Timezone: »
Despite overparameterization, deep networks trained via supervised learning are surprisingly easy to optimize and exhibit excellent generalization. One hypothesis to explain this is that overparameterized deep networks enjoy the benefits of implicit regularization induced by stochastic gradient descent, which favors parsimonious solutions that generalize well on test inputs. It is reasonable to surmise that deep reinforcement learning (RL) methods could also benefit from this effect. In this paper, we discuss how the implicit regularization effect of SGD seen in supervised learning could in fact be harmful in the offline deep RL setting, leading to poor generalization and degenerate feature representations. Our theoretical analysis shows that when existing models of implicit regularization are applied to temporal difference learning, the resulting derived regularizer favors degenerate solutions with excessive aliasing, in stark contrast to the supervised learning case. We back up these findings empirically, showing that feature representations learned by a deep network value function trained via bootstrapping can indeed become degenerate, aliasing the representations for state-action pairs that appear on either side of the Bellman backup. To address this issue, we derive the form of this implicit regularizer and, inspired by this derivation, propose a simple and effective explicit regularizer, called DR3, that counteracts the undesirable effects of this implicit regularizer. When combined with existing offline RL methods, DR3 substantially improves performance and stability, alleviating unlearning in Atari 2600 games, D4RL domains, and robotic manipulation from images.
Author Information
Aviral Kumar (UC Berkeley)
Rishabh Agarwal (Google Research, Brain Team)
Tengyu Ma (Stanford University)
Aaron Courville (U. Montreal)
George Tucker (Google Brain)
Sergey Levine (UC Berkeley)
Related Events (a corresponding poster, oral, or spotlight)
-
2021 : DR3: Value-Based Deep Reinforcement Learning Requires Explicit Regularization »
Mon. Dec 13th 05:15 -- 05:27 PM Room
More from the Same Authors
-
2021 Spotlight: Robust Predictable Control »
Ben Eysenbach · Russ Salakhutdinov · Sergey Levine -
2021 Spotlight: Offline Reinforcement Learning as One Big Sequence Modeling Problem »
Michael Janner · Qiyang Li · Sergey Levine -
2021 Spotlight: Neural Additive Models: Interpretable Machine Learning with Neural Nets »
Rishabh Agarwal · Levi Melnick · Nicholas Frosst · Xuezhou Zhang · Ben Lengerich · Rich Caruana · Geoffrey Hinton -
2021 Spotlight: Why Do Pretrained Language Models Help in Downstream Tasks? An Analysis of Head and Prompt Tuning »
Colin Wei · Sang Michael Xie · Tengyu Ma -
2021 Spotlight: Pragmatic Image Compression for Human-in-the-Loop Decision-Making »
Sid Reddy · Anca Dragan · Sergey Levine -
2021 Spotlight: A Variational Perspective on Diffusion-Based Generative Models and Score Matching »
Chin-Wei Huang · Jae Hyun Lim · Aaron Courville -
2021 : Calibrated Ensembles: A Simple Way to Mitigate ID-OOD Accuracy Tradeoffs »
Ananya Kumar · Aditi Raghunathan · Tengyu Ma · Percy Liang -
2021 : Extending the WILDS Benchmark for Unsupervised Adaptation »
Shiori Sagawa · Pang Wei Koh · Tony Lee · Irena Gao · Sang Michael Xie · Kendrick Shen · Ananya Kumar · Weihua Hu · Michihiro Yasunaga · Henrik Marklund · Sara Beery · Ian Stavness · Jure Leskovec · Kate Saenko · Tatsunori Hashimoto · Sergey Levine · Chelsea Finn · Percy Liang -
2021 : Test Time Robustification of Deep Models via Adaptation and Augmentation »
Marvin Zhang · Sergey Levine · Chelsea Finn -
2021 : Self-supervised Learning is More Robust to Dataset Imbalance »
Hong Liu · Jeff Z. HaoChen · Adrien Gaidon · Tengyu Ma -
2021 : Plan Better Amid Conservatism: Offline Multi-Agent Reinforcement Learning with Actor Rectification »
Ling Pan · Longbo Huang · Tengyu Ma · Huazhe Xu -
2021 : Value Function Spaces: Skill-Centric State Abstractions for Long-Horizon Reasoning »
· Ted Xiao · Alexander Toshev · Sergey Levine · brian ichter -
2021 : Data Sharing without Rewards in Multi-Task Offline Reinforcement Learning »
Tianhe Yu · Aviral Kumar · Yevgen Chebotar · Chelsea Finn · Sergey Levine · Karol Hausman -
2021 : Should I Run Offline Reinforcement Learning or Behavioral Cloning? »
Aviral Kumar · Joey Hong · Anikait Singh · Sergey Levine -
2021 : Offline Policy Selection under Uncertainty »
Mengjiao Yang · Bo Dai · Ofir Nachum · George Tucker · Dale Schuurmans -
2021 : Offline Reinforcement Learning with In-sample Q-Learning »
Ilya Kostrikov · Ashvin Nair · Sergey Levine -
2021 : C-Planning: An Automatic Curriculum for Learning Goal-Reaching Tasks »
Tianjun Zhang · Ben Eysenbach · Russ Salakhutdinov · Sergey Levine · Joseph Gonzalez -
2021 : Behavior Predictive Representations for Generalization in Reinforcement Learning »
Siddhant Agarwal · Aaron Courville · Rishabh Agarwal -
2021 : The Information Geometry of Unsupervised Reinforcement Learning »
Ben Eysenbach · Russ Salakhutdinov · Sergey Levine -
2021 : Mismatched No More: Joint Model-Policy Optimization for Model-Based RL »
Ben Eysenbach · Alexander Khazatsky · Sergey Levine · Russ Salakhutdinov -
2021 : Offline Meta-Reinforcement Learning with Online Self-Supervision »
Vitchyr Pong · Ashvin Nair · Laura Smith · Catherine Huang · Sergey Levine -
2021 : Hybrid Imitative Planning with Geometric and Predictive Costs in Offroad Environments »
Daniel Shin · · Ali Agha · Nicholas Rhinehart · Sergey Levine -
2021 : CoMPS: Continual Meta Policy Search »
Glen Berseth · Zhiwei Zhang · Grace Zhang · Chelsea Finn · Sergey Levine -
2021 : MIDI-DDSP: Hierarchical Modeling of Music for Detailed Control »
Yusong Wu · Ethan Manilow · Kyle Kastner · Tim Cooijmans · Aaron Courville · Cheng-Zhi Anna Huang · Jesse Engel -
2021 : Speaker Intro »
Aviral Kumar · George Tucker -
2021 : Speaker Intro »
Aviral Kumar · George Tucker -
2021 : Speaker Intro »
Rishabh Agarwal · Aviral Kumar -
2021 : Speaker Intro »
Rishabh Agarwal · Aviral Kumar -
2021 Workshop: Offline Reinforcement Learning »
Rishabh Agarwal · Aviral Kumar · George Tucker · Justin Fu · Nan Jiang · Doina Precup · Aviral Kumar -
2021 : Opening Remarks »
Rishabh Agarwal · Aviral Kumar -
2021 : Behavior Predictive Representations for Generalization in Reinforcement Learning »
Siddhant Agarwal · Aaron Courville · Rishabh Agarwal -
2021 : Offline Meta-Reinforcement Learning with Online Self-Supervision Q&A »
Vitchyr Pong · Ashvin Nair · Laura Smith · Catherine Huang · Sergey Levine -
2021 : Offline Meta-Reinforcement Learning with Online Self-Supervision »
Vitchyr Pong · Ashvin Nair · Laura Smith · Catherine Huang · Sergey Levine -
2021 : Offline Meta-Reinforcement Learning with Online Self-Supervision »
Vitchyr Pong · Ashvin Nair · Laura Smith · Catherine Huang · Sergey Levine -
2021 : DR3: Value-Based Deep Reinforcement Learning Requires Explicit Regularization Q&A »
Aviral Kumar · Rishabh Agarwal · Tengyu Ma · Aaron Courville · George Tucker · Sergey Levine -
2021 Oral: Replacing Rewards with Examples: Example-Based Policy Search via Recursive Classification »
Ben Eysenbach · Sergey Levine · Russ Salakhutdinov -
2021 Poster: Robust Predictable Control »
Ben Eysenbach · Russ Salakhutdinov · Sergey Levine -
2021 Poster: Which Mutual-Information Representation Learning Objectives are Sufficient for Control? »
Kate Rakelly · Abhishek Gupta · Carlos Florensa · Sergey Levine -
2021 Poster: COMBO: Conservative Offline Model-Based Policy Optimization »
Tianhe Yu · Aviral Kumar · Rafael Rafailov · Aravind Rajeswaran · Sergey Levine · Chelsea Finn -
2021 Poster: Gradient Starvation: A Learning Proclivity in Neural Networks »
Mohammad Pezeshki · Oumar Kaba · Yoshua Bengio · Aaron Courville · Doina Precup · Guillaume Lajoie -
2021 Poster: Outcome-Driven Reinforcement Learning via Variational Inference »
Tim G. J. Rudner · Vitchyr Pong · Rowan McAllister · Yarin Gal · Sergey Levine -
2021 Poster: Label Noise SGD Provably Prefers Flat Global Minimizers »
Alex Damian · Tengyu Ma · Jason Lee -
2021 Poster: Learning Barrier Certificates: Towards Safe Reinforcement Learning with Zero Training-time Violations »
Yuping Luo · Tengyu Ma -
2021 Poster: Bayesian Adaptation for Covariate Shift »
Aurick Zhou · Sergey Levine -
2021 Poster: Coupled Gradient Estimators for Discrete Latent Variables »
Zhe Dong · Andriy Mnih · George Tucker -
2021 Poster: Calibrating Predictions to Decisions: A Novel Approach to Multi-Class Calibration »
Shengjia Zhao · Michael Kim · Roshni Sahoo · Tengyu Ma · Stefano Ermon -
2021 Poster: Offline Reinforcement Learning as One Big Sequence Modeling Problem »
Michael Janner · Qiyang Li · Sergey Levine -
2021 Poster: Why Do Pretrained Language Models Help in Downstream Tasks? An Analysis of Head and Prompt Tuning »
Colin Wei · Sang Michael Xie · Tengyu Ma -
2021 Poster: Pragmatic Image Compression for Human-in-the-Loop Decision-Making »
Sid Reddy · Anca Dragan · Sergey Levine -
2021 Poster: Pretraining Representations for Data-Efficient Reinforcement Learning »
Max Schwarzer · Nitarshan Rajkumar · Michael Noukhovitch · Ankesh Anand · Laurent Charlin · R Devon Hjelm · Philip Bachman · Aaron Courville -
2021 Poster: A Variational Perspective on Diffusion-Based Generative Models and Score Matching »
Chin-Wei Huang · Jae Hyun Lim · Aaron Courville -
2021 Poster: Replacing Rewards with Examples: Example-Based Policy Search via Recursive Classification »
Ben Eysenbach · Sergey Levine · Russ Salakhutdinov -
2021 Oral: Provable Guarantees for Self-Supervised Deep Learning with Spectral Contrastive Loss »
Jeff Z. HaoChen · Colin Wei · Adrien Gaidon · Tengyu Ma -
2021 Oral: Deep Reinforcement Learning at the Edge of the Statistical Precipice »
Rishabh Agarwal · Max Schwarzer · Pablo Samuel Castro · Aaron Courville · Marc Bellemare -
2021 Poster: Safe Reinforcement Learning by Imagining the Near Future »
Garrett Thomas · Yuping Luo · Tengyu Ma -
2021 Poster: Neural Additive Models: Interpretable Machine Learning with Neural Nets »
Rishabh Agarwal · Levi Melnick · Nicholas Frosst · Xuezhou Zhang · Ben Lengerich · Rich Caruana · Geoffrey Hinton -
2021 Poster: Provable Guarantees for Self-Supervised Deep Learning with Spectral Contrastive Loss »
Jeff Z. HaoChen · Colin Wei · Adrien Gaidon · Tengyu Ma -
2021 Poster: Information is Power: Intrinsic Control via Information Capture »
Nicholas Rhinehart · Jenny Wang · Glen Berseth · John Co-Reyes · Danijar Hafner · Chelsea Finn · Sergey Levine -
2021 Poster: Conservative Data Sharing for Multi-Task Offline Reinforcement Learning »
Tianhe Yu · Aviral Kumar · Yevgen Chebotar · Karol Hausman · Sergey Levine · Chelsea Finn -
2021 Poster: Deep Reinforcement Learning at the Edge of the Statistical Precipice »
Rishabh Agarwal · Max Schwarzer · Pablo Samuel Castro · Aaron Courville · Marc Bellemare -
2021 Poster: Why Generalization in RL is Difficult: Epistemic POMDPs and Implicit Partial Observability »
Dibya Ghosh · Jad Rahme · Aviral Kumar · Amy Zhang · Ryan Adams · Sergey Levine -
2021 Poster: Autonomous Reinforcement Learning via Subgoal Curricula »
Archit Sharma · Abhishek Gupta · Sergey Levine · Karol Hausman · Chelsea Finn -
2021 Poster: Provable Model-based Nonlinear Bandit and Reinforcement Learning: Shelve Optimism, Embrace Virtual Curvature »
Kefan Dong · Jiaqi Yang · Tengyu Ma -
2021 Poster: Adaptive Risk Minimization: Learning to Adapt to Domain Shift »
Marvin Zhang · Henrik Marklund · Nikita Dhawan · Abhishek Gupta · Sergey Levine · Chelsea Finn -
2020 : Design-Bench: Benchmarks for Data-Driven Offline Model-Based Optimization »
Brandon Trabucco · Aviral Kumar · XINYANG GENG · Sergey Levine -
2020 : Conservative Objective Models: A Simple Approach to Effective Model-Based Optimization »
Brandon Trabucco · Aviral Kumar · XINYANG GENG · Sergey Levine -
2020 : Panel »
Emma Brunskill · Nan Jiang · Nando de Freitas · Finale Doshi-Velez · Sergey Levine · John Langford · Lihong Li · George Tucker · Rishabh Agarwal · Aviral Kumar -
2020 Workshop: Offline Reinforcement Learning »
Aviral Kumar · Rishabh Agarwal · George Tucker · Lihong Li · Doina Precup · Aviral Kumar -
2020 : Introduction »
Aviral Kumar · George Tucker · Rishabh Agarwal -
2020 Workshop: AI for Earth Sciences »
Surya Karthik Mukkavilli · Johanna Hansen · Natasha Dudek · Tom Beucler · Kelly Kochanski · Mayur Mudigonda · Karthik Kashinath · Amy McGovern · Paul D Miller · Chad Frischmann · Pierre Gentine · Gregory Dudek · Aaron Courville · Daniel Kammen · Vipin Kumar -
2020 Poster: Unsupervised Learning of Dense Visual Representations »
Pedro O. Pinheiro · Amjad Almahairi · Ryan Benmalek · Florian Golemo · Aaron Courville -
2020 Poster: One Solution is Not All You Need: Few-Shot Extrapolation via Structured MaxEnt RL »
Saurabh Kumar · Aviral Kumar · Sergey Levine · Chelsea Finn -
2020 Tutorial: (Track3) Offline Reinforcement Learning: From Algorithm Design to Practical Applications »
Sergey Levine · Aviral Kumar -
2019 : Tengyu Ma, "Designing Explicit Regularizers for Deep Models" »
Tengyu Ma -
2019 Poster: Ordered Memory »
Yikang Shen · Shawn Tan · Arian Hosseini · Zhouhan Lin · Alessandro Sordoni · Aaron Courville -
2019 Poster: Graph Normalizing Flows »
Jenny Liu · Aviral Kumar · Jimmy Ba · Jamie Kiros · Kevin Swersky -
2019 Poster: MelGAN: Generative Adversarial Networks for Conditional Waveform Synthesis »
Kundan Kumar · Rithesh Kumar · Thibault de Boissiere · Lucas Gestin · Wei Zhen Teoh · Jose Sotelo · Alexandre de BrĂ©bisson · Yoshua Bengio · Aaron Courville -
2019 Poster: No-Press Diplomacy: Modeling Multi-Agent Gameplay »
Philip Paquette · Yuchen Lu · SETON STEVEN BOCCO · Max Smith · Satya O.-G. · Jonathan K. Kummerfeld · Joelle Pineau · Satinder Singh · Aaron Courville -
2018 : Poster Session »
Sujay Sanghavi · Vatsal Shah · Yanyao Shen · Tianchen Zhao · Yuandong Tian · Tomer Galanti · Mufan Li · Gilad Cohen · Daniel Rothchild · Aristide Baratin · Devansh Arpit · Evangelos Papalexakis · Michael Perlmutter · Ashok Vardhan Makkuva · Pim de Haan · Yingyan Lin · Wanmo Kang · Cheolhyoung Lee · Hao Shen · Sho Yaida · Dan Roberts · Nadav Cohen · Philippe Casgrain · Dejiao Zhang · Tengyu Ma · Avinash Ravichandran · Julian Emilio Salazar · Bo Li · Davis Liang · Christopher Wong · Glen Bigan Mbeng · Animesh Garg -
2018 : Spotlights »
Guangneng Hu · Ke Li · Aviral Kumar · Phi Vu Tran · Samuel Fadel · Rita Kuznetsova · Bong-Nam Kang · Behrouz Haji Soleimani · Jinwon An · Nathan de Lara · Anjishnu Kumar · Tillman Weyde · Melanie Weber · Kristen Altenburger · Saeed Amizadeh · Xiaoran Xu · Yatin Nandwani · Yang Guo · Maria Pacheco · William Fedus · Guillaume Jaume · Yuka Yoneda · Yunpu Ma · Yunsheng Bai · Berk Kapicioglu · Maximilian Nickel · Fragkiskos Malliaros · Beier Zhu · Aleksandar Bojchevski · Joshua Joseph · Gemma Roig · Esma Balkir · Xander Steenbrugge -
2018 Workshop: Visually grounded interaction and language »
Florian Strub · Harm de Vries · Erik Wijmans · Samyak Datta · Ethan Perez · Mateusz Malinowski · Stefan Lee · Peter Anderson · Aaron Courville · Jeremie MARY · Dhruv Batra · Devi Parikh · Olivier Pietquin · Chiori HORI · Tim Marks · Anoop Cherian -
2018 Poster: Improving Explorability in Variational Inference with Annealed Variational Objectives »
Chin-Wei Huang · Shawn Tan · Alexandre Lacoste · Aaron Courville -
2018 Poster: Towards Text Generation with Adversarially Learned Neural Outlines »
Sandeep Subramanian · Sai Rajeswar Mudumba · Alessandro Sordoni · Adam Trischler · Aaron Courville · Chris Pal -
2017 Workshop: Visually grounded interaction and language »
Florian Strub · Harm de Vries · Abhishek Das · Satwik Kottur · Stefan Lee · Mateusz Malinowski · Olivier Pietquin · Devi Parikh · Dhruv Batra · Aaron Courville · Jeremie Mary -
2017 Poster: Improved Training of Wasserstein GANs »
Ishaan Gulrajani · Faruk Ahmed · Martin Arjovsky · Vincent Dumoulin · Aaron Courville -
2017 Poster: On the Optimization Landscape of Tensor Decompositions »
Rong Ge · Tengyu Ma -
2017 Poster: GibbsNet: Iterative Adversarial Inference for Deep Graphical Models »
Alex Lamb · R Devon Hjelm · Yaroslav Ganin · Joseph Paul Cohen · Aaron Courville · Yoshua Bengio -
2017 Poster: Modulating early visual processing by language »
Harm de Vries · Florian Strub · Jeremie Mary · Hugo Larochelle · Olivier Pietquin · Aaron Courville -
2017 Spotlight: Modulating early visual processing by language »
Harm de Vries · Florian Strub · Jeremie Mary · Hugo Larochelle · Olivier Pietquin · Aaron Courville -
2017 Oral: On the Optimization Landscape of Tensor Decompositions »
Rong Ge · Tengyu Ma -
2016 : Discussion panel »
Ian Goodfellow · Soumith Chintala · Arthur Gretton · Sebastian Nowozin · Aaron Courville · Yann LeCun · Emily Denton -
2016 : Adversarially Learned Inference (ALI) and BiGANs »
Aaron Courville -
2016 Poster: Professor Forcing: A New Algorithm for Training Recurrent Networks »
Alex M Lamb · Anirudh Goyal · Ying Zhang · Saizheng Zhang · Aaron Courville · Yoshua Bengio -
2016 Oral: Matrix Completion has No Spurious Local Minimum »
Rong Ge · Jason Lee · Tengyu Ma -
2016 Poster: Matrix Completion has No Spurious Local Minimum »
Rong Ge · Jason Lee · Tengyu Ma -
2016 Poster: A Non-generative Framework and Convex Relaxations for Unsupervised Learning »
Elad Hazan · Tengyu Ma -
2015 : Introduction »
Aaron Courville -
2015 Workshop: Multimodal Machine Learning »
Louis-Philippe Morency · Tadas Baltrusaitis · Aaron Courville · Kyunghyun Cho -
2015 Poster: Sum-of-Squares Lower Bounds for Sparse PCA »
Tengyu Ma · Avi Wigderson -
2015 Poster: A Recurrent Latent Variable Model for Sequential Data »
Junyoung Chung · Kyle Kastner · Laurent Dinh · Kratarth Goel · Aaron Courville · Yoshua Bengio -
2014 Poster: On Communication Cost of Distributed Statistical Estimation and Dimensionality »
Ankit Garg · Tengyu Ma · Huy Nguyen -
2014 Oral: On Communication Cost of Distributed Statistical Estimation and Dimensionality »
Ankit Garg · Tengyu Ma · Huy Nguyen -
2014 Poster: Generative Adversarial Nets »
Ian Goodfellow · Jean Pouget-Abadie · Mehdi Mirza · Bing Xu · David Warde-Farley · Sherjil Ozair · Aaron Courville · Yoshua Bengio -
2013 Poster: Multi-Prediction Deep Boltzmann Machines »
Ian Goodfellow · Mehdi Mirza · Aaron Courville · Yoshua Bengio -
2011 Poster: On Tracking The Partition Function »
Guillaume Desjardins · Aaron Courville · Yoshua Bengio -
2009 Poster: An Infinite Factor Model Hierarchy Via a Noisy-Or Mechanism »
Aaron Courville · Douglas Eck · Yoshua Bengio -
2009 Session: Oral Session 3: Deep Learning and Network Models »
Aaron Courville -
2008 Session: Oral session 11: Attention and Mind »
Aaron Courville -
2007 Spotlight: The rat as particle filter »
Nathaniel D Daw · Aaron Courville -
2007 Poster: The rat as particle filter »
Nathaniel D Daw · Aaron Courville