摘要
遗传算法作为一种随机化优化搜索方法,已经在很多领域得到了成功应用,但其存在控制参数多且配置困难的问题.本文采用一类最新试验设计方法-计算机试验设计,对遗传算法的参数配置进行优化.结果表明,基于正交拉丁超立方设计的参数配置,其算法的计算精度和速度表现最佳.模拟结果进一步讨论了不同试验设计方案在遗传算法中的差别.
As a random searching method,genetic algorithm has been widely used in many aspects.However,the parameter configuration of genetic algorithm is still difficult to confirm.This paper proposed a new class of design of experiment-computer experiment to optimize the parameter settings in genetic algorithm.Simulations show that the parameter settings based on orthogonal Latin hypercube design has the best performances both in accuracy and convergence rates.In addition,simulation results also reveal the differences of effects of four commonly used designs in parameter configurations of genetic algorithm.
作者
朱文佳
范家旭
陈雪平
郭兵
ZHU Wen-jia;FAN Jia-xu;CHEN Xue-ping;GUO Bing(Fundamental Education Department,Changzhou Technician College Jiangsu Province,Changzhou 213032,China;Department of Mathematics,Jiangsu University of Technology,Changzhou,213001,China;College of Mathematics,Sichuan University,Chengdu 610064,China)
出处
《数学的实践与认识》
北大核心
2020年第5期112-118,共7页
Mathematics in Practice and Theory
基金
国家自然科学基金(11601195,11971204)
江苏省自然科学基金(BK20160289)
江苏高校“青蓝工程”。
关键词
遗传算法
正交设计
拉丁超立方设计
正交拉丁超立方设计
genetic algorithm
orthogonal design
latin hypercube design
orthogonal latin hypercube design
