Visualization and complexity reduction of neural networks

Tamas Kenesei, Balazs Feil, J. Abonyi

Research output: Chapter in Book/Report/Conference proceedingChapter

1 Citation (Scopus)

Abstract

The identification of the proper structure of nonlinear neural networks (NNs) is a difficult problem, since these black-box models are not interpretable. The aim of the paper is to propose a new approach that can be used for the analysis and the reduction of these models. It is shown that NNs with sigmoid transfer function can be transformed into fuzzy systems. Hence, with the use of this transformation NNs can be analyzed by human experts based on the extracted linguistic rules. Moreover, based on the similarity of the resulted membership functions the hidden neurons of the NNs can be mapped into a two dimensional space. The resulted map provides an easily interpretable figure about the redundancy of the neurons. Furthermore, the contribution of these neurons can be measured by orthogonal least squares technique that can be used for the ordering of the extracted fuzzy rules based on their importance. A practical example related to the dynamic modeling of a chemical process system is used to prove that synergistic combination of model transformation, visualization and reduction of NNs is an effective technique, that can be used for the structural and parametrical analysis of NNs.

Original languageEnglish
Title of host publicationAdvances in Soft Computing
Pages43-52
Number of pages10
Volume52
DOIs
Publication statusPublished - 2009

Publication series

NameAdvances in Soft Computing
Volume52
ISSN (Print)16153871
ISSN (Electronic)18600794

Fingerprint

Visualization
Neural networks
Neurons
Fuzzy rules
Fuzzy systems
Membership functions
Linguistics
Transfer functions
Redundancy

ASJC Scopus subject areas

  • Computational Mechanics
  • Computer Science Applications
  • Computer Science (miscellaneous)

Cite this

Kenesei, T., Feil, B., & Abonyi, J. (2009). Visualization and complexity reduction of neural networks. In Advances in Soft Computing (Vol. 52, pp. 43-52). (Advances in Soft Computing; Vol. 52). https://doi.org/10.1007/978-3-540-88079-0_5

Visualization and complexity reduction of neural networks. / Kenesei, Tamas; Feil, Balazs; Abonyi, J.

Advances in Soft Computing. Vol. 52 2009. p. 43-52 (Advances in Soft Computing; Vol. 52).

Research output: Chapter in Book/Report/Conference proceedingChapter

Kenesei, T, Feil, B & Abonyi, J 2009, Visualization and complexity reduction of neural networks. in Advances in Soft Computing. vol. 52, Advances in Soft Computing, vol. 52, pp. 43-52. https://doi.org/10.1007/978-3-540-88079-0_5
Kenesei T, Feil B, Abonyi J. Visualization and complexity reduction of neural networks. In Advances in Soft Computing. Vol. 52. 2009. p. 43-52. (Advances in Soft Computing). https://doi.org/10.1007/978-3-540-88079-0_5
Kenesei, Tamas ; Feil, Balazs ; Abonyi, J. / Visualization and complexity reduction of neural networks. Advances in Soft Computing. Vol. 52 2009. pp. 43-52 (Advances in Soft Computing).
@inbook{ca89cbcb79ac48c6af886befdd11f4cf,
title = "Visualization and complexity reduction of neural networks",
abstract = "The identification of the proper structure of nonlinear neural networks (NNs) is a difficult problem, since these black-box models are not interpretable. The aim of the paper is to propose a new approach that can be used for the analysis and the reduction of these models. It is shown that NNs with sigmoid transfer function can be transformed into fuzzy systems. Hence, with the use of this transformation NNs can be analyzed by human experts based on the extracted linguistic rules. Moreover, based on the similarity of the resulted membership functions the hidden neurons of the NNs can be mapped into a two dimensional space. The resulted map provides an easily interpretable figure about the redundancy of the neurons. Furthermore, the contribution of these neurons can be measured by orthogonal least squares technique that can be used for the ordering of the extracted fuzzy rules based on their importance. A practical example related to the dynamic modeling of a chemical process system is used to prove that synergistic combination of model transformation, visualization and reduction of NNs is an effective technique, that can be used for the structural and parametrical analysis of NNs.",
author = "Tamas Kenesei and Balazs Feil and J. Abonyi",
year = "2009",
doi = "10.1007/978-3-540-88079-0_5",
language = "English",
isbn = "9783540880783",
volume = "52",
series = "Advances in Soft Computing",
pages = "43--52",
booktitle = "Advances in Soft Computing",

}

TY - CHAP

T1 - Visualization and complexity reduction of neural networks

AU - Kenesei, Tamas

AU - Feil, Balazs

AU - Abonyi, J.

PY - 2009

Y1 - 2009

N2 - The identification of the proper structure of nonlinear neural networks (NNs) is a difficult problem, since these black-box models are not interpretable. The aim of the paper is to propose a new approach that can be used for the analysis and the reduction of these models. It is shown that NNs with sigmoid transfer function can be transformed into fuzzy systems. Hence, with the use of this transformation NNs can be analyzed by human experts based on the extracted linguistic rules. Moreover, based on the similarity of the resulted membership functions the hidden neurons of the NNs can be mapped into a two dimensional space. The resulted map provides an easily interpretable figure about the redundancy of the neurons. Furthermore, the contribution of these neurons can be measured by orthogonal least squares technique that can be used for the ordering of the extracted fuzzy rules based on their importance. A practical example related to the dynamic modeling of a chemical process system is used to prove that synergistic combination of model transformation, visualization and reduction of NNs is an effective technique, that can be used for the structural and parametrical analysis of NNs.

AB - The identification of the proper structure of nonlinear neural networks (NNs) is a difficult problem, since these black-box models are not interpretable. The aim of the paper is to propose a new approach that can be used for the analysis and the reduction of these models. It is shown that NNs with sigmoid transfer function can be transformed into fuzzy systems. Hence, with the use of this transformation NNs can be analyzed by human experts based on the extracted linguistic rules. Moreover, based on the similarity of the resulted membership functions the hidden neurons of the NNs can be mapped into a two dimensional space. The resulted map provides an easily interpretable figure about the redundancy of the neurons. Furthermore, the contribution of these neurons can be measured by orthogonal least squares technique that can be used for the ordering of the extracted fuzzy rules based on their importance. A practical example related to the dynamic modeling of a chemical process system is used to prove that synergistic combination of model transformation, visualization and reduction of NNs is an effective technique, that can be used for the structural and parametrical analysis of NNs.

UR - http://www.scopus.com/inward/record.url?scp=59549104662&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=59549104662&partnerID=8YFLogxK

U2 - 10.1007/978-3-540-88079-0_5

DO - 10.1007/978-3-540-88079-0_5

M3 - Chapter

AN - SCOPUS:59549104662

SN - 9783540880783

VL - 52

T3 - Advances in Soft Computing

SP - 43

EP - 52

BT - Advances in Soft Computing

ER -