量子粒子群优化算法的收缩-扩张系数控制策略研究  

Control strategy of contraction-expansion coefficient in quantum-behaved particle swarm optimization

作　　者：黄为勇[1] 徐晓菊[1] 潘晓博[1] 孙彦景[2] 李松[2] Huang Weiyong;Xu Xiaoju;Pan Xiaobo;Sun Yanjing;Li Song

机构地区：[1]徐州工程学院江苏省大型工程装备检测与控制重点建设实验室,江苏徐州221111 [2]中国矿业大学信息与电气工程学院,江苏徐州221116

出　　处：《计算机应用研究》2016年第9期2592-2595,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(51274202);江苏省基础研究计划(自然科学基金)资助项目(BK20131124);徐州工程学院江苏省大型工程装备检测与控制重点建设实验室开放基金资助项目(JSKLEDC201212)

摘　　要：收缩—扩张(CE)系数是量子粒子群优化算法(QPSO)需要人工设定的最核心参数,如何选择该参数成为一个重要的问题。为寻找更为有效的CE系数控制方法,根据CE系数递减思想,提出了一种凸凹性可变的指数型非线性下降CE系数控制策略。采用Sphere、Rastrigrin、Griewank和Ackley四种典型的单峰与多峰标准测试函数研究了CE系数的不同控制策略以及不同初始值对量子粒子群优化算法收敛精度与收敛速度的影响,并与线性下降CE系数及固定CE系数两种控制策略进行了对比分析,得出了CE系数控制策略选择的一般性指导准则,为量子粒子群优化算法的应用提供依据。Considering contraction-expansion( CE) coefficient is the most influential parameter needed to be set artificially in the application of quantum-behaved particle swarm optimization (QPSO) , how to select the parameter has become an important issue. In order to search for an effective control method of CE coefficient, this paper proposed the control trategy of descent exponential nonlinear CE coefficient according to the idea of decreasing CE coefficient. It tested the optimal performance of descent exponential CE coefficient with different convexity and different initial value respectively on 4 typical unimodal or multimodal benchmarks of Sphere, Rastrigrin, Griewank and Ackley function. It analysed the impact of differnent control strategy of CE coefficient on convergence accuracy and convergence speed, and compared the experimental results obtained tothose by the control strategy of linear descent CE coefficient and fixed CE coefficient. According to the test results, it draws some conclusions concerning the control trategy of CE coefficient, which provides guidance in the selection of CE coefficient when using QPSO algorithm.

关 键 词：量子粒子群优化算法 收缩一扩张系数 控制策略 指数型非线性 凸凹性 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]