一种综合学习粒子群优化算法  

An integrated learning particle swarm optimizer

作　　者：徐刚[1] 杨玉群[2] 刘昊[3] 刘斌斌[1] 宋军[1] 

机构地区：[1]南昌大学数学系,江西南昌330031 [2]南昌大学附属中学,江西南昌330047 [3]辽宁科技大学理学院,辽宁鞍山114051

出　　处：《南昌大学学报（理科版）》2013年第5期428-432,共5页Journal of Nanchang University(Natural Science)

基　　金：国家自然科学基金资助项目(61175127;11101204);江西省教育厅基金资助项目(GJJ12093)

摘　　要：针对复杂多峰函数优化,提出了一种综合学习粒子群优化算法(IELPSO)。该算法把基于超球坐标系的粒子更新和辨识、加速质量差的粒子两个策略引入基于例子学习粒子群优化算法(ELPSO)。本算法利用超球坐标操作改变粒子大小和方向,因而粒子在搜索过程中能覆盖局部极小,同时能发现最差粒子并且加速它们靠拢最优解。提出的算法与其他已有算法进行了比较,对几种典型函数的测试结果表明,IELPSO算法提高了收敛速度和精度,全局搜索能力有了显著提高。It proposed a novel comprehensive learning particle swarm optimizer（IELPSO）in this paper,to optimize complex multimodal functions.This novel study introduced the particle＇s updating strategy based on hyperspherical coordinate system.It also revealed how to find the diverged particles and accelerate them towards optimal solution into example-based learning particle swarm optimization for continuous optimization（ELPSO）.This new algorithm took advantages of hyperspherical manipulations to change the magnitude and the directions of particles.Hence,particles overrid the deep local minima,meanwhile,found the diverged particles and accelerated them towards optimal solution.We also did empirically testing and compared it with other published methods on benchmark functions.The experimental results illustrated that the proposed algorithm largely improved convergence speed and convergence accuracy.The global search capability had been significantly improved.

关 键 词：粒子群优化算法 种群多样性 超球坐标 辨识和加速最差粒子 

分 类 号：O242[理学—计算数学]