Bridging State and History Representations: Understanding Self-Predictive RL

TL;DR

This paper unifies various representation learning methods in reinforcement learning under the concept of self-predictive abstraction, providing theoretical insights and a minimalist algorithm applicable to different RL environments.

Contribution

It reveals the shared foundation of state and history representations in RL, introduces a theoretical framework, and proposes a simple algorithm for learning self-predictive representations.

Findings

Unified view of state and history representations as self-predictive abstractions

Theoretical analysis of objectives and optimization techniques like stop-gradient

Validated algorithm across MDPs, distractor environments, and POMDPs

Abstract

Representations are at the core of all deep reinforcement learning (RL) methods for both Markov decision processes (MDPs) and partially observable Markov decision processes (POMDPs). Many representation learning methods and theoretical frameworks have been developed to understand what constitutes an effective representation. However, the relationships between these methods and the shared properties among them remain unclear. In this paper, we show that many of these seemingly distinct methods and frameworks for state and history abstractions are, in fact, based on a common idea of self-predictive abstraction. Furthermore, we provide theoretical insights into the widely adopted objectives and optimization, such as the stop-gradient technique, in learning self-predictive representations. These findings together yield a minimalist algorithm to learn self-predictive representations for states and histories. We validate our theories by applying our algorithm to standard MDPs, MDPs with distractors, and POMDPs with sparse rewards. These findings culminate in a set of preliminary guidelines for RL practitioners.