基于带控制器并行结构模型的并行微粒群算法  

Parallel Particle Swarm Optimization Based on Parallel Model with Controller

作　　者：王元元[1] 曾建潮[1] 谭瑛[1] 

机构地区：[1]太原科技大学系统仿真与计算机应用研究所,山西太原030024

出　　处：《系统仿真学报》2007年第10期2171-2176,共6页Journal of System Simulation

基　　金：教育部重点科研项目资助(204018)

摘　　要：并行计算是解决复杂大规模工程计算问题的有效方法。现提出了一种基于带控制器并行结构模型的并行微粒群算法,它是一种粗粒度的并行。它将种群分为几个子种群,分别放在不同的处理器上,每个子种群独立、同时进化,周期性地交换、更新最优信息。实验结果表明:若选择合适的通讯周期时,该并行微粒群算法不仅具有理想的加速比,而且有效地提高解的质量。Parallel computation is an effective method for solving complex large-scale engineering computation problems.A parallel PSO was studied based on the parallel model with controller that is coarse-grain parallel. It divides the whole group into several sub-groups, Every subgroup evolves in different processors independently and synchronously, exchanges and updates the best information periodically, The experiment results show that if the period of communication is selected appropriate, this parallel PSO not only has perfect speedup, but also improves the quality of result.

关 键 词：并行计算 微粒群算法 并行结构模型 加速比 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]