A Hitchhiker's Guide to Statistical Comparisons of Reinforcement Learning Algorithms

Abstract

Consistently checking the statistical significance of experimental results is the first mandatory step towards reproducible science. This paper presents a hitchhiker's guide to rigorous comparisons of reinforcement learning algorithms. After introducing the concepts of statistical testing, we review the relevant statistical tests and compare them empirically in terms of false positive rate and statistical power as a function of the sample size (number of seeds) and effect size. We further investigate the robustness of these tests to violations of the most common hypotheses (normal distributions, same distributions, equal variances). Beside simulations, we compare empirical distributions obtained by running Soft-Actor Critic and Twin-Delayed Deep Deterministic Policy Gradient on Half-Cheetah. We conclude by providing guidelines and code to perform rigorous comparisons of RL algorithm performances.

8 pages + supplementary material

Cited by in corpus (12)

• Language as a Cognitive Tool to Imagine Goals in Curiosity-Driven Exploration
• Deep Reinforcement Learning at the Edge of the Statistical Precipice
• EpidemiOptim: A Toolbox for the Optimization of Control Policies in Epidemiological Models
• A Domain-Agnostic Approach for Characterization of Lifelong Learning Systems
• Grounding Language to Autonomously-Acquired Skills via Goal Generation
• Curriculum Learning for Safe Mapless Navigation
• Novelty Search in Representational Space for Sample Efficient Exploration
• Deep Sets for Generalization in RL
• Evaluating the Safety of Deep Reinforcement Learning Models using Semi-Formal Verification
• Detecting Rewards Deterioration in Episodic Reinforcement Learning
• Is High Variance Unavoidable in RL? A Case Study in Continuous Control
• Improved Cooperation by Exploiting a Common Signal