高维多峰优化的遗传算法设计  

Design of High-dimensional Multimodal Optimization of GA

作　　者：孙如祥[1] 黄春[2] 邓国斌[1] Sun Ruxiang;Huang Chun;Deng Guobin(Department of Computer and Electronic Information Engineering,Guangxi Vocational and Technical College,Nanning 530226,China;School of Computer,Electronics and Information in Guangxi University,Nanning 530000,China)

机构地区：[1]广西职业技术学院计算机与电子信息工程系,南宁530226 [2]广西大学计算机与电子信息学院,南宁530000

出　　处：《科技通报》2017年第8期197-201,229,共6页Bulletin of Science and Technology

基　　金：广西高校中青年教师基础能力提升项目(KY2016LX495;KY2016YB610)

摘　　要：针对传统遗传算法用于多峰值问题时容易出现的问题,提出了一种基于适应度自动调节的改进遗传算法(FMT-GA)。FMT-GA算法采用了与传统遗传算法不同的适应度评估方法以及选择算子,并设计了基于适应度值大小的类似于非均匀变异的自适应变异算子以及自适应交叉算子,在约束条件的处理上,与传统的做法也有较大差异。文章最后对2个多峰值函数进行了实验测试,测试结果表明,FMT-GA算法克服了传统遗传算法易停滞于局部极值的缺陷,收敛精度以及速度都有了比较明显的提高。Whentraditional genetic algorithm is used to solve multi-peak problems,some difficulties are easy to occur.For these difficulties,wepresent an improvedGA that can automatically adjust itself based on fitness(FMT-GA).It adoptsdifferent fitness evaluation methodand selection operator,and design selfadaptive mutation operator that are similar to the size of the non-uniform mutation and self-adaptive crossover operator.On the processing of constraints,we also have bigger difference with traditional practice.Finally,we take a test to2multi-peak functions and the result show that FMT-GA overcomes the limitation of stagnation in the local extremeof traditional GA,precision and convergence speed have obvious improvement.

关 键 词：多峰 遗传算法 适应度 迁徙 

分 类 号：TP302[自动化与计算机技术—计算机系统结构]