DocumentCode
2400384
Title
Training recurrent networks
Author
Pedersen, Morten With
Author_Institution
Dept. of Math. Modelling, Tech. Univ. Lyngby, Denmark
fYear
1997
fDate
24-26 Sep 1997
Firstpage
355
Lastpage
364
Abstract
Training recurrent networks is generally believed to be a difficult task. Excessive training times and lack of convergence to an acceptable solution are frequently reported. In this paper we seek to explain the reason for this from a numerical point of view and show how to avoid problems when training. In particular we investigate ill-conditioning, the need for and effect of regularization and illustrate the superiority of second-order methods for training
Keywords
learning (artificial intelligence); recurrent neural nets; ill-conditioning; recurrent neural network training; regularization; second-order methods; Computer networks; Iron; Laser feedback; Least squares methods; Mathematical model; Newton method; Output feedback; Recurrent neural networks; Recursive estimation; Stability;
fLanguage
English
Publisher
ieee
Conference_Titel
Neural Networks for Signal Processing [1997] VII. Proceedings of the 1997 IEEE Workshop
Conference_Location
Amelia Island, FL
ISSN
1089-3555
Print_ISBN
0-7803-4256-9
Type
conf
DOI
10.1109/NNSP.1997.622416
Filename
622416
Link To Document