在GPU上求解大规模优化问题的反向策略的PSO算法

在GPU上求解大规模优化问题的反向策略的PSO算法 Opposition-Based Particle Swarm Optimization for Solving Large Scale Optimization Problems on Graphic Process Unit 汪靖 WANG Jing1 1 2 first-author 吴志健 WU Zhijian1 1 武汉大学软件工程国家重点实验室 武汉大学软件工程国家重点实验室 State Key Laboratory of Software Engineering,Wuhan University State Key Laboratory of Software Engineering,Wuhan University 江西财经大学软件与通信工程学院 江西财经大学软件与通信工程学院 School of Software and Communication Engineering,Jiangxi University of Finance and Economics School of Software and Communication Engineering,Jiangxi University of Finance and Economics 本文通过对传统粒子群算法（PSO）的分析,在GPU（Graphic Process Unit）上设计了基于一般反向学习策略的粒子群算法,并用于求解大规模优化问题.主要思想是通过一般反向学习策略转化当前解空间,提高算法找到最优解的几率,同时使用GPU大量线程并行来加速收敛速度.对比数值实验表明,对于求解大规模高维的优化问题,本文算法比其他智能算法具有更好的精度和更快的收敛速度. Through an analysis of the traditional particle swarm algorithm,this paper presents particle swarm algorithm based on the generalized opposition-based particle（GOBL） swarm algorithm on Graphic Processing Unit（GPU）,and applies it to solve large scale optimization problem.The generalized opposition learning strategies transforms the current solution space to provide more chances of finding better solutions,and GPU in parallel accelerates the convergence rate.Experiment shows that this algorithm has better accuracy and convergence speed than other algorithm for solving large-scale and high-dimensional problems. 粒子群算法 反向策略 图形处理器 高维优化问题 particle swarm optimization opposition-based learning Graphic Process Unit high dimensional optimization 10.14188/j.1671-8836.2011.02.002 TP301.6 国家自然科学基金项目（61070008）;江西省科技支撑项目（2009BHB16400）;山西省高校基金资助项目（200900294） 2011-02-01 2021-04-01 2 57