Research map / Bandits and online learning / RL theory and MDPs
First-Order Regret in Reinforcement Learning with Linear Function Approximation: A Robust Estimation Approach
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 robust Catoni estimation instead of least-squares estimation for next-state value expectations | FORCE in linear MDPs · Linear MDPs | regret upper bound: LSVI-UCB: \widetilde{\cO}(\sqrt{d^3 H^4 K}) -> FORCE: \widetilde{\cO}(\sqrt{d^3 H^3 \cdot V^\star_1 \cdot K} + d^{3.5}H^ | Helped |
| Use covariance eigenvectors to form the linear approximation | FORCE · Linear MDPs | regret upper bound: \sqrt{d^3 H^3 V^\star_1 K} -> \sqrt{d^4 H^3 V^\star_1 K} | Mixed |
| Use variance-normalized weighted least squares with a Bernstein-style self-normalized bound | Linear MDP regression · Linear MDPs | regret upper bound: (\sqrt{d} + \frac{H}{\sigminb}) \sqrt{\sigminb^2 HK} \ge H \sqrt{HK} | Hurt |
Closest papers by meaning
- Tackling Heavy-Tailed Rewards in Reinforcement Learning with Function Approximation: Minimax Optimal and Instance-Dependent Regret Bounds (NeurIPS 2023 · similarity 0.92)
- Achieving Constant Regret in Linear Markov Decision Processes (NeurIPS 2024 · similarity 0.92)
- Minimax Regret Bounds for Reinforcement Learning (ICML 2017 · similarity 0.92)
- Regret Minimization for Reinforcement Learning by Evaluating the Optimal Bias Function (NeurIPS 2019 · similarity 0.92)
- Horizon-Free Regret for Linear Markov Decision Processes (ICLR 2024 · similarity 0.92)
- Local Linearity: the Key for No-regret Reinforcement Learning in Continuous MDPs (NeurIPS 2024 · similarity 0.90)
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.