Research map / Bandits and online learning / RL theory and MDPs
Provably Efficient Reinforcement Learning with Linear Function Approximation under Adaptivity Constraints
Tianhao Wang, Dongruo Zhou, Quanquan Gu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| limit adaptivity to B=sqrt(K) batches instead of full adaptivity | LSVI-UCB-Batch · Hard-to-learn linear MDP (H=10, K=2500, d=13) | average regret | No clear effect |
| use rare-switch policy updates with constant eta instead of updating every episode | LSVI-UCB-RareSwitch · Hard-to-learn linear MDP (H=10, K=2500, d=13) | average regret | No clear effect |
| use rare-switch updates instead of fixed-batch updates | LSVI-UCB-RareSwitch vs Batch · Hard-to-learn linear MDP (H=10, K=2500, d=13) | regret gap to LSVI-UCB | Improved |
| decrease number of batches B (10 to 50 tested) (ablation) | LSVI-UCB-Batch · Hard-to-learn linear MDP (H=10, K=2500, d=13) | average regret | Hurt |
| increase switching threshold eta (2 to 32 tested) (ablation) | LSVI-UCB-RareSwitch · Hard-to-learn linear MDP (H=10, K=2500, d=13) | average regret | Hurt |
| compare against uniformly random policy | LSVI-UCB-Batch and RareSwitch · Hard-to-learn linear MDP (H=10, K=2500, d=13) | average regret | Improved |
| introduce model misspecification zeta = 0.05, 0.1, 0.2, 0.4 (ablation) | LSVI-UCB-Batch (B=50), RareSwitch (eta=8) · Misspecified linear MDP | average regret | Mixed |
Closest papers by meaning
- A General Framework for Sequential Decision-Making under Adaptivity Constraints (ICML 2024 · similarity 0.91)
- Provably Efficient RL under Episode-Wise Safety in Constrained MDPs with Linear Function Approximation (NeurIPS 2025 · similarity 0.89)
- Near-Optimal Reinforcement Learning with Self-Play under Adaptivity Constraints (ICML 2024 · similarity 0.89)
- A Nearly Optimal and Low-Switching Algorithm for Reinforcement Learning with General Function Approximation (NeurIPS 2024 · similarity 0.89)
- Achieving Constant Regret in Linear Markov Decision Processes (NeurIPS 2024 · similarity 0.89)
- Sample-Efficiency in Multi-Batch Reinforcement Learning: The Need for Dimension-Dependent Adaptivity (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.
