基于小波网络的非线性逼近方法

利用小波空间中函数的多分辨率分解思想，结合神经网络构造了一种多尺度正交小波网络模型，利用最小二乘法对网络参数进行确定，对非线性函数进行了逼近的仿真研究，与前向BP神经网络进行了仿真比较。结果表明，这种新的小波网络具有良好的函数逼近能力，收敛速度快且逼近精度高。

Method of Approaching Nonlinear Function Based on Wavelet Network

Based on the theory of wavelet analysis and combined with neural network, A multiresolution orthogonal wavelet neural network is constructed, and the weights are confirmed by LS algorithm . Simulations on types of nonlinear function are studied. The comparison with the traditional BP neural network is also given. Simulation results indicate that the proposed wavelet network in the paper has a good approximation capability and has the advantages of rapid convergence rate and high approximation accuracy.