Publication detail

The Algorithm for Choice of Samples in Training Set for Neural Networks

VYCHODIL, H., PIVOŇKA, P., KRUPANSKÝ, P.

Original Title

The Algorithm for Choice of Samples in Training Set for Neural Networks

Type

conference paper

Language

English

Original Abstract

The paper deals with on-line system identification for adaptive controller construction. A regressive model is used for system modelling. The regressive function is realized in neural network. The problem of an increasing number of data for neural network learning is solved by the choice of patterns in the training set based on the distance between patterns in the space of regressive model state vectors. Furthermore, the choice of scale to set this distance is considered. The objective is minimization of empirical approaches to allow for automation and wide application.

Keywords

Neural nets, model NARMAX, choice of patterns

Authors

VYCHODIL, H., PIVOŇKA, P., KRUPANSKÝ, P.

RIV year

2002

Released

4. 9. 2002

Publisher

Rektor der Hochschule Zittau/Görlitz

Location

Zittau, Německo

ISBN

3-9808089-2-3

Book

Proceedings East West Fuzzy Colloquium 2002

Pages from

224

Pages to

229

Pages count

6

BibTex

@inproceedings{BUT4917,
  author="Hynek {Vychodil} and Petr {Pivoňka} and Petr {Krupanský}",
  title="The Algorithm for Choice of Samples in Training Set for Neural Networks",
  booktitle="Proceedings East West Fuzzy Colloquium 2002",
  year="2002",
  pages="6",
  publisher="Rektor der Hochschule Zittau/Görlitz",
  address="Zittau, Německo",
  isbn="3-9808089-2-3"
}