【摘要】 采用Mamdani模型作为模糊分类器 ,利用神经网络建立非线性模型 ,构造一种分布式神经网络。采用多组样本数据建模 ,根据各输入模糊子集和隶属度函数 ,将输入样本空间模糊分割成多个子空间 ,对每个子空间用一个神经网络模型建立映射关系。对每一组输入向量在确定归属类后 ,自动切换至对应的子网络作为输入 ,该子网络的输出值则作为分布式网络的输出。仿真结果表明 ,该方法与用单个神经元网络相比 ,明显提高了模型的精度和泛化能力。更多还原

【Abstract】 A kind of distributed neural networks is constructed by using Mamdani model as a fuzzy classification to establish a nonlinear model.With the help of a group of sample data from on the spot acquisition,the input sample space is split into several sub space according to each input fuzzy subset and membership function,and every mapping relation is set to each sub space by using a neural networks.The output values of corresponsive sub-networks are outputted as total output of the model.The results of simulation show that the precision and generalization ability of the model are upgraded with the help of this method comparative to the single neural network.更多还原