17
104

Is Reinforcement Learning More Difficult Than Bandits? A Near-optimal Algorithm Escaping the Curse of Horizon

Abstract

Episodic reinforcement learning and contextual bandits are two widely studied sequential decision-making problems. Episodic reinforcement learning generalizes contextual bandits and is often perceived to be more difficult due to long planning horizon and unknown state-dependent transitions. The current paper shows that the long planning horizon and the unknown state-dependent transitions (at most) pose little additional difficulty on sample complexity. We consider the episodic reinforcement learning with SS states, AA actions, planning horizon HH, total reward bounded by 11, and the agent plays for KK episodes. We propose a new algorithm, \textbf{M}onotonic \textbf{V}alue \textbf{P}ropagation (MVP), which relies on a new Bernstein-type bonus. Compared to existing bonus constructions, the new bonus is tighter since it is based on a well-designed monotonic value function. In particular, the \emph{constants} in the bonus should be subtly setting to ensure optimism and monotonicity. We show MVP enjoys an O((SAK+S2A)\polylog(SAHK))O\left(\left(\sqrt{SAK} + S^2A\right) \poly\log \left(SAHK\right)\right) regret, approaching the Ω(SAK)\Omega\left(\sqrt{SAK}\right) lower bound of \emph{contextual bandits} up to logarithmic terms. Notably, this result 1) \emph{exponentially} improves the state-of-the-art polynomial-time algorithms by Dann et al. [2019] and Zanette et al. [2019] in terms of the dependency on HH, and 2) \emph{exponentially} improves the running time in [Wang et al. 2020] and significantly improves the dependency on SS, AA and KK in sample complexity.

View on arXiv
Comments on this paper