Which Mutual-Information Representation Learning Objectives are Sufficient for Control?

Abstract

Mutual information maximization provides an appealing formalism for learning representations of data. In the context of reinforcement learning (RL), such representations can accelerate learning by discarding irrelevant and redundant information, while retaining the information necessary for control. Much of the prior work on these methods has addressed the practical difficulties of estimating mutual information from samples of high-dimensional observations, while comparatively less is understood about which mutual information objectives yield representations that are sufficient for RL from a theoretical perspective. In this paper, we formalize the sufficiency of a state representation for learning and representing the optimal policy, and study several popular mutual-information based objectives through this lens. Surprisingly, we find that two of these objectives can yield insufficient representations given mild and common assumptions on the structure of the MDP. We corroborate our theoretical results with empirical experiments on a simulated game environment with visual observations.

INFOrmation Prioritization through EmPOWERment in Visual Model-Based RL

“…Since this result holds for all Q-functions, it also holds for the optimal Q * and Q * . This result extends sufficiency results in prior works Rakelly et al (2021) to a setting with stochastic encoders and KL-divergence losses on forward dynamics and reward, as opposed to Wasserstein metrics (Gelada et al, 2019), and bisimulation metrics (Zhang et al, 2020).…”