Research map / Bandits and online learning / RL theory and MDPs
No-Regret Online Reinforcement Learning with Adversarial Losses and Transitions
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 |
|---|---|---|---|
| Add amortized bonuses and log-barrier regularization to UOB-REPS | UOB-REPS · Adversarial episodic MDPs with known transition corruption | regret: \order \rbr{ L|S| \sqrt{|A| T \log\rbr{\iota} } + L|S|^4|A|\log^2 (\iota) +\cortran L|S|^4|A| \log (\iota)} | Helped |
| Remove the requirement to know transition corruption using a black-box reduction | Modified UOB-REPS · Adversarial episodic MDPs with unknown transition corruption | regret: \otil\left(\sqrt{\beta_1 T} + \beta_2 + \beta_3 \cortran \right) | Helped |
| Use optimistic transitions with adaptive log-barrier FTRL | Modified UOB-REPS · Adversarial episodic MDPs satisfying the gap condition | regret against \pi^\star: U +\sqrt{ U\corloss}+ \rbr{\cortran+1} L^2 |S|^4 |A|^2 \log^2 (\iota) | Helped |
| Replace negative entropy with a log-barrier regularizer | UOB-REPS for adversarial transitions · Adversarial episodic MDPs | stability term | Helped |
| Use optimistic transitions instead of an additional transition bonus | Gap-dependent FTRL · Adversarial episodic MDPs with a gap condition | gap-dependent regret bound | Helped |
| Discretize feedback probabilities and equalize feedback rates across instances | STABILISE · Stable corruption-robust learning protocol | regret dependence on transition corruption | Helped |
Closest papers by meaning
- Towards Optimal Regret in Adversarial Linear MDPs with Bandit Feedback (ICLR 2024 · similarity 0.93)
- Beating Adversarial Low-Rank MDPs with Unknown Transition and Bandit Feedback (NeurIPS 2024 · similarity 0.91)
- Dynamic Regret of Adversarial MDPs with Unknown Transition and Linear Function Approximation (AAAI 2024 · similarity 0.91)
- Learning Adversarial MDPs with Stochastic Hard Constraints (ICML 2025 · similarity 0.90)
- No-Regret Learning Under Adversarial Resource Constraints: A Spending Plan Is All You Need! (NeurIPS 2025 · similarity 0.89)
- Adapting to Stochastic and Adversarial Losses in Episodic MDPs with Aggregate Bandit Feedback (NeurIPS 2025 · 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.