基于Hamming神经网络聚类分析的进化策略
Evolutionary Strategies with Clustering Analysis by Hamming Nets
-
摘要: 本文提出了一种基于Hamming神经网络聚类分析的进化策略,模糊自适应Hamming神经网络各类族的权重矢量纪录被进化搜索过的区域,并相应记录下该区域内最优个体和它的适应度,因此通过Hamming神经网络对进化个体的聚类分析,进化策略具有搜索记忆性,可以充分保证下一代遗传群体中个体遗传基因的丰富性,从而避免早熟现象的发生,这种进化策略还可以避免在被搜索过的区域内的无用搜索,进而加快进化策略的收敛速度,并可在收敛时同时给出解空间内的多个全局的局部最优解。Abstract: The evolutionary strategies with clustering analysis by Hamming,neural network is proposed in this paper.Cluster vectors of fuzzy adaptive Hamming neural network record the evolutionary search areas,and the optimum strings in those areas and their fitness are also recorded at the same time.Through the clustering analysis to the evolutionary strings by Hamming neural network,the search process has the ability of recalling.Therefore the chromosome information of evolutionary population can be maintained abundant enough to avoid the early convergence.This evolutionary strategies can also gain higher convergence speed by avoid nonsense search and can give multiple local optima at the same time.