A UNIFIED APPROACH ON FAST TRAINING OF FEEDFORWARD AND RECURRENT NETWORKS USING EM ALGORITHM

Authors
MA S JI CY
Citation
S. Ma et Cy. Ji, A UNIFIED APPROACH ON FAST TRAINING OF FEEDFORWARD AND RECURRENT NETWORKS USING EM ALGORITHM, IEEE transactions on signal processing, 46(8), 1998, pp. 2270-2274
Citations number
11
Categorie Soggetti
Engineering, Eletrical & Electronic
Journal title
IEEE transactions on signal processing → ACNP
ISSN journal
1053587X
Volume
46
Issue
8
Year of publication
1998
Pages
2270 - 2274
Database
ISI
SICI code
1053-587X(1998)46:8<2270:AUAOFT>2.0.ZU;2-J
Abstract
In this work, we provide a theoretical framework that unifies the noti ons of hidden representations and moving targets through the expectati on and maximization (EM) algorithm. Based on such a framework, two fas t training algorithms can be derived consistently for both feedforward networks and recurrent networks.