基于模糊系统的新型CMAC神经网络及学习算法

A New CAMC and Its Learning Algorithm based on Fuzzy System

提出了一种基于模糊系统的新型CMAC神经网络,该神经网络与 CMAC相比,它不需要对输入分量进行量化,而且能够根据实际问题的性质来初始化网络参数,有利于提高网络的收敛速度.与一般的FCMAC相比,它的逼近精度更高,能够解决 CMAC系列网络逼近精度不高的弱点,所以此网络的实际应用前景更广阔.所做的大量仿真实验也证明了这一特性.

In this paper, a new CMAC and its learning algorithm based on the fuzzy system is presented. Compared with CMAC, there is no need for discretization of input components. Besides, the parameters of the presented network can be initiated according to the real problems. Therefore, its convergence rate will be promoted. Compared with general FCMAC, its effect of approximation is more accurate. It can solve the fatal false that the neural network of the CMAC and general FCMAC can not avoid. The neural network presented in the paper will be more applicable. Finally, many simulation results also demonstrate its advantage.