LS-based training algorithm for neural networks

Author

Claudio, E. D Di ; Parisi, R. ; Orlandi, G.

Author_Institution

Infocom Dept., Univ. of Rome "La Sapienza", Italy

fYear

1993

fDate

6-9 Sep 1993

Firstpage

22

Lastpage

29

Abstract

A new training algorithm is presented as a faster alternative to the backpropagation (BP) method. The new approach is based on the solution of a linear system at each step of the learning phase. The squared error at the output of each layer before the nonlinearity is minimized on the entire set of the learning patterns by a block least squares (LS) algorithm. The optimal weights for each layer are then computed by using the singular value decomposition (SVD) technique. The simulation results show considerable improvements from the point of view of both accuracy and speed of convergence

Keywords

learning (artificial intelligence); least squares approximations; neural nets; singular value decomposition; LS-based training algorithm; SVD; block least-squares algorithm; convergence speed; neural networks; optimal weights; singular value decomposition; squared error minimization; Backpropagation algorithms; Convergence; Least squares methods; Linear systems; Linearity; Minimization methods; Multilayer perceptrons; Neural networks; Neurons; Singular value decomposition;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks for Processing [1993] III. Proceedings of the 1993 IEEE-SP Workshop

Conference_Location

Linthicum Heights, MD

Print_ISBN

0-7803-0928-6

Type

conf

DOI

10.1109/NNSP.1993.471887

Filename

471887