Research map / Reinforcement learning / MDPs and planning
Reinforcement Learning with Feedback Graphs
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 feedback-graph side observations in optimistic model-based RL | ORLC in tabular episodic MDPs · Tabular episodic MDPs with feedback graph | worst-case regret bound: replaces a factor of S A ... with the mas-number M | Helped |
| Use feedback-graph side observations in optimistic model-based RL | ORLC in tabular episodic MDPs · Tabular episodic MDPs with symmetric feedback graph | worst-case regret: \tilde O(\sqrt{M H^2 T} + M \hat S H^2) versus \tilde \Omega(\sqrt{\alpha H^2 T}); M = \alpha | Helped |
| Learn tasks jointly using feedback observations across task copies | Multi-task tabular episodic MDPs · Multi-task tabular episodic MDPs | episodes to learn an epsilon-optimal policy for all tasks: \tilde O\left( \frac{ m ( 1 + \epsilon \wh S) M H^2}{\epsilon^2} \right) -> \tilde O\left( \frac{( 1 + \epsilon \wh S) M | Helped |
| Exploit a known, reachable small dominating set to collect side observations | RL Using Dominating Set · Tabular episodic MDPs with known dominating set | sample complexity: M is replaced by \frac{\gamma H}{p_0} in \operatorname{poly}(\epsilon^{-1}) terms | Helped |
| Rely on a small dominating set without accounting for the difficulty of reaching it | RL with feedback graphs · MDPs with a dominating set of size gamma | sample complexity: \Omega\left( \frac{\alpha H^2}{\epsilon^2}\ln\frac 1 \delta \wedge \left(\frac{\gamma H^2}{p_0 \epsilon^2}\ln\frac 1 \de | Hurt |
| Use biased side observations and adjust confidence bonuses for their bias | Optimistic model-based RL with feedback graphs · Tabular episodic MDPs with biased side observations | regret bound: O\left(\sqrt{M H^2 T} \ln \frac{H|\mathcal{X}| T}{\delta} + M \hat S H^2 \ln^3 \frac{H|\mathcal{X}| T}{\delta} + \sqrt{M | Hurt |
Closest papers by meaning
- Efficient Graph Bandit Learning with Side-Observations and Switching Constraints (AAAI 2025 · similarity 0.89)
- Sample Complexity Reduction via Policy Difference Estimation in Tabular Reinforcement Learning (NeurIPS 2024 · similarity 0.88)
- Stochastic contextual bandits with graph feedback: from independence number to MAS number (NeurIPS 2024 · similarity 0.88)
- Minimax Optimal Regret Bound for Reinforcement Learning with Trajectory Feedback (ICML 2025 · similarity 0.88)
- A Theoretical Framework for Partially-Observed Reward States in RLHF (ICLR 2025 · similarity 0.88)
- Sample-Efficient Learning of POMDPs with Multiple Observations In Hindsight (ICLR 2024 · similarity 0.88)
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.