摘要
针对小波神经网络训练不便的问题,构造了一种基于MATLAB神经网络工具箱的训练方法。首先通过公式推导简化了小波神经网络的学习参数,将伸缩因子和平移因子变换到了权重和阈值中。然后根据神经网络工具箱对激活函数的要求,采用MATLAB编制了适用于神经网络工具箱的小波基激活函数,简单描述了编写自定义激活函数的方法。最后将该小波神经网络应用于织物的热阻和湿阻预测中,采用贝叶斯正则化训练算法,取得了较好的效果。
One simple method for training wavelet neural network was discussed on the basis of MATLAB neural network toolbox.Firstly,the dilation factor and the translation factor of wavelet neural network were combined into the weight and the threshold by means of formula derivation respectively.Then the MATLAB codes of wavelet basis function suitable for MATLAB neural network toolbox were implemented.The idea of how to program the user-defined activation function was given.The prediction application to the fabric thermal resistance and wet resistance using Bayesian regulation algorithm verifies the effectiveness and feasibility of the method.
出处
《信息技术》
2011年第1期10-13,共4页
Information Technology
基金
河南省自然科学基金资助(082300410090)
