Introduces Good Policy Identification (GPI) and BEE-GPI algorithm whose sample complexity for positive instances has log(1/δ) coefficient O(H²/(V*−μ0)²) independent of state and action space sizes.
International Conference on Machine Learning , pages=
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.LG 2years
2026 2verdicts
UNVERDICTED 2representative citing papers
Presents an optimistic FTRL algorithm for online episodic tabular MDPs with unknown transitions that attains data-dependent regret bounds including first-order, second-order, path-length, and polylog(T) gap-dependent bounds in the stochastic case.
citing papers explorer
-
Pure Exploration for a Good Policy in Reinforcement Learning with Bandit Feedback
Introduces Good Policy Identification (GPI) and BEE-GPI algorithm whose sample complexity for positive instances has log(1/δ) coefficient O(H²/(V*−μ0)²) independent of state and action space sizes.
-
Policy Optimization Achieves Data-Dependent Regret Bounds in MDPs with Unknown Transitions
Presents an optimistic FTRL algorithm for online episodic tabular MDPs with unknown transitions that attains data-dependent regret bounds including first-order, second-order, path-length, and polylog(T) gap-dependent bounds in the stochastic case.