作者
Sergey Levine
Robotics / Reinforcement Learning
Implicit Under-Parameterization Inhibits Data-Efficient Deep Reinforcement Learning
Comments ICLR 2021. First two authors contributed equally. Website: https://agarwl.github.io/iup/
Offline Reinforcement Learning with Implicit Q-Learning
Legged Robots that Keep on Learning: Fine-Tuning Locomotion Policies in the Real World
Comments Project website: https://sites.google.com/berkeley.edu/fine-tuning-locomotion
The Information Geometry of Unsupervised Reinforcement Learning
Bridge Data: Boosting Generalization of Robotic Skills with Cross-Domain Datasets
Training on Test Data with Bayesian Adaptation for Covariate Shift
A Workflow for Offline Model-Free Robotic Reinforcement Learning
Comments CoRL 2021. Project Website: https://sites.google.com/view/offline-rl-workflow. First two authors contributed equally
Conservative Data Sharing for Multi-Task Offline Reinforcement Learning
Robust Predictable Control
Comments Project site with videos and code: https://ben-eysenbach.github.io/rpc
Pragmatic Image Compression for Human-in-the-Loop Decision-Making
Model-Based Reinforcement Learning via Latent-Space Collocation
Comments International Conference on Machine Learning (ICML), 2021. Videos and code at https://orybkin.github.io/latco/
Modularity in Reinforcement Learning via Algorithmic Independence in Credit Assignment
Comments Long Presentation at the Thirty-eighth International Conference on Machine Learning (ICML) 2021. 21 pages, 11 figures. v2: updated acknowledgments. v3: clarified that the internal function nodes of the credit assignment mechanism are not considered O(1)
Offline Meta-Reinforcement Learning with Advantage Weighting
Comments ICML 2021; for code & project info, see http://sites.google.com/view/macaw-metarl
MURAL: Meta-Learning Uncertainty-Aware Rewards for Outcome-Driven Reinforcement Learning
Comments Accepted to ICML 2021. First two authors contributed equally
WILDS: A Benchmark of in-the-Wild Distribution Shifts
Conservative Objective Models for Effective Offline Model-Based Optimization
Comments ICML 2021. First two authors contributed equally. Code at: https://github.com/brandontrabucco/design-baselines/blob/c65a53fe1e6567b740f0adf60c5db9921c1f2330/design_baselines/coms_cleaned/__init__.py
Why Generalization in RL is Difficult: Epistemic POMDPs and Implicit Partial Observability
Comments First two authors contributed equally
Simple and Effective VAE Training with Calibrated Decoders
Comments International Conference on Machine Learning (ICML), 2021. Project website is at https://orybkin.github.io/sigma-vae/
FitVid: Overfitting in Pixel-Level Video Prediction
Emergent Social Learning via Multi-agent Reinforcement Learning
Comments 14 pages, 19 figures. To be published in ICML 2021
Meta-World: A Benchmark and Evaluation for Multi-Task and Meta Reinforcement Learning
Comments This is an update version of a manuscript that originally appeared at CoRL 2019. Videos are here: meta-world.github.io, open-sourced code are available at: https://github.com/rlworkgroup/metaworld, and the baselines can be found at https://github.com/rlworkgroup/garage
Which Mutual-Information Representation Learning Objectives are Sufficient for Control?
Comments 18 pages, 11 figures
What Can I Do Here? Learning New Skills by Imagining Visual Affordances
Comments 10 pages, 10 figures. Presented at ICRA 2021. Project website: https://sites.google.com/view/val-rl
Actionable Models: Unsupervised Offline Reinforcement Learning of Robotic Skills
PsiPhi-Learning: Reinforcement Learning with Demonstrations using Successor Features and Inverse Temporal Difference Learning
Comments The last two authors contributed equally. Accepted at ICML 2021
Variational Empowerment as Representation Learning for Goal-Based Reinforcement Learning
Comments Accepted at International Conference on Machine Learning (ICML) 2021
Policy Information Capacity: Information-Theoretic Measure for Task Complexity in Deep Reinforcement Learning
Comments Accepted to ICML2021. The code is available at: https://github.com/frt03/pic
SimGAN: Hybrid Simulator Identification for Domain Adaptation via Adversarial Reinforcement Learning
Comments ICRA 2021, Code Available at: https://github.com/jyf588/SimGAN ; Accompanying Video: https://youtu.be/McKOGllO7nc
