Parallel and separable recursive Levenberg-Marquardt training algorithm

Author

Asirvadam, V.S. ; McLoone, S.F. ; Irwin, G.W.

Author_Institution

Sch. of Electr. & Electron. Eng., Queen´´s Univ., Belfast, UK

fYear

2002

fDate

2002

Firstpage

129

Lastpage

138

Abstract

A novel decomposed recursive Levenberg Marquardt (RLM) algorithm is derived for the training of feedforward neural networks. By neglecting interneuron weight correlations the recently proposed RLM training algorithm can be decomposed at neuron level enabling weights to be updated in an efficient parallel manner. A separable least squares implementation of decomposed RLM is also introduced. Experiment results for two nonlinear time series problems demonstrate the superiority of the new training algorithms.

Keywords

feedforward neural nets; learning (artificial intelligence); least squares approximations; nonlinear estimation; parallel algorithms; recursive estimation; time series; Levenberg-Marquardt training algorithm; decomposed RLM; feedforward neural networks; neuron level; nonlinear time series problems; parallel algorithm; separable least squares implementation; separable recursive training algorithm; weight updating; Backpropagation algorithms; Convergence; Cost function; Feedforward neural networks; Least squares methods; Neural networks; Neurons; Partitioning algorithms; Resonance light scattering; Training data;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks for Signal Processing, 2002. Proceedings of the 2002 12th IEEE Workshop on

Print_ISBN

0-7803-7616-1

Type

conf

DOI

10.1109/NNSP.2002.1030024

Filename

1030024