A fast learning algorithm of neural networks by changing error functions

In order to improve the training speed of multilayer feedforward neural networks, we propose and explore a new fast backpropagation (BP) algorithms obtained by changing the error functions, in case using the Fourier kernel function as alternative functions; to overcome the conventional BP learning problems of getting stuck into local minima or slow convergence. Our experimental results demonstrate the effectiveness of the modified error functions since the training speed is faster than that of existing fast methods.