正则模糊神经网络对保极大值函数类的泛逼近性  

Universal approximation of three-layer regular fuzzy neural networks for a class of functions

作　　者：段晨霞[1] 孙刚[2] 王贵君[1] 

机构地区：[1]天津师范大学数学科学学院,天津300387 [2]大连海事大学轮机工程学院,辽宁大连116026

出　　处：《山东大学学报（理学版）》2012年第3期81-86,共6页Journal of Shandong University(Natural Science)

基　　金：国家自然科学基金资助项目(60974144)

摘　　要：在模糊神经网络中将紧集上多元连续函数类推广为保极大值函数类,进而借助于保极大值性研究了紧集上扩展模糊函数类的特性和度量问题。最后,获得三层正则模糊神经网络关于其扩展模糊函数类具有泛逼近性,并通过实例分析了目标输出与正则模糊神经网络实际输出的逼近效果。In fuzzy neural networks, the multi-variable continuous functions class is extended as a functions class of pre- serving maximum values on compact sets. The properties and the metric problems of the extended functions class on compact sets are studied by means of the concept of preserving maximum values. Finally, it is found that three-layer regular fuzzy neural networks possess approximation with respect to the extended fuzzy function class. Also, the ap- proximation effect between the expected output and the real output of regular fuzzy neural networks is analysed and text is studied by an example.

关 键 词：正则模糊神经网络 保极大值 支撑函数 泛逼近性 

分 类 号：O159[理学—数学] TP183[理学—基础数学]