ADAPTIVE RETRAINING ALGORITHM WITH SHAKEN INITIALIZATION

The paper presents new specific aspects that could improve the adaptive retraining procedure of artificial neural networks (ANNs) for time series predictions. Usually, a retraining step starts from proportionally reduced values of the parameters (weights) used in the previous version of the ANN model. This time, the initial configuration of the weights is randomly “shaken” in order to further improve the model. The present results are promising and show a better adaptation of the forecasting system in a nonstationary environment.

[1]  Paul Dan Cristea,et al.  Signal representation and processing of nucleotide sequences , 2008, Int. J. Funct. Informatics Pers. Medicine.

[2]  P. Cristea,et al.  ANN Flexible Forecasting for the Adaptive Monitoring of a Multi-Tube Reactor , 2007, 2007 14th International Workshop on Systems, Signals and Image Processing and 6th EURASIP Conference focused on Speech and Image Processing, Multimedia Communications and Services.

[3]  Rui J. P. de Figueiredo.  A NEURAL-NETWORK-BASED APPROACH TO SPEECH SIGNAL PREDICTION , 2010 .

[4]  Martin Fodslette Møller,et al.  A scaled conjugate gradient algorithm for fast supervised learning , 1993, Neural Networks.

[5]  Paul Dan Cristea,et al.  An ANN - PCA adaptive forecasting model , 2012, 2012 19th International Conference on Systems, Signals and Image Processing (IWSSIP).

[6]  I. Jolliffe.  Principal Component Analysis , 2002 .

[7]  Dumitru Iulian Nastac,et al.  Neuro-Adaptive Model for Financial Forecasting , 2007 .

[8]  Adriana Florescu,et al.  COMPARISON AMONG COMPUTATIONAL INTELLIGENCE METHODS FOR ENGINE KNOCK DETECTION. PART 1 , 2011 .