Research map / Reinforcement learning / MDPs and planning
Non-stationary Reinforcement Learning under General Function Approximation
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 SW-OPEA's confidence-set planning with a sliding window and local transition variation compensation instead of UCB-type algorithms | Non-stationary linear MDPs · Non-stationary linear MDPs | dynamic regret: SW-OPEA: \widetilde{\mathcal{O}}\left(H^{\frac{3}{2}}T^{\frac{1}{2}}\widetilde{d}+HT\widetilde{d}^{\frac{3}{4}}L^{\frac{ | Mixed |
| use SW-OPEA's confidence-set planning rather than a kernel-based algorithm | Non-stationary tabular MDPs · Non-stationary tabular MDPs | dynamic regret: SW-OPEA: \widetilde{O}\left((SA)^{\frac{3}{4}}HTL^{\frac{1}{4}}+TL_\theta\right); kernel-based: \widetilde{O}\left(SA^{\ | Mixed |
| extend SW-OPEA to use bandit rather than full-information reward feedback | SW-OPEA · Non-stationary MDPs with bandit feedback | dynamic regret: \Delta_R(k)+H\Delta_P(k) +\mathcal{O}\left(\sqrt{w} +\frac{H^2k}{\sqrt{w}}\sqrt{d\log[KH|\mathcal{G}|/\delta]} +\frac{H^ | Mixed |
| include all episode-specific Bellman residuals in the complexity measure | Bellman Eluder dimension for non-stationary MDPs · Non-stationary MDPs | complexity dimension: If variations ... are relatively small compared to the universal gap, DBE dimension equals the BE dimension of the first | Helped |
Closest papers by meaning
- DARLING: Detection Augmented Reinforcement Learning with Non-Stationary Guarantees (NeurIPS 2026 · similarity 0.90)
- No-Regret Reinforcement Learning in Smooth MDPs (ICML 2024 · similarity 0.90)
- Local Linearity: the Key for No-regret Reinforcement Learning in Continuous MDPs (NeurIPS 2024 · similarity 0.89)
- NeoRL: Efficient Exploration for Nonepisodic RL (NeurIPS 2024 · similarity 0.89)
- A Nearly Optimal and Low-Switching Algorithm for Reinforcement Learning with General Function Approximation (NeurIPS 2024 · similarity 0.89)
- Provably Efficient RL under Episode-Wise Safety in Constrained MDPs with Linear Function Approximation (NeurIPS 2025 · 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.