一种基于网格划分的自适应粒子群优化算法  

Grid-based adaptive particle swarm optimization

作　　者：方冰[1] 李太勇[2] 吴江[2] 

机构地区：[1]贵阳学院计算机科学系,贵阳550005 [2]西南财经大学经济信息工程学院,成都610074

出　　处：《计算机应用研究》2010年第11期4136-4139,共4页Application Research of Computers

基　　金：西南财经大学金融智能与金融工程重点实验室校内公开项目(FIFE2010-P02)

摘　　要：为了提升粒子群优化算法(particle swarm optimization,PSO)的性能,提出了一种基于网格划分的自适应粒子群优化(grid-based adaptive particle swarm optimization,GAPSO)算法和基于网格划分的多样性计算方法,并设计了初始种群多样性最大化(maximal diversity algorithm for the initial swarm,MDAIS)算法。GAPSO分为两个阶段:根据粒子对多样性的贡献自适应调整粒子的进化方向,使种群维持较高的多样性;然后根据粒子对多样性的贡献自适应调整粒子的惯性权值。通过基准测试函数对算法进行了实验,实验表明,对于连续函数优化问题,提出的GAPSO算法优于经典PSO算法,GAPSO收敛到最优解的迭代次数比PSO平均减少了约60%。To improve the performance of PSO,this paper proposed aGAPSO,and developed a measurement method for warm diversity and a maximal diversity algorithm for the initial swarm （MDAIS）.Divided the GAPSO into two phases.In the first phase,tuned the evolving direction of a particle adaptively according to the contribution to the diversity.In the second phase,tuned the inertia weight of a particle adaptively according to the contribution to the diversity.Used several classic benchmark functions to evaluate the GAPSO.The experimental results show that for continuous optimization problems,the GAPSO outperforms the classic PSO.The iteration times for finding the best solutions in the GAPSO decrease about 60% averagely while compared with that in the classic PSO.

关 键 词：粒子群 网格 多样性 优化算法 惯性权值 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]