Title :
Analysis of the misadjustment of BP network and an improved algorithm
Author :
Jia, Yang ; Dali, Yang
Author_Institution :
Beijing Univ. of Posts & Telecommun., China
Abstract :
The authors analyze the misadjustment of the conventional backpropagation (BP) network introduced by the momentum factor α and its disadvantageous effect on the network, and propose an improved algorithm with stochastic attenuation momentum factor α(k). Analysis and simulations show that the momentum misadjustment is a nonnegligible factor affecting the convergence precision of the BP network. Compared with the conventional BP algorithm, the improved algorithm can effectively cancel the negative effect of α on a network and improve the convergence performance of the network
Keywords :
backpropagation; convergence; neural nets; stochastic processes; BP network; backpropagation; convergence performance; convergence precision; misadjustment; momentum factor; stochastic attenuation; Adaptive algorithm; Algorithm design and analysis; Convergence; Error correction; Feedforward neural networks; Least squares approximation; Multi-layer neural network; Neural networks; Performance analysis; Stochastic processes;
Conference_Titel :
Circuits and Systems, 1993., ISCAS '93, 1993 IEEE International Symposium on
Conference_Location :
Chicago, IL
Print_ISBN :
0-7803-1281-3
DOI :
10.1109/ISCAS.1993.394296
