ICML 2021 accepted papers on Reinforcement learning
138 of the 1,184 papers accepted at ICML 2021 (11.7%) are on Reinforcement learning, compared with 9.5% at ICML 2020. They fall into 5 topics.
See ICML 2021 in the interactive map
Submitting to ICML? Checklist →
Working on something in this area? Describe your idea in scime atlas to see the closest papers by meaning and how crowded the spot has become.
Topics and papers
Visual and representation RL · 49 papers
- Decoupling Representation Learning from Reinforcement Learning (63 citations)
- Reinforcement Learning with Prototypical Representations (39 citations)
- Fast active learning for pure exploration in reinforcement learning (25 citations)
- Goal-Conditioned Reinforcement Learning with Imagined Subgoals (24 citations)
- World Model as a Graph: Learning Latent Landmarks for Planning (21 citations)
- Multi-Task Reinforcement Learning with Context-based Representations (20 citations)
- Revisiting Rainbow: Promoting more insightful and inclusive deep reinforcement learning research (20 citations)
- State Entropy Maximization with Random Encoders for Efficient Exploration (20 citations)
- LTL2Action: Generalizing LTL Instructions for Multi-Task RL (19 citations)
- Discovering symbolic policies with deep reinforcement learning (16 citations)
- Grounding Language to Entities and Dynamics for Generalization in Reinforcement Learning (15 citations)
- PEBBLE: Feedback-Efficient Interactive Reinforcement Learning via Relabeling Experience and Unsupervised Pre-training (13 citations)
and 37 more in the interactive map.
Policy gradient and actor-critic · 48 papers
- Phasic Policy Gradient (51 citations)
- SUNRISE: A Simple Unified Framework for Ensemble Learning in Deep Reinforcement Learning (45 citations)
- Batch Value-function Approximation with Only Realizability (32 citations)
- Generalizable Episodic Memory for Deep Reinforcement Learning (19 citations)
- Accelerating Safe Reinforcement Learning with Constraint-mismatched Baseline Policies (18 citations)
- CRPO: A New Approach for Safe Reinforcement Learning with Convergence Guarantee (16 citations)
- Learning and Planning in Complex Action Spaces (16 citations)
- Muesli: Combining Improvements in Policy Optimization (12 citations)
- On-Policy Deep Reinforcement Learning for the Average-Reward Criterion (12 citations)
- Spectral Normalisation for Deep Reinforcement Learning: An Optimisation Perspective (12 citations)
- Low-Precision Reinforcement Learning: Running Soft Actor-Critic in Half Precision (11 citations)
- Sparse Feature Selection Makes Batch Reinforcement Learning More Sample Efficient (11 citations)
and 36 more in the interactive map.
MDPs and planning · 16 papers
- Exponential Lower Bounds for Batch Reinforcement Learning: Batch RL can be Exponentially Harder than Online RL (16 citations)
- Learning and Planning in Average-Reward Markov Decision Processes (15 citations)
- Robust Reinforcement Learning using Least Squares Policy Iteration with Provable Performance Guarantees (15 citations)
- Quantum algorithms for reinforcement learning with a generative model (10 citations)
- PC-MLP: Model-based Reinforcement Learning with Policy Cover Guided Exploration (5 citations)
- Reward Identification in Inverse Reinforcement Learning (5 citations)
- Safe Reinforcement Learning Using Advantage-Based Intervention (5 citations)
- On Reward-Free RL with Kernel and Neural Function Approximations: Single-Agent MDP and Markov Game (4 citations)
- Combining Pessimism with Optimism for Robust and Efficient Model-Based Deep Reinforcement Learning (3 citations)
- Near Optimal Reward-Free Reinforcement Learning (3 citations)
- A Lower Bound for the Sample Complexity of Inverse Reinforcement Learning (2 citations)
- Decision-Making Under Selective Labels: Optimal Finite-Domain Policies and Beyond (2 citations)
and 4 more in the interactive map.
Imitation learning · 15 papers
- Adversarial Option-Aware Hierarchical Imitation Learning (8 citations)
- Inverse Constrained Reinforcement Learning (6 citations)
- Robust Asymmetric Learning in POMDPs (6 citations)
- Hyperparameter Selection for Imitation Learning (5 citations)
- PsiPhi-Learning: Reinforcement Learning with Demonstrations using Successor Features and Inverse Temporal Difference Learning (5 citations)
- Interactive Learning from Activity Description (3 citations)
- Cross-domain Imitation from Observations (2 citations)
- Demonstration-Conditioned Reinforcement Learning for Few-Shot Imitation (2 citations)
- Provably Efficient Learning of Transferable Rewards (2 citations)
- Imitation by Predicting Observations (1 citation)
- Efficient Performance Bounds for Primal-Dual Reinforcement Learning from Demonstrations
- Keyframe-Focused Visual Imitation Learning
and 3 more in the interactive map.
Offline RL · 10 papers
- Offline Reinforcement Learning with Fisher Divergence Critic Regularization (49 citations)
- Uncertainty Weighted Actor-Critic for Offline Reinforcement Learning (45 citations)
- EMaQ: Expected-Max Q-Learning Operator for Simple Yet Effective Offline and Online RL (24 citations)
- Is Pessimism Provably Efficient for Offline RL? (23 citations)
- Representation Matters: Offline Pretraining for Sequential Decision Making (22 citations)
- Actionable Models: Unsupervised Offline Reinforcement Learning of Robotic Skills (16 citations)
- Offline Meta-Reinforcement Learning with Advantage Weighting (11 citations)
- Instabilities of Offline RL with Pre-Trained Neural Representation (8 citations)
- OptiDICE: Offline Policy Optimization via Stationary Distribution Correction Estimation (7 citations)
- Offline Reinforcement Learning with Pseudometric Learning (5 citations)
