基于聚类分析和资源竞争模型的生境遗传算法

A Niching Genetic Algorithm Based on Clustering and Resource Competition

传统的遗传算法在处理多模函数优化问题时 ,只能收敛到单个极值点 使用生境机制的改进算法可以获得多个局部极值点 提出了一种新的基于聚类分析和资源竞争模型的生境遗传算法 这种算法将聚类分析、共享技术和拥挤技术有机地结合起来 ,可以有效地对多模函数进行优化 ,而无需事先确定生境的具体数目和生境半径的大小 通过数学分析 ,证明了这种算法可以控制收敛到的生境的数目 ,避免找到无效的极值点

A standard genetic algorithm only has the capability of converging to a single peak point eventually for a multimodal function optimization problem An improved GA used niche mechanism can gain a set of local peak A new niching GA based on clustering and resource competition is approved Clustering, a sharing model, and a crowd method are organized in this algorithm A multimodal function can be optimized efficiently by it without determining the count of niche and the value of niche radius It is proved by mathematics analysis that the number of converged peak can be controlled so as to avoid the inefficient points These conclusions are approved by the test of a typical problem