应用自适应指数比例变换的适应值共享遗传算法  

The Fitness Sharing Genetic Algorithms with Adaptive Power Law Scaling

作　　者：于歆杰[1] 王赞基[1] 

机构地区：[1]清华大学电机工程与应用电子技术系,北京100084

出　　处：《系统工程理论与实践》2002年第2期24-28,共5页Systems Engineering-Theory & Practice

基　　金：清华大学基础研究基金 ( JC2 0 0 1 0 1 9)

摘　　要：适应值共享遗传算法是用来解决多峰函数优化的常用手段 .本文提出了可以用来提高适应值共享遗传算法搜索能力的自适应比例变换方法 .针对复杂欺骗性问题 ,本文采用不同的指数比例变换方法进行了优化 .结果表明 ,本文提出的自适应指数比例变换方法能够迅速、稳定地找到全部全局峰 .The fitness sharing genetic algorithms are the common approaches to solve multi-modal optimization problems. In this paper, a new adaptive power law scaling method is suggested to improve the search ability of the fitness sharing genetic algorithms. Different power law scaling methods have been adopted to optimize the massive deceptive problem. The empirical results show that the new adaptive power law scaling method can find all the global peaks steadily and quickly. This method is especially suit for the problems whose radii of peaks are hard to define.

关 键 词：多峰函数优化 适应值共享 自适应比例变换 遗传算法 

分 类 号：O242.23[理学—计算数学]