A Comparative Study of Back-propagation Algorithms: Levenberg-Marquardt and BFGS for the Formation of Multilayer Neural Networks for Estimation of Fluoride

R. El Chaal, M. O. Aboutafail

2021

Abstract

This work describes a comparative approach between the twoback-propagation algorithms: the Levenberg-Marquardt (LM) and the Broyden Fletcher Goldfarb Shanno (BFGS); For the prediction of Fluoride in the Inaouène basin, using artificial neural networks (ANN) of the multilayer perceptron type(MLP), from the concentrations of sixteen physicochemical parameters. We have developed several models based on the evolution of combinations of activation functions andneurons number in the hidden layer.The performance of the different ANN model training algorithms was evaluated using the mean square error (MSE) and the correlation coefficient (R). The evaluation shows that the LM training algorithms perform better than the BFGS training algorithm. The results obtained demonstrate the efficiency of the LM algorithm for the prediction of Fluoride compared to the BFGS algorithm by MLP type neural networks, as shown by the statistical indicators ({R = 0.99 and MSE = 0.135 for LM} and {R = 0.95 and MSE = 41.22 for BFGS}).

Paper Citation

in Harvard Style

El Chaal R. and Aboutafail M. (2021). A Comparative Study of Back-propagation Algorithms: Levenberg-Marquardt and BFGS for the Formation of Multilayer Neural Networks for Estimation of Fluoride. In Proceedings of the 2nd International Conference on Big Data, Modelling and Machine Learning - Volume 1: BML, ISBN 978-989-758-559-3, pages 558-565. DOI: 10.5220/0010746300003101

in Bibtex Style

@conference{bml21,
author={R. El Chaal and M. O. Aboutafail},
title={A Comparative Study of Back-propagation Algorithms: Levenberg-Marquardt and BFGS for the Formation of Multilayer Neural Networks for Estimation of Fluoride},
booktitle={Proceedings of the 2nd International Conference on Big Data, Modelling and Machine Learning - Volume 1: BML,},
year={2021},
pages={558-565},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0010746300003101},
isbn={978-989-758-559-3},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 2nd International Conference on Big Data, Modelling and Machine Learning - Volume 1: BML,
TI - A Comparative Study of Back-propagation Algorithms: Levenberg-Marquardt and BFGS for the Formation of Multilayer Neural Networks for Estimation of Fluoride
SN - 978-989-758-559-3
AU - El Chaal R.
AU - Aboutafail M.
PY - 2021
SP - 558
EP - 565
DO - 10.5220/0010746300003101