Title :
Magnified gradient function in adaptive learning: the MGFPROP algorithm
Author :
Ng, Sin-Chun ; Cheung, Chi-Chung ; Leung, Shu-hung
Author_Institution :
Dept. of Comput. & Math., Inst. of Vocational Educ., Hong Kong, China
fDate :
1/4/2001 12:00:00 AM
Abstract :
A new algorithm is proposed to solve the “flat spot” problem in backpropagation neural networks by magnifying the gradient function. Simulation results show that, in terms of the convergence rate and the percentage of global convergence, the new algorithm consistently outperforms other traditional methods
Keywords :
backpropagation; convergence; feedforward neural nets; MGFPROP algorithm; adaptive learning; backpropagation neural networks; convergence rate; flat spot problem solution; global convergence; magnified gradient function;
Journal_Title :
Electronics Letters
DOI :
10.1049/el:20010002
