FAST TRAINING OF RECURRENT NETWORKS BASED ON THE EM ALGORITHM

Authors
MA S JI CY
Citation
S. Ma et Cy. Ji, FAST TRAINING OF RECURRENT NETWORKS BASED ON THE EM ALGORITHM, IEEE transactions on neural networks, 9(1), 1998, pp. 11-26
Citations number
35
Categorie Soggetti
Computer Science Artificial Intelligence","Computer Science Hardware & Architecture","Computer Science Theory & Methods","Computer Science Artificial Intelligence","Computer Science Hardware & Architecture","Computer Science Theory & Methods
Journal title
IEEE transactions on neural networks → ACNP
ISSN journal
10459227
Volume
9
Issue
1
Year of publication
1998
Pages
11 - 26
Database
ISI
SICI code
1045-9227(1998)9:1<11:FTORNB>2.0.ZU;2-B
Abstract
In this work, a probabilistic model is established for recurrent netwo rks, The EM (expectation-maximization) algorithm is then applied to de rive a new fast training algorithm for recurrent networks through mean -field approximation, This new algorithm converts training a complicat ed recurrent network into training an array of individual feedforward neurons, These neurons are then trained via a linear weighted regressi on algorithm. The training time has been improved by five to 15 times on benchmark problems.