Maximum Entropy Reinforcement Learning with Mixture Policies

TL;DR

This paper introduces a low-variance entropy estimator for mixture policies in MaxEnt reinforcement learning, enabling effective use of mixture models with Soft Actor-Critic in continuous control tasks.

Contribution

It develops a tractable entropy estimator for mixture policies and integrates it into a SAC variant, advancing MaxEnt RL with expressive policy classes.

Findings

The proposed estimator closely approximates true mixture entropy.

The SAC variant with mixture policies performs well on continuous control tasks.

Mixture policies enhance policy expressiveness in MaxEnt RL.

Abstract

Mixture models are an expressive hypothesis class that can approximate a rich set of policies. However, using mixture policies in the Maximum Entropy (MaxEnt) framework is not straightforward. The entropy of a mixture model is not equal to the sum of its components, nor does it have a closed-form expression in most cases. Using such policies in MaxEnt algorithms, therefore, requires constructing a tractable approximation of the mixture entropy. In this paper, we derive a simple, low-variance mixture-entropy estimator. We show that it is closely related to the sum of marginal entropies. Equipped with our entropy estimator, we derive an algorithmic variant of Soft Actor-Critic (SAC) to the mixture policy case and evaluate it on a series of continuous control tasks.