LES CHERCHEURS DE MILA PRÉSENTERONT LEURS TRAVAUX À L’OCCASION DE LA CONFÉRENCE ICLR 2021

Mila est fier de partager les efforts et les réalisations de ses chercheurs, qui présenteront leurs travaux à l’International Conference on Learning Representations(ICLR). L’événement se déroulera en ligne du 4 au 8 mai.

Les publications acceptées des chercheurs de Mila feront l’objet de présentations orales, d’affiches et même de « spotlights ». Leurs travaux, ainsi que leurs ateliers, couvrent un large éventail de sujets, y compris, mais sans s’y limiter, de nouvelles représentations autoprédictives, des algorithmes d’optimisation des politiques à partir de données fixes, et une approche novatrice pour la prévision de l’infectiosité par une méthode PCT (Proactive Contact Tracing).

La conférence ICLR est reconnue mondialement pour la présentation de recherches de pointe dans le domaine de l’apprentissage automatique, de l’IA et d’autres domaines connexes. C’est l’une des conférences les mieux classées sur la base de son h5-index et son score d’impact (voir : Google Scholar and Guide2Research). Parmi les 2 997 soumissions de cette année, 860 publications ont été acceptées (28,7 %). De ce chiffre, 31 ont été co-rédigées par des membres de Mila.

Voici la liste complète des publications acceptées, ainsi qu’un aperçu des ateliers co-organisés par nos chercheurs :

Publications « spotlight » :

Parameterization of Hypercomplex Multiplications
Aston Zhang, Yi Tay, Shuai Zhang, Anh Tuan Luu, Siu Hui, Jie Fu

Data-Efficient Reinforcement Learning with Self-Predictive Representations
Max Schwarzer, Ankesh Anand, Rishab Goel, R. Devon Hjelm, Aaron Courville, Philip Bachman
https://arxiv.org/abs/2007.05929

Predicting Infectiousness for Proactive Contact Tracing
Yoshua Bengio, Prateek Gupta, Tegan Maharaj, Nasim Rahaman, Martin Weiss, Tristan Deleu, Eilif Muller, Meng Qu, Victor Schmidt, Pierre-Luc St-Charles, Hannah Alsdurf, Olexa Bilanuik, David Buckeridge, Gáetan Marceau Caron, Pierre-Luc Carrier, Joumana Ghosn, Satya Ortiz-Gagne, Chris Pal, Irina Rish, Bernhard Schölkopf, Abhinav Sharma, Jian Tang, Andrew Williams
https://arxiv.org/abs/2010.12536

Recurrent Independent Mechanisms
Anirudh Goyal, Alex Lamb, Jordan Hoffman, Shagun Sodhani, Yoshua Bengio, Bernhard Schölkopf
https://arxiv.org/abs/1909.10893v6

Regularized Inverse Reinforcement Learning
Wonseok Jeon, Chen-Yang Su, Paule Barde, Thang Doan, Derek Nowrouzezahrai, Joelle Pineau
https://arxiv.org/abs/2010.03691

Systematic generalization with group invariant predictions
Faruk Ahmed, Yoshua Bengio, Harm van Seijen, Aaron Courville

Neural Approximate Sufficient Statistics for Implicit Models
Yanzhi Chen, Dinghuai Zhang, Michael U. Gutmann, Aaron Courville, Zhanxing Zhu
https://arxiv.org/abs/2010.10079

Contrastive Similarity Embeddings for Generalization in Reinforcement Learning
Rishabh Agarwal, Marlos Machado, Pablo Samuel Castro, Marc Bellemare
https://agarwl.github.io/pse/

Présentation orale :

Iterated Learning for Emergent Systematicity in VQA
Ankit Vani, Max Schwarzer, Yuchen Lu, Eeshan Dhekane, Aaron Courville

Présentations par affiches :

CoCon: A Self-Supervised Approach for Controlled Text Generation
Alvin Chan, Yew-Soon Ong, Bill Pung, Aston Zhang, Jie Fu
https://arxiv.org/abs/2006.03535

Adversarial score matching and improved sampling for image generation
Alexia Jolicoeur-Martineau, Rémi Piché-Taillefer, Rémi Tachet des Combes, Ioannis Mitliagkas
https://arxiv.org/abs/2009.05475

The Importance of Pessimism in Fixed-Dataset Policy Optimization
Jacob Buckman, Carles Gelada, Marc G. Bellemare
https://arxiv.org/abs/2009.06799

A Universal Representation Transformer Layer for Few-Shot Image Classification
Lu Liu, William Hamilton, Guodong Long, Jing Jiang, Hugo Larochelle
https://arxiv.org/abs/2006.11702

Saliency is a Possible Red Herring When Diagnosing Poor Generalization
Joseph Viviano, Becks Simpson, Francis Dutil, Yoshua Bengio, Joseph Paul Cohen
https://arxiv.org/abs/1910.00199v2

Integrating Categorical Semantics into Unsupervised Domain Translation
Samuel Lavoie-Marchildon, Faruk Ahmed, Aaron Courville
https://arxiv.org/abs/2010.01262v1

CausalWorld: A Robotic Manipulation Benchmark for Causal Structure and Transfer Learning
Ossama Ahmed, Frederik Träuble, Anirudh Goyal, Alexander Neitz, Manuel Wuthrich, Yoshua Bengio, Bernard Schölkopf, Stefan Bauer
https://arxiv.org/abs/2010.04296

Learning Neural Generative Dynamics for Molecular Conformation Generation
Minkai Xu, Shitong Luo, Yoshua Bengio, Jian Peng, Jian Tang

Meta Attention Networks: Meta-Learning Attention to Modulate Information Between Recurrent Independent Mechanisms
Kanika Madan, Nan Rosemary Ke, Anirudh Goyal, Bernard Schölkopf, Yoshua Bengio

Learning Robust State Abstractions for Hidden-Parameter Block MDPs
Amy Zhang, Shagun Sodhani, Khimya Khetarpal, Joelle Pineau

DC3: A learning method for optimization with hard constraints
Priya L. Donti, David Rolnick, J. Zico Kolter

Conditionally Adaptive Multi-Task Learning: Improving Transfer Learning in NLP Using Fewer Parameters & Less Data
Jonathan Pilault, Amine El hattami, Christopher Pal
https://arxiv.org/abs/2009.09139

Factorizing Declarative and Procedural Knowledge in Structured, Dynamical Systems
Anirudh Goyal, Alex Lamb, Phanideep Gampa, Philippe Beaudoin, Charles Blundell, Sergey Levine, Yoshua Bengio, Michael Curtis Mozer

Learning to live with Dale’s principle: ANNs with separate excitatory and inhibitory units
Jonathan Cornford, Damjan Kalajdzievski, Marco Leite, Amélie Lamarquelle, Dimitri Michael Kullmann, Blake Richards
https://www.biorxiv.org/content/10.1101/2020.11.02.364968v1

Reinforcement Learning with Random Delays
Yann Bouteiller, Simon Ramstedt, Giovanni Beltrame, Christopher Pal, Jonathan Binas
https://arxiv.org/abs/2010.02966

gradSim: Differentiable simulation for system identification and visuomotor control
Krishna Murthy, Miles Macklin, Florian Golemo, Vikram Voleti, Linda Petrini, Martin Weiss, Breandan Considine, Jérôme Parent-Lévesque, Kevin Xie, Kenny Erleben, Liam Paull, Florian Shkurti, Derek Nowrouzezahrai, Sanja Fidler

RNNLogic: Learning Logic Rules for Reasoning on Knowledge Graphs
Meng Qu, Junkun Chen, Louis-Pascal Xhonneux, Yoshua Bengio, Jian Tang
https://arxiv.org/abs/2010.04029

Neural representation and generation for RNA secondary structures
Zichao Yan, William Hamilton, Mathieu Blanchette
https://www.biorxiv.org/content/10.1101/2020.02.11.931030v1

Convex Potential Flows: Universal Probability Distributions with Optimal Transport and Convex Optimization
Chin-Wei Huang, Ricky T. Q. Chen, Christos Tsirigotis, Aaron Courville
https://arxiv.org/abs/2012.05942

Repurposing Pretrained Models for Robust Out-of-domain Few-Shot Learning
Namyeong Kwon, Hwidong Na, Gabriel Huang, Simon Lacoste-Julien

Spatially Structured Recurrent Modules
Nasim Rahaman, Anirudh Goyal, Muhammad Waleed Gondal, Manuel Wuthrich, Stefan Bauer, Yash Sharma, Yoshua Bengio, Bernhard Schölkopf
https://arxiv.org/abs/2007.06533

Implicit Under-Parameterization Inhibits Data-Efficient Deep Reinforcement Learning
Aviral Kumar, Rishabh Agarwal, Dibya Ghosh, Sergey Levine
https://arxiv.org/abs/2010.14498

Ateliers :

Geometric and Topological Representation Learning
Guy Wolf, Xiuyuan Cheng, Smita Krishnaswamy, Jure Leskovec, Bastian A. Rieck, Soledad Villar
https://gt-rl.github.io/

A Roadmap to Never-Ending RL
Feryal Behbahani, Khimya Khetarpal, Louis Kirsch, Rose Wang, Annie Xie, Adam White, Doina Precup

Beyond Static Papers: Rethinking How we Share Scientific Understanding in ML
Krishna Murthy Jatavallabhula, Bhairav Mehta, Tegan Maharaj, Amy Tabb, Khimya Khetarpal, Aditya Kusupati, Anna Rogers, Sara Hooker, Breandan Considine, Devi Parikh, Derek Nowrouzezahrai, Yoshua Bengio
https://rethinkingmlpapers.github.io/

Self-Supervision for Reinforcement Learning
Ankesh Anand, Bogdan Mazoure, Amy Zhang, Thang Doan, Khurram Javed, Devon Hjelm, Martha White

Hardware-Aware Efficient Training of Deep Learning Models
Ghouthi Boukli Hacene, Vincent Gripon, François Leduc-Primeau, Vahid Partovi Nia, Andreas Moshovos, Fan Yang, Yoshua Bengio
https://haet2021.github.io/