Improving Value Estimation Critically Enhances Vanilla Policy Gradient

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wang, Tao, Zhang, Ruipeng, Gao, Sicun
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910967572463616
author Wang, Tao
Zhang, Ruipeng
Gao, Sicun
author_facet Wang, Tao
Zhang, Ruipeng
Gao, Sicun
contents Modern policy gradient algorithms, such as TRPO and PPO, outperform vanilla policy gradient in many RL tasks. Questioning the common belief that enforcing approximate trust regions leads to steady policy improvement in practice, we show that the more critical factor is the enhanced value estimation accuracy from more value update steps in each iteration. To demonstrate, we show that by simply increasing the number of value update steps per iteration, vanilla policy gradient itself can achieve performance comparable to or better than PPO in all the standard continuous control benchmark environments. Importantly, this simple change to vanilla policy gradient is significantly more robust to hyperparameter choices, opening up the possibility that RL algorithms may still become more effective and easier to use.
format Preprint
id arxiv_https___arxiv_org_abs_2505_19247
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Improving Value Estimation Critically Enhances Vanilla Policy Gradient
Wang, Tao
Zhang, Ruipeng
Gao, Sicun
Machine Learning
Artificial Intelligence
Robotics
I.2.6
Modern policy gradient algorithms, such as TRPO and PPO, outperform vanilla policy gradient in many RL tasks. Questioning the common belief that enforcing approximate trust regions leads to steady policy improvement in practice, we show that the more critical factor is the enhanced value estimation accuracy from more value update steps in each iteration. To demonstrate, we show that by simply increasing the number of value update steps per iteration, vanilla policy gradient itself can achieve performance comparable to or better than PPO in all the standard continuous control benchmark environments. Importantly, this simple change to vanilla policy gradient is significantly more robust to hyperparameter choices, opening up the possibility that RL algorithms may still become more effective and easier to use.
title Improving Value Estimation Critically Enhances Vanilla Policy Gradient
topic Machine Learning
Artificial Intelligence
Robotics
I.2.6
url https://arxiv.org/abs/2505.19247