针对目前用于信号（函数）逼近的子波神经网络的收敛速度慢的问题，提出一种子波网络的初值设定方法，使子波网络的收敛速度大大提高，通过对线性多项式、指数、三角函数及某些分段函数的实验以及与不同网络的对比，表明此方法具有很强的普适性 Aiming at the problem of slow convergence of the wavelet neural network which is used for signal (function) approximation, an initial value setting method for wavelet network is proposed. The convergence speed of wavelet network is greatly improved. By linear polynomial, exponential, The experiments of trigonometric functions and some piecewise functions and the comparison with different networks show that this method is very universal