Regret Bounds for Kernel-Based Reinforcement Learning

04/12/2020 ∙ by Omar Darwiche Domingues, et al. ∙ 0

We consider the exploration-exploitation dilemma in finite-horizon reinforcement learning problems whose state-action space is endowed with a metric. We introduce Kernel-UCBVI, a model-based optimistic algorithm that leverages the smoothness of the MDP and a non-parametric kernel estimator of the rewards and transitions to efficiently balance exploration and exploitation. Unlike existing approaches with regret guarantees, it does not use any kind of partitioning of the state-action space. For problems with K episodes and horizon H, we provide a regret bound of O( H^3 K^max(1/2, 2d/2d+1)), where d is the covering dimension of the joint state-action space. We empirically validate Kernel-UCBVI on discrete and continuous MDPs.



There are no comments yet.


page 1

page 2

page 3

page 4

This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.