Research map / Reinforcement learning / Policy gradient and actor-critic
Statistics and Samples in Distributional Reinforcement Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use sample imputation instead of directly treating expectile estimates as target samples (ablation) | EDRL on N-Chain · 15-Chain policy evaluation | expectile estimation error | Helped |
| Increase the number of learned expectiles in EDRL-Naive (ablation) | EDRL-Naive · N-Chain policy evaluation | expectile estimation error | Hurt |
| Use an imputation strategy when learning Huber quantiles (ablation) | Huber-QDRL · 15-Chain policy evaluation | Huber quantile estimation error | Helped |
| Use ER-DQN with imputation rather than the naive ER-DQN without imputation (ablation) | ER-DQN · Atari-57 | mean human normalised score | Helped |
| Use ER-DQN with expectile statistics instead of QR-DQN with Huber quantile statistics | Atari-57 · Atari-57 | mean human normalised score | Helped |
| Include the mean as a statistic in EDRL | EDRL · Tabular control | expected returns | Helped |
| Use EDRL rather than CDRL or QDRL for estimating action means | Tabular control · Five-state MDP | greedy policy optimality | Helped |
Closest papers by meaning
- Near-Minimax-Optimal Distributional Reinforcement Learning with a Generative Model (NeurIPS 2024 · similarity 0.90)
- Action Gaps and Advantages in Continuous-Time Distributional Reinforcement Learning (NeurIPS 2024 · similarity 0.89)
- A Distributional Analogue to the Successor Representation (ICML 2024 · similarity 0.89)
- Distributional Reinforcement Learning with Regularized Wasserstein Loss (NeurIPS 2024 · similarity 0.88)
- Categorical Distributional Reinforcement Learning with Kullback-Leibler Divergence: Convergence and Asymptotics (ICML 2025 · similarity 0.88)
- Diverse Projection Ensembles for Distributional Reinforcement Learning (ICLR 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.