English

Kernel-Based Reinforcement Learning: A Finite-Time Analysis

Machine Learning 2022-03-25 v3 Machine Learning

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.

Keywords

Cite

@article{arxiv.2004.05599,
  title  = {Kernel-Based Reinforcement Learning: A Finite-Time Analysis},
  author = {Omar Darwiche Domingues and Pierre Ménard and Matteo Pirotta and Emilie Kaufmann and Michal Valko},
  journal= {arXiv preprint arXiv:2004.05599},
  year   = {2022}
}

Comments

Update following the publication in ICML 2021, including fixed typos

R2 v1 2026-06-23T14:48:29.652Z