Title :
Speeding up the convergence of backpropagation networks
Author :
Sureerattanan, Songyot ; Phien, Huynh Ngoc
Author_Institution :
Asian Inst. of Technol., Pathumthani, Thailand
Abstract :
A new algorithm is proposed for speeding up the convergence of backpropagation (BP) networks. This algorithm is obtained by applying the momentum term and adaptive neuron model with temperature momentum term to the Kalman filter (KF) algorithm. It is found that this algorithm can perform satisfactorily in all cases considered. Not only the convergence rate can be improved, but also the sum of squared errors can be further reduced
Keywords :
Kalman filters; backpropagation; convergence; neural nets; Kalman filter algorithm; adaptive neuron model; backpropagation networks; convergence; convergence rate; momentum term; squared errors sum; temperature momentum term; Backpropagation algorithms; Convergence; Electronic mail; Equations; Multi-layer neural network; Multilayer perceptrons; Neurons; Nonhomogeneous media; Supervised learning; Temperature;
Conference_Titel :
Circuits and Systems, 1998. IEEE APCCAS 1998. The 1998 IEEE Asia-Pacific Conference on
Conference_Location :
Chiangmai
Print_ISBN :
0-7803-5146-0
DOI :
10.1109/APCCAS.1998.743905
