Functional Acceleration for Policy Mirror Descent

Veronica Chelu, Doina Precup

Keywords: functional acceleration, Policy Mirror Descent (PMD), momentum, extrapolation

TL;DR: We apply functional acceleration to Policy Mirror Descent (PMD) leading to a momentum-based PMD update with learning speed-ups on ill-conditioned policy optimization landscapes.

Abstract: We apply functional acceleration to the Policy Mirror Descent (PMD) general family of algorithms, which cover a wide range of novel and fundamental methods in Reinforcement Learning (RL). Leveraging duality, we propose a momentum-based PMD update. By taking the functional route, our approach is independent of the policy parametrization and applicable to large-scale optimization, covering previous applications of momentum at the level of policy parameters as a special case. We theoretically analyze several properties of this approach and complement with a numerical ablation study, which serves to illustrate the policy optimization dynamics on the value polytope, relative to different algorithmic design choices in this space. We further characterize numerically several features of the problem setting relevant for functional acceleration, and lastly, we investigate the impact of approximation on their learning mechanics.

Submission Number: 17