基于小世界模型动态演化邻域的微粒群算法

Particle Swarm Optimization with Dynamic Evolutionary Neighbourhood of Small-world Model

分析了影响网络中信息传播的主要因素,并结合小世界网络的形成机制,提出了一种具有动态邻域结构的微粒群算法。该算法初始化群体拓扑结构为"聚集系数大,平均最短路径长"的环形规则网络,以降低邻域间信息交流的速度,保持种群的多样性。在算法进化过程中,当邻域多样性小于给定阈值时,以小概率向网络随机增加长距离边,逐步形成"聚集系数大,平均最短路径小"的小世界网络,加快邻域间信息交流的速度。仿真结果表明,结合适当的惯性策略,该算法能获得更好的收敛性能和收敛速度。

The major factors influencing information transmission in networks were analyzed.Combined with small-world network formation mechanism, a novel particle swarm optimization with dynamic neighbourhood structure was proposed. In this algorithm, the population topology is initialized as regular ring lattice of ＂high clustering coefficient, long average path length＂, to slow down the information exchange between different neighbourhoods and maintain the diversity of the population. In evolution process, while neighbourhood diversity is smaller than the threshold value, long-distance edges are added into networks with a small probability, gradually small-world network of ＂high clustering coefficient, short average path length＂ is formed, and the information exchange between neighborhoods is speeded up. Experimental simulations show that with appropriate inertial strategy, the proposed method can obtain better convergence performance and convergence rate.