作为（离散）势理论的强化学习
Reinforcement Learning as (Discrete) Potential Theory
浏览论文内容
中文总结 AI 辅助
本文回顾强化学习与概率论、势理论的联系，在固定策略假设下从势理论视角研究强化学习的表示与算法，该视角或可提升样本效率并施加形式约束，且该框架可扩展至非线性情形。
中文摘要 AI 辅助
强化学习（RL）理论通过马尔可夫链从根本上依赖于概率论，而概率论与势理论之间存在深刻联系。本文回顾该联系，并在固定策略假设下，从势理论视角探讨核心强化学习表示与算法，该视角或能为提升样本效率及施加可应用于RL的形式约束提供路径；当放松固定策略假设时，线性势理论框架可自然扩展至非线性情形。
英文摘要
Reinforcement learning (RL) theory fundamentally depends on probability theory through the Markov chain. There is a deep connection between probability theory and potential theory. This paper reviews that connection and explores the potential-theoretic viewpoint for core reinforcement learning representations and algorithms under a fixed-policy assumption. This viewpoint may offer a path for improved sample efficiency and formal constraints that can be applied to RL. When the fixed-policy assumption is relaxed, the linear potential theory framework can be naturally extended to the nonlinear case.
发表机构
- SRI International（SRI国际）
- Computer Science Laboratory（计算机科学实验室）
机构由 AI 辅助整理，请以论文原文为准。
