具有动态拓扑结构的聚类粒子群算法研究  

Improved Clustering PSO with Dynamic Topology Structure

作　　者：梁晓磊[1] 李文锋[1] 张煜[1] 李斌[2] 

机构地区：[1]武汉理工大学物流工程学院,湖北武汉430063 [2]福建工程学院经济管理系,福建福州350108

出　　处：《武汉理工大学学报（信息与管理工程版）》2011年第1期22-26,共5页Journal of Wuhan University of Technology：Information & Management Engineering

基　　金：国家十一五科技支撑计划基金资助项目(2006BAH02A06);国家自然科学基金资助项目(60475031);湖北省自然科学基金资助项目(2009CDB108);福建省教育厅A类科技基金资助项目(JA10214)

摘　　要：从增加种群多样性和信息交流能力出发,结合K-means聚类算法和Ring型拓扑结构的特点,提出了一种动态拓扑结构的改进算法(KPSO)。在粒子信息交流中,提出两种位置和速度更新方式,并通过Bench-mark函数优化问题测试,比较了KPSO算法与经典PSO的各种性能。对KPSO的重要参数聚类数K、种群规模N的选择进行了组合实验测试。结果表明,改进后的KPSO求解复杂优化问题性能良好,在搜索精度和速度上优势明显,对于不同的种群规模选择不同的聚类数对算法有重要影响。In order to improve the diversity of the swarm and ability to communication of particles,an improved particle swarm optimization with dynamic topology structure was proposed by using K-means clustering method and Ring structure,called KPSO.Two kinds of patterns of updating position and speed of every particle were given.Through the tests of Benchmark Functions,the performance of KSPO was compared with that of PSO.Then an experiment was carried out for choosing the parameters of KPSO：clustering number K and population size N.The results of experiments show that KPSO does well in complex optimization problem.It also can be seen from the experiments that different combinations of K and N have important effect on KPSO.

关 键 词：粒子群算法 动态拓扑结构 K-MEANS聚类 函数优化 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]