An Empirical Analysis of Proximal Policy Optimization with Kronecker-factored Natural Gradients

TL;DR

This paper empirically evaluates PPO combined with K-FAC natural gradient optimization, showing improvements in sample efficiency and training speed over PPO in MuJoCo environments, but with some limitations in scalability and epoch sensitivity.

Contribution

It introduces PPOKFAC, a novel combination of PPO and K-FAC natural gradients, and provides comprehensive empirical analysis of its performance and limitations.

Findings

PPOKFAC outperforms PPO in sample complexity and speed.

PPOKFAC is scalable with batch size.

Adding more epochs does not improve sample efficiency.

Abstract

In this technical report, we consider an approach that combines the PPO objective and K-FAC natural gradient optimization, for which we call PPOKFAC. We perform a range of empirical analysis on various aspects of the algorithm, such as sample complexity, training speed, and sensitivity to batch size and training epochs. We observe that PPOKFAC is able to outperform PPO in terms of sample complexity and speed in a range of MuJoCo environments, while being scalable in terms of batch size. In spite of this, it seems that adding more epochs is not necessarily helpful for sample efficiency, and PPOKFAC seems to be worse than its A2C counterpart, ACKTR.