论文部分内容阅读
提出了一种模糊ＣＭＡＣ（小脑模型关节控制器）神经网络，它由输入层、模糊化层、模糊相联层、模糊后相联层与输出层等５层节点组成，具有与ＣＭＡＣ相似的单层连接权，可通过ＢＰ算法学习推论参数或模糊规则．给出了网络的连接结构与学习算法，并将其应用于函数逼近问题中仿真结果验证了该方法较之ＣＭＡＣ的优越性．
A fuzzy CMAC (cerebellar model joint controller) neural network is proposed. It consists of 5 layers of nodes, such as input layer, fuzzy layer, fuzzy associative layer, post-fuzzy associative layer and output layer, Layer connection rights can be learned by BP algorithm inference parameters or fuzzy rules. The connection structure and learning algorithm of the network are given and applied to the approximation of function approximation. The simulation results verify the superiority of this method over CMAC.