计算机工程与应用 ›› 2011, Vol. 47 ›› Issue (1): 46-49.DOI: 10.3778/j.issn.1002-8331.2011.01.013

• 研究、探讨 • 上一篇    下一篇

改进的粒子群算法及收敛性分析

谢铮桂1,2,钟少丹1,2,韦玉科2   

  1. 1.韩山师范学院 数学与信息技术系,广东 潮州 521041
    2.广东工业大学 计算机学院,广州 510090

  • 收稿日期:2010-08-17 修回日期:2010-10-18 出版日期:2011-01-01 发布日期:2011-01-01
  • 通讯作者: 谢铮桂

Modified particle swarm optimization algorithm and its convergence analysis

XIE Zhenggui1,2,ZHONG Shaodan1,2,WEI Yuke2   

  1. 1.Mathematics and Information Technology Department,Hanshan Normal University,Chaozhou,Guangdong 521041,China
    2.Faculty of Computer,Guangdong University of Technology,Guangzhou 510090,China
  • Received:2010-08-17 Revised:2010-10-18 Online:2011-01-01 Published:2011-01-01
  • Contact: XIE Zhenggui

摘要: 针对PSO算法对多峰值函数搜索易陷入局部极值点的缺点,提出一种改进的粒子群(MPSO)算法。MPSO算法采用逃逸策略和免疫学习策略来保证种群多样性,使算法能有效进行全局搜索。并讨论MPSO算法的收敛性,证明其能以概率1全局收敛。最后用3个常用的测试函数进行仿真,实验结果表明MPSO算法比PSO算法有更好的收敛性和更快的收敛速度。

关键词: 粒子群算法, 逃逸, 免疫学习, 全局优化, 收敛性

Abstract: This paper proposes a modified particle swarm algorithm in allusion to the defect that the PSO algorithm easily plunges into the local optimization for multi-peak function optimization problem.MPSO adopts escape strategy and immune learning strategy to guarantee the particles diversity and to make particles explore the global optimization more efficiently.The convergence of MPSO algorithm is discussed and is proved to converge to the global optimization with probability one.Finally,three familiar test functions are simulated to show that MPSO achieves better and faster convergence.

Key words: particle swarm algorithm, escape, immune learning, global optimization, convergence

中图分类号: