4.75628152e-02 7.97112226e-01
4.75465387e-01 1.09863019e+00 -6.19004726e-01 2.39750315e-02
2.86635399e-01 -5.07199883e-01 -4.99600172e-01 4.87567000e-02
-2.88949937e-01 -6.78813756e-01 -2.52353311e-01 8.64233971e-01
7.51317501e-01 2.65052646e-01 7.52814710e-01 -1.41806448e+00
-3.19422215e-01 6.32234514e-01 -6.17194593e-01 5.93993425e-01
1.36774927e-01 2.57081479e-01 6.38165057e-01 5.03438003e-02
1.96210802e-01 1.36395073e+00 6.74243569e-01 6.30642593e-01
-1.36877346e+00 -5.62306762e-01 4.93972629e-01 -7.12270290e-02
-7.68678844e-01 -3.19105238e-01 2.55472690e-01 -3.75954241e-01
8.96274745e-01 -2.02542618e-01 5.47257125e-01 1.53369296e+00
6.22384906e-01 1.13554990e+00 1.66465127e+00 -5.83005130e-01
6.03566289e-01 1.05998650e-01 2.51351595e-01 4.14580256e-01
2.79891133e-01 6.34292066e-01 -4.65883702e-01 -4.78850543e-01
1.23676550e+00 1.55593112e-01 -3.95356417e-02 -5.48694670e-01
-1.06817353e+00 8.88096035e-01 -3.32635254e-01 -4.56873029e-02
-9.70980406e-01 4.15756881e-01 1.82477489e-01 5.38112164e-01
4.46836561e-01 1.08253740e-01 -4.50938702e-01 -4.70593989e-01
-8.04533362e-01 3.01654667e-01 8.91211152e-01 1.14437187e+00
1.58717960e-01 1.57634392e-01 -5.11458576e-01 3.98816884e-01
1.26130655e-01 6.97836459e-01 5.91514409e-01 -1.38568020e+00
3.68206292e-01 -3.01715404e-01 9.41502213e-01 -2.73743004e-01
-6.60011649e-01 -6.24643981e-01 -3.92447621e-01 2.15488877e-02
5.12484074e-01 -5.07194102e-01 -7.35698283e-01 2.04521894e+00
1.96971536e-01 4.51909035e-01 2.98992306e-01 4.59979802e-01
7.33940005e-02 1.50389984e-01 9.81896818e-02 -9.98504639e-01
9.57197070e-01 -7.09658504e-01 -1.22571087e+00 1.30067930e-01
4.22545433e-01 -5.12977540e-01 9.28001344e-01 5.22729158e-01
-1.48321533e+00 -1.22789115e-01 -5.50481319e-01 8.06488812e-01
3.04470599e-01 -2.84288734e-01 6.05586886e-01 6.33090854e-01
-1.02904916e+00 7.05592334e-01 -1.33242905e+00 -3.60238582e-01
2.29531914e-01 3.89737666e-01 3.28441232e-01 1.64151564e-02
-1.16080260e+00 8.40730190e-01 -6.77675456e-02 -9.39705595e-02
-1.34876788e+00 -6.78165317e-01 -4.31407809e-01 -1.16387270e-01
1.04740763e+00 -6.89332366e-01 1.88909268e+00 -1.10241461e+00
-1.79878783e+00 4.75765020e-02 1.16863782e-02 -6.35304570e-01
8.94549251e-01 -2.20430955e-01 -2.62831300e-01 1.87613681e-01
7.33035207e-02 -2.94050537e-02 8.35076451e-01 -1.26557028e+00
-6.57197118e-01 -1.38634697e-01 7.16090873e-02 4.02151287e-01
9.07557011e-02 3.70734208e-03 1.36735037e-01 -5.08379400e-01
-3.73349756e-01 -1.08624327e+00 -4.50224549e-01 -6.74222767e-01
-2.87171692e-01 -1.48947060e-01 1.69722110e-01 -2.77383089e-01
1.21393752e+00 -1.94667459e+00 -7.94859529e-02 2.32733428e-01
-2.36994233e-02 -2.48672292e-01 -2.34734081e-03 8.44511569e-01
1.72135964e-01 -2.04781860e-01 -1.93522677e-01 -4.33949471e-01
2.17948332e-01 6.47822917e-01 -4.12809670e-01 7.65861690e-01
-1.78508431e-01 7.74518073e-01 -1.23597944e+00 -3.74218643e-01
5.58555312e-02 2.56081000e-02 -6.37380064e-01 1.30085930e-01
-2.04009324e-01 7.03851700e-01 -6.33693278e-01 3.53320301e-01
2.90304303e-01 -2.80562460e-01 4.67526317e-01 6.22795880e-01
-1.30033568e-01 5.01541533e-02 -1.18431044e+00 1.15017021e+00
-3.23124200e-01 5.45829237e-02 6.67862147e-02 -1.17415893e+00
2.42093906e-01 5.65368116e-01 1.02753079e+00 -6.26772761e-01
2.31638536e-01 1.57071218e-01 3.79012302e-02 -6.83050990e-01
1.93523407e-01 -6.89261019e-01 -6.30950406e-02 5.66366613e-01
1.54007208e-02 3.12065899e-01 -5.67115881e-02 -1.78879686e-03
1.33252609e+00 3.91203821e-01 4.03354615e-01 -5.06065071e-01
-2.51311868e-01 -2.85007745e-01 7.26430058e-01 1.47591102e+00
-7.63480842e-01 -5.58344834e-02 7.66007781e-01 4.21625450e-02
-6.92717791e-01 -1.18158543e+00 -2.08321601e-01 9.21068192e-01
1.28721353e-02 8.64889920e-02 -5.13306856e-01 -5.42287827e-01
1.34801790e-01 7.15948105e-01 -8.81945491e-01 -1.34178579e-01
-2.26181313e-01 -1.15005231e+00 3.71069729e-01 5.55295467e-01
1.49660319e-01 -1.03154135e+00 -1.00366724e+00 4.77378279e-01
-1.74220279e-01 -9.52861488e-01 -2.36798137e-01 2.34279111e-01
-1.09730887e+00 -9.50484693e-01 -9.36035573e-01 -1.01489477e-01
4.18024391e-01 6.53544441e-02 1.03843534e+00 -3.52110237e-01
9.49599817e-02 1.16174603e+00 -3.60025465e-01 -6.62161469e-01
-4.84096974e-01 -2.62178838e-01 4.03203130e-01 6.47278577e-02
2.90965587e-01 -3.87030423e-01 -8.21657062e-01 3.59890342e-01
-9.45020735e-01 -3.07152033e-01 4.88115579e-01 9.46932316e-01
6.52260721e-01 2.46156275e-01 8.33787680e-01 -9.98603404e-01
9.46820140e-01 -6.64569497e-01 -7.48239219e-01 3.74056160e-01
-9.06923175e-01 1.44642711e-01 3.37255806e-01 -7.20800281e-01
-1.20694554e+00 -3.41216445e-01 4.08215821e-01 -6.37641907e-01
-1.58940077e-01 4.33078408e-01 3.70155334e-01 4.44722950e-01
3.46406549e-01 8.19189250e-02 1.08915426e-01 -2.93164551e-01
-8.50286558e-02 3.43923062e-01 2.17096165e-01 -8.57277155e-01
2.30531961e-01 4.40221012e-01 1.90242723e-01 -5.62650084e-01
-7.53655016e-01 -4.89653379e-01 -2.83567458e-01 -1.86825082e-01
6.77378774e-01 -9.98023272e-01 -1.12672830e+00 2.43702814e-01
-5.13765693e-01 -1.20560312e+00 -7.34915078e-01 1.13021195e+00
-1.51249909e+00 3.96663666e-01 -8.17810476e-01 -1.56737030e+00
4.09374118e-01 -1.08196568e+00 8.18488896e-01 3.75682972e-02
2.04108387e-01 -1.31123853e+00 6.93703234e-01 -9.08086449e-02
2.29368150e-01 6.83890209e-02 7.08338678e-01 -4.16844577e-01
-3.84514898e-01 3.04621637e-01 3.71196359e-01 5.71156032e-02
1.99737191e-01 -9.95810181e-02 -7.19783485e-01 -7.89985418e-01
2.07302287e-01 -3.29636693e-01 6.96044922e-01 1.19256413e+00
9.05901790e-01 -3.37793767e-01 -1.74832642e-01 1.96533680e-01
1.37248981e+00 5.94971895e-01 3.87831002e-01 4.74857062e-01
2.15717535e-02 6.98070824e-01 9.38156426e-01 1.17319739e+00
4.93628651e-01 5.00877976e-01 3.36247951e-01 -1.28652528e-01
4.94811356e-01 -2.32913643e-01 7.12164283e-01 6.26149178e-01
-2.96348095e-01 -5.46501160e-01 -5.06147742e-01 5.11597216e-01
-2.34601641e+00 -1.24701822e+00 1.06717475e-01 2.65932560e+00
7.31196523e-01 -2.19188072e-02 6.63945913e-01 -2.48425782e-01
6.49496198e-01 -2.26451337e-01 -9.37102079e-01 -2.89747417e-01
9.46558490e-02 2.07056701e-01 1.05521429e+00 5.82117558e-01
-9.33083951e-01 6.33520901e-01 7.61555433e+00 4.69039142e-01
-5.33806384e-01 2.98645347e-01 3.22201669e-01 -2.82809198e-01
-8.72729048e-02 -9.10255164e-02 -6.35673046e-01 4.22398865e-01
1.47965491e+00 9.61337760e-02 5.46847999e-01 3.85347337e-01
6.24752522e-01 -3.08886617e-01 -1.00095212e+00 6.21759176e-01
-2.61276603e-01 -3.76113087e-01 -6.34902537e-01 3.58128875e-01
8.48342717e-01 1.35135517e-01 4.59260315e-01 4.80590284e-01
1.07668269e+00 -6.94380522e-01 5.79835296e-01 8.16314995e-01
4.85188782e-01 -8.88190508e-01 5.90322256e-01 6.20419860e-01
-5.76721311e-01 -4.61914390e-01 -2.66470134e-01 6.89003915e-02
1.26175553e-01 2.23741353e-01 -3.92300040e-01 5.59238911e-01
5.57666361e-01 6.39517367e-01 9.31634903e-02 1.03795314e+00
-8.60109851e-02 8.00429761e-01 -1.58948362e-01 3.52578104e-01
2.82674760e-01 -5.03119826e-02 5.47455609e-01 8.57293785e-01
3.12493861e-01 -5.95196076e-02 3.45381409e-01 3.99872631e-01
4.49404716e-01 -1.41426141e-03 -5.32464921e-01 -3.45094465e-02
2.83718437e-01 7.51678944e-01 -7.68189549e-01 -3.69910508e-01
-4.20928270e-01 8.46449494e-01 1.12496324e-01 7.62395918e-01
-9.82424021e-01 1.13740481e-01 3.46659839e-01 -8.22128206e-02
3.83923024e-01 -2.11840779e-01 3.98738414e-01 -1.30920923e+00
-3.19512784e-01 -1.01985824e+00 7.48899341e-01 -2.50398457e-01
-1.45091975e+00 1.06402747e-01 6.74380064e-01 -1.08014023e+00
-5.78524768e-01 -4.08716381e-01 -1.71550632e-01 5.37125826e-01
-2.03080940e+00 -4.55792308e-01 3.91798288e-01 9.58727896e-01
5.30948579e-01 1.95485540e-02 5.49615026e-01 1.03358597e-01
-7.50828326e-01 4.16552335e-01 8.75117421e-01 -5.41756451e-01
6.69781208e-01 -1.46749973e+00 -3.56415123e-01 5.25782228e-01
-3.09453309e-01 6.25738382e-01 9.52374101e-01 -7.87791669e-01
-1.17179620e+00 -7.04986572e-01 2.07407981e-01 -3.82676631e-01
7.17581272e-01 -1.32780701e-01 -7.23203003e-01 1.03554082e+00
2.15805247e-01 -3.07190418e-01 8.58001709e-01 1.41306430e-01
2.38854632e-01 2.65113682e-01 -9.08308744e-01 4.77866411e-01
8.96814227e-01 -2.07850426e-01 -3.79641563e-01 4.70626980e-01
4.81253535e-01 -2.75542885e-01 -1.00786924e+00 3.69816631e-01
6.11332536e-01 -9.01100218e-01 7.50455022e-01 -1.03215837e+00
-8.56954083e-02 2.61975467e-01 -1.98502705e-01 -1.36266613e+00
-2.98849195e-01 -1.20672798e+00 -3.46424222e-01 7.75157094e-01
1.98118657e-01 -1.05055416e+00 3.90772521e-01 6.29520595e-01
3.55564058e-02 -5.08147240e-01 -1.07088482e+00 -1.32745302e+00
3.03966731e-01 -3.53569418e-01 3.64495516e-01 8.37035954e-01
3.21705602e-02 -8.01365972e-02 -7.95740306e-01 1.66871026e-01
8.48524630e-01 8.76954198e-02 3.49218518e-01 -1.01071501e+00
-9.61096048e-01 -2.68303186e-01 2.44357213e-01 -1.00725484e+00
3.92155230e-01 -1.31052539e-01 2.47304723e-01 -1.19473779e+00
5.00250220e-01 -5.78058183e-01 -9.33318615e-01 2.41422698e-01
-1.60922155e-01 -3.80940974e-01 -3.81896719e-02 3.64318073e-01
-8.86720657e-01 8.32418919e-01 1.20304441e+00 3.93361509e-01
-5.58326542e-01 5.79510033e-01 -5.21993876e-01 5.25655329e-01
9.82375324e-01 -7.60390103e-01 -8.46590400e-01 3.82361598e-02
6.35733455e-02 7.32812107e-01 3.22006613e-01 -4.90322709e-01
-1.32391602e-01 -8.54785264e-01 -9.31727588e-02 -3.78977090e-01
1.12495422e-01 -8.02486718e-01 8.12872350e-02 8.28006089e-01
-6.81110561e-01 3.64516646e-01 2.89599210e-01 1.07921445e+00
3.06232452e-01 -1.30854249e-01 6.67702496e-01 -2.64636666e-01
-1.04620881e-01 4.54421550e-01 -7.48751700e-01 2.48488978e-01
1.01482868e+00 1.51990384e-01 -4.13988620e-01 -8.79972756e-01
-8.60278785e-01 3.53713870e-01 2.49927506e-01 -4.30757329e-02
3.68380576e-01 -9.99090314e-01 -5.61298668e-01 -3.01561002e-02
-2.01104835e-01 -5.75205743e-01 5.67647517e-01 1.45526671e+00
3.16709429e-02 4.46556658e-01 2.39147386e-03 -5.13794839e-01
-9.58680630e-01 1.04987121e+00 4.06450212e-01 -6.95529580e-01
-7.30972767e-01 2.29157418e-01 4.88416582e-01 -5.48956506e-02
3.25294167e-01 -3.68562549e-01 -6.21526577e-02 -2.29307309e-01
3.41080040e-01 5.13314843e-01 -3.51621836e-01 -3.49883765e-01
-5.13583645e-02 2.37696156e-01 8.60342477e-03 -6.78729892e-01
1.19256961e+00 -5.03711700e-01 5.01082838e-01 9.75848079e-01
4.30026323e-01 -3.05381745e-01 -1.90755665e+00 -3.94851357e-01
-6.51848922e-03 -4.62252885e-01 1.76448926e-01 -7.67980635e-01
-6.06669903e-01 8.33034456e-01 8.87243629e-01 3.90418917e-01
1.09878325e+00 -1.83550671e-01 2.09010348e-01 1.80333823e-01
5.75241387e-01 -1.35286641e+00 1.93113908e-01 2.04896823e-01
4.71219122e-01 -1.00133789e+00 -9.15586799e-02 1.27828017e-01
-9.85056281e-01 5.71633875e-01 2.31289953e-01 -1.32582679e-01
6.61349893e-01 2.86706775e-01 -2.06455365e-01 -2.73167063e-02
-1.16624236e+00 -4.56837267e-01 -1.31848946e-01 7.53822565e-01
3.50030810e-02 2.86099702e-01 -3.34703147e-01 3.35127264e-01
2.92849481e-01 3.47564310e-01 8.02926898e-01 1.42062247e+00
-2.33787924e-01 -1.24290240e+00 -2.57330269e-01 3.98152977e-01
-8.67629826e-01 -7.75226504e-02 1.60840720e-01 9.50091243e-01
-4.36011136e-01 1.00586116e+00 7.53013864e-02 2.56014735e-01
4.08309042e-01 2.64057405e-02 9.20217872e-01 -3.15800637e-01
-4.29805666e-01 7.04686701e-01 -1.20114043e-01 -6.16923451e-01
-8.23490322e-01 -9.33490336e-01 -1.01608682e+00 -2.64263600e-01
-2.87502676e-01 1.52069315e-01 3.00477386e-01 9.13226902e-01
1.83141708e-01 8.20181370e-01 8.20279241e-01 -5.18062115e-01
-1.17402959e+00 -7.57542968e-01 -1.10072792e+00 1.39277592e-01
7.67651200e-01 -1.21841371e+00 -4.51160997e-01 -3.15473884e-01]
|
[4.210254669189453, 2.515092611312866]
|
73db0788-d4a7-406f-a3d3-40db6223dc4a
|
learning-sparse-and-low-rank-priors-for-image
|
2304.10536
| null |
https://arxiv.org/abs/2304.10536v1
|
https://arxiv.org/pdf/2304.10536v1.pdf
|
Learning Sparse and Low-Rank Priors for Image Recovery via Iterative Reweighted Least Squares Minimization
|
We introduce a novel optimization algorithm for image recovery under learned sparse and low-rank constraints, which we parameterize as weighted extensions of the $\ell_p^p$-vector and $\mathcal S_p^p$ Schatten-matrix quasi-norms for $0\!<p\!\le1$, respectively. Our proposed algorithm generalizes the Iteratively Reweighted Least Squares (IRLS) method, used for signal recovery under $\ell_1$ and nuclear-norm constrained minimization. Further, we interpret our overall minimization approach as a recurrent network that we then employ to deal with inverse low-level computer vision problems. Thanks to the convergence guarantees that our IRLS strategy offers, we are able to train the derived reconstruction networks using a memory-efficient implicit back-propagation scheme, which does not pose any restrictions on their effective depth. To assess our networks' performance, we compare them against other existing reconstruction methods on several inverse problems, namely image deblurring, super-resolution, demosaicking and sparse recovery. Our reconstruction results are shown to be very competitive and in many cases outperform those of existing unrolled networks, whose number of parameters is orders of magnitude higher than that of our learned models.
|
['Iaroslav Koshelev', 'Stamatios Lefkimmiatis']
|
2023-04-20
| null | null | null | null |
['demosaicking', 'deblurring']
|
['computer-vision', 'computer-vision']
|
[ 4.42041248e-01 5.42133711e-02 2.57429350e-02 -2.00994551e-01
-9.51080918e-01 1.91544946e-02 2.76554257e-01 -5.07312894e-01
-5.38994014e-01 8.54600370e-01 3.24090004e-01 -1.09264158e-01
-5.85870445e-01 -6.00681782e-01 -9.07641828e-01 -9.03096914e-01
-2.81966478e-01 1.26973525e-01 -2.51320571e-01 -3.47382247e-01
2.89074957e-01 4.70628202e-01 -1.30655622e+00 9.75016505e-02
7.98258364e-01 1.09948659e+00 2.68522501e-01 5.38149059e-01
3.26835662e-01 1.22003651e+00 1.86078802e-01 1.39669385e-02
3.46177340e-01 -4.23830360e-01 -8.94696712e-01 1.98189229e-01
5.44689357e-01 -4.06410158e-01 -6.08609557e-01 1.19536245e+00
4.63884771e-01 4.34127957e-01 6.28471613e-01 -4.18620318e-01
-7.14405358e-01 6.01021588e-01 -8.55870783e-01 3.97622705e-01
2.22365454e-01 -1.23730011e-01 8.26104462e-01 -1.38439322e+00
7.76911855e-01 9.97374117e-01 1.03768623e+00 1.34330958e-01
-1.47729337e+00 -6.02114320e-01 1.80353038e-02 2.97867924e-01
-1.53097451e+00 -9.69199479e-01 8.44017625e-01 -2.92807966e-01
9.12296057e-01 1.94694117e-01 2.49534100e-01 7.36398757e-01
-2.73647103e-02 5.53510189e-01 1.11846757e+00 -4.89907712e-01
1.14674874e-01 -1.04369968e-01 -4.13519740e-02 8.11066568e-01
5.24055995e-02 1.02091640e-01 -5.82662046e-01 -1.50637761e-01
1.19047451e+00 -2.45472625e-01 -7.10791349e-01 -3.56929749e-01
-1.44419920e+00 8.90553176e-01 5.62655330e-01 5.40044546e-01
-6.20023012e-01 4.20387685e-01 5.65251559e-02 2.71597981e-01
8.02969515e-01 2.70963907e-01 -1.12625219e-01 4.28155452e-01
-1.36817813e+00 1.09339185e-01 7.37084270e-01 6.86474323e-01
8.75930727e-01 5.40986836e-01 1.36144280e-01 1.17123783e+00
3.82378131e-01 5.72674274e-01 3.08723658e-01 -1.50892055e+00
2.36926094e-01 -6.46698102e-02 2.85603851e-01 -1.48966539e+00
-2.93592602e-01 -8.10078561e-01 -1.61094511e+00 4.36717533e-02
1.12304285e-01 -3.63245569e-02 -5.89799345e-01 1.93348360e+00
1.67683195e-02 6.78119957e-01 6.71779141e-02 1.11233556e+00
5.67910492e-01 6.81240201e-01 -3.54646474e-01 -6.63209975e-01
1.04781067e+00 -7.30657458e-01 -7.08076358e-01 -4.54818368e-01
3.61470193e-01 -7.19548881e-01 5.40030658e-01 3.75513762e-01
-1.45551109e+00 -3.99877995e-01 -9.07996535e-01 -3.19234021e-02
4.01856840e-01 1.46476850e-01 6.02573156e-01 2.87163287e-01
-1.39658415e+00 9.57667112e-01 -7.54982531e-01 5.57852313e-02
4.32502031e-01 4.06824917e-01 -5.14369845e-01 -2.91626722e-01
-9.99615133e-01 8.73211205e-01 -2.09779665e-02 6.40480757e-01
-8.67764413e-01 -6.37784958e-01 -9.53173339e-01 -1.41008288e-01
2.49904707e-01 -7.67919838e-01 6.94304824e-01 -8.89647067e-01
-1.39856124e+00 1.01363921e+00 -2.95794845e-01 -4.43470091e-01
2.92737514e-01 -1.51184872e-01 -1.95462838e-01 2.27454573e-01
1.57070115e-01 2.96050310e-01 1.22939241e+00 -1.36469805e+00
-8.36120471e-02 -3.00562173e-01 -3.75524729e-01 2.16329411e-01
-1.18127055e-01 -7.58472458e-02 -2.70178884e-01 -8.88985813e-01
7.93188691e-01 -7.32204616e-01 -7.32651174e-01 5.66540435e-02
-3.55315804e-01 4.35659081e-01 2.64088482e-01 -1.10733497e+00
1.06291413e+00 -2.06631875e+00 8.12080264e-01 4.32163358e-01
3.41918528e-01 2.65782345e-02 -3.48944783e-01 2.95881480e-01
-3.87250096e-01 -2.67101467e-01 -7.37450719e-01 -3.93846184e-01
-2.65310168e-01 1.09951124e-01 -3.02290112e-01 9.72684979e-01
-2.27408379e-01 5.21690488e-01 -6.92929745e-01 -1.89829454e-01
9.42800790e-02 9.09495056e-01 -6.61535084e-01 1.22823559e-01
2.76713632e-02 6.79891765e-01 -2.81829715e-01 2.22319275e-01
7.91932702e-01 -5.36247075e-01 2.67694861e-01 -6.22903109e-01
-2.46892616e-01 -1.88851096e-02 -1.55252802e+00 2.10592747e+00
-6.91615403e-01 3.54736030e-01 8.47212851e-01 -1.53112948e+00
6.69670463e-01 2.51241714e-01 6.48338795e-01 -7.10587800e-01
1.42662702e-02 5.00702918e-01 -3.64603847e-01 -4.65844959e-01
4.56484228e-01 -5.20860553e-01 4.07184720e-01 4.28699195e-01
3.31741944e-02 -7.69250318e-02 6.43586963e-02 2.53477633e-01
1.05127156e+00 1.14718780e-01 1.60143062e-01 -6.49292111e-01
8.25648367e-01 -3.18409830e-01 4.50766712e-01 7.92180359e-01
2.10884064e-01 6.29416645e-01 1.99161425e-01 -2.49781460e-01
-1.05109501e+00 -8.37861598e-01 -3.17928612e-01 1.07422042e+00
2.27928758e-02 6.80607930e-02 -5.63294768e-01 -2.05411557e-02
-3.84667248e-01 5.51706076e-01 -3.31218064e-01 2.71490067e-01
-9.09193695e-01 -1.10234237e+00 3.41464430e-01 1.97500557e-01
6.19135022e-01 -1.01008749e+00 -2.21513689e-01 4.20223951e-01
-4.28143740e-01 -1.04810238e+00 -4.76028472e-01 6.69896975e-02
-1.11328578e+00 -9.70761240e-01 -1.09845388e+00 -9.22794580e-01
7.93732524e-01 4.13072467e-01 1.14770484e+00 1.02562070e-01
-1.64037615e-01 5.07630289e-01 -1.39168128e-01 5.27158618e-01
-1.59269929e-01 -1.70128867e-01 1.51511788e-01 2.79760212e-01
-3.63335401e-01 -1.27805936e+00 -8.67561519e-01 2.28131190e-01
-9.56818163e-01 8.22282583e-02 5.73338568e-01 1.14148951e+00
8.44027996e-01 2.32456475e-02 3.40643525e-01 -9.17779207e-01
4.82687682e-01 -4.51662242e-01 -5.17822027e-01 -6.05180077e-02
-4.89355057e-01 2.66325772e-01 5.51730156e-01 -3.04313689e-01
-9.05894339e-01 1.20191328e-01 -3.50922763e-01 -5.94576716e-01
3.10647458e-01 7.39809811e-01 2.68314570e-01 -5.54320216e-01
8.00613344e-01 6.29440367e-01 1.49264932e-01 -6.20733261e-01
5.80130279e-01 2.63397396e-01 7.00605452e-01 -4.68237966e-01
9.50849712e-01 7.99746037e-01 -3.65938358e-02 -1.07003820e+00
-9.11158085e-01 -2.56909102e-01 -4.53543246e-01 2.28292518e-03
6.62095964e-01 -1.18580842e+00 -7.07049012e-01 5.08632064e-01
-1.03791523e+00 -3.07302982e-01 -3.03961128e-01 6.06720567e-01
-8.46129298e-01 7.17985392e-01 -9.16831434e-01 -6.38145566e-01
-4.89855886e-01 -1.06034696e+00 7.49700546e-01 -3.28878134e-01
8.98484364e-02 -9.08446670e-01 3.35122757e-02 3.97001505e-01
8.75014007e-01 1.45599395e-01 7.23829091e-01 1.60968062e-02
-6.26089871e-01 1.17321953e-01 -6.11273646e-01 6.37726665e-01
-2.61165738e-01 -8.00022006e-01 -6.07752860e-01 -6.85852408e-01
5.95799685e-01 -2.73360550e-01 1.12295175e+00 7.08004117e-01
1.12279117e+00 -7.96823204e-01 -1.06344230e-01 1.09299982e+00
1.62041676e+00 -4.40558851e-01 7.79531419e-01 2.08915666e-01
7.03400552e-01 4.16429043e-01 -1.32258505e-01 5.11984408e-01
2.65906602e-01 5.83254576e-01 5.27499199e-01 -8.91474262e-02
-1.62970528e-01 1.93588525e-01 3.68330300e-01 1.23974729e+00
-4.82400209e-01 1.87890470e-01 -6.26252592e-01 5.65937579e-01
-1.79855084e+00 -1.15196490e+00 -1.87537655e-01 2.14461565e+00
8.75267804e-01 -3.51293594e-01 -3.41473401e-01 6.59846514e-02
6.05642736e-01 6.79011345e-01 -5.29423416e-01 7.27259144e-02
-2.48632908e-01 5.88656366e-01 7.27217555e-01 9.40103829e-01
-9.98482525e-01 7.34101772e-01 6.43423986e+00 9.48505044e-01
-9.10309315e-01 3.18307072e-01 6.34054959e-01 -8.16989541e-02
-3.38553667e-01 -1.90569803e-01 -2.04043388e-01 2.89692909e-01
7.52471447e-01 2.21842065e-01 1.05751717e+00 5.53001046e-01
3.36576730e-01 -1.84101630e-02 -9.55947936e-01 1.26683593e+00
3.12170297e-01 -1.62371600e+00 -2.23692834e-01 -8.75613913e-02
7.77556181e-01 2.06666559e-01 1.95844546e-01 -3.02303247e-02
3.59305203e-01 -1.34880877e+00 5.35756171e-01 5.55754304e-01
9.10102367e-01 -5.66530704e-01 3.97281796e-01 4.86777931e-01
-9.83364463e-01 -7.42023252e-03 -4.50075686e-01 -4.43249233e-02
3.94323349e-01 9.42656755e-01 -4.82450463e-02 4.34486359e-01
6.80570722e-01 9.95514452e-01 1.27107546e-01 6.78040266e-01
-1.31217465e-01 3.93266797e-01 -3.95464420e-01 6.53868377e-01
8.34042430e-02 -6.59811199e-01 8.36887717e-01 1.03411555e+00
6.62395656e-01 5.42423546e-01 -1.04786627e-01 1.00216246e+00
-2.46748686e-01 6.29726872e-02 -4.34863865e-01 3.74278754e-01
1.26604870e-01 1.01490271e+00 -4.91108447e-01 -2.23191276e-01
-2.04077795e-01 1.11658835e+00 2.56628454e-01 5.79825282e-01
-5.18764675e-01 -1.20753735e-01 5.69827795e-01 1.95307419e-01
4.71293926e-01 -4.34059650e-01 -8.40241462e-02 -1.64325356e+00
-1.10425934e-01 -1.09352648e+00 1.05413668e-01 -8.29338729e-01
-1.28862166e+00 7.60470986e-01 -1.74525350e-01 -1.11189473e+00
-1.68837190e-01 -3.80593061e-01 -2.37287089e-01 9.76442456e-01
-1.77771771e+00 -8.84645879e-01 -3.24975625e-02 7.82332242e-01
3.44656795e-01 8.09383858e-03 7.26844192e-01 5.98412037e-01
-5.21839797e-01 1.67545497e-01 3.65689874e-01 5.79933822e-02
3.30122113e-01 -6.60489738e-01 -2.28247419e-01 1.00956249e+00
-8.95620286e-02 6.59498990e-01 9.66016591e-01 -3.37669194e-01
-1.74855208e+00 -8.70907247e-01 6.54526472e-01 1.59376115e-01
6.81784093e-01 3.80842313e-02 -1.07277644e+00 9.02241647e-01
2.34622762e-01 3.71072918e-01 3.49939674e-01 -8.51823539e-02
-4.74420726e-01 -8.12252313e-02 -1.15974724e+00 2.44120136e-01
1.24752462e+00 -5.15593648e-01 -3.44068646e-01 5.57514131e-01
2.59983331e-01 -5.48490286e-01 -9.71185923e-01 5.08239269e-01
3.94184917e-01 -1.20567286e+00 1.54549849e+00 -1.59568831e-01
6.61182821e-01 -9.52357650e-02 -4.51173544e-01 -1.13102686e+00
-6.94428802e-01 -5.08087754e-01 -2.95780487e-02 6.60306811e-01
3.63246620e-01 -6.08552217e-01 7.80023813e-01 3.57288599e-01
-1.19777113e-01 -5.81378639e-01 -1.18216693e+00 -3.87519628e-01
1.63726527e-02 -5.98609090e-01 -1.94726381e-02 1.16247475e+00
-2.82809973e-01 1.90545037e-01 -1.00705254e+00 3.47517967e-01
1.26068926e+00 8.75036642e-02 1.92671686e-01 -8.81787717e-01
-6.20147467e-01 -3.49929452e-01 2.89199464e-02 -1.41186976e+00
3.66986901e-01 -1.08793223e+00 1.59990445e-01 -1.34426105e+00
2.05297038e-01 -5.65237045e-01 -2.93820500e-01 2.12353364e-01
2.73866653e-01 6.41491473e-01 9.16521102e-02 5.40651381e-01
-4.90483612e-01 6.32572711e-01 1.24218953e+00 -2.05776885e-01
-9.06408876e-02 -1.79672062e-01 -6.76611781e-01 8.80202472e-01
2.97180504e-01 -3.74635756e-01 -1.37813330e-01 -8.21018815e-01
5.84588051e-01 7.31973112e-01 3.98162007e-01 -9.02451098e-01
4.04765189e-01 -5.36469519e-02 2.91133791e-01 -1.08864479e-01
5.04402876e-01 -6.19816542e-01 4.52256799e-01 3.62068504e-01
-4.20723915e-01 -3.40604246e-01 -1.10914283e-01 5.03057718e-01
-2.25338563e-01 -1.85704216e-01 1.12421727e+00 -4.12352473e-01
-5.18013060e-01 3.15757245e-01 -4.01512831e-01 -1.45769864e-01
3.72366279e-01 -1.21885635e-01 8.75146315e-02 -6.30904138e-01
-9.18757260e-01 -8.23119953e-02 3.11660349e-01 -3.05444837e-01
8.70551109e-01 -1.27134204e+00 -1.10304868e+00 3.31842035e-01
-4.12595570e-01 2.16082763e-02 6.55697465e-01 1.33011687e+00
-6.42095804e-01 6.56547174e-02 -4.01473790e-02 -4.91752416e-01
-7.93460548e-01 3.38228106e-01 5.06054819e-01 -4.66459572e-01
-8.79689932e-01 1.14580297e+00 -2.85609476e-02 -4.04087394e-01
3.53781879e-02 9.07031670e-02 -2.30441138e-01 -5.21758534e-02
5.17312229e-01 6.06295824e-01 -1.13254167e-01 -9.88344908e-01
-3.00160617e-01 9.54978526e-01 1.63636237e-01 -2.03052446e-01
1.91648388e+00 -3.14280421e-01 -8.29861581e-01 -5.16936705e-02
1.33321548e+00 -1.66759582e-03 -1.25217545e+00 -6.05227828e-01
-3.16656500e-01 -3.85190785e-01 4.18163419e-01 -3.43512923e-01
-1.48685181e+00 5.74829102e-01 5.97468078e-01 -7.47775286e-02
1.19619966e+00 -1.94795012e-01 7.18268275e-01 3.60868990e-01
5.02684474e-01 -7.60647297e-01 -7.59494379e-02 6.18227243e-01
1.20490932e+00 -1.15433872e+00 3.61404181e-01 -4.24586445e-01
-1.26632601e-01 8.72410536e-01 -1.12530431e-02 -6.59880638e-01
7.90744662e-01 1.85003951e-01 -2.20672354e-01 -2.03257099e-01
-3.83380026e-01 1.94406509e-02 2.78902948e-01 2.44630486e-01
5.93558967e-01 -3.50780964e-01 -4.54506040e-01 1.47227749e-01
9.28625371e-03 2.52721995e-01 4.35782999e-01 4.80363965e-01
-5.19576371e-01 -9.39236701e-01 -5.52856684e-01 4.72736925e-01
-4.27577168e-01 -3.07534486e-01 4.34376478e-01 4.47354317e-01
-1.30976036e-01 7.43056595e-01 -2.19607264e-01 -2.91096885e-02
1.48536205e-01 -1.91668168e-01 6.25681758e-01 -3.35205168e-01
-2.80035913e-01 2.14740247e-01 4.22814023e-03 -8.06529164e-01
-8.16828609e-01 -5.82326353e-01 -1.06028390e+00 -4.62296188e-01
-1.58165574e-01 1.09926231e-01 4.27306890e-01 8.85313034e-01
3.00143093e-01 3.07792723e-01 6.13693535e-01 -1.45663774e+00
-6.32373333e-01 -9.04402077e-01 -8.13766897e-01 3.46759886e-01
4.80120122e-01 -3.51371855e-01 -5.99608779e-01 8.48274007e-02]
|
[11.55685043334961, -2.29243803024292]
|
d462f77b-c8d0-4409-96bc-bb680537bbe0
|
unsupervised-learning-of-depth-and-ego-motion-2
|
1802.05522
| null |
http://arxiv.org/abs/1802.05522v2
|
http://arxiv.org/pdf/1802.05522v2.pdf
|
Unsupervised Learning of Depth and Ego-Motion from Monocular Video Using 3D Geometric Constraints
|
We present a novel approach for unsupervised learning of depth and ego-motion
from monocular video. Unsupervised learning removes the need for separate
supervisory signals (depth or ego-motion ground truth, or multi-view video).
Prior work in unsupervised depth learning uses pixel-wise or gradient-based
losses, which only consider pixels in small local neighborhoods. Our main
contribution is to explicitly consider the inferred 3D geometry of the scene,
enforcing consistency of the estimated 3D point clouds and ego-motion across
consecutive frames. This is a challenging task and is solved by a novel
(approximate) backpropagation algorithm for aligning 3D structures.
We combine this novel 3D-based loss with 2D losses based on photometric
quality of frame reconstructions using estimated depth and ego-motion from
adjacent frames. We also incorporate validity masks to avoid penalizing areas
in which no useful information exists.
We test our algorithm on the KITTI dataset and on a video dataset captured on
an uncalibrated mobile phone camera. Our proposed approach consistently
improves depth estimates on both datasets, and outperforms the state-of-the-art
for both depth and ego-motion. Because we only require a simple video, learning
depth and ego-motion on large and varied datasets becomes possible. We
demonstrate this by training on the low quality uncalibrated video dataset and
evaluating on KITTI, ranking among top performing prior methods which are
trained on KITTI itself.
|
['Anelia Angelova', 'Reza Mahjourian', 'Martin Wicke']
|
2018-02-15
|
unsupervised-learning-of-depth-and-ego-motion-3
|
http://openaccess.thecvf.com/content_cvpr_2018/html/Mahjourian_Unsupervised_Learning_of_CVPR_2018_paper.html
|
http://openaccess.thecvf.com/content_cvpr_2018/papers/Mahjourian_Unsupervised_Learning_of_CVPR_2018_paper.pdf
|
cvpr-2018-6
|
['depth-and-camera-motion']
|
['computer-vision']
|
[ 2.30116665e-01 -1.25718728e-01 -9.35227349e-02 -4.91353929e-01
-8.74598145e-01 -6.03872240e-01 4.56065029e-01 -3.92258853e-01
-6.89945281e-01 7.59358943e-01 2.14165583e-01 2.21389830e-01
6.61299005e-02 -7.09658742e-01 -1.09266555e+00 -8.09544683e-01
2.70960126e-02 4.13157046e-01 3.95377189e-01 3.84315342e-01
3.54286015e-01 6.16731524e-01 -1.69557643e+00 2.13272050e-01
4.96593744e-01 9.83394027e-01 3.10381144e-01 9.91378009e-01
1.46026880e-01 1.01794434e+00 -1.84112310e-01 -1.20470650e-01
5.92598736e-01 -2.65862107e-01 -6.39755964e-01 4.74964023e-01
1.24992287e+00 -8.50444436e-01 -5.23655653e-01 8.68894160e-01
4.48537022e-01 2.60922313e-01 4.88263726e-01 -9.66741562e-01
1.64237559e-01 -1.45734712e-01 -5.48474133e-01 1.62572592e-01
5.63751400e-01 2.29393482e-01 8.32791150e-01 -9.37815130e-01
1.01566184e+00 1.03847373e+00 7.41212070e-01 3.37329865e-01
-1.19108474e+00 -2.46161908e-01 1.70357808e-01 1.87312588e-01
-1.29824102e+00 -6.74269021e-01 9.07976747e-01 -6.11045003e-01
1.20359015e+00 -1.62595436e-01 5.54639578e-01 8.46715152e-01
9.14911330e-02 6.94146574e-01 9.27562773e-01 -4.34303939e-01
3.03964585e-01 6.96582720e-02 -3.52381319e-01 7.67681897e-01
-2.51912456e-02 3.54897588e-01 -7.71094084e-01 2.21409112e-01
1.06333327e+00 -3.72241512e-02 -5.06379545e-01 -8.71564269e-01
-1.26207840e+00 5.04705429e-01 3.05036455e-01 -1.58035457e-01
-3.49121630e-01 3.88713360e-01 -3.36113758e-02 1.74222916e-01
7.05718577e-01 -5.38711809e-02 -5.26386201e-01 -2.58671016e-01
-1.23361325e+00 -2.09417120e-02 6.36810064e-01 9.74043906e-01
1.33541882e+00 2.08248124e-02 3.13991725e-01 5.55919707e-01
3.27173263e-01 5.46460509e-01 2.37280235e-01 -1.56475663e+00
5.93081176e-01 2.22326651e-01 1.99689284e-01 -8.23988557e-01
-2.10112929e-01 -1.81522936e-01 -4.44227278e-01 6.27818167e-01
6.30928874e-01 -1.01868652e-01 -8.91887605e-01 1.58340061e+00
5.39504707e-01 5.98676324e-01 1.12746367e-02 1.00328910e+00
6.40019298e-01 4.71700966e-01 -6.50257707e-01 -2.06401125e-01
5.29186308e-01 -8.60851169e-01 -3.91235709e-01 -4.73336816e-01
5.59888899e-01 -5.85590959e-01 7.26524591e-01 5.70379436e-01
-1.30303419e+00 -5.13145030e-01 -9.88298774e-01 -4.07695383e-01
-2.07113964e-03 6.80979863e-02 3.57488960e-01 4.96717393e-01
-1.28919387e+00 7.30577469e-01 -1.02592802e+00 -2.58770317e-01
3.56342643e-01 3.94626379e-01 -7.36521125e-01 -4.16083068e-01
-7.79206932e-01 6.79246664e-01 1.66114271e-01 -1.91563033e-02
-1.14756143e+00 -7.22232938e-01 -1.21600294e+00 -3.55868518e-01
3.05874169e-01 -9.61707175e-01 8.80263627e-01 -1.09144902e+00
-1.61086452e+00 1.17677224e+00 -3.24327439e-01 -5.16563058e-01
8.41024518e-01 -4.51601863e-01 2.85489708e-01 6.90593243e-01
4.07040492e-02 9.82103586e-01 9.12695885e-01 -1.30796897e+00
-7.27860391e-01 -3.94056231e-01 3.55802804e-01 5.26280284e-01
-5.92641197e-02 -5.45071125e-01 -6.95571184e-01 -2.40774721e-01
4.86301094e-01 -8.73229086e-01 2.96510546e-03 3.80922109e-01
-1.51343361e-01 4.75830495e-01 8.07141721e-01 -6.87933505e-01
5.79925120e-01 -1.93269980e+00 3.76063257e-01 4.88676615e-02
-7.19241649e-02 -3.06286514e-01 1.18297115e-01 -2.91931350e-02
1.24394387e-01 -4.65660214e-01 -5.64765573e-01 -8.00604284e-01
-4.00317401e-01 4.25378263e-01 -9.12008882e-02 8.62429738e-01
-4.05527987e-02 5.72937727e-01 -9.76870716e-01 -4.51056927e-01
7.76263952e-01 5.81604958e-01 -8.85718524e-01 2.97300607e-01
-1.73467696e-01 7.82921314e-01 1.31332219e-01 6.81741595e-01
8.26454163e-01 -7.44183362e-02 -1.08864367e-01 -2.30757684e-01
-1.51741296e-01 3.76793683e-01 -1.30403852e+00 2.26002622e+00
-5.12393594e-01 8.58387113e-01 1.81796983e-01 -9.15338933e-01
5.44869602e-01 6.05259947e-02 7.24708557e-01 -4.00100023e-01
-1.22624725e-01 2.14238778e-01 -4.41053152e-01 -5.83511591e-01
3.67182165e-01 -2.31748119e-01 3.49738508e-01 2.60495186e-01
3.45980823e-01 -5.48917592e-01 -1.27260685e-01 9.93532836e-02
1.13392913e+00 7.67368078e-01 1.03201801e-02 -2.91773621e-02
5.85315228e-01 -1.15396492e-01 6.64926171e-01 5.02369583e-01
-1.43158093e-01 1.20295203e+00 2.11451739e-01 -2.76503086e-01
-1.08330131e+00 -1.07210410e+00 -1.46382317e-01 4.25683081e-01
3.13585818e-01 -9.65208709e-02 -6.44069612e-01 -7.08628893e-01
-9.11309049e-02 4.56005961e-01 -4.97210592e-01 2.28508547e-01
-5.57191312e-01 -3.88036460e-01 2.20148578e-01 4.95057106e-01
5.76085269e-01 -6.22454166e-01 -7.74213612e-01 4.35315631e-02
-3.49038839e-01 -1.56009281e+00 -3.38191330e-01 1.35877013e-01
-1.25665784e+00 -1.12254441e+00 -6.97406054e-01 -4.65063065e-01
7.29712069e-01 3.75548184e-01 1.20928586e+00 -2.81416625e-01
-3.61294392e-03 9.58562493e-01 -1.14875305e-02 1.52477145e-01
2.08487883e-02 -2.64608920e-01 1.27554581e-01 1.09769449e-01
2.04699993e-01 -9.05715227e-01 -7.34647334e-01 3.26985091e-01
-7.88897455e-01 7.79273063e-02 2.00830206e-01 7.54996479e-01
9.27296996e-01 9.70808323e-03 -2.02979043e-01 -8.32041502e-01
-5.24747550e-01 -3.12518001e-01 -9.21338558e-01 -3.23840767e-01
-2.54539639e-01 -1.24704674e-01 2.21566692e-01 -7.97415599e-02
-1.11218536e+00 5.43076396e-01 -1.84945583e-01 -1.00644231e+00
-2.20429704e-01 8.95193368e-02 -3.67392927e-01 -3.47405970e-01
5.64495683e-01 2.90819913e-01 -8.34824219e-02 -3.09642166e-01
2.37988189e-01 2.80168504e-01 8.00972104e-01 -4.32649285e-01
7.34387994e-01 1.27433312e+00 2.15563580e-01 -9.81465280e-01
-7.36736000e-01 -7.48933554e-01 -1.08248508e+00 -3.30929279e-01
9.80190098e-01 -1.37158859e+00 -3.79264474e-01 6.15937948e-01
-1.20078146e+00 -6.43037498e-01 -2.94984400e-01 8.14367414e-01
-9.90295827e-01 7.63355374e-01 -5.89824438e-01 -7.93052077e-01
-1.33926319e-02 -1.08479142e+00 1.30434251e+00 -2.10471779e-01
1.84867550e-02 -1.28433728e+00 2.33380333e-01 4.61139202e-01
-1.06086321e-01 4.21772420e-01 2.06864536e-01 3.04516673e-01
-1.15051270e+00 5.15532568e-02 -5.95818274e-02 6.95143580e-01
1.20434918e-01 -5.37496991e-02 -1.44799340e+00 -2.70643860e-01
3.36322248e-01 -3.21366251e-01 1.04020655e+00 7.52511501e-01
9.17066932e-01 3.61311436e-02 -3.19633670e-02 1.29044461e+00
1.74166822e+00 -8.11413825e-02 8.37623656e-01 5.11497259e-01
1.19614387e+00 8.69708657e-01 6.52712405e-01 3.52053434e-01
5.39803624e-01 6.56950533e-01 8.12626481e-01 1.08218841e-01
-1.77411020e-01 -3.06923509e-01 6.03558838e-01 5.52587569e-01
-2.20763236e-01 -9.11190137e-02 -6.72327161e-01 6.84346318e-01
-1.74948907e+00 -9.96607780e-01 -1.80555701e-01 2.63689017e+00
5.57996154e-01 2.84215420e-01 -9.32649001e-02 1.77761644e-01
3.53380263e-01 1.91554055e-01 -7.36294150e-01 1.33044392e-01
-4.10854399e-01 2.11243983e-02 7.45176911e-01 1.11345851e+00
-1.14486468e+00 8.08761537e-01 5.80663776e+00 2.73968518e-01
-1.05817139e+00 1.26330465e-01 6.17853582e-01 -5.33675611e-01
-3.37766558e-01 1.89128757e-01 -7.84857810e-01 2.93517768e-01
5.61848700e-01 3.40017706e-01 3.11253905e-01 8.04565847e-01
4.49901104e-01 -5.93942523e-01 -1.51590037e+00 1.40186691e+00
3.03617179e-01 -1.38584447e+00 -2.42034063e-01 1.62864432e-01
1.18699801e+00 4.23492789e-01 -2.84379691e-01 -2.67951727e-01
-9.14739072e-02 -7.50453413e-01 8.99988294e-01 6.16202593e-01
8.61504257e-01 -5.94427109e-01 6.48286760e-01 3.64123762e-01
-9.50596690e-01 1.35828167e-01 -3.67199451e-01 -2.72123218e-01
3.25775385e-01 9.35385525e-01 -3.25553149e-01 5.41517258e-01
8.88904810e-01 1.38918829e+00 -2.28997543e-01 9.37538147e-01
-4.12225276e-01 2.30707765e-01 -6.52182043e-01 7.15291262e-01
2.01908618e-01 -3.49430233e-01 7.36217022e-01 1.04283357e+00
3.02979827e-01 1.10339500e-01 -4.19541001e-02 6.63998604e-01
-6.94506392e-02 -3.34741652e-01 -8.56994212e-01 6.54056668e-01
1.90905467e-01 1.04609549e+00 -5.83333790e-01 -4.31388706e-01
-5.50938785e-01 1.28347611e+00 2.03178555e-01 4.54193950e-01
-5.35813570e-01 7.64390603e-02 7.74863541e-01 4.16288316e-01
5.09593010e-01 -4.52094525e-01 -2.99246699e-01 -1.51293051e+00
1.99535057e-01 -4.33418959e-01 2.98578441e-01 -1.09942031e+00
-1.00263679e+00 3.35145772e-01 8.53841454e-02 -1.52143216e+00
-5.88676631e-01 -6.51044905e-01 -4.10181105e-01 7.28157997e-01
-1.86424112e+00 -6.89562142e-01 -6.56098425e-01 6.58505142e-01
5.69662154e-01 1.59380645e-01 3.13066781e-01 4.27479118e-01
-6.57301694e-02 2.95419544e-01 2.21705139e-01 -1.74991041e-01
8.37752879e-01 -1.24504125e+00 1.90118089e-01 9.88590121e-01
2.82199591e-01 2.12640598e-01 5.27552724e-01 -4.53308642e-01
-1.35798907e+00 -9.40569103e-01 6.39669001e-01 -8.92804086e-01
2.48266146e-01 -3.62511754e-01 -8.52674782e-01 7.75511503e-01
-1.60462722e-01 2.91383713e-01 1.66244581e-01 -4.46266353e-01
-5.70406616e-02 -2.15682581e-01 -1.26115525e+00 1.79075226e-01
1.37450302e+00 -7.71832228e-01 -2.52626419e-01 2.28449315e-01
4.64380294e-01 -7.79789329e-01 -6.24736428e-01 3.62560511e-01
5.28741777e-01 -1.59767354e+00 1.13349974e+00 3.30492146e-02
7.36196637e-01 -4.92712855e-01 -4.83129799e-01 -1.01572537e+00
2.33783424e-01 -5.53394914e-01 -2.42893264e-01 9.74775076e-01
2.65676179e-03 -3.50975543e-01 1.31192350e+00 6.29051328e-01
-1.77798480e-01 -4.92525458e-01 -9.80612457e-01 -5.81332088e-01
-1.61395386e-01 -8.32067966e-01 -8.48832503e-02 1.00386763e+00
-4.47969019e-01 5.64080738e-02 -4.57665473e-01 3.32082719e-01
1.03573477e+00 -1.42603204e-01 1.07946539e+00 -8.38851929e-01
-5.53072512e-01 -6.23509884e-02 -7.62233436e-01 -1.50917721e+00
1.50215149e-01 -6.82027757e-01 1.55171826e-01 -1.48579323e+00
-1.58844277e-01 -1.51657656e-01 1.09615892e-01 1.18312411e-01
1.75022408e-01 4.67459828e-01 -1.80574507e-01 3.53214681e-01
-4.14154083e-01 5.54923594e-01 9.09685075e-01 -2.75188982e-02
-2.68685639e-01 -1.69397235e-01 1.14974543e-01 1.11596024e+00
3.69502723e-01 -3.95638168e-01 -6.31075561e-01 -8.19767594e-01
1.49980217e-01 2.03544959e-01 6.33239865e-01 -1.18967736e+00
3.10753793e-01 3.72369774e-02 6.44957244e-01 -8.27212334e-01
7.29009092e-01 -9.19432759e-01 1.11565217e-01 1.13356605e-01
-3.56944762e-02 -1.24654695e-01 1.46130905e-01 6.91317260e-01
-1.95926890e-01 -2.30312869e-01 8.54713202e-01 -2.76445895e-01
-9.88642097e-01 3.65398645e-01 -3.90306264e-02 5.11768572e-02
7.69084692e-01 -7.51317382e-01 2.11192563e-01 -7.13098109e-01
-6.33437872e-01 3.09150685e-02 1.02658689e+00 3.43330167e-02
9.34327006e-01 -1.09870124e+00 -5.92047930e-01 3.96738589e-01
5.45446537e-02 5.39599299e-01 3.44494730e-01 7.90358722e-01
-1.05445075e+00 1.92802906e-01 -1.97825506e-01 -1.19849908e+00
-1.18030953e+00 2.66945273e-01 6.48302078e-01 5.66550270e-02
-8.12986672e-01 1.02126026e+00 4.93315488e-01 -4.87390161e-01
4.68743354e-01 -4.44136560e-01 1.47098020e-01 -2.00569183e-01
3.77130598e-01 4.28280413e-01 1.21308446e-01 -8.20735991e-01
-3.89612734e-01 1.04649162e+00 1.60278678e-01 -4.27406341e-01
1.30024850e+00 -5.45159221e-01 3.26549970e-02 5.77684164e-01
1.47383082e+00 1.29475474e-01 -2.16698766e+00 -2.11836398e-01
-3.08819622e-01 -9.08403337e-01 3.75283420e-01 -2.93231219e-01
-1.14471853e+00 1.02546239e+00 6.60362720e-01 -5.31327307e-01
1.14654410e+00 -1.99203253e-01 5.95279813e-01 3.29054296e-01
6.11665845e-01 -1.05177307e+00 1.40431195e-01 5.20128787e-01
4.37718630e-01 -1.56198323e+00 3.06870192e-01 -4.14175928e-01
-2.29113355e-01 1.00332201e+00 6.00861847e-01 -2.81658798e-01
5.50400496e-01 7.58107603e-02 9.71690118e-02 2.82812808e-02
-5.94245076e-01 -2.32965708e-01 2.24575326e-01 6.84015334e-01
1.92906916e-01 -5.76472104e-01 2.53244340e-01 -3.62042964e-01
-4.23684493e-02 5.64161241e-02 7.08757997e-01 9.48590636e-01
-2.68968731e-01 -8.42131972e-01 -5.06480098e-01 1.05189614e-01
-2.75824428e-01 -1.19579539e-01 -1.43812060e-01 7.02244520e-01
1.96137100e-01 7.92819738e-01 2.94396281e-01 -3.02682251e-01
1.52588204e-01 -1.83730334e-01 9.70761538e-01 -6.19295835e-01
1.85264163e-02 2.75711358e-01 9.54941735e-02 -8.62681448e-01
-9.58949924e-01 -7.85828888e-01 -1.14612889e+00 -2.59606957e-01
2.76830699e-03 -2.71204114e-01 7.28147328e-01 1.00241160e+00
1.31769404e-01 1.33589119e-01 7.02235341e-01 -1.48901093e+00
8.54078084e-02 -6.26635253e-01 -4.67807919e-01 5.32409370e-01
6.36929393e-01 -6.53551877e-01 -8.04237962e-01 4.60821867e-01]
|
[8.590929985046387, -2.360869884490967]
|
5f73452f-2e98-4d8f-bf93-876c87b54632
|
moviepuzzle-visual-narrative-reasoning
|
2306.02252
| null |
https://arxiv.org/abs/2306.02252v2
|
https://arxiv.org/pdf/2306.02252v2.pdf
|
MoviePuzzle: Visual Narrative Reasoning through Multimodal Order Learning
|
We introduce MoviePuzzle, a novel challenge that targets visual narrative reasoning and holistic movie understanding. Despite the notable progress that has been witnessed in the realm of video understanding, most prior works fail to present tasks and models to address holistic video understanding and the innate visual narrative structures existing in long-form videos. To tackle this quandary, we put forth MoviePuzzle task that amplifies the temporal feature learning and structure learning of video models by reshuffling the shot, frame, and clip layers of movie segments in the presence of video-dialogue information. We start by establishing a carefully refined dataset based on MovieNet by dissecting movies into hierarchical layers and randomly permuting the orders. Besides benchmarking the MoviePuzzle with prior arts on movie understanding, we devise a Hierarchical Contrastive Movie Clustering (HCMC) model that considers the underlying structure and visual semantic orders for movie reordering. Specifically, through a pairwise and contrastive learning approach, we train models to predict the correct order of each layer. This equips them with the knack for deciphering the visual narrative structure of movies and handling the disorder lurking in video data. Experiments show that our approach outperforms existing state-of-the-art methods on the \MoviePuzzle benchmark, underscoring its efficacy.
|
['Zilong Zheng', 'Dongyan Zhao', 'Yuxuan Wang', 'Jianghui Wang']
|
2023-06-04
| null | null | null | null |
['video-understanding']
|
['computer-vision']
|
[ 2.20536128e-01 -1.08731762e-01 -3.25788677e-01 -2.52732188e-01
-2.23643586e-01 -1.05613756e+00 7.84924090e-01 -7.08409250e-02
-4.90966439e-02 1.39504299e-01 7.62191355e-01 -2.71462470e-01
-1.75972849e-01 -2.51572847e-01 -8.52513909e-01 -3.04187328e-01
-3.81606743e-02 3.05343240e-01 1.68604776e-01 -1.70722753e-01
4.80062872e-01 -2.66285278e-02 -1.68542337e+00 1.29532707e+00
2.95646220e-01 1.07092369e+00 2.28347152e-01 8.91307533e-01
-8.99451897e-02 1.67806649e+00 -4.11633998e-01 -5.43477297e-01
1.34556487e-01 -7.67449498e-01 -1.14234626e+00 7.00330794e-01
8.55884671e-01 -5.36192894e-01 -6.31294131e-01 6.16902530e-01
-1.49976313e-01 3.13747734e-01 6.32938027e-01 -1.15709209e+00
-5.27207136e-01 8.80043030e-01 -4.26554948e-01 5.81186593e-01
6.42809272e-01 1.70336351e-01 1.37724411e+00 -9.21679735e-01
1.29874492e+00 1.18646348e+00 5.98740816e-01 4.94383633e-01
-1.10144043e+00 -3.08772564e-01 5.94071507e-01 1.06278312e+00
-1.10310709e+00 -5.02106726e-01 8.85281444e-01 -1.03553736e+00
9.25926805e-01 3.41172516e-01 8.53543282e-01 1.40766394e+00
-1.52036712e-01 9.88660514e-01 8.20634604e-01 -8.62327069e-02
7.44486600e-02 -1.45137176e-01 1.06773809e-01 6.30705774e-01
-7.41468251e-01 -1.79307252e-01 -9.74556327e-01 5.75947821e-01
6.14647210e-01 2.04210915e-02 -2.40781516e-01 -6.10307157e-01
-1.30122864e+00 5.51848412e-01 1.62751824e-01 3.20853680e-01
-2.46079460e-01 9.90588404e-03 7.52394795e-01 2.31326476e-01
2.31919080e-01 5.09478867e-01 -1.55057788e-01 -4.12047029e-01
-1.11836672e+00 3.47526014e-01 6.20594025e-01 6.90066695e-01
2.57019162e-01 -1.61717504e-01 -2.98170567e-01 5.50385237e-01
-1.13540336e-01 -2.00572938e-01 5.38970418e-02 -1.35144615e+00
5.12234807e-01 5.42109013e-01 -1.79005474e-01 -1.29064488e+00
-3.35897267e-01 -3.10123395e-02 -6.23008609e-01 -1.35028511e-01
5.07088184e-01 2.48447716e-01 -6.52713776e-01 1.43270719e+00
1.34125188e-01 6.30400598e-01 -1.86219856e-01 1.00291312e+00
1.02627671e+00 9.18498397e-01 -1.61597214e-03 -3.41666847e-01
1.35983086e+00 -1.25262523e+00 -7.17682302e-01 1.83183029e-02
3.08603853e-01 -4.66279954e-01 1.32831717e+00 8.64837050e-01
-1.11139607e+00 -7.95086801e-01 -1.10318506e+00 -4.41590160e-01
-1.37466624e-01 -1.98004711e-02 4.02573824e-01 1.10154457e-01
-7.94746757e-01 7.38202512e-01 -6.49872959e-01 -2.62247086e-01
6.07952476e-01 -1.70629367e-01 -5.75795710e-01 -3.89619857e-01
-8.47932041e-01 6.11896098e-01 4.69182253e-01 5.75274006e-02
-1.43054998e+00 -9.65695620e-01 -8.97026837e-01 4.58455086e-02
8.12925160e-01 -4.99004066e-01 9.96612430e-01 -1.29384613e+00
-1.24963820e+00 8.36218417e-01 -9.72635001e-02 -4.62272793e-01
3.52224320e-01 -1.73796520e-01 -1.32100075e-01 6.71609521e-01
-1.17600106e-01 6.39159679e-01 1.00460875e+00 -1.50467157e+00
-5.66352665e-01 -1.88447073e-01 5.25535762e-01 2.89848536e-01
-3.85191232e-01 2.59670109e-01 -6.38137877e-01 -9.06857848e-01
-2.43223727e-01 -6.53798282e-01 9.72144827e-02 -3.69397402e-01
-4.81791973e-01 -9.01570544e-02 7.40414143e-01 -9.82003093e-01
1.65255499e+00 -2.24942470e+00 9.67836559e-01 -2.12406188e-01
5.91767907e-01 -1.56906605e-01 -1.78165883e-01 6.24036789e-01
-1.19974829e-01 1.41179085e-01 -2.80236118e-02 -6.16552114e-01
2.75649857e-02 7.63843432e-02 -5.36478043e-01 2.10608840e-01
5.53861521e-02 8.18165064e-01 -8.69185090e-01 -5.19660056e-01
2.25416377e-01 2.86802620e-01 -1.00096548e+00 3.91630024e-01
-4.91719395e-01 7.10213661e-01 9.67307240e-02 5.32029390e-01
1.26834050e-01 -5.88032782e-01 5.57518899e-01 -5.24556518e-01
-7.75491372e-02 1.79169014e-01 -9.69248116e-01 1.91201174e+00
-6.66525587e-02 9.59630251e-01 -1.27157658e-01 -1.17868876e+00
3.06569219e-01 1.89489841e-01 5.64031303e-01 -6.53074801e-01
1.37783557e-01 -4.85570788e-01 -2.20899880e-01 -1.13018072e+00
5.79077661e-01 -2.54999697e-01 -2.56956071e-01 2.78133661e-01
4.43525046e-01 1.15081705e-01 5.73100865e-01 6.92430377e-01
9.35930669e-01 5.28020084e-01 3.65112908e-02 1.57112643e-01
4.21807021e-01 2.56753922e-01 2.39550710e-01 4.71905649e-01
-2.41906326e-02 7.69013166e-01 1.01392388e+00 -7.29613125e-01
-1.13145781e+00 -1.04570663e+00 3.79457027e-01 1.64404130e+00
2.73807615e-01 -1.08668923e+00 -7.47871637e-01 -7.63968527e-01
-3.51229161e-01 5.76530516e-01 -9.88067269e-01 1.27074599e-01
-6.49499357e-01 -2.50046283e-01 2.18124881e-01 5.65964401e-01
6.86377659e-02 -9.36048090e-01 -3.84392500e-01 4.83015133e-03
-5.61680496e-01 -1.53263736e+00 -4.44457084e-01 -3.06622982e-01
-3.83104086e-01 -1.42225206e+00 -5.29806018e-02 -5.92140138e-01
3.29423308e-01 1.99754760e-01 1.32692122e+00 7.99027681e-02
-1.43330485e-01 6.77955985e-01 -7.68017173e-01 1.85009018e-01
-4.20968324e-01 -1.63597569e-01 -1.39537588e-01 3.15033555e-01
1.50757715e-01 -5.49087584e-01 -6.22932374e-01 1.90692171e-01
-9.60417986e-01 4.39477324e-01 1.93891600e-02 5.58719158e-01
4.82847989e-01 7.08065256e-02 2.32871249e-01 -8.11087489e-01
2.02500433e-01 -7.11487710e-01 -9.27882195e-02 3.32165271e-01
-1.44301623e-01 -3.46618146e-01 8.40041876e-01 -4.55547214e-01
-8.95749986e-01 7.64381066e-02 1.71077549e-01 -8.61510098e-01
-2.79108942e-01 3.46226960e-01 -1.29625708e-01 4.44202542e-01
3.27748597e-01 1.56250104e-01 -2.17858896e-01 -5.69422960e-01
8.57430339e-01 1.72242373e-01 1.01522946e+00 -4.10654068e-01
5.97201109e-01 7.16825247e-01 -3.28841895e-01 -8.97031367e-01
-1.08740199e+00 -4.80575502e-01 -1.09529674e+00 -8.88718367e-01
1.53873956e+00 -9.22425568e-01 -8.71799111e-01 1.31400093e-01
-1.08360016e+00 -2.40580067e-01 -1.53666828e-02 1.23903356e-01
-7.49908805e-01 6.87236786e-01 -7.04723060e-01 -4.29052979e-01
3.17333907e-01 -1.04653466e+00 8.10437918e-01 -1.55066311e-01
-5.37323534e-01 -9.20539141e-01 -6.92387670e-02 1.03811491e+00
-2.18445823e-01 5.24760067e-01 1.15428829e+00 -5.34682214e-01
-6.63677871e-01 2.42219850e-01 -9.06258374e-02 1.12464838e-01
-4.12109286e-01 1.47920221e-01 -7.94374406e-01 -9.03477445e-02
-8.40186924e-02 -5.44769049e-01 1.00453734e+00 2.94547439e-01
1.33572114e+00 -5.31734645e-01 3.05175073e-02 6.75056577e-01
1.10273242e+00 1.29747733e-01 5.39512277e-01 3.85552168e-01
1.07699299e+00 9.31974828e-01 4.55640614e-01 6.58808231e-01
8.72436106e-01 7.00471818e-01 8.00809026e-01 1.99793547e-01
-2.11582437e-01 -6.39555097e-01 5.30943930e-01 8.53278399e-01
-2.09177211e-01 -3.11403185e-01 -8.37794840e-01 6.34311616e-01
-2.01487422e+00 -1.35768044e+00 -6.20371215e-02 1.57888639e+00
5.52322328e-01 3.59011143e-02 4.58387852e-01 1.52964011e-01
3.94416541e-01 6.62711978e-01 -2.31655091e-01 -3.02796096e-01
-1.80200115e-01 -2.09845364e-01 -1.70601279e-01 4.49330986e-01
-1.25974691e+00 9.70534861e-01 5.84447241e+00 6.76437855e-01
-8.24156582e-01 5.36382794e-02 7.19922185e-01 -4.99032170e-01
-2.42386803e-01 1.35979414e-01 -4.56348419e-01 3.35686505e-01
5.92742801e-01 2.20640436e-01 7.48122573e-01 5.24302483e-01
4.17683184e-01 5.93919419e-02 -1.49207795e+00 9.52367842e-01
7.57027209e-01 -1.88476801e+00 3.30298483e-01 -4.15719658e-01
6.71447635e-01 -3.94746393e-01 -9.89332721e-02 3.43339086e-01
-5.66672459e-02 -1.36735976e+00 1.07848918e+00 6.12525284e-01
5.92898309e-01 -4.25734729e-01 3.80840003e-01 2.51621097e-01
-1.17411160e+00 -4.25171703e-01 6.17933385e-02 -4.15989548e-01
4.31656122e-01 -2.50773281e-01 -5.93713045e-01 4.94726628e-01
9.09825683e-01 1.39087331e+00 -7.93198764e-01 5.20435750e-01
-1.19321600e-01 3.96432877e-01 3.57760429e-01 3.66018474e-01
2.87854701e-01 -1.36054128e-01 5.51827013e-01 1.15307498e+00
-1.07836165e-01 3.25233966e-01 2.48741642e-01 6.36522412e-01
-7.52001852e-02 -2.81371502e-03 -3.93612087e-01 -3.35408449e-01
2.53147483e-01 1.15408218e+00 -8.76993716e-01 -4.31020230e-01
-3.97865832e-01 1.07143509e+00 2.46024549e-01 2.72582084e-01
-9.53813195e-01 4.31703895e-01 4.17196900e-01 3.70108604e-01
7.52027452e-01 -3.31333190e-01 -1.86259776e-01 -1.38592052e+00
-9.53649133e-02 -1.23362184e+00 6.69699252e-01 -9.29369509e-01
-1.20931137e+00 7.11811543e-01 2.26670250e-01 -1.16138387e+00
-3.57508391e-01 -5.94981670e-01 -5.95971704e-01 -2.26862818e-01
-1.02932036e+00 -1.34963059e+00 -3.72359365e-01 7.73829222e-01
1.28966260e+00 4.31180485e-02 2.68116027e-01 1.84451550e-01
-5.29649913e-01 3.09865057e-01 -1.95915505e-01 1.32450715e-01
5.18060684e-01 -1.09536326e+00 1.11047380e-01 9.38396215e-01
5.53953230e-01 3.19204032e-01 9.29242671e-01 -4.05661970e-01
-1.45720220e+00 -7.20279336e-01 6.07953906e-01 -1.02690458e+00
1.16013277e+00 -8.45396161e-01 -7.45493233e-01 7.65553296e-01
5.65252841e-01 -2.80789047e-01 8.60206783e-01 9.85926688e-02
-7.87718236e-01 1.66050032e-01 -3.55355084e-01 5.37480772e-01
1.33076036e+00 -8.36566508e-01 -8.89283597e-01 3.58910561e-01
7.21673489e-01 -2.53049463e-01 -1.00791466e+00 3.81034523e-01
7.34983265e-01 -1.16453004e+00 1.00041437e+00 -1.19568455e+00
1.39886320e+00 -1.93108246e-01 -2.70507336e-01 -8.91307831e-01
-1.98718891e-01 -6.60394967e-01 -3.19973797e-01 1.38690579e+00
2.18640164e-01 6.46917820e-01 7.47206450e-01 2.44776607e-01
-1.80762932e-01 -7.32001007e-01 -6.77591503e-01 -3.70869666e-01
-1.73892200e-01 -6.38188422e-01 3.56372893e-02 1.19307399e+00
4.25186843e-01 6.38679445e-01 -8.96544397e-01 -5.96928895e-02
3.47963601e-01 2.31626540e-01 8.33148420e-01 -8.07980955e-01
-5.97906470e-01 -5.64471841e-01 -1.99512869e-01 -1.16672087e+00
2.44116277e-01 -7.23131359e-01 -1.11531042e-01 -1.48501945e+00
5.48893631e-01 3.56571496e-01 -1.92470998e-01 2.44126827e-01
-8.24929997e-02 5.21366656e-01 8.33799303e-01 2.64495820e-01
-1.44631004e+00 3.61439407e-01 1.13876319e+00 -1.60756618e-01
4.96441685e-02 -5.11214793e-01 -5.30865371e-01 8.85019243e-01
1.73718378e-01 -8.82252902e-02 -6.13497555e-01 -5.82136095e-01
5.70474982e-01 1.90334186e-01 5.88151157e-01 -7.62929738e-01
2.14577973e-01 -1.25403732e-01 3.11652154e-01 -6.90568447e-01
3.72905970e-01 -6.55210376e-01 1.81015551e-01 -9.17967632e-02
-5.75576127e-01 2.69462734e-01 -3.03273723e-02 7.26857841e-01
-4.31762308e-01 1.96107879e-01 3.76634419e-01 -1.63017616e-01
-1.28310621e+00 2.23784909e-01 -5.24170637e-01 1.72723427e-01
1.07172835e+00 -4.64875966e-01 -5.03871739e-01 -5.36380947e-01
-1.14097035e+00 3.75012577e-01 5.48342943e-01 9.00890172e-01
7.27997541e-01 -1.11354661e+00 -5.44072628e-01 -7.16021284e-02
2.21393853e-01 -2.93406844e-01 9.36024189e-01 8.78436863e-01
-4.14860249e-01 1.06395192e-01 -3.89092743e-01 -6.14885092e-01
-1.32803512e+00 1.03284633e+00 1.01468898e-01 -1.14843592e-01
-6.59356177e-01 9.35714960e-01 5.33743322e-01 1.47785097e-01
3.60917658e-01 -2.15473875e-01 -8.40905488e-01 6.05239689e-01
6.03088617e-01 3.23410332e-01 -3.72420639e-01 -8.47895741e-01
-1.43890589e-01 5.83948016e-01 -7.16898665e-02 6.42019585e-02
1.49528134e+00 -4.86054599e-01 -1.65106848e-01 6.80456698e-01
1.14146054e+00 -1.00409515e-01 -1.70677364e+00 7.11431652e-02
1.15034468e-01 -3.56582582e-01 -3.97422463e-01 -6.23494804e-01
-7.85010159e-01 9.72737312e-01 -4.45405692e-02 3.25306177e-01
1.18377388e+00 4.37920749e-01 6.85126603e-01 -3.18659209e-02
-2.27999970e-01 -1.06532383e+00 6.63303435e-01 6.50747240e-01
9.15136695e-01 -1.09935772e+00 7.74948066e-03 -2.89609462e-01
-1.15909803e+00 1.11202133e+00 5.83763480e-01 -1.95982426e-01
4.07782614e-01 6.87050223e-02 -4.48759906e-02 -5.41596591e-01
-1.11328363e+00 -2.32414659e-02 5.95072329e-01 3.45864207e-01
1.36440203e-01 -1.80648163e-01 -1.14323199e-01 1.10530877e+00
-2.38349721e-01 -2.26033106e-01 5.54961443e-01 4.00446743e-01
-2.59098023e-01 -6.80059552e-01 -1.17484085e-01 -1.00381505e-02
-5.51033318e-01 -1.02557771e-01 -7.81767607e-01 7.66799986e-01
4.75095421e-01 9.27444756e-01 2.97191441e-01 -7.90036440e-01
1.67234927e-01 4.46609873e-03 5.31876922e-01 -4.88558322e-01
-6.47499919e-01 1.11187451e-01 3.97315860e-01 -8.69372606e-01
-7.24879026e-01 -6.54793859e-01 -9.52297032e-01 -1.86214522e-01
3.91815484e-01 -1.07213080e-01 2.81178027e-01 1.33666515e+00
2.66001225e-01 6.25577807e-01 5.64757943e-01 -1.07822311e+00
5.85988909e-02 -6.29725277e-01 -2.49162257e-01 9.56386745e-01
3.20964813e-01 -5.59556246e-01 -2.63280034e-01 7.10910857e-01]
|
[10.222158432006836, 0.8335487842559814]
|
7bd4ac95-fb4e-48c8-9e3a-119f1547248b
|
comparison-and-analysis-of-deep-audio
|
2104.06517
| null |
https://arxiv.org/abs/2104.06517v1
|
https://arxiv.org/pdf/2104.06517v1.pdf
|
Comparison and Analysis of Deep Audio Embeddings for Music Emotion Recognition
|
Emotion is a complicated notion present in music that is hard to capture even with fine-tuned feature engineering. In this paper, we investigate the utility of state-of-the-art pre-trained deep audio embedding methods to be used in the Music Emotion Recognition (MER) task. Deep audio embedding methods allow us to efficiently capture the high dimensional features into a compact representation. We implement several multi-class classifiers with deep audio embeddings to predict emotion semantics in music. We investigate the effectiveness of L3-Net and VGGish deep audio embedding methods for music emotion inference over four music datasets. The experiments with several classifiers on the task show that the deep audio embedding solutions can improve the performances of the previous baseline MER models. We conclude that deep audio embeddings represent musical emotion semantics for the MER task without expert human engineering.
|
['Shlomo Dubnov', 'Eunjeong Koh']
|
2021-04-13
| null | null | null | null |
['music-emotion-recognition']
|
['music']
|
[-2.03776807e-01 -2.24294081e-01 2.25498229e-01 -2.25928113e-01
-7.64189243e-01 -5.47001958e-01 2.14082435e-01 -2.10124720e-02
-2.83019662e-01 9.10514668e-02 5.29547930e-01 3.56139779e-01
-2.62068331e-01 -5.98389983e-01 -4.64706063e-01 -5.10964990e-01
-2.40404427e-01 3.25477034e-01 -2.82648265e-01 -4.10410315e-01
1.64140970e-01 2.73629099e-01 -2.10621476e+00 7.32370734e-01
1.06725223e-01 1.52177560e+00 -3.34190458e-01 1.00506139e+00
-1.15849234e-01 8.14084411e-01 -7.76243269e-01 -4.14695829e-01
1.72910485e-02 -1.94280103e-01 -5.91377676e-01 -4.59630936e-01
2.60226607e-01 2.62732282e-02 -2.66789854e-01 8.32016349e-01
9.30983961e-01 3.22999328e-01 8.80301416e-01 -1.43784857e+00
-8.43401790e-01 8.79572213e-01 -6.85325116e-02 -1.41109496e-01
3.76240760e-01 -2.46671930e-01 1.59970999e+00 -1.06387186e+00
3.28150600e-01 1.20908225e+00 1.02802968e+00 3.75177532e-01
-1.01524627e+00 -8.19089770e-01 -7.86907673e-02 7.40601063e-01
-1.32764411e+00 -2.75780737e-01 1.23268545e+00 -6.15717769e-01
1.12885988e+00 4.67375547e-01 1.11528575e+00 1.40053213e+00
8.20015464e-03 8.65685701e-01 6.15251124e-01 -4.11790609e-01
1.48577869e-01 3.66144031e-02 1.73038766e-01 4.10557210e-01
-5.79897463e-01 1.08916178e-01 -1.05574489e+00 -5.03320098e-01
5.55704892e-01 -1.10458672e-01 -7.37585276e-02 -1.91958219e-01
-1.23825538e+00 1.04670024e+00 3.45843315e-01 4.74425435e-01
-5.53121150e-01 7.73270965e-01 1.10057235e+00 7.54419804e-01
5.70331872e-01 7.34659791e-01 -6.10405862e-01 -7.91315377e-01
-9.07237232e-01 2.49901921e-01 6.85495794e-01 5.57969034e-01
3.56700897e-01 3.19012135e-01 -1.29317984e-01 1.13196504e+00
2.91130930e-01 -6.69573620e-02 8.12202930e-01 -1.01029658e+00
-2.61894614e-01 5.42096078e-01 -2.35754490e-01 -1.07729363e+00
-1.31099820e-01 -5.60228825e-01 -7.87372291e-01 2.91748792e-01
-6.61907867e-02 1.66426897e-01 -2.40695357e-01 1.57868862e+00
2.01111939e-02 7.31395543e-01 8.76851678e-02 8.73853087e-01
7.17051327e-01 8.60130191e-01 -1.03653036e-02 2.60720253e-01
1.52221334e+00 -1.12800729e+00 -9.21709359e-01 3.12870592e-01
5.19749403e-01 -1.03278530e+00 1.59598291e+00 9.88819361e-01
-8.30446720e-01 -9.96170878e-01 -1.17408526e+00 -3.51090908e-01
-5.59524000e-01 4.89614844e-01 8.27173233e-01 5.97272277e-01
-7.57483542e-01 9.06627953e-01 -5.06336153e-01 1.14716746e-01
2.95523137e-01 3.61434191e-01 -3.36940289e-01 3.85030001e-01
-1.34519029e+00 2.91768163e-01 2.51496375e-01 1.44560918e-01
-1.08012497e+00 -9.12707806e-01 -7.55425155e-01 4.57837611e-01
-1.50862008e-01 -7.31234431e-01 1.31023741e+00 -1.16764593e+00
-1.78652370e+00 7.72279143e-01 1.94561481e-01 -2.51839846e-01
-3.71671584e-03 -6.32827640e-01 -5.06199002e-01 1.56518430e-01
-2.24134773e-01 6.48274362e-01 1.01645207e+00 -7.82739758e-01
-3.04025292e-01 -3.61805230e-01 -2.21627384e-01 -7.50458613e-03
-7.60327280e-01 1.68083355e-01 -2.02102121e-02 -1.21873248e+00
-3.61605525e-01 -9.86814022e-01 9.93618816e-02 6.98447749e-02
-1.80814728e-01 -5.31169593e-01 8.59559774e-01 -4.22100216e-01
1.26429677e+00 -2.77432537e+00 4.44142342e-01 1.65653955e-02
-1.13141045e-01 -6.48376122e-02 -4.39950556e-01 6.24075234e-01
-3.09361279e-01 -1.02130741e-01 1.21494442e-01 -4.86244023e-01
8.35002780e-01 1.05422400e-01 -8.79196525e-01 5.80689311e-02
1.28696471e-01 9.66695666e-01 -5.36324561e-01 -1.96309745e-01
1.05646424e-01 9.54276741e-01 -8.77865076e-01 4.00982559e-01
-2.67070949e-01 3.32306400e-02 -1.03507973e-01 5.90252399e-01
5.91648743e-02 1.71319649e-01 -1.62127852e-01 -4.88410681e-01
1.85028896e-01 3.52922678e-01 -1.43647265e+00 2.35523152e+00
-5.01834393e-01 6.69784606e-01 -2.18182385e-01 -8.81797075e-01
1.22724164e+00 7.64948249e-01 6.34149492e-01 -1.65631965e-01
2.98658848e-01 1.51904210e-01 -1.97949857e-01 -3.79547268e-01
7.17833936e-01 -6.73605382e-01 -5.77172041e-01 6.32872462e-01
5.64479530e-01 -3.43447141e-02 -4.04322118e-01 -2.24891633e-01
9.20898974e-01 -1.03362359e-01 -1.16200214e-02 1.81270298e-02
3.79796296e-01 -3.43625724e-01 5.31909049e-01 3.90758753e-01
-7.61729851e-02 4.93425757e-01 4.58518714e-01 -5.46689391e-01
-8.05728018e-01 -9.64729428e-01 -5.19391373e-02 1.81040752e+00
-4.56514210e-01 -1.07012594e+00 -3.18285465e-01 -3.82242471e-01
1.64887369e-01 3.90647799e-01 -7.55153894e-01 -5.00724256e-01
-1.41808823e-01 -2.57459283e-01 1.11980414e+00 8.48253071e-01
-1.38087973e-01 -1.53080153e+00 -3.13715190e-01 5.84673762e-01
-1.71369985e-01 -7.37545371e-01 -2.46942177e-01 3.84704083e-01
-5.89310408e-01 -8.57119739e-01 -6.76984906e-01 -9.86343682e-01
-4.41790670e-01 -4.84757483e-01 1.14914560e+00 -2.84864217e-01
-6.13023758e-01 6.55100763e-01 -6.34294152e-01 -8.60247135e-01
-1.46608472e-01 6.88923597e-02 1.33708358e-01 1.65984020e-01
6.01145864e-01 -9.56981063e-01 -4.67308879e-01 5.14092445e-02
-8.87022734e-01 -4.90667284e-01 1.46116242e-01 8.51984441e-01
8.10159624e-01 1.51718602e-01 9.52762365e-01 -3.78527343e-01
9.85740960e-01 -1.26539156e-01 1.62283540e-01 -5.61522730e-02
-1.27238393e-01 -4.64350469e-02 3.93748879e-01 -9.03807282e-01
-3.29666287e-01 1.22312233e-01 -5.62061787e-01 -8.17487061e-01
-1.97787464e-01 5.78493237e-01 -1.12976968e-01 1.57412514e-01
5.05071878e-01 -1.83457788e-02 -2.29307771e-01 -8.46650481e-01
7.81423151e-01 9.86422062e-01 5.70621610e-01 -7.52817750e-01
1.27263993e-01 2.50192612e-01 -2.74822593e-01 -8.11359465e-01
-8.30606580e-01 -5.30583143e-01 -4.39013273e-01 -2.02087268e-01
9.78390038e-01 -1.19418395e+00 -9.75650847e-01 1.31548598e-01
-1.03550005e+00 -1.95072547e-01 -8.21195662e-01 6.57837391e-01
-1.15010881e+00 3.22899550e-01 -1.01450086e+00 -8.28647852e-01
-6.01964355e-01 -9.71925020e-01 1.51824534e+00 -4.22098011e-01
-7.20058858e-01 -7.51373827e-01 6.14376545e-01 4.25670967e-02
1.52010262e-01 1.95829228e-01 1.16151297e+00 -4.75914657e-01
-1.44622266e-01 -2.02810869e-01 1.55747846e-01 6.14003956e-01
-3.27261686e-01 2.08451157e-03 -1.74153435e+00 1.88516136e-02
-1.46164581e-01 -8.19934607e-01 1.08551776e+00 2.95489669e-01
1.60870051e+00 7.78304040e-02 3.69624943e-01 7.18805492e-01
1.19337606e+00 -1.51829079e-01 7.13042796e-01 3.10936123e-01
6.08118951e-01 3.64381045e-01 7.13901639e-01 8.38626742e-01
1.18481576e-01 8.09818864e-01 3.93531233e-01 2.40489185e-01
6.42236471e-02 -3.29476416e-01 6.47113740e-01 1.44331789e+00
-2.87595261e-02 8.77503976e-02 -4.50172871e-01 5.59502304e-01
-2.00854540e+00 -9.89911199e-01 2.37924829e-01 1.50198650e+00
8.58606160e-01 -3.30961376e-01 3.08678359e-01 9.86075222e-01
2.68507332e-01 1.10395811e-01 -2.40089387e-01 -8.94848585e-01
9.15604644e-03 6.87908292e-01 -5.54726303e-01 9.90508646e-02
-1.15455329e+00 9.94616985e-01 6.49339151e+00 8.58993709e-01
-1.09053719e+00 2.83165544e-01 -2.71737665e-01 -3.47826451e-01
-2.54948109e-01 -3.40738595e-01 -3.38183165e-01 1.04301430e-01
1.20370924e+00 3.05591412e-02 4.98911202e-01 1.05063009e+00
-2.82878637e-01 8.87121260e-01 -1.35229897e+00 1.66165531e+00
1.85266584e-01 -1.18415654e+00 2.59872407e-01 -8.16366449e-02
3.63863587e-01 -2.01504007e-01 2.67848432e-01 8.42105746e-01
-2.04357922e-01 -1.26597178e+00 7.70672083e-01 7.05426455e-01
8.96280706e-01 -1.09532094e+00 6.31834626e-01 -5.02787344e-02
-1.48149467e+00 -3.86063576e-01 -6.84418201e-01 -3.12559158e-01
1.66755259e-01 4.49028701e-01 -4.55868959e-01 3.96020085e-01
8.84628594e-01 1.16343391e+00 -3.61525536e-01 8.33090186e-01
-1.72617793e-01 7.03924537e-01 -8.96613598e-02 4.55529541e-02
2.28677958e-01 2.41895728e-02 5.02931237e-01 1.34353018e+00
6.28681898e-01 -3.67377639e-01 -1.56383276e-01 9.48586345e-01
-7.01777115e-02 4.29895610e-01 -5.53355038e-01 -6.70953870e-01
-6.02228520e-03 1.22997332e+00 -1.18759580e-01 -2.85888821e-01
2.34195385e-02 1.18636298e+00 2.44750202e-01 8.76618847e-02
-8.28148723e-01 -7.35914111e-01 1.43371904e+00 -2.55042344e-01
5.98765492e-01 -6.28497172e-03 -6.45864464e-04 -1.17766547e+00
-1.67959884e-01 -9.87485707e-01 4.99688655e-01 -1.17429841e+00
-1.65950382e+00 7.28968501e-01 -7.29178667e-01 -1.28605616e+00
-3.64146262e-01 -8.63992512e-01 -7.95833886e-01 5.19709647e-01
-1.37721550e+00 -1.24219251e+00 -9.48801860e-02 9.63823617e-01
3.66855979e-01 -5.91479003e-01 1.65854251e+00 6.38116777e-01
-7.14097545e-02 7.40082085e-01 -5.89501560e-02 1.41063854e-01
1.01458561e+00 -1.43674624e+00 4.72976500e-03 -2.77464569e-01
1.03249717e+00 2.55956680e-01 4.11402047e-01 -3.53292227e-02
-1.31741095e+00 -9.07573104e-01 8.59256148e-01 -4.09439743e-01
8.69183540e-01 -5.84956765e-01 -8.83473575e-01 6.65193856e-01
2.28361353e-01 -1.36169910e-01 1.63903570e+00 7.63083816e-01
-6.99348748e-01 -9.44222808e-02 -5.60001314e-01 2.57680148e-01
7.34315395e-01 -1.33255684e+00 -9.99041080e-01 -1.18576743e-01
9.51795816e-01 2.83581227e-01 -1.43075716e+00 4.32209969e-01
1.00542474e+00 -6.28657699e-01 1.28449547e+00 -1.07358658e+00
5.15774310e-01 -1.61145702e-01 -5.14834642e-01 -1.48629642e+00
-4.64404106e-01 -5.30747890e-01 -2.45328024e-01 1.12162852e+00
1.26313761e-01 -5.27576432e-02 5.04864097e-01 -2.23722875e-01
-1.62570179e-01 -5.70317626e-01 -8.74242127e-01 -6.38287127e-01
6.53150007e-02 -1.08199072e+00 8.38277996e-01 1.26707244e+00
2.75781810e-01 6.32343769e-01 -4.15763289e-01 -1.41975028e-03
1.85175315e-01 4.74273503e-01 8.63577068e-01 -1.70897925e+00
-6.66476607e-01 -5.37486136e-01 -1.19128537e+00 -4.97371256e-01
6.83029175e-01 -1.22681868e+00 -4.16221440e-01 -1.19333398e+00
-1.31838307e-01 -2.04784945e-01 -1.05899143e+00 4.75437224e-01
2.81348258e-01 6.89565539e-01 2.25824192e-01 -2.74568528e-01
-6.19476795e-01 1.22547972e+00 8.47643197e-01 -3.85082215e-01
-9.03742090e-02 -2.98236310e-01 -6.76591694e-01 6.74884379e-01
7.39691973e-01 -6.56652868e-01 -2.73301512e-01 -1.57136142e-01
8.23018014e-01 -1.19918823e-01 5.99126518e-01 -1.05317211e+00
-6.30052313e-02 5.41722119e-01 3.77993286e-01 -5.28893530e-01
9.18455362e-01 -9.32493627e-01 1.53977513e-01 -7.58050159e-02
-5.79193354e-01 5.83646260e-03 5.43003142e-01 6.02686346e-01
-8.76244426e-01 -1.47457451e-01 3.14394802e-01 2.47784629e-01
-6.62628055e-01 1.78772733e-01 -2.95639485e-01 -5.34444191e-02
5.27120590e-01 6.80507794e-02 2.28800341e-01 -4.13912684e-01
-1.46970451e+00 -3.86957288e-01 -6.28078282e-02 8.03373218e-01
7.69959152e-01 -2.10157752e+00 -5.82330167e-01 3.25259924e-01
5.67541540e-01 -6.62208915e-01 4.22753632e-01 5.36402881e-01
-1.72844768e-01 6.92085028e-02 -5.32803476e-01 -4.32221055e-01
-1.52598310e+00 5.73849678e-01 1.78255603e-01 -9.65137687e-03
-6.81948841e-01 8.94177079e-01 5.85945882e-02 -7.40541875e-01
7.33438909e-01 -5.39401233e-01 -1.91476658e-01 4.19228107e-01
7.00550497e-01 2.36137167e-01 -7.87911564e-02 -3.61457556e-01
-1.82086766e-01 8.54029119e-01 4.08979863e-01 -3.23519051e-01
1.80350542e+00 3.15648735e-01 -8.73362049e-02 1.26735198e+00
1.38194954e+00 -2.28699166e-02 -7.35617757e-01 8.82638544e-02
-1.49286613e-01 -3.11871350e-01 3.38838726e-01 -7.76994824e-01
-8.53470385e-01 1.56774223e+00 8.07728112e-01 1.69357345e-01
1.14766502e+00 -4.27910313e-02 9.04397190e-01 5.20917773e-01
-8.83036666e-03 -1.20593250e+00 4.44508046e-01 5.52642643e-01
1.25798237e+00 -6.98940396e-01 -5.74054122e-01 1.02172956e-01
-8.94550622e-01 1.35123658e+00 6.18238039e-02 -5.16074002e-01
1.13768804e+00 2.87490875e-01 2.83423781e-01 -4.57842916e-01
-9.81012762e-01 -2.20637470e-01 7.30916560e-01 2.81339377e-01
6.10722303e-01 3.05933148e-01 1.51077345e-01 1.45785427e+00
-7.35241890e-01 9.36016887e-02 1.29514471e-01 5.34652591e-01
-3.09217989e-01 -1.32812071e+00 -1.35837883e-01 -6.40523946e-03
-5.72473347e-01 -4.09173220e-02 -9.03221965e-01 4.84218925e-01
2.70452350e-01 6.20130420e-01 1.56091809e-01 -6.45462692e-01
5.40808439e-01 7.45061874e-01 5.93559802e-01 -5.07724524e-01
-1.08385253e+00 4.64218467e-01 -4.23150212e-02 -7.04326332e-01
-3.11981440e-01 -2.25639477e-01 -1.11558998e+00 1.07295461e-01
-1.26138493e-01 2.40452439e-01 7.92547464e-01 3.44197661e-01
6.24927640e-01 9.54024434e-01 3.91916484e-01 -9.54897344e-01
-5.90648651e-01 -1.09763968e+00 -1.23634756e+00 5.01956582e-01
1.59898028e-01 -6.98112011e-01 -4.44074064e-01 7.42017403e-02]
|
[15.729960441589355, 5.194106578826904]
|
26a67488-33b4-4915-8a98-0e35e7e3b559
|
multi-tasks-retinanet-for-mitosis-detection
|
2208.12657
| null |
https://arxiv.org/abs/2208.12657v1
|
https://arxiv.org/pdf/2208.12657v1.pdf
|
Multi tasks RetinaNet for mitosis detection
|
The account of mitotic cells is a key feature in tumor diagnosis. However, due to the variability of mitotic cell morphology, it is a highly challenging task to detect mitotic cells in tumor tissues. At the same time, although advanced deep learning method have achieved great success in cell detection, the performance is often unsatisfactory when tested data from another domain (i.e. the different tumor types and different scanners). Therefore, it is necessary to develop algorithms for detecting mitotic cells with robustness in domain shifts scenarios. Our work further proposes a foreground detection and tumor classification task based on the baseline(Retinanet), and utilizes data augmentation to improve the domain generalization performance of our model. We achieve the state-of-the-art performance (F1 score: 0.5809) on the challenging premilary test dataset.
|
['Zhang Yongbing', 'Bian Hao', 'Fang Zijie', 'Wang Ziyue', 'Chen Yang']
|
2022-08-26
| null | null | null | null |
['cell-detection', 'mitosis-detection']
|
['computer-vision', 'medical']
|
[ 2.05327168e-01 -3.04124206e-01 -2.19537437e-01 1.34597421e-01
-9.57531869e-01 -3.75744104e-01 6.09094977e-01 1.86060846e-01
-5.58095634e-01 8.78160596e-01 -1.59805030e-01 -3.37553769e-01
4.17845786e-01 -4.50591952e-01 -3.62168789e-01 -1.33868694e+00
5.52188337e-01 6.11293316e-01 6.56293571e-01 1.22670449e-01
3.46309841e-01 6.32214963e-01 -1.22791994e+00 2.44514972e-01
1.04185045e+00 9.06361818e-01 2.12684244e-01 7.41449833e-01
-2.78708756e-01 5.40154874e-01 -7.00662136e-01 -2.13501886e-01
-2.00935915e-01 -2.35723898e-01 -6.93156540e-01 1.48437038e-01
4.07406867e-01 -3.45712341e-02 -3.10308844e-01 1.23732340e+00
5.48173308e-01 -4.86988485e-01 1.04416442e+00 -9.50507998e-01
-4.58712637e-01 1.29885316e-01 -9.89309609e-01 6.54370666e-01
-3.45849127e-01 2.42793992e-01 5.16642272e-01 -8.19221675e-01
7.88871050e-01 6.31080925e-01 4.13923889e-01 8.35965991e-01
-1.22399580e+00 -6.83579683e-01 -9.63096023e-02 3.07815313e-01
-1.43336046e+00 -4.35816824e-01 4.44009274e-01 -5.03754497e-01
4.48819488e-01 -1.01614699e-01 5.21792710e-01 1.28436220e+00
3.35475594e-01 9.77691770e-01 1.24221301e+00 -2.21221402e-01
2.60609686e-01 2.18600348e-01 1.26427352e-01 5.72430372e-01
5.58478534e-01 -2.38736406e-01 -4.52394307e-01 2.54853964e-01
7.67187655e-01 -6.38990551e-02 -4.25990969e-01 -1.80483982e-01
-1.33312011e+00 5.41701198e-01 2.61562765e-01 5.47830582e-01
9.58859250e-02 -1.08818308e-01 4.84394342e-01 -7.29668140e-02
4.51609969e-01 1.44556016e-01 -3.08112264e-01 -4.05315794e-02
-9.83773351e-01 5.77051751e-02 4.48439002e-01 5.61696529e-01
2.72675276e-01 -1.18629478e-01 -1.85413226e-01 6.85172737e-01
8.76631439e-02 4.12607372e-01 8.27559352e-01 -2.47487590e-01
1.67983174e-02 7.39961386e-01 -1.99720472e-01 -4.65319484e-01
-5.84592283e-01 -9.93735015e-01 -1.15624332e+00 2.64042616e-01
1.10136092e+00 2.50610262e-01 -1.32876575e+00 1.55524588e+00
3.14628512e-01 3.26988071e-01 1.75053462e-01 7.72753060e-01
9.46357191e-01 2.56783903e-01 2.84222364e-01 -1.46173075e-01
1.50904572e+00 -8.11059117e-01 -6.58550620e-01 -3.44298750e-01
9.52070117e-01 -6.37199521e-01 1.07691789e+00 2.64219373e-01
-7.33029246e-01 -2.37256408e-01 -1.08417487e+00 -1.26124546e-01
-4.94394839e-01 2.03218907e-01 5.87091923e-01 6.57707572e-01
-8.46888244e-01 -4.92546000e-02 -1.02203989e+00 -7.21408963e-01
8.20401669e-01 3.47983778e-01 -2.66965687e-01 -8.49943459e-02
-5.80723286e-01 7.90029228e-01 3.60188365e-01 -3.09064776e-01
-9.04997587e-01 -9.43415403e-01 -3.92382264e-01 -8.44435319e-02
2.46052191e-01 -7.08602548e-01 1.30568326e+00 -5.90240657e-01
-1.16222703e+00 1.39722955e+00 -3.47886592e-01 -4.07810897e-01
6.62186980e-01 4.55224276e-01 -2.75320381e-01 8.04461986e-02
-4.35779914e-02 8.05466235e-01 5.10976970e-01 -9.31052446e-01
-1.03008831e+00 -4.17136252e-01 -5.01246393e-01 -1.14087582e-01
-3.25842321e-01 -2.36492842e-01 -5.73891521e-01 -5.94095767e-01
7.98102543e-02 -1.04896581e+00 7.32826814e-02 2.20311686e-01
-2.11505190e-01 -1.41875386e-01 9.69479322e-01 -6.34068370e-01
8.62967432e-01 -2.32875156e+00 -8.62502679e-02 -1.33161277e-01
3.45861495e-01 4.07204777e-01 2.28304453e-02 -4.09824729e-01
5.58546558e-02 1.15975365e-01 -1.99867606e-01 -1.42371833e-01
-2.26469547e-01 -8.27731639e-02 1.39111534e-01 6.55775547e-01
2.51934499e-01 9.18862581e-01 -6.44673049e-01 -7.83263624e-01
-1.47737190e-01 4.40853268e-01 -2.79543936e-01 -1.85726389e-01
-1.59229383e-01 6.74479485e-01 -1.36905819e-01 1.23856652e+00
7.99443364e-01 -5.89149058e-01 -1.39690906e-01 -1.86100751e-01
1.32990435e-01 -1.79687649e-01 -6.96150243e-01 1.51257622e+00
2.08800696e-02 9.57826734e-01 7.91794658e-02 -9.68597293e-01
6.82005584e-01 3.11583459e-01 4.76733059e-01 -6.90184414e-01
3.07649076e-01 5.56046784e-01 4.38848406e-01 -4.18310791e-01
1.85813650e-01 -2.38101870e-01 1.81684703e-01 -4.64382619e-02
-6.31932393e-02 1.79917552e-02 2.78074861e-01 3.74471396e-02
1.18498814e+00 -2.80917495e-01 3.20632398e-01 -3.45947832e-01
6.71999812e-01 1.98702991e-01 8.36938202e-01 2.24929303e-01
-7.73836493e-01 6.62785053e-01 7.07735240e-01 -3.20157379e-01
-9.64163125e-01 -1.04563069e+00 -4.45795655e-01 6.81084514e-01
1.03434913e-01 3.14820856e-01 -6.75630867e-01 -6.54285789e-01
-5.91520295e-02 2.12394893e-01 -7.17810035e-01 -2.20160306e-01
-2.70841718e-01 -1.31562257e+00 7.33860493e-01 6.23173714e-01
9.87917602e-01 -4.98519093e-01 -3.16915840e-01 -2.12697648e-02
-2.05433235e-01 -1.43714654e+00 -2.50253886e-01 2.91428089e-01
-7.25488007e-01 -1.14942825e+00 -1.05299616e+00 -1.11163843e+00
8.12115252e-01 3.22609961e-01 8.96322429e-01 1.09736964e-01
-4.94198203e-01 -2.80682474e-01 -6.77183270e-02 -6.35469913e-01
-4.87372339e-01 2.82171309e-01 -2.93945640e-01 4.56644185e-02
6.68565273e-01 -1.22058786e-01 -6.82036757e-01 3.82476956e-01
-9.01898146e-01 1.94803372e-01 8.65035534e-01 1.07716918e+00
9.45092678e-01 1.87971905e-01 5.03165245e-01 -9.71094608e-01
3.49334516e-02 -2.27746889e-01 -5.49665272e-01 1.37519091e-01
-3.65512609e-01 -2.10595444e-01 3.87740999e-01 -6.89436316e-01
-1.02611470e+00 1.48053825e-01 -2.77991742e-02 -5.46606891e-02
-3.38781446e-01 3.26733649e-01 -2.73555607e-01 -3.10672849e-01
7.65680790e-01 2.54452467e-01 1.83807492e-01 4.19356413e-02
-3.14047128e-01 5.79287410e-01 8.41353893e-01 -1.57398880e-01
6.71040118e-01 7.35986233e-01 3.01775604e-01 -1.01164961e+00
-7.35312819e-01 -4.38771188e-01 -3.91507566e-01 -1.52190238e-01
9.36799467e-01 -9.85407174e-01 -3.64393383e-01 9.17601645e-01
-8.02298427e-01 -4.59925503e-01 1.59572929e-01 4.31626827e-01
-2.75171131e-01 2.86245018e-01 -6.97409391e-01 -3.27300489e-01
-2.44495839e-01 -1.17286646e+00 9.43582177e-01 6.53249443e-01
-8.54993090e-02 -1.02320373e+00 -1.07822165e-01 5.69663227e-01
3.79760802e-01 5.63729525e-01 1.17958272e+00 -6.50399208e-01
-7.44403362e-01 -1.66984722e-01 -5.04953742e-01 -2.28827253e-01
2.29295090e-01 4.23629612e-01 -1.30266535e+00 -2.43919596e-01
-3.85212034e-01 -1.22114010e-01 1.23030066e+00 6.90293312e-01
1.22327781e+00 4.03946072e-01 -9.80284631e-01 8.55313420e-01
1.26353431e+00 2.06746459e-01 8.56217325e-01 6.53690219e-01
3.22764188e-01 4.14452612e-01 5.16903818e-01 -7.38639012e-02
9.46283937e-02 3.35438609e-01 3.62965405e-01 -4.49567378e-01
-5.73951721e-01 2.39707977e-01 -6.03236407e-02 2.33822823e-01
1.89163163e-01 -2.69757181e-01 -1.28511465e+00 7.72803366e-01
-1.63361990e+00 -7.73841143e-01 -1.65122733e-01 1.95369208e+00
7.92879760e-01 3.73339623e-01 1.97624162e-01 2.53303677e-01
8.24564517e-01 -2.56091744e-01 -9.51870084e-01 1.18716389e-01
-4.63206887e-01 2.06566080e-02 5.46891093e-01 3.18216272e-02
-1.22853446e+00 8.58597338e-01 6.16875648e+00 1.01187420e+00
-1.43064821e+00 3.80632207e-02 1.15009117e+00 -4.46860604e-02
1.91763237e-01 -4.34430212e-01 -1.00905049e+00 6.25491798e-01
5.74987113e-01 -1.55724958e-01 -1.38336346e-01 4.77769852e-01
-5.16209714e-02 -3.57978731e-01 -9.94754970e-01 1.17279255e+00
5.24359532e-02 -1.34815466e+00 -3.24009627e-01 3.52333963e-01
7.09871888e-01 -3.33519280e-02 4.60068971e-01 3.99191797e-01
-9.61462855e-02 -9.92170632e-01 1.08999074e-01 1.52488664e-01
8.95305455e-01 -6.00379646e-01 9.76206541e-01 4.92121190e-01
-9.18520510e-01 4.70959619e-02 -4.13909107e-01 3.29910368e-01
-2.25076124e-01 7.09470212e-01 -1.23182809e+00 -6.46802187e-02
4.54986513e-01 5.48723221e-01 -8.80158842e-01 1.55085719e+00
1.08214051e-01 5.57696104e-01 -1.55848682e-01 -1.28401950e-01
-1.91295370e-01 2.33966321e-01 2.19373137e-01 1.21053946e+00
3.82518858e-01 -2.30236053e-01 -1.27442047e-01 6.52174830e-01
-2.03542158e-01 -1.53989673e-01 -3.85117024e-01 -1.28836632e-01
4.98633534e-01 1.53739130e+00 -1.39379418e+00 -1.18349850e-01
-4.55454290e-01 6.42986059e-01 2.89215833e-01 3.16868782e-01
-9.74454999e-01 -1.32632300e-01 7.57872581e-01 1.51753932e-01
1.23637356e-01 -1.88522395e-02 -5.02927959e-01 -1.01469445e+00
-3.12736243e-01 -7.67978489e-01 4.69986409e-01 -3.73151541e-01
-1.26220465e+00 2.35542610e-01 -5.32922745e-01 -1.18525589e+00
3.29401702e-01 -9.15080905e-01 -6.95797205e-01 5.07501125e-01
-1.74810863e+00 -1.06954730e+00 -7.58232951e-01 5.08606732e-01
5.49860477e-01 -2.72632957e-01 5.44904232e-01 2.84003317e-01
-7.85215914e-01 8.32974494e-01 1.44053623e-01 1.83604673e-01
1.07175660e+00 -1.32919252e+00 1.08064927e-01 8.81240487e-01
-2.22498074e-01 1.04694687e-01 5.85827947e-01 -4.72882986e-01
-1.05671418e+00 -1.18149674e+00 4.85558867e-01 -2.61727065e-01
6.52653456e-01 -3.02231550e-01 -1.08735168e+00 3.80789608e-01
1.13649979e-01 2.11169884e-01 9.06380951e-01 -1.63717523e-01
-1.04922488e-01 -1.04260594e-01 -1.23361313e+00 7.81325042e-01
7.51063824e-01 -3.40007454e-01 -1.55248597e-01 3.94675195e-01
3.46117467e-01 -7.31839478e-01 -6.60595715e-01 4.93419826e-01
4.18453544e-01 -7.27929711e-01 7.23622620e-01 -1.93284154e-01
2.53281236e-01 -5.21282732e-01 7.73847774e-02 -1.17267132e+00
-5.12774289e-01 -8.29254761e-02 -1.15672700e-01 1.16371441e+00
4.91208017e-01 -6.83243632e-01 1.38731182e+00 3.37573469e-01
-2.03325525e-01 -8.21537554e-01 -1.06634998e+00 -7.19798744e-01
2.90575832e-01 2.12028548e-01 2.28193015e-01 8.16703677e-01
-1.35841057e-01 3.41030031e-01 2.56019682e-01 3.53522837e-01
5.47688007e-01 -1.07625417e-01 7.68594205e-01 -1.44943380e+00
3.50811556e-02 -8.39843214e-01 -8.00360382e-01 -6.41502500e-01
1.88050449e-01 -9.09412563e-01 -6.55390173e-02 -1.29845679e+00
4.54414278e-01 -2.38439783e-01 -4.40320075e-01 3.02215278e-01
-3.42045158e-01 4.05033350e-01 -1.69632465e-01 1.78578883e-01
-5.12671590e-01 2.71417826e-01 1.46419668e+00 -5.26539505e-01
1.12473547e-01 -1.99774057e-01 -7.72845864e-01 7.21440315e-01
9.05852795e-01 -2.66871393e-01 -1.77875027e-01 -3.38911563e-01
-1.67320341e-01 -2.55976617e-01 2.67903477e-01 -1.39870739e+00
5.33061981e-01 -8.87627527e-02 8.43677878e-01 -8.57642055e-01
1.77406952e-01 -5.35244465e-01 -1.25453308e-01 7.26615608e-01
-1.10250108e-01 -1.99711069e-01 3.77267033e-01 6.78642571e-01
-1.63527042e-01 5.98986074e-02 1.17093003e+00 3.10052000e-02
-7.33490109e-01 4.30517882e-01 -5.52071869e-01 1.58255398e-01
1.20942712e+00 -5.81141949e-01 -9.61367369e-01 2.19070390e-01
-3.95841330e-01 2.34047800e-01 7.69740343e-01 -2.35324875e-02
3.86542678e-01 -1.10241449e+00 -7.35037386e-01 1.26253515e-01
4.60071981e-01 2.97901571e-01 6.81662560e-02 1.26473689e+00
-7.71203339e-01 3.60823065e-01 -2.56100208e-01 -1.12458301e+00
-1.51012492e+00 3.39825243e-01 4.99236733e-01 -4.22666013e-01
-3.62913311e-01 1.09221256e+00 5.65730691e-01 1.68053076e-01
3.41639370e-01 -3.90708238e-01 -2.81353444e-01 -3.19471322e-02
6.33697569e-01 2.31319517e-01 3.72004211e-01 -3.00976157e-01
-4.59512085e-01 5.41520059e-01 -5.27069092e-01 2.65624255e-01
9.35022473e-01 1.81591570e-01 8.65980312e-02 2.16805235e-01
9.11097944e-01 -9.61903185e-02 -1.28917897e+00 -1.79860190e-01
9.36005712e-02 -3.91694218e-01 1.32313251e-01 -7.97942817e-01
-1.19746184e+00 9.26691175e-01 9.06713963e-01 -1.14374273e-01
1.27478623e+00 1.82558447e-02 7.79219031e-01 2.36234784e-01
1.26428127e-01 -1.09633589e+00 1.83365777e-01 4.94947940e-01
1.89341798e-01 -1.54057038e+00 -1.94958463e-01 -4.59930480e-01
-4.81736451e-01 1.06740510e+00 8.43465149e-01 2.10964471e-01
4.29073602e-01 6.09580398e-01 3.46928179e-01 -4.31544259e-02
-8.84148180e-01 -3.66077304e-01 -3.05808578e-02 8.34601104e-01
5.82027912e-01 -1.40619397e-01 -3.28542650e-01 3.12781751e-01
1.28087446e-01 2.25050487e-02 7.05111980e-01 7.60717094e-01
-6.16949081e-01 -8.79642367e-01 -3.24000001e-01 4.41054106e-01
-7.28552878e-01 3.21138322e-01 -5.97936630e-01 1.13366735e+00
8.33563507e-03 6.12183154e-01 1.57587692e-01 -3.61120962e-02
-2.64093280e-02 1.44301459e-01 5.19931316e-01 -4.72369790e-01
-8.27283412e-02 2.79169321e-01 -3.20897520e-01 -2.11377651e-03
-3.95158082e-01 -6.75645351e-01 -1.48191941e+00 -3.36759031e-01
-2.58187741e-01 -2.67587274e-01 4.14846420e-01 7.47866571e-01
1.37388617e-01 6.61365211e-01 1.26350611e-01 -3.02129209e-01
-1.79962888e-01 -8.72023284e-01 -7.97955930e-01 4.11593705e-01
4.70868707e-01 -8.76262844e-01 -2.44026929e-01 2.15406492e-01]
|
[15.098042488098145, -3.127023696899414]
|
231d435a-41e3-4668-a4fe-db0365dd9da2
|
memorizing-complementation-network-for-few
|
2208.0561
| null |
https://arxiv.org/abs/2208.05610v1
|
https://arxiv.org/pdf/2208.05610v1.pdf
|
Memorizing Complementation Network for Few-Shot Class-Incremental Learning
|
Few-shot Class-Incremental Learning (FSCIL) aims at learning new concepts continually with only a few samples, which is prone to suffer the catastrophic forgetting and overfitting problems. The inaccessibility of old classes and the scarcity of the novel samples make it formidable to realize the trade-off between retaining old knowledge and learning novel concepts. Inspired by that different models memorize different knowledge when learning novel concepts, we propose a Memorizing Complementation Network (MCNet) to ensemble multiple models that complements the different memorized knowledge with each other in novel tasks. Additionally, to update the model with few novel samples, we develop a Prototype Smoothing Hard-mining Triplet (PSHT) loss to push the novel samples away from not only each other in current task but also the old distribution. Extensive experiments on three benchmark datasets, e.g., CIFAR100, miniImageNet and CUB200, have demonstrated the superiority of our proposed method.
|
['Xuelong Li', 'Yanwei Pang', 'Xiyao Liu', 'Zhishen Hou', 'Zhong Ji']
|
2022-08-11
| null | null | null | null |
['few-shot-class-incremental-learning', 'novel-concepts']
|
['methodology', 'reasoning']
|
[ 1.30708411e-01 -1.00711100e-01 1.01913884e-01 -2.33903781e-01
-9.50520560e-02 1.42188175e-02 3.49088371e-01 2.02440605e-01
-7.28483915e-01 1.32993698e+00 -2.32320696e-01 1.59637645e-01
-2.26644978e-01 -7.95732677e-01 -7.20913410e-01 -9.27239537e-01
-3.47106233e-02 4.02939230e-01 7.21573293e-01 -1.37319833e-01
1.01884820e-01 1.48210523e-03 -1.84164715e+00 2.83892691e-01
1.49547958e+00 1.02222764e+00 7.00455725e-01 -2.88028810e-02
-3.82557571e-01 6.97521210e-01 -6.90461993e-01 -4.38929319e-01
3.50349583e-02 -4.20712113e-01 -5.52114964e-01 -4.29765210e-02
2.50233889e-01 -1.10998243e-01 -2.67656088e-01 9.93385613e-01
5.49565256e-01 5.77296197e-01 3.13905329e-01 -1.20551276e+00
-1.12754261e+00 8.75265062e-01 -7.87627220e-01 4.89193261e-01
-1.74954101e-01 2.15150803e-01 3.47833008e-01 -1.32666218e+00
5.57259977e-01 1.19931471e+00 7.74857402e-01 7.76087761e-01
-8.72292280e-01 -9.15811539e-01 7.24804461e-01 7.07367718e-01
-1.46694708e+00 -3.47623348e-01 6.60279572e-01 -7.63418898e-02
6.16469920e-01 -6.00949787e-02 8.77895236e-01 1.18634701e+00
2.43798748e-01 1.10304391e+00 9.99148726e-01 -4.09185052e-01
3.77212882e-01 4.97414529e-01 3.64705712e-01 4.85261738e-01
5.66438675e-01 -6.86991168e-03 -6.79150105e-01 8.07124153e-02
2.72712260e-01 6.14371181e-01 -3.23938936e-01 -4.98617232e-01
-1.00947583e+00 5.40581346e-01 4.34061766e-01 3.33410174e-01
-1.04305133e-01 -2.36210674e-01 5.07626116e-01 5.37182450e-01
4.40691501e-01 1.39965221e-01 -5.76905131e-01 1.66305229e-01
-7.03678846e-01 -1.07667871e-01 9.48393196e-02 1.04321897e+00
1.05182827e+00 2.77113020e-01 -4.10425514e-01 1.01002991e+00
-1.91311762e-01 2.61607289e-01 1.34612846e+00 -1.90953657e-01
3.75929505e-01 5.54684758e-01 -1.01457067e-01 -8.59590828e-01
-2.94417948e-01 -9.80904162e-01 -9.79430497e-01 -2.23922148e-01
-9.62066948e-02 -2.07054183e-01 -1.05861342e+00 1.89712119e+00
3.51584047e-01 5.79399943e-01 7.93255195e-02 4.65434313e-01
7.94806719e-01 6.58158660e-01 3.23587239e-01 -8.55201125e-01
7.67234743e-01 -9.83562589e-01 -6.04666770e-01 -3.48363400e-01
3.03955108e-01 -2.65249014e-01 1.26654410e+00 4.71182138e-01
-9.23242509e-01 -1.10754097e+00 -1.24713600e+00 9.14030969e-02
-3.85887265e-01 -2.27860749e-01 6.36463642e-01 3.49225134e-01
-5.53509891e-01 6.91079974e-01 -6.49511516e-01 -1.42727375e-01
1.01531076e+00 5.51434495e-02 -1.52784865e-02 -4.40373033e-01
-1.18348956e+00 7.36989439e-01 1.08062446e+00 -2.20028218e-02
-9.35736477e-01 -9.69119906e-01 -4.32081103e-01 2.45976284e-01
5.55623591e-01 -6.85487866e-01 9.94750321e-01 -1.10551512e+00
-1.11082768e+00 2.85137206e-01 -8.63199160e-02 -5.99742591e-01
5.80678523e-01 -2.75606900e-01 -7.70821333e-01 -2.27075115e-01
-1.78083766e-03 7.25304008e-01 1.06940532e+00 -1.22545218e+00
-9.12239611e-01 -4.37870979e-01 -4.81153429e-01 3.58307660e-01
-9.24294651e-01 -7.66931236e-01 -2.04410806e-01 -8.11186969e-01
1.99997425e-01 -5.29370606e-01 7.17317238e-02 -8.64578858e-02
-4.07922491e-02 -4.02090609e-01 9.65061128e-01 -1.80058405e-01
1.25429118e+00 -2.12195349e+00 -8.23343694e-02 -1.45544082e-01
2.82296717e-01 5.78424335e-01 -3.65524143e-01 -8.18820968e-02
-1.34296808e-02 -3.42620611e-01 -2.25086913e-01 -2.11525202e-01
-4.17865396e-01 3.48089397e-01 -5.65498114e-01 -6.24133348e-02
4.37909365e-02 9.53104675e-01 -1.34374332e+00 -3.29922765e-01
-1.40678436e-01 1.39725983e-01 -2.49947950e-01 -1.05154060e-01
-3.07532996e-01 3.09614360e-01 -2.23114714e-01 6.69971645e-01
1.00542927e+00 -1.31704122e-01 -1.30323380e-01 1.63339004e-01
2.63062596e-01 -2.62867063e-01 -1.19024146e+00 1.86041069e+00
-2.56267875e-01 5.54772541e-02 -5.85498095e-01 -1.11978936e+00
1.03980505e+00 -4.61476743e-02 -1.10305147e-02 -8.42968583e-01
-8.99421982e-03 1.24320604e-01 6.97684363e-02 -5.33854425e-01
4.72700059e-01 -5.59676528e-01 3.63347948e-01 4.31201994e-01
5.49588144e-01 5.05719244e-01 1.90448701e-01 1.83049232e-01
7.25916088e-01 -1.85559392e-02 2.42877468e-01 -1.38687894e-01
3.87201995e-01 -2.77188689e-01 1.06841397e+00 1.00131071e+00
-3.55800986e-01 3.66334587e-01 -1.52124614e-01 -6.15529537e-01
-4.92537588e-01 -1.52862859e+00 -1.11450039e-01 1.34548557e+00
2.32841372e-01 -1.45755336e-02 -1.77157670e-01 -8.46450746e-01
1.45780042e-01 9.17264402e-01 -7.12553024e-01 -1.03589749e+00
-4.03615147e-01 -9.54307497e-01 -2.08770437e-03 5.23425102e-01
7.22718120e-01 -1.19628859e+00 -5.80508530e-01 5.00340462e-01
-1.46540046e-01 -4.87558067e-01 -3.81003976e-01 3.11877608e-01
-1.30239737e+00 -1.02883327e+00 -8.83587778e-01 -9.40200806e-01
8.91186595e-01 7.10835636e-01 9.69205856e-01 2.23176152e-01
-3.92810673e-01 1.78309828e-01 -5.63583910e-01 -8.02997351e-01
1.59875587e-01 6.33884296e-02 4.73472476e-01 1.32886216e-01
6.64449215e-01 -7.68361986e-01 -4.67470855e-01 6.20170794e-02
-9.47991610e-01 -1.32101774e-01 6.06505334e-01 1.30912876e+00
7.07695425e-01 5.03191471e-01 1.36194003e+00 -1.04195583e+00
6.83075130e-01 -8.16819429e-01 -9.40971747e-02 5.88611841e-01
-1.04138732e+00 -2.39329413e-01 8.72902095e-01 -1.06176150e+00
-1.60107636e+00 -2.95710325e-01 3.66715491e-01 -5.88253081e-01
2.65377969e-01 5.54660082e-01 -1.48505092e-01 1.45129040e-01
6.62662685e-01 7.40726411e-01 -1.15991876e-01 -6.38384759e-01
4.03078526e-01 2.46334806e-01 7.50379801e-01 -6.23614371e-01
8.22899163e-01 4.14918512e-01 -4.97363567e-01 -5.45902789e-01
-1.10735822e+00 -1.22623637e-01 -5.84765732e-01 -1.39677292e-02
8.11271891e-02 -9.89094198e-01 -4.02895093e-01 7.57743776e-01
-7.92680502e-01 7.52956793e-02 -9.09954846e-01 4.30291444e-01
-2.05291316e-01 2.76939154e-01 -2.32624114e-01 -8.61759782e-01
-3.62275898e-01 -5.53138196e-01 2.05214038e-01 7.66239047e-01
1.89474896e-01 -6.95161879e-01 -5.70309646e-02 4.75492235e-03
5.95075488e-01 -1.06785916e-01 1.04199958e+00 -5.37422776e-01
-4.80355978e-01 -6.80394322e-02 -6.98305890e-02 3.89023542e-01
2.82985896e-01 -5.39979100e-01 -1.01226687e+00 -7.65277803e-01
2.77283430e-01 -7.25389063e-01 1.67020261e+00 5.55128343e-02
1.44675672e+00 -3.70264143e-01 -4.54700440e-01 4.02160019e-01
1.25869620e+00 4.35087085e-01 7.05624282e-01 1.47826821e-01
4.63027447e-01 2.93376356e-01 8.29430342e-01 5.67522407e-01
3.26238811e-01 3.53618599e-02 -1.83384549e-02 4.64139134e-01
-2.67634273e-01 -6.23008370e-01 1.04313992e-01 1.26323116e+00
1.74090769e-02 6.53998032e-02 -5.73969364e-01 7.40134239e-01
-1.89260805e+00 -1.13381851e+00 5.23303866e-01 2.25042033e+00
1.31385326e+00 4.36965048e-01 -2.61186898e-01 7.60051310e-02
7.02160001e-01 -2.14396819e-01 -1.10795975e+00 2.07172886e-01
-3.75072539e-01 4.01845306e-01 -1.25405490e-01 -5.36465608e-02
-8.10331047e-01 9.43752646e-01 4.85386324e+00 1.28274167e+00
-9.79541302e-01 4.69675303e-01 7.63853967e-01 -5.38867474e-01
-3.22146684e-01 1.62072200e-02 -9.07085061e-01 7.51562655e-01
4.65378433e-01 -5.42744696e-01 2.93834537e-01 8.07679415e-01
-4.18646425e-01 -3.48404527e-01 -7.60833442e-01 1.09960234e+00
3.11532289e-01 -1.25384045e+00 4.49734807e-01 -7.88425863e-01
1.07882774e+00 -1.87516242e-01 3.53688091e-01 9.77207124e-01
1.61582470e-01 -7.03921258e-01 5.89889705e-01 1.10011780e+00
5.98472238e-01 -9.03955877e-01 6.03716195e-01 5.89280963e-01
-9.96019959e-01 -6.80717111e-01 -9.87270415e-01 -8.00621808e-02
-1.32607177e-01 8.23072374e-01 -4.97445136e-01 4.23299313e-01
1.00897706e+00 8.73378515e-01 -9.62347388e-01 1.36216569e+00
-4.50403169e-02 4.13057923e-01 -1.63841732e-02 5.56466281e-02
-1.11853264e-01 1.90498278e-04 3.81347388e-01 6.20120704e-01
4.87805188e-01 2.30750769e-01 2.06911787e-01 7.54579365e-01
-3.42381328e-01 1.09870479e-01 -2.89865047e-01 9.02400687e-02
9.08842742e-01 8.96693587e-01 -6.18144095e-01 -5.72210431e-01
-1.57154202e-01 1.02257264e+00 7.04126775e-01 4.85632449e-01
-4.16290134e-01 -7.32210755e-01 1.11359388e-01 -4.18031067e-02
4.63781506e-01 7.27083161e-02 -2.16995537e-01 -1.31007004e+00
3.28615814e-01 -5.37507653e-01 6.75175726e-01 -6.59095585e-01
-1.69812322e+00 4.90575075e-01 -5.25282621e-02 -1.39369321e+00
4.09885108e-01 2.74195187e-02 -9.39812601e-01 6.10977471e-01
-1.67110395e+00 -7.50431180e-01 -4.74877924e-01 6.25224710e-01
6.65039539e-01 -7.18138993e-01 4.65207219e-01 3.71537209e-01
-5.13109684e-01 9.37789857e-01 4.25450832e-01 -4.02276456e-01
7.56170034e-01 -7.26421177e-01 4.72801588e-02 5.47112942e-01
-7.13690603e-03 8.30916524e-01 3.83880496e-01 -8.91040742e-01
-9.72285450e-01 -1.35204077e+00 7.76244462e-01 -2.46805832e-01
2.51521915e-01 -3.41190487e-01 -1.41708112e+00 5.81048369e-01
-1.08817108e-01 4.27952409e-02 6.53587759e-01 1.28425434e-01
-4.26955312e-01 -6.06047451e-01 -1.10302854e+00 4.96087492e-01
1.21667445e+00 -4.67071943e-02 -1.05301940e+00 3.02341372e-01
1.06558013e+00 5.91251142e-02 -2.21733168e-01 4.44451034e-01
3.73221844e-01 -9.62215483e-01 9.37611878e-01 -8.14763308e-01
4.13507447e-02 -3.24865878e-01 1.66984260e-01 -1.65116203e+00
-5.31282127e-01 -2.32330516e-01 -4.96502280e-01 1.33872545e+00
1.92845523e-01 -8.24671328e-01 6.69244707e-01 2.04502732e-01
-2.57429808e-01 -8.14316809e-01 -1.13600588e+00 -1.30771756e+00
2.48589769e-01 1.48456069e-02 7.28222370e-01 1.19054329e+00
-2.60439254e-02 5.28673530e-01 -5.42373717e-01 -4.01409239e-01
6.65156364e-01 6.68927878e-02 3.04808527e-01 -1.59563518e+00
-1.97690621e-01 -1.25771523e-01 -2.16366991e-01 -8.31303835e-01
1.91951320e-02 -1.02172363e+00 3.44333127e-02 -1.14453340e+00
7.01773465e-01 -7.81813085e-01 -1.05690920e+00 7.10837185e-01
-7.30621636e-01 -1.27410322e-01 2.03887358e-01 1.99938610e-01
-1.09542000e+00 1.31457162e+00 1.39907491e+00 -4.18296665e-01
-3.95821035e-01 -1.49695233e-01 -1.03694248e+00 5.68503439e-01
6.67970479e-01 -7.06155002e-01 -1.00060511e+00 -5.98905444e-01
-5.33501478e-03 -4.19692606e-01 4.85810302e-02 -1.43233192e+00
5.26937842e-01 -4.87608686e-02 1.03988504e+00 -7.29287982e-01
1.28673911e-01 -6.51557386e-01 -5.22175916e-02 6.51599228e-01
-2.41928026e-01 -1.57787591e-01 3.21995378e-01 1.11333847e+00
4.68750484e-02 -4.55262184e-01 9.52555001e-01 -3.70948076e-01
-1.01937342e+00 5.85994720e-01 1.72192976e-01 4.66362476e-01
1.18145895e+00 -3.94866407e-01 -5.74053466e-01 1.88412666e-01
-7.82962561e-01 5.97844541e-01 5.74609414e-02 8.24606478e-01
1.14820635e+00 -1.53630149e+00 -6.21046841e-01 3.73733401e-01
3.11976969e-01 4.77915294e-02 9.44627643e-01 6.58838511e-01
2.19188824e-01 2.18903646e-03 -5.50945580e-01 -2.60023266e-01
-8.22764695e-01 1.05254567e+00 5.57517856e-02 3.57445367e-02
-6.54691041e-01 1.25270772e+00 1.65101245e-01 -3.99556011e-01
4.19899076e-01 4.98561375e-03 -1.74173996e-01 3.55534941e-01
9.96430874e-01 4.34995085e-01 4.88304235e-02 6.77636936e-02
-1.69970632e-01 1.04877844e-01 -8.74401212e-01 4.65434462e-01
1.33800483e+00 -2.93728560e-01 -7.19668493e-02 9.42345381e-01
9.02557194e-01 -5.34240544e-01 -1.29935646e+00 -7.74714410e-01
9.98406205e-03 -5.17969012e-01 -3.13694805e-01 -9.22082663e-01
-9.62997854e-01 9.27025795e-01 9.90285397e-01 -3.24418247e-01
1.14422393e+00 -3.08922857e-01 1.13422561e+00 8.34043384e-01
6.55483723e-01 -1.63638008e+00 5.41506231e-01 6.12752855e-01
6.53493702e-01 -1.16759562e+00 1.65008858e-01 1.36648402e-01
-5.48180640e-01 8.87479484e-01 1.13867784e+00 8.68522450e-02
7.52309561e-01 -4.23087329e-01 -2.69844592e-01 1.66943595e-01
-1.00660563e+00 -8.41394663e-02 -2.57478710e-02 8.48878860e-01
-1.18839175e-01 -2.63921879e-02 -3.46407771e-01 1.20396507e+00
1.08602270e-01 4.11061853e-01 4.52259481e-01 1.12293530e+00
-1.04124844e+00 -7.73712397e-01 -2.12052222e-02 9.78717029e-01
1.97574690e-01 -2.93309540e-01 -2.31225276e-03 4.23428923e-01
6.79421604e-01 7.96071529e-01 9.41514820e-02 -4.26732063e-01
2.34393552e-01 3.57639492e-01 5.12513459e-01 -7.99107552e-01
-1.48608267e-01 -3.44306409e-01 -6.17865980e-01 5.50416857e-02
-1.26619890e-01 -3.86366546e-01 -1.26882541e+00 -4.59951051e-02
-4.43552107e-01 3.58732283e-01 -1.51168555e-01 9.25384343e-01
4.83695596e-01 7.23127782e-01 6.36144161e-01 -1.97175071e-01
-9.46066916e-01 -1.03789890e+00 -8.71692419e-01 3.25267941e-01
1.58661932e-01 -1.10677493e+00 -2.07650989e-01 -1.55695928e-02]
|
[9.840677261352539, 3.404435634613037]
|
73037ab5-66c6-4a7e-8a71-82fb889111b8
|
dual-skipping-networks
|
1710.10386
| null |
http://arxiv.org/abs/1710.10386v3
|
http://arxiv.org/pdf/1710.10386v3.pdf
|
Dual Skipping Networks
|
Inspired by the recent neuroscience studies on the left-right asymmetry of
the human brain in processing low and high spatial frequency information, this
paper introduces a dual skipping network which carries out coarse-to-fine
object categorization. Such a network has two branches to simultaneously deal
with both coarse and fine-grained classification tasks. Specifically, we
propose a layer-skipping mechanism that learns a gating network to predict
which layers to skip in the testing stage. This layer-skipping mechanism endows
the network with good flexibility and capability in practice. Evaluations are
conducted on several widely used coarse-to-fine object categorization
benchmarks, and promising results are achieved by our proposed network model.
|
['xiangyang xue', 'Jianfeng Feng', 'Yu-Gang Jiang', 'Yanwei Fu', 'Wenlian Lu', 'Wei Liu', 'Changmao Cheng']
|
2017-10-28
|
dual-skipping-networks-1
|
http://openaccess.thecvf.com/content_cvpr_2018/html/Cheng_Dual_Skipping_Networks_CVPR_2018_paper.html
|
http://openaccess.thecvf.com/content_cvpr_2018/papers/Cheng_Dual_Skipping_Networks_CVPR_2018_paper.pdf
|
cvpr-2018-6
|
['object-categorization']
|
['computer-vision']
|
[ 9.57576036e-02 -1.41122192e-01 -3.84492964e-01 -8.87448967e-01
6.92316666e-02 -4.35472786e-01 6.18410408e-01 -8.27718824e-02
-6.02170885e-01 5.21726012e-01 1.47258818e-01 -1.24806173e-01
-4.27543253e-01 -8.47965419e-01 -3.15683246e-01 -5.31000257e-01
-2.39542991e-01 1.81933135e-01 3.74693483e-01 -5.11714704e-02
6.15804315e-01 5.33454597e-01 -1.60547924e+00 8.68039787e-01
7.07436681e-01 1.47653306e+00 3.26627254e-01 2.80496806e-01
-1.03788540e-01 6.75085187e-01 -2.61222094e-01 -1.67929441e-01
3.92322272e-01 -9.68221426e-02 -9.53328848e-01 -2.10645840e-01
3.78354937e-01 -1.40172124e-01 -1.31810829e-01 1.16660118e+00
4.31437612e-01 3.49238887e-02 7.70342529e-01 -1.32769418e+00
-8.64249468e-01 8.14665377e-01 -8.27796996e-01 9.78954792e-01
-5.25374115e-02 2.11945549e-01 1.24081385e+00 -9.37860966e-01
2.15672061e-01 1.48219585e+00 3.41341048e-01 5.13303220e-01
-1.27312756e+00 -8.82546663e-01 6.31674290e-01 3.19257081e-01
-1.52428591e+00 -2.13397369e-01 6.38919592e-01 -7.73209870e-01
1.03822315e+00 -4.67849113e-02 6.77963197e-01 9.23688471e-01
6.54049218e-01 6.11636519e-01 1.43337107e+00 2.31968574e-02
2.25134432e-01 -3.27824533e-01 6.12389922e-01 5.15537977e-01
5.44200063e-01 2.37893432e-01 -6.87285960e-01 1.03891246e-01
1.03324187e+00 1.97459608e-01 -1.47530613e-02 -2.08810925e-01
-1.35342288e+00 7.20241666e-01 1.02407038e+00 4.58373249e-01
-4.92969602e-01 1.36111125e-01 6.09024763e-01 4.55973566e-01
1.94602266e-01 4.78026241e-01 -4.83236820e-01 4.48253959e-01
-1.05758321e+00 1.15376422e-02 3.26538146e-01 5.12786686e-01
5.90190232e-01 3.13849188e-02 -7.05033600e-01 8.60425055e-01
3.57874960e-01 -4.67421524e-02 8.09539735e-01 -7.73844481e-01
5.63338161e-01 6.90135419e-01 -2.36377656e-01 -8.12959611e-01
-8.35635841e-01 -6.41214371e-01 -1.36983323e+00 2.58029699e-01
2.42482767e-01 2.98561305e-01 -9.89143550e-01 1.88993335e+00
-5.11682741e-02 9.04555991e-02 -3.75185341e-01 1.09009719e+00
8.37304771e-01 1.98284805e-01 6.28332794e-01 9.97816473e-02
1.65303004e+00 -8.34826112e-01 -4.28012818e-01 -5.06442726e-01
1.29154488e-01 -2.18522459e-01 1.16555500e+00 3.67324889e-01
-1.03329086e+00 -1.06484497e+00 -9.33676600e-01 -1.10551026e-02
-5.29355764e-01 1.44606277e-01 9.20908749e-01 4.73100364e-01
-1.02104020e+00 7.08845615e-01 -7.13187575e-01 -3.53084207e-02
1.00044692e+00 5.13303518e-01 -3.60476643e-01 7.72488192e-02
-1.29117644e+00 6.36823356e-01 8.04117024e-01 2.90078759e-01
-6.90577149e-01 -8.80288780e-01 -6.13715947e-01 5.81922412e-01
-1.97117865e-01 -9.23673213e-01 8.23729157e-01 -8.42180967e-01
-1.15949345e+00 1.05539453e+00 -5.77854514e-02 -4.64265406e-01
2.22047210e-01 1.99950054e-01 -3.00763011e-01 1.61465019e-01
2.02869982e-01 1.10940933e+00 1.01703584e+00 -8.60005200e-01
-6.98576987e-01 -6.29940033e-01 1.79749429e-01 5.30099077e-03
-3.00338954e-01 -1.03762455e-01 -5.31507693e-02 -1.20326638e+00
3.99478823e-01 -6.33564532e-01 -3.27622145e-02 -1.10931378e-02
-3.37181091e-01 -7.04108119e-01 3.92426044e-01 -2.87791461e-01
1.11959040e+00 -2.16219664e+00 -2.57158484e-02 3.16146046e-01
4.97199953e-01 -1.17829204e-01 -5.11007488e-01 -8.67992863e-02
-5.08910716e-01 1.26569107e-01 -6.29293621e-02 4.00294550e-02
1.97597638e-01 2.12853793e-02 -5.11315227e-01 3.12663585e-01
3.08487892e-01 1.13440537e+00 -5.82011461e-01 -2.69986957e-01
-5.58669567e-02 8.77763405e-02 -8.80600989e-01 7.36219063e-02
2.44454682e-01 4.44511920e-01 -4.41339791e-01 7.31035590e-01
9.57146168e-01 -5.66665292e-01 1.56517357e-01 -4.83899534e-01
1.12108715e-01 1.95285901e-01 -1.02451992e+00 1.49650156e+00
-3.32817584e-01 3.32922578e-01 8.14346001e-02 -1.10783637e+00
6.12634301e-01 -4.87303734e-02 -3.32612097e-02 -1.05817258e+00
4.08990055e-01 6.72595277e-02 6.97305977e-01 -8.62991437e-02
1.25048786e-01 -2.28517652e-01 -2.19864890e-01 4.66674030e-01
3.66410941e-01 1.71440080e-01 2.25796744e-01 -9.06030163e-02
7.60634840e-01 -1.10145502e-01 4.44484919e-01 -8.34360480e-01
6.43620610e-01 -4.60458577e-01 6.02809310e-01 9.05393958e-01
-5.30802369e-01 2.54223496e-01 4.74802434e-01 -6.82110846e-01
-5.85587740e-01 -1.12271559e+00 -2.91765869e-01 1.87219858e+00
2.02223465e-01 -8.16418529e-02 -7.04338253e-01 -4.90178019e-01
1.05158404e-01 2.26957843e-01 -1.09564185e+00 -4.61237639e-01
-4.56541061e-01 -7.30751216e-01 3.96142989e-01 1.04685044e+00
8.34865630e-01 -1.40603101e+00 -6.50263786e-01 -1.52979139e-02
-1.15819685e-01 -1.09419775e+00 -4.60114151e-01 4.06579554e-01
-9.11048591e-01 -9.58060622e-01 -5.44008255e-01 -9.67676997e-01
6.37207806e-01 2.66582638e-01 1.07735991e+00 1.27402827e-01
-4.02469873e-01 -1.42997846e-01 5.45055047e-02 -2.11047590e-01
4.11937952e-01 3.47229898e-01 2.05370262e-01 1.40199199e-01
5.30897200e-01 -5.16733944e-01 -9.76838946e-01 6.01105928e-01
-8.47830117e-01 -3.59060466e-02 8.09541821e-01 1.03290594e+00
5.62695563e-01 1.54252216e-01 8.40071380e-01 -6.49401188e-01
7.83765376e-01 -4.39848840e-01 -6.18526638e-01 9.22739953e-02
-5.07122934e-01 1.42101645e-01 8.33835900e-01 -4.43723083e-01
-1.02102304e+00 -1.80111989e-01 -1.81172386e-01 -2.11730555e-01
-3.26750964e-01 2.65857458e-01 8.19348916e-03 -4.35956270e-01
4.04567242e-01 1.47205353e-01 -4.55066770e-01 -4.84748453e-01
1.64775118e-01 4.14165527e-01 3.61519367e-01 -4.58689600e-01
6.72643602e-01 5.40369153e-01 -1.73112378e-01 -4.48958337e-01
-1.15780401e+00 -2.61385441e-01 -1.07705712e+00 2.20107839e-01
1.08335614e+00 -9.80484128e-01 -1.30843735e+00 6.00900531e-01
-8.57539654e-01 -3.85893703e-01 -2.54221201e-01 4.35176104e-01
-4.82812375e-01 -3.54126282e-02 -6.07348263e-01 -1.39250070e-01
-2.63517946e-01 -9.16243851e-01 9.39395666e-01 3.15688699e-01
-2.30841503e-01 -8.46575201e-01 -3.61261398e-01 1.22767873e-01
6.19018197e-01 -4.32109743e-01 1.12805450e+00 -6.69934869e-01
-5.54102778e-01 1.85408995e-01 -8.06774378e-01 2.14684233e-01
-1.46434680e-01 -5.96465826e-01 -1.06914020e+00 -4.64329749e-01
7.04522803e-02 -5.28698564e-01 1.53293824e+00 5.61274111e-01
2.11045170e+00 -1.41310170e-01 -3.69437277e-01 8.61889362e-01
1.20022786e+00 -8.48486498e-02 5.24961054e-01 1.57817006e-01
5.91380954e-01 5.19905686e-01 2.90018916e-01 4.08928841e-01
4.63186920e-01 3.46626401e-01 3.78230035e-01 2.26932447e-02
-2.13340819e-01 -1.88001007e-01 -2.95179218e-01 5.33679128e-01
-1.35526344e-01 2.35571414e-02 -7.42824554e-01 5.61611831e-01
-1.44218063e+00 -1.11542261e+00 2.97832906e-01 1.74757576e+00
7.03302145e-01 4.71870810e-01 -6.78525344e-02 2.92102545e-01
7.13898003e-01 5.01945019e-01 -6.66474462e-01 -4.19111609e-01
-6.94562041e-04 4.39846426e-01 2.92136878e-01 1.51422709e-01
-1.25641477e+00 9.90623295e-01 7.18967199e+00 9.86193776e-01
-1.31379437e+00 1.40066579e-01 7.58647084e-01 1.23144388e-01
3.72832716e-02 -4.00606722e-01 -8.09765458e-01 3.65285397e-01
6.68358088e-01 1.10343888e-01 5.22599280e-01 5.98668814e-01
7.39175975e-02 -2.99254432e-02 -1.14532852e+00 9.66290534e-01
-1.65111065e-01 -1.45740104e+00 3.67395550e-01 -2.18366176e-01
5.24428546e-01 2.12514296e-01 5.81773557e-02 3.45106184e-01
1.12354100e-01 -1.21177268e+00 7.48689651e-01 4.15477604e-01
8.83725882e-01 -4.88458276e-01 5.96979916e-01 4.40496445e-01
-1.59732747e+00 -6.94286466e-01 -6.11444771e-01 -3.64167899e-01
-2.23350018e-01 4.16778117e-01 -2.88590878e-01 1.35830054e-02
1.00413144e+00 5.93471885e-01 -7.17426300e-01 1.05614901e+00
-3.22356075e-01 4.69223917e-01 1.75269544e-01 9.49217007e-02
1.22068547e-01 1.32724494e-01 -1.03343837e-03 1.28805518e+00
1.06699532e-02 2.47975841e-01 1.70368120e-01 9.75611329e-01
-3.98878455e-01 -2.06524491e-01 -2.52129316e-01 9.81594697e-02
3.92011464e-01 1.39919090e+00 -1.11503279e+00 -3.72599632e-01
-2.61772364e-01 7.70439446e-01 6.00594401e-01 3.69556963e-01
-5.23316741e-01 -6.26682222e-01 6.24229491e-01 1.70316964e-01
6.86010242e-01 -2.07581684e-01 -7.35458374e-01 -1.27836466e+00
-2.70772964e-01 -5.21048009e-01 6.54576063e-01 -6.50483787e-01
-1.69343150e+00 7.63078332e-01 -1.19900838e-01 -9.01521027e-01
1.63196251e-01 -9.97977793e-01 -7.19271004e-01 1.07724118e+00
-1.92213356e+00 -9.66242552e-01 -4.51359183e-01 9.09218252e-01
5.88663936e-01 -1.46147355e-01 6.94267213e-01 3.36192787e-01
-3.21048141e-01 6.50751352e-01 -4.71639037e-01 4.35391635e-01
3.69776070e-01 -1.05003369e+00 4.41129357e-01 5.50991893e-01
-1.14542902e-01 1.00305927e+00 1.32075474e-02 -3.95545006e-01
-8.16699386e-01 -1.14538097e+00 7.97566414e-01 -4.15719822e-02
6.17929280e-01 -6.22239411e-01 -9.86065507e-01 6.52176976e-01
2.31593743e-01 5.94052613e-01 6.23837113e-01 1.68443248e-01
-7.70264983e-01 -3.40178311e-01 -1.21689034e+00 1.15549766e-01
1.22230613e+00 -6.68693960e-01 -1.13409340e+00 1.29797295e-01
3.75455588e-01 -9.38404202e-02 -8.03099573e-01 3.38517070e-01
7.69031286e-01 -9.70571697e-01 1.16371381e+00 -8.06763649e-01
3.29846531e-01 -1.43265367e-01 -1.47629797e-01 -1.42422867e+00
-1.14749908e+00 -2.66374890e-02 3.74064147e-02 6.71070635e-01
1.15129538e-01 -8.52193117e-01 5.59773564e-01 7.76851997e-02
-4.49053198e-02 -8.09556842e-01 -1.02244282e+00 -7.39265561e-01
2.66490906e-01 -2.04014048e-01 6.62442505e-01 7.84145355e-01
-2.69386858e-01 3.78741562e-01 2.06687059e-02 -8.45419895e-03
6.08925641e-01 4.91639286e-01 -4.83805053e-02 -1.82306552e+00
-6.10740706e-02 -8.29333007e-01 -4.34121728e-01 -1.12022769e+00
3.91458958e-01 -1.22065568e+00 -6.80063590e-02 -1.39296699e+00
5.48742652e-01 -2.76388437e-01 -8.14716756e-01 7.30634630e-01
-2.74919681e-02 8.22154999e-01 1.75759733e-01 3.40109944e-01
-4.85989302e-01 3.40614527e-01 1.43622828e+00 -1.22327678e-01
2.15678036e-01 2.60193720e-02 -1.19992411e+00 1.07028913e+00
7.90563762e-01 -4.15964961e-01 -3.37966323e-01 -3.93095464e-01
8.94220844e-02 -2.48213336e-01 5.21418095e-01 -9.58277583e-01
1.37316585e-01 -3.70802492e-01 9.54039276e-01 -4.26179856e-01
8.21177065e-02 -5.46500444e-01 -5.56724966e-01 7.07358956e-01
-5.84864855e-01 9.50407907e-02 2.03337401e-01 3.66153568e-01
-4.19315010e-01 7.20024183e-02 1.38330865e+00 -3.03675860e-01
-1.08931005e+00 6.91042185e-01 -1.58930868e-01 4.15641665e-01
9.45463777e-01 -1.16685309e-01 -6.16797030e-01 2.58339554e-01
-9.50508177e-01 2.06673890e-01 -2.27012947e-01 6.18530512e-01
5.82151651e-01 -1.51318407e+00 -4.91527706e-01 6.92065239e-01
1.59724683e-01 -6.18053079e-01 5.07526815e-01 8.42037559e-01
-1.47067860e-01 8.90869021e-01 -1.00357044e+00 -6.10797942e-01
-7.73963869e-01 6.20250583e-01 4.17233139e-01 -2.96290725e-01
-2.27866113e-01 1.11602402e+00 7.21774459e-01 -3.64780337e-01
5.34511618e-02 -6.39146626e-01 -5.93168497e-01 2.77375221e-01
7.10079014e-01 2.09546328e-01 1.00847550e-01 -5.87529540e-01
-7.13034749e-01 6.46658242e-01 -1.11526281e-01 3.16318631e-01
1.37717140e+00 -2.12470233e-01 -3.37451369e-01 3.86872798e-01
1.01085019e+00 -3.09057921e-01 -1.06616139e+00 -2.44001016e-01
-7.95577541e-02 -4.71481711e-01 1.90412831e-02 -8.56571794e-01
-1.18233383e+00 1.29828441e+00 7.32432485e-01 3.83992970e-01
1.24338663e+00 3.41777802e-02 4.09084529e-01 4.49126124e-01
3.82102460e-01 -1.26011360e+00 1.15842693e-01 8.28139544e-01
1.05848694e+00 -1.28631544e+00 -1.43412516e-01 -2.15859607e-01
-5.40005982e-01 9.81781542e-01 1.00572062e+00 -5.06400347e-01
1.07743669e+00 5.15151918e-02 -2.86124289e-01 -2.10487649e-01
-9.05602217e-01 -1.43110231e-01 6.50664032e-01 4.34148610e-01
5.70308030e-01 1.52365208e-01 -3.66434693e-01 9.97490525e-01
-8.81305784e-02 3.03126991e-01 4.20600586e-02 6.84623361e-01
-5.38065553e-01 -6.65588319e-01 -1.06221832e-01 9.85213220e-01
-3.85528773e-01 -2.15852261e-01 6.05562851e-02 5.17201662e-01
3.96437019e-01 4.53131080e-01 3.75788718e-01 -2.71157593e-01
2.55744666e-01 5.71790822e-02 5.16522169e-01 -6.52539253e-01
-7.92404592e-01 -2.90027648e-01 -5.22971392e-01 -1.01363111e+00
-3.74410480e-01 -3.18582475e-01 -1.25988960e+00 -8.90312567e-02
2.03122422e-02 -1.12198427e-01 2.10200563e-01 9.68355954e-01
5.89232385e-01 8.56125534e-01 4.69078839e-01 -1.10492706e+00
-6.98085010e-01 -1.06311738e+00 -8.85911167e-01 3.23546648e-01
5.17557561e-01 -1.06582642e+00 -2.99952477e-01 -2.23585874e-01]
|
[9.55135440826416, 2.1599888801574707]
|
2f332aac-3cbe-424a-aadc-5c3e0f1ed0d9
|
hypergraph-and-protein-function-prediction
|
1212.0388
| null |
http://arxiv.org/abs/1212.0388v1
|
http://arxiv.org/pdf/1212.0388v1.pdf
|
Hypergraph and protein function prediction with gene expression data
|
Most network-based protein (or gene) function prediction methods are based on
the assumption that the labels of two adjacent proteins in the network are
likely to be the same. However, assuming the pairwise relationship between
proteins or genes is not complete, the information a group of genes that show
very similar patterns of expression and tend to have similar functions (i.e.
the functional modules) is missed. The natural way overcoming the information
loss of the above assumption is to represent the gene expression data as the
hypergraph. Thus, in this paper, the three un-normalized, random walk, and
symmetric normalized hypergraph Laplacian based semi-supervised learning
methods applied to hypergraph constructed from the gene expression data in
order to predict the functions of yeast proteins are introduced. Experiment
results show that the average accuracy performance measures of these three
hypergraph Laplacian based semi-supervised learning methods are the same.
However, their average accuracy performance measures of these three methods are
much greater than the average accuracy performance measures of un-normalized
graph Laplacian based semi-supervised learning method (i.e. the baseline method
of this paper) applied to gene co-expression network created from the gene
expression data.
|
['Loc Tran']
|
2012-12-03
| null | null | null | null |
['protein-function-prediction']
|
['medical']
|
[ 4.26628888e-01 4.26981270e-01 -2.53523201e-01 -4.13503349e-01
-4.98560257e-02 -5.11879563e-01 9.09825936e-02 2.71789670e-01
-1.21935792e-02 1.13938415e+00 -3.11760455e-01 -8.61648843e-02
-5.85384786e-01 -8.54239523e-01 -6.41314387e-01 -1.15054846e+00
-4.79521424e-01 6.03756785e-01 3.71412516e-01 7.03486204e-02
1.41207017e-02 3.15086752e-01 -1.15045035e+00 1.76649615e-01
7.60327339e-01 4.05209422e-01 -9.89568010e-02 4.68152404e-01
-2.74516284e-01 5.56167185e-01 -3.13467085e-01 -1.81119815e-01
-2.82993894e-02 -9.04729962e-01 -9.66710389e-01 2.99997121e-01
1.86076954e-01 7.24638224e-01 -3.56608212e-01 1.41974509e+00
2.22682506e-01 8.24761763e-02 8.90208721e-01 -1.71606493e+00
-3.94798726e-01 6.13694668e-01 -8.44613671e-01 -2.42149562e-01
3.41636747e-01 -3.89347523e-01 1.22436333e+00 -5.07598639e-01
8.72177184e-01 1.36772215e+00 6.96911335e-01 1.03531353e-01
-1.73608851e+00 -4.05906349e-01 -1.36313319e-01 2.00886503e-01
-1.69369471e+00 -5.95631115e-02 6.78658187e-01 -4.00755554e-01
8.32507014e-01 1.03360087e-01 5.86107194e-01 3.26666415e-01
2.78598011e-01 1.66659310e-01 9.71590281e-01 -5.59658170e-01
2.18122929e-01 -1.50479265e-02 5.72339714e-01 1.15829837e+00
5.41845918e-01 -5.13590276e-01 -3.74175757e-01 -8.40472937e-01
2.53960282e-01 8.48114043e-02 -5.62675595e-01 -7.54640043e-01
-1.07975435e+00 7.31354177e-01 2.52887726e-01 7.47397184e-01
-1.18071973e-01 -2.04153791e-01 4.40298975e-01 5.49533725e-01
5.40643573e-01 3.75283450e-01 -6.23512506e-01 3.47317249e-01
-6.67959809e-01 -2.99147040e-01 1.19772506e+00 9.33350801e-01
1.38524151e+00 -2.35354394e-01 4.03750926e-01 7.98431814e-01
3.41286361e-01 -1.68261543e-01 5.72297096e-01 -5.75685084e-01
-1.79386556e-01 1.16258812e+00 -3.72674257e-01 -1.34806001e+00
-5.51611960e-01 5.81466535e-04 -1.09537363e+00 6.27337843e-02
6.16094410e-01 -2.07924187e-01 -7.97286749e-01 1.85103297e+00
2.55409300e-01 3.79357547e-01 2.29717195e-01 4.54987615e-01
5.81498206e-01 6.36267364e-01 6.43688887e-02 -6.98836207e-01
8.21505308e-01 -6.94876611e-01 -8.29683900e-01 3.77901495e-01
1.12930489e+00 -5.30722260e-01 7.48177588e-01 9.61588845e-02
-6.48633540e-01 -3.33612293e-01 -1.09930623e+00 3.93413782e-01
-6.65233374e-01 -1.19594477e-01 7.92621791e-01 2.91369468e-01
-1.04468715e+00 1.11430597e+00 -7.27136910e-01 -9.10605490e-01
1.04419224e-01 7.13042796e-01 -8.16850901e-01 -1.23240367e-01
-1.12940431e+00 6.13131106e-01 6.66649163e-01 -3.48258853e-01
-3.18283349e-01 -4.13069248e-01 -6.68258786e-01 1.88117653e-01
1.54579610e-01 -3.27624112e-01 2.49191523e-01 -1.21665692e+00
-1.03928924e+00 9.78791118e-01 -2.98510313e-01 9.77537129e-03
-6.57981634e-02 5.00393093e-01 -2.49889672e-01 -4.70508914e-03
-1.15798088e-03 4.46692437e-01 9.61131677e-02 -1.08491683e+00
-3.52143854e-01 -6.36248350e-01 -3.77736241e-01 4.24632281e-02
-3.72910589e-01 -3.19817960e-01 -3.46349180e-01 -1.65437341e-01
4.53462571e-01 -1.30941331e+00 -1.89430878e-01 -1.61396358e-02
-7.07211912e-01 -6.20181978e-01 9.75904584e-01 -2.12945640e-01
1.14159954e+00 -2.18834352e+00 4.75794375e-01 7.99138844e-01
3.13520342e-01 6.64141476e-02 -2.11063519e-01 5.76601565e-01
-7.89635420e-01 3.10874313e-01 -5.31409383e-01 3.81283253e-01
-4.42297488e-01 4.36027586e-01 4.74178523e-01 9.01853144e-01
-1.53693855e-01 6.42545044e-01 -9.80865657e-01 -6.21186316e-01
-1.50978580e-01 2.52096832e-01 5.69465570e-02 9.21642557e-02
1.54610211e-02 3.57704222e-01 -5.04245043e-01 4.99616385e-01
6.74263120e-01 -5.40403485e-01 1.06952858e+00 -5.36668062e-01
2.83021927e-01 -2.91319132e-01 -1.00592935e+00 1.44548118e+00
2.62880266e-01 5.79968214e-01 -1.47856399e-01 -1.37122238e+00
9.55672741e-01 3.81298065e-01 9.26930428e-01 1.97826385e-01
-2.47503698e-01 -8.51709247e-02 3.89729917e-01 -4.36964959e-01
-5.36228895e-01 1.27306417e-01 3.32255840e-01 4.78430212e-01
2.92505801e-01 2.40983129e-01 4.34393078e-01 4.73699272e-01
1.36249793e+00 7.17664063e-02 7.19143867e-01 -7.21375525e-01
8.30018401e-01 1.49833739e-01 9.77112055e-01 7.84327462e-02
-7.08207265e-02 2.26331070e-01 1.22084498e+00 -2.10018843e-01
-1.09151196e+00 -7.33551025e-01 -9.95715335e-02 8.28545630e-01
3.40949595e-02 -6.94173336e-01 -8.64881992e-01 -8.68359685e-01
2.88197473e-02 2.07278430e-01 -6.10215127e-01 -3.56011689e-01
-1.66124076e-01 -1.14337420e+00 5.95106244e-01 -5.69791943e-02
2.04418987e-01 -5.50424218e-01 4.71411049e-01 -1.51012661e-02
-1.30924016e-01 -6.59111083e-01 -3.14081609e-01 4.90395129e-01
-1.09115839e+00 -1.67343390e+00 -4.99840170e-01 -1.23160303e+00
1.38031173e+00 2.55449355e-01 7.75318205e-01 1.40301764e-01
-2.96788245e-01 5.88826574e-02 -2.89517939e-01 8.82019773e-02
-4.54257846e-01 -1.59240425e-01 1.90792605e-01 1.11607701e-01
6.29158914e-01 -9.38638091e-01 -2.14661211e-01 5.52665472e-01
-9.42420900e-01 -1.36499330e-01 4.82577890e-01 9.85385716e-01
1.06517947e+00 4.92318511e-01 6.59202874e-01 -1.45238125e+00
5.19469976e-01 -6.28925920e-01 -4.19616640e-01 8.04958344e-01
-1.03942871e+00 2.64768630e-01 5.56349277e-01 -4.29882646e-01
-8.55952740e-01 7.23989904e-01 4.00564104e-01 -3.10890153e-02
-2.34875917e-01 7.38115489e-01 -4.73943859e-01 -3.62895131e-01
5.56425214e-01 1.61329865e-01 1.37292922e-01 -4.21226770e-01
2.82487184e-01 6.09154344e-01 -9.97870974e-03 -4.49157208e-02
4.84013468e-01 3.96307230e-01 7.47445345e-01 -9.05774653e-01
-5.07918775e-01 -6.70313537e-01 -1.07558203e+00 -1.50569519e-02
8.51622343e-01 -5.46083808e-01 -7.77884126e-01 2.60713249e-01
-8.55532765e-01 1.82531685e-01 2.68485963e-01 4.80839372e-01
-5.48780620e-01 9.60357010e-01 -6.20346606e-01 -5.39463997e-01
-1.89176928e-02 -7.61737764e-01 4.69421685e-01 -6.82841465e-02
-3.67068112e-01 -1.29166734e+00 3.95535231e-01 -5.11983410e-03
-1.74870178e-01 5.71746171e-01 1.66858661e+00 -1.06837797e+00
-2.99715940e-02 -2.88325429e-01 -1.39145404e-01 1.77822471e-01
7.49489009e-01 3.11407954e-01 -5.55204332e-01 -4.52637464e-01
-4.33716685e-01 -1.92993149e-01 8.80674601e-01 1.94254726e-01
8.37652683e-01 -1.96501404e-01 -9.41365182e-01 2.56079435e-01
1.57913411e+00 1.93228066e-01 5.29206336e-01 -3.84998649e-01
6.93213105e-01 9.79220271e-01 4.59337443e-01 -7.77357072e-02
-3.17902863e-01 6.94387555e-01 1.40031382e-01 -1.82718575e-01
9.29195955e-02 -2.29995728e-01 4.11159873e-01 1.07784510e+00
-8.67626369e-02 -4.33018893e-01 -7.41694093e-01 2.95742750e-01
-2.34304190e+00 -6.16462827e-01 -7.48522341e-01 2.29250479e+00
9.65368629e-01 -1.37278408e-01 -1.01851299e-01 9.33039859e-02
1.11857021e+00 -3.52722108e-01 -7.10901201e-01 -2.80894518e-01
-4.22797233e-01 -1.60927042e-01 4.36145931e-01 4.22180772e-01
-9.41265047e-01 7.46089101e-01 6.46872330e+00 8.60033572e-01
-4.40952808e-01 -1.36611998e-01 6.51359499e-01 4.00941998e-01
1.55313417e-01 2.68615693e-01 -3.38719428e-01 1.97755128e-01
1.06221175e+00 -5.53888202e-01 2.00005993e-01 7.96034455e-01
1.34936258e-01 -1.81536168e-01 -1.17375338e+00 7.54307628e-01
4.84802611e-02 -7.63647020e-01 -1.01805724e-01 1.39675900e-01
9.30291295e-01 -1.06520228e-01 -4.98181701e-01 -2.49999836e-01
4.91599560e-01 -8.21732104e-01 -5.30927777e-01 7.62772918e-01
6.25702739e-01 -8.55994284e-01 9.10336971e-01 3.69392604e-01
-1.21802473e+00 3.51230949e-01 -8.09973061e-01 1.19878808e-02
-5.35437427e-02 1.11557531e+00 -8.06635618e-01 5.15146077e-01
4.39160556e-01 1.05249596e+00 -4.10598367e-01 9.29937005e-01
-1.53315336e-01 8.52666795e-01 4.05419664e-03 -1.51259825e-02
6.49440438e-02 -5.99403381e-01 3.26275975e-01 1.13829172e+00
1.44428849e-01 6.47431687e-02 3.20879608e-01 6.77957058e-01
-1.31271437e-01 4.59751248e-01 -9.41701353e-01 -3.83460224e-01
4.21020389e-02 1.49424279e+00 -9.35033143e-01 -2.53372043e-01
-8.32875371e-01 1.07771051e+00 5.70130289e-01 4.56341654e-01
-4.54478174e-01 -7.45381057e-01 8.09565544e-01 1.57421902e-01
-1.46574914e-01 -9.71744508e-02 2.19862625e-01 -8.10831308e-01
-4.64386880e-01 -4.30267394e-01 5.34491003e-01 -6.54403687e-01
-1.72163987e+00 2.81027913e-01 -1.47574186e-01 -1.02307701e+00
3.15844566e-02 -8.01504016e-01 -5.30549884e-01 6.80658817e-01
-1.19532120e+00 -7.71206439e-01 -8.27484298e-03 5.35075188e-01
-2.60906294e-02 -2.33694658e-01 1.20956838e+00 2.32391357e-01
-5.51854134e-01 2.94177949e-01 5.49481690e-01 1.34983987e-01
7.99496591e-01 -1.31425786e+00 -1.73531473e-01 3.33584130e-01
1.56699166e-01 7.22061038e-01 3.98195803e-01 -9.27237272e-01
-1.34863365e+00 -1.31114936e+00 1.14578354e+00 2.97778901e-02
5.69486082e-01 -2.30275005e-01 -1.25686622e+00 7.39517093e-01
1.84866175e-01 3.98595678e-03 1.14861405e+00 -3.94138396e-02
-5.09174652e-02 -2.95148995e-02 -1.11864221e+00 3.36942911e-01
1.14641333e+00 -3.21402282e-01 -2.91285336e-01 9.19187009e-01
5.77834547e-01 6.33780837e-01 -1.19006968e+00 5.41316748e-01
2.98270196e-01 -6.61786497e-01 6.98998690e-01 -1.04685116e+00
-3.64370309e-02 -5.97785652e-01 -7.86816850e-02 -1.45774603e+00
-8.45606744e-01 -4.25649732e-01 1.76121369e-01 1.19648504e+00
7.60570228e-01 -6.14327133e-01 9.61195588e-01 1.79679424e-01
4.32009488e-01 -5.22810221e-01 -7.86628604e-01 -7.75255561e-01
-2.52526820e-01 4.11346704e-01 8.76767039e-02 1.57084215e+00
8.88282061e-01 8.46496701e-01 -2.07399234e-01 -6.17936440e-02
7.69175649e-01 2.19883129e-01 4.08268809e-01 -1.61030829e+00
-7.31521621e-02 -3.68801430e-02 -1.04191685e+00 -6.01339042e-01
6.35122418e-01 -1.31507206e+00 -3.32098901e-02 -1.35581183e+00
7.29625583e-01 -1.91780347e-02 -5.75335622e-01 7.77632773e-01
-3.53658795e-02 9.97294337e-02 -5.83251297e-01 2.96687692e-01
-6.01085424e-01 1.96861491e-01 9.66874957e-01 -1.58147827e-01
-2.99285680e-01 -9.48948378e-04 -4.06018287e-01 1.02201962e+00
9.02714491e-01 -5.83379209e-01 -6.47194386e-01 3.65723044e-01
1.73707277e-01 -5.67121170e-02 -2.17775315e-01 -7.51332819e-01
5.28916717e-01 -1.94243073e-01 3.03191364e-01 -1.41519248e-01
-2.04852279e-02 -9.64770734e-01 5.24988413e-01 6.38054848e-01
-5.78455925e-01 1.16829082e-01 -1.84057266e-01 9.74157453e-01
-4.04443264e-01 -2.06582531e-01 7.95695543e-01 -3.60714376e-01
-5.10726750e-01 2.49726146e-01 -5.24063468e-01 -3.09731275e-01
1.11363363e+00 -2.53737807e-01 -2.96158623e-02 -4.74647909e-01
-1.13945270e+00 6.30036592e-02 4.05275196e-01 -1.75174698e-02
5.58393359e-01 -1.30256033e+00 -6.14766598e-01 -1.05573051e-01
3.46022248e-01 -5.59546888e-01 -1.60497278e-01 8.52215409e-01
-2.31024787e-01 5.35980940e-01 -2.49162927e-01 -3.15575093e-01
-1.84701860e+00 6.16995871e-01 2.27087051e-01 -2.91391402e-01
-6.29716218e-02 5.25999606e-01 4.41539645e-01 -6.76220238e-01
2.49387175e-02 3.07300955e-01 -4.60749537e-01 -1.11786067e-01
3.36426385e-02 6.08021319e-01 -1.33129507e-01 -8.82639349e-01
-3.73455554e-01 7.51553595e-01 5.60583249e-02 4.41547066e-01
1.31508696e+00 -7.91986585e-02 -1.11218154e+00 5.88935077e-01
1.55842388e+00 -1.86860695e-01 -5.05024433e-01 -4.71710205e-01
3.40152204e-01 -2.67739594e-01 -1.92823023e-01 -6.45273268e-01
-1.07619417e+00 5.76774597e-01 5.29839814e-01 1.25487149e-01
9.75687623e-01 1.40558675e-01 2.42727455e-02 7.42408991e-01
2.29602486e-01 -8.76238823e-01 -1.52743593e-01 2.47988895e-01
4.05619740e-01 -9.34241474e-01 2.35997647e-01 -1.20196903e+00
-4.56479818e-01 1.26057279e+00 7.19418108e-01 1.28150955e-01
9.32891607e-01 1.78304896e-01 -1.79037347e-01 -3.78139108e-01
-7.93226719e-01 -1.85354650e-01 1.97489381e-01 7.21183717e-01
1.08068573e+00 4.12877090e-02 -8.48833501e-01 2.21697405e-01
3.20887536e-01 -8.53430554e-02 3.48768145e-01 7.87980855e-01
-5.98993361e-01 -1.25626314e+00 1.45897150e-01 6.51320100e-01
-3.62271786e-01 -8.71424191e-03 -1.07148635e+00 4.62241679e-01
5.54719791e-02 8.86569023e-01 -2.43382648e-01 -4.58696514e-01
1.23402044e-01 5.22187293e-01 1.85924828e-01 -5.92115998e-01
6.86934739e-02 3.29304545e-04 1.27746344e-01 -4.46305335e-01
-6.63421392e-01 -3.19452584e-01 -1.74812210e+00 -2.75864869e-01
-7.07894027e-01 5.90644598e-01 3.06260437e-01 6.02645993e-01
5.79277515e-01 4.00001466e-01 5.78832448e-01 -1.72473222e-01
-1.60069585e-01 -9.34042573e-01 -1.31568849e+00 6.71745241e-01
-3.84560108e-01 -5.01260400e-01 -3.89080107e-01 2.86535144e-01]
|
[6.655246734619141, 5.479518413543701]
|
037d906a-e48c-4c64-8c27-4d762a2f5652
|
activity-auto-completion-predicting-human
| null | null |
http://openaccess.thecvf.com/content_iccv_2015/html/Xu_Activity_Auto-Completion_Predicting_ICCV_2015_paper.html
|
http://openaccess.thecvf.com/content_iccv_2015/papers/Xu_Activity_Auto-Completion_Predicting_ICCV_2015_paper.pdf
|
Activity Auto-Completion: Predicting Human Activities From Partial Videos
|
In this paper, we propose an activity auto-completion (AAC) model for human activity prediction by formulating activity prediction as a query auto-completion (QAC) problem in information retrieval. First, we extract discriminative patches in frames of videos. A video is represented based on these patches and divided into a collection of segments, each of which is regarded as a character typed in the search box. Then a partially observed video is considered as an activity prefix, consisting of one or more characters. Finally, the missing observation of an activity is predicted as the activity candidates provided by the auto-completion model. The candidates are matched against the activity prefix on-the-fly and ranked by a learning-to-rank algorithm. We validate our method on UT-Interaction Set #1 and Set #2 [19]. The experimental results show that the proposed activity auto-completion model achieves promising performance.
|
['Zhen Xu', 'Laiyun Qing', 'Jun Miao']
|
2015-12-01
| null | null | null |
iccv-2015-12
|
['activity-prediction', 'activity-prediction']
|
['computer-vision', 'time-series']
|
[ 6.03260815e-01 -9.05638486e-02 -6.16486669e-01 -2.83016145e-01
-8.75608087e-01 -2.89828509e-01 5.02625942e-01 6.67877942e-02
-3.24483931e-01 5.90211153e-01 5.35394609e-01 4.30440426e-01
1.70560658e-03 -2.08081678e-01 -5.91052651e-01 -5.03362656e-01
-4.32346761e-01 3.26069176e-01 5.56361437e-01 3.62949252e-01
2.35843301e-01 2.40834624e-01 -1.67392945e+00 7.19488859e-01
6.97073519e-01 9.75691378e-01 5.46141744e-01 5.72346270e-01
1.13153838e-01 1.09418631e+00 -4.78730768e-01 3.78465876e-02
1.27955437e-01 -7.56600082e-01 -8.29353273e-01 5.37851572e-01
2.36014858e-01 -5.16001165e-01 -5.68239808e-01 6.53051257e-01
7.64916986e-02 4.03539211e-01 4.22568232e-01 -1.26400530e+00
1.09547354e-01 3.59289423e-02 -4.87623781e-01 3.13867629e-01
1.01661849e+00 -2.05795750e-01 1.07437503e+00 -1.36646569e+00
9.93417203e-01 9.13572013e-01 4.12458628e-02 4.37473059e-01
-8.40553939e-01 -3.78336847e-01 2.89682537e-01 6.23022497e-01
-1.42858517e+00 -2.30258584e-01 8.45596075e-01 -5.04311085e-01
7.23749101e-01 2.46553838e-01 1.25370097e+00 1.02733064e+00
2.77295429e-02 1.53400469e+00 5.69323719e-01 -2.64469951e-01
3.51778626e-01 -3.85530919e-01 4.65253256e-02 7.95468867e-01
5.66009209e-02 -2.77915835e-01 -9.38495874e-01 -3.87182385e-01
6.99550450e-01 3.84636760e-01 -4.55779910e-01 -6.38077676e-01
-1.54030597e+00 3.84789616e-01 3.06247585e-02 -2.58717798e-02
-7.77912199e-01 1.08526545e-02 3.54368776e-01 5.62864766e-02
2.16218725e-01 7.40870461e-03 -2.44216830e-01 -4.28143680e-01
-1.04326105e+00 3.60885859e-01 7.17952609e-01 1.10351658e+00
8.97087991e-01 -3.45889807e-01 -5.39395869e-01 7.88055420e-01
2.96915740e-01 1.92226261e-01 4.58449513e-01 -1.20299149e+00
5.64769506e-01 7.74531782e-01 4.17398125e-01 -9.96372581e-01
1.07267819e-01 -2.88478613e-01 -4.28698063e-01 -2.34990209e-01
2.09092364e-01 1.41268343e-01 -4.93886769e-01 1.33600664e+00
5.09393990e-01 7.03681588e-01 -1.91521272e-02 1.07112908e+00
3.05260926e-01 1.00266814e+00 2.29220726e-02 -7.88867831e-01
1.35906732e+00 -1.37687361e+00 -7.57075727e-01 2.65881047e-02
6.18504584e-01 -6.47958040e-01 7.87511289e-01 7.87980139e-01
-1.13389289e+00 -7.88883567e-01 -1.06302798e+00 1.88862681e-01
1.96887091e-01 5.44954181e-01 1.41482860e-01 1.73366275e-02
-5.44763029e-01 5.12841284e-01 -1.13366652e+00 -3.91859740e-01
1.92506731e-01 -1.62897725e-03 -4.06446010e-01 -2.96340764e-01
-8.01494658e-01 1.66973159e-01 4.20505345e-01 -3.00668240e-01
-1.51632309e+00 -1.71950370e-01 -9.09800172e-01 7.11507723e-02
7.28301346e-01 -2.05657199e-01 1.25550723e+00 -1.18159592e+00
-1.13217306e+00 6.01728797e-01 -7.00350761e-01 -4.69974935e-01
4.45910692e-01 -5.59808314e-01 -7.91634202e-01 6.28223419e-01
2.52986878e-01 4.93057847e-01 9.49752152e-01 -1.13262010e+00
-1.17592239e+00 -1.58121720e-01 -3.17218490e-02 5.79597473e-01
-3.31795007e-01 4.91983332e-02 -1.29302025e+00 -7.49920189e-01
2.04417363e-01 -9.91533160e-01 2.04112157e-02 -4.97705974e-02
-1.59229323e-01 -3.88757646e-01 1.03294134e+00 -9.45077419e-01
1.82268083e+00 -2.07214022e+00 1.96117595e-01 3.57327759e-01
-4.17804234e-02 1.46351323e-01 -2.75711477e-01 5.41791141e-01
-4.16261442e-02 -3.69960785e-01 1.68388877e-02 -2.21569613e-01
-4.72722113e-01 1.30265117e-01 5.48251271e-02 5.14901221e-01
-6.95951730e-02 4.70547259e-01 -1.23553061e+00 -9.96440351e-01
1.31627485e-01 2.52523988e-01 -5.69899142e-01 4.33891892e-01
-3.55965465e-01 4.13070887e-01 -8.22209895e-01 6.64204299e-01
2.50817537e-01 -2.74599075e-01 3.41812313e-01 -2.07091719e-01
-1.85801074e-01 1.11744940e-01 -1.25791240e+00 2.07774591e+00
1.55601352e-01 4.03782636e-01 -3.87569875e-01 -8.12061191e-01
6.57374263e-01 4.20470744e-01 1.01061249e+00 -5.26415408e-01
-4.90239978e-01 2.98687331e-02 -2.93543756e-01 -6.99085712e-01
6.56895757e-01 8.39591146e-01 2.38841865e-02 3.81844819e-01
2.22726360e-01 7.44822383e-01 6.03211522e-01 3.51993024e-01
1.26758325e+00 5.27973473e-01 3.88009876e-01 1.76914513e-01
9.06189978e-01 1.03469469e-01 1.02326882e+00 5.67772508e-01
-3.78671706e-01 5.95511794e-01 4.40611035e-01 -4.15490627e-01
-1.03016841e+00 -8.67567062e-01 3.10252547e-01 1.08452749e+00
1.93163782e-01 -1.13516200e+00 -8.32831264e-01 -8.82492721e-01
-4.84244555e-01 2.41558552e-01 -4.97256070e-01 5.65780737e-02
-6.94252193e-01 1.66316167e-01 2.01617166e-01 4.06615525e-01
6.59488320e-01 -1.18222833e+00 -7.60445058e-01 4.36235070e-01
-6.32086575e-01 -8.57363045e-01 -9.19744611e-01 -3.22283119e-01
-9.01015222e-01 -1.31581175e+00 -8.26926112e-01 -1.05457771e+00
7.30562091e-01 7.44800806e-01 8.51119161e-01 2.97014445e-01
-9.78844091e-02 5.80514967e-01 -7.06906438e-01 2.23228812e-01
-1.79367155e-01 -2.72788376e-01 8.88093784e-02 7.14940012e-01
4.26448137e-01 -2.70801038e-01 -1.05763113e+00 7.05708623e-01
-6.84858322e-01 1.58142839e-02 6.59565568e-01 6.56480312e-01
1.29896271e+00 -1.06026143e-01 1.59205839e-01 -6.02948606e-01
3.23790640e-01 -4.88591611e-01 -4.71887857e-01 4.64286715e-01
-4.16534334e-01 -1.99039117e-01 2.43128762e-01 -4.98346150e-01
-1.10940278e+00 6.72209203e-01 5.02649903e-01 -7.61246085e-01
-1.83998570e-01 7.25277603e-01 -5.03407657e-01 5.24679005e-01
4.14776564e-01 5.49596667e-01 -3.85660440e-01 -5.86727202e-01
9.23536345e-02 4.62540090e-01 7.28713334e-01 -4.00389701e-01
5.04713774e-01 4.22998369e-01 -2.85623610e-01 -9.61828947e-01
-6.47415817e-01 -9.82260764e-01 -6.30649805e-01 -6.69348240e-01
1.04438329e+00 -1.30667436e+00 -3.02010745e-01 1.78796306e-01
-1.09295917e+00 3.42131928e-02 -1.59070507e-01 8.12879264e-01
-6.27786636e-01 8.33752155e-01 -6.03162050e-01 -7.98452020e-01
2.96044210e-03 -9.76243556e-01 9.46685791e-01 3.36366326e-01
-3.62283230e-01 -3.92744958e-01 3.63775313e-01 6.08444571e-01
-5.31825244e-01 1.10526472e-01 5.46389997e-01 -8.47475648e-01
-7.42837846e-01 -5.39426386e-01 2.99825698e-01 9.31814015e-02
2.24833819e-03 -1.59385979e-01 -5.32798290e-01 -4.37764227e-01
-3.60981315e-01 -2.35665828e-01 5.52835107e-01 1.73505530e-01
1.43638372e+00 -4.01375234e-01 -6.98300600e-01 1.61761880e-01
1.11075437e+00 5.71138561e-01 8.79846454e-01 8.00938755e-02
5.21204412e-01 3.18577915e-01 1.39365375e+00 7.77919054e-01
4.38378714e-02 9.35201466e-01 2.31701061e-01 2.38052174e-01
3.40379477e-02 -7.96003759e-01 5.06793916e-01 7.93418407e-01
-4.10289735e-01 -2.90672094e-01 -5.39326489e-01 5.24447501e-01
-2.22359252e+00 -1.44812226e+00 -2.02602834e-01 2.51541376e+00
4.83721435e-01 1.38932198e-01 3.56790721e-01 1.66244045e-01
7.59593248e-01 2.39426896e-01 -5.40830135e-01 3.53448004e-01
8.81921649e-02 -2.06519663e-01 -6.84022382e-02 2.54073173e-01
-1.15540838e+00 7.85319328e-01 5.63292599e+00 9.66024041e-01
-2.85585374e-01 1.54028749e-02 3.35909098e-01 -9.45490077e-02
1.43904109e-02 2.33180121e-01 -5.93002141e-01 6.02598667e-01
4.82655406e-01 7.15129897e-02 3.26186687e-01 1.16018212e+00
5.74784636e-01 -5.82177281e-01 -1.29828656e+00 1.13878107e+00
4.25806850e-01 -1.01481175e+00 2.10943341e-01 1.08221442e-01
8.49126458e-01 -3.41743290e-01 -3.35302740e-01 2.41966128e-01
-2.12263122e-01 -4.18684036e-01 6.86598837e-01 6.42497122e-01
6.08966947e-01 -7.32234240e-01 3.08762193e-01 5.41003644e-01
-1.67399633e+00 -1.42608017e-01 -2.35492751e-01 1.52135998e-01
1.16055429e-01 2.60038435e-01 -6.09981894e-01 3.65689665e-01
6.90766633e-01 1.06614637e+00 -4.03851002e-01 1.36410773e+00
-2.57525176e-01 8.33072007e-01 3.10443752e-02 1.75234210e-02
-1.73172541e-02 -4.79342192e-01 6.62145615e-01 1.09479392e+00
3.96464974e-01 4.49798673e-01 6.87787771e-01 3.27350199e-01
-4.71666977e-02 1.56582907e-01 -5.19230664e-01 -8.80301446e-02
5.82832515e-01 1.23324132e+00 -5.25854409e-01 -8.26693177e-01
-7.41292417e-01 1.49153388e+00 1.45947650e-01 4.65891033e-01
-7.15080738e-01 -2.75267661e-01 3.82621080e-01 1.46185532e-01
3.23029131e-01 1.99701842e-02 6.20072305e-01 -1.26563263e+00
3.29501569e-01 -9.58875060e-01 7.92477310e-01 -1.10815680e+00
-6.80361807e-01 3.72763246e-01 8.66490528e-02 -2.06411672e+00
-4.08103943e-01 -8.62359256e-03 -4.12916034e-01 3.74190450e-01
-8.51951122e-01 -9.81036603e-01 -6.82127178e-01 9.75180984e-01
1.14760232e+00 -2.67113954e-01 7.09409118e-01 2.86172926e-01
-4.74345773e-01 2.56791532e-01 4.38342318e-02 3.72213602e-01
4.28915262e-01 -8.40780735e-01 -1.48866206e-01 9.99760032e-01
3.27217519e-01 3.95171583e-01 4.31428105e-01 -1.09576857e+00
-1.28678834e+00 -1.03468192e+00 9.97770846e-01 -2.60562778e-01
3.08979928e-01 -3.07500422e-01 -9.59124982e-01 6.90131366e-01
6.28230721e-02 -1.46431193e-01 8.08433294e-01 -3.42551589e-01
-6.80920035e-02 -5.22380807e-02 -6.31383061e-01 5.96093595e-01
1.32306552e+00 -5.67090988e-01 -7.95747042e-01 3.73048395e-01
5.85641086e-01 -2.38781959e-01 -7.26068676e-01 1.70389488e-01
6.50606573e-01 -8.05897832e-01 8.55896831e-01 -6.91682816e-01
3.55144531e-01 -5.17825544e-01 -7.29163587e-02 -9.31383550e-01
-3.95406127e-01 -6.65865541e-01 -5.99237621e-01 9.48473096e-01
2.11664557e-01 3.69611889e-01 1.18733215e+00 2.70800382e-01
-1.86283156e-01 -6.84173942e-01 -7.57002771e-01 -6.83529377e-01
-1.02891743e+00 -2.85312772e-01 3.22127908e-01 7.21374512e-01
4.79904503e-01 3.27514023e-01 -7.13914037e-01 2.14228481e-02
5.93451262e-01 1.02183186e-01 8.26613545e-01 -1.19797897e+00
-4.48222220e-01 2.23148659e-01 -2.63703555e-01 -1.63137484e+00
-9.17266607e-02 -5.68817317e-01 1.28693134e-01 -1.49785137e+00
6.06481612e-01 1.62751213e-01 -4.21221972e-01 3.28017086e-01
-3.40239443e-02 -1.51638657e-01 2.34771017e-02 8.81601512e-01
-1.30162454e+00 5.16079426e-01 1.04684985e+00 -2.41783053e-01
-4.35283214e-01 2.44886756e-01 1.67371392e-01 7.96746075e-01
5.17680347e-01 -4.37918097e-01 -6.03356779e-01 8.25978443e-02
-3.27908210e-02 7.66556978e-01 1.51487350e-01 -1.32486105e+00
2.28149682e-01 -2.57148087e-01 6.10753417e-01 -1.21789169e+00
5.39442182e-01 -8.75823617e-01 2.80609101e-01 6.84226632e-01
-5.03243685e-01 7.89465383e-02 -4.74256575e-01 1.11887741e+00
-4.32231098e-01 -1.06504694e-01 2.07036957e-01 -9.86716747e-02
-1.08933365e+00 5.93395472e-01 -6.03577554e-01 -1.89071208e-01
1.14672613e+00 -5.61972678e-01 1.23694912e-01 -4.48194057e-01
-9.37829316e-01 3.56872946e-01 3.30702513e-01 4.61686105e-01
1.04799879e+00 -1.67967427e+00 -4.55462188e-01 1.09664969e-01
6.65088654e-01 -1.70872688e-01 3.18994552e-01 7.07611740e-01
-3.95001680e-01 3.08759630e-01 -7.34238699e-02 -6.76548362e-01
-1.59589648e+00 7.12400436e-01 -4.87604402e-02 -5.02068758e-01
-5.27491212e-01 3.41290027e-01 1.48581013e-01 4.26442325e-01
6.80423617e-01 -1.10208385e-01 -4.42177504e-01 4.10160935e-03
7.48822212e-01 5.33698201e-01 -3.15537453e-01 -7.83227503e-01
-3.83642435e-01 4.22515035e-01 -5.07850647e-02 -4.16526049e-01
1.01020813e+00 5.98467945e-04 1.60123780e-01 2.14508981e-01
1.21127141e+00 -7.80410469e-02 -1.61851251e+00 -5.26880920e-01
1.91996142e-01 -7.74221480e-01 -4.20144558e-01 -5.84539533e-01
-9.07961488e-01 4.83702272e-01 6.92308247e-01 -3.20673585e-01
1.14801919e+00 9.92130786e-02 7.47572184e-01 5.69228768e-01
5.08520544e-01 -1.44510698e+00 9.31107044e-01 2.13186786e-01
8.59579623e-01 -7.78042316e-01 4.22141561e-03 -2.82349855e-01
-8.96218598e-01 7.23769307e-01 7.90228724e-01 -3.22184972e-02
4.81852949e-01 -3.53700578e-01 -4.75554258e-01 -1.45560682e-01
-7.28225946e-01 -8.55801925e-02 3.04651916e-01 5.44984698e-01
1.47147179e-01 -9.51370597e-02 -6.20625734e-01 5.07762372e-01
5.39756477e-01 3.83799672e-01 2.86207378e-01 1.30915344e+00
-7.54989386e-01 -9.63966608e-01 -3.70927066e-01 3.24516743e-01
-2.04897076e-01 3.50554109e-01 -4.20764506e-01 3.88392776e-01
7.52472430e-02 9.54295456e-01 1.11662120e-01 -6.54780567e-01
4.27405089e-01 2.24692017e-01 3.04315954e-01 -5.39951622e-01
-1.20268814e-01 6.91817462e-01 1.71267778e-01 -9.43857551e-01
-4.46628958e-01 -1.01490843e+00 -1.31975543e+00 3.37026209e-01
-1.09890364e-01 4.57283944e-01 1.28439948e-01 7.16538966e-01
3.70309860e-01 1.59725308e-01 7.70416021e-01 -5.97108841e-01
-2.29922622e-01 -9.06817257e-01 -5.75460732e-01 7.44742751e-01
5.75136393e-02 -5.92723906e-01 -1.86484680e-02 6.31081820e-01]
|
[8.573851585388184, 0.5433752536773682]
|
fb4e5fd7-b40e-4465-9470-ac95b29b5e0d
|
economic-impacts-of-ai-augmented-r-d
|
2212.08198
| null |
https://arxiv.org/abs/2212.08198v2
|
https://arxiv.org/pdf/2212.08198v2.pdf
|
Economic impacts of AI-augmented R&D
|
Since its emergence around 2010, deep learning has rapidly become the most important technique in Artificial Intelligence (AI), producing an array of scientific firsts in areas as diverse as protein folding, drug discovery, integrated chip design, and weather prediction. As more scientists and engineers adopt deep learning, it is important to consider what effect widespread deployment would have on scientific progress and, ultimately, economic growth. We assess this impact by estimating the idea production function for AI in two computer vision tasks that are considered key test-beds for deep learning and show that AI idea production is notably more capital-intensive than traditional R&D. Because increasing the capital-intensity of R&D accelerates the investments that make scientists and engineers more productive, our work suggests that AI-augmented R&D has the potential to speed up technological change and economic growth.
|
['Neil Thompson', 'Nicholas Emery-Xu', 'Tamay Besiroglu']
|
2022-12-15
| null | null | null | null |
['protein-folding']
|
['natural-language-processing']
|
[-3.24418783e-01 1.94009498e-01 -2.19117522e-01 2.20006138e-01
-1.32103398e-01 -5.02681732e-01 5.83949804e-01 2.08766103e-01
-4.95026231e-01 6.63185716e-01 5.38801849e-02 -8.04766953e-01
4.10706066e-02 -1.00730562e+00 -9.11983609e-01 -3.85557175e-01
1.44987591e-02 4.15784746e-01 -4.79310334e-01 -5.02201617e-02
3.71183485e-01 8.40081871e-01 -1.03699088e+00 -1.76754817e-01
8.97854030e-01 9.94469702e-01 3.08600087e-02 2.95083493e-01
-1.66738346e-01 7.22408473e-01 -6.32456601e-01 -4.66845274e-01
4.19204772e-01 -1.93394408e-01 -5.68308890e-01 -1.96194977e-01
-8.30716491e-02 6.97490796e-02 -3.98041129e-01 7.55885601e-01
2.55716592e-01 -2.45093759e-02 5.66968501e-01 -9.18251455e-01
-8.83227825e-01 5.82947612e-01 -7.40166605e-01 2.38284729e-02
-3.39249253e-01 6.92508936e-01 9.46342766e-01 -6.75054014e-01
4.84321445e-01 1.03611887e+00 7.18548954e-01 2.22582445e-02
-8.91710699e-01 -9.90031600e-01 -1.40827537e-01 5.09840176e-02
-8.21877182e-01 -2.36558661e-01 7.79020786e-01 -5.82502186e-01
9.20296848e-01 -2.49763310e-01 9.94527519e-01 5.29635727e-01
7.92590618e-01 7.61957765e-01 7.61579335e-01 -4.16077107e-01
6.88837171e-01 -9.17623714e-02 -1.19320326e-01 8.97843421e-01
8.48317444e-01 -1.06750699e-02 -3.18531960e-01 6.78205863e-02
1.13186157e+00 2.68258899e-01 2.15709999e-01 -7.28057623e-02
-1.41255891e+00 9.39552069e-01 5.74367821e-01 4.57683861e-01
-7.50178933e-01 6.15994990e-01 2.77262300e-01 4.65918094e-01
4.07333106e-01 1.46484959e+00 -7.76483417e-01 -2.56434411e-01
-7.06217229e-01 4.72177774e-01 5.78357756e-01 3.34777236e-01
7.31978714e-01 4.48041707e-01 4.03228939e-01 4.25069064e-01
-1.65293694e-01 5.14682591e-01 3.49509299e-01 -1.48483562e+00
5.23303123e-03 1.03383601e+00 1.44446760e-01 -8.34988952e-01
-4.45532888e-01 -9.51725364e-01 -9.85092461e-01 2.16973305e-01
3.79752070e-01 -6.42821252e-01 -7.74129629e-01 1.38794196e+00
-1.95044070e-01 3.03315893e-02 -1.56618357e-02 6.33670628e-01
3.79267842e-01 8.19623768e-01 -2.28259750e-02 -8.32859725e-02
1.02307403e+00 -8.43081057e-01 -2.57763118e-01 -3.52898479e-01
8.16919386e-01 -4.33426797e-01 7.18408167e-01 5.80289781e-01
-1.07031178e+00 -6.32444561e-01 -8.79863739e-01 9.41631719e-02
-4.58385497e-01 -1.28384635e-01 1.38183367e+00 6.71907842e-01
-6.98222697e-01 7.08427846e-01 -7.12950528e-01 -1.49104908e-01
8.93939853e-01 4.27901596e-01 -1.78251565e-01 5.46373846e-03
-9.09892499e-01 1.05096424e+00 -2.89688781e-02 -3.72412801e-01
-7.38693714e-01 -1.02371716e+00 -2.62408882e-01 3.86237383e-01
6.08849674e-02 -9.78904963e-01 8.97153139e-01 -1.09292924e+00
-1.20435798e+00 5.58634758e-01 3.07118744e-01 -8.32558811e-01
-2.05729492e-02 -1.99366510e-01 -3.75964418e-02 -3.40862215e-01
-2.66942054e-01 8.03656578e-01 5.10333896e-01 -6.32315576e-01
-6.57590747e-01 -6.69364154e-01 -1.16944090e-01 -1.49871647e-01
-6.49262488e-01 -2.12851688e-01 9.56307724e-02 -4.62628603e-01
-3.41586560e-01 -1.03534925e+00 -4.97683018e-01 -4.61914912e-02
1.38279751e-01 -3.86776030e-01 4.65798646e-01 -5.75848997e-01
5.58976769e-01 -1.72239912e+00 3.04705292e-01 -3.95259000e-02
5.35102129e-01 4.82871324e-01 -1.37814894e-01 4.50398214e-02
2.62258053e-01 2.30116531e-01 1.55580357e-01 3.34670693e-01
-3.56273949e-01 -2.05681950e-01 -1.88156262e-01 2.34147161e-01
4.65270728e-01 1.21788573e+00 -8.92426074e-01 2.70385593e-01
8.79306421e-02 1.58225194e-01 -4.29930508e-01 -7.68501833e-02
-5.68154931e-01 3.41860831e-01 -5.89544415e-01 7.00098872e-01
2.05489710e-01 -4.92328882e-01 1.36096001e-01 3.70923191e-01
-1.53661892e-01 6.65791007e-03 -2.69279242e-01 1.45179975e+00
-5.43458998e-01 1.16442943e+00 -2.81411260e-01 -1.46824789e+00
1.12926447e+00 -1.85109454e-03 6.46036148e-01 -1.02189171e+00
1.56165048e-01 7.46285394e-02 5.72745740e-01 -2.12429568e-01
1.96103022e-01 -1.83169559e-01 1.05795879e-02 4.88545150e-01
-1.73183233e-01 -2.83930093e-01 -1.02240004e-01 5.35418652e-02
1.54686701e+00 -2.33901069e-01 1.11421034e-01 -3.74529004e-01
-4.06573266e-02 2.82690883e-01 5.91425836e-01 4.21648383e-01
-1.15256481e-01 -3.32611352e-01 4.79767442e-01 -8.64765942e-01
-1.48687840e+00 -9.09483075e-01 3.28006178e-01 9.31430161e-01
-3.24692875e-01 2.67041326e-01 -5.38062394e-01 -3.82331252e-01
5.08242607e-01 4.42337483e-01 -3.49157065e-01 -3.12336713e-01
-3.79095316e-01 -6.37400687e-01 5.34640610e-01 4.90239531e-01
5.25441885e-01 -1.07777154e+00 -6.36937976e-01 1.29949674e-01
4.78535384e-01 -7.76119590e-01 2.20181778e-01 2.36410245e-01
-1.17908227e+00 -6.03046298e-01 -1.17812836e+00 -7.08368957e-01
5.32683671e-01 1.47024646e-01 1.15215671e+00 -1.57039016e-01
-6.72461748e-01 1.23955302e-01 -1.30438894e-01 -1.08322048e+00
-3.81409317e-01 2.63711065e-01 3.77923787e-01 -4.37817723e-01
4.03549552e-01 -5.35485744e-01 -6.60560250e-01 -2.28610858e-01
-3.68050367e-01 1.30262256e-01 1.00798011e+00 5.75495899e-01
9.17685926e-02 3.72121155e-01 1.04987860e+00 -6.76266253e-01
6.30451560e-01 -6.08840823e-01 -7.33888447e-01 2.85369232e-02
-8.67735922e-01 4.23615128e-01 6.94275796e-01 -3.03971380e-01
-8.05673718e-01 -1.96218327e-01 1.81894332e-01 -2.85543054e-01
8.50114003e-02 8.87400627e-01 2.01591313e-01 -4.55218442e-02
7.64981270e-01 -1.43195674e-01 1.06946595e-01 -4.51638937e-01
2.83405066e-01 4.93347347e-01 2.68561274e-01 -3.82612973e-01
6.06954336e-01 -1.83516759e-02 4.52560335e-01 -1.07446313e+00
-3.71648192e-01 -3.51824462e-02 -1.53874152e-03 -2.75827765e-01
5.54244101e-01 -9.98252034e-01 -9.01273847e-01 3.04468244e-01
-1.03518331e+00 -3.74127477e-01 -2.24627525e-01 4.14643466e-01
-2.34497786e-02 -1.73972383e-01 -3.72179776e-01 -5.75361669e-01
-4.65069652e-01 -9.88467634e-01 5.55455923e-01 6.34413779e-01
-2.70895183e-01 -9.90727246e-01 -1.58432667e-04 4.95721132e-01
5.51893711e-01 2.14218348e-01 1.36644554e+00 -2.86542803e-01
-7.87927568e-01 -2.31467187e-01 -4.32080299e-01 2.65563458e-01
8.06523636e-02 -9.82954502e-02 -7.75943100e-01 -2.48230528e-02
-3.06283116e-01 -3.57088298e-01 9.10716057e-01 7.20848680e-01
1.09100628e+00 -1.36823710e-02 -4.13602203e-01 4.94868070e-01
1.10810864e+00 6.45453572e-01 4.94218409e-01 3.57664108e-01
6.39542282e-01 5.08287787e-01 3.17221612e-01 2.95478821e-01
1.46514311e-01 2.06666648e-01 1.03059553e-01 -4.09563959e-01
2.91849971e-01 1.87682256e-03 3.36481899e-01 8.03505421e-01
-3.77512991e-01 -7.85619244e-02 -1.42544341e+00 5.77366948e-01
-1.67518938e+00 -7.76683450e-01 1.88910991e-01 1.99089313e+00
6.18764818e-01 3.46725285e-01 2.34979928e-01 -1.17034130e-01
1.01859607e-01 -2.85845101e-01 -1.04334700e+00 -6.11237764e-01
-9.41328481e-02 6.63190067e-01 7.36244857e-01 -8.13313127e-02
-5.14492512e-01 9.57827806e-01 6.56623745e+00 4.99703407e-01
-1.53092861e+00 7.99308880e-04 1.38553917e+00 -2.48246238e-01
-3.51769805e-01 -2.37830400e-01 -3.46327990e-01 3.32379609e-01
1.25884521e+00 -4.58132237e-01 5.96003592e-01 9.40213621e-01
3.10926914e-01 -8.81629586e-02 -1.02454567e+00 7.11138666e-01
-5.59362054e-01 -2.11635756e+00 -1.67952225e-01 5.41014552e-01
1.02341735e+00 1.94730818e-01 4.56203312e-01 3.36829811e-01
6.94887638e-01 -1.18738890e+00 -2.97790878e-02 5.83765805e-01
6.48689032e-01 -1.13809311e+00 7.26675034e-01 2.71129400e-01
-6.60221696e-01 -4.05810744e-01 -4.77516860e-01 -9.07575428e-01
-3.86290967e-01 1.09965456e+00 -9.92345035e-01 -7.75956586e-02
3.78154755e-01 4.29712236e-01 -2.17720389e-01 7.91875958e-01
1.53761148e-01 9.89993691e-01 -7.91766569e-02 -3.49198997e-01
2.70792693e-01 -1.39302343e-01 1.54410034e-01 8.83326232e-01
3.03817421e-01 3.36669758e-02 -1.18065290e-01 9.85273719e-01
-7.23694205e-01 -3.13394845e-01 -7.55952418e-01 -1.02513742e+00
6.01129472e-01 9.93193746e-01 -9.33122873e-01 -1.74620882e-01
-2.12796658e-01 6.97677732e-01 1.61164612e-01 3.39601994e-01
-5.23857594e-01 -5.17134011e-01 9.80448365e-01 6.07284121e-02
5.47653921e-02 -5.70883870e-01 -9.89021242e-01 -6.92057610e-01
-4.75587338e-01 -8.75310600e-01 -3.48902375e-01 -4.69186276e-01
-7.69685030e-01 -1.32765742e-02 -6.38373435e-01 -4.39450115e-01
-1.30024448e-01 -7.77098536e-01 -6.25824809e-01 7.34578013e-01
-1.17637634e+00 -7.57865846e-01 2.48299316e-02 -4.02759522e-01
3.76735479e-01 -7.97827721e-01 5.22267401e-01 6.67271838e-02
-7.63354540e-01 4.79774058e-01 5.61854303e-01 1.39369220e-02
3.31593543e-01 -8.20452273e-01 8.84262204e-01 6.90333605e-01
3.17332953e-01 6.69590950e-01 4.73385930e-01 -6.18177533e-01
-1.99494672e+00 -9.71151412e-01 6.01049960e-01 -3.67917389e-01
6.97971523e-01 -1.83721989e-01 -4.19676095e-01 5.69053352e-01
1.77125800e-02 -5.24490952e-01 4.19940472e-01 4.94617671e-01
-1.00835159e-01 -4.38255727e-01 -8.12746823e-01 5.28200328e-01
6.21371210e-01 -3.95441055e-01 -7.45305941e-02 3.67593408e-01
9.53169048e-01 4.06384677e-01 -1.03701580e+00 2.00240433e-01
6.54906154e-01 -5.55170655e-01 9.58764911e-01 -5.99371910e-01
9.19431686e-01 3.05151165e-01 2.99559772e-01 -1.23485446e+00
-6.86128139e-01 -6.58945143e-01 -1.98558107e-01 7.13827908e-01
6.62214100e-01 -4.61317062e-01 1.27382195e+00 7.17666030e-01
-3.57518792e-02 -1.08614504e+00 -5.60697556e-01 -8.20736289e-01
5.73076487e-01 -7.98724219e-02 4.80779618e-01 1.14215600e+00
1.41408503e-01 5.14292836e-01 -9.81684253e-02 -4.22601283e-01
3.91518980e-01 7.96263143e-02 8.63735735e-01 -1.72982621e+00
-3.35384309e-01 -8.85953128e-01 -6.02859318e-01 -7.70204246e-01
1.14213258e-01 -5.64255416e-01 -3.91678661e-01 -1.41682589e+00
1.86130375e-01 -5.46396494e-01 -4.27744538e-01 5.66609383e-01
-1.38952598e-01 1.43851936e-01 1.14030398e-01 4.39425148e-02
-2.77027488e-01 2.93727100e-01 1.04578233e+00 -3.63957018e-01
-3.56264621e-01 -1.28404215e-01 -1.09012783e+00 6.35256708e-01
8.92063022e-01 -5.42723201e-02 3.86213586e-02 -3.68410826e-01
5.23352385e-01 4.32937257e-02 -1.23430081e-01 -1.25691545e+00
-1.70988068e-02 -5.75008869e-01 6.31274998e-01 1.38521582e-01
-2.10568262e-03 -5.36415398e-01 6.60902113e-02 9.34449673e-01
-3.72106791e-01 7.10021779e-02 3.15055400e-01 1.80728614e-01
1.70453981e-01 2.19021931e-01 7.61966169e-01 -2.63436794e-01
-4.12881255e-01 2.77445227e-01 -6.59903109e-01 -2.62810975e-01
9.59908903e-01 -7.95568526e-02 -1.70393273e-01 -2.52179086e-01
-4.42444347e-02 6.03293963e-02 6.01732433e-01 3.47692698e-01
3.67177665e-01 -7.72785068e-01 -6.08543515e-01 -6.99382648e-02
-2.88747609e-01 -6.94132447e-02 -1.48703858e-01 3.87228876e-01
-7.84395814e-01 7.46035874e-01 -6.09928310e-01 -1.24831311e-02
-7.91618049e-01 5.66105306e-01 5.47605567e-02 -3.68193418e-01
-3.83830816e-01 1.25627232e+00 2.15466782e-01 6.45100176e-02
2.89716810e-01 -2.24911347e-01 2.60443658e-01 -2.40579948e-01
3.11278433e-01 7.35440969e-01 -1.81418419e-01 2.32791781e-01
-1.49675295e-01 1.30822048e-01 -2.74951100e-01 -6.88267266e-03
1.56732285e+00 6.03660762e-01 -6.49668649e-02 4.44522649e-01
7.20356703e-01 -3.34340602e-01 -1.12901330e+00 2.01197919e-02
1.95577919e-01 3.66389123e-03 6.67597592e-01 -9.79945183e-01
-1.05850387e+00 1.14610624e+00 6.28819823e-01 1.05000667e-01
9.04200613e-01 -2.56506413e-01 1.13780904e+00 1.02022028e+00
5.39223313e-01 -1.22759283e+00 3.93557936e-01 5.76286733e-01
4.90861475e-01 -1.15679955e+00 3.21744740e-01 4.03953701e-01
-3.21814895e-01 9.31599379e-01 6.20777488e-01 -5.71815446e-02
4.72755373e-01 2.98560679e-01 -4.49733406e-01 -2.53362060e-01
-7.37319469e-01 6.70285448e-02 -7.63425678e-02 4.52153772e-01
7.49003530e-01 3.60473305e-01 -1.09953389e-01 3.88533741e-01
7.57761374e-02 3.92739773e-01 3.77309680e-01 7.38833666e-01
-9.96259689e-01 -9.58278298e-01 -2.12463558e-01 1.00902414e+00
-3.70122075e-01 -1.23250194e-01 -6.89472437e-01 3.22394997e-01
2.59638608e-01 5.48701167e-01 4.33737576e-01 -5.47287881e-01
-2.81667262e-01 3.72139476e-02 5.10251939e-01 -4.44249570e-01
-4.64395911e-01 -4.72244173e-01 -8.11406225e-03 -1.04712315e-01
-6.93253651e-02 -4.30493504e-01 -1.33093953e+00 -8.24033499e-01
-3.13542426e-01 1.07593074e-01 1.00586116e+00 1.12152195e+00
9.06243980e-01 5.63604474e-01 7.22024977e-01 -8.18785369e-01
-1.10702537e-01 -8.66408229e-01 -2.37244010e-01 -1.87565058e-01
2.18355507e-02 -4.87456560e-01 1.52070716e-01 -1.18430845e-01]
|
[8.90942668914795, 6.343876361846924]
|
3cc64475-e3d1-4a26-8e84-196eb48dd9da
|
pareto-policy-pool-for-model-based-offline
| null | null |
https://openreview.net/forum?id=OqcZu8JIIzS
|
https://openreview.net/pdf?id=OqcZu8JIIzS
|
Pareto Policy Pool for Model-based Offline Reinforcement Learning
|
Online reinforcement learning (RL) can suffer from poor exploration, sparse reward, insufficient data, and overhead caused by inefficient interactions between an immature policy and a complicated environment. Model-based offline RL instead trains an environment model using a dataset of pre-collected experiences so online RL methods can learn in an offline manner by solely interacting with the model. However, the uncertainty and accuracy of the environment model can drastically vary across different state-action pairs so the RL agent may achieve high model return but perform poorly in the true environment. Unlike previous works that need to carefully tune the trade-off between the model return and uncertainty in a single objective, we study a bi-objective formulation for model-based offline RL that aims at producing a pool of diverse policies on the Pareto front performing different levels of trade-offs, which provides the flexibility to select the best policy for each realistic environment from the pool. Our method, ``Pareto policy pool (P3)'', does not need to tune the trade-off weight but can produce policies allocated at different regions of the Pareto front. For this purpose, we develop an efficient algorithm that solves multiple bi-objective optimization problems with distinct constraints defined by reference vectors targeting diverse regions of the Pareto front. We theoretically prove that our algorithm can converge to the targeted regions. In order to obtain more Pareto optimal policies without linearly increasing the cost, we leverage the achieved policies as initialization to find more Pareto optimal policies in their neighborhoods. On the D4RL benchmark for offline RL, P3 substantially outperforms several recent baseline methods over multiple tasks, especially when the quality of pre-collected experiences is low.
|
['Yuhui Shi', 'Jie Ma', 'Tianyi Zhou', 'Jing Jiang', 'Yijun Yang']
|
2021-09-29
| null | null | null |
iclr-2022-4
|
['d4rl']
|
['robots']
|
[-2.26849258e-01 -2.42119223e-01 -3.26960623e-01 3.46909091e-02
-9.79849219e-01 -9.23182547e-01 2.95491844e-01 1.93104178e-01
-8.04005086e-01 9.53251123e-01 8.30569863e-03 -1.70258433e-01
-2.98991948e-01 -6.05219901e-01 -9.52147424e-01 -8.41088176e-01
-3.28761935e-01 6.94242597e-01 3.36312805e-03 -1.60979047e-01
2.35334218e-01 2.61836231e-01 -1.54480958e+00 -2.61212164e-03
1.23133802e+00 9.61451471e-01 6.90881908e-01 5.99616587e-01
7.35093206e-02 5.60688853e-01 -7.92706728e-01 1.44229814e-01
5.61719835e-01 -4.35318768e-01 -3.90027314e-01 -2.11142242e-01
-2.31569916e-01 -5.02526581e-01 7.30471760e-02 1.12361324e+00
7.23465860e-01 4.98267978e-01 2.81654298e-01 -1.27688813e+00
-1.06793180e-01 7.03238189e-01 -4.22528893e-01 -1.07745141e-01
1.06996834e-01 6.45931304e-01 8.51745069e-01 -2.90291965e-01
5.89302897e-01 1.26982498e+00 1.83118418e-01 7.10559607e-01
-1.44890714e+00 -7.01611042e-01 7.45699823e-01 1.78415682e-02
-1.11750650e+00 -1.96936965e-01 5.48236251e-01 -9.53980088e-02
7.13714898e-01 3.08679305e-02 8.46670866e-01 1.31685257e+00
2.72239119e-01 8.89123619e-01 1.40838671e+00 -1.47557631e-01
1.03982282e+00 3.80096287e-02 -4.06481445e-01 5.73392153e-01
1.78877488e-01 6.35424614e-01 -6.05450988e-01 -2.86592096e-01
6.71957254e-01 -1.43396974e-01 -2.75889903e-01 -6.82949424e-01
-1.06036067e+00 6.96381688e-01 2.96443343e-01 1.17229894e-02
-5.97270608e-01 3.64364296e-01 2.67095864e-01 5.21066904e-01
-2.07474846e-02 1.10300732e+00 -6.32045686e-01 -5.39165556e-01
-5.86413264e-01 4.99879837e-01 7.68172622e-01 6.46169007e-01
7.45389521e-01 1.09993003e-01 -5.66925049e-01 6.11013949e-01
1.02703914e-01 5.49325228e-01 3.62247556e-01 -1.29279423e+00
7.54874349e-01 3.07440400e-01 8.69600534e-01 -6.29517615e-01
-2.08055168e-01 -7.74021626e-01 -1.97494954e-01 5.95844030e-01
5.25621951e-01 -6.94568574e-01 -8.66543949e-01 2.13191438e+00
5.34382641e-01 6.41720816e-02 2.03712955e-01 1.02448070e+00
-9.00242999e-02 6.29552364e-01 -1.36248440e-01 -3.19542080e-01
7.39424229e-01 -1.07380211e+00 -5.92306316e-01 -4.45408881e-01
4.15446728e-01 -2.12209061e-01 1.33991694e+00 4.47747141e-01
-1.11539662e+00 -1.85963914e-01 -9.87484157e-01 6.50917053e-01
-1.01656914e-01 2.06226427e-02 3.82041037e-01 3.18051755e-01
-1.00765157e+00 7.92987585e-01 -8.85934114e-01 7.61624426e-02
3.97981644e-01 4.92668927e-01 1.59397826e-01 1.09043941e-02
-9.58317697e-01 9.65732753e-01 5.54872096e-01 -3.75819989e-02
-1.66116810e+00 -6.63005829e-01 -4.73579705e-01 1.80995494e-01
1.08145595e+00 -4.41343367e-01 1.42570412e+00 -1.22635567e+00
-2.08245897e+00 -5.43858185e-02 1.39235780e-01 -4.65478450e-01
7.83611655e-01 -3.05492133e-01 2.24291552e-02 -2.05002114e-01
-1.85980216e-01 5.87882280e-01 9.67548013e-01 -1.64420736e+00
-8.22261930e-01 -7.00433776e-02 5.25719464e-01 6.55589998e-01
-8.36736709e-02 -2.74732411e-01 -3.82604361e-01 -3.14966053e-01
-4.62448061e-01 -9.88892555e-01 -6.13712311e-01 -4.07755196e-01
-1.10251687e-01 1.22152187e-01 2.92178571e-01 -2.43920833e-01
1.10957479e+00 -1.83949792e+00 3.72620940e-01 3.65971446e-01
-1.31247818e-01 1.42853245e-01 -5.02649307e-01 4.19842839e-01
5.86673439e-01 6.59826249e-02 -5.65714054e-02 -3.82587790e-01
3.33164304e-01 4.88320500e-01 -4.11652923e-01 3.00107867e-01
-1.74771100e-01 7.41508722e-01 -1.42786610e+00 -1.77564859e-01
-3.88298370e-02 4.23740149e-02 -7.22867668e-01 4.27060425e-01
-8.59251916e-01 7.13721454e-01 -7.72711098e-01 4.18746054e-01
3.28572661e-01 3.53369378e-02 4.77062285e-01 3.96365136e-01
-2.31289268e-01 2.66463496e-02 -1.31718791e+00 1.58992112e+00
-8.17600250e-01 5.16611971e-02 4.11646038e-01 -8.90987873e-01
8.65068913e-01 3.02508045e-02 5.64058125e-01 -8.97086561e-01
1.80948958e-01 2.77148545e-01 1.41209453e-01 -2.28460476e-01
3.49129409e-01 3.46015468e-02 -1.77301928e-01 6.68353975e-01
-1.73717216e-02 9.42849461e-03 1.46853551e-01 -2.46903226e-01
1.18137479e+00 4.74642128e-01 1.04999498e-01 -1.85918257e-01
1.71412274e-01 -3.36437300e-02 8.99926364e-01 1.31194305e+00
-3.28352243e-01 8.05969611e-02 7.37994790e-01 -2.79176533e-01
-8.79447401e-01 -9.41297650e-01 3.64581436e-01 1.29085326e+00
4.25078601e-01 -4.65642065e-02 -4.89203513e-01 -9.15468335e-01
1.35257110e-01 9.89565194e-01 -6.44242406e-01 -1.89329222e-01
-5.02813697e-01 -4.70472515e-01 1.60668492e-01 3.04425865e-01
3.66576612e-01 -1.15311277e+00 -1.19090235e+00 3.33357662e-01
-8.13671295e-03 -7.07565129e-01 -6.32193267e-01 4.92827624e-01
-7.32505798e-01 -8.44852328e-01 -5.68102062e-01 -3.06717634e-01
8.00428092e-01 -1.02459058e-01 1.06390691e+00 -2.47167319e-01
1.68889537e-01 5.02547145e-01 -1.60725415e-01 -4.83713120e-01
-3.83874387e-01 -5.02302591e-03 2.14114383e-01 -1.43582806e-01
-2.28022888e-01 -4.53613371e-01 -6.12882435e-01 4.58576977e-01
-6.80312037e-01 -4.11312021e-02 5.46972215e-01 9.59559858e-01
7.76852369e-01 2.40902424e-01 7.40293264e-01 -4.47210938e-01
8.96336436e-01 -5.93481481e-01 -1.16679192e+00 5.14829040e-01
-6.42699480e-01 6.74492180e-01 1.11102450e+00 -8.63192916e-01
-1.00439227e+00 -9.03817862e-02 3.75593662e-01 -6.80344760e-01
2.51204133e-01 2.54795700e-01 -1.68356776e-01 1.03222504e-01
5.97183526e-01 1.67582810e-01 1.32130399e-01 -3.50254089e-01
3.62951458e-01 1.20482236e-01 2.38607153e-01 -1.28983843e+00
6.71877325e-01 2.01565549e-01 -1.70085266e-01 4.92046960e-02
-8.72428060e-01 -7.66997263e-02 -1.08117703e-02 -4.37010795e-01
3.29646766e-01 -7.30376005e-01 -9.62094784e-01 1.78101540e-01
-7.29853570e-01 -9.99451637e-01 -5.87526381e-01 4.15801436e-01
-9.01928544e-01 -2.62191772e-01 -1.48065807e-02 -1.19975603e+00
-1.74352497e-01 -1.33719039e+00 7.04972029e-01 4.15345430e-01
1.26928434e-01 -8.13758016e-01 3.76941621e-01 -1.49963498e-01
5.54610670e-01 3.69713336e-01 7.12673128e-01 -3.86244148e-01
-4.86208826e-01 3.31774712e-01 3.22506309e-01 9.99426320e-02
5.54795079e-02 -2.90688396e-01 -5.98347425e-01 -7.86791027e-01
-1.25319794e-01 -7.49094009e-01 6.39308572e-01 3.03722709e-01
1.22697079e+00 -8.00812602e-01 -2.21496165e-01 5.82086682e-01
1.59473801e+00 6.54330194e-01 2.35232651e-01 7.39287376e-01
1.52830988e-01 4.16104496e-01 1.05615425e+00 8.00612271e-01
3.05777460e-01 6.26183867e-01 7.08047271e-01 3.12505066e-01
4.46478754e-01 -6.00398958e-01 8.42415750e-01 1.53138295e-01
7.52819031e-02 -2.08938971e-01 -6.81123793e-01 5.24595857e-01
-2.27794743e+00 -8.79264653e-01 9.78428245e-01 2.75806046e+00
1.03010368e+00 1.99914783e-01 4.11233127e-01 -4.63470191e-01
4.98177201e-01 2.50749476e-02 -1.39084566e+00 -5.27705371e-01
2.61001945e-01 1.71053723e-01 6.26062810e-01 5.36360502e-01
-7.28209257e-01 8.77421081e-01 6.18875170e+00 7.81032741e-01
-1.34328616e+00 -8.34647343e-02 5.86028337e-01 -7.45305479e-01
-4.24146950e-01 4.86050509e-02 -6.85594440e-01 6.65529728e-01
9.09479678e-01 -3.59391779e-01 1.21366441e+00 8.91300738e-01
5.56184590e-01 -2.23019093e-01 -1.07678246e+00 7.11638451e-01
-5.31217933e-01 -1.11286640e+00 -3.93963486e-01 6.01319894e-02
1.07780206e+00 1.63588509e-01 2.66274035e-01 7.12648869e-01
1.01336956e+00 -1.03349721e+00 1.01245892e+00 5.66239595e-01
4.98450279e-01 -1.10812128e+00 5.02281964e-01 7.67588794e-01
-8.28827381e-01 -7.37064123e-01 -2.85075963e-01 1.09122120e-01
-6.10998683e-02 9.23493132e-02 -7.37250686e-01 3.70365709e-01
5.58850944e-01 1.55605346e-01 -8.48744437e-02 9.87465560e-01
-1.54731259e-01 3.84482741e-01 -4.19970393e-01 -3.21646184e-01
6.39158309e-01 -4.16529268e-01 5.71111917e-01 5.85274756e-01
2.94709653e-01 -1.81572065e-01 6.47801340e-01 1.00399721e+00
2.17099875e-01 5.62419696e-03 -4.62566853e-01 -1.86332524e-01
7.68213332e-01 1.03916681e+00 -4.86411065e-01 -8.46443139e-03
1.84338003e-01 5.31410217e-01 7.35408127e-01 6.05287373e-01
-1.03401601e+00 1.58412848e-02 8.09043050e-01 -4.52890635e-01
4.15963113e-01 -3.02637458e-01 -4.00407687e-02 -9.58124697e-01
-1.13245353e-01 -1.24387085e+00 2.11768225e-01 -2.43774042e-01
-1.00487673e+00 3.74835849e-01 9.06621143e-02 -1.11070049e+00
-4.00094628e-01 -2.71619350e-01 -5.68732560e-01 6.20575428e-01
-1.46211243e+00 -4.40019697e-01 4.48745526e-02 3.74291331e-01
4.84035611e-01 -2.43535072e-01 5.03147781e-01 -2.43934512e-01
-7.52072453e-01 5.70034385e-01 6.63538694e-01 -5.69444537e-01
5.76490164e-01 -1.34594619e+00 -1.12266004e-01 5.41347682e-01
-1.95454836e-01 4.50590372e-01 9.54403043e-01 -5.46979606e-01
-1.67732918e+00 -1.00098383e+00 -1.38254479e-01 -1.16453625e-01
6.51338041e-01 -3.16381395e-01 -5.25072575e-01 2.89495587e-01
-1.30000412e-01 -3.83888744e-02 2.26084903e-01 -1.46122091e-02
7.80550018e-02 -3.37418497e-01 -1.28640640e+00 9.15753663e-01
1.02748346e+00 -4.73668426e-03 -4.16566767e-02 1.06789343e-01
7.27975428e-01 -6.02432549e-01 -6.66291833e-01 2.05774814e-01
5.56658328e-01 -7.80827045e-01 7.39886820e-01 -7.32140183e-01
1.25614300e-01 -3.44101429e-01 -7.18962029e-02 -1.87249398e+00
-5.05900569e-02 -1.11130130e+00 -4.46026474e-01 9.39413488e-01
3.70128512e-01 -8.52405190e-01 5.31077266e-01 5.22531688e-01
5.84925115e-02 -1.28266454e+00 -9.23212647e-01 -1.05556905e+00
1.36955708e-01 -2.81758875e-01 9.20608401e-01 3.40839714e-01
-2.04057872e-01 -8.05206038e-03 -4.04700428e-01 1.88018084e-01
7.00137138e-01 4.19876158e-01 7.55896688e-01 -6.43816352e-01
-7.73693979e-01 -5.01017094e-01 5.25858402e-01 -1.11209846e+00
3.55556399e-01 -4.14204806e-01 5.24416029e-01 -1.36491191e+00
2.49966960e-02 -1.12400651e+00 -5.80429077e-01 4.90998298e-01
-1.20493509e-01 -7.22473681e-01 4.69956428e-01 1.18926316e-01
-9.61973250e-01 9.59447145e-01 1.51516604e+00 -3.43028866e-02
-7.27328897e-01 3.62303406e-02 -7.18280435e-01 4.91231799e-01
9.31078792e-01 -4.71253097e-01 -7.63781786e-01 -5.30060649e-01
3.21233541e-01 3.89632612e-01 -9.63426009e-02 -9.56630647e-01
7.89815411e-02 -8.68709385e-01 1.31706536e-01 -2.94512302e-01
3.01350266e-01 -7.91507900e-01 1.30382434e-01 5.71789145e-01
-6.05724096e-01 4.97611091e-02 1.61735252e-01 9.10482764e-01
2.40251601e-01 -3.04405600e-01 7.10748434e-01 -3.68024290e-01
-4.86405194e-01 3.43969226e-01 -3.75911772e-01 3.27033103e-01
1.17656326e+00 9.68267918e-02 -2.43163154e-01 -3.68664414e-01
-3.76060694e-01 1.04555988e+00 7.13626266e-01 4.29746598e-01
4.06547844e-01 -1.06490588e+00 -3.41185808e-01 -1.28920808e-01
-1.56607568e-01 4.92141731e-02 7.22609693e-03 5.32937348e-01
-6.74425364e-02 1.38004169e-01 -3.17205399e-01 -3.11504006e-01
-6.61903143e-01 6.01291478e-01 6.85797572e-01 -7.70960093e-01
-4.33881313e-01 5.27559757e-01 1.03759225e-02 -4.62606966e-01
5.04517853e-01 -3.42254490e-01 -1.50503814e-01 -7.46073127e-02
4.45044070e-01 4.17576879e-01 -3.16555649e-01 -9.62453708e-03
-2.49389440e-01 1.90081090e-01 1.28200978e-01 -5.03556013e-01
1.36419058e+00 3.68543081e-02 3.91106218e-01 3.91731620e-01
7.56000996e-01 -4.05278429e-02 -2.11791420e+00 -3.00029628e-02
-7.40079582e-02 -6.06016695e-01 2.54956752e-01 -1.16805112e+00
-1.00221717e+00 1.99716121e-01 6.02126122e-01 -6.55951276e-02
1.09185171e+00 -3.96130770e-01 5.94089150e-01 5.05855381e-01
9.78714943e-01 -1.65302646e+00 3.40339214e-01 6.09552801e-01
8.03443968e-01 -1.09584439e+00 -1.17866851e-01 4.90210027e-01
-1.00527394e+00 8.88001621e-01 8.29587817e-01 -1.95633695e-01
1.34699330e-01 3.07125449e-01 -2.25050617e-02 1.98488787e-01
-1.28931332e+00 -3.48872542e-01 -5.54361530e-02 5.72538793e-01
-3.65546376e-01 1.73759535e-01 -1.57382354e-01 4.89829749e-01
-1.60362553e-02 -7.14985058e-02 2.84235328e-01 1.17738783e+00
-6.03995144e-01 -1.32756019e+00 -4.72826600e-01 4.05441761e-01
-1.76545352e-01 3.33046257e-01 -9.86980796e-02 5.33047616e-01
3.58769596e-02 8.44380200e-01 -1.64667349e-02 -1.76226705e-01
1.27578720e-01 -2.10242972e-01 5.99063933e-01 -4.00587648e-01
-7.79616177e-01 1.73613653e-02 1.56482488e-01 -1.06936753e+00
5.34318499e-02 -6.39202476e-01 -1.34088445e+00 -2.67923400e-02
-1.10436365e-01 3.17197800e-01 5.69658458e-01 9.62365806e-01
5.91172814e-01 5.01143217e-01 9.94831562e-01 -8.75164211e-01
-1.29396486e+00 -4.56699371e-01 -3.77431542e-01 7.52331018e-02
4.94253516e-01 -1.07889676e+00 -2.35364646e-01 -6.70297205e-01]
|
[4.128280162811279, 2.1908161640167236]
|
f79248f7-5d87-48ea-922d-424cff0c8f93
|
updated-version-a-video-anomaly-detection
|
2303.05109
| null |
https://arxiv.org/abs/2303.05109v1
|
https://arxiv.org/pdf/2303.05109v1.pdf
|
Updated version: A Video Anomaly Detection Framework based on Appearance-Motion Semantics Representation Consistency
|
Video anomaly detection is an essential but challenging task. The prevalent methods mainly investigate the reconstruction difference between normal and abnormal patterns but ignore the semantics consistency between appearance and motion information of behavior patterns, making the results highly dependent on the local context of frame sequences and lacking the understanding of behavior semantics. To address this issue, we propose a framework of Appearance-Motion Semantics Representation Consistency that uses the gap of appearance and motion semantic representation consistency between normal and abnormal data. The two-stream structure is designed to encode the appearance and motion information representation of normal samples, and a novel consistency loss is proposed to enhance the consistency of feature semantics so that anomalies with low consistency can be identified. Moreover, the lower consistency features of anomalies can be used to deteriorate the quality of the predicted frame, which makes anomalies easier to spot. Experimental results demonstrate the effectiveness of the proposed method.
|
['Zhiqiang Wu', 'Caidan Zhao', 'Xiangyu Huang']
|
2023-03-09
| null | null | null | null |
['video-anomaly-detection']
|
['computer-vision']
|
[ 7.61755109e-02 -4.97509539e-01 -8.39104652e-02 -4.35115278e-01
3.79049569e-01 -1.66652799e-01 3.65218163e-01 9.59832743e-02
4.61514182e-02 2.09133383e-02 1.52992144e-01 2.08377942e-01
-7.68464431e-03 -6.65220022e-01 -2.81141669e-01 -6.84935212e-01
-1.83321834e-02 -5.09172380e-01 8.27413380e-01 -1.99917585e-01
2.93322116e-01 1.83283851e-01 -1.86276472e+00 5.02163768e-01
6.82077110e-01 1.35882998e+00 1.11105546e-01 1.57019913e-01
-5.87467849e-01 1.00591612e+00 -4.21161979e-01 1.89562201e-01
1.39803678e-01 -6.96402371e-01 -2.01937199e-01 5.59103966e-01
2.93646485e-01 -5.55578053e-01 -5.31276882e-01 1.43308866e+00
8.44200850e-02 7.30536133e-02 1.23342209e-01 -1.67008841e+00
-3.56661618e-01 -2.32933879e-01 -4.71704036e-01 8.03959489e-01
4.34310168e-01 1.99385539e-01 5.71922541e-01 -5.72836995e-01
4.50639039e-01 1.46552205e+00 4.40655231e-01 5.30940473e-01
-5.22664666e-01 -5.03429890e-01 8.04678619e-01 9.86892641e-01
-1.14713335e+00 -4.46647614e-01 1.00123203e+00 -4.58074749e-01
5.01671672e-01 3.70823324e-01 9.00283754e-01 1.02976978e+00
6.30365461e-02 9.07098055e-01 5.28078377e-01 -1.81051493e-02
1.56177610e-01 -2.20312133e-01 2.42942244e-01 7.21358955e-01
4.16642904e-01 -1.63300619e-01 -5.32882154e-01 -1.11160353e-01
5.81686080e-01 5.91227591e-01 -4.97046322e-01 -3.63975257e-01
-1.04451108e+00 2.88631380e-01 -2.63559502e-02 4.54091698e-01
-2.07184523e-01 -1.53057694e-01 8.50940168e-01 4.39449430e-01
4.39984500e-01 -3.71957064e-01 -1.93418160e-01 -3.41789365e-01
-5.00569344e-01 9.02013555e-02 9.76816863e-02 9.57242608e-01
5.66060841e-01 3.64371270e-01 -1.89959884e-01 5.81087649e-01
5.67765236e-01 5.12944937e-01 7.21584678e-01 -7.80104458e-01
3.24262619e-01 1.07446706e+00 4.78843646e-03 -1.78099954e+00
-1.86149493e-01 1.11420810e-01 -7.82532513e-01 -2.68724305e-03
2.00133070e-01 4.07476515e-01 -7.23536074e-01 1.54015446e+00
5.20671189e-01 8.29537630e-01 -1.29603595e-01 1.01217103e+00
7.18424976e-01 6.87086523e-01 6.47484288e-02 -5.69563270e-01
1.18199694e+00 -6.88797116e-01 -1.20667171e+00 -7.28203729e-02
6.54075325e-01 -7.15712905e-01 9.91552472e-01 1.37474716e-01
-9.90092099e-01 -8.46596777e-01 -1.09462035e+00 3.38631064e-01
-1.39826417e-01 -2.67920524e-01 2.76103437e-01 3.76492679e-01
-6.46134853e-01 3.04658532e-01 -9.15254176e-01 -2.73589551e-01
2.06663832e-01 -1.57214120e-01 -3.07418913e-01 -3.02730680e-01
-1.14957118e+00 4.51157451e-01 5.84317923e-01 3.82778674e-01
-3.94706607e-01 -3.76926303e-01 -9.70894933e-01 -4.97339815e-02
3.52658033e-01 -2.44959220e-01 7.98251271e-01 -1.52421284e+00
-9.09952879e-01 3.36316109e-01 -4.10070121e-01 -7.04216659e-02
5.10960579e-01 -1.49581045e-01 -1.13150740e+00 2.61877775e-01
1.20871112e-01 -1.01972334e-01 1.00749516e+00 -1.00792468e+00
-1.16922390e+00 -3.92613977e-01 -2.38723025e-01 1.85738757e-01
-4.52782214e-01 -2.87014134e-02 -6.26858473e-01 -8.75412941e-01
7.35130250e-01 -4.86064404e-01 1.93213329e-01 3.33390594e-01
5.67733943e-02 -4.38211374e-02 1.66604066e+00 -9.01686370e-01
1.60843158e+00 -2.58503556e+00 -7.24061653e-02 5.33415616e-01
1.84700638e-02 2.18083248e-01 -9.99668837e-02 5.59763834e-02
-1.50326565e-01 -1.28465220e-01 -1.99367702e-01 2.43784979e-01
-4.09662157e-01 4.36064869e-01 -3.88285905e-01 5.47179818e-01
2.09425166e-01 2.73899078e-01 -9.89180088e-01 -5.99946260e-01
3.71789396e-01 3.79221775e-02 -4.55396235e-01 4.73057657e-01
-8.87541994e-02 6.69719815e-01 -8.27360392e-01 8.42624187e-01
8.12683105e-01 -6.96635805e-03 -9.53887999e-02 -4.74735022e-01
1.23864256e-01 -2.25067273e-01 -1.56257415e+00 1.57371938e+00
4.18133475e-02 3.25354725e-01 -1.84138268e-01 -1.04901850e+00
8.21731150e-01 2.62277991e-01 8.08531821e-01 -1.01587486e+00
-1.59497008e-01 -3.45413908e-02 -3.76329459e-02 -1.16780281e+00
1.43208504e-01 3.74575406e-01 4.13870126e-01 7.23203644e-02
-4.93977934e-01 5.64244330e-01 1.49259672e-01 -2.01973673e-02
1.06563783e+00 3.42918575e-01 1.26987368e-01 3.52177359e-02
9.75511909e-01 -1.72602460e-01 1.21769476e+00 2.20354021e-01
-4.14028645e-01 5.50268829e-01 3.80108356e-01 -8.38460386e-01
-9.01599705e-01 -9.74662364e-01 4.55123484e-02 6.07142031e-01
8.91527295e-01 -5.20871580e-01 -4.97266561e-01 -8.49666715e-01
-1.26447499e-01 5.45067251e-01 -4.32236075e-01 -7.38717377e-01
-7.44268358e-01 -6.61300302e-01 8.77710804e-02 3.04341555e-01
8.67464125e-01 -8.58193099e-01 -5.00255644e-01 4.52492610e-02
-5.40758967e-01 -1.14634824e+00 -4.27399665e-01 -9.13625360e-01
-9.44759548e-01 -1.24858046e+00 -1.95803031e-01 -5.97817302e-01
9.22535956e-01 6.01057172e-01 7.03005850e-01 7.74821818e-01
-2.50994653e-01 5.47633171e-01 -7.56723583e-01 -4.38630432e-02
-4.09895360e-01 -8.52696598e-01 1.17527053e-01 6.59594655e-01
5.18189788e-01 -4.12058979e-01 -7.46971369e-01 4.22107011e-01
-1.28439438e+00 1.00971321e-02 2.48608276e-01 7.15986907e-01
4.78372663e-01 1.76212296e-01 3.68882298e-01 -4.53915715e-01
2.33667985e-01 -6.62716925e-01 -2.97859699e-01 3.40443045e-01
-6.54297471e-01 -1.41538724e-01 5.59758782e-01 -3.93705279e-01
-1.08315456e+00 -2.90850103e-01 1.19104087e-01 -7.03572989e-01
-3.53295267e-01 8.01967084e-02 -4.97735053e-01 1.59120411e-01
1.63710155e-02 6.99083090e-01 1.60759598e-01 -4.26722318e-01
-1.18945792e-01 4.00562674e-01 3.72111797e-01 -3.85889828e-01
4.63202864e-01 6.90080345e-01 1.33389041e-01 -7.92213202e-01
-4.44135904e-01 -5.83832502e-01 -3.40045214e-01 -5.38575292e-01
8.10510278e-01 -7.49263108e-01 -2.49050424e-01 5.71131647e-01
-1.16090584e+00 4.30205286e-01 -2.66004086e-01 4.20120507e-01
-3.92864078e-01 1.23574913e+00 -4.89666611e-01 -7.37588525e-01
1.29606560e-01 -1.03725064e+00 9.11236882e-01 1.14608347e-01
-2.28597857e-02 -8.29688251e-01 -1.61223829e-01 -1.48341209e-01
2.07880646e-01 3.19689393e-01 9.59772706e-01 -4.43682462e-01
-6.16947830e-01 -1.78959981e-01 -2.57826865e-01 3.86665732e-01
4.81338143e-01 2.31496006e-01 -6.77253425e-01 -1.26765266e-01
4.61536080e-01 3.20134848e-01 6.13899946e-01 2.22372308e-01
1.58320820e+00 -3.50088775e-01 -3.07292819e-01 5.68723619e-01
1.20306182e+00 4.92006153e-01 8.90958726e-01 5.12987792e-01
8.61418068e-01 7.34851003e-01 1.05950606e+00 6.56873763e-01
2.92121112e-01 6.39880478e-01 5.39013922e-01 1.08648777e-01
-5.46586839e-03 -7.59266764e-02 5.93300343e-01 9.84428287e-01
-2.09291995e-01 3.33037488e-02 -6.00475132e-01 5.02530634e-01
-2.34215021e+00 -1.35572362e+00 -3.11696917e-01 2.01609659e+00
3.88989896e-01 8.30871239e-02 -6.37419522e-02 3.31347615e-01
9.37828898e-01 3.59484553e-01 -3.96139592e-01 -1.19635552e-01
-9.00502279e-02 -6.72415555e-01 -6.20323904e-02 5.04431687e-03
-9.06762421e-01 3.96878362e-01 5.33096457e+00 9.89181936e-01
-1.08764029e+00 5.07793762e-02 5.42448461e-01 1.08056419e-01
-3.37192327e-01 7.21352398e-02 -3.48189682e-01 1.17740393e+00
3.69700402e-01 3.50613296e-02 6.00421019e-02 7.69841909e-01
5.89553416e-01 -9.59572345e-02 -1.03924930e+00 1.23697734e+00
7.91366920e-02 -8.93860400e-01 4.50271696e-01 -4.24523294e-01
3.33295524e-01 -6.52876139e-01 -1.05483413e-01 -6.63356707e-02
-4.92626131e-01 -5.57609200e-01 7.56702423e-01 9.67632055e-01
3.00122619e-01 -5.30197859e-01 7.71389902e-01 2.00318888e-01
-1.54476368e+00 -2.42178276e-01 -4.85866398e-01 4.28837277e-02
2.27751404e-01 5.49091876e-01 -1.49639651e-01 7.94651985e-01
1.01983964e+00 1.31213760e+00 -6.87483370e-01 9.72989500e-01
1.89335689e-01 5.03326893e-01 -1.66021228e-01 4.91759270e-01
1.48647070e-01 -4.41116810e-01 8.16634595e-01 9.88358319e-01
6.02751434e-01 -3.03024575e-02 5.73207796e-01 5.76723814e-01
6.25517368e-01 2.43306711e-01 -8.38948846e-01 3.12694132e-01
3.29946607e-01 8.71124327e-01 -6.43343091e-01 -4.17354256e-01
-9.27598536e-01 1.12126172e+00 -1.74028158e-01 4.19385314e-01
-9.06397104e-01 5.90109229e-02 7.54795134e-01 1.56428799e-01
1.16069570e-01 -1.50001019e-01 -1.16931386e-01 -1.38165534e+00
6.13434017e-01 -7.87436187e-01 7.57507741e-01 -5.29987991e-01
-1.41335344e+00 2.67665327e-01 1.20763108e-01 -1.86544883e+00
-1.80754796e-01 -2.08896071e-01 -9.44792807e-01 3.93011451e-01
-1.44343841e+00 -9.72879291e-01 -7.08437085e-01 8.78524959e-01
7.90286362e-01 -2.90917158e-01 3.93387467e-01 3.91412586e-01
-8.12008262e-01 3.01908553e-01 -2.04674993e-02 7.67187104e-02
4.59156007e-01 -6.70581341e-01 7.08778482e-03 1.33985949e+00
-2.43000954e-01 2.99563378e-01 5.97677410e-01 -8.73521388e-01
-1.09613955e+00 -1.26709580e+00 3.39629024e-01 -9.79088992e-02
4.32656616e-01 2.01404899e-01 -1.39769673e+00 3.50260913e-01
-1.92358166e-01 4.16553557e-01 4.17009771e-01 -4.15135294e-01
-2.86418557e-01 -3.18363518e-01 -1.18143880e+00 8.45366895e-01
1.32955575e+00 -2.30749011e-01 -6.27716601e-01 -7.16883391e-02
6.35098696e-01 -1.04590960e-01 -5.06448388e-01 8.30754995e-01
4.47171867e-01 -1.15512705e+00 9.04698610e-01 -6.93993390e-01
6.12978525e-02 -8.51950407e-01 -3.59027892e-01 -7.65515566e-01
-2.95017034e-01 -3.88029814e-02 -5.61054230e-01 1.28480291e+00
-3.20291549e-01 -4.62951362e-01 4.78033125e-01 5.19017041e-01
-2.28035241e-01 -5.54009914e-01 -9.46669042e-01 -7.96079397e-01
-8.25660765e-01 -4.12180811e-01 6.66006565e-01 1.10942924e+00
-6.63102418e-02 -3.09423357e-01 -3.13744724e-01 4.32261884e-01
4.04067189e-01 4.18828055e-02 5.35608172e-01 -9.51407552e-01
5.55049106e-02 -3.78975183e-01 -1.14916432e+00 -7.52141893e-01
1.28562912e-01 -4.73136157e-01 -1.16463974e-01 -1.12551451e+00
7.61036277e-02 -2.39444539e-01 -6.23416722e-01 2.77874887e-01
-3.92776221e-01 -4.93940711e-02 -6.89647347e-02 4.73334789e-01
-8.48587751e-01 7.42633283e-01 1.09480929e+00 -1.69361338e-01
-2.25284416e-02 -1.56940401e-01 -9.26025584e-02 1.11069214e+00
6.37337506e-01 -1.44513026e-01 -8.60210359e-01 -3.92795891e-01
-1.34488702e-01 -1.67256236e-01 4.49824780e-01 -1.21074736e+00
1.60117909e-01 -5.80286622e-01 5.92703164e-01 -4.35717285e-01
2.37013921e-02 -1.27107453e+00 8.81623924e-02 5.34955859e-01
-3.07084732e-02 5.65174878e-01 1.07510619e-01 1.07509065e+00
-5.43655455e-01 -8.97528678e-02 5.59771836e-01 -1.39033556e-01
-1.54299271e+00 6.06787801e-01 -2.46855095e-01 8.53982717e-02
1.32197964e+00 -5.92739105e-01 -1.05280384e-01 -3.26956421e-01
-6.84612870e-01 2.17380971e-01 6.41291797e-01 8.09469759e-01
1.03775966e+00 -1.72980511e+00 -4.99583513e-01 7.27974117e-01
4.11353052e-01 -2.84392107e-02 7.55913436e-01 8.37026000e-01
-6.01917863e-01 -2.06773221e-01 -3.65849078e-01 -9.28975701e-01
-1.23173320e+00 6.55754924e-01 2.65235335e-01 2.38482475e-01
-9.59434569e-01 1.30671173e-01 4.13708806e-01 4.18760270e-01
1.45080850e-01 -3.57495695e-01 -4.89421993e-01 -3.25253934e-01
8.59471142e-01 5.82903206e-01 -1.21496208e-01 -9.08032894e-01
-3.74440551e-01 5.98966539e-01 -2.23937798e-02 2.84747869e-01
8.10524642e-01 -6.73143268e-01 -3.29888672e-01 5.92101991e-01
8.75983596e-01 -1.48853391e-01 -1.24916053e+00 -2.12973863e-01
2.14261144e-01 -1.09841394e+00 -1.44662663e-01 -7.04154447e-02
-1.16262770e+00 6.67120516e-01 1.06937337e+00 3.55035454e-01
1.39437747e+00 -3.55607480e-01 1.03260529e+00 9.09953266e-02
1.92284495e-01 -1.15192103e+00 1.91886023e-01 1.82338938e-01
6.65939748e-01 -1.20781994e+00 -1.70522809e-01 -5.57070076e-01
-6.28272712e-01 1.24309051e+00 1.13032734e+00 -2.04793066e-02
6.62162662e-01 -1.10986091e-01 2.43330076e-02 -1.79215893e-01
-5.48495591e-01 -1.18992865e-01 4.93015796e-01 6.55263662e-01
1.16239250e-01 -4.93979752e-01 -4.01691049e-01 5.83378792e-01
6.71028197e-01 -2.67773986e-01 2.69008100e-01 1.09027350e+00
-5.80707908e-01 -9.54214990e-01 -3.68480712e-01 3.23020220e-01
-4.52100545e-01 2.87063628e-01 9.46620330e-02 4.10876900e-01
5.02459407e-01 8.96733284e-01 4.33534086e-01 -3.79198343e-01
3.58260185e-01 1.33533001e-01 1.21498883e-01 -2.88366199e-01
2.76655585e-01 6.12921193e-02 -2.09965348e-01 -9.91800010e-01
-6.77209675e-01 -6.08907163e-01 -1.35536718e+00 -2.32679516e-01
-5.66550456e-02 4.32006130e-03 1.82284400e-01 1.07685900e+00
4.13615763e-01 4.99604464e-01 8.72692227e-01 -2.22643092e-01
-3.10520053e-01 -5.89753330e-01 -6.60924375e-01 1.21254766e+00
5.24797022e-01 -7.70927489e-01 -2.78238624e-01 3.12942058e-01]
|
[7.899126052856445, 1.5612841844558716]
|
a170c41f-b214-46af-b5f7-a9c846400b76
|
source-free-domain-adaptation-requires
|
2304.02798
| null |
https://arxiv.org/abs/2304.02798v2
|
https://arxiv.org/pdf/2304.02798v2.pdf
|
Source-free Domain Adaptation Requires Penalized Diversity
|
While neural networks are capable of achieving human-like performance in many tasks such as image classification, the impressive performance of each model is limited to its own dataset. Source-free domain adaptation (SFDA) was introduced to address knowledge transfer between different domains in the absence of source data, thus, increasing data privacy. Diversity in representation space can be vital to a model`s adaptability in varied and difficult domains. In unsupervised SFDA, the diversity is limited to learning a single hypothesis on the source or learning multiple hypotheses with a shared feature extractor. Motivated by the improved predictive performance of ensembles, we propose a novel unsupervised SFDA algorithm that promotes representational diversity through the use of separate feature extractors with Distinct Backbone Architectures (DBA). Although diversity in feature space is increased, the unconstrained mutual information (MI) maximization may potentially introduce amplification of weak hypotheses. Thus we introduce the Weak Hypothesis Penalization (WHP) regularizer as a mitigation strategy. Our work proposes Penalized Diversity (PD) where the synergy of DBA and WHP is applied to unsupervised source-free domain adaptation for covariate shift. In addition, PD is augmented with a weighted MI maximization objective for label distribution shift. Empirical results on natural, synthetic, and medical domains demonstrate the effectiveness of PD under different distributional shifts.
|
['Mohammad Havaei', 'Thomas Fevens', 'Samira Ebrahimi Kahou', 'Alexandre See', 'Farhood Farahnak', 'Ivaxi Sheth', 'Laya Rafiee Sevyeri']
|
2023-04-06
| null | null | null | null |
['source-free-domain-adaptation']
|
['computer-vision']
|
[ 6.59391820e-01 1.20380864e-01 -3.00612986e-01 -5.07864356e-01
-5.92517793e-01 -3.45429391e-01 6.33698583e-01 2.62191385e-01
-3.99758011e-01 1.07550275e+00 2.69232541e-01 -4.21158299e-02
-3.80587965e-01 -7.10186005e-01 -5.98853290e-01 -9.88283336e-01
1.00597747e-01 8.35446790e-02 -2.96087209e-02 -1.52039722e-01
-1.68016747e-01 3.56669307e-01 -1.19544661e+00 1.34498462e-01
1.29103351e+00 8.41719210e-01 1.79549560e-01 4.10289243e-02
-7.92589635e-02 4.12757695e-01 -6.70187891e-01 -4.63189155e-01
6.35283947e-01 -5.53446770e-01 -3.76196861e-01 1.23236358e-01
2.26845533e-01 -3.82564180e-02 -1.76952735e-01 1.04611325e+00
7.63963282e-01 1.30375773e-01 9.04729486e-01 -1.37287652e+00
-6.71155810e-01 5.78894913e-01 -4.53813583e-01 -1.44127328e-02
-2.82640494e-02 -5.46688493e-03 8.55365217e-01 -8.64684582e-01
5.03198683e-01 8.87144268e-01 5.91032863e-01 5.80400765e-01
-1.58411455e+00 -9.26919878e-01 1.02722891e-01 -2.29834557e-01
-1.46691477e+00 -4.08789665e-01 9.72568452e-01 -5.32215357e-01
5.13269126e-01 1.86040521e-01 2.07253203e-01 1.21339750e+00
1.58237785e-01 7.10779548e-01 1.01892042e+00 -4.17948544e-01
5.12859106e-01 5.81120372e-01 -1.00993793e-02 4.03772831e-01
4.03418213e-01 1.19086824e-01 -7.08089232e-01 -5.65125108e-01
5.90990603e-01 1.01689905e-01 -2.91000992e-01 -9.50210214e-01
-1.14400387e+00 9.66837525e-01 4.51477200e-01 6.92004561e-02
-5.51266611e-01 -6.10031724e-01 3.38728786e-01 4.59966660e-01
4.66224015e-01 7.05045283e-01 -5.40727854e-01 5.60253143e-01
-8.35143089e-01 3.70827019e-01 6.11996949e-01 8.74271452e-01
7.97213137e-01 1.40283778e-01 -4.04982686e-01 9.95757341e-01
7.57927895e-02 5.38418114e-01 8.88198495e-01 -7.73791611e-01
3.69641095e-01 7.97872424e-01 -6.90254867e-02 -1.05708325e+00
-3.59309763e-01 -9.40157950e-01 -1.29390931e+00 1.49849519e-01
3.89672130e-01 -2.66570091e-01 -7.62379229e-01 2.28435898e+00
4.42150831e-01 -1.77872442e-02 4.18074369e-01 6.82618678e-01
5.74522197e-01 3.35326850e-01 2.66078800e-01 -4.23956245e-01
1.05205131e+00 -4.10638869e-01 -5.98040044e-01 -1.63424075e-01
4.94299442e-01 -2.97070742e-01 7.29439497e-01 2.56530076e-01
-7.03272343e-01 -3.20373416e-01 -1.12595868e+00 2.22489730e-01
-2.67525017e-01 -2.02000618e-01 3.73810679e-01 7.50802338e-01
-5.73479116e-01 4.02460694e-01 -5.00184476e-01 -3.05335075e-01
9.26239967e-01 5.74022830e-01 -5.28018296e-01 -7.42873996e-02
-1.35863972e+00 7.37279594e-01 5.17701089e-01 -3.72468054e-01
-5.34423172e-01 -8.76160383e-01 -7.65675128e-01 2.50190161e-02
3.29979211e-01 -1.01793754e+00 6.43244624e-01 -1.25569630e+00
-1.50196803e+00 5.77001274e-01 3.39906923e-02 -8.32987666e-01
6.10486269e-01 -9.79945213e-02 -4.54181224e-01 -7.90640116e-02
5.26968241e-02 7.29692817e-01 1.06667340e+00 -1.21817851e+00
-4.63654488e-01 -3.95306796e-01 -4.85362202e-01 3.76050979e-01
-7.99548984e-01 -3.63915026e-01 1.97139740e-01 -9.87313211e-01
2.01503634e-02 -8.00256014e-01 -3.93082768e-01 2.50395238e-02
-4.00676996e-01 6.06612116e-02 5.97611725e-01 -5.53369224e-01
1.40821946e+00 -2.41101670e+00 2.42518052e-01 5.89966714e-01
2.89307863e-01 4.24321800e-01 -2.24376068e-01 2.29213417e-01
-4.02346402e-02 -1.88363254e-01 -8.96174490e-01 -1.02824539e-01
-1.80457756e-01 2.39850879e-01 -3.55712771e-01 3.87089282e-01
4.28126693e-01 4.94112909e-01 -7.24895179e-01 -2.87977844e-01
-2.92062256e-02 4.63134140e-01 -7.81692982e-01 9.84979887e-03
-8.19531456e-02 6.40438735e-01 -4.91235286e-01 4.54927623e-01
8.29599082e-01 -3.66107494e-01 3.58334064e-01 -4.86667044e-02
2.19456673e-01 -1.04046024e-01 -1.26666498e+00 1.63915455e+00
-1.14036791e-01 1.35862395e-01 -1.55880585e-01 -1.01622295e+00
1.23656404e+00 1.92380637e-01 6.72790051e-01 -6.02828145e-01
2.27030553e-02 2.70287544e-01 2.30410680e-01 -1.48207501e-01
2.71388590e-01 -2.50664175e-01 -7.06931204e-02 1.91224247e-01
1.17029019e-01 2.22359657e-01 -2.11095721e-01 1.12725168e-01
1.12984574e+00 -2.16072455e-01 7.50368059e-01 -3.74219865e-01
4.87259209e-01 -1.95283651e-01 9.63515401e-01 8.32562983e-01
-4.01379138e-01 6.41692817e-01 1.97749928e-01 -1.33557782e-01
-9.62180912e-01 -1.19336021e+00 -4.91498709e-01 1.18711185e+00
6.76712394e-02 -2.86629163e-02 -4.86757785e-01 -9.20582414e-01
3.52054715e-01 7.63117194e-01 -5.82260132e-01 -4.87449855e-01
-2.67214775e-01 -9.54191744e-01 5.45528769e-01 3.16562533e-01
6.13927186e-01 -7.81763554e-01 -4.86780405e-01 2.37025946e-01
1.17343094e-03 -6.00317776e-01 -4.69504952e-01 5.52011609e-01
-7.94760346e-01 -7.74276614e-01 -9.28416729e-01 -5.36070168e-01
7.67391384e-01 9.89392847e-02 7.21466660e-01 -6.76470339e-01
2.29082210e-03 1.82350978e-01 -2.61532217e-01 -6.33212984e-01
-5.13565898e-01 2.08356574e-01 3.78613859e-01 3.25286865e-01
4.28093821e-01 -7.33965397e-01 -6.25654757e-01 3.68422955e-01
-1.09103978e+00 -1.48664206e-01 7.39574969e-01 1.25768089e+00
5.00904202e-01 -2.73576360e-02 1.10104227e+00 -1.09563446e+00
7.82899141e-01 -9.70700920e-01 -1.74582437e-01 1.78010836e-01
-8.63885343e-01 1.95615247e-01 6.44932091e-01 -6.97733164e-01
-1.20007789e+00 3.08673561e-01 2.46393681e-01 -4.75808442e-01
-1.21068053e-01 5.05847633e-01 -4.50507849e-01 1.37077361e-01
1.22445655e+00 3.08542430e-01 4.77728516e-01 -3.66469115e-01
3.05138320e-01 6.73641205e-01 3.00124645e-01 -4.01399493e-01
6.37018800e-01 3.40553761e-01 -6.24836348e-02 -7.54181147e-01
-6.68942153e-01 -3.43024611e-01 -5.30095339e-01 2.84005374e-01
4.83891815e-01 -1.04510462e+00 -5.27295396e-02 4.05654728e-01
-5.73961556e-01 -1.17390193e-02 -5.76012909e-01 5.62410653e-01
-3.94344181e-01 2.92938411e-01 2.40417734e-01 -5.45905411e-01
-3.48858505e-01 -8.83238912e-01 4.20948744e-01 2.43053436e-01
-3.88529122e-01 -9.35558319e-01 1.29552811e-01 1.57102466e-01
4.73155349e-01 4.03113604e-01 8.69390786e-01 -1.37161481e+00
-4.23483066e-02 -1.95990041e-01 -1.14969807e-02 7.37223506e-01
4.70415950e-01 -5.75953841e-01 -9.95773435e-01 -4.43748236e-01
5.97438700e-02 -9.28760841e-02 8.98226857e-01 4.36485052e-01
1.04564965e+00 -5.27642310e-01 -2.81872243e-01 5.58384061e-01
1.20174634e+00 2.19017461e-01 3.49857330e-01 3.07702959e-01
5.28516233e-01 5.69259524e-01 3.90699774e-01 7.56206512e-01
1.52753845e-01 5.58008552e-01 1.70841858e-01 1.11228991e-02
-4.09018900e-03 -2.78890431e-01 3.16628665e-01 4.28708017e-01
2.87494957e-01 -1.88261285e-01 -8.23506534e-01 5.41568577e-01
-1.84321678e+00 -7.70697773e-01 2.90860891e-01 2.44723988e+00
1.01699829e+00 -2.55322438e-02 2.88586229e-01 8.79381523e-02
7.18992233e-01 -6.21812344e-02 -9.95253026e-01 -9.74023268e-02
-5.66216826e-01 5.08077070e-02 5.44405937e-01 5.86737283e-02
-1.08102059e+00 4.31504846e-01 5.75036287e+00 9.33093131e-01
-1.05456793e+00 -2.27334211e-03 5.40676773e-01 -1.44335479e-01
-3.99984062e-01 -2.41191477e-01 -5.88791132e-01 5.76165199e-01
6.29371524e-01 -3.60246837e-01 1.29109278e-01 9.20879006e-01
-1.31847695e-01 1.32219821e-01 -1.03331566e+00 8.05426419e-01
2.49874238e-02 -1.11963832e+00 2.54507273e-01 1.93473607e-01
8.79736066e-01 -6.65609464e-02 3.89991999e-01 2.25302339e-01
3.24878812e-01 -7.52707541e-01 3.08868766e-01 4.28313792e-01
7.21178174e-01 -8.61203015e-01 5.74045122e-01 4.67636526e-01
-5.33699155e-01 -4.10094380e-01 -2.79341936e-01 1.79621369e-01
-2.16300696e-01 7.37616122e-01 -1.08614600e+00 6.25566065e-01
3.56654406e-01 6.25611782e-01 -4.24025595e-01 8.36705208e-01
2.29739457e-01 5.69769084e-01 -3.58605236e-01 2.21760407e-01
-1.65833473e-01 -1.75011400e-02 8.76115084e-01 1.04591751e+00
3.17235351e-01 1.25388158e-02 1.49532050e-01 6.28539622e-01
-1.51918858e-01 2.40739644e-01 -8.59398186e-01 5.69849089e-02
7.88019419e-01 8.40407133e-01 -9.54788327e-02 5.81865199e-02
-3.09971690e-01 8.86896133e-01 2.50245333e-01 3.92804325e-01
-4.73012716e-01 -3.10684741e-01 8.24324012e-01 3.05218279e-01
3.05605650e-01 1.39098549e-02 -6.15399480e-01 -1.03401029e+00
-8.02998468e-02 -1.06657338e+00 7.53193438e-01 -8.03972334e-02
-1.77535331e+00 3.94312620e-01 7.97368661e-02 -1.50618005e+00
-1.19707752e-02 -2.41843998e-01 -2.75657713e-01 9.77221429e-01
-1.49835908e+00 -9.44072008e-01 1.47165433e-01 8.69937360e-01
2.48357877e-01 -7.70747066e-01 8.81233037e-01 2.53653467e-01
-5.24410307e-01 1.15603662e+00 6.00376785e-01 -1.30540803e-01
1.18408883e+00 -9.73694682e-01 -2.82406926e-01 6.38975441e-01
-1.53830901e-01 8.47170293e-01 6.76871359e-01 -7.20673025e-01
-1.03039360e+00 -1.41345322e+00 6.88100457e-01 -2.70324260e-01
2.33182400e-01 -2.33124450e-01 -1.25291729e+00 3.71190310e-01
-7.58372098e-02 -6.86666146e-02 1.31647813e+00 8.82807001e-02
-6.08244658e-01 -3.78096849e-01 -1.65540278e+00 4.38750535e-01
8.07223737e-01 -2.01288819e-01 -4.48681086e-01 1.31491631e-01
6.68329298e-01 -8.84387642e-02 -9.81512666e-01 5.05292118e-01
3.98314625e-01 -7.19783485e-01 9.64656889e-01 -6.13377988e-01
3.56915981e-01 -3.17390501e-01 -4.81849909e-01 -1.40460825e+00
-5.69139898e-01 -4.85007554e-01 -2.25957125e-01 1.34240568e+00
3.84808332e-01 -8.63372207e-01 5.74995041e-01 8.47596765e-01
1.15773767e-01 -5.42347908e-01 -1.05120599e+00 -8.63413572e-01
2.05895886e-01 -1.31373096e-03 6.95013702e-01 1.30796897e+00
2.01652152e-03 2.69394159e-01 -5.96932113e-01 2.90332228e-01
7.35695899e-01 -2.38667458e-01 8.36604714e-01 -1.34208870e+00
-4.24396247e-01 -2.60514289e-01 -3.83810669e-01 -5.20468771e-01
6.29529683e-03 -1.22005415e+00 -1.66257262e-01 -8.59978199e-01
2.14441985e-01 -4.32715178e-01 -8.55188966e-01 5.43196738e-01
-3.21410805e-01 -1.57013282e-01 1.76721647e-01 3.70702446e-01
-2.24105760e-01 7.32838869e-01 9.79245961e-01 -6.09814078e-02
-5.87875962e-01 8.63291770e-02 -1.08175051e+00 5.73633552e-01
8.84766638e-01 -6.77707672e-01 -6.32304430e-01 -1.71146363e-01
-4.39684354e-02 -3.24671090e-01 2.81533509e-01 -9.64571893e-01
1.80264249e-01 -2.83985049e-01 4.49029535e-01 -1.46690428e-01
-1.17372423e-02 -9.23835933e-01 2.50652969e-01 5.00772178e-01
-6.22415900e-01 -5.11101305e-01 1.66950136e-01 8.94254029e-01
-1.68621913e-01 1.58284813e-01 1.07696700e+00 9.89150852e-02
-4.69886154e-01 1.90716878e-01 7.85956159e-04 2.90305037e-02
1.08114159e+00 -3.15041929e-01 -1.31822705e-01 -4.15958390e-02
-5.57333171e-01 1.89589456e-01 3.42289358e-01 3.60111952e-01
5.69802165e-01 -1.39482093e+00 -9.44283664e-01 5.17987072e-01
2.72448301e-01 8.26341659e-02 3.15623879e-01 5.69506228e-01
2.13519737e-01 7.26014301e-02 -3.48063409e-01 -5.33414364e-01
-1.14957941e+00 5.39255440e-01 1.54112816e-01 -2.92471081e-01
-4.15213555e-01 8.14582288e-01 5.67129493e-01 -6.84152126e-01
1.77382767e-01 1.83475554e-01 -1.93499908e-01 1.35100216e-01
3.75246704e-01 3.79087836e-01 -5.01874536e-02 -3.74904662e-01
-3.62632930e-01 6.57113492e-02 -4.76749331e-01 5.60781099e-02
1.35289860e+00 -2.66553760e-01 2.47901678e-01 2.40787044e-01
1.02963006e+00 -6.48119077e-02 -1.38252318e+00 -8.46960068e-01
-1.24551639e-01 -4.10798192e-01 -1.02844507e-01 -1.05730486e+00
-7.40983427e-01 5.02249360e-01 7.86825776e-01 -6.96387067e-02
1.45431268e+00 -3.06651115e-01 5.11598527e-01 3.11945677e-01
1.28453597e-01 -1.09638691e+00 -8.86288844e-03 3.09235185e-01
8.31373751e-01 -1.35312939e+00 5.67519516e-02 -5.76930344e-02
-1.04217613e+00 6.45394027e-01 4.36092943e-01 3.89669910e-02
6.57177567e-01 3.19756707e-03 -1.39642954e-01 2.20322043e-01
-5.06400526e-01 4.56801504e-02 3.76818538e-01 7.70111382e-01
1.16500825e-01 9.97693017e-02 -3.99370253e-01 9.94751692e-01
5.97411417e-04 1.25643641e-01 1.70402527e-01 9.72523689e-01
-2.84785181e-01 -1.28212857e+00 -2.92324901e-01 5.99556148e-01
-3.26595336e-01 -1.57672092e-02 -3.55130970e-01 5.03396213e-01
2.83347338e-01 7.02907443e-01 -1.16137020e-01 -3.33600879e-01
2.14376643e-01 3.51565957e-01 9.96013954e-02 -5.35432637e-01
-6.58106089e-01 -1.89344212e-02 -2.54181892e-01 -1.45725161e-01
-3.43397468e-01 -8.56627345e-01 -9.75920022e-01 9.95465741e-02
-1.67393520e-01 6.24095090e-02 5.19933820e-01 8.16685677e-01
8.53970528e-01 3.63873035e-01 8.71397972e-01 -1.10046282e-01
-8.53148103e-01 -7.35279739e-01 -7.40465224e-01 6.05452716e-01
5.24452150e-01 -6.23040855e-01 -2.23973781e-01 4.65511084e-02]
|
[10.356512069702148, 3.239576816558838]
|
33bd9b0b-a740-45d1-af3b-55a2d2c2825e
|
graph-sequential-neural-ode-process-for-link
|
2211.08568
| null |
https://arxiv.org/abs/2211.08568v1
|
https://arxiv.org/pdf/2211.08568v1.pdf
|
Graph Sequential Neural ODE Process for Link Prediction on Dynamic and Sparse Graphs
|
Link prediction on dynamic graphs is an important task in graph mining. Existing approaches based on dynamic graph neural networks (DGNNs) typically require a significant amount of historical data (interactions over time), which is not always available in practice. The missing links over time, which is a common phenomenon in graph data, further aggravates the issue and thus creates extremely sparse and dynamic graphs. To address this problem, we propose a novel method based on the neural process, called Graph Sequential Neural ODE Process (GSNOP). Specifically, GSNOP combines the advantage of the neural process and neural ordinary differential equation that models the link prediction on dynamic graphs as a dynamic-changing stochastic process. By defining a distribution over functions, GSNOP introduces the uncertainty into the predictions, making it generalize to more situations instead of overfitting to the sparse data. GSNOP is also agnostic to model structures that can be integrated with any DGNN to consider the chronological and geometrical information for link prediction. Extensive experiments on three dynamic graph datasets show that GSNOP can significantly improve the performance of existing DGNNs and outperform other neural process variants.
|
['Shirui Pan', 'Reza Haffari', 'Linhao Luo']
|
2022-11-15
| null | null | null | null |
['graph-mining']
|
['graphs']
|
[-2.77969420e-01 6.17089942e-02 -9.88565236e-02 9.26361233e-02
2.33640894e-01 -2.36212865e-01 4.39262092e-01 3.11508566e-01
2.22928584e-01 6.70111477e-01 -1.01667121e-01 -5.66692412e-01
-4.86706227e-01 -1.46828210e+00 -6.23280108e-01 -6.61774635e-01
-4.33385134e-01 7.80217826e-01 4.57796425e-01 -3.50901097e-01
-2.77205944e-01 5.07479787e-01 -1.01532876e+00 -3.29389304e-01
1.25524771e+00 8.78584445e-01 1.76182002e-01 4.28943366e-01
-6.28425181e-01 6.62935138e-01 -2.78849602e-01 -5.69780946e-01
4.20064658e-01 2.15587318e-02 -3.40266109e-01 -2.66566664e-01
-1.18965775e-01 -2.27429178e-02 -9.53714490e-01 1.06822002e+00
4.07360464e-01 3.22684497e-01 6.34671807e-01 -1.49139404e+00
-8.56259048e-01 7.79950202e-01 -6.31747305e-01 2.60416001e-01
1.05801210e-01 1.55414954e-01 9.33289766e-01 -3.90260071e-01
4.89104182e-01 1.49346530e+00 1.21989012e+00 1.28920123e-01
-1.28121912e+00 -7.16966093e-01 6.19905829e-01 1.84857979e-01
-1.27426016e+00 2.94930786e-01 1.39079356e+00 -5.22210598e-01
5.57306111e-01 -1.56411715e-02 1.04313302e+00 1.22688103e+00
5.53235114e-01 6.09947860e-01 6.57684922e-01 2.54269302e-01
4.64509912e-02 -4.83451813e-01 2.63905019e-01 6.28171861e-01
4.52721387e-01 2.17025757e-01 -1.88210517e-01 -2.17214063e-01
1.04179823e+00 4.00698006e-01 -3.23531568e-01 -4.06774789e-01
-7.95348465e-01 6.09203517e-01 6.20636702e-01 3.31246033e-02
-4.19112235e-01 2.12248519e-01 3.77780288e-01 4.62683141e-01
6.93235993e-01 1.13325097e-01 -4.41629261e-01 -2.78184116e-02
-5.72926164e-01 1.08450137e-01 1.21595454e+00 8.21399570e-01
6.48029029e-01 3.06482226e-01 -4.41168509e-02 7.94323325e-01
3.33157033e-01 3.83097172e-01 3.87339443e-01 -2.86597103e-01
5.86024582e-01 1.07744861e+00 -2.15714440e-01 -1.84792626e+00
-6.17298007e-01 -7.47249007e-01 -1.44544017e+00 -2.86010236e-01
3.20191622e-01 -4.77068424e-01 -9.72810686e-01 1.58862960e+00
3.33246946e-01 3.89648408e-01 -1.27052605e-01 4.49980080e-01
1.00612986e+00 1.07588112e+00 1.58989742e-01 -3.72305870e-01
5.08331478e-01 -9.21925068e-01 -7.44573712e-01 -4.21854146e-02
3.31000060e-01 -3.01684022e-01 6.12475753e-01 1.91212922e-01
-6.54326320e-01 -5.51647305e-01 -7.88070083e-01 3.31630498e-01
-3.78032416e-01 -3.35443467e-01 1.18552160e+00 2.41739213e-01
-1.15562618e+00 1.05560958e+00 -9.04946089e-01 -3.34108293e-01
2.01680437e-01 5.08484960e-01 -1.41889170e-01 -9.17689130e-03
-1.39562023e+00 5.35144925e-01 7.24445939e-01 5.66928983e-01
-4.66626197e-01 -6.63983405e-01 -7.68210351e-01 3.16654623e-01
7.54613698e-01 -6.51673973e-01 6.74127996e-01 -7.94765472e-01
-1.47506988e+00 -4.81404960e-02 1.22499160e-01 -3.81094396e-01
7.42902100e-01 1.57146141e-01 -7.47366786e-01 -1.39845476e-01
-2.21479893e-01 2.91150790e-02 7.50360668e-01 -1.10527456e+00
-1.97882876e-01 -7.58747011e-02 -3.92441787e-02 4.09341045e-02
-2.69199401e-01 -6.83309436e-01 -6.33493006e-01 -8.95533860e-01
3.59205902e-01 -1.03113985e+00 -5.30026734e-01 -2.05820292e-01
-6.59256697e-01 -4.76956785e-01 9.90096986e-01 -8.93490374e-01
1.37673616e+00 -1.75258625e+00 1.71432674e-01 4.93615538e-01
5.76987803e-01 1.73567697e-01 -1.46896884e-01 7.60336041e-01
-1.28755093e-01 1.69748098e-01 -5.26706576e-02 6.62936568e-02
-1.91258147e-01 4.03762579e-01 -5.18963858e-02 5.01643829e-02
-5.51892631e-02 1.07970917e+00 -1.06502295e+00 -4.33526784e-01
-4.31875400e-02 3.43992203e-01 -2.45655507e-01 -5.17042316e-02
-4.55574989e-01 3.93631101e-01 -6.96876585e-01 5.90369701e-01
7.57465303e-01 -5.30928791e-01 2.54566222e-01 -4.91561405e-02
1.47247583e-01 -2.10018352e-01 -1.39669085e+00 1.25483108e+00
-2.63066471e-01 4.13710833e-01 -2.14215800e-01 -1.13063478e+00
1.34410918e+00 2.71101505e-01 8.21619332e-01 -3.64644408e-01
3.62203233e-02 -3.73720564e-02 2.55963773e-01 -3.41587156e-01
2.64450610e-01 -7.04867905e-03 1.43032789e-01 1.55022144e-01
6.60329536e-02 2.43277282e-01 3.95375371e-01 1.76323637e-01
1.34236705e+00 8.97815749e-02 5.54515794e-03 -2.38616019e-01
5.05901933e-01 -1.97639167e-01 1.10445869e+00 7.37180531e-01
3.10325492e-02 1.60149127e-01 9.71601605e-01 -6.40044630e-01
-7.94328451e-01 -1.03762496e+00 3.16083491e-01 5.32273173e-01
2.78089285e-01 -3.61223876e-01 -1.92131862e-01 -7.05615103e-01
3.44845891e-01 2.64361858e-01 -4.93751794e-01 -3.15056324e-01
-4.65099484e-01 -9.84318197e-01 4.50318903e-02 6.37877405e-01
4.41733807e-01 -9.97332335e-01 6.06917202e-01 6.32033169e-01
1.08075909e-01 -9.47053730e-01 -2.23943844e-01 1.58278439e-02
-1.07681978e+00 -1.11489964e+00 -4.95541453e-01 -7.02257693e-01
6.58475399e-01 1.11984998e-01 9.91207778e-01 2.20912352e-01
8.78701508e-02 2.30382323e-01 -1.91792741e-01 -3.82756352e-01
-3.03674072e-01 3.64865541e-01 4.18052189e-02 1.99439391e-01
1.22651532e-01 -1.25840211e+00 -3.59799147e-01 1.23972856e-01
-5.45391083e-01 2.23318383e-01 7.02557504e-01 8.10192466e-01
5.38505495e-01 7.79613674e-01 5.98613620e-01 -1.00032997e+00
9.30591464e-01 -7.86914408e-01 -7.45597661e-01 2.29631364e-01
-8.49161327e-01 1.75007641e-01 9.64324534e-01 -7.64187515e-01
-9.85799432e-01 -7.77073801e-02 3.76061872e-02 -6.19173646e-01
3.03593099e-01 1.00790477e+00 -3.35168362e-01 -2.46889606e-01
1.98933855e-01 2.15619922e-01 3.05246282e-02 -5.13008773e-01
1.73658691e-02 -8.74719694e-02 2.68966347e-01 -4.63692546e-01
1.32417440e+00 2.14383677e-01 5.47728837e-01 -6.08250737e-01
-4.36608821e-01 -2.33423963e-01 -5.28400242e-01 -4.40772057e-01
4.12937611e-01 -6.46394432e-01 -8.18741083e-01 6.60324454e-01
-1.09362054e+00 -4.31606680e-01 -1.19562536e-01 3.96882504e-01
-1.65808290e-01 5.17706394e-01 -1.00487268e+00 -7.59295464e-01
-3.78930897e-01 -7.69109964e-01 3.35613221e-01 4.17181820e-01
1.82816759e-01 -1.51622784e+00 2.24539805e-02 -2.96581119e-01
3.61048490e-01 7.84155965e-01 1.07943869e+00 -8.53594601e-01
-7.70580232e-01 -5.97577751e-01 -2.78329521e-01 1.18688583e-01
1.88122541e-01 3.13059062e-01 -2.37511799e-01 -9.65507329e-03
-9.01881456e-02 3.53783399e-01 6.89809203e-01 4.82638806e-01
1.20187247e+00 -3.91939372e-01 -6.07496500e-01 6.14270389e-01
1.51306391e+00 3.46566647e-01 4.44144100e-01 1.07411994e-02
1.33379889e+00 6.79050922e-01 2.60641873e-01 1.43972352e-01
6.33777559e-01 2.75854617e-01 5.24724960e-01 -5.99500574e-02
1.64624989e-01 -6.30368710e-01 8.17015320e-02 1.27018154e+00
-3.49205315e-01 -4.24503773e-01 -1.17620802e+00 3.40985477e-01
-2.23040962e+00 -9.72057760e-01 -7.51933455e-01 1.84241307e+00
4.64238226e-01 3.82240117e-01 -1.25616387e-01 2.87495535e-02
1.11262453e+00 3.72147143e-01 -7.29236841e-01 -1.49114937e-01
1.92905031e-02 -7.11170733e-02 5.76315880e-01 3.25758636e-01
-9.72980559e-01 8.21680486e-01 5.64102554e+00 9.06509936e-01
-1.11958122e+00 -3.58623080e-02 4.93477553e-01 2.30438560e-01
-2.77596444e-01 1.54975072e-01 -7.10803032e-01 7.27135420e-01
7.81269729e-01 -5.27112544e-01 5.00636756e-01 9.03674245e-01
3.54410350e-01 2.40146071e-01 -8.66845250e-01 9.07501638e-01
-2.89930135e-01 -1.15377665e+00 2.08217308e-01 3.15677911e-01
9.44605172e-01 -5.40859699e-02 -1.90522343e-01 6.56982958e-01
8.39690626e-01 -1.00556469e+00 2.71145087e-02 1.07870924e+00
1.84504032e-01 -8.30885470e-01 7.80386150e-01 4.30688381e-01
-1.59467411e+00 -1.87192142e-01 -4.09360677e-01 -1.11667581e-01
5.42552590e-01 1.07265913e+00 -5.95540822e-01 1.08020031e+00
4.99476999e-01 1.14126372e+00 -5.76296568e-01 1.31932521e+00
-2.27086350e-01 6.28518105e-01 -5.51420093e-01 -2.05039948e-01
1.12527080e-01 -5.53965330e-01 8.51356447e-01 6.27750814e-01
4.87350255e-01 -2.02360630e-01 6.66096628e-01 6.90556645e-01
-4.42988500e-02 2.46214032e-01 -6.80542946e-01 -2.72614807e-01
3.16520482e-01 1.24599290e+00 -8.85069847e-01 -6.17729947e-02
-3.94005507e-01 6.27942264e-01 2.92945862e-01 6.65371597e-01
-7.21000373e-01 -1.69900045e-01 3.39549541e-01 2.35673979e-01
2.30704695e-01 -4.91978705e-01 -5.66315353e-02 -1.00740409e+00
2.22235173e-01 -4.11700487e-01 5.62939525e-01 -5.51358581e-01
-1.80053627e+00 5.48164129e-01 -1.60930440e-01 -1.25608933e+00
-1.20746970e-01 -5.33829927e-01 -1.01471770e+00 8.04239154e-01
-1.35189092e+00 -1.30174339e+00 -5.96100509e-01 4.54079598e-01
2.56102979e-01 2.33628955e-02 3.67741048e-01 2.86380619e-01
-8.60191882e-01 1.57972157e-01 2.01628745e-01 1.78066924e-01
3.34379643e-01 -1.32919478e+00 5.68686664e-01 8.43638062e-01
-1.31351620e-01 4.23627973e-01 6.32725716e-01 -1.33613086e+00
-1.49609005e+00 -1.33500111e+00 4.84287530e-01 -1.34514526e-01
1.28392899e+00 -1.84659615e-01 -1.28588021e+00 7.55157471e-01
-3.18167686e-01 1.87249944e-01 3.97418767e-01 5.12569785e-01
1.93489239e-01 -3.23534429e-01 -7.91258991e-01 6.13349020e-01
1.42250335e+00 -1.75055042e-01 -1.32675126e-01 4.69938070e-01
9.11079049e-01 -4.48864758e-01 -1.14174449e+00 7.06832767e-01
1.79752111e-01 -5.49339652e-01 7.96616018e-01 -4.38131273e-01
3.17220181e-01 -2.60670424e-01 5.12966633e-01 -1.55517447e+00
-5.32620192e-01 -8.05579662e-01 -9.03164148e-01 1.42128098e+00
3.18811208e-01 -9.98120844e-01 9.82141018e-01 5.28000057e-01
1.52427070e-02 -9.43045616e-01 -6.38967872e-01 -1.03323066e+00
-1.04290649e-01 -3.97696942e-01 8.01997185e-01 1.10014844e+00
-4.83947277e-01 2.31203780e-01 -4.79621679e-01 3.50164592e-01
5.32826245e-01 3.62324603e-02 8.11115503e-01 -2.04645658e+00
-5.29858768e-01 -5.06500185e-01 -6.52743340e-01 -9.79467750e-01
1.98992729e-01 -1.01098847e+00 -2.05013618e-01 -1.98923564e+00
-1.97712854e-01 -6.20283127e-01 -1.85672641e-01 2.08453223e-01
-2.99214482e-01 -4.08113807e-01 -8.84280354e-02 3.23451847e-01
-2.98295796e-01 8.87692690e-01 1.45553863e+00 -1.54368892e-01
-5.70042312e-01 3.10774356e-01 -5.15251398e-01 8.15350175e-01
7.32389152e-01 -3.81179988e-01 -6.45219982e-01 -9.57334414e-02
4.33831215e-01 2.76053756e-01 2.36650869e-01 -1.22221076e+00
5.07401943e-01 -2.09342718e-01 3.92050743e-01 -8.48419070e-01
2.81697921e-02 -1.02103734e+00 6.73595846e-01 5.01410782e-01
1.88113615e-01 3.03678721e-01 1.12389863e-01 1.32891798e+00
-1.05433427e-01 -1.79854706e-02 1.34694636e-01 -1.28114328e-01
-7.22725630e-01 1.09497118e+00 -1.33067831e-01 -2.25581720e-01
1.04760647e+00 -1.84256107e-01 -2.06801608e-01 -5.52864194e-01
-1.00619352e+00 7.43142664e-01 2.63753116e-01 4.28574890e-01
2.72481740e-01 -1.46062469e+00 -2.99603909e-01 -1.40276551e-01
-2.31401637e-01 4.16032076e-01 4.80816662e-01 8.66522133e-01
-5.68227410e-01 1.07760161e-01 1.65578946e-01 -4.83046532e-01
-7.79010594e-01 8.23873937e-01 3.22508305e-01 -8.70348513e-01
-1.05268180e+00 3.79036784e-01 -7.19841495e-02 -5.30025244e-01
8.42606872e-02 -1.52803376e-01 -3.59078348e-01 1.18990630e-01
-1.76389381e-01 2.89151430e-01 -1.96176767e-01 -2.03974172e-01
2.52655506e-01 3.52055341e-01 4.87254672e-02 3.96771610e-01
1.49872589e+00 -1.08759627e-02 -2.65967667e-01 7.51925290e-01
8.40214074e-01 -8.52347836e-02 -1.43006968e+00 -2.99279839e-01
6.37476519e-02 -2.67285146e-02 -1.40448168e-01 -4.52717930e-01
-1.57643521e+00 5.69918752e-01 2.08222717e-01 7.66568780e-01
8.33722532e-01 -2.87335604e-01 1.06281972e+00 4.61193860e-01
3.28200489e-01 -1.09443140e+00 -5.92869590e-04 7.05652535e-01
8.06432843e-01 -9.88860130e-01 -3.11270114e-02 -9.11305428e-01
-2.85949320e-01 1.23499346e+00 9.14231062e-01 -4.27291483e-01
1.14516914e+00 2.03509405e-02 -4.70435143e-01 -2.52236933e-01
-6.25916362e-01 4.71473252e-03 2.51396567e-01 6.46813333e-01
-9.71731544e-02 3.73687223e-02 -3.93792242e-01 6.40355110e-01
-2.13138700e-01 -2.49613211e-01 3.83531898e-01 7.31427848e-01
-2.02835612e-02 -1.08962870e+00 -6.99307323e-02 9.36285615e-01
-7.60873854e-02 -6.30562380e-02 -2.34896749e-01 8.84308398e-01
-1.14596784e-01 5.51993191e-01 6.47379905e-02 -4.87820029e-01
1.81094885e-01 1.24551132e-02 -4.26501743e-02 -4.71412718e-01
-2.70399421e-01 -2.72536606e-01 1.33648574e-01 -3.88322413e-01
-6.17895089e-02 -5.25495768e-01 -1.06364179e+00 -6.03863478e-01
-4.22916263e-01 2.37422790e-02 4.22642857e-01 7.58238673e-01
4.21199322e-01 9.04919267e-01 5.07718027e-01 -7.32535839e-01
-1.18143752e-01 -9.36153650e-01 -7.16263771e-01 2.36051545e-01
1.47991301e-02 -8.86902452e-01 -3.72809380e-01 -4.15285319e-01]
|
[7.234675884246826, 5.931657314300537]
|
b33fd621-848a-43f2-b951-7bfa3567ccdb
|
a-multi-domain-vne-algorithm-based-on-multi
|
2202.1283
| null |
https://arxiv.org/abs/2202.12830v1
|
https://arxiv.org/pdf/2202.12830v1.pdf
|
A multi-domain VNE algorithm based on multi-objective optimization for IoD architecture in Industry 4.0
|
Unmanned aerial vehicle (UAV) has a broad application prospect in the future, especially in the Industry 4.0. The development of Internet of Drones (IoD) makes UAV operation more autonomous. Network virtualization technology is a promising technology to support IoD, so the allocation of virtual resources becomes a crucial issue in IoD. How to rationally allocate potential material resources has become an urgent problem to be solved. The main work of this paper is presented as follows: (1) In order to improve the optimization performance and reduce the computation time, we propose a multi-domain virtual network embedding algorithm (MP-VNE) adopting the centralized hierarchical multi-domain architecture. The proposed algorithm can avoid the local optimum through incorporating the genetic variation factor into the traditional particle swarm optimization process. (2) In order to simplify the multi-objective optimization problem, we transform the multi-objective problem into a single-objective problem through weighted summation method. The results prove that the proposed algorithm can rapidly converge to the optimal solution. (3) In order to reduce the mapping cost, we propose an algorithm for selecting candidate nodes based on the estimated mapping cost. Each physical domain calculates the estimated mapping cost of all nodes according to the formula of the estimated mapping cost, and chooses the node with the lowest estimated mapping cost as the candidate node. The simulation results show that the proposed MP-VNE algorithm has better performance than MC-VNM, LID-VNE and other algorithms in terms of delay, cost and comprehensive indicators.
|
['Haotong Cao', 'Zeyu Qin', 'Chao Wang', 'Peiying Zhang']
|
2022-02-08
| null | null | null | null |
['network-embedding']
|
['methodology']
|
[-2.42465660e-01 -4.06526268e-01 -1.86075434e-01 4.55720305e-01
3.90137672e-01 -2.71943092e-01 -8.87112841e-02 -6.82566315e-02
-4.25561011e-01 9.15384352e-01 -5.02486408e-01 -2.64891684e-01
-8.27684045e-01 -1.15963399e+00 -2.25152429e-02 -9.38189089e-01
-5.33699542e-02 3.89626175e-01 4.64443833e-01 -4.39204872e-01
1.96274698e-01 5.72913170e-01 -1.28153646e+00 -7.72176504e-01
1.20462799e+00 1.14416420e+00 7.67220795e-01 1.02995440e-01
3.80752608e-02 1.77941218e-01 -1.09124541e+00 1.64293677e-01
4.46851224e-01 -3.92462671e-01 -6.32287681e-01 2.19401777e-01
-7.94346035e-01 -2.87324995e-01 -5.88426366e-02 1.25981843e+00
6.06405199e-01 3.13873261e-01 3.52010518e-01 -1.70064712e+00
-1.98954403e-01 2.00180873e-01 -7.83530116e-01 3.23751241e-01
-2.08867997e-01 -2.14738041e-01 5.50284028e-01 -3.83096129e-01
5.46158493e-01 1.04694676e+00 1.80300593e-01 1.02334104e-01
-6.42515302e-01 -7.69799829e-01 2.69054383e-01 2.90711790e-01
-1.77039456e+00 1.47983342e-01 6.36250675e-01 -1.32679626e-01
4.25057471e-01 3.34035516e-01 9.14774120e-01 -6.83412924e-02
4.78749514e-01 5.77623472e-02 6.49758995e-01 -3.77539903e-01
2.61884660e-01 3.62485275e-02 -4.49127406e-01 6.21716738e-01
9.19959784e-01 5.39435409e-02 2.13323370e-01 -2.13057503e-01
8.95062685e-01 6.10841475e-02 -4.87549186e-01 -4.00398046e-01
-1.27271712e+00 7.65563488e-01 5.03807783e-01 3.09973210e-01
-6.13893867e-01 1.84098934e-03 3.38439047e-01 2.12849900e-01
3.60035807e-01 5.17512143e-01 -4.00284588e-01 4.61259373e-02
-6.53567612e-01 1.01827748e-01 5.61883509e-01 8.93334448e-01
6.52523935e-01 2.80130148e-01 3.37794811e-01 6.37701631e-01
3.33278537e-01 5.43123007e-01 2.06400469e-01 -8.70876431e-01
3.34354490e-01 6.86385095e-01 3.06903332e-01 -1.31967449e+00
-4.94055718e-01 -5.85856557e-01 -8.87022734e-01 3.91538799e-01
-2.95984715e-01 -6.88040495e-01 -3.48310024e-01 1.37844276e+00
6.22276962e-01 2.86776274e-01 2.18899280e-01 1.07218564e+00
4.61710840e-01 1.02168286e+00 -2.07437471e-01 -8.38769913e-01
1.13488388e+00 -7.55585074e-01 -8.77753794e-01 2.51565874e-01
3.18677694e-01 -9.90403771e-01 3.10771316e-01 2.26023510e-01
-7.69287467e-01 -4.00282323e-01 -1.37957251e+00 8.23362708e-01
-3.00992817e-01 3.02989453e-01 4.37447220e-01 6.06750429e-01
-9.80298042e-01 2.62807280e-01 -3.84151816e-01 -5.46850979e-01
1.09752081e-01 7.03519642e-01 5.41580468e-02 -3.57079059e-02
-1.33984792e+00 7.90297270e-01 7.86008060e-01 1.34511024e-01
-6.55440748e-01 -3.70900631e-01 -5.00861704e-01 4.71202061e-02
7.77336955e-01 -6.41591668e-01 6.57473207e-01 -6.41424716e-01
-1.45619404e+00 -7.17759132e-03 6.38063177e-02 -1.05370119e-01
9.27618593e-02 5.05065858e-01 -6.51783288e-01 1.07473768e-01
1.31593630e-01 3.12399656e-01 5.01690447e-01 -1.16200840e+00
-1.06660295e+00 -1.60941761e-03 3.41677219e-01 4.17903155e-01
-6.47855341e-01 -5.35006216e-03 -2.37258166e-01 -4.41635191e-01
-5.69076352e-02 -9.54256654e-01 -5.22401989e-01 -2.10053548e-01
-1.12336904e-01 -1.88445926e-01 1.26876962e+00 -2.09898248e-01
1.44699192e+00 -1.95060396e+00 6.17254913e-01 4.41347629e-01
3.22587609e-01 4.58978862e-01 3.04592829e-02 4.47203994e-01
4.02321994e-01 4.29050215e-02 1.13923348e-01 4.22605306e-01
-2.87902087e-01 1.19967490e-01 1.86862782e-01 3.06583762e-01
-1.87022820e-01 1.39080405e-01 -8.61639500e-01 -5.97477317e-01
3.82617176e-01 2.16064855e-01 -4.25274372e-01 -4.58891504e-02
-7.51892626e-02 1.08464949e-01 -8.40783596e-01 7.91197300e-01
1.27144206e+00 -2.61982560e-01 4.51796949e-01 -3.68697613e-01
-5.95136106e-01 -5.03134012e-01 -1.49082053e+00 1.11739278e+00
-2.89872169e-01 2.85916090e-01 6.05719984e-01 -9.40060854e-01
1.02234602e+00 1.99439645e-01 9.04125571e-01 -2.93393731e-01
5.43593287e-01 3.99182498e-01 1.79708511e-01 -3.01907241e-01
5.49501836e-01 -6.78580534e-03 -3.66933197e-02 3.23020220e-01
-2.34588429e-01 -1.05893519e-02 2.96139836e-01 7.59859905e-02
8.10279489e-01 -2.82577723e-01 4.09169048e-01 -4.58364964e-01
9.16543424e-01 3.38553280e-01 1.05106330e+00 -5.86021543e-02
-3.50725800e-01 -3.55657548e-01 4.04380560e-01 -3.11152160e-01
-7.10357547e-01 -6.40055954e-01 8.47126096e-02 2.35785767e-01
1.26711321e+00 -1.88226119e-01 -4.92746919e-01 -3.61913174e-01
4.01009107e-04 3.42416734e-01 -2.83603035e-02 -2.52684265e-01
-2.23916769e-01 -9.31377470e-01 4.79345433e-02 -2.39838302e-01
9.13850963e-01 -5.33741415e-01 -6.43210411e-01 4.23664033e-01
-1.19728386e-01 -9.74055827e-01 -2.11352319e-01 -3.13419193e-01
-6.86405599e-01 -9.61045146e-01 -6.72498941e-01 -9.61249292e-01
7.09305167e-01 8.27032745e-01 4.61431593e-01 4.31709856e-01
-2.39548653e-01 8.67053121e-02 -5.60438395e-01 -3.52741182e-01
4.22309376e-02 1.93223342e-01 4.53783542e-01 -1.51364207e-01
-2.08929125e-02 -4.30992991e-01 -4.24732059e-01 7.36657262e-01
-7.77693629e-01 -2.70409971e-01 5.42569697e-01 7.78810203e-01
7.80644953e-01 1.26878047e+00 6.54040515e-01 -1.22522801e-01
1.04324305e+00 -6.26548171e-01 -1.15744340e+00 2.73679376e-01
-6.25803173e-01 -3.64524454e-01 8.20072770e-01 -2.28022844e-01
-7.31811047e-01 -3.04880679e-01 3.99177462e-01 -5.32574475e-01
4.35756683e-01 6.72465265e-01 -5.23172021e-01 -7.11839080e-01
-2.40972444e-01 5.55806756e-02 2.37155721e-01 -1.00644186e-01
-2.00335950e-01 6.79549336e-01 -8.82400274e-02 -2.81630993e-01
9.86576676e-01 2.12867737e-01 6.47636116e-01 -8.93263936e-01
-1.07176043e-01 -3.28431875e-02 -1.45789668e-01 -5.53837359e-01
7.82747865e-01 -8.30830216e-01 -1.18850398e+00 9.02114809e-02
-1.15452385e+00 7.66322538e-02 2.24158570e-01 8.11471224e-01
-2.13138144e-02 4.33391541e-01 -1.06713839e-01 -7.07058787e-01
-2.20629439e-01 -1.46136761e+00 2.72245824e-01 6.24784350e-01
2.58391470e-01 -8.63894820e-01 -2.41807386e-01 -1.17833823e-01
4.33996528e-01 4.91137445e-01 8.63178194e-01 -5.65949418e-02
-1.00069594e+00 1.24261297e-01 -3.64832550e-01 1.91727579e-01
4.29440320e-01 2.93203175e-01 9.47153568e-02 -6.08011484e-01
-8.14650729e-02 3.41710836e-01 1.22375228e-01 5.73815405e-01
8.84298980e-01 -9.99889672e-02 -7.30045617e-01 7.07503796e-01
1.93635714e+00 8.73037696e-01 4.24060464e-01 7.41253078e-01
4.66914892e-01 3.63148004e-01 1.53575242e+00 9.24459398e-01
2.95582682e-01 6.99308455e-01 1.08055305e+00 -4.74101976e-02
4.06925678e-01 2.02301115e-01 1.44429848e-01 9.53787804e-01
-2.23707154e-01 -7.45385170e-01 -5.89801133e-01 3.08170795e-01
-1.89022517e+00 -8.37016761e-01 -1.85436215e-02 2.16833568e+00
-4.35262322e-02 2.82663163e-02 6.67746440e-02 1.93258941e-01
1.12561679e+00 1.90065250e-01 -4.31856841e-01 -3.93825442e-01
-3.82080153e-02 -2.02995732e-01 7.87041247e-01 3.14724922e-01
-9.15962875e-01 6.80472374e-01 4.88197041e+00 1.15071023e+00
-1.29364657e+00 1.14924470e-02 2.32699551e-02 -1.45962745e-01
-1.07227191e-01 1.80989243e-02 -6.60711586e-01 9.18326378e-01
5.15152752e-01 -7.78070092e-01 7.28546917e-01 6.16840720e-01
4.97601211e-01 -1.46605149e-01 -9.91297364e-02 9.51794624e-01
-6.82339370e-02 -1.41559899e+00 5.95640764e-02 5.75580955e-01
7.01656997e-01 -3.03310752e-01 -3.06209400e-02 -1.28558889e-01
7.38913119e-02 -5.10104775e-01 2.95193553e-01 1.79292887e-01
7.80990422e-01 -1.39590085e+00 1.00217676e+00 3.19909036e-01
-1.68085742e+00 -3.53804708e-01 -8.31112504e-01 -1.34874538e-01
3.85055631e-01 6.45164788e-01 -5.93706846e-01 1.38335121e+00
5.50154924e-01 6.21685445e-01 6.29494861e-02 1.32329178e+00
2.06598625e-01 -1.41094804e-01 -2.88146734e-01 -4.02235627e-01
2.39088118e-01 -8.13320518e-01 9.09886301e-01 1.29003957e-01
9.06671822e-01 1.99223995e-01 5.00530124e-01 4.49846804e-01
6.55867532e-02 3.10034484e-01 -5.01168072e-01 -1.90313578e-01
1.15578902e+00 1.30328465e+00 -8.89504135e-01 4.42469753e-02
-3.19651723e-01 6.80585444e-01 -2.96724558e-01 1.07974082e-01
-9.74365234e-01 -1.15827429e+00 8.84712398e-01 -1.02992378e-01
3.09255481e-01 -3.02228183e-01 6.73674941e-02 -5.86174965e-01
-1.13660470e-01 -6.29935026e-01 1.48019642e-01 -6.53649688e-01
-6.02263033e-01 8.81992936e-01 -7.71245137e-02 -2.02466011e+00
2.31175661e-01 -5.11694610e-01 -5.68755686e-01 9.16089058e-01
-1.64543211e+00 -6.48778677e-01 -5.76068938e-01 5.13162315e-01
2.05240414e-01 -5.90830147e-01 7.06210434e-01 5.70784390e-01
-9.55714166e-01 1.36295825e-01 2.47702196e-01 -4.03521597e-01
3.66614699e-01 -4.54284668e-01 -4.67915982e-01 1.01889014e+00
-7.73813725e-01 4.07018006e-01 6.79388821e-01 -7.79478610e-01
-1.42319107e+00 -1.01640594e+00 5.16692638e-01 4.93059158e-01
6.02989435e-01 2.77687550e-01 -1.59752622e-01 2.26376891e-01
3.98826897e-01 -2.12580845e-01 4.30284292e-01 -4.52682257e-01
8.62452745e-01 -4.79541123e-01 -1.40562844e+00 5.61782479e-01
7.78650105e-01 2.88028955e-01 1.27909601e-01 3.74883026e-01
1.07343328e+00 -3.17538589e-01 -1.05159307e+00 4.98841435e-01
6.85328245e-02 -6.51124537e-01 8.63511086e-01 4.59461510e-02
-7.82121345e-02 -9.76149738e-01 -6.73396662e-02 -1.60878325e+00
-5.54205120e-01 -6.50320590e-01 3.19238245e-01 1.24503899e+00
9.21469629e-02 -1.13591349e+00 6.37503624e-01 2.53335666e-02
-1.03756398e-01 -9.41642463e-01 -9.82189536e-01 -9.85918462e-01
-6.59170449e-01 3.67728323e-01 1.19047713e+00 7.64380813e-01
-7.97288194e-02 2.06997827e-01 -2.78996736e-01 5.39136469e-01
6.11329556e-01 2.97699988e-01 6.66179597e-01 -1.67116058e+00
-5.19600697e-03 -1.62654534e-01 -5.18633962e-01 -8.97943914e-01
1.91867068e-01 -6.10261023e-01 -4.16978180e-01 -1.86511433e+00
-3.26148659e-01 -8.62241447e-01 -3.59911114e-01 -6.08766153e-02
1.10269956e-01 -3.34727764e-01 2.63236910e-01 2.84907162e-01
-5.96738398e-01 7.19850838e-01 1.62568963e+00 6.96525201e-02
-2.43676335e-01 1.46773234e-01 -6.13869667e-01 3.83097172e-01
9.82261896e-01 -3.37901354e-01 -8.29649925e-01 -4.51162934e-01
3.61163281e-02 5.00390947e-01 -1.27663324e-02 -1.08228326e+00
2.79002219e-01 -5.71410358e-01 1.31670371e-01 -6.10003948e-01
5.27195156e-01 -1.47775841e+00 3.32212836e-01 8.37994456e-01
6.41849279e-01 5.95839143e-01 2.27827579e-01 7.36047089e-01
-4.19149637e-01 -2.19734818e-01 5.15600085e-01 1.57498926e-01
-1.03566933e+00 5.52489579e-01 -5.28583527e-01 -2.36199424e-01
1.75533843e+00 -3.45816344e-01 -5.38218737e-01 -4.91024256e-02
-2.83720374e-01 7.85774112e-01 6.11921251e-01 1.92708224e-01
7.12986052e-01 -1.34858346e+00 -3.64503860e-01 -2.43516210e-02
-2.18447983e-01 -2.45260477e-01 5.08501768e-01 8.98834527e-01
-1.10934854e+00 3.71378243e-01 -6.22385502e-01 -3.21753144e-01
-1.35349262e+00 4.97680217e-01 3.12282681e-01 -2.04859823e-01
7.66559318e-02 7.76716828e-01 -3.75000536e-01 -1.37555525e-01
-2.20238760e-01 2.36966386e-01 -3.48817855e-01 -1.33117773e-02
1.65683940e-01 8.69064152e-01 -2.06216276e-01 -6.35077953e-01
-6.55576587e-01 7.37254739e-01 2.42709696e-01 9.32446346e-02
1.22446871e+00 -4.33040738e-01 -7.87749350e-01 -5.08660316e-01
9.03793573e-01 -5.94497425e-03 -6.48853242e-01 2.92613059e-01
-4.93038327e-01 -7.89408028e-01 5.31851947e-01 -4.23804969e-01
-1.33370590e+00 3.26425016e-01 4.79604900e-01 3.98030490e-01
1.52232480e+00 -7.17860401e-01 8.75367105e-01 5.28552122e-02
9.83135879e-01 -1.16792691e+00 -2.37115353e-01 3.13832551e-01
2.90059119e-01 -7.29584873e-01 4.32576269e-01 -8.57987404e-01
-4.97890443e-01 1.19982064e+00 9.52734888e-01 -8.42545778e-02
8.25284779e-01 1.64651901e-01 -1.64874434e-01 -9.37826559e-02
-5.57967782e-01 -2.33206883e-01 -3.91950876e-01 5.71626902e-01
-2.76843369e-01 1.74368307e-01 -8.50745857e-01 4.19538319e-01
1.50206253e-01 3.32630686e-02 7.33134806e-01 9.20894861e-01
-7.43149936e-01 -1.47563696e+00 -4.59786445e-01 3.89825791e-01
-8.10293630e-02 2.67638713e-01 1.90905228e-01 8.58387113e-01
5.31351328e-01 1.23761344e+00 9.78081375e-02 -8.58318210e-01
2.52866536e-01 -7.56325841e-01 1.80731848e-01 -3.95188570e-01
-1.12085447e-01 2.42935847e-02 8.95463228e-02 -1.62872300e-01
-4.83053178e-01 -2.95003861e-01 -1.34060562e+00 -7.08065867e-01
-6.02248013e-01 7.43991613e-01 6.16085887e-01 5.55975020e-01
5.24869204e-01 1.00355566e+00 9.53459680e-01 -6.88244581e-01
-2.04341799e-01 -2.86468893e-01 -9.61626709e-01 -5.48610628e-01
-1.00214332e-01 -1.16788888e+00 -3.98938745e-01 -7.18590915e-01]
|
[5.877322673797607, 1.713700532913208]
|
444d0599-7610-4409-be5e-140a399d0686
|
gapoera-application-programming-interface-for
|
2110.11924
| null |
https://arxiv.org/abs/2110.11924v1
|
https://arxiv.org/pdf/2110.11924v1.pdf
|
Gapoera: Application Programming Interface for AI Environment of Indonesian Board Game
|
Currently, the development of computer games has shown a tremendous surge. The ease and speed of internet access today have also influenced the development of computer games, especially computer games that are played online. Internet technology has allowed computer games to be played in multiplayer mode. Interaction between players in a computer game can be built in several ways, one of which is by providing balanced opponents. Opponents can be developed using intelligent agents. On the other hand, research on developing intelligent agents is also growing rapidly. In computer game development, one of the easiest ways to measure the performance of an intelligent agent is to develop a virtual environment that allows the intelligent agent to interact with other players. In this research, we try to develop an intelligent agent and virtual environment for the board game. To be easily accessible, the intelligent agent and virtual environment are then developed into an Application Programming Interface (API) service called Gapoera API. The Gapoera API service that is built is expected to help game developers develop a game without having to think much about the artificial intelligence that will be embedded in the game. This service provides a basic multilevel intelligent agent that can provide users with playing board games commonly played in Indonesia. Although the Gapoera API can be used for various types of games, in this paper, we will focus on the discussion on a popular traditional board game in Indonesia, namely Mancala. The test results conclude that the multilevel agent concept developed has worked as expected. On the other hand, the development of the Gapoera API service has also been successfully accessed on several game platforms.
|
['Galang Prihadi Mahardhika', 'Rian Adam Rajagede']
|
2021-10-22
| null | null | null | null |
['board-games']
|
['playing-games']
|
[-5.59602380e-01 -7.88319930e-02 3.90794396e-01 1.70730859e-01
1.36138275e-01 -6.81396663e-01 1.78882569e-01 -1.62165090e-01
-6.20975614e-01 6.36743426e-01 -4.78957593e-01 -7.12751567e-01
-1.52505636e-01 -1.34809709e+00 2.89642299e-03 -5.33657134e-01
1.33333296e-01 7.22082257e-01 6.47151113e-01 -8.92088711e-01
2.42256552e-01 2.80565232e-01 -1.76378369e+00 -1.45824879e-01
7.29299664e-01 5.10279775e-01 7.74320126e-01 9.75530624e-01
-1.02621555e-01 1.14638042e+00 -9.58393097e-01 -3.40762407e-01
4.71982539e-01 -4.78991896e-01 -5.93912005e-01 -1.76713914e-01
-8.22075844e-01 -5.32303333e-01 2.15928271e-01 9.70149875e-01
5.27293324e-01 3.10740769e-01 3.03784162e-01 -1.59036791e+00
2.49126822e-01 2.88145214e-01 -3.21607649e-01 5.51945157e-02
5.92634380e-01 1.77567393e-01 7.03245223e-01 -1.71033591e-01
6.42974079e-01 8.45374703e-01 2.64590561e-01 5.23369431e-01
-4.64532852e-01 -9.84733760e-01 -3.90366256e-01 2.94540584e-01
-1.22755706e+00 3.69306892e-01 6.28397405e-01 -2.64709145e-01
1.12422776e+00 4.52678621e-01 1.11208832e+00 1.15259334e-01
4.39181030e-01 2.41740376e-01 1.13084733e+00 -7.37455070e-01
4.37724054e-01 4.75983322e-01 -2.71391682e-02 4.67581689e-01
3.75494957e-01 -1.12333797e-01 1.14220880e-01 7.30418647e-03
1.06554246e+00 -4.47625339e-01 1.99042916e-01 -9.50722117e-03
-2.43796706e-01 1.17008114e+00 1.22179888e-01 4.82609749e-01
-5.96769154e-01 -2.46187478e-01 5.20574450e-01 2.98602432e-01
5.58118820e-02 6.36935413e-01 -2.79679298e-01 -1.06904733e+00
-1.65469006e-01 5.55716395e-01 1.27535880e+00 2.12605000e-01
4.28434074e-01 2.29871422e-01 9.54479992e-01 9.84303653e-01
6.96840644e-01 2.34138921e-01 6.13174260e-01 -9.18582082e-01
7.99761415e-02 9.72807109e-01 -1.35224871e-03 -1.13002324e+00
-4.06139284e-01 9.33246166e-02 -8.19961876e-02 1.28571188e+00
4.71012115e-01 -5.34556925e-01 -6.47280574e-01 1.25179076e+00
4.30057377e-01 -1.66772246e-01 4.05203342e-01 6.54691756e-01
1.13454807e+00 8.97981822e-01 7.32845515e-02 1.94608286e-01
1.50021243e+00 -9.10036266e-01 -6.25838399e-01 -5.24026081e-02
6.21482968e-01 -7.65609384e-01 9.95798707e-01 6.04555130e-01
-1.19035792e+00 -2.80028045e-01 -1.24516892e+00 5.43289304e-01
-5.98926425e-01 -5.75877368e-01 8.95285606e-01 1.16630983e+00
-1.15748918e+00 4.79760766e-02 -8.99973869e-01 -4.79403406e-01
-3.13568652e-01 8.00136387e-01 -3.99759114e-01 2.61673927e-01
-1.28575921e+00 1.31708515e+00 5.43590486e-01 -3.83010119e-01
-2.84222484e-01 -3.40872705e-02 -8.40294719e-01 -4.53172177e-02
3.41864139e-01 -4.07496303e-01 1.24947822e+00 -1.39810038e+00
-1.91981089e+00 6.76182985e-01 5.92324138e-01 -1.53849512e-01
3.20934117e-01 5.38382888e-01 -5.01112819e-01 -5.67066446e-02
3.03398192e-01 2.83706427e-01 -1.47517130e-01 -9.06274021e-01
-1.38419461e+00 -1.68748215e-01 9.49200809e-01 8.02274704e-01
-2.11575739e-02 5.57837844e-01 -5.82148433e-01 -7.70955086e-02
-3.70767266e-01 -7.27924764e-01 -3.01248074e-01 -6.41156077e-01
5.19504726e-01 -2.96317011e-01 7.75086582e-01 -4.46901143e-01
1.25079417e+00 -1.84973073e+00 -2.78267235e-01 4.17836279e-01
1.49178714e-01 5.25556087e-01 1.95074767e-01 6.79330707e-01
8.87038857e-02 5.01170158e-02 5.36093116e-01 6.76138520e-01
8.24838430e-02 3.43080759e-01 4.36599165e-01 -2.18416303e-01
-4.70514238e-01 1.43354714e-01 -7.78068244e-01 -3.03256452e-01
4.04777616e-01 2.94586360e-01 -6.03702009e-01 3.52248326e-02
1.09600678e-01 -8.74077305e-02 -4.69814777e-01 4.14243698e-01
5.82373738e-01 1.14665404e-01 2.47599080e-01 5.51108301e-01
-4.07984197e-01 2.43712381e-01 -1.52516246e+00 9.82049286e-01
-4.81780201e-01 5.76839209e-01 3.02970082e-01 -7.92214155e-01
9.03569877e-01 7.53218055e-01 4.69469637e-01 -8.23142529e-01
3.92084301e-01 4.16104496e-01 8.51445556e-01 -5.62389314e-01
6.64812744e-01 -1.86429083e-01 -6.67807553e-03 7.11620688e-01
-3.38557661e-01 -4.36162472e-01 7.54744411e-01 1.58496976e-01
1.06045187e+00 7.81948715e-02 5.81165910e-01 -5.59725007e-03
5.11605918e-01 5.02698362e-01 4.61479515e-01 5.63149452e-01
-1.35374039e-01 -1.92348942e-01 4.59397316e-01 -4.07293886e-01
-9.38160062e-01 -8.74227643e-01 2.50937432e-01 1.08222902e+00
3.35899860e-01 -5.28472126e-01 -8.41672421e-01 -1.20244347e-01
-5.46816349e-01 5.40080607e-01 1.91168673e-02 2.38746598e-01
-2.01040953e-01 -5.19641578e-01 4.14661318e-01 4.81794700e-02
8.24645996e-01 -1.47379231e+00 -1.06880963e+00 7.59672999e-01
8.92426521e-02 -7.37934053e-01 4.05099481e-01 -8.15162286e-02
-5.07599115e-01 -1.00968969e+00 -1.90043628e-01 -8.99973691e-01
2.60372698e-01 4.80868638e-01 8.56992245e-01 5.07814407e-01
-1.40541360e-01 3.78045976e-01 -8.11830342e-01 -1.10781264e+00
-5.60315609e-01 -3.50544378e-02 -1.25990838e-01 -8.51847470e-01
6.01483345e-01 -5.66118717e-01 -1.79143652e-01 3.80301237e-01
-9.46653724e-01 5.08361757e-01 1.76806331e-01 4.29627717e-01
-1.44180134e-01 9.72279668e-01 6.48691475e-01 -7.93311059e-01
1.09374809e+00 -3.92333359e-01 -7.96086192e-01 -1.60099626e-01
-3.03286374e-01 -5.94326615e-01 5.99941134e-01 -2.54900426e-01
-1.04734397e+00 -4.17902857e-01 -6.40117347e-01 5.20999134e-01
-1.90735072e-01 1.09271359e+00 -4.01631176e-01 -2.31080621e-01
3.90800595e-01 -1.53373539e-01 4.19035286e-01 6.81517944e-02
-3.23492140e-01 1.12854862e+00 3.57408971e-02 -4.96743202e-01
4.82357025e-01 1.23177528e-01 -3.71697515e-01 -8.26605499e-01
4.67665255e-01 -2.87804276e-01 2.69618154e-01 -8.77877891e-01
8.11869740e-01 -6.43191278e-01 -1.05063415e+00 8.46086144e-01
-7.11420178e-01 -4.53620821e-01 1.66121960e-01 6.39775634e-01
-2.62932628e-01 -1.24654710e-01 -5.33580899e-01 -9.72397447e-01
-1.37904242e-01 -1.45479453e+00 -7.56253973e-02 1.10194170e+00
-4.87160116e-01 -1.12944794e+00 4.69976425e-01 6.51888609e-01
3.42983246e-01 1.07540227e-01 6.89983666e-01 -6.77306056e-01
-2.74789333e-01 -5.75608253e-01 1.51188076e-01 1.49655595e-01
4.25132930e-01 3.02293658e-01 -3.78684431e-01 4.80610393e-02
2.21501794e-02 3.27861719e-02 -4.84592438e-01 3.03439468e-01
-1.92963872e-02 1.80789992e-01 -1.14409760e-01 5.35865650e-02
1.57426882e+00 1.49344444e+00 1.10218382e+00 1.34120023e+00
1.07202232e-01 4.47670162e-01 9.96896446e-01 3.32296789e-01
6.13979340e-01 7.55299032e-01 4.84988391e-01 -3.49740833e-01
4.44568574e-01 7.74631053e-02 4.64123011e-01 7.15566814e-01
-7.62031257e-01 -5.35372309e-02 -9.87275243e-01 1.01166986e-01
-1.75793910e+00 -1.14598536e+00 -2.05054283e-01 1.92443681e+00
4.29536819e-01 3.30316991e-01 5.78306377e-01 3.07734191e-01
5.89747608e-01 -3.27477604e-01 3.08501925e-02 -1.27108443e+00
3.51077348e-01 7.77325273e-01 2.01887578e-01 6.94268882e-01
-7.22449422e-01 1.02558732e+00 4.90200377e+00 8.48998666e-01
-1.31449389e+00 1.88148141e-01 2.41788715e-01 6.21509142e-02
2.08141163e-01 8.63531604e-02 -3.27128798e-01 4.70434368e-01
6.78469360e-01 -7.62132764e-01 5.32859683e-01 1.04851127e+00
7.44493842e-01 -8.22829068e-01 -8.21816921e-02 8.42301488e-01
-3.60344589e-01 -1.34986997e+00 -2.09078401e-01 4.55556154e-01
4.43724930e-01 -3.03965628e-01 -2.61883885e-01 4.18297589e-01
7.42730618e-01 -8.22948098e-01 4.77451980e-01 -2.42976636e-01
1.07210018e-01 -1.24824262e+00 1.18378091e+00 3.83095175e-01
-1.34272826e+00 2.14453060e-02 -2.29796618e-01 -1.14005804e+00
1.80248488e-02 -4.94069159e-01 -8.75119805e-01 5.29937446e-01
8.10647011e-01 -1.72258690e-01 -1.60812661e-01 1.44535971e+00
-2.11970564e-02 4.55886543e-01 -3.68271440e-01 -4.59785551e-01
5.55881441e-01 -7.43202388e-01 4.02554274e-01 4.76821601e-01
2.34028131e-01 5.44671535e-01 -9.09445509e-02 2.86422104e-01
6.27396584e-01 5.69622993e-01 -6.63140297e-01 -5.23468032e-02
4.73462105e-01 1.26521492e+00 -1.21382415e+00 -1.73888654e-01
-7.01849401e-01 7.39002168e-01 -2.32204556e-01 -5.10227494e-02
-8.98992419e-01 -8.65081012e-01 8.66367936e-01 2.64231771e-01
-6.79971352e-02 -8.46436806e-03 -1.71563961e-02 -5.11189401e-01
-4.02029663e-01 -1.59387529e+00 2.17599273e-01 -9.33020592e-01
-6.45618141e-01 6.77979589e-01 2.63878703e-02 -8.97490144e-01
-6.36692047e-01 -7.82974541e-01 -1.10732532e+00 9.84173000e-01
-4.15531814e-01 -9.10726726e-01 -1.74172565e-01 4.97709811e-01
6.59570873e-01 -5.69212496e-01 1.00937760e+00 2.28235796e-01
-5.27758718e-01 1.22332871e-01 6.26826659e-02 2.74138927e-01
1.21768318e-01 -9.16208863e-01 -1.78852960e-01 7.52668262e-01
-1.69690758e-01 4.02864754e-01 7.68003583e-01 -6.85045838e-01
-9.87677872e-01 1.06922928e-02 3.52861226e-01 1.75061226e-01
6.51306033e-01 -6.80808350e-03 -3.47954154e-01 3.50361824e-01
3.98112118e-01 -9.11566377e-01 8.27203870e-01 -8.76858681e-02
4.28451151e-01 2.22164462e-03 -1.32896686e+00 9.61431503e-01
4.59561162e-02 -1.74894258e-01 -3.27712476e-01 -3.40437412e-01
-7.56808892e-02 -4.40105259e-01 -4.06462431e-01 -1.98478371e-01
8.01391602e-01 -1.18038929e+00 5.91917217e-01 -2.90212840e-01
2.25802898e-01 -6.11541033e-01 1.43028408e-01 -1.35644531e+00
-2.20587868e-02 -4.55500692e-01 9.54168200e-01 1.08613598e+00
5.66835582e-01 -1.20383656e+00 1.12240481e+00 1.35868263e+00
1.88878179e-01 -3.10101211e-01 -6.83214068e-01 -5.38732588e-01
-3.05306204e-02 -7.89837062e-01 6.28627956e-01 6.69853091e-01
7.59055912e-01 1.96466357e-01 1.25215296e-02 -9.76642687e-03
1.16013132e-01 -5.12401104e-01 1.02509749e+00 -1.18639660e+00
-6.80555940e-01 -5.28969646e-01 -1.14485037e+00 -3.37175280e-01
-3.30766618e-01 -3.38426590e-01 -3.27702850e-01 -1.77792990e+00
1.48016140e-01 -4.57602799e-01 2.80053437e-01 2.26748988e-01
-4.43857498e-02 3.07271063e-01 5.99232554e-01 -1.02505103e-01
-6.80912426e-03 -4.20805752e-01 1.11922324e+00 2.68013239e-01
-5.61438560e-01 2.38807872e-01 -7.39490986e-01 9.62883174e-01
1.02143836e+00 -2.48512357e-01 -6.51477337e-01 7.69841447e-02
6.71673119e-01 2.27494568e-01 -1.41527653e-01 -1.02595174e+00
3.80211949e-01 -3.35975528e-01 -1.22983836e-01 -9.55697373e-02
4.18839127e-01 -1.08607149e+00 7.71160543e-01 6.95930541e-01
4.80067402e-01 4.06017631e-01 4.22556221e-01 -3.37349027e-01
-2.83226430e-01 -8.89866829e-01 4.40207660e-01 -4.53687191e-01
-9.55920696e-01 -2.90735483e-01 -1.42432058e+00 -3.32354188e-01
1.60997617e+00 -9.55853581e-01 5.27529418e-03 -9.87721562e-01
-6.26417637e-01 1.55406341e-01 7.07834542e-01 6.38602600e-02
3.60613078e-01 -9.98853505e-01 -3.15175444e-01 8.84568971e-03
-3.36282641e-01 -4.54152495e-01 1.86096877e-01 2.95935690e-01
-1.70366764e+00 1.49177745e-01 -9.96826947e-01 2.14625672e-01
-1.86858499e+00 2.92793848e-02 3.82701576e-01 -5.74585319e-01
-3.83936465e-01 5.25632501e-01 9.26919505e-02 -1.99083999e-01
-1.84930727e-01 3.51526052e-01 -7.79064357e-01 -4.60405461e-02
8.14474881e-01 3.83645058e-01 -1.85458556e-01 -6.20721638e-01
-2.98170984e-01 1.23265721e-01 -5.77195808e-02 -8.01520526e-01
1.51065469e+00 8.86390656e-02 -3.83491099e-01 2.17106298e-01
3.49999607e-01 1.54172227e-01 -6.79997385e-01 7.48766065e-01
-3.13543171e-01 -6.47634804e-01 2.42555630e-03 -8.95771623e-01
-1.04118276e+00 1.83848396e-01 6.41997635e-01 8.19076240e-01
1.23011744e+00 -4.49487865e-01 3.17573547e-01 -4.95366044e-02
8.87401760e-01 -1.36736548e+00 -3.40629369e-01 5.63886285e-01
3.63914728e-01 -8.50658894e-01 -2.81176656e-01 -2.50115782e-01
-9.27219391e-01 1.21788979e+00 8.98261547e-01 -1.20606758e-01
6.79018497e-01 8.40915203e-01 6.08982265e-01 -2.50400513e-01
-6.01227760e-01 -4.32508379e-01 -5.86023390e-01 1.10451210e+00
4.33793455e-01 2.86224604e-01 -9.02965844e-01 8.02185237e-01
-5.55185795e-01 2.91191280e-01 1.15662456e+00 1.07278848e+00
-5.61076880e-01 -1.60359454e+00 -8.06712329e-01 2.95488775e-01
-7.55029619e-01 -7.15226168e-03 -3.94696921e-01 1.36074376e+00
1.89290136e-01 1.39189911e+00 1.96825594e-01 -5.32381833e-01
3.02972794e-01 -4.34913069e-01 3.32967430e-01 -5.69069982e-01
-1.07831407e+00 4.13782615e-03 5.14365017e-01 -2.54275687e-02
-2.20317796e-01 -3.04446161e-01 -1.42443371e+00 -9.88878727e-01
-5.67644417e-01 7.44894087e-01 9.30878997e-01 8.08517814e-01
-2.98675716e-01 4.18145359e-01 2.94499934e-01 -7.88290501e-01
3.01828355e-01 -7.47889161e-01 -9.98508096e-01 -1.30887315e-01
-8.69706154e-01 -6.55892730e-01 -5.03328256e-02 -4.78475392e-01]
|
[3.4727137088775635, 1.4863237142562866]
|
2c17d207-42cd-49ad-b3b7-bb7e0d4eff26
|
actrce-augmenting-experience-via-teachers
|
1902.04546
| null |
http://arxiv.org/abs/1902.04546v1
|
http://arxiv.org/pdf/1902.04546v1.pdf
|
ACTRCE: Augmenting Experience via Teacher's Advice For Multi-Goal Reinforcement Learning
|
Sparse reward is one of the most challenging problems in reinforcement
learning (RL). Hindsight Experience Replay (HER) attempts to address this issue
by converting a failed experience to a successful one by relabeling the goals.
Despite its effectiveness, HER has limited applicability because it lacks a
compact and universal goal representation. We present Augmenting experienCe via
TeacheR's adviCE (ACTRCE), an efficient reinforcement learning technique that
extends the HER framework using natural language as the goal representation. We
first analyze the differences among goal representation, and show that ACTRCE
can efficiently solve difficult reinforcement learning problems in challenging
3D navigation tasks, whereas HER with non-language goal representation failed
to learn. We also show that with language goal representations, the agent can
generalize to unseen instructions, and even generalize to instructions with
unseen lexicons. We further demonstrate it is crucial to use hindsight advice
to solve challenging tasks, and even small amount of advice is sufficient for
the agent to achieve good performance.
|
['Sanja Fidler', 'Jimmy Ba', 'Yuhuai Wu', 'Jamie Kiros', 'Harris Chan']
|
2019-02-12
| null | null | null | null |
['multi-goal-reinforcement-learning']
|
['methodology']
|
[-1.32594451e-01 4.52181041e-01 3.95815521e-02 -1.37948647e-01
-8.27545464e-01 -7.80971110e-01 4.23899531e-01 -3.32849286e-02
-6.52958572e-01 1.11839151e+00 2.58176118e-01 -4.70603228e-01
-1.76311627e-01 -6.61562622e-01 -6.28661931e-01 -6.57373130e-01
-3.28565389e-01 4.83504385e-01 1.25991315e-01 -8.30417097e-01
3.77490461e-01 3.37599039e-01 -1.54025006e+00 -3.85339744e-02
7.93818235e-01 4.30634201e-01 6.70350611e-01 6.94186091e-01
-2.54219353e-01 1.39103675e+00 -5.76807380e-01 1.42557248e-01
5.04532635e-01 -7.13696301e-01 -1.09073997e+00 -4.87562343e-02
7.95898885e-02 -5.39992809e-01 -2.85590738e-01 1.03600645e+00
4.43529576e-01 7.22047865e-01 6.53625667e-01 -1.28889060e+00
-6.66334152e-01 6.22235954e-01 -2.98909426e-01 2.63366885e-02
6.95553958e-01 1.45965487e-01 9.26194131e-01 -4.69367862e-01
5.40413618e-01 1.17191231e+00 4.11100537e-01 1.05891788e+00
-1.05018091e+00 -3.87004942e-01 4.27569568e-01 2.75470316e-02
-8.58821034e-01 -1.51502907e-01 4.32952225e-01 -6.84283301e-02
1.39257741e+00 -1.65995076e-01 7.57238567e-01 9.76763606e-01
1.38187453e-01 9.90957975e-01 1.40464067e+00 -5.68310380e-01
5.86342096e-01 -2.71558493e-01 -4.57887091e-02 9.22749639e-01
-1.23944297e-01 7.70392656e-01 -4.33352470e-01 5.12384772e-02
9.04689550e-01 -5.00398986e-02 -2.45035395e-01 -7.05743492e-01
-9.77469802e-01 1.00915062e+00 6.18030369e-01 2.55522907e-01
-4.47253376e-01 3.64150256e-01 2.65615135e-01 1.15762460e+00
-3.24428529e-01 1.06092787e+00 -6.57067120e-01 -2.94591159e-01
-2.46256858e-01 5.40641367e-01 8.61180425e-01 1.11749053e+00
1.08926129e+00 4.80329335e-01 1.04432285e-01 7.76667893e-01
2.68954560e-02 4.63352412e-01 8.58496666e-01 -1.46147013e+00
2.36438319e-01 2.93769807e-01 2.61755913e-01 -3.92343372e-01
-6.74027920e-01 -4.14436102e-01 -2.58835584e-01 9.47581708e-01
5.19429624e-01 -4.12965298e-01 -9.26653206e-01 2.18029141e+00
1.21350378e-01 -7.09790513e-02 7.66707301e-01 8.35040569e-01
7.86387563e-01 4.78791207e-01 2.09754467e-01 -2.86573738e-01
8.05463791e-01 -1.25764823e+00 -3.90319824e-01 -6.72442675e-01
1.20329821e+00 -2.72920579e-01 1.30368769e+00 5.83132386e-01
-9.44060802e-01 -4.43638325e-01 -1.07599890e+00 -6.48195520e-02
-3.43044221e-01 -3.82421762e-01 1.12936020e+00 3.29187006e-01
-1.22634017e+00 6.16648138e-01 -4.96135771e-01 -4.89995211e-01
1.37819275e-01 4.96966988e-01 -6.09656811e-01 -3.78376931e-01
-1.00903332e+00 1.21338069e+00 3.61714214e-01 -4.56889987e-01
-1.53552878e+00 -1.10568851e-01 -1.28937650e+00 -5.42656239e-03
7.40842164e-01 -6.71173573e-01 1.86038494e+00 -8.07441473e-01
-1.77387989e+00 6.15422428e-01 3.20354015e-01 -5.50392330e-01
6.07754989e-03 -2.16967195e-01 3.55904102e-02 1.98306829e-01
4.12030190e-01 9.58196759e-01 7.49849260e-01 -1.23161411e+00
-7.42829680e-01 -3.61139774e-01 6.02776587e-01 7.34054387e-01
1.41271845e-01 -4.43193078e-01 1.32573336e-01 -4.44839239e-01
6.06718957e-02 -9.81895030e-01 -7.01026201e-01 -4.77505356e-01
2.53471792e-01 -4.65562612e-01 3.60701501e-01 -1.14213817e-01
5.87994158e-01 -2.10101724e+00 3.18854123e-01 4.14091209e-03
2.66092777e-01 -3.55603024e-02 -7.19446898e-01 5.13111591e-01
4.48418409e-02 -3.38528723e-01 -1.05635859e-01 -4.90500219e-03
1.51759699e-01 8.35483432e-01 -4.15111929e-01 2.28736788e-01
-6.88844025e-02 9.42169011e-01 -1.37545490e+00 -1.43132970e-01
3.84403244e-02 -1.30856961e-01 -1.03520739e+00 5.69954991e-01
-5.00367880e-01 5.73078871e-01 -7.08902836e-01 4.41715539e-01
7.00496957e-02 -3.41550000e-02 1.67330042e-01 5.77197552e-01
1.04540803e-01 3.54138345e-01 -1.12731373e+00 2.06280208e+00
-6.67834103e-01 1.78545818e-01 7.01593161e-02 -1.19885886e+00
1.01315057e+00 2.83415675e-01 2.59903669e-01 -9.56674159e-01
9.02843028e-02 2.98338890e-01 1.59265250e-01 -6.15110099e-01
5.68304360e-01 -5.75090945e-01 -4.11740810e-01 8.23580682e-01
4.59989339e-01 -6.38871372e-01 1.91187471e-01 2.66414821e-01
1.39891517e+00 6.07684910e-01 6.96371675e-01 -1.22585081e-01
2.83384740e-01 2.17220053e-01 4.68313754e-01 1.31669927e+00
-3.88401985e-01 3.30862045e-01 3.42711896e-01 -5.01089275e-01
-6.66486442e-01 -1.09593916e+00 5.51629663e-01 1.57406354e+00
1.18799217e-01 -3.09356660e-01 -4.80271906e-01 -1.03514898e+00
-3.76997553e-02 9.36776102e-01 -7.27022171e-01 -2.41554871e-01
-5.71395218e-01 -2.33300686e-01 2.94264883e-01 5.44176161e-01
4.78957564e-01 -1.75830185e+00 -9.35896635e-01 3.66361082e-01
5.60353212e-02 -7.14139700e-01 -1.87270060e-01 9.18700397e-01
-8.25019479e-01 -1.19487572e+00 -7.42398858e-01 -1.15453148e+00
6.22094154e-01 3.91413033e-01 1.26273084e+00 3.04080963e-01
-2.57211756e-02 9.86310422e-01 -7.28390813e-01 -1.84144959e-01
-4.79271173e-01 -5.81431314e-02 2.25637794e-01 -9.49478626e-01
4.28077877e-01 -8.26285601e-01 -2.67900079e-01 1.44813418e-01
-7.76332140e-01 -2.21233502e-01 6.39618099e-01 1.15177286e+00
2.50223160e-01 -1.04980335e-01 1.03660893e+00 -7.92332411e-01
1.06660342e+00 -4.74559307e-01 -5.17768323e-01 -1.11370087e-02
-6.40446961e-01 6.23131692e-01 7.75045216e-01 -3.39617640e-01
-9.19011414e-01 -2.24506464e-02 -4.33861732e-01 6.32331148e-02
-3.81641448e-01 4.61378872e-01 2.43752718e-01 -1.71377972e-01
1.18430710e+00 3.24064821e-01 1.36285558e-01 -3.19740266e-01
6.07610464e-01 2.08484694e-01 4.36019868e-01 -1.09034181e+00
4.71536815e-01 -1.51824467e-02 -6.84886128e-02 -5.82977951e-01
-9.46699023e-01 -2.99069881e-01 -1.83920294e-01 -3.32997516e-02
6.02636635e-01 -8.20095897e-01 -9.34638381e-01 -3.95776480e-02
-7.07462966e-01 -1.11345530e+00 -8.08985293e-01 6.53399646e-01
-1.44174087e+00 1.76355869e-01 -6.29067779e-01 -7.14407265e-01
2.75801569e-02 -1.13969648e+00 6.19323194e-01 2.87098587e-01
-1.41738668e-01 -8.08471203e-01 2.32502818e-01 -1.57984212e-01
3.80872130e-01 -1.08766787e-01 1.02121782e+00 -7.60591745e-01
-3.44196528e-01 3.12812656e-01 9.96184051e-02 1.06657960e-01
1.89324394e-02 -9.91264820e-01 -5.37819028e-01 -4.98965889e-01
2.65187085e-01 -1.21829760e+00 7.13616192e-01 2.44388074e-01
7.09170640e-01 -9.68035981e-02 2.20442172e-02 4.27767277e-01
1.39219868e+00 4.34240401e-01 6.15938425e-01 6.44832790e-01
2.63607115e-01 5.84911346e-01 8.76831710e-01 3.79450440e-01
4.20565069e-01 2.53425360e-01 5.80590427e-01 3.43370348e-01
-7.48235732e-02 -5.40314257e-01 5.74648499e-01 5.92812896e-01
-6.70303702e-02 1.33575380e-01 -6.07495189e-01 4.32543784e-01
-1.84338760e+00 -8.14317107e-01 5.71011424e-01 1.86703229e+00
8.93685639e-01 4.82325815e-03 2.41776049e-01 2.98416195e-03
1.00177184e-01 -5.74192703e-02 -7.31270790e-01 -7.24382997e-01
1.01274535e-01 4.26982820e-01 1.81670934e-01 8.36789548e-01
-5.88169515e-01 1.53466511e+00 7.48412704e+00 5.95935166e-01
-5.98677576e-01 -1.86005086e-02 -1.45618813e-02 2.14525476e-01
-2.16694102e-01 1.04438856e-01 -6.40024364e-01 -3.03767145e-01
4.27785426e-01 -9.91664156e-02 1.00842559e+00 1.10699975e+00
-2.36908853e-01 -2.78435886e-01 -1.25309229e+00 9.02211487e-01
-8.51927251e-02 -8.85879576e-01 -1.12567559e-01 -2.08639652e-01
7.65679061e-01 1.13837607e-01 -6.36454523e-02 1.36622667e+00
1.28180110e+00 -1.23815131e+00 4.11439925e-01 6.05819561e-02
5.35090268e-01 -7.61486769e-01 4.81072575e-01 7.72796810e-01
-7.69483745e-01 -5.14696598e-01 -6.81977510e-01 -7.67506540e-01
-7.75103793e-02 -3.40675443e-01 -8.80650818e-01 2.45478958e-01
4.20071661e-01 5.59211195e-01 -3.14680040e-01 7.57790506e-01
-7.57223010e-01 1.34770960e-01 -2.05741346e-01 -3.08956653e-01
7.65027463e-01 -1.71159655e-01 4.10011888e-01 6.30645573e-01
5.57981491e-01 4.59566563e-01 6.81516886e-01 4.22597557e-01
1.45279229e-01 1.67495266e-01 -1.14111388e+00 4.68891114e-02
1.81302592e-01 8.31639051e-01 -4.77759242e-01 -1.07138269e-01
-3.98049951e-01 1.04557729e+00 9.98320103e-01 6.25501931e-01
-3.24941456e-01 -2.93505758e-01 4.41792190e-01 -3.16398293e-01
3.33684564e-01 -3.74254674e-01 1.74921677e-01 -1.02964234e+00
-5.62793434e-01 -1.28436792e+00 6.09029949e-01 -1.04063404e+00
-1.09621429e+00 7.08145440e-01 -1.35190398e-01 -1.17445958e+00
-8.83344948e-01 -7.63546824e-01 -2.64390677e-01 5.97081721e-01
-1.59455240e+00 -8.98795962e-01 -4.42207791e-02 1.03829312e+00
6.89401031e-01 -6.55495882e-01 1.30034554e+00 -2.90802985e-01
1.12355895e-01 4.09468621e-01 -2.72886634e-01 -9.08875838e-02
5.22615016e-01 -1.77272904e+00 1.53233886e-01 3.41544926e-01
3.35275859e-01 6.07873559e-01 8.96960974e-01 -4.73329961e-01
-1.29931033e+00 -5.90317070e-01 4.16655809e-01 -1.30998790e-01
4.59533155e-01 4.54757223e-03 -5.39475083e-01 1.03789353e+00
3.28698158e-01 -1.46685883e-01 6.44655526e-01 4.35159564e-01
-3.81448478e-01 2.03992859e-01 -1.14306021e+00 1.09051478e+00
1.24598706e+00 -3.53788942e-01 -1.35903525e+00 2.60788560e-01
7.59449065e-01 -4.67339575e-01 -3.96339148e-01 1.48151711e-01
3.08500767e-01 -1.06715095e+00 7.17740178e-01 -1.08209968e+00
3.86108398e-01 -2.91549534e-01 -3.63630146e-01 -1.94350028e+00
-3.44879001e-01 -8.50272119e-01 -1.82574335e-02 5.32761872e-01
2.81032890e-01 -6.72462106e-01 8.89317513e-01 6.80998638e-02
-3.51592630e-01 -6.64171338e-01 -5.94822884e-01 -9.99334514e-01
5.03796935e-01 -5.06395459e-01 5.19968569e-01 7.70316601e-01
5.17021179e-01 6.31454170e-01 -4.86833483e-01 -1.04121707e-01
3.85235369e-01 3.07970315e-01 9.92467463e-01 -8.99061501e-01
-5.74010074e-01 -1.96488738e-01 -1.29955828e-01 -1.71267295e+00
4.33663964e-01 -1.11717629e+00 4.27459180e-01 -1.58725429e+00
-1.94948599e-01 -7.44161129e-01 -2.73538291e-01 7.05747366e-01
7.10798204e-02 4.81024981e-02 2.63837337e-01 -1.18527845e-01
-9.39973414e-01 6.82401001e-01 1.77011395e+00 6.38724118e-02
-4.58502233e-01 8.83633643e-02 -1.08180499e+00 7.95570195e-01
9.88546491e-01 -5.77702701e-01 -7.58388519e-01 -3.81176889e-01
5.54343104e-01 3.64060581e-01 9.24787372e-02 -9.38712656e-01
1.98161677e-01 -2.81525463e-01 1.87687054e-01 -1.71691611e-01
1.85498729e-01 -6.97925806e-01 -5.54966390e-01 6.30597115e-01
-5.69034815e-01 2.76280075e-01 2.24164918e-01 6.08651817e-01
-1.36510357e-01 -7.47012377e-01 5.68758547e-01 -7.83038139e-01
-1.14574957e+00 1.76304579e-01 -7.74504423e-01 4.10700142e-01
9.53662574e-01 1.96815655e-02 -2.43233606e-01 -7.83177376e-01
-1.25497770e+00 5.32778859e-01 4.06041056e-01 1.64380923e-01
8.28463137e-01 -1.36456966e+00 -4.83331472e-01 2.81285346e-01
2.48674273e-01 -6.22440800e-02 6.41726032e-02 3.24768364e-01
-4.53874767e-01 2.21098170e-01 -5.74418128e-01 -2.56187081e-01
-7.92409897e-01 8.78238142e-01 3.40325385e-01 -4.54468071e-01
-8.50340784e-01 8.71507883e-01 3.25717032e-01 -8.65625262e-01
4.22331899e-01 -2.39911780e-01 -4.88925546e-01 -3.42618585e-01
5.60260236e-01 -6.52049333e-02 -1.75377816e-01 -1.07648604e-01
1.12805858e-01 4.59783524e-01 -3.30228359e-01 -3.73431265e-01
1.52229095e+00 -1.79293960e-01 3.54076922e-01 2.15258583e-01
7.33822107e-01 -1.18670970e-01 -1.50570381e+00 -3.05027127e-01
-4.91803885e-02 -2.67019004e-01 -2.25502804e-01 -8.40152264e-01
-5.66882849e-01 7.65317917e-01 3.58412713e-01 2.90503234e-01
1.02142370e+00 8.38642474e-03 5.98672450e-01 1.26929200e+00
8.41788828e-01 -1.14799738e+00 6.74570441e-01 1.02603030e+00
1.07360244e+00 -1.26447427e+00 -1.40786260e-01 1.34004578e-01
-9.17456925e-01 9.78851855e-01 8.39571059e-01 -4.72000986e-01
2.84883618e-01 1.68312997e-01 3.01567763e-01 -2.99750328e-01
-7.82042444e-01 -7.68942297e-01 -3.52427423e-01 1.14316535e+00
-1.79796549e-03 -1.31564066e-01 -2.09589183e-01 5.41619778e-01
-5.71558833e-01 -3.07761226e-02 6.19326890e-01 1.23969996e+00
-8.83844614e-01 -1.21528637e+00 -1.50429055e-01 2.46857241e-01
-2.99949706e-01 -1.10574752e-01 -3.27926353e-02 8.97789061e-01
-1.41779125e-01 9.76784527e-01 -2.76953787e-01 -3.18618089e-01
2.21451715e-01 4.06776294e-02 1.09187055e+00 -1.15124297e+00
-6.30243838e-01 -1.81869552e-01 -1.28811533e-02 -7.21253574e-01
-2.75580019e-01 -3.81874353e-01 -1.70690644e+00 -9.68616158e-02
9.39158052e-02 4.60778445e-01 2.66484231e-01 1.17981946e+00
-3.14871855e-02 4.32371020e-01 5.89743972e-01 -6.52108133e-01
-1.03143942e+00 -7.69557416e-01 -9.03577507e-01 2.81598419e-01
6.07098520e-01 -9.38214302e-01 -4.86735821e-01 -3.05654317e-01]
|
[4.091815948486328, 1.4609720706939697]
|
74f558b9-884f-419b-a0ec-646d48963fe1
|
video-cloze-procedure-for-self-supervised
|
2001.00294
| null |
https://arxiv.org/abs/2001.00294v1
|
https://arxiv.org/pdf/2001.00294v1.pdf
|
Video Cloze Procedure for Self-Supervised Spatio-Temporal Learning
|
We propose a novel self-supervised method, referred to as Video Cloze Procedure (VCP), to learn rich spatial-temporal representations. VCP first generates "blanks" by withholding video clips and then creates "options" by applying spatio-temporal operations on the withheld clips. Finally, it fills the blanks with "options" and learns representations by predicting the categories of operations applied on the clips. VCP can act as either a proxy task or a target task in self-supervised learning. As a proxy task, it converts rich self-supervised representations into video clip operations (options), which enhances the flexibility and reduces the complexity of representation learning. As a target task, it can assess learned representation models in a uniform and interpretable manner. With VCP, we train spatial-temporal representation models (3D-CNNs) and apply such models on action recognition and video retrieval tasks. Experiments on commonly used benchmarks show that the trained models outperform the state-of-the-art self-supervised models with significant margins.
|
['Can Ma', 'Yu Zhou', 'Chang Liu', 'Qixiang Ye', 'Dongbao Yang', 'Dezhao Luo', 'Weiping Wang']
|
2020-01-02
| null | null | null | null |
['self-supervised-action-recognition']
|
['computer-vision']
|
[ 2.85385847e-01 -1.25312909e-01 -7.39422739e-01 -5.01644671e-01
-6.73108816e-01 -4.20781583e-01 6.58474684e-01 -1.53120264e-01
-7.18110204e-02 3.50128889e-01 5.76925814e-01 -3.28681134e-02
1.84211716e-01 -6.16366386e-01 -9.64593649e-01 -5.02802730e-01
-1.09049082e-01 2.81203896e-01 1.42890483e-01 6.64658099e-02
2.41834164e-01 5.29224724e-02 -1.58554828e+00 9.85833824e-01
7.42329001e-01 1.14074886e+00 1.62401423e-01 3.43369097e-01
-1.69514731e-01 1.48529732e+00 -4.68006939e-01 4.83078659e-02
6.30970374e-02 -4.87473458e-01 -7.80113339e-01 4.09348100e-01
3.06674838e-01 -6.12714410e-01 -8.76524627e-01 7.46553659e-01
1.13884076e-01 3.16027194e-01 8.42533112e-01 -1.18899322e+00
-1.14175665e+00 5.68903089e-01 -3.99390370e-01 2.57052004e-01
7.51032412e-01 3.13670784e-01 1.06946945e+00 -1.00700021e+00
7.59653926e-01 1.15657151e+00 3.39432567e-01 8.47362936e-01
-1.15355718e+00 -7.54118502e-01 3.26752067e-01 3.66757661e-01
-1.28538728e+00 -5.57517231e-01 9.83332038e-01 -6.55722380e-01
7.89364100e-01 1.68058872e-01 5.78300893e-01 1.48802567e+00
-2.05602750e-01 1.42035854e+00 8.03517938e-01 -1.52233049e-01
4.58978623e-01 -8.87739658e-02 2.07385924e-02 5.49431622e-01
-3.61499578e-01 1.04764305e-01 -7.21703827e-01 1.52480081e-01
9.87536132e-01 5.28279841e-01 -2.87663758e-01 -6.04439974e-01
-1.35123575e+00 7.17602372e-01 6.63469732e-01 2.11477697e-01
-2.58272499e-01 9.18052793e-02 5.53633213e-01 3.17439944e-01
6.81829333e-01 4.67897296e-01 -6.80176839e-02 -6.14812709e-02
-1.10857463e+00 2.15480790e-01 2.55550504e-01 1.19956195e+00
7.08506882e-01 2.38774955e-01 -6.49617195e-01 7.64453351e-01
-1.42365471e-01 3.02863508e-01 8.03826809e-01 -7.24407434e-01
6.67365611e-01 8.01506698e-01 1.85046256e-01 -9.06844616e-01
-1.62002653e-01 -1.17272437e-01 -8.06831658e-01 -6.79299459e-02
1.04890286e-03 1.82016984e-01 -1.14380944e+00 1.53998172e+00
-1.27371997e-01 4.59459662e-01 -8.74286488e-05 1.06207585e+00
1.17651546e+00 9.64895010e-01 2.23433599e-01 -2.35668272e-01
1.10534215e+00 -1.47044253e+00 -9.18013632e-01 -3.27201396e-01
6.64775074e-01 -1.80255622e-01 1.33095992e+00 2.60540783e-01
-1.12913036e+00 -7.37493038e-01 -9.59001899e-01 -9.50144902e-02
-2.05252945e-01 3.41851890e-01 6.28689647e-01 4.65038829e-02
-9.00438786e-01 8.15979719e-01 -8.35701108e-01 -2.35147238e-01
8.39920521e-01 -2.76481267e-04 -5.51546574e-01 -2.46245101e-01
-1.02057683e+00 5.49111307e-01 1.80859849e-01 -2.82411456e-01
-1.53078640e+00 -7.06439078e-01 -1.21047354e+00 1.07778579e-01
5.18751264e-01 -3.15903217e-01 1.17859304e+00 -1.36059737e+00
-1.48155606e+00 1.03365564e+00 -2.25688487e-01 -4.69093204e-01
5.88045478e-01 -3.85763377e-01 -6.37751222e-01 5.07146299e-01
3.12573195e-01 8.11537325e-01 1.29073453e+00 -1.02995527e+00
-3.80489111e-01 -1.23712339e-01 2.67864525e-01 7.84591958e-02
-3.79369467e-01 1.32293124e-02 -6.97491884e-01 -1.03035748e+00
1.32237256e-01 -9.11830604e-01 6.86168969e-02 3.22948545e-01
-3.79773945e-01 -2.19784528e-01 7.65383840e-01 -4.22709048e-01
1.44057548e+00 -2.62420368e+00 3.72375637e-01 -2.32044518e-01
1.51118025e-01 2.41033986e-01 -4.81813818e-01 3.67468894e-01
-4.98472333e-01 6.48539513e-02 -8.32637697e-02 -4.95309293e-01
-1.47299305e-01 2.68218756e-01 -6.18238866e-01 4.52268481e-01
4.90439981e-01 9.22399163e-01 -1.19614518e+00 -3.13084096e-01
2.96560735e-01 2.21853599e-01 -7.85426915e-01 6.50148869e-01
-4.98859137e-01 5.13826609e-01 -5.02973258e-01 7.42452919e-01
4.01947588e-01 -5.17470777e-01 5.08599654e-02 -2.16542438e-01
-4.06018645e-02 4.60195273e-01 -7.57896245e-01 1.99834239e+00
-2.75482714e-01 6.65034950e-01 -5.65796137e-01 -1.19895351e+00
8.51505458e-01 2.35490441e-01 3.80210996e-01 -9.00649905e-01
-7.08703846e-02 -1.89495757e-01 -6.71033859e-01 -9.09625411e-01
4.87315327e-01 1.89001963e-01 -1.58894822e-01 3.85575622e-01
2.49279365e-01 2.34066725e-01 6.64186925e-02 3.70837539e-01
1.19903946e+00 5.45829654e-01 3.31013918e-01 -1.96332615e-02
3.42686534e-01 1.02731679e-03 4.76156265e-01 5.11404395e-01
-1.72346890e-01 8.72680724e-01 5.54660380e-01 -6.77025497e-01
-7.45428264e-01 -1.13372004e+00 2.56067842e-01 1.47364616e+00
1.14226751e-01 -6.99577272e-01 -4.92100388e-01 -8.64092708e-01
-3.64192128e-02 5.88917553e-01 -9.58255291e-01 -3.68739218e-01
-4.73035723e-01 1.54894844e-01 1.00655392e-01 1.04172468e+00
5.29906392e-01 -1.44024670e+00 -5.41221380e-01 -1.01328485e-01
-1.96657822e-01 -1.08855689e+00 -6.08880639e-01 1.32209212e-01
-8.12432170e-01 -1.04215527e+00 -8.11565816e-01 -9.23277795e-01
9.38445210e-01 7.06027269e-01 1.00237465e+00 3.46535072e-02
1.45315304e-02 2.45307028e-01 -9.16115999e-01 9.37541872e-02
-7.05048293e-02 -2.28167221e-01 6.45700172e-02 2.81543642e-01
3.20726275e-01 -4.87643659e-01 -7.82060087e-01 3.42483103e-01
-1.11074388e+00 2.86333948e-01 3.56723070e-01 9.18624818e-01
7.09918916e-01 -3.06030989e-01 3.54357898e-01 -1.14938521e+00
3.26971024e-01 -7.35619426e-01 -2.27073699e-01 2.19863892e-01
-2.08289083e-02 9.75068361e-02 8.05001080e-01 -7.37829328e-01
-7.91941822e-01 1.78713039e-01 2.75278270e-01 -1.17405343e+00
-1.67204037e-01 5.87010503e-01 -1.80136144e-01 3.23021680e-01
7.45446920e-01 4.96072590e-01 -4.60664071e-02 -5.33500075e-01
4.42920625e-01 5.87763250e-01 7.71601260e-01 -4.29890066e-01
7.84855604e-01 6.00926518e-01 -5.27746320e-01 -5.08433700e-01
-1.06695271e+00 -3.68915349e-01 -6.07686639e-01 -2.02122197e-01
1.00621474e+00 -1.26338172e+00 -2.89756775e-01 3.09214294e-01
-9.67112422e-01 -6.11212730e-01 -2.73246646e-01 2.77300835e-01
-7.77420104e-01 1.54973790e-01 -4.23477411e-01 -5.97336292e-01
1.15861677e-01 -9.54178751e-01 1.14575362e+00 1.40076101e-01
-4.59786803e-01 -7.51494527e-01 -2.05861926e-01 4.16611999e-01
2.26631880e-01 3.11337560e-01 8.54254007e-01 -6.29778564e-01
-6.71941280e-01 -3.14739764e-01 -1.58517897e-01 3.71134520e-01
1.80353314e-01 -1.05728485e-01 -9.23218548e-01 -2.52942383e-01
-3.09888184e-01 -6.76206768e-01 1.09494293e+00 1.20812848e-01
1.85219693e+00 -5.69861174e-01 -2.95955956e-01 8.90600741e-01
1.00414026e+00 1.38563290e-01 8.95787179e-01 3.33261818e-01
6.60086155e-01 4.05788332e-01 6.57432795e-01 5.81021905e-01
4.23853457e-01 7.29009748e-01 4.79610860e-01 3.72674852e-03
-1.24322966e-01 -8.56296301e-01 6.04141891e-01 6.06060147e-01
4.43516858e-02 -9.64618325e-02 -6.37292862e-01 5.81451893e-01
-2.11685061e+00 -1.25601268e+00 3.82494986e-01 1.97206974e+00
7.61217356e-01 7.42264763e-02 2.25063443e-01 1.61149278e-01
7.20398366e-01 7.43350744e-01 -7.86163688e-01 -4.04189080e-02
9.52649191e-02 1.16902448e-01 6.00585006e-02 -4.13157903e-02
-1.48200536e+00 1.18719661e+00 6.39818430e+00 6.92458808e-01
-1.23584676e+00 7.14749191e-03 4.54104692e-01 -3.97001088e-01
-2.23252237e-01 -1.54302761e-01 -3.16917211e-01 7.45625198e-01
5.65041840e-01 -5.22371605e-02 5.14833510e-01 1.06035340e+00
2.95475781e-01 1.73654050e-01 -1.70361042e+00 1.26259601e+00
3.92604321e-01 -1.60688424e+00 4.38159823e-01 -5.38032293e-01
7.82754600e-01 -3.59832525e-01 2.87423491e-01 6.15437329e-01
1.29793733e-01 -1.10409915e+00 9.45742011e-01 5.10684192e-01
1.12917972e+00 -2.39162102e-01 1.62860408e-01 1.84514984e-01
-1.17035675e+00 -3.16807479e-01 -3.27174991e-01 -2.83034921e-01
-8.73471517e-03 -5.47305914e-04 -1.30477384e-01 2.29493499e-01
7.60476947e-01 1.40168512e+00 -5.83662152e-01 8.87412250e-01
-5.85656166e-01 6.12445474e-01 2.37010479e-01 1.50395408e-01
3.06830794e-01 -3.74101289e-02 2.48993382e-01 1.14142978e+00
1.88171983e-01 3.18492591e-01 2.39387944e-01 9.27965343e-01
-2.19464943e-01 -9.28852335e-02 -6.37309790e-01 -4.14387614e-01
6.12771988e-01 7.74803698e-01 -2.60172248e-01 -4.66264874e-01
-4.50166345e-01 1.21132874e+00 5.73621154e-01 5.99792719e-01
-8.23242128e-01 -1.16146058e-01 5.20737946e-01 3.65128487e-01
5.67419767e-01 -1.19393365e-03 1.24028593e-01 -1.61260307e+00
1.63491905e-01 -9.49897230e-01 4.75422978e-01 -1.26714325e+00
-1.23964655e+00 5.57538390e-01 4.11521122e-02 -1.96327829e+00
-2.89092839e-01 -5.05308807e-01 -7.75431693e-01 1.98366567e-01
-1.51426971e+00 -1.19029546e+00 -5.40227473e-01 9.68861043e-01
9.30979013e-01 -4.15352046e-01 8.82477403e-01 1.58182308e-01
-7.58388638e-01 6.99883938e-01 1.09254373e-02 4.03971553e-01
7.19648540e-01 -8.91931713e-01 1.55610770e-01 5.65210521e-01
1.78766444e-01 5.59707224e-01 3.89591306e-01 -4.25895512e-01
-1.41621804e+00 -1.37868798e+00 5.72373092e-01 -3.20289969e-01
6.17073536e-01 -4.88141865e-01 -9.41841900e-01 8.81583512e-01
1.88782886e-02 4.90135133e-01 7.55844533e-01 -1.53430596e-01
-8.01952302e-01 -6.22331090e-02 -7.30967343e-01 5.90308309e-01
1.36887765e+00 -1.03294623e+00 -9.09452081e-01 6.67120934e-01
7.48522818e-01 -6.43981814e-01 -3.38130355e-01 2.64984727e-01
5.01740515e-01 -9.99356866e-01 8.88026595e-01 -1.15370405e+00
1.15345454e+00 -9.22901183e-02 -6.57548159e-02 -1.39536250e+00
-5.12088954e-01 -6.31061792e-01 -4.06229287e-01 1.08918321e+00
2.21352398e-01 -7.37401471e-02 6.90326750e-01 5.06044984e-01
-1.84938908e-01 -7.58172035e-01 -7.17455328e-01 -8.16828668e-01
-1.37535825e-01 -3.66139680e-01 5.63322425e-01 1.14519322e+00
2.21547753e-01 1.35364514e-02 -5.58773160e-01 -1.96877241e-01
2.57335424e-01 1.67972863e-01 7.34854281e-01 -8.87477160e-01
-2.08410993e-01 -2.50410497e-01 -4.34786320e-01 -1.51658344e+00
5.58547556e-01 -8.97555411e-01 -1.41580120e-01 -1.34448338e+00
3.37538034e-01 -7.44510815e-02 -5.08123696e-01 6.71839178e-01
-4.58923578e-02 1.86866239e-01 3.18028122e-01 5.82257926e-01
-1.13250792e+00 6.15142405e-01 1.08226407e+00 -4.51385707e-01
-1.92768127e-01 -7.76472911e-02 -5.65319777e-01 8.04499865e-01
4.28578764e-01 -2.12191150e-01 -5.76663375e-01 -8.03473651e-01
7.23575950e-02 2.82778978e-01 4.23170120e-01 -1.06484747e+00
2.22800002e-01 -2.16925830e-01 4.89167303e-01 -5.50932407e-01
2.59228110e-01 -6.29290462e-01 -1.56762168e-01 -3.69659849e-02
-9.15344357e-01 -7.51381069e-02 -1.17337490e-02 6.21656954e-01
-4.39599901e-01 -1.40998224e-02 4.70850170e-01 -2.77147621e-01
-1.05843699e+00 4.79749233e-01 -3.08452636e-01 -1.63716767e-02
1.05599856e+00 -3.41959387e-01 -3.92967045e-01 -7.25334704e-01
-9.24918354e-01 3.82082850e-01 5.36624730e-01 7.68194437e-01
8.53698790e-01 -1.68112195e+00 -3.96726429e-01 4.29428458e-01
5.97210765e-01 4.33791764e-02 5.16749144e-01 3.61669928e-01
-4.27937359e-01 3.34479958e-01 -3.88099968e-01 -5.66009223e-01
-9.09439325e-01 9.52644408e-01 -5.06674536e-02 4.43588980e-02
-7.63972938e-01 8.21525514e-01 5.40589631e-01 1.50894998e-02
6.47880495e-01 -2.96538502e-01 -4.70242590e-01 2.76910737e-02
8.44730616e-01 -3.29129547e-02 -4.21807498e-01 -5.94474137e-01
-2.12866470e-01 4.57219899e-01 -2.54355133e-01 3.44548792e-01
1.54736900e+00 1.03511095e-01 2.18134925e-01 4.30885136e-01
1.35321522e+00 -4.56483126e-01 -1.80721736e+00 -4.76427346e-01
-1.02837011e-01 -6.80706084e-01 -7.60338902e-02 -5.65753341e-01
-1.20061445e+00 9.91755903e-01 2.72476286e-01 -1.50129199e-01
1.18747497e+00 1.02914944e-01 5.69415510e-01 4.21759814e-01
3.72918129e-01 -1.04754901e+00 7.61048913e-01 4.10699040e-01
1.27577126e+00 -1.17434931e+00 -5.71695343e-02 -3.81065607e-01
-1.18492699e+00 8.65178645e-01 8.19199622e-01 -4.70754981e-01
4.18609917e-01 -1.09859541e-01 -4.60822135e-02 -5.29292077e-02
-7.92989075e-01 -1.38491720e-01 4.06857789e-01 7.70468473e-01
3.75876933e-01 6.60203993e-02 6.02760352e-02 9.77806866e-01
7.28832856e-02 3.26423258e-01 3.81652415e-01 9.28748012e-01
-1.34851232e-01 -6.95862114e-01 6.47031963e-02 4.13857549e-01
4.59386371e-02 -5.61660416e-02 -3.66575956e-01 5.74917614e-01
-3.75171043e-02 6.90088332e-01 3.59678894e-01 -7.47798920e-01
3.69271696e-01 1.31016672e-01 1.52418867e-01 -9.31915402e-01
-4.01805639e-01 -8.48922580e-02 -1.06465556e-01 -9.74968016e-01
-5.68610370e-01 -4.60633397e-01 -1.08566570e+00 1.13188170e-01
1.07669175e-01 -1.26912773e-01 8.34935829e-02 1.04045916e+00
4.99340117e-01 5.75710475e-01 9.20647502e-01 -1.13307881e+00
-7.49252975e-01 -8.88768673e-01 -4.82937545e-01 1.04740524e+00
4.20659065e-01 -9.08447683e-01 -3.23936492e-01 3.36922646e-01]
|
[8.692266464233398, 0.7276906371116638]
|
ae30e49c-38d0-4892-8e25-4500ab6bd6fe
|
neural-user-simulation-for-corpus-based
| null | null |
https://aclanthology.org/W18-5007
|
https://aclanthology.org/W18-5007.pdf
|
Neural User Simulation for Corpus-based Policy Optimisation of Spoken Dialogue Systems
|
User Simulators are one of the major tools that enable offline training of task-oriented dialogue systems. For this task the Agenda-Based User Simulator (ABUS) is often used. The ABUS is based on hand-crafted rules and its output is in semantic form. Issues arise from both properties such as limited diversity and the inability to interface a text-level belief tracker. This paper introduces the Neural User Simulator (NUS) whose behaviour is learned from a corpus and which generates natural language, hence needing a less labelled dataset than simulators generating a semantic output. In comparison to much of the past work on this topic, which evaluates user simulators on corpus-based metrics, we use the NUS to train the policy of a reinforcement learning based Spoken Dialogue System. The NUS is compared to the ABUS by evaluating the policies that were trained using the simulators. Cross-model evaluation is performed i.e. training on one simulator and testing on the other. Furthermore, the trained policies are tested on real users. In both evaluation tasks the NUS outperformed the ABUS.
|
["Milica Ga{\\v{s}}i{\\'c}", 'Pawe{\\l} Budzianowski', 'I{\\~n}igo Casanueva', 'Florian Kreyssig']
|
2018-07-01
| null | null | null |
ws-2018-7
|
['user-simulation']
|
['natural-language-processing']
|
[ 2.57411867e-01 7.06141710e-01 1.91203237e-01 -4.57350850e-01
-5.40338218e-01 -5.61143160e-01 9.51532841e-01 -4.16317210e-03
-7.83686996e-01 1.19130623e+00 2.07486212e-01 -5.25910497e-01
3.05522114e-01 -4.85322446e-01 -5.33426642e-01 -3.87716085e-01
9.94253010e-02 1.09118462e+00 4.04363722e-01 -7.39436388e-01
2.31561139e-01 -6.89769909e-02 -1.48119426e+00 1.89695746e-01
1.06942987e+00 6.41166210e-01 6.82060421e-01 1.22617829e+00
-8.15495625e-02 9.71449494e-01 -1.05055988e+00 -3.43719162e-02
-7.54126236e-02 -1.01478076e+00 -1.06922460e+00 -8.23341161e-02
-9.67413858e-02 -3.97481352e-01 4.46976162e-02 8.42428923e-01
6.16944671e-01 5.26809692e-01 5.77183604e-01 -1.26287818e+00
8.03814530e-02 4.26489025e-01 4.12032872e-01 -1.70415804e-01
8.16333532e-01 3.63703519e-01 7.09400415e-01 -3.04260492e-01
6.61875010e-01 1.69989204e+00 3.64029318e-01 1.21254551e+00
-1.28019643e+00 -2.12832108e-01 -5.35744168e-02 -4.02732611e-01
-9.30299222e-01 -4.75903958e-01 2.46215403e-01 -4.12827671e-01
1.15061927e+00 2.15622813e-01 4.57283020e-01 1.39386189e+00
-3.99615616e-02 9.99196947e-01 1.46802747e+00 -7.89237618e-01
6.63073123e-01 9.02317107e-01 -2.72845924e-02 7.87758112e-01
-2.84990191e-01 5.77958822e-01 -2.91359603e-01 -3.84330601e-01
7.40722775e-01 -8.01113784e-01 -3.94295365e-01 -3.48658979e-01
-8.93821836e-01 1.06138301e+00 -3.87477800e-02 2.33861968e-01
-4.56047565e-01 -1.14998452e-01 6.24129534e-01 6.21332049e-01
1.04787506e-01 9.62795615e-01 -5.79873502e-01 -5.92883468e-01
-6.46171808e-01 6.04788661e-01 1.72682405e+00 8.26919019e-01
3.02724093e-01 3.00012231e-01 -4.43325669e-01 9.25511718e-01
5.13121367e-01 3.51698369e-01 8.48742306e-01 -9.57233489e-01
2.35060468e-01 4.01658386e-01 4.89370108e-01 -2.21336842e-01
-3.85937452e-01 1.34625524e-01 -2.00395539e-01 3.85511369e-01
5.56753695e-01 -8.33079755e-01 -7.95128047e-01 1.86685801e+00
2.80614018e-01 6.77291080e-02 5.63122690e-01 7.89182723e-01
8.12671065e-01 8.48432422e-01 3.57912034e-01 -5.99935986e-02
1.11892664e+00 -9.16721821e-01 -7.02779114e-01 -1.82507977e-01
9.43863809e-01 -3.59065622e-01 1.21520543e+00 5.21902800e-01
-1.11726367e+00 -5.53028286e-01 -9.32738841e-01 4.36951101e-01
-4.67100054e-01 -2.26884365e-01 2.65292853e-01 7.41476178e-01
-1.42083561e+00 6.38407409e-01 -5.83584011e-01 -6.29632056e-01
-3.52546245e-01 4.98352498e-01 7.54930153e-02 4.30586249e-01
-1.77257061e+00 1.56401098e+00 8.18479836e-01 -4.25292075e-01
-1.16845405e+00 -1.14171676e-01 -1.01783264e+00 -7.26557374e-02
3.30097347e-01 -4.57083583e-01 2.39327192e+00 -1.14275873e+00
-2.47964168e+00 5.83358824e-01 2.17240065e-01 -6.92109346e-01
6.86257064e-01 -3.64813916e-02 -1.94041342e-01 -4.30110618e-02
-8.64953920e-02 8.38713348e-01 5.21380007e-01 -1.44634664e+00
-8.94684196e-01 2.07422316e-01 3.49158168e-01 7.23413229e-01
7.47824609e-02 -7.41722435e-02 -1.57252610e-01 -2.06825778e-01
-8.11849713e-01 -1.09107161e+00 -2.31962726e-01 -7.10259616e-01
-2.52391785e-01 -5.44125378e-01 7.46848822e-01 -5.21148622e-01
1.16518414e+00 -1.65431225e+00 2.07318813e-01 1.18339106e-01
-3.49526465e-01 6.97314084e-01 -5.47666065e-02 7.29860485e-01
2.27781907e-01 -1.74656913e-01 -1.54909194e-01 -2.84163058e-01
1.55133560e-01 5.75098932e-01 -1.71628237e-01 -1.73597723e-01
3.17830853e-02 5.79919517e-01 -1.16202128e+00 -3.65654796e-01
2.36472011e-01 1.02900036e-01 -5.91303110e-01 9.38798726e-01
-8.41782033e-01 7.06176221e-01 -4.72305775e-01 -1.66083872e-01
-1.09429888e-01 -7.26536214e-02 3.76550436e-01 4.14194584e-01
-8.51061344e-02 6.60631716e-01 -9.36499059e-01 1.68861008e+00
-9.38501894e-01 3.71634722e-01 1.66123018e-01 -7.85738349e-01
9.04435933e-01 8.58968496e-01 -2.05941334e-01 -7.96929479e-01
2.77671665e-01 1.90852568e-01 3.78392696e-01 -5.74456036e-01
6.21903658e-01 -2.38086507e-01 -2.15046972e-01 8.43710899e-01
5.36308765e-01 -4.10459548e-01 2.48818606e-01 4.16202873e-01
8.52230012e-01 5.88536263e-01 4.69912142e-01 -3.78493637e-01
7.24136531e-01 3.04378808e-01 1.09145686e-01 1.17187536e+00
-1.11279331e-01 2.56065160e-01 5.39643049e-01 -3.51271741e-02
-9.90425229e-01 -6.03249371e-01 1.39672667e-01 1.48768926e+00
-1.23824276e-01 -2.68304408e-01 -1.39071023e+00 -8.12786281e-01
-2.14518726e-01 1.54775977e+00 -4.43002194e-01 -1.81635112e-01
-4.36294556e-01 -2.81375647e-01 6.22711182e-01 2.26468191e-01
4.34280455e-01 -1.57742858e+00 -9.72470820e-01 5.73189080e-01
9.61449370e-03 -9.23435867e-01 -2.47833908e-01 1.46719664e-01
-6.47166967e-01 -8.53922963e-01 -6.72310829e-01 -5.87806523e-01
4.01985228e-01 -2.42797017e-01 1.35154092e+00 1.36574268e-01
2.82828897e-01 6.89546168e-01 -3.86475891e-01 -9.25759852e-01
-1.40973938e+00 1.41367346e-01 1.72354266e-01 -2.76077360e-01
2.72745639e-01 -1.46513015e-01 -2.21356735e-01 3.19247305e-01
-8.58883917e-01 3.62256318e-01 4.81296718e-01 1.20036709e+00
-2.16074228e-01 -2.95046359e-01 8.50394130e-01 -1.35858083e+00
1.43210709e+00 -3.48084390e-01 -6.42945766e-01 2.64986217e-01
-7.35167623e-01 3.67987216e-01 7.66640007e-01 -3.80669147e-01
-1.36667097e+00 5.09195216e-02 -2.71316171e-01 4.21450101e-02
-5.66751063e-01 3.61900985e-01 -3.47681306e-02 2.71661729e-01
9.57884729e-01 2.60794818e-01 4.35117900e-01 -2.98243791e-01
1.93167344e-01 1.09158683e+00 3.11893374e-01 -8.73169422e-01
1.96400210e-01 -3.68648052e-01 -6.42404020e-01 -1.13005555e+00
-6.62976146e-01 -1.74348533e-01 -2.48800844e-01 -3.58081251e-01
6.98680878e-01 -6.23356462e-01 -7.36666858e-01 3.61161739e-01
-1.14997566e+00 -1.10123575e+00 -2.00814933e-01 3.82796675e-01
-1.00737250e+00 -7.52477273e-02 -4.66410190e-01 -1.29348791e+00
-3.06824476e-01 -1.25632143e+00 9.31467891e-01 4.40882415e-01
-7.91735768e-01 -1.39699984e+00 4.73357111e-01 6.49320930e-02
3.87417853e-01 -5.80100976e-02 7.96240747e-01 -1.45424342e+00
1.90106347e-01 -1.29810899e-01 3.06308448e-01 4.36085224e-01
-1.74390435e-01 -3.63090038e-01 -1.18980360e+00 -3.37295979e-01
7.03788325e-02 -9.47734416e-01 1.00011513e-01 5.48528135e-02
4.56896365e-01 -4.67181534e-01 -1.01898208e-01 -1.36023030e-01
8.97419631e-01 6.84893847e-01 3.39972407e-01 3.34514380e-01
1.44785076e-01 8.95339131e-01 6.49831116e-01 1.56145796e-01
4.06570137e-01 6.91920102e-01 9.57524255e-02 -3.95443775e-02
3.65052849e-01 -4.80560929e-01 6.30958259e-01 5.32892585e-01
1.63247645e-01 -4.46200758e-01 -8.95667315e-01 2.14288116e-01
-2.07789898e+00 -6.33661389e-01 4.10750657e-01 2.21394348e+00
1.18958819e+00 3.52007478e-01 5.33151686e-01 -1.84384093e-01
5.25329173e-01 -7.29579926e-02 -4.39142704e-01 -1.06445944e+00
4.14731145e-01 3.82800549e-01 1.82919607e-01 1.06113887e+00
-7.45565712e-01 1.06226659e+00 5.98057699e+00 3.75794709e-01
-9.00394201e-01 7.43159791e-03 4.16598141e-01 2.89362520e-01
7.01501295e-02 -9.81970727e-02 -8.68608892e-01 4.97654915e-01
1.70503354e+00 -1.53865665e-01 5.07952809e-01 8.77854645e-01
5.26314795e-01 -5.60321391e-01 -1.15922606e+00 3.33278686e-01
-5.36772273e-02 -9.10262704e-01 -1.12812899e-01 -1.48831218e-01
5.17757654e-01 -1.83271378e-01 -3.17185670e-01 1.02566361e+00
1.02624094e+00 -1.15873742e+00 5.39733410e-01 4.34574187e-01
6.01814330e-01 -6.50998414e-01 8.72141123e-01 9.75255609e-01
-1.71582356e-01 2.71943450e-01 -2.96099801e-02 -6.93838447e-02
2.04121590e-01 -4.62624997e-01 -1.72156513e+00 2.44655207e-01
1.68728963e-01 -8.70253742e-02 -1.63636535e-01 7.02352107e-01
-2.28944793e-01 8.24223280e-01 -3.42246443e-01 -6.22201145e-01
6.97335958e-01 -3.14636171e-01 4.26545769e-01 1.41261566e+00
-1.56109214e-01 1.32589936e-01 5.54177165e-01 6.04605973e-01
2.75234938e-01 1.20738782e-02 -8.46143723e-01 -1.90857902e-01
3.27548355e-01 1.00168145e+00 -2.25576103e-01 -6.28571093e-01
-2.42365584e-01 1.02978456e+00 1.11729927e-01 2.71428138e-01
-4.60556746e-01 -3.32660854e-01 3.38510424e-01 -6.93387389e-02
-5.79604320e-02 1.61722556e-01 2.38101929e-01 -7.21801758e-01
-8.10281873e-01 -1.34739447e+00 1.88543856e-01 -7.62542784e-01
-9.42665458e-01 9.40070271e-01 2.98460960e-01 -7.92368770e-01
-1.19854355e+00 -6.16337717e-01 -5.86612463e-01 1.31778026e+00
-1.08816230e+00 -4.14861262e-01 -8.42185467e-02 3.44428152e-01
8.35727513e-01 -5.27449369e-01 1.39881563e+00 -2.80815929e-01
-4.88685906e-01 4.73841995e-01 -6.23375028e-02 -4.80997264e-02
4.98607159e-01 -1.60850537e+00 4.55937624e-01 1.45333886e-01
-1.77203029e-01 4.91005808e-01 1.22182965e+00 -6.10643387e-01
-9.37109232e-01 -7.50582337e-01 6.79899454e-01 -4.86654222e-01
4.79903907e-01 -3.79507512e-01 -1.03937054e+00 5.99056602e-01
6.18411243e-01 -6.84626818e-01 6.19784117e-01 -1.67406008e-01
3.66158128e-01 5.84935963e-01 -1.24851346e+00 7.34961450e-01
4.44697320e-01 -4.06300336e-01 -8.50194991e-01 4.40886050e-01
4.76723075e-01 -8.30292881e-01 -7.44279802e-01 2.25687353e-03
3.55786055e-01 -1.00217688e+00 4.88808811e-01 -1.00567091e+00
1.90167248e-01 5.01688756e-02 1.36087731e-01 -1.97927976e+00
2.15795457e-01 -9.33011353e-01 -8.09393227e-02 9.74056244e-01
5.97586513e-01 -6.70621693e-01 5.84355891e-01 7.51160622e-01
-3.28591140e-03 -6.79324269e-01 -4.80825126e-01 -6.88360155e-01
-3.51565029e-03 1.64409522e-02 3.85575891e-01 5.97576380e-01
1.26243189e-01 1.02192712e+00 -3.76472414e-01 -2.61420190e-01
1.57691449e-01 -4.80141163e-01 9.19954300e-01 -1.16734612e+00
-4.98317391e-01 -3.27565461e-01 1.59136266e-01 -1.09636807e+00
2.51296103e-01 -5.57549357e-01 5.67882180e-01 -1.25454938e+00
-5.05480170e-01 -3.37814689e-01 3.96420360e-02 1.11023448e-01
-1.32236764e-01 -5.43687940e-01 1.34394497e-01 -1.44273907e-01
-5.53455174e-01 6.75217688e-01 1.21992469e+00 4.05872852e-01
-5.89141965e-01 3.13390255e-01 -4.62480128e-01 8.83919299e-01
1.07225049e+00 -3.88030946e-01 -8.59770894e-01 1.37400270e-01
-1.14288747e-01 5.16389847e-01 4.14301716e-02 -8.54775190e-01
2.57646650e-01 -1.60599515e-01 -5.62571594e-03 4.60621789e-02
3.20254564e-01 -4.89800334e-01 -3.13048095e-01 5.61580837e-01
-8.31801951e-01 8.19485784e-02 2.56206095e-01 4.40201908e-01
-1.24411322e-01 -8.17445576e-01 8.58599484e-01 -6.41853571e-01
-7.37755120e-01 -2.99366742e-01 -8.40104818e-01 2.72949040e-01
9.58098650e-01 -2.09508255e-01 1.11245960e-01 -1.01463866e+00
-6.06333435e-01 4.61484551e-01 2.21546575e-01 4.10822183e-01
3.82419050e-01 -7.77170420e-01 -6.57015681e-01 1.69024870e-01
7.72218630e-02 -6.42976724e-03 -2.61627942e-01 3.98547500e-01
-6.03517354e-01 7.18438864e-01 -2.08760470e-01 -4.68636006e-01
-1.07075953e+00 1.00509919e-01 6.95472419e-01 -4.92462426e-01
-3.86397749e-01 6.26554608e-01 1.45052880e-01 -1.09599102e+00
4.91947681e-01 1.43367797e-01 -5.24052978e-01 -1.51378915e-01
4.13882494e-01 4.87494059e-02 -9.23218280e-02 -4.18170780e-01
2.14227065e-01 -5.53890705e-01 -1.18644029e-01 -9.41026866e-01
1.05854511e+00 1.66047975e-01 4.53523993e-01 7.57760525e-01
6.39054537e-01 -5.05660594e-01 -1.28230119e+00 -2.48242229e-01
2.83865213e-01 -6.72676135e-03 -1.86554324e-02 -1.26837254e+00
-1.40764818e-01 7.63664961e-01 5.25524318e-01 6.09080732e-01
4.62726891e-01 -5.34863293e-01 4.66425419e-01 6.73709512e-01
5.13269067e-01 -1.53642154e+00 1.17486380e-02 8.54845464e-01
1.03855526e+00 -1.28028405e+00 -4.09722298e-01 2.12308168e-01
-1.42704189e+00 8.32219899e-01 1.02112496e+00 -1.09535046e-01
1.71811089e-01 1.63252577e-01 3.35812420e-01 -1.19036287e-02
-1.11881566e+00 -1.84843928e-01 -4.40312885e-02 7.08941579e-01
5.29608011e-01 4.50044163e-02 -2.35907063e-01 3.09671611e-01
-4.22380954e-01 1.67030692e-01 6.05824351e-01 1.16352725e+00
-5.25969863e-01 -1.35622978e+00 -2.66808063e-01 5.21502137e-01
-1.25022650e-01 2.51241941e-02 -6.86317205e-01 9.68315005e-01
-4.60117698e-01 1.03020310e+00 -8.28216448e-02 -3.48205268e-01
5.99050820e-01 5.68447769e-01 3.87973070e-01 -1.08674192e+00
-1.16279805e+00 -2.16167003e-01 6.98967457e-01 -4.65613306e-01
-1.10998854e-01 -5.40820301e-01 -1.14487255e+00 9.73253548e-02
-3.34960461e-01 8.03732753e-01 8.11100066e-01 9.78591859e-01
6.16257265e-02 5.02711356e-01 6.90889120e-01 -8.59875441e-01
-1.19419146e+00 -1.44172287e+00 -3.74342591e-01 3.81280154e-01
2.44426936e-01 -6.55481279e-01 -1.54614478e-01 -2.38738731e-01]
|
[13.050532341003418, 8.027019500732422]
|
cffbccb1-a4fb-4b80-a3d1-df7680ba523f
|
deep-convolutional-neural-network-for-multi
|
1910.04066
| null |
https://arxiv.org/abs/1910.04066v1
|
https://arxiv.org/pdf/1910.04066v1.pdf
|
Deep Convolutional Neural Network for Multi-modal Image Restoration and Fusion
|
In this paper, we propose a novel deep convolutional neural network to solve the general multi-modal image restoration (MIR) and multi-modal image fusion (MIF) problems. Different from other methods based on deep learning, our network architecture is designed by drawing inspirations from a new proposed multi-modal convolutional sparse coding (MCSC) model. The key feature of the proposed network is that it can automatically split the common information shared among different modalities, from the unique information that belongs to each single modality, and is therefore denoted with CU-Net, i.e., Common and Unique information splitting network. Specifically, the CU-Net is composed of three modules, i.e., the unique feature extraction module (UFEM), common feature preservation module (CFPM), and image reconstruction module (IRM). The architecture of each module is derived from the corresponding part in the MCSC model, which consists of several learned convolutional sparse coding (LCSC) blocks. Extensive numerical results verify the effectiveness of our method on a variety of MIR and MIF tasks, including RGB guided depth image super-resolution, flash guided non-flash image denoising, multi-focus and multi-exposure image fusion.
|
['Pier Luigi Dragotti', 'Xin Deng']
|
2019-10-09
| null | null | null | null |
['multi-exposure-image-fusion']
|
['computer-vision']
|
[ 3.81344765e-01 -2.38322034e-01 1.29543439e-01 -5.65855391e-02
-9.23722267e-01 1.03515036e-01 2.58027494e-01 -4.08027321e-01
-1.86696827e-01 6.51662052e-01 5.43751359e-01 2.12058425e-01
-2.55850405e-01 -6.86519921e-01 -6.61612391e-01 -1.09792066e+00
4.70298856e-01 -2.98289299e-01 1.92999333e-01 -3.51059943e-01
1.93531036e-01 2.82826453e-01 -1.68144453e+00 4.97902244e-01
1.00934017e+00 1.28307498e+00 3.33129764e-01 3.34302366e-01
-3.93680893e-02 1.28347743e+00 -2.10364476e-01 4.43578735e-02
6.85554370e-02 -6.04916930e-01 -8.21217418e-01 1.70117930e-01
2.25157633e-01 -5.54326236e-01 -6.70069098e-01 1.11200321e+00
6.78817213e-01 2.40865216e-01 4.61649030e-01 -1.18648553e+00
-7.26739526e-01 2.77030677e-01 -8.38980138e-01 2.54666716e-01
2.01454848e-01 -7.45283514e-02 2.41499439e-01 -7.89872408e-01
5.05871773e-01 1.08181953e+00 7.13765919e-01 4.10350770e-01
-7.60481179e-01 -6.12675011e-01 -1.07414745e-01 3.19423199e-01
-1.25652015e+00 -4.48754936e-01 8.31140041e-01 -3.94375920e-01
5.32516837e-01 6.92904070e-02 4.46668983e-01 4.65072632e-01
4.08574224e-01 6.71549797e-01 1.23923111e+00 -3.35652620e-01
-4.70569078e-03 -2.83745527e-01 7.87902176e-02 7.27209866e-01
-4.59398553e-02 1.75086543e-01 -5.34164846e-01 1.27049342e-01
9.64906514e-01 3.55654895e-01 -7.02554047e-01 -1.88621596e-01
-1.10957015e+00 7.11861968e-01 5.25195062e-01 5.30890644e-01
-3.95943850e-01 5.66319041e-02 2.47803286e-01 1.39833286e-01
2.77540505e-01 -4.32395667e-01 -1.45029679e-01 3.37023437e-01
-9.19979095e-01 -2.34903451e-02 3.57236326e-01 6.67475283e-01
1.26837182e+00 3.13623965e-01 -1.72974110e-01 9.86068130e-01
3.09928685e-01 5.31436324e-01 8.87970090e-01 -1.24098217e+00
4.67180982e-02 6.13258719e-01 -1.15409821e-01 -9.64971900e-01
-3.34654838e-01 -3.95634979e-01 -1.53509521e+00 1.40463382e-01
-3.74912351e-01 -2.13305756e-01 -8.48436952e-01 1.71396732e+00
3.45983654e-01 4.68972087e-01 2.44292125e-01 1.06232703e+00
1.44943273e+00 6.80747509e-01 -2.28265628e-01 -3.88651520e-01
1.10239935e+00 -8.83061767e-01 -9.56933618e-01 2.22075023e-02
2.71645039e-01 -6.84200227e-01 4.76431787e-01 2.93933809e-01
-1.16494644e+00 -6.60214901e-01 -1.09941506e+00 -3.57565522e-01
-4.56691533e-02 9.71781835e-02 5.92956305e-01 1.39724225e-01
-1.13954878e+00 3.69452477e-01 -4.20809567e-01 -1.36188060e-01
5.06246567e-01 3.18692207e-01 -6.30753934e-01 -5.49070299e-01
-1.10261655e+00 6.86336756e-01 3.52505296e-01 1.87652335e-01
-9.51479197e-01 -6.69511437e-01 -9.28277075e-01 9.71123055e-02
1.44036680e-01 -8.41356158e-01 7.74663925e-01 -1.13143659e+00
-1.37568045e+00 5.24724662e-01 -2.31135562e-01 -1.32444888e-01
-1.80318415e-01 1.54654300e-02 -5.55666208e-01 5.08275568e-01
2.49600753e-01 5.41867733e-01 1.01904368e+00 -1.67873096e+00
-6.39405966e-01 -2.42787227e-01 -4.95216139e-02 1.69960469e-01
-1.64367735e-01 1.19260497e-01 -4.18537050e-01 -6.46168411e-01
2.65472800e-01 -5.05452752e-01 -9.55694988e-02 -2.10673973e-01
-4.45808023e-01 2.50746340e-01 1.02465260e+00 -9.54382420e-01
1.28167963e+00 -2.27522421e+00 5.11893690e-01 -1.15867861e-01
3.49754989e-01 1.43124610e-01 -1.47719115e-01 4.19845700e-01
-4.45153803e-01 -1.96359903e-01 -6.86380684e-01 -3.80216271e-01
-4.72931206e-01 1.87713280e-01 -9.23317596e-02 6.01586819e-01
-8.61680508e-02 8.45067382e-01 -7.01302409e-01 -6.29313052e-01
5.26222467e-01 7.61697829e-01 -3.10114175e-01 1.50486231e-01
2.27694914e-01 6.88911200e-01 -1.92647263e-01 8.71748090e-01
1.21457982e+00 -3.99032086e-01 -2.01223105e-01 -7.65838504e-01
-3.53561342e-01 -4.34161812e-01 -1.49536228e+00 2.05272150e+00
-4.99061674e-01 2.34164089e-01 4.55794126e-01 -1.11529040e+00
6.19575441e-01 3.33129644e-01 7.29743361e-01 -9.86998141e-01
2.99641520e-01 1.09994985e-01 -3.86827677e-01 -5.20038605e-01
4.98789996e-01 -5.16647279e-01 2.68363446e-01 5.40392399e-01
4.89598751e-01 2.31330767e-01 1.89685002e-02 3.54584008e-01
8.28461707e-01 -3.33991110e-01 8.18600357e-02 -1.86302871e-01
1.03017020e+00 -3.24390560e-01 7.21723676e-01 2.96646982e-01
-3.53427008e-02 9.77468491e-01 2.00352848e-01 -2.75815934e-01
-7.40127385e-01 -9.98569965e-01 -5.71300015e-02 4.03968096e-01
5.48863232e-01 -3.54703814e-01 -4.82273936e-01 -2.57982731e-01
-2.45751545e-01 8.81474912e-02 -4.83088404e-01 -3.18941474e-01
-5.97613454e-01 -7.36069202e-01 4.59345020e-02 3.66306156e-01
9.87761319e-01 -1.04383588e+00 -2.52120286e-01 2.66248323e-02
-5.83104730e-01 -1.04802334e+00 -3.98419470e-01 -8.35921690e-02
-7.25595772e-01 -1.38946259e+00 -7.15625942e-01 -9.42178667e-01
4.20928448e-01 8.03072274e-01 8.02466154e-01 1.90633744e-01
-2.52933115e-01 5.30727386e-01 -3.79727513e-01 8.81137326e-02
-1.08735748e-01 -3.63839865e-01 -1.05752774e-01 6.24744534e-01
-1.66762561e-01 -7.25279391e-01 -7.80175269e-01 1.15334116e-01
-1.57381535e+00 2.82727748e-01 7.91473091e-01 1.03022814e+00
7.98087358e-01 4.30510640e-01 4.63962525e-01 -6.07802927e-01
3.74246508e-01 -5.71423054e-01 -9.91963223e-02 2.74301589e-01
-2.47755319e-01 -2.99880326e-01 5.81368268e-01 -1.32791430e-01
-1.25513399e+00 6.96533844e-02 -2.34869823e-01 -6.50790870e-01
-3.55778784e-02 5.32050431e-01 -3.64074677e-01 -5.51387668e-01
9.27186757e-02 7.68102884e-01 2.67654479e-01 -5.27608275e-01
4.92710292e-01 8.07270110e-01 8.20797622e-01 -1.82288989e-01
7.48892426e-01 8.32623422e-01 9.69153568e-02 -8.44113708e-01
-9.36130702e-01 -4.34882641e-01 -4.98709977e-01 -3.62528414e-01
1.06152987e+00 -1.39020491e+00 -8.14328313e-01 9.56305027e-01
-1.05060792e+00 5.23393080e-02 -2.42842749e-01 3.78776759e-01
-4.16462749e-01 6.64700150e-01 -7.49821961e-01 -3.57692808e-01
-4.18851107e-01 -1.22514915e+00 1.26383531e+00 6.71476901e-01
6.35121167e-01 -9.07627881e-01 4.48146686e-02 6.62579656e-01
6.12585068e-01 4.38322902e-01 9.40202713e-01 9.47552100e-02
-7.12852240e-01 3.95589322e-02 -4.63653356e-01 8.62283885e-01
1.33022085e-01 -4.59601194e-01 -8.69744480e-01 -5.47599077e-01
4.03949171e-01 -4.89593357e-01 1.22001743e+00 6.01254761e-01
1.22773802e+00 -8.42704922e-02 -1.03694730e-01 1.02657902e+00
1.96520829e+00 1.26177743e-01 1.06981575e+00 2.88193673e-01
8.62091124e-01 2.57345676e-01 2.05936462e-01 4.07556027e-01
5.52908182e-01 4.29257184e-01 7.59312570e-01 -4.41821575e-01
-3.27297419e-01 2.06436783e-01 3.23557258e-01 1.04369783e+00
1.54894397e-01 2.61356551e-02 -4.76037025e-01 4.26072031e-01
-1.77365577e+00 -9.52154756e-01 -1.74587563e-01 1.78225541e+00
4.43135679e-01 -5.64869881e-01 -2.68091381e-01 1.46270886e-01
7.45895684e-01 2.85607219e-01 -4.70298797e-01 4.47555855e-02
-6.43796802e-01 3.51353168e-01 2.26106122e-01 5.35750926e-01
-1.13341045e+00 4.29813623e-01 5.49153376e+00 9.99743223e-01
-9.74956989e-01 4.37018156e-01 6.03791595e-01 9.69554484e-02
-3.91770840e-01 -7.83209503e-03 -4.52940524e-01 6.63859189e-01
5.61975420e-01 1.17505312e-01 4.51946527e-01 2.15033486e-01
-4.78615947e-02 -4.18766171e-01 -4.46222693e-01 1.35498369e+00
4.34774280e-01 -1.58757663e+00 2.82102823e-01 -2.49102697e-01
8.50086987e-01 -6.66218102e-02 5.12752570e-02 4.18223664e-02
-1.69834588e-02 -9.47270811e-01 4.22473013e-01 9.01494145e-01
9.70655084e-01 -8.99965167e-01 8.50368559e-01 1.57313317e-01
-1.34362674e+00 -4.01988447e-01 -3.27300638e-01 3.11621755e-01
2.44229436e-01 8.76049042e-01 5.83186328e-01 1.31585562e+00
8.37525547e-01 1.22854996e+00 -3.16905558e-01 9.19473588e-01
2.92285550e-02 1.83516964e-01 3.38617191e-02 8.88738036e-01
4.71361950e-02 -3.19763899e-01 3.58655393e-01 8.16451371e-01
3.65872532e-01 3.51932377e-01 8.77137855e-03 6.98798120e-01
6.57813549e-02 -1.41718000e-01 -2.33600602e-01 3.63922179e-01
2.09470659e-01 1.55158305e+00 -3.07372987e-01 -2.71354079e-01
-7.31134951e-01 1.04261458e+00 6.96067736e-02 4.17334706e-01
-6.03041172e-01 -4.61282313e-01 5.15565515e-01 -1.07610241e-01
4.09886062e-01 8.87002274e-02 -1.06597222e-01 -1.38588011e+00
4.33665467e-03 -8.67309809e-01 3.68817598e-01 -1.14103341e+00
-1.16959178e+00 6.65854156e-01 -2.07425430e-01 -1.30242753e+00
2.75845677e-01 -2.51802951e-01 -7.69020259e-01 1.20514989e+00
-2.08376575e+00 -1.44312024e+00 -7.78032243e-01 1.19305527e+00
1.79847062e-01 -1.63384303e-01 4.22783166e-01 5.88859081e-01
-8.31557572e-01 1.53546363e-01 3.67523879e-01 5.79040311e-03
5.06139636e-01 -7.74232328e-01 -5.33409894e-01 1.05681920e+00
-5.61645865e-01 4.09847975e-01 -2.71693915e-02 -4.66343075e-01
-1.55298877e+00 -1.21474421e+00 4.88213599e-01 2.56117791e-01
3.97680789e-01 1.16498053e-01 -8.75242889e-01 4.64220583e-01
3.43658954e-01 1.53583884e-01 5.74292004e-01 -5.00181317e-01
-1.49896428e-01 -3.52372825e-01 -1.15680683e+00 4.90676947e-02
6.65237010e-01 -5.71772814e-01 -2.99781054e-01 3.65347564e-01
7.88633406e-01 -3.93104017e-01 -1.04696715e+00 5.15211940e-01
2.64910668e-01 -1.29907691e+00 1.29803264e+00 -8.22353512e-02
8.99427891e-01 -6.73225164e-01 -5.32767534e-01 -1.10243332e+00
-4.87243146e-01 -2.27264017e-01 -3.07965159e-01 1.22126353e+00
-3.55916768e-01 -4.03069198e-01 4.18729335e-01 -5.29744141e-02
-6.40904725e-01 -8.05480957e-01 -1.13646984e+00 -3.90974939e-01
-7.77425990e-02 -1.58629902e-02 6.35078371e-01 9.11765754e-01
-4.90967244e-01 3.66102248e-01 -7.28409111e-01 3.60042483e-01
8.24938595e-01 3.45686316e-01 2.62931585e-01 -1.12293613e+00
-2.74487138e-01 -1.09136112e-01 -1.13338336e-01 -8.77317905e-01
7.72854239e-02 -9.18630779e-01 -1.13475688e-01 -1.88214386e+00
6.06201947e-01 -1.64040864e-01 -4.67604011e-01 3.36189181e-01
-8.10716972e-02 3.77371877e-01 3.06963533e-01 4.36309695e-01
-7.43294656e-01 8.81390750e-01 1.44983017e+00 -1.94212928e-01
7.02964291e-02 -4.18904692e-01 -9.89958346e-01 4.83532280e-01
4.25651044e-01 -1.10447958e-01 -2.91956067e-01 -6.33747220e-01
4.51071225e-02 5.69933832e-01 8.10759366e-01 -1.23111033e+00
5.77729881e-01 5.20207845e-02 4.64631438e-01 -6.64066792e-01
4.25254524e-01 -7.50569642e-01 2.30249599e-01 4.48098511e-01
8.02186131e-02 -3.12989891e-01 6.21805862e-02 5.60971677e-01
-7.28066146e-01 -3.50645818e-02 1.10942602e+00 -3.32081378e-01
-9.55816507e-01 4.42545861e-01 -1.89589038e-01 -8.87291580e-02
8.46616685e-01 -1.84018180e-01 -5.71382523e-01 -2.75120944e-01
-5.66445649e-01 2.23552987e-01 3.74826521e-01 1.08058013e-01
1.08236504e+00 -1.64210880e+00 -6.66240096e-01 3.52773994e-01
2.21338845e-03 1.87106952e-01 1.18417346e+00 1.17277575e+00
-3.40254396e-01 1.87202752e-01 -3.16474438e-01 -5.06416678e-01
-9.72539127e-01 5.04938245e-01 5.26039481e-01 -1.72328219e-01
-7.48665810e-01 7.18132317e-01 3.13927710e-01 -1.84301168e-01
2.62206346e-02 -1.18030431e-02 -4.67059016e-01 -8.86098146e-02
7.50392258e-01 4.57275689e-01 3.11557520e-02 -9.87758040e-01
-2.13652968e-01 9.20411646e-01 1.21374298e-02 2.21601114e-01
1.56622231e+00 -3.63349140e-01 -8.30660760e-01 2.08915904e-01
1.37637842e+00 -4.94817495e-01 -1.17770123e+00 -4.18211669e-01
-6.82869852e-01 -3.18703592e-01 4.90038335e-01 -8.01591933e-01
-1.71282160e+00 7.70418108e-01 8.54316175e-01 -2.75783420e-01
1.80759335e+00 -1.49533108e-01 1.23107815e+00 -1.16501831e-01
2.86765903e-01 -8.11341107e-01 1.99035063e-01 3.64033818e-01
7.44255245e-01 -1.21793866e+00 3.98156829e-02 -2.32311234e-01
-4.37073052e-01 1.00764251e+00 5.85814059e-01 -1.10598586e-01
1.02746880e+00 1.70837551e-01 -2.62692004e-01 -3.20040196e-01
-4.48127776e-01 -1.53754309e-01 1.26830563e-01 5.37454426e-01
1.23094968e-01 -2.87066251e-01 -2.34088972e-01 7.74609685e-01
3.39060187e-01 2.77545512e-01 7.20569193e-01 9.68286395e-01
-6.00640297e-01 -8.19097281e-01 -3.98357153e-01 5.36610186e-01
-2.30425805e-01 -2.13005900e-01 1.51759550e-01 5.77364206e-01
6.00942552e-01 1.06719625e+00 -4.56148200e-02 -6.97755694e-01
1.14161111e-01 -3.00938308e-01 3.20078224e-01 -2.62759209e-01
-5.78196943e-01 3.00813243e-02 -4.51227486e-01 -7.47567952e-01
-1.06121922e+00 -5.20191789e-01 -1.13611341e+00 -3.23131680e-01
-1.88278168e-01 -1.00925565e-01 3.76129031e-01 1.00928712e+00
3.12041163e-01 8.01285207e-01 7.91964412e-01 -9.59679365e-01
4.17411104e-02 -8.64872694e-01 -9.58481789e-01 3.41298997e-01
7.44973361e-01 -7.16991842e-01 -3.39567840e-01 -2.50653494e-02]
|
[10.601655960083008, -1.863613486289978]
|
63d1b273-c8e8-4767-a276-5998aa418620
|
data-driven-computational-imaging-for
|
2210.16709
| null |
https://arxiv.org/abs/2210.16709v1
|
https://arxiv.org/pdf/2210.16709v1.pdf
|
Data-Driven Computational Imaging for Scientific Discovery
|
In computational imaging, hardware for signal sampling and software for object reconstruction are designed in tandem for improved capability. Examples of such systems include computed tomography (CT), magnetic resonance imaging (MRI), and superresolution microscopy. In contrast to more traditional cameras, in these devices, indirect measurements are taken and computational algorithms are used for reconstruction. This allows for advanced capabilities such as super-resolution or 3-dimensional imaging, pushing forward the frontier of scientific discovery. However, these techniques generally require a large number of measurements, causing low throughput, motion artifacts, and/or radiation damage, limiting applications. Data-driven approaches to reducing the number of measurements needed have been proposed, but they predominately require a ground truth or reference dataset, which may be impossible to collect. This work outlines a self-supervised approach and explores the future work that is necessary to make such a technique usable for real applications. Light-emitting diode (LED) array microscopy, a modality that allows visualization of transparent objects in two and three dimensions with high resolution and field-of-view, is used as an illustrative example. We release our code at https://github.com/vganapati/LED_PVAE and our experimental data at https://doi.org/10.6084/m9.figshare.21232088 .
|
['Vidya Ganapati', 'Yolanda Hu', 'Andrew Olsen']
|
2022-10-29
| null | null | null | null |
['transparent-objects', 'object-reconstruction']
|
['computer-vision', 'computer-vision']
|
[ 5.36905766e-01 -2.93289185e-01 1.39916837e-01 -1.65621862e-01
-6.07444763e-01 -3.56801271e-01 3.21343333e-01 -5.69584109e-02
-4.76857126e-01 7.46803045e-01 -9.94148254e-02 -2.26077959e-01
1.14581786e-01 -5.62202096e-01 -4.08635914e-01 -1.06327701e+00
1.54420047e-03 2.93458760e-01 4.55072135e-01 2.89901942e-01
3.13588142e-01 6.16612196e-01 -1.39200616e+00 1.28890917e-01
5.25397837e-01 7.96576202e-01 7.82497466e-01 6.35691047e-01
-2.88278684e-02 2.84936816e-01 -4.88955408e-01 1.42331883e-01
1.39035761e-01 -5.22346437e-01 -5.75029373e-01 -2.95551606e-02
1.53523996e-01 -5.04785717e-01 -3.77623253e-02 8.43581438e-01
8.52535307e-01 -2.01557651e-01 4.46788222e-01 -6.74562395e-01
-3.68761718e-01 -4.65570427e-02 -8.06818724e-01 5.57043850e-01
5.81877172e-01 3.18699211e-01 1.46309510e-01 -9.50621903e-01
7.34430373e-01 7.59338796e-01 2.34815180e-01 7.81601071e-01
-1.51037300e+00 -5.54544687e-01 -5.61131477e-01 8.73962119e-02
-1.12879431e+00 -6.38412893e-01 6.24518573e-01 -5.58537364e-01
6.51924253e-01 3.69149446e-01 7.40617394e-01 1.07760823e+00
4.66526717e-01 2.81256825e-01 1.66565156e+00 -4.94769961e-01
3.21062893e-01 2.75959015e-01 -1.85157031e-01 5.64824998e-01
4.40473676e-01 6.15816861e-02 -6.38058662e-01 -1.15919344e-01
1.18111837e+00 2.76951909e-01 -5.64079046e-01 -3.72921020e-01
-1.38114572e+00 2.98432469e-01 1.06337324e-01 5.30686021e-01
-3.32197100e-01 -1.16729543e-01 1.70729995e-01 1.42280132e-01
3.22058678e-01 4.26076144e-01 9.65010971e-02 -1.87534407e-01
-8.12341630e-01 -3.37886065e-01 5.12778759e-01 6.40901744e-01
5.19771397e-01 -1.69065207e-01 3.79247755e-01 5.09394169e-01
8.59637484e-02 4.03778374e-01 3.97134960e-01 -1.22951460e+00
-1.89024583e-01 3.94791245e-01 1.42547220e-01 -6.79814219e-01
-5.52390218e-01 -2.31228769e-01 -9.82434750e-01 7.51033843e-01
5.81233680e-01 1.65459290e-01 -6.40046954e-01 1.26459205e+00
6.01907969e-01 1.00383684e-01 -2.05320567e-01 1.19347084e+00
8.63319576e-01 4.18288410e-01 -2.87201166e-01 -5.43434083e-01
1.46246171e+00 -4.45534855e-01 -6.41755164e-01 1.80955112e-01
3.36047173e-01 -7.77004838e-01 1.28588402e+00 6.77518725e-01
-1.17399633e+00 -1.37968451e-01 -8.13405514e-01 -7.29619414e-02
-1.51136428e-01 -1.64856821e-01 4.65335429e-01 3.78644586e-01
-8.82098138e-01 4.74723905e-01 -1.16946602e+00 -3.59749764e-01
5.62140703e-01 2.47446314e-01 -3.86398643e-01 -1.95165128e-01
-5.13535738e-01 7.67419100e-01 -2.38792673e-01 1.24631799e-03
-6.88557446e-01 -9.53208327e-01 -2.37589553e-01 -3.84376734e-01
6.59637898e-02 -6.94212019e-01 7.39295185e-01 -3.19602370e-01
-1.70632696e+00 1.14083087e+00 -2.27882192e-01 -4.36417758e-03
3.27496231e-01 9.16607231e-02 -1.97099686e-01 6.55216455e-01
-1.05486341e-01 4.49046493e-01 5.88869154e-01 -1.31942248e+00
-2.00210407e-01 -6.46528125e-01 -1.63716733e-01 -3.41053084e-02
-1.76035821e-01 2.52998859e-01 -1.70825377e-01 -2.26213500e-01
2.39664942e-01 -8.79723489e-01 -2.47138277e-01 4.44306105e-01
-2.34247938e-01 2.48510301e-01 8.62200439e-01 -4.70226943e-01
7.16891408e-01 -2.00642252e+00 -1.36416778e-01 -9.71654281e-02
4.95090097e-01 1.35708347e-01 2.50209659e-01 5.52522540e-01
4.75065038e-02 -1.88602209e-02 -2.97337204e-01 -2.86104560e-01
-7.57434726e-01 -1.53636903e-01 1.71772346e-01 9.63771522e-01
-3.27260196e-01 6.32099867e-01 -7.02394664e-01 -5.01913786e-01
5.53599298e-01 8.32749009e-01 -4.96265404e-02 6.47783875e-02
2.36587495e-01 1.48293233e+00 -5.18560946e-01 7.74679661e-01
6.68292999e-01 -5.49166799e-01 1.97036713e-01 -2.65962273e-01
-5.33847570e-01 3.56292874e-02 -1.11855352e+00 1.68595171e+00
-3.29578310e-01 5.05667448e-01 4.94636238e-01 -8.19431841e-01
6.77215695e-01 4.63139147e-01 7.92819381e-01 -8.89492989e-01
1.13199636e-01 2.95572340e-01 -4.83231544e-02 -8.20513368e-01
-1.86642725e-02 -3.76697123e-01 4.38971877e-01 6.06090069e-01
-3.90361488e-01 -3.84657323e-01 3.50929163e-02 4.90276478e-02
1.08021176e+00 1.51919588e-01 2.55716473e-01 -2.62653142e-01
2.51355976e-01 1.46686539e-01 3.49640548e-01 4.69892472e-01
-5.38212880e-02 7.72379816e-01 4.85614911e-02 -4.93194073e-01
-1.28835022e+00 -9.54291999e-01 -6.82389140e-01 3.91566545e-01
2.79011190e-01 1.57933645e-02 -6.04246736e-01 1.43772900e-01
-3.31472784e-01 1.98951721e-01 -2.70163149e-01 3.72662514e-01
-5.16360223e-01 -7.44244277e-01 1.19641669e-01 4.07799035e-02
2.21983701e-01 -8.40535581e-01 -1.29211009e+00 9.15748477e-02
-3.90454717e-02 -1.15690494e+00 9.25398842e-02 7.07037598e-02
-1.15118229e+00 -1.03782892e+00 -8.84307146e-01 -3.24146569e-01
9.62843120e-01 5.22900522e-01 7.81480074e-01 1.50752971e-02
-8.65894139e-01 5.89909017e-01 -2.10279509e-01 -3.26229572e-01
-1.37408704e-01 -5.03034890e-01 3.79791319e-01 -2.40409076e-01
6.04757145e-02 -1.00860357e+00 -1.11809206e+00 3.01059961e-01
-9.50806558e-01 3.43535542e-01 6.79711223e-01 6.96092188e-01
9.94661331e-01 -2.41383061e-01 3.54839206e-01 -7.92337894e-01
3.02739084e-01 -1.98803127e-01 -8.65761936e-01 -1.07413776e-01
-5.32532990e-01 -3.82094175e-01 8.02122176e-01 -4.88884717e-01
-9.69006002e-01 -7.93220624e-02 1.76935241e-01 -4.38615292e-01
-5.20170569e-01 1.83830321e-01 2.07106143e-01 -3.51055533e-01
8.35597336e-01 2.92383045e-01 4.52168792e-01 -4.68215287e-01
-9.15240645e-02 7.02963352e-01 4.63326663e-01 -2.54059196e-01
4.62086916e-01 1.18902647e+00 2.96442270e-01 -1.14099991e+00
-4.11070704e-01 -5.05953252e-01 -6.19877577e-01 -3.68387967e-01
7.63670206e-01 -6.50654852e-01 -7.60029852e-01 2.59521663e-01
-7.66765594e-01 -3.64804834e-01 -2.55751997e-01 9.16628599e-01
-4.12516624e-01 4.05272394e-01 -4.99963164e-01 -7.82320738e-01
-3.59491706e-01 -1.08083868e+00 1.03940248e+00 2.98202932e-01
-2.18397900e-01 -7.88692415e-01 1.18301973e-01 6.62170410e-01
6.73050761e-01 4.83543545e-01 5.38183510e-01 1.45894200e-01
-8.28403294e-01 -1.05575688e-01 -2.70018494e-03 2.63378527e-02
2.09028110e-01 -8.28991979e-02 -1.11826396e+00 -3.97540092e-01
4.60964292e-01 -3.35049927e-01 4.52719092e-01 6.89474285e-01
1.39982557e+00 4.39094789e-02 -4.95938599e-01 7.43845761e-01
1.51234925e+00 2.64211178e-01 6.71184838e-01 2.12454885e-01
5.24158299e-01 7.19136834e-01 4.51578438e-01 4.69847858e-01
-2.03113481e-02 8.54420900e-01 3.95106524e-01 -2.55073875e-01
-2.40351066e-01 2.60654181e-01 2.95647252e-02 6.77688003e-01
-4.82420713e-01 5.93605451e-02 -9.04957354e-01 2.18606472e-01
-1.13313520e+00 -9.45660055e-01 -4.79566753e-01 2.42763376e+00
7.39179611e-01 -3.01237494e-01 2.19496246e-02 1.29762515e-01
4.77917314e-01 -2.91549236e-01 -7.05885231e-01 -7.75699839e-02
8.16424564e-02 5.56279384e-02 3.90281469e-01 3.32965821e-01
-4.73763764e-01 3.51480484e-01 5.85672569e+00 5.02010047e-01
-1.61884677e+00 2.84024417e-01 4.74560380e-01 -4.88155186e-01
-3.81431878e-01 -2.90246699e-02 -4.07402247e-01 6.40502810e-01
9.94178116e-01 5.59906997e-02 4.94793713e-01 3.07082117e-01
6.34705007e-01 -8.21427882e-01 -8.92541945e-01 1.28638852e+00
-1.73646331e-01 -1.38238323e+00 -3.51133645e-01 3.48416477e-01
2.95139313e-01 2.68189251e-01 5.44810016e-03 -7.37204731e-01
-3.10262382e-01 -8.60249460e-01 1.84981436e-01 6.48142517e-01
1.28668106e+00 -2.52011329e-01 3.86191636e-01 5.12747586e-01
-6.56458616e-01 1.98501006e-01 -3.80698293e-01 8.91902111e-03
4.24257934e-01 1.14118373e+00 -6.91624701e-01 2.56797016e-01
9.57746387e-01 2.29841143e-01 -2.73963779e-01 1.06877434e+00
6.01286106e-02 4.96454865e-01 -4.19391096e-01 7.45994672e-02
-3.65037024e-01 -5.45034170e-01 7.23240376e-01 8.41477156e-01
4.13021654e-01 4.70948488e-01 -7.15983286e-02 8.46925259e-01
1.96624056e-01 -1.95815369e-01 -7.51164019e-01 8.95416811e-02
5.35489738e-01 1.66186929e+00 -1.06660044e+00 -2.02910572e-01
-6.67719245e-01 8.31619799e-01 2.48875525e-02 1.89434558e-01
-5.03112733e-01 -8.90496224e-02 2.74171799e-01 6.35126531e-01
-1.32474989e-01 -4.44990963e-01 -2.78151304e-01 -1.24532330e+00
9.24868286e-02 -6.11554444e-01 7.62608424e-02 -9.05445874e-01
-9.13742661e-01 3.22843999e-01 -7.65332580e-02 -1.40286255e+00
2.03992659e-03 -5.39301515e-01 -4.83141541e-01 7.39587486e-01
-1.33659220e+00 -7.76879549e-01 -4.45015103e-01 3.96750510e-01
3.31038386e-01 2.39920482e-01 9.61586237e-01 2.85114110e-01
-3.73805314e-01 -1.22818217e-01 4.92657632e-01 -3.23415697e-01
6.48236275e-01 -9.26624477e-01 -1.41555041e-01 6.33243263e-01
-9.23554823e-02 5.41130900e-01 7.28794515e-01 -4.38802421e-01
-1.91926098e+00 -4.68716055e-01 1.68575376e-01 -4.37142670e-01
5.49560785e-01 -4.05989945e-01 -9.40833390e-01 3.16926271e-01
1.78437531e-01 3.92121643e-01 9.91899431e-01 -4.03248101e-01
1.15754351e-01 -1.65890511e-02 -1.58812284e+00 4.43798572e-01
9.27672565e-01 -3.57834816e-01 -7.45544210e-02 4.28663105e-01
5.40016368e-02 -5.55721223e-01 -1.10156512e+00 1.76711511e-02
9.13822472e-01 -1.29570901e+00 7.66953111e-01 2.38098040e-01
3.23757291e-01 -4.01714265e-01 1.71799317e-01 -9.68033552e-01
-2.40802258e-01 -4.66675252e-01 -8.33179951e-02 7.88410425e-01
1.14334054e-01 -9.50779915e-01 8.37077975e-01 5.49616575e-01
-1.44679591e-01 -9.04619157e-01 -1.08839512e+00 -9.09057379e-01
-2.18457669e-01 -2.38398880e-01 1.72746062e-01 8.84639561e-01
1.57743007e-01 1.49574563e-01 -3.37164663e-02 5.66153713e-02
1.02638447e+00 4.45266813e-01 5.73813736e-01 -1.10458589e+00
-2.20991537e-01 -2.06095606e-01 -5.14060438e-01 -7.44694471e-01
-4.13711786e-01 -7.21151710e-01 -2.79277593e-01 -1.60158050e+00
4.58555967e-01 -4.69872832e-01 2.17195809e-01 1.15125824e-03
3.42346311e-01 5.17062008e-01 -1.43789738e-01 6.83617413e-01
-3.82806897e-01 1.34025902e-01 1.53960752e+00 2.98334837e-01
-7.80017525e-02 -1.61468446e-01 -5.67420423e-01 5.97774446e-01
1.07909203e+00 -4.54348356e-01 -3.15812707e-01 -5.29045522e-01
-2.80714706e-02 2.96987683e-01 5.77528834e-01 -1.04462004e+00
5.43985963e-01 -1.06956273e-01 5.80132484e-01 -3.28306705e-01
6.30877137e-01 -9.04370487e-01 5.45519233e-01 5.76848805e-01
-1.89698860e-02 -1.40790492e-01 -1.62819520e-01 3.98021251e-01
4.25429121e-02 -1.95374861e-01 1.06196427e+00 -4.57993835e-01
-3.71067882e-01 1.78721711e-01 -5.12953758e-01 -2.24138051e-01
1.10374880e+00 -5.92154443e-01 -4.94546354e-01 -9.97540578e-02
-6.82010829e-01 -4.43676226e-02 8.12958300e-01 -2.56926298e-01
8.61145198e-01 -8.36013854e-01 -5.06087005e-01 1.27377465e-01
-1.92795787e-02 2.79702663e-01 5.54275692e-01 1.43277979e+00
-5.85829496e-01 2.51410365e-01 -3.66935998e-01 -8.07197928e-01
-1.45205057e+00 4.78017569e-01 3.03863317e-01 8.43049660e-02
-1.07465792e+00 5.21106064e-01 1.11532055e-01 -1.74945131e-01
-1.83479920e-01 -5.13785966e-02 -7.20784217e-02 -3.44759077e-01
8.94960046e-01 5.75183809e-01 5.64784482e-02 -3.37851703e-01
-3.28027636e-01 7.53426075e-01 6.35727197e-02 -1.48170859e-01
1.56290841e+00 -4.99439806e-01 -3.15387011e-01 6.38173699e-01
8.55070949e-01 1.10371031e-01 -1.07116544e+00 -1.12151913e-01
-3.21032614e-01 -6.38905704e-01 1.84799492e-01 -5.89515150e-01
-8.40651512e-01 1.08677089e+00 7.87338257e-01 2.45951697e-01
1.26339138e+00 2.55767733e-01 5.89006186e-01 4.95211035e-02
7.64208317e-01 -8.70834112e-01 -8.58166069e-02 -1.04321294e-01
7.01989651e-01 -1.24173057e+00 3.18117440e-01 -5.08093596e-01
-2.55830914e-01 1.16363573e+00 3.40720594e-01 1.57974541e-01
4.79838431e-01 8.36867929e-01 3.85718979e-02 -5.14782429e-01
-6.99648321e-01 1.06374636e-01 -2.26292193e-01 7.55268097e-01
6.58261836e-01 -5.39931357e-02 -3.38290542e-01 -2.61574592e-02
9.76258293e-02 3.28653514e-01 7.51311839e-01 1.18257654e+00
-3.39052290e-01 -9.18822110e-01 -5.66802323e-01 7.24622309e-01
-6.43284261e-01 1.41140684e-01 -1.39158545e-02 4.31551576e-01
-1.80341527e-01 8.33553851e-01 -3.09392400e-02 1.21429414e-01
6.37499616e-02 -1.79497004e-01 8.22460353e-01 -6.10645294e-01
-7.57841468e-02 2.60210127e-01 -1.91627532e-01 -6.80993676e-01
-7.41682172e-01 -6.66306376e-01 -1.34818053e+00 -3.61133128e-01
-2.73726702e-01 -1.74389388e-02 8.75477672e-01 5.37729681e-01
6.33633494e-01 3.96697670e-01 4.58262563e-01 -1.09272230e+00
4.81963083e-02 -8.11573207e-01 -8.11216593e-01 6.99297786e-02
4.01267767e-01 -5.91392636e-01 -3.87929738e-01 2.43571937e-01]
|
[12.913479804992676, -2.7983791828155518]
|
65b1b70c-2b66-4a88-a83a-3a7bf6b6329f
|
video-face-manipulation-detection-through
|
2004.07676
| null |
https://arxiv.org/abs/2004.07676v1
|
https://arxiv.org/pdf/2004.07676v1.pdf
|
Video Face Manipulation Detection Through Ensemble of CNNs
|
In the last few years, several techniques for facial manipulation in videos have been successfully developed and made available to the masses (i.e., FaceSwap, deepfake, etc.). These methods enable anyone to easily edit faces in video sequences with incredibly realistic results and a very little effort. Despite the usefulness of these tools in many fields, if used maliciously, they can have a significantly bad impact on society (e.g., fake news spreading, cyber bullying through fake revenge porn). The ability of objectively detecting whether a face has been manipulated in a video sequence is then a task of utmost importance. In this paper, we tackle the problem of face manipulation detection in video sequences targeting modern facial manipulation techniques. In particular, we study the ensembling of different trained Convolutional Neural Network (CNN) models. In the proposed solution, different models are obtained starting from a base network (i.e., EfficientNetB4) making use of two different concepts: (i) attention layers; (ii) siamese training. We show that combining these networks leads to promising face manipulation detection results on two publicly available datasets with more than 119000 videos.
|
['Nicolò Bonettini', 'Luca Bondi', 'Paolo Bestagini', 'Stefano Tubaro', 'Sara Mandelli', 'Edoardo Daniele Cannas']
|
2020-04-16
| null | null | null | null |
['localization-in-video-forgery', 'image-manipulation-detection', 'gan-image-forensics', 'video-forensics', 'fake-image-detection', 'detecting-image-manipulation']
|
['computer-vision', 'computer-vision', 'computer-vision', 'computer-vision', 'computer-vision', 'computer-vision']
|
[ 2.12635472e-01 7.13380799e-02 6.73535392e-02 -4.04846296e-02
-8.63377098e-03 -4.42430913e-01 8.62863958e-01 -2.73176432e-01
-4.57250953e-01 8.07924688e-01 -2.28733763e-01 2.62176059e-02
1.82085127e-01 -7.48312294e-01 -1.07438612e+00 -6.50478125e-01
-3.91959362e-02 7.98406452e-02 1.55647367e-01 -4.30007935e-01
4.75784868e-01 8.28729510e-01 -1.76356828e+00 1.77665770e-01
5.69470286e-01 1.17859483e+00 -1.07986048e-01 4.05245364e-01
7.95912296e-02 9.91665184e-01 -8.26879144e-01 -9.75092709e-01
2.49476001e-01 -4.18956220e-01 -5.35553515e-01 1.49782553e-01
8.33031774e-01 -6.89375699e-01 -3.32792699e-01 1.50442576e+00
1.72381222e-01 -1.67012885e-01 3.44075054e-01 -1.44462323e+00
-4.92599994e-01 3.26721191e-01 -6.54358327e-01 2.68449903e-01
4.50078815e-01 3.18969816e-01 2.88141072e-01 -8.08149695e-01
8.20791602e-01 1.50942409e+00 5.61969101e-01 6.86662138e-01
-8.55438590e-01 -1.07109928e+00 -3.25773388e-01 4.41846192e-01
-1.20949543e+00 -8.32360029e-01 7.35206366e-01 -5.23831427e-01
4.89996463e-01 1.94535315e-01 6.81521833e-01 1.39996564e+00
2.16618761e-01 5.20869374e-01 8.28166962e-01 -1.50758535e-01
8.90613068e-03 2.54661679e-01 -4.36007231e-01 8.93140256e-01
3.86766672e-01 8.50560591e-02 -6.17702484e-01 -1.09067075e-01
6.17690027e-01 -8.57794732e-02 -4.83923137e-01 -6.37196898e-02
-9.50548053e-01 8.37526262e-01 4.69339877e-01 4.51795518e-01
-2.86177903e-01 4.53018248e-01 6.99870169e-01 5.41742682e-01
3.96364421e-01 4.36258018e-01 9.61655576e-04 -8.34536105e-02
-9.20375466e-01 3.32261741e-01 5.68184495e-01 6.74076855e-01
4.18607712e-01 8.45771357e-02 1.29755571e-01 4.49602425e-01
1.07947245e-01 5.61277211e-01 4.17767644e-01 -6.30322099e-01
4.82947648e-01 1.00482680e-01 2.91176826e-01 -1.80168009e+00
-1.43547982e-01 1.92399248e-01 -8.75825405e-01 3.76342654e-01
6.04920447e-01 -1.08652286e-01 -6.31314158e-01 1.58180535e+00
3.83988917e-01 4.93517697e-01 -4.45895463e-01 8.48799288e-01
6.93643808e-01 4.97500211e-01 3.88987409e-03 -4.64348570e-02
1.31153357e+00 -9.68207836e-01 -9.88725305e-01 1.46353170e-01
3.78348023e-01 -7.34106302e-01 5.43763459e-01 7.64374137e-01
-1.03802657e+00 -4.41346765e-01 -9.81088519e-01 8.10558423e-02
-6.56369865e-01 8.69222134e-02 4.40780371e-01 1.07007563e+00
-1.00809002e+00 9.49852347e-01 -6.60312712e-01 -3.13819110e-01
9.90454912e-01 4.02865499e-01 -7.40471900e-01 8.72359052e-03
-1.18584538e+00 1.01746428e+00 2.54645944e-01 4.27034736e-01
-1.30387270e+00 -2.90840566e-01 -5.97542167e-01 9.49898362e-02
6.24491394e-01 -1.80128559e-01 8.28422964e-01 -1.58579397e+00
-1.46759057e+00 1.01206589e+00 2.69453794e-01 -4.15685624e-01
9.62800324e-01 -4.91388887e-01 -3.86953890e-01 4.62793887e-01
-2.57587343e-01 6.76905751e-01 1.72834682e+00 -1.06946611e+00
-3.36411893e-01 -3.30586761e-01 2.74089783e-01 -3.76487166e-01
-7.94769883e-01 5.52936435e-01 -7.00033680e-02 -6.92113280e-01
-5.18422902e-01 -8.28182697e-01 2.94120640e-01 5.31487107e-01
-4.44801867e-01 -4.93719652e-02 1.09820104e+00 -1.14281988e+00
1.00506473e+00 -2.06142640e+00 3.75840038e-01 -6.50415048e-02
3.80865246e-01 8.04814756e-01 -1.67969719e-01 2.87181646e-01
-1.87571749e-01 3.60199720e-01 -1.35968313e-01 -1.09637007e-01
-3.04145128e-01 -1.67174250e-01 -1.36571497e-01 8.69655609e-01
3.57095540e-01 8.31261694e-01 -9.91752982e-01 -3.63855541e-01
2.11117327e-01 5.57069242e-01 -6.61428392e-01 1.14455760e-01
-2.03918740e-01 5.16259313e-01 -2.74102300e-01 7.21436679e-01
8.27156425e-01 1.01584509e-01 9.79739875e-02 -1.45593271e-01
5.61840385e-02 -2.33545974e-01 -7.11272776e-01 1.31944907e+00
-2.94206291e-01 9.01102781e-01 1.67330921e-01 -9.09501255e-01
4.90256459e-01 4.10367221e-01 1.25336677e-01 -3.17312539e-01
6.57995105e-01 1.81220204e-01 3.49007435e-02 -8.44894826e-01
4.39552665e-01 -6.78268448e-03 2.72155762e-01 2.84331650e-01
2.31593594e-01 5.16719148e-02 1.92766920e-01 -9.25977454e-02
1.11331666e+00 -4.77227457e-02 2.87779212e-01 1.84809957e-02
8.17813993e-01 -3.46135676e-01 1.00103982e-01 4.56294507e-01
-4.44744408e-01 2.31497467e-01 7.05393970e-01 -4.93962675e-01
-9.99717236e-01 -6.15254164e-01 1.50300339e-01 8.19851637e-01
2.45336611e-02 -5.07369488e-02 -9.76413190e-01 -8.63397241e-01
1.55429378e-01 2.60344207e-01 -7.55253553e-01 -4.85903740e-01
-6.77399695e-01 -1.32093683e-01 9.20023143e-01 3.30497697e-02
7.84156322e-01 -1.42662859e+00 -7.46717870e-01 -6.11041747e-02
-1.20014019e-01 -1.11801541e+00 -1.34296671e-01 -4.79722291e-01
-4.74782825e-01 -1.09253395e+00 -8.49168181e-01 -5.68498075e-01
5.78220248e-01 3.60714912e-01 8.83721828e-01 4.86117095e-01
-4.14918482e-01 -3.96655267e-03 -4.71508473e-01 -3.17561477e-01
-5.73088109e-01 -2.70619184e-01 1.98696584e-01 3.39659959e-01
2.00539380e-01 -3.55037540e-01 -3.21490377e-01 1.67982355e-01
-1.08979046e+00 -3.42768133e-01 3.29723001e-01 7.58950174e-01
-4.12738979e-01 1.89746335e-01 4.69982952e-01 -8.29120159e-01
4.01270956e-01 -5.58022261e-01 -7.59192228e-01 9.20027271e-02
2.53478549e-02 -3.22555244e-01 7.45382428e-01 -5.98342240e-01
-8.39250684e-01 -1.82004139e-01 -2.12800875e-01 -9.31292057e-01
-3.10345441e-01 1.82980284e-01 -2.29763955e-01 -6.16824985e-01
4.85383898e-01 6.93924055e-02 2.47572809e-01 -3.05216223e-01
3.55448544e-01 6.10320687e-01 3.28983366e-01 -1.53492719e-01
1.07871616e+00 7.41907656e-01 3.49200293e-02 -1.05790496e+00
-5.17030776e-01 2.04844251e-02 -4.55700308e-01 -7.00616658e-01
9.03787315e-01 -6.29565537e-01 -9.63569403e-01 1.03902483e+00
-1.29916239e+00 8.24358035e-03 3.37814450e-01 1.20037876e-01
-2.58210629e-01 7.66568244e-01 -5.09819150e-01 -7.17957795e-01
3.78733687e-02 -1.02793264e+00 9.34541404e-01 8.39071497e-02
-4.60136216e-03 -7.01427817e-01 -4.09699529e-01 4.79132563e-01
4.18995470e-01 6.14745080e-01 6.09656274e-01 -3.49075198e-01
-6.72152221e-01 -2.83409834e-01 -3.09739202e-01 6.56105280e-01
2.72270739e-01 2.16460213e-01 -1.06511045e+00 -3.77976567e-01
9.27716345e-02 -5.99658191e-01 8.35865617e-01 -1.27846990e-02
1.39719939e+00 -5.39959967e-01 -2.89429367e-01 3.73762965e-01
1.27424157e+00 2.13238224e-01 1.03219831e+00 1.16254173e-01
6.63550496e-01 7.73464084e-01 5.89272082e-01 4.81601685e-01
-3.70039254e-01 7.66819835e-01 9.03669238e-01 3.96998495e-01
3.87697928e-02 9.66733508e-03 5.89683354e-01 4.72138405e-01
-3.73769760e-01 -4.38885540e-01 -6.52037203e-01 3.99407417e-01
-1.43891191e+00 -1.25385809e+00 3.31482664e-02 2.18295097e+00
4.63921309e-01 1.01242408e-01 3.66268642e-02 1.71305045e-01
9.63719428e-01 2.28264660e-01 -3.08422834e-01 -4.58996922e-01
1.05774663e-01 2.73300022e-01 3.61260027e-01 1.04520999e-01
-1.33405840e+00 9.95657504e-01 4.88020754e+00 8.34833682e-01
-1.38968194e+00 3.65011215e-01 4.90972489e-01 -1.33448258e-01
2.97787517e-01 -5.10743797e-01 -5.26105762e-01 9.26861167e-01
7.82598913e-01 1.25184149e-01 6.14446819e-01 7.87105680e-01
2.10227132e-01 -9.12987068e-02 -9.61299717e-01 1.19536614e+00
6.47533774e-01 -1.29205096e+00 7.15267733e-02 4.22941037e-02
5.28904080e-01 -3.39543492e-01 2.01149046e-01 1.13399856e-01
-1.34056240e-01 -1.22574782e+00 8.34805310e-01 4.21253741e-01
9.19507504e-01 -8.67672384e-01 7.69286633e-01 2.86867708e-01
-6.24802113e-01 -1.88792408e-01 -3.52519065e-01 8.57620686e-02
1.92963257e-01 5.38620293e-01 -4.22789782e-01 4.15654778e-01
7.05418289e-01 7.29888916e-01 -3.36322129e-01 9.73515391e-01
-3.52931172e-01 3.84729832e-01 -2.26597533e-01 -3.74721475e-02
2.00340271e-01 -1.41712964e-01 6.48033857e-01 8.24103475e-01
3.37599903e-01 -2.59695917e-01 -4.80090320e-01 7.18691409e-01
-4.99018520e-01 -5.14483787e-02 -9.53133047e-01 -4.69372839e-01
3.93483713e-02 1.32879531e+00 -7.09956825e-01 -2.02413157e-01
-4.22478139e-01 1.20467055e+00 3.46697986e-01 -8.62295702e-02
-1.28736162e+00 -3.47159475e-01 7.63942719e-01 2.86500096e-01
4.78664279e-01 -6.09580353e-02 3.98520678e-01 -1.27859223e+00
2.10557831e-03 -1.13244998e+00 -1.65610909e-01 -7.04542220e-01
-1.17620003e+00 4.46790665e-01 -2.53762275e-01 -1.04901052e+00
-8.41238722e-02 -8.11815083e-01 -2.88486719e-01 4.62180555e-01
-1.37012434e+00 -7.78460920e-01 -5.44953525e-01 5.24851382e-01
4.20793146e-01 -2.32494667e-01 4.40660298e-01 7.13933527e-01
-5.47596455e-01 6.09712660e-01 -1.35467783e-01 3.64407688e-01
6.86807096e-01 -5.78203797e-01 3.94614279e-01 6.41269982e-01
4.86848727e-02 3.96356076e-01 6.22602999e-01 -5.80425799e-01
-1.41299689e+00 -1.09418929e+00 6.26710594e-01 -3.85819674e-01
7.65618384e-01 -4.12278086e-01 -8.02202523e-01 5.90016246e-01
1.29825473e-01 -4.07549180e-02 5.71828149e-02 -3.57017875e-01
-3.42752308e-01 1.68486908e-01 -1.51197100e+00 5.41813433e-01
9.45171118e-01 -4.48614985e-01 -3.43602329e-01 5.50758839e-01
2.14542538e-01 -1.78275019e-01 -4.13200945e-01 1.69613689e-01
7.75787830e-01 -1.23708737e+00 8.40099037e-01 -8.34295392e-01
8.97156119e-01 -4.06017303e-02 3.51464540e-01 -1.17106795e+00
2.56202221e-02 -6.42677963e-01 -2.77937353e-01 9.31155682e-01
-1.90984309e-01 -6.54284775e-01 7.93838203e-01 1.02570087e-01
2.54442990e-01 -4.65646625e-01 -1.11048937e+00 -8.64435136e-01
-1.52084917e-01 -4.21641618e-02 3.36134642e-01 1.17980552e+00
-3.29862982e-01 -1.24926873e-01 -8.84692311e-01 1.04448445e-01
4.92211550e-01 -4.41763490e-01 9.24549639e-01 -1.04710984e+00
-1.29411533e-01 -4.89013314e-01 -1.03736532e+00 -8.22874248e-01
4.41520721e-01 -5.83453596e-01 -1.15732424e-01 -6.56638384e-01
8.27738121e-02 -4.02054116e-02 7.17096031e-02 5.06268501e-01
7.17352927e-02 6.51428521e-01 3.70659739e-01 -4.17872109e-02
-2.26853490e-01 4.20025587e-01 1.27369916e+00 -1.72850177e-01
3.65518719e-01 -2.24223942e-01 -1.85324788e-01 8.55931640e-01
8.24819326e-01 -5.44849992e-01 4.05515991e-02 -2.65681714e-01
3.82231444e-01 1.26793468e-02 7.59028614e-01 -1.21244776e+00
-1.88250095e-01 1.56374291e-01 2.60376871e-01 -1.03795193e-01
4.73165810e-01 -9.34802115e-01 3.05633526e-02 7.15460420e-01
-1.57226715e-02 -1.94697812e-01 1.33016184e-01 6.74308658e-01
-2.60750115e-01 -5.12206852e-01 1.11786342e+00 -2.06899956e-01
-5.98429918e-01 3.36543992e-02 -3.90383989e-01 -4.25826102e-01
1.28588116e+00 6.17194623e-02 -3.81395578e-01 -5.55768907e-01
-6.38811588e-01 -1.40214831e-01 3.50995094e-01 6.26102865e-01
5.61089456e-01 -1.17189944e+00 -5.97637475e-01 9.44970325e-02
-7.23497346e-02 -5.98547399e-01 2.20444143e-01 8.45624864e-01
-9.46093917e-01 2.24240810e-01 -6.55602038e-01 -3.35213065e-01
-1.50405717e+00 8.64081502e-01 2.55600274e-01 1.38862640e-01
-3.91936570e-01 1.00714123e+00 3.06237824e-02 -1.12559617e-01
9.64317247e-02 -1.35049209e-01 -2.15106487e-01 2.93124378e-01
8.02462041e-01 5.44511735e-01 3.03927045e-02 -9.38028872e-01
-2.59237468e-01 1.95029721e-01 -1.65715128e-01 4.08115387e-01
1.19668579e+00 1.46069199e-01 -4.85670358e-01 1.02130905e-01
1.13721323e+00 4.65297475e-02 -9.79144037e-01 3.56590390e-01
-1.43168330e-01 -8.99331629e-01 -4.84137237e-02 -4.97481287e-01
-1.45605767e+00 9.16361153e-01 5.20274818e-01 3.11072469e-01
9.75058496e-01 -2.50811309e-01 8.99302661e-01 4.55668420e-01
6.42107844e-01 -8.19027185e-01 3.74507844e-01 1.35486886e-01
1.04383516e+00 -1.28190243e+00 -2.35360283e-02 -6.16194844e-01
-2.60986239e-01 1.26756537e+00 3.92071277e-01 -2.98140854e-01
6.82841003e-01 6.75991401e-02 -1.49822414e-01 -3.07876647e-01
-3.10822040e-01 9.64895785e-02 -7.62191862e-02 4.58466172e-01
8.48388523e-02 -1.14181884e-01 -3.27583909e-01 -1.26854196e-01
-2.81216819e-02 2.12491259e-01 7.65993476e-01 7.95117438e-01
-3.92928511e-01 -6.72926784e-01 -5.76041698e-01 5.10709703e-01
-6.64181054e-01 1.23431735e-01 -4.50048536e-01 9.06498432e-01
3.37960899e-01 8.63298357e-01 -1.48084536e-01 -2.42609575e-01
-8.18442702e-02 -2.31314059e-02 6.92089915e-01 -3.47686768e-01
-7.60210514e-01 -4.32748854e-01 3.12474556e-02 -8.43135595e-01
-6.05954051e-01 -6.14935458e-01 -4.66452241e-01 -6.58185899e-01
-4.74410146e-01 -3.65930378e-01 7.32429564e-01 8.81931007e-01
5.03057428e-02 2.41303369e-01 6.55521274e-01 -1.17058051e+00
-5.70525527e-01 -1.05121672e+00 -6.06249869e-01 5.83092570e-01
2.36587122e-01 -1.03147280e+00 -5.00191331e-01 1.51640058e-01]
|
[12.60118293762207, 1.1074293851852417]
|
43a6b002-647f-4b8e-8955-74b782ca0500
|
online-to-pac-conversions-generalization
|
2305.19674
| null |
https://arxiv.org/abs/2305.19674v1
|
https://arxiv.org/pdf/2305.19674v1.pdf
|
Online-to-PAC Conversions: Generalization Bounds via Regret Analysis
|
We present a new framework for deriving bounds on the generalization bound of statistical learning algorithms from the perspective of online learning. Specifically, we construct an online learning game called the "generalization game", where an online learner is trying to compete with a fixed statistical learning algorithm in predicting the sequence of generalization gaps on a training set of i.i.d. data points. We establish a connection between the online and statistical learning setting by showing that the existence of an online learning algorithm with bounded regret in this game implies a bound on the generalization error of the statistical learning algorithm, up to a martingale concentration term that is independent of the complexity of the statistical learning method. This technique allows us to recover several standard generalization bounds including a range of PAC-Bayesian and information-theoretic guarantees, as well as generalizations thereof.
|
['Gergely Neu', 'Gábor Lugosi']
|
2023-05-31
| null | null | null | null |
['generalization-bounds']
|
['methodology']
|
[-1.04213424e-01 2.88205445e-01 -7.17951804e-02 -2.94412971e-01
-1.18675804e+00 -8.45439374e-01 -8.81459787e-02 3.82841855e-01
-6.21588051e-01 6.42058909e-01 -5.59656739e-01 -4.92457598e-01
-7.13695824e-01 -7.09386408e-01 -1.03462648e+00 -9.54003513e-01
-5.36687553e-01 4.66534674e-01 3.74999344e-01 1.55432075e-01
2.57526308e-01 3.02838594e-01 -1.54733348e+00 -3.15889180e-01
7.38288879e-01 1.26509333e+00 1.67926386e-01 1.17158103e+00
-1.45292193e-01 2.45568946e-01 -3.52493554e-01 -6.10069811e-01
6.86685145e-01 -7.84646749e-01 -7.12065041e-01 1.37041032e-01
4.84101564e-01 -1.83390632e-01 -4.37281728e-01 1.42411149e+00
4.87312168e-01 3.75621736e-01 7.26627648e-01 -1.21834242e+00
-2.75077522e-01 8.50066423e-01 -5.54188311e-01 3.08673292e-01
2.20396742e-01 -4.36919421e-01 1.30108380e+00 -3.19229692e-01
3.09183091e-01 9.43400145e-01 6.29347086e-01 7.94857144e-01
-1.13626528e+00 -5.52205443e-01 6.30691350e-02 1.61214694e-01
-1.13585174e+00 1.99230351e-02 2.99098611e-01 -4.44606215e-01
2.81514019e-01 5.28293177e-02 4.16717052e-01 6.48478925e-01
2.62868345e-01 1.00017059e+00 8.57806504e-01 -6.76882327e-01
8.99432421e-01 1.49391174e-01 2.26370573e-01 7.72148848e-01
4.56300408e-01 4.83179420e-01 -6.55896842e-01 -2.12962598e-01
4.88825262e-01 4.40010689e-02 -3.07891015e-02 -7.87093580e-01
-1.99122876e-01 9.89102066e-01 1.67722642e-01 -1.37325786e-02
-4.89045009e-02 1.23064689e-01 3.44999582e-01 7.06030309e-01
5.41206241e-01 1.20633066e-01 -6.41501069e-01 -1.87043354e-01
-4.67952490e-01 2.83047318e-01 1.15788007e+00 9.77028072e-01
7.93535352e-01 -2.98880816e-01 1.76228270e-01 5.72981298e-01
1.35731518e-01 7.00285256e-01 1.66981623e-01 -8.90739143e-01
4.97592419e-01 1.33329378e-02 3.25637966e-01 -3.46926272e-01
-1.90756172e-01 -4.93159324e-01 -3.82424086e-01 2.11195037e-01
8.77199113e-01 -4.27103639e-01 -2.59752631e-01 1.93273461e+00
4.38233912e-01 1.73896864e-01 5.98179474e-02 5.54736376e-01
-2.00564727e-01 4.45581287e-01 -2.66450137e-01 -4.96283412e-01
5.41242003e-01 -3.15577358e-01 -5.28903008e-01 -1.75020009e-01
9.78628218e-01 3.21688736e-03 9.12936151e-01 6.68474078e-01
-1.01110077e+00 -2.93609221e-02 -9.98597145e-01 2.33250171e-01
-2.30540156e-01 -6.47539914e-01 6.34645641e-01 1.10267580e+00
-8.83881569e-01 8.92962217e-01 -8.87988925e-01 -2.02031642e-01
4.55388427e-01 2.50246733e-01 -1.98796570e-01 -1.83529720e-01
-6.95954740e-01 3.34654301e-01 7.12527275e-01 -1.80217668e-01
-1.11756432e+00 -7.87622750e-01 -6.27402961e-01 -2.84020025e-02
7.07136393e-01 -4.48531926e-01 1.58798146e+00 -1.16677487e+00
-1.48983788e+00 7.34523654e-01 2.36600861e-02 -7.63371468e-01
8.16393495e-01 -2.96586841e-01 2.16538399e-01 -2.14205869e-02
-1.18923604e-01 -1.89147830e-01 5.87052226e-01 -7.39046335e-01
-1.11890113e+00 -9.32817996e-01 3.86361927e-01 1.65288016e-01
-2.97534853e-01 -3.15481603e-01 -2.74050236e-03 -8.57714936e-02
5.17099202e-02 -8.64938200e-01 -3.81741911e-01 1.90446414e-02
-6.59116451e-03 -4.35743749e-01 2.74317592e-01 -1.79134548e-01
9.49339867e-01 -2.35494304e+00 3.17777805e-02 3.29855859e-01
-7.28134215e-02 -3.59947741e-01 3.45190503e-02 3.21575314e-01
1.49133772e-01 2.06096694e-01 -8.12622011e-02 -1.37564152e-01
1.68478966e-01 1.48770407e-01 -4.21064019e-01 7.47221589e-01
-4.95651335e-01 5.81189573e-01 -9.79209185e-01 -4.26439308e-02
-2.45618403e-01 -4.69329476e-01 -8.93803298e-01 2.99617112e-01
-3.19807529e-01 3.31998281e-02 -5.65392077e-01 5.70574626e-02
7.87484884e-01 -2.43535131e-01 3.04663807e-01 9.80955184e-01
2.29436755e-01 1.09916836e-01 -1.46035492e+00 1.53124094e+00
-5.76511562e-01 4.07760978e-01 2.63105631e-01 -1.07990801e+00
4.29773510e-01 6.49748966e-02 2.40643904e-01 -1.38926253e-01
2.26106927e-01 3.07895809e-01 -1.80089682e-01 -2.32656196e-01
1.47091210e-01 -6.21845126e-01 -2.33751297e-01 5.88628888e-01
4.06684399e-01 3.76322538e-01 9.20097828e-02 5.72875142e-02
1.07557857e+00 -1.24857277e-01 1.13164090e-01 -5.20513952e-01
2.86939293e-01 -4.21589494e-01 4.22113270e-01 1.59069157e+00
-9.66112912e-02 -1.22679830e-01 7.09063888e-01 -1.49693996e-01
-9.92973685e-01 -1.37494171e+00 -9.54822376e-02 1.69143784e+00
4.61398959e-02 -2.93619245e-01 -8.85632932e-01 -8.12443435e-01
3.52260619e-01 4.27558422e-01 -7.95633256e-01 -1.75752848e-01
3.05040210e-01 -4.73890901e-01 7.88429379e-02 3.66120130e-01
2.57305533e-01 -3.40398282e-01 -3.89317155e-01 1.35874618e-02
4.39931780e-01 -9.48748171e-01 -5.19106805e-01 5.08732796e-01
-1.20241773e+00 -1.35388124e+00 -2.74233460e-01 -3.82688224e-01
3.42423320e-01 4.39423084e-01 6.66013300e-01 -1.62124828e-01
2.39452757e-02 7.44962037e-01 -2.05884352e-01 -9.34099734e-01
-5.28949440e-01 1.77850768e-01 2.05620229e-01 -1.19668068e-02
3.59915316e-01 -4.06081408e-01 -3.85713845e-01 5.31402342e-02
-8.78369987e-01 -3.26412201e-01 3.26927125e-01 7.43949592e-01
3.94634604e-01 2.60559738e-01 5.71173251e-01 -1.03163505e+00
3.80888671e-01 -5.62799394e-01 -1.50716841e+00 2.86922514e-01
-6.51037753e-01 3.35147947e-01 9.35604393e-01 -3.82407159e-01
-6.89333260e-01 1.97462037e-01 3.21233898e-01 -2.58341074e-01
5.57294823e-02 4.78860706e-01 -3.21472853e-01 -1.69390753e-01
7.94640720e-01 4.52358931e-01 6.25013933e-02 -5.16033888e-01
3.73790771e-01 6.08699679e-01 4.67695326e-01 -7.86286592e-01
8.72556388e-01 2.46618718e-01 3.17387551e-01 -6.49490535e-01
-1.50062203e+00 -6.03752315e-01 -6.32402360e-01 -1.14014730e-01
4.35653687e-01 -6.73554182e-01 -1.15953171e+00 4.45793897e-01
-7.11167216e-01 -7.13946879e-01 -4.32207793e-01 5.82862318e-01
-1.08638096e+00 3.18176538e-01 -2.43560508e-01 -1.53837752e+00
3.23278517e-01 -3.85797143e-01 4.80812907e-01 4.22520250e-01
6.75085843e-01 -1.28041446e+00 1.54828429e-01 1.59012303e-01
-3.50626931e-02 -2.24345103e-02 6.66874468e-01 -1.08975422e+00
-4.60768580e-01 -4.88676488e-01 3.53690386e-02 4.70713735e-01
-2.96677560e-01 -2.71411449e-01 -8.84411335e-01 -3.61792594e-01
1.50247648e-01 -3.34153533e-01 7.24603951e-01 5.40470421e-01
1.53939140e+00 -6.18404567e-01 -8.27022865e-02 4.82066542e-01
1.71591222e+00 1.31660789e-01 2.35861868e-01 1.49420291e-01
2.09687278e-01 3.85225534e-01 5.00404894e-01 8.00096810e-01
-1.19308516e-01 1.88685507e-01 7.46993721e-02 8.69224548e-01
7.98911512e-01 -6.46571636e-01 4.57624108e-01 3.60528886e-01
3.03456753e-01 -3.02542392e-02 -7.70363212e-01 3.04010123e-01
-1.78462708e+00 -9.53593910e-01 3.63258958e-01 3.17058444e+00
1.08524990e+00 1.69331402e-01 5.84525406e-01 1.15468301e-01
7.92798281e-01 -2.99202144e-01 -9.41217303e-01 -5.34706116e-01
6.89562857e-02 3.94490391e-01 1.19327319e+00 7.92322099e-01
-9.95078743e-01 6.07451439e-01 7.24151468e+00 1.09774280e+00
-6.20695353e-01 3.21036816e-01 5.69919944e-01 -1.62701178e-02
-9.66372192e-02 -3.88073511e-02 -8.86437833e-01 2.91495264e-01
1.43068635e+00 -7.26518095e-01 7.52567947e-01 1.37227201e+00
8.12799297e-03 -3.10039252e-01 -1.39461493e+00 9.16789353e-01
-3.40724885e-01 -1.19150615e+00 -3.48877221e-01 5.30186176e-01
8.02782953e-01 -8.80131572e-02 -8.68297294e-02 3.48708957e-01
5.61715126e-01 -7.49044359e-01 5.97935200e-01 2.95999795e-01
4.80690002e-01 -1.10958338e+00 6.81891859e-01 8.17976773e-01
-7.89938748e-01 -6.19377434e-01 -5.04868031e-01 -2.31029749e-01
-4.94457990e-01 4.74238127e-01 -4.49541003e-01 5.39417088e-01
3.83226991e-01 4.08244342e-01 -2.58985788e-01 1.46546173e+00
4.94449446e-03 6.96235240e-01 -6.13937497e-01 -4.33409631e-01
1.15039624e-01 -5.53379297e-01 2.92400658e-01 9.96783912e-01
1.69266984e-01 2.71475643e-01 6.53776601e-02 4.38804358e-01
-3.28671426e-01 2.82917172e-01 -6.52761042e-01 -7.83562511e-02
5.21730840e-01 7.52748549e-01 -3.96678656e-01 3.79221775e-02
-4.08836842e-01 8.57800901e-01 4.17736053e-01 3.12438677e-03
-6.30077362e-01 -4.18625891e-01 4.51125175e-01 -5.69657981e-02
6.18794024e-01 -3.63396972e-01 -3.05581152e-01 -8.80746424e-01
1.95305854e-01 -4.08334136e-01 8.43186140e-01 -1.69105940e-02
-1.52776444e+00 -3.05985987e-01 -7.58895352e-02 -8.27973783e-01
-2.14697510e-01 -8.29178572e-01 -5.04515231e-01 5.13808668e-01
-1.00475883e+00 -1.73704788e-01 1.01443432e-01 4.41442728e-01
1.71597570e-01 -1.57690838e-01 5.74082792e-01 -2.20820263e-01
-3.39836687e-01 9.90339696e-01 9.90613997e-01 2.75908150e-02
1.45037800e-01 -1.85111678e+00 -2.32631028e-01 8.03893089e-01
3.50099146e-01 4.92807291e-02 8.09806108e-01 -1.28444716e-01
-1.76711810e+00 -8.32554579e-01 4.32552755e-01 -2.93714434e-01
1.18047214e+00 -5.21058857e-01 -7.39697516e-01 6.95009470e-01
-5.37554502e-01 1.48106039e-01 9.97269869e-01 6.38298273e-01
-5.36464155e-01 -3.30720127e-01 -1.17926431e+00 3.42567176e-01
1.08108139e+00 -7.50557005e-01 -3.66151869e-01 5.16218960e-01
5.66418469e-01 -4.51219887e-01 -7.59905577e-01 -3.46001416e-01
6.41421318e-01 -8.52089465e-01 3.81858140e-01 -1.21274149e+00
1.24245457e-01 3.32579374e-01 -3.19487840e-01 -9.63774323e-01
4.67234366e-02 -1.21380854e+00 -2.17112243e-01 6.27657413e-01
5.78294516e-01 -8.08014631e-01 9.98680234e-01 7.14047790e-01
3.55366796e-01 -7.40655303e-01 -1.33120358e+00 -1.53023303e+00
6.47208035e-01 -7.77911961e-01 -1.15712974e-02 3.49891722e-01
3.60978514e-01 5.60615361e-02 -1.12575836e-01 3.85457426e-01
9.45092857e-01 -7.00322390e-02 7.18147814e-01 -1.34230566e+00
-7.52304196e-01 -4.73207235e-01 -7.09355533e-01 -1.38327348e+00
3.69813830e-01 -8.82028162e-01 2.14316621e-01 -7.46497929e-01
5.72670937e-01 -3.88327152e-01 -6.18064106e-01 6.45535765e-04
1.93567928e-02 -5.53467333e-01 9.04138908e-02 -2.49186352e-01
-1.02286041e+00 4.16429847e-01 7.93189526e-01 4.51032907e-01
-2.20610082e-01 6.67894661e-01 -7.23701537e-01 4.90088493e-01
6.99038804e-01 -7.63544798e-01 -3.91900510e-01 9.93726552e-02
4.58843052e-01 3.20313931e-01 1.42115250e-01 -7.56360471e-01
4.46537435e-01 -3.75500917e-01 -1.14329711e-01 -2.57363439e-01
-3.21613491e-01 -5.31674623e-01 -4.13309842e-01 7.70900011e-01
-9.63665307e-01 -4.66186345e-01 1.75466221e-02 1.36241972e+00
4.10436869e-01 -7.18093753e-01 9.84622478e-01 1.75141841e-01
-2.24505872e-01 7.09315240e-01 -3.89082164e-01 6.81605756e-01
9.63921309e-01 -3.82337123e-02 2.42458254e-01 -6.01406097e-01
-7.79453397e-01 2.00500861e-01 1.71754181e-01 -6.19596727e-02
2.30808809e-01 -9.94321704e-01 -5.69835007e-01 -3.07508558e-02
5.39644286e-02 -5.00288084e-02 2.70741522e-01 5.75161994e-01
-1.68087706e-01 2.58745104e-01 2.20279068e-01 -2.73145765e-01
-8.88703585e-01 7.69796252e-01 6.50585055e-01 -5.44719100e-02
-3.93349200e-01 7.97431946e-01 1.64701536e-01 -1.00937240e-01
6.51004016e-01 -1.70491617e-02 6.87243283e-01 -2.12125778e-01
7.84966230e-01 5.45264959e-01 -3.01501602e-02 3.56450766e-01
-6.29356429e-02 1.08681113e-01 -8.30786899e-02 -4.40604985e-01
1.11323059e+00 -3.52009624e-01 2.42343724e-01 1.09290588e+00
1.34673309e+00 -4.39472385e-02 -1.46905029e+00 -7.69315302e-01
6.12246513e-01 -5.44906259e-01 -1.97148826e-02 -5.48875451e-01
-8.03070784e-01 7.66876578e-01 8.26826155e-01 5.35057366e-01
1.01896632e+00 2.98965573e-01 4.19208705e-01 8.28055143e-01
7.62368262e-01 -1.39495575e+00 -1.48728192e-01 6.21932030e-01
3.94604653e-01 -1.23778498e+00 -2.51586109e-01 -2.08088215e-02
-1.25922650e-01 1.48173845e+00 2.27920581e-02 -4.38258052e-01
1.04947639e+00 1.42642155e-01 -4.59798574e-01 2.10564688e-01
-9.73683596e-01 -2.47548684e-01 1.07236281e-01 4.51579988e-01
3.49948928e-02 2.77540624e-01 -2.03869790e-01 9.42397118e-01
-2.79968798e-01 -2.76044384e-03 6.25702202e-01 6.60064161e-01
-1.03124774e+00 -9.26545680e-01 1.25115439e-01 3.12745303e-01
-7.70774245e-01 1.17503077e-01 -2.82179266e-01 6.24208570e-01
-4.07937109e-01 8.10352802e-01 8.90802294e-02 -1.98684320e-01
-1.10471159e-01 2.89000273e-01 8.07822049e-01 -7.24577487e-01
1.36293620e-01 -1.53890699e-01 -4.16043133e-01 -4.76862848e-01
-7.75690079e-02 -7.85361528e-01 -9.39474344e-01 -4.86277312e-01
-4.99894768e-01 5.21298170e-01 6.59116805e-01 1.34520483e+00
9.72595182e-04 -3.55685383e-01 1.32166052e+00 -2.03398958e-01
-1.49592781e+00 -6.72968447e-01 -1.34304452e+00 -3.56285386e-02
2.20736802e-01 -2.74866164e-01 -1.00783730e+00 -3.40345740e-01]
|
[4.848182678222656, 3.5158028602600098]
|
3f7b0f10-fdcc-4936-a490-5f62d3ee56f0
|
discord-questions-a-computational-approach-to
|
2211.05007
| null |
https://arxiv.org/abs/2211.05007v1
|
https://arxiv.org/pdf/2211.05007v1.pdf
|
Discord Questions: A Computational Approach To Diversity Analysis in News Coverage
|
There are many potential benefits to news readers accessing diverse sources. Modern news aggregators do the hard work of organizing the news, offering readers a plethora of source options, but choosing which source to read remains challenging. We propose a new framework to assist readers in identifying source differences and gaining an understanding of news coverage diversity. The framework is based on the generation of Discord Questions: questions with a diverse answer pool, explicitly illustrating source differences. To assemble a prototype of the framework, we focus on two components: (1) discord question generation, the task of generating questions answered differently by sources, for which we propose an automatic scoring method, and create a model that improves performance from current question generation (QG) methods by 5%, (2) answer consolidation, the task of grouping answers to a question that are semantically similar, for which we collect data and repurpose a method that achieves 81% balanced accuracy on our realistic test set. We illustrate the framework's feasibility through a prototype interface. Even though model performance at discord QG still lags human performance by more than 15%, generated questions are judged to be more interesting than factoid questions and can reveal differences in the level of detail, sentiment, and reasoning of sources in news coverage.
|
['Caiming Xiong', "Xiang 'Anthony' Chen", "Lidiya Murakhovs'ka", 'Chien-Sheng Wu', 'Philippe Laban']
|
2022-11-09
| null | null | null | null |
['question-generation']
|
['natural-language-processing']
|
[-8.72753337e-02 6.74938321e-01 -9.94363353e-02 -2.61647850e-01
-1.86830521e+00 -1.06176388e+00 8.39447558e-01 6.37702227e-01
-1.54443756e-01 9.12045956e-01 1.17601871e+00 -3.22780371e-01
-6.82238489e-02 -7.83433914e-01 -6.35544181e-01 1.44854724e-01
5.61879992e-01 7.67547846e-01 4.99343097e-01 -7.92483926e-01
5.69030404e-01 -2.52171576e-01 -1.82406950e+00 1.14331317e+00
1.33800280e+00 9.82909441e-01 -7.90284351e-02 6.93311930e-01
-8.30269575e-01 1.22143149e+00 -1.22878265e+00 -8.32168341e-01
-8.28499794e-02 -9.91069317e-01 -1.38256192e+00 -1.21717066e-01
6.77242815e-01 3.46258096e-02 5.25494337e-01 8.83339763e-01
4.73205268e-01 -6.30963594e-04 6.29180193e-01 -9.06159043e-01
-8.68052185e-01 1.16037881e+00 1.20372735e-01 4.63573903e-01
1.01151192e+00 -3.46532881e-01 1.43876302e+00 -7.70152807e-01
9.46704805e-01 1.26888669e+00 6.69420302e-01 3.96278262e-01
-1.01299882e+00 -1.70474946e-01 6.75840070e-03 9.91637856e-02
-7.42734671e-01 -5.26007652e-01 3.31706792e-01 -4.77038115e-01
7.62754500e-01 8.44864309e-01 4.34816033e-01 9.66844201e-01
-6.52162284e-02 7.59739876e-01 1.19110465e+00 -2.62374401e-01
4.07777607e-01 5.79056025e-01 2.02478319e-01 1.52014807e-01
1.71007752e-01 -5.95276117e-01 -7.04565406e-01 -4.62023318e-01
-3.18566412e-02 -3.83392543e-01 -4.97241199e-01 4.49571431e-01
-1.23194432e+00 1.07357621e+00 4.98337030e-01 9.31086093e-02
-3.39543909e-01 -4.53042030e-01 2.35947728e-01 6.20188415e-01
7.79453278e-01 1.28695118e+00 -4.64850813e-01 -1.20322146e-01
-9.61895704e-01 1.06313431e+00 1.59848177e+00 1.00077915e+00
3.49243939e-01 -5.68459868e-01 -8.12016964e-01 8.61018777e-01
-1.56724490e-02 4.11684871e-01 6.75836802e-01 -1.02246892e+00
6.14867985e-01 9.16270614e-01 6.87906504e-01 -1.11152339e+00
-2.76446134e-01 -8.00236642e-01 -1.18521839e-01 -2.97770370e-02
6.59031570e-01 -3.23195696e-01 -3.06611121e-01 1.26965022e+00
2.89464772e-01 -7.09452152e-01 1.46159753e-01 7.96824276e-01
1.42045188e+00 8.83171022e-01 -1.35567516e-01 -1.31806906e-03
1.79281306e+00 -9.88884032e-01 -6.81710243e-01 -1.56735718e-01
4.12066281e-01 -1.24898398e+00 1.29195118e+00 1.46890908e-01
-1.41436744e+00 -5.49893141e-01 -7.42772460e-01 -2.23348498e-01
-4.11807805e-01 -7.21030608e-02 9.67619270e-02 4.83885646e-01
-9.97491837e-01 3.62180382e-01 1.24990724e-01 -3.63398790e-01
8.07145089e-02 -5.01020312e-01 2.76038349e-01 1.71847895e-01
-1.33056164e+00 9.91733134e-01 4.19833437e-02 -8.87450457e-01
-2.82640874e-01 -1.02970850e+00 -4.71666336e-01 1.07563086e-01
3.99190128e-01 -9.62160110e-01 1.69537961e+00 -8.55119467e-01
-1.04596412e+00 7.98492670e-01 -3.50852609e-01 -5.17240107e-01
6.02156281e-01 -2.97135741e-01 -6.41513765e-01 2.16085345e-01
8.19932222e-01 5.20795524e-01 4.51686978e-01 -1.30122137e+00
-9.43094671e-01 4.27743196e-02 3.63378763e-01 3.00416023e-01
6.51853830e-02 3.46254438e-01 8.77467766e-02 -8.00052285e-01
-1.80749409e-02 -3.76902521e-01 -4.33858894e-02 -3.16999435e-01
-3.49051565e-01 -3.69340807e-01 2.78981864e-01 -9.82481718e-01
1.32613158e+00 -1.54539895e+00 -3.89993042e-01 -1.04638547e-01
3.66068661e-01 -3.47795069e-01 1.00869024e-02 8.10956120e-01
3.28862637e-01 2.83651292e-01 9.75668281e-02 5.40048480e-02
2.04334497e-01 -2.66327947e-01 -1.01967597e+00 -3.87115538e-01
3.51681709e-02 9.68308687e-01 -1.11120474e+00 -4.23837721e-01
-8.04450214e-01 -1.52828112e-01 -6.25159264e-01 1.17440701e-01
-8.76164079e-01 1.53874472e-01 -3.54520440e-01 5.34667373e-01
1.54078558e-01 -6.41619563e-01 -3.30568433e-01 -6.64547533e-02
-1.76138014e-01 1.20012105e+00 -1.03188992e+00 1.16147137e+00
-2.99284399e-01 5.98578632e-01 -2.00060189e-01 -1.96787030e-01
9.98481989e-01 2.77568877e-01 -2.38973051e-02 -6.93123579e-01
-9.41944420e-02 3.35920572e-01 -1.85803890e-01 -5.32780826e-01
1.03520060e+00 -1.01885483e-01 -3.81945193e-01 9.67124760e-01
-1.21539384e-01 -4.19191062e-01 5.72958469e-01 6.14153028e-01
1.24254823e+00 -2.16943517e-01 2.23512530e-01 -4.51138318e-01
7.92719945e-02 7.53883481e-01 -1.31201774e-01 1.11025584e+00
3.83508503e-01 6.92631662e-01 6.37164176e-01 -1.82946116e-01
-9.35644269e-01 -1.03147840e+00 7.98611064e-03 1.26129222e+00
1.79842234e-01 -6.19172752e-01 -7.51124978e-01 -7.91830480e-01
-3.61716673e-02 1.29256701e+00 -6.82189167e-01 4.15825158e-01
-2.95971006e-01 -6.08301520e-01 3.87816668e-01 2.33568370e-01
3.25911492e-01 -9.34697270e-01 -7.01190829e-01 4.80680794e-01
-1.11012411e+00 -7.73229241e-01 -5.79973936e-01 -2.17622846e-01
-4.77868140e-01 -1.08776879e+00 -7.96551764e-01 -5.49316227e-01
3.35860133e-01 3.73546958e-01 1.92265332e+00 -1.37778288e-02
1.73828036e-01 4.56872642e-01 -8.26041579e-01 -6.88844562e-01
-8.63004744e-01 2.51849055e-01 -6.15558743e-01 -3.20212126e-01
1.94545716e-01 -2.27911144e-01 -7.37074018e-01 4.24143791e-01
-9.95440841e-01 4.76822965e-02 3.91024679e-01 5.53072691e-01
2.46384203e-01 -3.18960816e-01 1.31344426e+00 -1.25073779e+00
1.46754897e+00 -1.15822077e+00 -3.92466262e-02 3.59021157e-01
-5.89926362e-01 4.09023836e-02 6.85630023e-01 -1.14663944e-01
-1.18991137e+00 -7.61288166e-01 -4.96206701e-01 6.92400098e-01
1.03033319e-01 8.06593418e-01 1.00847773e-01 4.50540185e-01
1.50860310e+00 -3.64636853e-02 -5.17448559e-02 -7.10492671e-01
6.79194152e-01 6.06514812e-01 4.72195089e-01 -5.65057993e-01
5.13030052e-01 3.11584383e-01 -9.01813686e-01 -4.39561695e-01
-1.55328548e+00 -5.91835141e-01 1.12138316e-01 -3.14963400e-01
4.93107378e-01 -9.10538316e-01 -1.59485072e-01 -1.51291505e-01
-1.19091940e+00 3.64033021e-02 -8.39966297e-01 4.62794397e-03
-2.57941902e-01 3.62709612e-02 -4.40475881e-01 -3.88604403e-01
-4.44143146e-01 -5.28117299e-01 7.64960647e-01 2.82030046e-01
-9.31456864e-01 -8.84487152e-01 2.82559484e-01 8.53199482e-01
7.76994348e-01 3.47323626e-01 1.11586726e+00 -1.29912257e+00
-4.18533444e-01 -1.05022311e-01 -2.01633722e-01 -2.46053133e-02
2.87990030e-02 -2.64787912e-01 -7.64249563e-01 2.38735914e-01
2.66096264e-01 -5.02045631e-01 8.53674769e-01 7.75607079e-02
6.74208045e-01 -9.68354106e-01 -1.81320846e-01 -1.88254133e-01
1.03302538e+00 -1.22380376e-01 5.27847052e-01 4.99622673e-01
1.14161015e-01 1.07777953e+00 2.85531789e-01 3.55079621e-01
9.11445141e-01 4.59144771e-01 1.59354597e-01 2.04232574e-01
-4.97634321e-01 -5.72412372e-01 2.38996193e-01 8.54759693e-01
2.93298185e-01 -4.73143697e-01 -5.61603606e-01 8.46369386e-01
-1.61951458e+00 -1.40399742e+00 -3.61095309e-01 1.81642020e+00
1.23086166e+00 3.18885446e-01 4.75835860e-01 -6.11099228e-02
4.90423203e-01 7.12652057e-02 -3.55875075e-01 -3.37278962e-01
-3.25163126e-01 1.64807379e-01 -5.18644936e-02 6.19057834e-01
-4.78130490e-01 4.09404099e-01 6.78262615e+00 7.57592201e-01
-5.35066128e-01 1.42283767e-01 7.08889127e-01 1.42077571e-02
-1.21542871e+00 1.77230135e-01 -9.79537487e-01 7.02297211e-01
8.13202977e-01 -6.28432393e-01 -9.44543332e-02 8.60198855e-01
-7.41165131e-02 -7.08057806e-02 -9.39926147e-01 3.78233016e-01
4.73461956e-01 -1.98537600e+00 3.45819652e-01 -4.51430559e-01
9.95904922e-01 -2.23178238e-01 -2.47414410e-01 5.95335126e-01
7.62400985e-01 -8.69105339e-01 1.06767714e+00 4.79476869e-01
3.76056284e-01 -5.04924893e-01 6.46371901e-01 5.46845376e-01
-6.27375185e-01 -4.08176631e-02 -2.35942230e-01 -2.12823361e-01
4.21573460e-01 7.54301786e-01 -9.39838529e-01 3.28234911e-01
6.76449060e-01 1.06526911e-01 -8.76026750e-01 1.07271361e+00
-3.35536480e-01 7.26448655e-01 -6.02496080e-02 -6.05255485e-01
1.04771733e-01 3.41716081e-01 5.99217176e-01 1.32368314e+00
4.79722768e-01 8.41265544e-02 7.86741748e-02 9.87306893e-01
-3.67800057e-01 3.81049633e-01 -1.05537206e-01 3.56747985e-01
7.40751743e-01 1.14399016e+00 -7.73987055e-01 -6.20633245e-01
-2.09656715e-01 6.01058125e-01 1.92043304e-01 1.52073517e-01
-5.48555493e-01 -5.50142169e-01 1.93933442e-01 4.58353937e-01
2.95328535e-02 4.71078932e-01 -4.67176676e-01 -1.19683874e+00
1.98314607e-01 -1.38180196e+00 5.64811468e-01 -9.91641760e-01
-1.59284496e+00 9.27067757e-01 -6.54565170e-04 -1.12843871e+00
-5.37462711e-01 -8.44280273e-02 -7.21441269e-01 1.03827715e+00
-1.29232132e+00 -6.49007082e-01 -5.20992279e-01 1.32811993e-01
8.69210005e-01 7.00039603e-03 6.36715531e-01 5.54375499e-02
3.18161100e-01 4.37258720e-01 -2.96489894e-02 -1.39719337e-01
8.50979328e-01 -1.61739612e+00 7.72298932e-01 5.98847926e-01
2.71818995e-01 5.39103031e-01 1.04908419e+00 -7.34748840e-01
-6.92204118e-01 -9.58465219e-01 1.45467865e+00 -9.97408271e-01
7.38632321e-01 -2.60818228e-02 -9.39588249e-01 2.98044980e-01
5.79357207e-01 -8.30011725e-01 1.13273263e+00 2.71131635e-01
-6.10891461e-01 1.69202819e-01 -1.21950555e+00 7.11449623e-01
7.29023337e-01 -3.70146841e-01 -1.33008075e+00 6.50255561e-01
9.59574103e-01 -4.43406016e-01 -5.89991033e-01 -8.97168815e-02
3.77763212e-01 -1.18060112e+00 7.30575919e-01 -6.16265535e-01
9.06982243e-01 -3.96496564e-01 -5.33483401e-02 -1.61447155e+00
-1.80622548e-01 -7.02481806e-01 -5.55649325e-02 1.43536878e+00
9.69829619e-01 -5.31686902e-01 3.49130988e-01 6.40808463e-01
-3.12799245e-01 -8.02635074e-01 -6.23588324e-01 -4.57066596e-01
1.06539711e-01 -2.94942677e-01 8.40541661e-01 8.83009791e-01
1.79350883e-01 8.67749810e-01 8.69208202e-02 -2.17118293e-01
1.85633004e-01 4.95302916e-01 7.75817692e-01 -1.14668179e+00
-3.76967907e-01 -6.61976814e-01 4.09405053e-01 -1.41857076e+00
-4.70826983e-01 -9.59757209e-01 4.26272973e-02 -2.10554743e+00
1.28874838e-01 -3.19738626e-01 2.63838023e-01 -4.41615954e-02
-4.62557197e-01 1.65139735e-01 7.49736056e-02 4.15644765e-01
-8.07041943e-01 1.88919410e-01 1.12824798e+00 9.38492939e-02
-8.31978768e-02 1.77225739e-01 -1.78476262e+00 6.89907789e-01
5.86170316e-01 -4.30019796e-01 -4.89095181e-01 -4.50871468e-01
1.05636358e+00 1.55031100e-01 4.14652288e-01 -9.71536398e-01
2.32289106e-01 -2.49777995e-02 4.71215367e-01 -6.59482121e-01
-1.22714952e-01 -1.17756858e-01 -1.85377523e-02 1.81769550e-01
-9.93486166e-01 4.10792857e-01 -9.05991495e-02 3.32619309e-01
-3.09438705e-01 -5.11055231e-01 3.89500678e-01 -4.43677545e-01
-2.94099808e-01 -3.24956894e-01 -4.66659188e-01 1.02588248e+00
3.32097620e-01 -2.41181534e-03 -1.04093814e+00 -1.02713370e+00
-4.25859183e-01 1.77272141e-01 4.22307253e-01 5.62452435e-01
4.15253013e-01 -1.27209246e+00 -1.27473891e+00 -2.64462590e-01
3.91983688e-01 -3.70298445e-01 1.12036556e-01 3.59494567e-01
-4.66666460e-01 2.77817130e-01 1.55802146e-01 -4.50687148e-02
-7.07361877e-01 2.39968449e-01 4.26809937e-02 -4.68651325e-01
-3.96017611e-01 1.05560386e+00 -1.36322409e-01 -2.66228110e-01
-1.23743340e-01 -5.45972407e-01 -7.38693833e-01 7.37531662e-01
1.08171737e+00 5.75488031e-01 2.32024342e-01 -3.32094193e-01
7.11307228e-02 7.40996674e-02 -1.42207146e-01 -4.24499035e-01
9.74245191e-01 -3.62054110e-01 -4.33475897e-02 3.76017928e-01
8.50912094e-01 6.98997200e-01 -8.41647029e-01 -2.89603382e-01
3.32814962e-01 -3.29008549e-01 -4.96328026e-01 -1.43891823e+00
-1.62118524e-01 2.21160069e-01 -2.20279451e-02 1.13877034e+00
7.93179035e-01 6.02466226e-01 1.18262041e+00 2.70021290e-01
2.23277971e-01 -9.81618345e-01 2.84081578e-01 6.43855155e-01
1.38657129e+00 -1.11629748e+00 -1.10341184e-01 -2.00043112e-01
-8.02456200e-01 9.74219501e-01 3.49634409e-01 3.74574400e-02
3.22088242e-01 -3.34825404e-02 4.92526174e-01 -4.47325855e-01
-1.00483906e+00 -2.12586880e-01 7.51220405e-01 2.98028827e-01
6.04405463e-01 -2.55683184e-01 -4.67298687e-01 9.23881948e-01
-1.05463362e+00 -3.64335001e-01 6.54866695e-01 6.09455526e-01
-1.02957606e+00 -6.40466928e-01 -4.65318143e-01 8.38228106e-01
-8.52507770e-01 -1.32484287e-01 -6.25463247e-01 4.18800890e-01
-1.48517285e-02 1.38063669e+00 -7.53637850e-02 -1.41616642e-01
5.64602315e-01 8.86577368e-02 5.38270101e-02 -9.56013083e-01
-1.05775249e+00 -2.47106254e-01 6.80382192e-01 -1.67502165e-01
-2.39277735e-01 -5.47058105e-01 -7.96991765e-01 -2.39630356e-01
-9.48176384e-02 9.31801617e-01 5.60120940e-01 9.45325971e-01
6.40724897e-01 4.37545419e-01 4.75819409e-01 -1.70723051e-01
-7.89132476e-01 -1.05798435e+00 -1.04512811e-01 7.02658117e-01
2.80000627e-01 -2.71251202e-01 -4.91762221e-01 2.27000251e-01]
|
[11.586124420166016, 8.198283195495605]
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.