机构地区: 西安电子科技大学
出 处: 《电子学报》 2001年第z1期1878-1885,共8页
摘 要: 本文首次对子波神经网络的分类机理进行了详细地研究 ,在此基础上将免疫进化算法与子波神经网络相集成 ,并提出了免疫进化子波网络模型及其学习算法 .其中 ,免疫进化算法是一种集免疫机制和进化机制于一体的一种新的全局并行算法 ,它可以通过对进化环境的自适应和自学习 ,有针对性地抑制由原进化算子操作的盲目性而引起的退化现象 .理论分析和用于双螺旋线分类的仿真结果表明 ,免疫进化子波网络不仅是可行的 ,也是十分有效的 .由于免疫进化算法本身所固有的并行运算规则、智能搜索方式和概率判断准则 ,从而有利于弱化子波网络的应用条件 。 The classifiable function of the wavelet neural network (WNN) is studied in this paper,and an integration of the immune evolutionary algorithm and WNN is proposed,based on the model of the immune evolutionary wavelet neural network (IEWNN).In this model,the immune evolutionary algorithm(IEA) used for training the network is a globally parallel algorithm which integrates the immune and the evolutionary mechanisms.The IEA is conducive to alleviating the undulating phenomenon produced by the blindfold behaviors of the original operators in the existent evolutionary algorithms through adaptively learning form the evolutionary environment.The analysis in theory and simulations show that WNN based on IEA is not only feasible but also effective.Because IEA has the parallel operation rules,the intelligent searching behavior and the criterion based on probability,it is conducive to relaxing the conditions of applying WNN and improving the ability of association memory and information processing.