22 papers accepted at the International Conference on Machine Learning (ICML)

We're happy to announce that Mila students and faculty had no fewer than 22 papers accepted at the International Conference on Machine Learning (ICML), which will take place in Long Beach, California this week. Here is the complete list:


Greedy Layerwise Learning Can Scale To ImageNet

Eugene Belilovsky (Mila, Université de Montréal) · Michael Eickenberg (UC Berkeley) · Edouard Oyallon (CentraleSupélec)



TarMAC: Targeted Multi-Agent Communication

Abhishek Das (Georgia Tech) · Theophile Gervet (Carnegie Mellon University) · Joshua Romoff (Mila, McGill University) · Dhruv Batra (Georgia Institute of Technology / Facebook AI Research) · Devi Parikh (Georgia Institute of Technology / Facebook AI Research) · Michael Rabbat (Facebook, Mila) · Joelle Pineau (Facebook, Mila)



Off-Policy Deep Reinforcement Learning without Exploration

Scott Fujimoto (Mila, McGill University) · David Meger (Mila, McGill University) · Doina Precup (Mila, McGill University / DeepMind)



GEOMetrics: Exploiting Geometric Structure for Graph-Encoded Objects

Edward Smith (McGill University) · Adriana Romero (FAIR, Mila) · Scott Fujimoto (Mila, McGill University) · David Meger (Mila, McGill University)



Safe Policy Improvement with Baseline Bootstrapping

Romain Laroche (Microsoft Research) · Paul Trichelair (Mila, McGill University) · Remi Tachet des Combes (Microsoft Research Montreal)



Garbage In, Reward Out: Bootstrapping Exploration in Multi-Armed Bandits

Branislav Kveton (Google Research) · Csaba Szepesvari (DeepMind / University of Alberta) · Sharan Vaswani (Mila, Université de Montréal) · Zheng Wen (Adobe Research) · Tor Lattimore (DeepMind) · Mohammad Ghavamzadeh (Facebook AI Research)



Manifold Mixup: Better Representations by Interpolating Hidden States

Vikas Verma (Mila, Aalto University) · Alex Lamb (Mila, Université de Montréal) · Christopher Beckham (Mila, Ecole Polytechnique de Montréal) · Amir Najafi (Sharif University of Technology) · Ioannis Mitliagkas (Mila, Université de Montréal) · David Lopez-Paz (Facebook AI Research) · Yoshua Bengio (Mila / Université de Montréal)



Per-Decision Option Discounting

Anna Harutyunyan (DeepMind) · Peter Vrancx (PROWLER.io) · Philippe Hamel (Deepmind) · Ann Nowe (VU Brussel) · Doina Precup (DeepMind, Mila)



Stochastic Gradient Push for Distributed Deep Learning

Mahmoud Assran (McGill University / Facebook FAIR) · Nicolas Loizou (The University of Edinburgh) · Nicolas Ballas (Facebook FAIR) · Michael Rabbat (Facebook, Mila)



Compositional Fairness Constraints for Graph Embeddings

Avishek Bose (McGill/Mila) · William Hamilton (McGill University)



State-Reification Networks: Improving Generalization by Modeling the Distribution of Hidden Representations

Alex Lamb (Université de Montréal) · Jonathan Binas (Mila, Université de Montréal) · Anirudh Goyal (Mila, Université de Montréal) · Sandeep Subramanian (Mila, Université de Montréal) · Denis Kasakov (University of Colorado Boulder) · Ioannis Mitliagkas (Mila, Université de Montréal) · Yoshua Bengio (Mila, Université de Montreal) · Michael Mozer (Google Research & University of Colorado Boulder)



The Value Function Polytope in Reinforcement Learning

Robert Dadashi (Google AI Residency Program) · Marc Bellemare (Google Brain, Mila) · Adrien Ali Taiga (Mila, Université de Montréal) · Nicolas Le Roux (Google) · Dale Schuurmans (Google / University of Alberta)



DeepMDP: Learning Continuous Latent Space Models for Representation Learning

Carles Gelada (Google Brain) · Saurabh Kumar (Google Brain) · Jacob Buckman (Johns Hopkins University) · Ofir Nachum (Google Brain) · Marc Bellemare (Google Brain, Mila)



Understanding the Impact of Entropy on Policy Optimization

Zafarali Ahmed (Mila, McGill University) · Nicolas Le Roux (Google, Mila) · Mohammad Norouzi (Google Brain) · Dale Schuurmans (Google / University of Alberta)



Multi-objective training of Generative Adversarial Networks with multiple discriminators

Isabela Albuquerque (Institut National de la Recherche Scientifique) · Joao Monteiro (Institut National de la Recherche Scientifique) · Thang Doan (McGill University) · Breandan Considine (Mila, Université de Montréal) · Tiago Falk (INRS-EMT) · Ioannis Mitliagkas (Mila, Université de Montréal)



GMNN: Graph Markov Neural Networks

Meng Qu (Mila, Université de Montréal) · Yoshua Bengio (Mila, Université de Montréal) · Jian Tang (Mila, HEC Montréal)



Unreproducible Research is Reproducible

Xavier Bouthillier (Mila, Université de Montréal) · César Laurent (Mila, Université de Montréal) · Pascal Vincent (Mila, Université de Montréal)



On the Spectral Bias of Neural Networks

Nasim Rahaman (Mila, University of Heidelberg) · Aristide Baratin (Mila, Université de Montréal) · Devansh Arpit (Mila, Université de Montréal) · Felix Draxler (Heidelberg University) · Min Lin (Mila, Université de Montréal) · Fred Hamprecht (Heidelberg Collaboratory for Image Processing) · Yoshua Bengio (Mila, Université de Montréal) · Aaron Courville (Université de Montréal)



Separable value functions across time-scales

Joshua Romoff  (Mila, McGill University) · Peter Henderson (Stanford University) · Ahmed Touati (FAIR, Mila) · Yann Ollivier (Facebook Artificial Intelligence Research) · Joelle Pineau (Mila, McGill University / Facebook) · Emma Brunskill (Stanford University)



Fairwashing: the risk of rationalization

Ulrich Aïvodji (UQAM) · Hiromi Arai (RIKEN AIP) · Olivier Fortineau (Ensta Paristech) · Sébastien Gambs (UQAM) · Satoshi Hara (Osaka University) · Alain Tapp (Mila, Université de Montréal)



Hierarchical Importance Weighted Autoencoders

Chin-Wei Huang (Mila, Université de Montréal) · Kris Sankaran (Mila, Université de Montréal) · Eeshan Dhekane (Mila, Université de Montréal) · Alexandre Lacoste (Element AI) · Aaron Courville (Mila, Université de Montréal)1



On Variational Bounds of Mutual Information

Ben Poole (Google Brain) · Sherjil Ozair (Mila, University of Montreal) · Aäron van den Oord (Google Deepmind) · Alexander Alemi (Google) · George Tucker (Google Brain)