一种具有动态拓扑结构的粒子群算法研究  

Research on a Novel Particle Swarm Algorithm with Dynamic Topology

作　　者：王雪飞[1] 王芳[1] 邱玉辉[1] 

机构地区：[1]西南大学,重庆400715

出　　处：《计算机科学》2007年第3期205-207,233,共4页Computer Science

摘　　要：受小世界网络模型的启发,提出了一种具有动态拓扑结构的新颖粒子群算法。该算法通过对每个粒子邻域的记忆和更新,模拟小世界网络模型中的信息传播方式。在大量基准问题上的实验结果显示,提出的算法能有效保持优秀粒子与非优粒子所占比例的均衡性,维持了种群的多样性,避免了经典粒子群算法在高维、多峰问题上的早熟收敛现象。Inspired by the model of small-world networks, a novel particle swarm optimizer （PSO） with dynamic topology is proposed, which mimics the information transferring style of small-world networks by remembering and updating individual particle＇s neighborhood during the process is running. Experimental results on variety benchmark problems demonstrate that the proposed algorithm effectively balances the proportion of optimal and non-optimal particles, thus can obtain better population diversity and restrain the premature convergence encountered by the standard PSO algorithm when solving high-dimensional multimodal problems.

关 键 词：粒子群 小世界网络 拓扑结构 多峰函数 

分 类 号：TP393.1[自动化与计算机技术—计算机应用技术]