Superior training of artificial neural networks using weight-space partitioning

Author

Gupta, Hoshin V. ; Hsu, Kuo-lin ; Sorooshian, Soroosh

Author_Institution

Dept. of Hydrology & Water Resources, Arizona Univ., Tucson, AZ, USA

Volume

3

fYear

1997

fDate

9-12 Jun 1997

Firstpage

1919

Abstract

Linear least squares simplex (LLSSIM) is a new algorithm for batch training of three-layer feedforward artificial neural networks (ANN), based on a partitioning of the weight space. The input-hidden weights are trained using a “multi-start downhill simplex” global search algorithm, and the hidden-output weights are estimated using “conditional linear least squares”. Monte-Carlo testing shows that LLSSIM provides globally superior weight estimates with significantly fewer function evaluations than the conventional backpropagation, adaptive backpropagation, and conjugate gradient strategies

Keywords

feedforward neural nets; learning (artificial intelligence); least squares approximations; optimisation; search problems; batch training; conditional linear least squares; feedforward neural networks; global search algorithm; input-hidden weights; linear least squares simplex; weight-space partitioning; Artificial neural networks; Backpropagation algorithms; Convergence; Joining processes; Least squares methods; Logistics; Neurons; Partitioning algorithms; Testing; Transfer functions;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks,1997., International Conference on

Conference_Location

Houston, TX

Print_ISBN

0-7803-4122-8

Type

conf

DOI

10.1109/ICNN.1997.614192

Filename

614192