Regret Bounds for Kernel-Based Reinforcement Learning

by   Omar Darwiche Domingues, et al.

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.


page 1

page 2

page 3

page 4


A Kernel-Based Approach to Non-Stationary Reinforcement Learning in Metric Spaces

In this work, we propose KeRNS: an algorithm for episodic reinforcement ...

A Tractable Algorithm For Finite-Horizon Continuous Reinforcement Learning

We consider the finite horizon continuous reinforcement learning problem...

Exploration in Structured Reinforcement Learning

We address reinforcement learning problems with finite state and action ...

Exploration in Action Space

Parameter space exploration methods with black-box optimization have rec...

Reinforcement Leaning in Feature Space: Matrix Bandit, Kernels, and Regret Bound

Exploration in reinforcement learning (RL) suffers from the curse of dim...

Zooming for Efficient Model-Free Reinforcement Learning in Metric Spaces

Despite the wealth of research into provably efficient reinforcement lea...

No-Regret Exploration in Goal-Oriented Reinforcement Learning

Many popular reinforcement learning problems (e.g., navigation in a maze...