基于博弈策略强化学习的函数优化算法.pdfVIP

基于博弈策略强化学习的函数优化算法 叶 俊 1 刘贤德 1 韩 露 2 1!华中科技大学光电系武汉 430074# 2!华中科技大学电信系武汉 430074$ E-maii tianyuan@ 摘 要 该文提出了一种基于博弈论的函数优化算法% 算法将优化问题的搜索空间映射为博弈的策略组合空间优化目 标函数映射为博弈的效用函数通过博弈策略的强化学习过程智能地求解函数优化问题% 文章给出了算法的形式定义及 描述然后在一组标准的函数优化测试集上进行了仿真运算验证了算法的有效性% 关键词 博弈 函数优化 强化学习 策略组合 效用函数 文章编号 1002-8331-!2004$17-0067-02 文献标识码 A 中图分类号 TP18 A Function Optimization Algorithm Based on Reinforcement Learning of Game Strategies Ye Jun1 Liu Xiande1 Han Lu2 1(Optoeiectronics Engineering Dept.,HUST,Wuhan 430074 2(Eiectronics Information Engineering Dept.,HUST,Wuhan 430074 Abstract In this paper,we propose a novei optimization probiem based on Game Theory.It maps the search space and objective function of optimization probiem to the strategy profiie space and utiiity function of non -cooperative game respectiveiy,and soives the probiem through the dynamic evoiution process with reinforcement iearning on corresponding game strategies inteiiigentiy.In the articie,we present the formai definition and detaiied description of the aigorithm,and test it on a set of standard function optimization suite.The efficiency of the proposed aigorithm is proved by the performance comparison with genetic aigorithm. Keywords game,function optimization,reinforcement iearning,strategy profiie,utiiity function 1 引言 标准的函数优化测试集上进行了仿真计算,然后与并行遗传算