基于Nelder-Mead单纯形法的改进量子行为粒子群算法  

A Improved Quantum Behaved Particle Swarm Optimization Algorithm Using Nelder and Mead′s Simplex Algorithm

作　　者：郑伟博[1] 张纪会[1] 

机构地区：[1]青岛大学复杂性科学研究所,山东青岛266071

出　　处：《复杂系统与复杂性科学》2016年第2期97-104,共8页Complex Systems and Complexity Science

基　　金：山东省自然科学基金(ZR2010GM006)

摘　　要：针对PSO算法搜索精度较低,并且在复杂多模态函数优化中,容易陷入局部极值的问题,提出了一种改进的量子行为粒子群优化算法。研究了该算法的基本原理、给出了算法流程并采用正交试验的方式获得了一套通用性较强的算法参数。并以CEC’13的28个测试函数作为测试集,采用Wilcoxon符号秩检验将NM-QPSO算法分别与PSO算法和QPSO算法的误差进行比较试验。试验表明:NM-QPSO算法在统计意义上优于传统的PSO算法和QPSO算法,并且在高维函数优化中,具有显著优势。PSO algorithm is poor in search accuracy and prone to fall into the local extremum when solving complex multimodal function optimization problem. So, we propose an improved quantum behaved particle swarm optimization algorithm. This paper studies the fundamentals and basic procedure of that algorithm, An orthogonal test for parameter selection is designed to select a set of reasonable control parameters. We use a suite of 28 test functions from CEC^13 as test set. NM-QPSO is compared with both of traditional PSO and QPSO by using the Wileoxon Signed Ranks Test respectively. Tests show that the NM-QPSO algorithm has better performance than the traditional PSO and QPSO algorithms in statistical sense, and it has obvious advantages in the high-dimensional function optimization.

关 键 词：群体智能 粒子群优化算法 量子行为粒子群优化算法 Nelder Mead单纯形法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]