TY - GEN
T1 - A new adaptive learning algorithm using magnified gradient function
AU - Ng, S. C.
AU - Cheung, C. C.
AU - Leung, S. H.
AU - Luk, A.
PY - 2001
Y1 - 2001
N2 - In this paper, a new algorithm is proposed to solve the "flat spot" problem in back-propagation networks by magnifying the gradient function. The idea of the new learning algorithm is to vary the gradient of the activation function so as to magnify the backward propagated error signal gradient function especially when the output approaches a wrong value, thus the convergence rate can be accelerated and the flat spot problem can be eliminated. Simulation results show that, in terms of the convergence rate and global search capability, the new algorithm always outperforms the other traditional methods.
AB - In this paper, a new algorithm is proposed to solve the "flat spot" problem in back-propagation networks by magnifying the gradient function. The idea of the new learning algorithm is to vary the gradient of the activation function so as to magnify the backward propagated error signal gradient function especially when the output approaches a wrong value, thus the convergence rate can be accelerated and the flat spot problem can be eliminated. Simulation results show that, in terms of the convergence rate and global search capability, the new algorithm always outperforms the other traditional methods.
UR - https://www.scopus.com/pages/publications/0034844242
UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-0034844242&origin=recordpage
M3 - RGC 32 - Refereed conference paper (with host publication)
VL - 1
SP - 156
EP - 159
BT - Proceedings of the International Joint Conference on Neural Networks
T2 - International Joint Conference on Neural Networks (IJCNN'01)
Y2 - 15 July 2001 through 19 July 2001
ER -