Release Time:2019-03-11  Hits:

Indexed by: Journal Article

Date of Publication: 2007-03-01

Journal: Journal of Information and Computational Science

Included Journals: EI

Volume: 4

Issue: 1

Page Number: 251-255

ISSN: 15487741

Abstract: Gradient algorithm has been widely used as a learning algorithm for training the weights of feedforward neural networks. A simple criterion is used to choose the learning rate during each cycle of training iteration. A convergence result of the corresponding batch gradient algorithm and the monotonicity error function in the iteration are proved, which are stronger than the usual results for general optimization problems. A supporting numerical experiment is also presented.