论文部分内容阅读
给出了一个关于多层感知器（ＭＬＰ）连续函数逼近能力更为普遍的新结论，表明ＭＬＰ的函数逼近能力与隐层数目及隐层神经元特性函数的选取关系不大，而主要是决定于ＭＬＰ的隐层神经元数目
A new conclusion about MLP continuous function approximation ability is given. It shows that the function approximation ability of MLP has little to do with the number of hidden layers and the characteristic function of hidden layer neurons, but mainly determines The number of hidden layer neurons in MLP