静态与动态SF拓扑邻域对PSO改进算法的分析

Improved Particle Swarm Optimization Algorithm Study Based on static and Dynamic Scale-Free Network Neighborhood Topology

本文探讨以静态无标度网为拓扑邻域的PSO,分析在不同的网络平均度条件下PSO的寻优效果.结果表明在平均度较大的范围内,粒子寻优基于静态无标度网络表现效果较好.提出以无标度网络为粒子群的初始邻域,在寻优过程中网络呈有向动态变化的PSO改进算法,并以三个测试函数为例,分析了算法的有效性.本文还分析了基于有向动态网络的改进算法在不同网络拓扑即平均度条件下粒子群的寻优效果.结果表明粒子寻优效果受网络的平均度尤其是出度的影响,然而度值越大或者越小并不一定使寻优效果越好,如对于Rosenbrock函数保持较小的平均度会使粒群寻优效果更好,而对Rastrigrin函数的测试显示平均度对粒群寻优结果的影响差别不大.

The paper first discusses the PSO improved algorithm based on a static scale-free network topology neighborhood and analyzes the results in different condition of the network＇s average degree.The results show that the particle optimization performs better when the average degree varies in a wide range.Secondly we briefly introduce the improved algorithm which based on dynamic and directed scale-free network.Taking three different benchmark functions as testing functions,we summary the influence of scale-free network with different average degree.Finally,we analyze the optimization of the particles with different average degree topology structures based on the improved dynamic and directed scale-free network.It shows that the optimization effects is influenced significantly by the average degree especially by the out degree factor,however,it is not necessary better optimization will be made under the higher or lower degree,for example the particles optimization will be better for Rosenbrock function to maintain a smaller average degree,while the impact shows not very different when Rastrigrin function tested with groups of different average degree.