Research map / Reinforcement learning / MDPs and planning
Improved Sample Complexity for Reward-free Reinforcement Learning under Low-rank MDPs
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 RAFFLE's model-based reward-free exploration | Reward-free RL under low-rank MDPs · ε-optimal policy | sample complexity: \tilde{O}(\frac{H^{22}K^9d^7}{\epsilon^{10}}) -> \tilde{O}(\frac{H^3d^2K(d^2+K)}{\epsilon^2}) | Helped |
| Use RAFFLE for system identification | Reward-free low-rank MDPs · ε-accurate system identification | sample complexity: \tilde{O}(\frac{H^{22}K^9d^7}{\epsilon^{10}}) -> \tilde{O}(\frac{H^3d^2K(d^2+K)}{\epsilon^2}) | Helped |
| Use RAFFLE rather than the linear-MDP reward-known bound | Reward-known RL under low-rank MDPs · Low-rank MDPs | sample complexity dependence on H: \Theta(H^2) | Helped |
| Add a trajectory-wise exploration-driven reward based on feature uncertainty | RAFFLE · Reward-free low-rank MDPs | sample complexity | Helped |
| Truncate the estimated value function | RAFFLE · Reward-free low-rank MDPs | sample complexity dependence on H: H^2 | Helped |
| Terminate exploration when the estimated uncertainty value is below a threshold | RAFFLE · Reward-free low-rank MDPs | sample complexity | Helped |
| Collect only one sample trajectory per exploration policy at each episode | RAFFLE · Reward-free low-rank MDPs | sample complexity | Helped |
| Learn the representation by regression on targets computed from RAFFLE's estimated model, without further environment interaction | RepLearn · Near-accurate representation learning | representation divergence: O(\frac{\epsilon d C_\mathrm{min}}{C_D}+\frac{d}{C_D}\sqrt{\frac{\log{\frac{2}{\delta}}}{TN_f}}) | Helped |
Closest papers by meaning
- Reward-Free RL is No Harder Than Reward-Aware RL in Linear Markov Decision Processes (ICML 2022 · similarity 0.92)
- Towards Minimax Optimal Reward-free Reinforcement Learning in Linear MDPs (ICLR 2023 · similarity 0.92)
- Sample-Efficient Reinforcement Learning for Linearly-Parameterized MDPs with a Generative Model (NeurIPS 2021 · similarity 0.91)
- Model-Free Reinforcement Learning: from Clipped Pseudo-Regret to Sample Complexity (ICML 2021 · similarity 0.91)
- Multi-Reward Best Policy Identification (NeurIPS 2024 · similarity 0.90)
- Optimal Single-Policy Sample Complexity and Transient Coverage for Average-Reward Offline RL (NeurIPS 2025 · 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.