改进的K-均值聚类排挤小生境遗传算法  

The Improved Crowding Niching Genetic Algorithm

作　　者：杨新春[1,2] 许峰[1] 

机构地区：[1]安徽理工大学理学院,安徽淮南232001 [2]安徽省产品质量监督检验研究院,安徽合肥230051

出　　处：《软件导刊》2009年第6期54-57,共4页Software Guide

基　　金：安徽省教育厅自然科学基金项目(KJ2007B216)

摘　　要：针对基于罚函数的排挤小生境遗传算法的缺陷,提出了改进的K-均值聚类排挤小生境遗传算法。该算法将改进的K-均值聚类分析与排挤技术有机地结合起来,无需事先确定小生境的数目和半径,可以通过调节最小聚类距离控制收敛到的小生境的数目,避免找到无效的极值点。理论分析和数值实验表明,该算法对于非均匀分布的多峰函数具有较好的搜索性能。The paper presents a crowding niching genetic algorithm which based on improved K-means clustering for overcoming the defects of crowding niching genetic algorithm based on penalty function. The algorithm organically combines the clustering and technique of crowding, which not only can search validly more extreme points within space of muhimodal function but also can control the number of converged niche through adjusting the minimal distance of clustering to avoid find the invalid extreme points. The improved K-means clustering crowding niching genetic algorithm takes no need to know the concrete number of niche and the size of the niche radium in advance and fit in optimization of various problems. Theoretical analysis and numeral experiment indicate that it has good searching function on multiple hump function which distributes non-uniform.

关 键 词：遗传算法 多峰函数优化 排挤小生境 聚类分析 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]