Technical Communique: Modified AIC rule for model selection in combination with prior estimated noise models

Authors:
J. Schoukens;Y. Rolain;R. Pintelon
Affiliations:
Department ELEC, Vrije Universiteit Brussel, Pleinlaan 2, 1050 Brussels, Belgium;Department ELEC, Vrije Universiteit Brussel, Pleinlaan 2, 1050 Brussels, Belgium;Department ELEC, Vrije Universiteit Brussel, Pleinlaan 2, 1050 Brussels, Belgium
Venue:
Automatica (Journal of IFAC)
Year:
2002

Citing 2
Cited 2

Frequency-domain system identification using non-parametric noise models estimated from a small number of data sets

Automatica (Journal of IFAC)
Paper: Modeling by shortest data description

Automatica (Journal of IFAC)

Uncertainty of transfer function modelling using prior estimated noise models

Automatica (Journal of IFAC)
Identification of linear systems with nonlinear distortions

Automatica (Journal of IFAC)

Quantified Score

Hi-index	22.15

Visualization

Abstract

In this paper we adapt the AIC model selection rule to be used in combination with prior estimated noise models in the presence of model errors. It is shown that under these conditions a slightly modified cost function extension is needed, resulting in a multiplicative complexity term instead of an additive one as would be found for fixed noise models and no model errors. The equivalency of this result with the classical prediction error framework is discussed.