混合邻域结构的粒子群算法  

A PSO algorithm based on mixture neighborhood structure

作　　者：张泽星[1] 

机构地区：[1]石家庄经济学院网络信息安全实验室,河北石家庄050031

出　　处：《河北工业科技》2015年第3期219-223,共5页Hebei Journal of Industrial Science and Technology

基　　金：河北省科技厅计划项目(13210702D)

摘　　要：为了克服标准粒子群算法的早熟、停滞进化或易于陷入局部最优的现象,提出了一种混合模型(简称NSPO)。NSPO将一个粒子映射到无标度网络的多个网络节点上,借助网络结构获得该粒子的邻域拓扑。对粒子的更新,NSPO既考虑种群的最优,又考虑邻域的最优。在3个具有不同难度特点的测试函数上,将NSPO与标准粒子群算法进行了比较。实验结果表明:对于全局最优和梯度信息明显的函数,NSPO具有非常优越的表现;对于具有诸多局部最优的函数,NSPO逃逸局部最优的能力要强于标准粒子群算法;对于具有误导性梯度信息的函数,NSPO偶尔表现优异。In order to overcome the phenomena of precocious,stagnant or falling into local optimal of the standard particle swarm algorithm(PSO),a mixture model(NPSO)is given.NPSO maps a particle to more than one network node of a scalefree network.Thus,the particle's neighborhood topology is obtained by the network's structure.The update method of the speed of the particle considers not only the optimal among the population but also the optimal neighborhood.On three test functions with different characteristics,a series of experiments are conducted to compare the proposed algorithm with the standard PSO.The experimental results show that:for the function with obvious global optimum and gradient information,NPSO has excellent performance;for functions with many local optima,this new algorithm to escape from local optimum is stronger than the standard PSO;and for functions with misleading gradient information,the proposed new model has excellent performance occasionally.

关 键 词：算法理论 粒子群 演化计算 复杂网络 邻域结构 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]