A generalization error estimate for nonlinear systems

Author

Larsen, Jan

Author_Institution

Tech. Univ. of Denmark, Lyngby, Denmark

fYear

1992

fDate

31 Aug-2 Sep 1992

Firstpage

29

Lastpage

38

Abstract

A new estimate (GEN) of the generalization error is presented. The estimator is valid for both incomplete and nonlinear models. An incomplete model is characterized in that it does not model the actual nonlinear relationship perfectly. The GEN estimator has been evaluated by simulating incomplete models of linear and simple neural network systems. Within the linear system GEN is compared to the final prediction error criterion and the leave-one-out cross-validation technique. It was found that the GEN estimate of the true generalization error is less biased on the average. It is concluded that GEN is an applicable alternative in estimating the generalization at the expense of an increased complexity

Keywords

generalisation (artificial intelligence); neural nets; nonlinear systems; parameter estimation; final prediction error criterion; generalization error estimate; incomplete model; leave-one-out cross-validation technique; linear system; neural network systems; nonlinear models; nonlinear systems; parameter estimation; Buildings; Computer architecture; Computer networks; Cost function; Neural networks; Noise generators; Nonlinear systems; Parameter estimation; Signal processing; Testing;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks for Signal Processing [1992] II., Proceedings of the 1992 IEEE-SP Workshop

Conference_Location

Helsingoer

Print_ISBN

0-7803-0557-4

Type

conf

DOI

10.1109/NNSP.1992.253710

Filename

253710