AN EFFICIENT EM-BASED TRAINING ALGORITHM FOR FEEDFORWARD NEURAL NETWORKS

Authors
MA S JI CY FARMER J
Citation
S. Ma et al., AN EFFICIENT EM-BASED TRAINING ALGORITHM FOR FEEDFORWARD NEURAL NETWORKS, Neural networks, 10(2), 1997, pp. 243-256
Citations number
21
Categorie Soggetti
Mathematical Methods, Biology & Medicine","Computer Sciences, Special Topics","Computer Science Artificial Intelligence",Neurosciences,"Physics, Applied
Journal title
Neural networks → ACNP
ISSN journal
08936080
Volume
10
Issue
2
Year of publication
1997
Pages
243 - 256
Database
ISI
SICI code
0893-6080(1997)10:2<243:AEETAF>2.0.ZU;2-Q
Abstract
A fast training algorithm is developed for two-layer feedforward neura l networks based on a probabilistic model for hidden representations a nd the EM algorithm. The algorithm decomposes training the original tw o-layer networks into training a set of single neurons. The individual neurons are then trained via a linear weighted regression algorithm. Significant improvement on training speed has been made using this alg orithm for several bench-mark problems. (C) 1997 Elsevier Science Ltd. All Rights Reserved.