Online Bootstrap Inference For Policy Evaluation in Reinforcement Learning

TL;DR

This paper investigates the use of the online bootstrap method for statistical inference in reinforcement learning, specifically for policy evaluation using TD and GTD algorithms, demonstrating its consistency and effectiveness.

Contribution

It introduces the online bootstrap approach for RL policy evaluation, extending its applicability to Markov noise settings and providing theoretical and empirical validation.

Findings

The online bootstrap is distributionally consistent for RL policy evaluation.

The method performs well across various real RL environments.

It extends bootstrap inference to Markov noise scenarios in RL.

Abstract

The recent emergence of reinforcement learning has created a demand for robust statistical inference methods for the parameter estimates computed using these algorithms. Existing methods for statistical inference in online learning are restricted to settings involving independently sampled observations, while existing statistical inference methods in reinforcement learning (RL) are limited to the batch setting. The online bootstrap is a flexible and efficient approach for statistical inference in linear stochastic approximation algorithms, but its efficacy in settings involving Markov noise, such as RL, has yet to be explored. In this paper, we study the use of the online bootstrap method for statistical inference in RL. In particular, we focus on the temporal difference (TD) learning and Gradient TD (GTD) learning algorithms, which are themselves special instances of linear stochastic approximation under Markov noise. The method is shown to be distributionally consistent for statistical inference in policy evaluation, and numerical experiments are included to demonstrate the effectiveness of this algorithm at statistical inference tasks across a range of real RL environments.