改善微粒群算法多样性的多子群动态聚合

针对微粒群算法的群体多样性问题，将无标度网络形成过程中的择优连接机制引入微粒群算法的群体构造过程中，提出了一种多子群动态聚合的微粒群算法．依据算法搜索的状态，当某个子群面临搜索困境时，择优选择优秀子群建立连接，随着搜索的不断推进，微粒群体由多个独立的子群逐步聚合为一个类无标度群体网络．通过仿真实验发现，该算法通过调整微粒的邻域规模和连接方式平衡了算法的勘探和开发的能力，获得了较好的收敛性能．

Multi-swarm Dynamic Polymerization to Improve the Diversity of Particle Swarm Optimization

To improve the diversity of the swarm for particle swarm optimization, preferential attachment mechanism in the process of scale-free network formation is introduced into particle swarm optimization, and a particle swarm optimization with dynamic polymerized multi-swarms is proposed. Based on the searching state, a subpopulation will connect to the more excellent one when facing search predicament. With the continuous advance of search, population structure is aggregated into a scale-free like population network from multiple independent subpopulation. Good balance between exploration and development of the PSO-DPM is maintained by adjusting the particle size of the neighborhood and connection mode. Experimental results reveal that the new algorithm obtained a better convergence performance.