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 the fuzzy neural network, the class of compact continuous congruous functions is generalized as preserving maximally valued functions, and then the properties and metrological problems of expanding ambiguous function classes are studied by means of preserving maximalities. Finally, we obtain the general approximation of the extended fuzzy functions class of the three-layer regular fuzzy neural network and analyze the approximation effect of the target output and the actual output of the fuzzy neural network by examples.