Predictive minimum description length criterion for time series modeling with neural networks

Authors:
Mikko Lehtokangas;Jukka Saarinen;Pentti Huuhtanen;Kimmo Kaski
Affiliations:
Tampere University of Technology, Microelectronics Laboratory, P.O. Box 692, FIN-33101 Tampere, Finland;Tampere University of Technology, Microelectronics Laboratory, P.O. Box 692, FIN-33101 Tampere, Finland;University of Tampere, Department of Mathematical Sciences, P.O. Box 607, FIN-33101 Tampere, Finland;Tampere University of Technology, Microelectronics Laboratory, P.O. Box 692, FIN-33101 Tampere, Finland
Venue:
Neural Computation
Year:
1996

Citing 0
Cited 2

Prequential and Cross-Validated Regression Estimation

Machine Learning
A novel approach for estimation of optimal embedding parameters of nonlinear time series by structural learning of neural network

Neurocomputing

Quantified Score

Hi-index	0.00

Visualization

Abstract

Nonlinear time series modeling with a multilayer perceptron network is presented. An important aspect of this modeling is the model selection, i.e., the problem of determining the size as well as the complexity of the model. To overcome this problem we apply the predictive minimum description length (PMDL) principle as a minimization criterion. In the neural network scheme it means minimizing the number of input and hidden units. Three time series modeling experiments are used to examine the usefulness of the PMDL model selection scheme. A comparison with the widely used cross-validation technique is also presented. In our experiments the PMDL scheme and the cross-validation scheme yield similar results in terms of model complexity. However, the PMDL method was found to be two times faster to compute. This is significant improvement since model selection in general is very time consuming.