203
v1v2v3 (latest)

Kernel-Based Reinforcement Learning: A Finite-Time Analysis

International Conference on Machine Learning (ICML), 2020
Abstract

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. For problems with KK episodes and horizon HH, we provide a regret bound of O~(H3K2d2d+1)\widetilde{O}\left( H^3 K^{\frac{2d}{2d+1}}\right), where dd is the covering dimension of the joint state-action space. This is the first regret bound for kernel-based RL using smoothing kernels, which requires very weak assumptions on the MDP and has been previously applied to a wide range of tasks. We empirically validate our approach in continuous MDPs with sparse rewards.

View on arXiv
Comments on this paper