用于多峰值函数优化的对数自适应排挤遗传算法  

Logarithmic adaption crowding genetic algorithm for multimodal function optimization

作　　者：刘文涛[1] 胡家宝[2] 

机构地区：[1]武汉轻工大学数学与计算机学院,武汉430023 [2]武汉理工大学计算机科学与技术学院,武汉430070

出　　处：《计算机应用》2014年第6期1645-1648,1675,共5页journal of Computer Applications

摘　　要：排挤遗传算法能够比较稳定地获取多个峰值,但其求解效率不高,在有限的遗传代数下无法获得较高的求解精度,需要较多的迭代次数。为了快速求出多峰函数的所有最优解,提出了一种基于对数自适应的排挤遗传算法。该算法结合小生境排挤遗传和爬山算子,根据遗传代数对爬山算子的距离值进行对数自适应计算,使种群在遗传过程中保持多样性。通过对多个一维和二维多峰函数的实验和比较分析,测试结果表明,该算法在有限的遗传代数下既能保证求解精度又能提高收敛速度,能够比较稳定地求得所有最优解,是求解多峰函数问题的有效算法。Crowding genetic algorithm can obtain multiple optima of multimodal functions, but it has low efficiency, and cannot get a higher precision in limited iterations. In order to obtain all optima of the muhimodal function quickly, the crowding genetic algorithm based on logarithmic adaption was presented combined with niche crowding genetic and climbing operators. The algorithm computed the distance values of climbing operators by logarithmic adaption according to the iterations, which made the population maintain genetic diversity in the process. According to the experiments and comparative analysis of several one-dimensional and two-dimensional multimodal functions, the test results show that the algorithm can ensure both the solution accuracy rate and the convergence speed in the limited iterations, and obtain all optimal solutions more stably. It is proved to be an effective algorithm for the multimodal function problems.

关 键 词：排挤遗传算法 多峰函数优化 爬山算子 对数自适应 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]