Nouvel article accepté dans la revue "Turkish Journal of Forecasting"

Multi-layer Perceptron and Pruning


A Multi-Layer Perceptron (MLP) defines a family of artificial neural networks often
used in TS modeling and forecasting. Because of its “black box” aspect, many
researchers refuse to use it. Moreover, the optimization (often based on the exhaustive
approach where “all” configurations are tested) and learning phases of this artificial
intelligence tool (often based on the Levenberg-Marquardt algorithm; LMA) are
weaknesses of this approach (exhaustively and local minima). These two tasks must be
repeated depending on the knowledge of each new problem studied, making the process,
long, laborious and not systematically robust. In this short communication, a pruning
process is presented. This method allows, during the training phase, to carry out an
inputs selecting method activating (or not) inter-nodes connections in order to verify if
forecasting is improved. We propose to use iteratively the popular damped least-squares
method to activate inputs and neurons. A first pass is applied to 10% of the learning
sample to determine weights significantly different from 0 and delete other. Then a
classical batch process based on LMA is used with the new MLP. The validation is done
using 25 measured meteorological TS and cross-comparing the prediction results of the
classical LMA and the 2-stage LMA

https://www.researchgate.net/publication/317740944_Multi-layer_Perceptron_and_Pruning https://www.researchgate.net/publication/317740944_Multi-layer_Perceptron_and_Pruning


Rédigé par Cyril VOYANT le Jeudi 22 Juin 2017 à 09:43 | Lu 27 fois