计算机工程与应用 ›› 2013, Vol. 49 ›› Issue (6): 1-7.

• 博士论坛 • 上一篇    下一篇

自适应信息选择PSO算法及其特性分析

吴  涛1,严余松1,2,陈  曦1   

  1. 1.西南交通大学 信息科学与技术学院,成都 610031
    2.四川师范大学,成都 610068
  • 出版日期:2013-03-15 发布日期:2013-03-14

Adaptive partly informed PSO algorithm and its characteristics analysis

WU Tao1, YAN Yusong1,2, CHEN Xi1   

  1. 1.School of Information Science & Technology, Southwest Jiaotong University, Chengdu 610031, China
    2.Sichuan Normal University, Chengdu 610068, China
  • Online:2013-03-15 Published:2013-03-14

摘要: 针对粒子群优化过程中容易出现早熟收敛或停滞的问题,在全信息粒子群优化(FIPSO)算法的基础上结合社会心理学原理提出了一种新的粒子群优化算法——自适应信息选择粒子群优化算法(API-PSO)。在API-PSO算法中,粒子根据其邻域粒子不同表现,自适应地选择群体共享经验。实验表明,新的优化算法具有较好的收敛精度和收敛速度。分别对API-PSO算法的种群多样性和收敛性进行了数学分析,分析结果为合理选择算法参数,解决算法种群多样性匮乏,促进种群进化发展,改善算法性能提供了理论依据。

关键词: 群体智能, 粒子群优化, 早熟收敛, 种群多样性

Abstract: To avoid the premature convergence or stagnation during particle swarm optimization process, on the basis of Fully-Informed Particle Swarm Optimization(FIPSO) in conjunction with social psychological principles, the novel Adaptive Partly Informed Particle Swarm Optimization(API-PSO) algorithm is proposed. In API-PSO algorithm, a particle adaptively selects swarm-shared experience according to performances of its neighboring particles. Experiment results show that API-PSO offers satisfactory convergence accuracy and speed. Probabilistic analysis on swarm diversity and convergence analysis of API-PSO is conducted which serves as the theoretic basis to select the algorithm parameters, solve swarm diversity lack, facilitate swarm evolution development and improve algorithm performance.

Key words: swarm intelligence, Particle Swarm Optimization(PSO), premature convergence, population diversity