Title :
Convergence and generalization properties of multilayer feedforward neural networks
Author :
Tang, Chuan Zhang ; Kwan, Hon Keung
Author_Institution :
Dept. of Electr. Eng., Windsor Univ., Ont., Canada
Abstract :
The effects of various parameters of multilayer feedforward neural networks on their convergence speed and generalization capability are analyzed and compared. Based on these analyses, an enhanced version of the backpropagation algorithm, BPABAS, is proposed. It combines backpropagation with adaptive slope of the activation function and adaptive bias of the neuron. The algorithm is shown to speed up the learning process while increasing generalization capability
Keywords :
backpropagation; feedforward neural nets; BPABAS; activation function; adaptive bias; adaptive slope; backpropagation algorithm; convergence speed; generalization properties; learning process; multilayer feedforward neural networks; Algorithm design and analysis; Artificial neural networks; Backpropagation algorithms; Convergence; Feedforward neural networks; Impedance matching; Multi-layer neural network; Neural networks; Neurons; Pattern matching;
Conference_Titel :
Circuits and Systems, 1992. ISCAS '92. Proceedings., 1992 IEEE International Symposium on
Conference_Location :
San Diego, CA
Print_ISBN :
0-7803-0593-0
DOI :
10.1109/ISCAS.1992.230013
