类似CMAC的模糊神经网络及其在控制中的应用

提出的一种模糊神经元网络是模糊逻辑的一种网络结构的实现。该网络由特征网络 和功能网络两部分组成。特征网络用来产生模糊规则的前件，相应于每条规则的适用度。功能 网络用来实现模糊规则的后件。最后的输出则为各模糊规则后件的加权和。该网络具有小脑模 型关节控制器（ＣＭＡＣ）的一些性质，具有神经元网络和模糊逻辑两者的优点。它既可以容易地 表示模糊和定性的知识，又具有较好的学习能力。文章同时给出了网络在控制中应用的两种实 现结构以及用作非线性映射的一个算例。

CMAC-like fuzzy neural network and its application to controls

A fuzzy neural network is presented. It is essentially a network implementation of fuzzy logic. The network is comprised of two parts:a context network and a function network. The context network matches the premises of fuzzy rules and produces a matching factor as outputs for each rules. The function network performs the consequences of fuzzy rules. The total output is the weighting sum of the consequences with matching factors as weighting coefficients. The network has a similar structure to the Cerebella Model Atriculation Controller (CMAC) in some aspects. It has advantages both from the neural network and the fuzzy logic. The fuzzy neural network can express the fuzzy and quantitative knowledge easily and has good learning abilities. Two control structures based on the fuzzy neural network are given. An example of nonlinear function mapping is presented to show advantages of the network.