A modified back-propagation method to avoid false local minima

Abstract

The back-propagation method encounters two problems in practice, i.e., slow learning progress and convergence to a false local minimum. The present study addresses the latter problem and proposes a modified back-propagation method. The basic idea of the method is to keep the sigmoid derivative relatively large while some of the error signals are large. For… (More)
DOI: 10.1016/S0893-6080(98)00087-2

Topics

Cite this paper

@article{Fukuoka1998AMB, title={A modified back-propagation method to avoid false local minima}, author={Yutaka Fukuoka and Hideo Matsuki and Haruyuki Minamitani and Akimasa Ishida}, journal={Neural networks : the official journal of the International Neural Network Society}, year={1998}, volume={11 6}, pages={1059-1072} }