PPO in the Fisher-Rao geometry

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Lascu, Razvan-Andrei, Šiška, David, Szpruch, Łukasz
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908799352176640
author Lascu, Razvan-Andrei
Šiška, David
Szpruch, Łukasz
author_facet Lascu, Razvan-Andrei
Šiška, David
Szpruch, Łukasz
contents Proximal Policy Optimization (PPO) is widely used in reinforcement learning due to its strong empirical performance, yet it lacks formal guarantees for policy improvement and convergence. PPO's clipped surrogate objective is motivated by a lower bound on linearization of the value function in flat geometry setting. We derive a tighter surrogate objective and introduce Fisher-Rao PPO (FR-PPO) by leveraging the Fisher-Rao (FR) geometry. Our scheme provides strong theoretical guarantees, including monotonic policy improvement. In the direct parametrization setting, we show that FR-PPO achieves sub-linear convergence with no dependence on action or state space dimensions, and for parametrized policies we further obtain sub-linear convergence up to the compatible function approximation error. Finally, although our primary focus is theoretical, we also demonstrate empirically that FR-PPO performs well across a range of standard reinforcement learning tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2506_03757
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle PPO in the Fisher-Rao geometry
Lascu, Razvan-Andrei
Šiška, David
Szpruch, Łukasz
Machine Learning
Optimization and Control
Proximal Policy Optimization (PPO) is widely used in reinforcement learning due to its strong empirical performance, yet it lacks formal guarantees for policy improvement and convergence. PPO's clipped surrogate objective is motivated by a lower bound on linearization of the value function in flat geometry setting. We derive a tighter surrogate objective and introduce Fisher-Rao PPO (FR-PPO) by leveraging the Fisher-Rao (FR) geometry. Our scheme provides strong theoretical guarantees, including monotonic policy improvement. In the direct parametrization setting, we show that FR-PPO achieves sub-linear convergence with no dependence on action or state space dimensions, and for parametrized policies we further obtain sub-linear convergence up to the compatible function approximation error. Finally, although our primary focus is theoretical, we also demonstrate empirically that FR-PPO performs well across a range of standard reinforcement learning tasks.
title PPO in the Fisher-Rao geometry
topic Machine Learning
Optimization and Control
url https://arxiv.org/abs/2506.03757