A Framework for Policy Evaluation Enhancement by Diffusion Models

Tao Ma, Xuzhi Yang

Keywords: Reinforcement Learning, Policy Evaluation, Generative Models, Diffusion Models

TL;DR: We propose a framework to generate synthetic data in reinforcement learning to enhance policy evaluation.

Abstract: Reinforcement learning plays an important role in various fields, and has fast development in policy evaluation and learning methods, enjoying the advantages of large data size. However, when data are limited, directly applying evaluation methods does not necessarily result in a good policy evaluation. In this work, we provide a framework to generate synthetic data with diffusion models, to enhance data-efficient policy evaluation, which is supported by experiments.

Submission Number: 121