Embed to Control Partially Observed Systems: Representation Learning with Provable Sample Efficiency

TL;DR

This paper introduces Embed to Control (ETC), a novel reinforcement learning algorithm that learns low-dimensional representations of observations and histories in POMDPs, achieving polynomial sample complexity and bridging representation learning with policy optimization.

Contribution

The paper presents ETC, the first sample-efficient algorithm that combines representation learning and policy optimization in POMDPs with infinite observation and state spaces.

Findings

ETC attains $O(1/\epsilon^2)$ sample complexity for low-rank POMDPs.

ETC scales polynomially with the horizon and intrinsic dimension.

The algorithm effectively learns minimal sufficient representations for control.

Abstract

Reinforcement learning in partially observed Markov decision processes (POMDPs) faces two challenges. (i) It often takes the full history to predict the future, which induces a sample complexity that scales exponentially with the horizon. (ii) The observation and state spaces are often continuous, which induces a sample complexity that scales exponentially with the extrinsic dimension. Addressing such challenges requires learning a minimal but sufficient representation of the observation and state histories by exploiting the structure of the POMDP. To this end, we propose a reinforcement learning algorithm named Embed to Control (ETC), which learns the representation at two levels while optimizing the policy.~(i) For each step, ETC learns to represent the state with a low-dimensional feature, which factorizes the transition kernel. (ii) Across multiple steps, ETC learns to represent the full history with a low-dimensional embedding, which assembles the per-step feature. We integrate (i) and (ii) in a unified framework that allows a variety of estimators (including maximum likelihood estimators and generative adversarial networks). For a class of POMDPs with a low-rank structure in the transition kernel, ETC attains an $O(1/\epsilon^2)$ sample complexity that scales polynomially with the horizon and the intrinsic dimension (that is, the rank). Here $\epsilon$ is the optimality gap. To our best knowledge, ETC is the first sample-efficient algorithm that bridges representation learning and policy optimization in POMDPs with infinite observation and state spaces.