Derivation and global convergence for pseudo - Newton - δ class 伪Newton-δ族的导出和全局收敛性
Derivation and Property of Algorithm of Pseudo - Newton - R Class 伪Newton-R族算法的导出及其性质
The results show that, for a wavelet neural network, the pseudo - Newton algorithm makes the error function descend fastest, and gives the best result when using the trained neural network to express a highly nonlinear function. 结果证明,采用小波神经网络结合拟牛顿学习算法,可以得到最快的学习收敛速度,并且具有最好的学习效果,即对强非线性函数的拟合精度最高。