基于高聚集性无标度网络模型的微粒群算法

Particle Swarm Optimization with Highly-Clustered Scale-Free Network Model

受无标度网络结构特性的启发,将BA模型的"择优连接"机制进行扩展,引入微粒群群体组织方式的构造过程,提出基于高聚集性的无标度网络模型的微粒群算法。算法初期微粒被随机分布在环形结构中,随着搜索的进行不断增加新的微粒,并依据节点度和节点间的距离增加新的连接,最终形成具有高聚集性的无标度网络模型。这样,群体中多数微粒进行局部范围的搜索,而少量微粒按照全局模式搜索,两种方式相互制衡。仿真实验表明,改进后的算法能获得更好的收敛精度和进化速度。

Enlightened by the properties of scale-free network model,＂preferential attachment＂ mechanism of the BA model is extended and introduced into particle swarm optimization,and a novel particle swarm optimization with highly-clustered scale-free network model（PSO-HCSF） is proposed.At the early stage of the algorithm,particles were randomly distributed in a ring,new particles are continuously added into the population with searching,and based on the node degree and the distance between nodes new connections are produced,and a high aggregation degree of scale-free network model are formed in the end.In this way,the majority of particles search in local scope and a small amount of particles search with the overall pattern,two ways check and balance.Experimental simulations show that the new method obtains better evolution speed and convergence performance.