多核环境下并行粒子群算法parallel particle swarm optimization .pdfVIP

多核环境下并行粒子群算法 1* 1 1 2 何莉 ，刘晓东 ，李松阳 ，张倩 (1.河南工程学院 计算机学院, 河南 郑州 451191;2.河南工程学院 材料与化学工程学院, 河南 郑州 451191) （*通信作者电子邮箱engineerheli@126.com） 摘 要：针对串行粒子群算法在解决大任务耗时过长的问题，提出一种共享并行粒子群算法（Shared-PSO）。充分利用 多核处理能力缩短问题处理运行时间，设置共享区和采取粒子随机替换策略有效促进粒子信息的交流，其算法流程具有较好 的通用性，允许利用多种串行粒子群算法完成粒子信息更新工作。在标准优化测试集CEC2014上的实验结果显示新算法的 执行时间是串行算法的四分之一。新算法能够有效地改善串行粒子群的执行效率，扩展粒子群算法的应用范围。 关键词:并行粒子群算法；多核计算；并行算法；最优化；群智能 中图分类号: TP301.6 文献标志码:A ParallelParticleSwarmOptimizationAlgorithminMulticore computingenvironment 1 1 1 2 HELi ,LIUXiaodong ,LiSongyang,ZhangQian (1. SchoolofComputer,HenanInstituteofEngineering,ZhengzhouHenan451191,China) (2.SchoolofMaterialsandchemicalengineering,HenanInstituteofEngineering,ZhengzhouHenan451191,China) Abstract: Aiming at the problem that serial particle swarm optimization (PSO) algorithms are time-consuming to dealwith big tasks, anovel sharedparallel PSO algorithm (Shared-PSO)was proposed. The multi-core processing power was made use of to reduce to resolution. In order to facilitate communication of particles, a shared area was set up and a random strategy was applied to switch particles. Several serial PSO algorithms could be permitted to update particle information because of the universality of its algorithm flow. Shared-PSOwasappliedonthestandardoptimizationtestsetCEC(CongressonEvolutionaryComputation)2014. TheexperimentresultsshowthattheexecutiontimeofShared-PSOisaquarteroftheserialPSO’s.Theproposed algorithmcaneffectivelyimprovetheexecutionefficiencyofserialPSOs,andexpandappliedrangeofPSOs. Key words: Parallel particle swarm optimization; Multicore computing; Parallel algorithm; Optimization; Swarmintelligence 0 引言 粒子群算法自 1995年由文献[1]提出后