作　　者： ; ; ;

机构地区： 武汉理工大学计算机科学与技术学院

出　　处： 《计算机工程与应用》 2005年第23期80-83,共4页

摘　　要： 针对原始PSO算法存在的各种缺点,论文提出了一种新型的PSO算法。该算法在搜索的中期进行变异,并依据搜索开始后的收敛情况来决定首次变异的时机、依据每次变异后的收敛情况来决定下次变异的时机,以便最大限度地增加变异次数,并充分利用每次变异带来的避免局部最优、求得全局最优的机会,从而提高搜索成功的概率;在后期则在一定的循环次数内不变异,以确保搜索精度。每次变异只选择那些收敛程度相对比较大的维来进行,这样可以在不破坏粒子群的结构的情况下扩大搜索空间、提高收敛速度。通过对四个多峰的测试函数和一个问题空间为非凸集的实例所做的对比实验,表明改进的PSO算法增强了全局搜索能力,搜索成功率大为提高,克服了原始的PSO算法易于收敛到局部最优点的缺点。 This paper proposes a new PSO which aims at the existing disadvantages of the original PSO.The algorithm processes mutation in the middle stage of searching,determines the time of the first mutation according to the situation of convergence with the beginning of searching,and then determines the frequency of mutation according to each situation of convergence after mutation in order to increase the number of mutation to the highest limit and completely utilizes the chance of the global optimization resulted from each mutation,avoiding local optimization,to enhance the probability of the success of searching;whereas mutation is not processed within certain cycle indices in the late stage to ensure the accuracy of searching.Those dimensions with relatively larger convergence are chosen during each mutation,which can extend the space of searching,increase the velocity of convergence while not destroying the structure of the particle swarm.The control experiment of one testing function for four multiple-modal and the example of the problem space which is not convex set indicates that the improved PSO greatly enhances the ability of global searching and the probability of the success of searching and gets over the original PSO＇s disadvantage of easily converging to local optimal point。

关 键 词： 粒子群优化算法 变异 布局优化 全局最优

领　　域： [自动化与计算机技术] [自动化与计算机技术]