Research map / Reinforcement learning / MDPs and planning
A General Framework for Sample-Efficient Function Approximation in 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 OPERA under the ABC framework with low FE dimension | Low Witness rank MDPs · Low Witness rank | sample complexity: by a factor of dH | Helped |
| Use OPERA with the ABC framework | KNR · KNR | regret: OPERA: \tilde{\cO}\left(H^2 \sqrt{d_\phi T\beta}/\sigma\right); prior attempt: \cO(T^{3/4}) | Helped |
| Use OPERA rather than BiLin-UCB | Linear mixture MDPs · Linear Mixture MDPs | sample complexity: d^3 H^4/\epsilon^2 -> d_{\text{FE}}^2 H^2/\epsilon^2; dH^2 worse | Helped |
| Use OPERA rather than BiLin-UCB | Linear Q*/V* · Linear Q*/V* | sample complexity: \tilde{\cO}(\frac{d^3 H^4}{\epsilon^2}) -> \tilde{\cO}(\frac{d^2 H^2}{\epsilon^2}) | Helped |
| Use OPERA rather than BiLin-UCB | Low occupancy complexity models · Low occupancy complexity | sample complexity: d^3 H^4/\epsilon^2 -> d^2 H^2/\epsilon^2 | Helped |
| Use OPERA with the ABC framework for low eluder dimension models | Low eluder dimension models · Low eluder dimension | regret dependence on eluder dimension: \tilde{\cO}(\sqrt{d}) -> \tilde{\cO}(d) | Helped |
| Use OPERA rather than the cited prior approach | Linear MDPs · Linear MDPs | regret | No clear effect |
| Use OPERA rather than the cited prior approach | Low Bellman rank models · Low Bellman rank | sample complexity dependence on rank: linear in d vs. quadratic | Helped |
Closest papers by meaning
- Reinforcement Learning with General Value Function Approximation: Provably Efficient Approach via Bounded Eluder Dimension (NeurIPS 2020 · similarity 0.91)
- Contextual Decision Processes with low Bellman rank are PAC-Learnable (ICML 2017 · similarity 0.91)
- Efficient Model-Free Exploration in Low-Rank MDPs (NeurIPS 2023 · similarity 0.91)
- A Nearly Optimal and Low-Switching Algorithm for Reinforcement Learning with General Function Approximation (NeurIPS 2024 · similarity 0.90)
- Outcome-Based Online Reinforcement Learning: Algorithms and Fundamental Limits (NeurIPS 2025 · similarity 0.89)
- Reward-Free Kernel-Based Reinforcement Learning (ICML 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.