arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.06567math.OC
收藏

第三类马尔可夫决策过程:通过策略梯度下降学习分布

Markov Decision Processes of the Third Kind: Learning Distributions by Policy Gradient Descent

Karlsruhe Institute of Technology (KIT)(卡尔斯鲁厄理工学院)

Nicole Bäuerle, Athanasios Vasileiadis

AI总结:

本文提出了一种基于神经网络的策略梯度算法,用于学习分布性马尔可夫决策过程中的策略,以使累积奖励分布接近目标分布。

摘要:

本文的目标是分析分布性马尔可夫决策过程,作为一类控制问题,其目标是学习策略,使累积奖励的分布向指定的目标分布靠近,而不是优化期望值或风险功能。为在无模型设置下解决由此产生的分布性控制问题,我们提出了一种基于随机马尔可夫策略的神经网络参数化策略梯度算法,该策略定义在扩展的状态空间上,并基于样本评估的特征函数损失。在温和的正则性和增长假设下,我们利用随机逼近技术证明了该算法收敛到平稳点。若干数值实验展示了该方法匹配复杂目标分布、在存在经典最优策略时恢复其以及揭示分布性控制特有的内在非唯一性现象的能力。

英文摘要:

The goal of this paper is to analyze distributional Markov Decision Processes as a class of control problems in which the objective is to learn policies that steer the distribution of a cumulative reward toward a prescribed target law, rather than optimizing an expected value or a risk functional. To solve the resulting distributional control problem in a model-free setting, we propose a policy-gradient algorithm based on neural-network parameterizations of randomized Markov policies, defined on an augmented state space and a sample-based evaluation of the characteristic-function loss. Under mild regularity and growth assumptions, we prove convergence of the algorithm to stationary points using stochastic approximation techniques. Several numerical experiments illustrate the ability of the method to match complex target distributions, recover classical optimal policies when they exist, and reveal intrinsic non-uniqueness phenomena specific to distributional control.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑