基于自动定义函数GP的自适应建模研究  

Study on Adaptive Modeling Based on Genetic Programming with Automatically Defined Function

作　　者：唐丽珏[1] 李淼[1] 张建[1] 张勇进[1] 

机构地区：[1]中国科学院合肥智能机械研究所,安徽合肥230031

出　　处：《小型微型计算机系统》2005年第6期1000-1003,共4页Journal of Chinese Computer Systems

基　　金：国家"八六三"高技术研究发展计划信息技术项目基金(2001AA115180)资助

摘　　要：遗传程序设计(GeneticProgramming,简称GP)在进化过程中由于种群多样性的损失,常导致低收敛性.本文尝试将自动定义函数引入到GP中克服这个问题,并应用于数据的自适应建模.文中将两者的性能进行了比较,实验表明自动定义函数的发现和使用增加了种群的多样性.它不仅降低了整个遗传程序的大小,还增加了GP搜索的计算有效性,提高了收敛性能,取得了满意的结果.Genetic Programming (GP) often results in low convergence because of the loss of population diversity during the evolutionary process. This paper attempts to introduce Automatically Defined Function into GP to overcome it and apply it to adaptive modeling of data. The performance between them is compared. The result shows that the discovery and use of automatically defined functions increase the diversity of population. It also reduces the complexity of individuals, increases computation validity of GP training, strengthens the convergence performance and gets satisfying results.

关 键 词：自动定义函数 遗传程序设计 自适应建模 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]