Contrastive Behavioral Similarity Embeddings for Generalization in Reinforcement Learning

TL;DR

This paper introduces a new policy similarity metric and a contrastive learning method to embed state similarities, significantly enhancing reinforcement learning generalization across diverse benchmarks.

Contribution

The paper proposes a theoretically grounded policy similarity metric and a contrastive embedding method, explicitly leveraging the sequential structure in reinforcement learning for better generalization.

Findings

PSEs improve generalization on LQR with spurious correlations

PSEs enhance performance on pixel-based jumping tasks

PSEs outperform baselines on Distracting DM Control Suite

Abstract

Reinforcement learning methods trained on few environments rarely learn policies that generalize to unseen environments. To improve generalization, we incorporate the inherent sequential structure in reinforcement learning into the representation learning process. This approach is orthogonal to recent approaches, which rarely exploit this structure explicitly. Specifically, we introduce a theoretically motivated policy similarity metric (PSM) for measuring behavioral similarity between states. PSM assigns high similarity to states for which the optimal policies in those states as well as in future states are similar. We also present a contrastive representation learning procedure to embed any state similarity metric, which we instantiate with PSM to obtain policy similarity embeddings (PSEs). We demonstrate that PSEs improve generalization on diverse benchmarks, including LQR with spurious correlations, a jumping task from pixels, and Distracting DM Control Suite.