【摘要】 对于多输入 CMAC神经网络 ,概念映射决定了网络泛化特性 .从而直接影响到网络学习速度和网络的逼近精度 .本文在对多输入 CMAC网络进行深入分析的基础上 ,指出了概念映射的重要性及现有映射算法存在的不足 ;巧妙地将网络存储器由原来的一维改为多维从而得到了一类新型的多输入 CMAC神经网络 ,同时给出了新型网络的概念映射算法并分析论证了算法的合理性、严密性 .计算机仿真结果验证了所提出新型网络及其概念映射算法的有效性更多还原

【Abstract】 For the multi input CMAC neural network, the conception mapping algorithm have a strong impact on its generalization and learning rate and learning precision. In this paper, the multi input CMAC neural network is comprehensively investigated. The vital importance of the conception mapping for a CMAC neural network and its drawbacks in existing CMAC neural network are pointed out. From this basis, by replace the existing one dimensional memory architecture with multi dimensional architecture, a new CMAC neural network and its conception mapping algorithm are presented. Additionally, the rationality and rigor of the new algorithm are analyzed. Simulation results demonstrate that efficiency of the new network and the new algorithm for conception mapping.更多还原