Sample complexity for learning recurrent perceptron mappings

Author

DasGupta, Bhaskar ; Sontag, Eduardo D.

Author_Institution

Dept. of Comput. Sci., Waterloo Univ., Ont., Canada

Volume

42

Issue

5

fYear

1996

fDate

9/1/1996 12:00:00 AM

Firstpage

1479

Lastpage

1487

Abstract

Recurrent perceptron classifiers generalize the usual perceptron model. They correspond to linear transformations of input vectors obtained by means of “autoregressive moving-average schemes”, or infinite impulse response filters, and take into account those correlations and dependences among input coordinates which arise from linear digital filtering. This paper provides tight bounds on the sample complexity associated to the fitting of such models to experimental data. The results are expressed in the context of the theory of probably approximately correct (PAC) learning

Keywords

IIR filters; autoregressive moving average processes; correlation methods; digital filters; filtering theory; learning (artificial intelligence); multilayer perceptrons; pattern classification; recurrent neural nets; signal sampling; autoregressive moving-average; correlations; experimental data; infinite impulse response filters; input coordinates; input vectors; linear digital filtering; linear transformations; perceptron model; probably approximately correct learning; recurrent perceptron classifiers; recurrent perceptron mappings; sample complexity; tight bounds; Digital filters; Filtering; IIR filters; Information processing; Input variables; Linear programming; Neural networks; Nonlinear filters; Recurrent neural networks; Vectors;

fLanguage

English

Journal_Title

Information Theory, IEEE Transactions on

Publisher

ieee

ISSN

0018-9448

Type

jour

DOI

10.1109/18.532888

Filename

532888