非线性函数逼近的神经网络方法

An Approximation of Nonlinear Function with Neural Network

针对工程实际中一类典型的非线性函数逼近问题,本文阐述了单层函数型网络及双层网络在应用时各自的特点,并提出了对网络输入节点进行优化选择,可提高网络的逼近精度,简化网络结构.文中给出了具体的选择方法,计算机模拟结果验证了试方法的有效性.

A method for the approximation of a typical nonlinear function is presented by neural network theory,in which the characteristics of a kind of function-linked neural network with one or two layers are studied in detail.The results show that the optimum selection of the input nodes is important in getting high accuracy and simplicity structure of neural network.Therefore a selection proceeding of input node is given,the emulation shows its availability.