Contrastive Behavioral Similarity Embeddings for Generalization in Reinforcement Learning
Preprint
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.Keywords
All Related Versions
This publication has 4 references indexed in Scilit:
- Rotation, Translation, and Cropping for Zero-Shot Generalization2020
- Neuroevolution of self-interpretable agents2020
- A Tour of Reinforcement Learning: The View from Continuous ControlAnnual Review of Control, Robotics, and Autonomous Systems, 2019
- Equivalence notions and model minimization in Markov decision processesArtificial Intelligence, 2003