Research map / Bandits and online learning / RL theory and MDPs
Reinforcement Learning for Non-Stationary Markov Decision Processes: The Blessing of (More) Optimism
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 |
|---|---|---|---|
| Add confidence widening to transition confidence regions | Sliding-window UCRL2 for non-stationary MDPs · Non-stationary MDPs | dynamic regret: SWUCRL2-CW: \tilde{O}\left( D_\text{max} (B_r + B_p+1)^{\frac{1}{4}} S^{\frac{2}{3}}A^{\frac{1}{2}}T^{\frac{3}{4}}\right | Helped |
| Use a fixed, oblivious choice of window and widening parameters rather than adaptively tuning them | SWUCRL2-CW · Non-stationary MDPs with unknown variation budgets | dynamic regret: \tilde{O}\left( D_\text{max} (B_r + B_p+1) S^{\frac{2}{3}}A^{\frac{1}{2}}T^{\frac{3}{4}}\right) -> \tilde{O}(D_{\max}(B_ | Hurt |
| Remove confidence widening (set η to 0) (ablation) | Sliding-window UCRL2 · Non-stationary MDPs | diameter of MDP in confidence region: \Omega(\sqrt{W/\log W}) | Hurt |
| Use sliding-window estimates without confidence widening (ablation) | Naive sliding-window UCRL2 · Non-stationary MDPs | diameter of empirical MDP: \Theta(W) | Hurt |
| Use the parameter-free BORL tuning framework | SWUCRL2-CW · Non-stationary MDPs with unknown variation budgets | dynamic regret: \tilde{O}(D_{\max}(B_r+B_p+1)^{\frac{1}{4}}S^{\frac{2}{3}}A^{\frac{1}{2}}T^{\frac{3}{4}}) | Helped |
Closest papers by meaning
- DARLING: Detection Augmented Reinforcement Learning with Non-Stationary Guarantees (NeurIPS 2026 · similarity 0.91)
- MetaCURL: Non-stationary Concave Utility Reinforcement Learning (NeurIPS 2024 · similarity 0.90)
- Efficient Exploration in Average-Reward Constrained Reinforcement Learning: Achieving Near-Optimal Regret With Posterior Sampling (ICML 2024 · similarity 0.90)
- No-Regret Reinforcement Learning in Smooth MDPs (ICML 2024 · similarity 0.90)
- Pausing Policy Learning in Non-stationary Reinforcement Learning (ICML 2024 · similarity 0.89)
- Horizon-Free Regret for Linear Markov Decision Processes (ICLR 2024 · similarity 0.89)
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.