基于混沌变量的模式搜索法及其应用  

Pattern Search Algorithm Using Chaos and Its Application

作　　者：袁小芳[1] 王耀南[1] 吴亮红[1] 

机构地区：[1]湖南大学电气与信息工程学院,湖南长沙410082

出　　处：《湖南大学学报（自然科学版）》2007年第9期30-33,共4页Journal of Hunan University:Natural Sciences

基　　金：国家自然科学基金资助项目(60375001);高校博士点基金资助项目(20030532004)

摘　　要：提出了基于混沌变量的模式搜索法.在探索移动中,由混沌遍历性来生成移动方向和步长;在模式移动中,针对粗搜索与细搜索分别采取最优一维搜索和直接模式搜索.将全局优化能力强的混沌优化融入到模式搜索法中,且采取二级模式移动,搜索效率高.算法应用于模糊神经网络权值优化,仿真验证了其优良性能.A new pattern search algorithm using chaos was proposed, in which the ergodicity of chaos was applied to produce stochastic move direction and length in exploratory move. Moreover, two-staged pattern move was employed for different search phases. The global search property of chaotic optimization was utilized in the proposed algorithmand it could realize global optimum, and two-staged pattern move had fast convergence velocity. The training of fuzzy neural networks was implemented by using the proposed optimization algorithm, and simulation results have demonstrated that it has efficient search ability.

关 键 词：神经网络 混沌 模式搜索法 最优化 全局优化 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]