Measuring the Reliability of Reinforcement Learning Algorithms

TL;DR

This paper introduces a set of quantitative metrics and statistical tools to evaluate the reliability of reinforcement learning algorithms, focusing on variability and risk during and after training, with open-source implementation and practical recommendations.

Contribution

The paper presents a novel, general-purpose suite of metrics and statistical tests for assessing RL reliability, along with an open-source library for practical application.

Findings

Metrics effectively quantify variability and risk in RL.

Comparative analysis of common RL algorithms using the metrics.

Open-source tools facilitate rigorous reliability evaluation.

Abstract

Lack of reliability is a well-known issue for reinforcement learning (RL) algorithms. This problem has gained increasing attention in recent years, and efforts to improve it have grown substantially. To aid RL researchers and production users with the evaluation and improvement of reliability, we propose a set of metrics that quantitatively measure different aspects of reliability. In this work, we focus on variability and risk, both during training and after learning (on a fixed policy). We designed these metrics to be general-purpose, and we also designed complementary statistical tests to enable rigorous comparisons on these metrics. In this paper, we first describe the desired properties of the metrics and their design, the aspects of reliability that they measure, and their applicability to different scenarios. We then describe the statistical tests and make additional practical recommendations for reporting results. The metrics and accompanying statistical tools have been made available as an open-source library at https://github.com/google-research/rl-reliability-metrics. We apply our metrics to a set of common RL algorithms and environments, compare them, and analyze the results.