摘要
对多模态函数优化问题,分析了各种小生境策略;将拥挤模型与聚类算法相结合,提出了一种拥挤聚类遗传算法.拥挤模型在适应值曲面上形成多个小生境,聚类算法消除了每个小生境内部的基因漂移现象.理论分析证明了算法的收敛性能.数值实例表明,拥挤聚类模型在多极值搜索的数量、质量和精度上都优于拥挤模型与确定性拥挤模型.将拥挤聚类遗传算法应用于国家同步辐射实验室变间距全息光栅的设计,取得了满意的效果.*
For muhimodal function optimization problems, this paper analyzes several niching strategies, combines the crowding model with the clustering algorithm, and proposes a crowding clustering genetic algorithm. Crowding model is used to form multiple niches in fitness landscape, while clustering algorithm eliminates genetic drift in each inner niche. Theoretical analysis proves the convergence property of the proposed algorithm. Numerical results indicate that crowding clustering model is superior to both crowding model and deterministic crowding model in quantity, quality and accuracy of multi-optima searching. The crowding clustering genetic algorithm has been applied to the varied-line-spacing holographic grating design in the National Synchrotron Radiation Laboratory, and achieves satisfactory results.
出处
《信息与控制》
CSCD
北大核心
2006年第6期715-720,共6页
Information and Control
基金
安徽省优秀青年科技基金(04042046)
关键词
多模态函数优化
拥挤聚类遗传算法
基因漂移
变间距全息光栅
muhimodal function optimization
crowding clustering genetic algorithm
genetic drift
varied-line- spacing holographic grating
