用于多峰函数优化的改进小生境微粒群算法  

A modified niching particle swarm optimization algorithm for multimodal function

作　　者：杨诗琴[1] 须文波[1] 孙俊[1] 

机构地区：[1]江南大学信息工程学院,江苏无锡214122

出　　处：《计算机应用》2007年第5期1191-1193,1200,共4页journal of Computer Applications

基　　金：国家自然科学基金资助项目(60474030)

摘　　要：针对小生境微粒群算法在处理复杂多峰函数优化问题中存在的一些缺陷,提出一种改进的小生境SNPSO算法。SNPSO算法将顺序小生境的思想引入其中,首先在主群体中应用Stretching技术,其次对子群体采用解散策略,即当在子群体中找到一个极值点后把子群体解散回归主群体,最后设置子群体创建时的半径阈值,避免子群体半径过大。该算法解决了标准的NichePSO算法在处理多峰函数时,极值点的个数依赖于子群体个数及极值点容易出现重复、遗漏等问题。对3个常用的基本测试函数的实验表明,新算法(SNPSO)在多峰函数寻优中解的稳定性、收敛性和覆盖率均优于标准NichePSO。A modified niching Particle Swarm Optimization （PSO） algorithm was constructed which allowed unimodal function optimization methods to efficiently locate all optima of multimodal problems that the Niche PSO cannot reach. In the new algorithm, the sequential niche technique was introduced. Firstly, a stretching technique was adopted in main swarm. Secondly, the dismissal mechanism was used in sub-swarms namely when a local extreme point of value was found in subswarms, the sub-swarms would be dismissed and regressed to the main swarm. At last, the radius of created sub-swarms was confined in order to avoid the excessive of radius. The new Stretching-Niche PSO （SNPSO） algorithm could resolve the disadvantage of standard Niche PSO that the local best of value depends on the number of sub-swarms and easily has the problem of iteration and pretermission. Testing of the algorithm by using three benchmark functions indicate that the modified niching PSO has better performance than standard Niche PSO in terms of the stability, convergence and coverage in searching a better value.

关 键 词：微粒群算法 小生境 Stretching技术 子群解散策略 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP301.6[自动化与计算机技术—控制科学与工程]