The analysis of artificial neural network data models

C. M. Roadknight; D. Palmer-Brown; G. E. Mills

doi:10.1007/bfb0052852

The analysis of artificial neural network data models

C. M. Roadknight, D. Palmer-Brown, G. E. Mills

Research output: Chapter in Book/Conference proceeding › Conference contribution › peer-review

1 Citation (Scopus)

Abstract

Artificial neural networks are good non-linear function approximators but their multi-layer, non-linear form gives little immediate indication of the features they have learnt. Several methods are put forward in this paper that reduce the complexity of the network or give simplified equations that are easier to interpret. Relative weight analysis and equation synthesis are summarised while correlated activity pruning is introduced and explained in detail. The former techniques use the weights of a trained network to assign importance to inputs or groups of inputs. The latter algorithm reduces complexity of a network by merging hidden units that have correlated activations. This procedure also allows the relationship between detected features to be evaluated. Data from pollutant impact studies are used but the techniques developed are applicable to many scientific data modelling environments.

Original language	English
Title of host publication	Advances in Intelligent Data Analysis
Subtitle of host publication	Reasoning about Data - 2nd International Symposium, IDA-1997, Proceedings
Editors	Xiaohui Liu, Paul Cohen, Michael Berthold
Publisher	Springer Verlag
Pages	337-346
Number of pages	10
ISBN (Print)	9783540633464
DOIs	https://doi.org/10.1007/bfb0052852
Publication status	Published - 1997
Externally published	Yes
Event	2nd International Symposium on Intelligent Data Analysis, IDA 1997 - London, United Kingdom Duration: 4 Aug 1997 → 6 Aug 1997

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	1280
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	2nd International Symposium on Intelligent Data Analysis, IDA 1997
Country/Territory	United Kingdom
City	London
Period	4/08/97 → 6/08/97

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1007/bfb0052852

Cite this

Roadknight, C. M., Palmer-Brown, D., & Mills, G. E. (1997). The analysis of artificial neural network data models. In X. Liu, P. Cohen, & M. Berthold (Eds.), Advances in Intelligent Data Analysis: Reasoning about Data - 2nd International Symposium, IDA-1997, Proceedings (pp. 337-346). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1280). Springer Verlag. https://doi.org/10.1007/bfb0052852

Roadknight, C. M. ; Palmer-Brown, D. ; Mills, G. E. / The analysis of artificial neural network data models. Advances in Intelligent Data Analysis: Reasoning about Data - 2nd International Symposium, IDA-1997, Proceedings. editor / Xiaohui Liu ; Paul Cohen ; Michael Berthold. Springer Verlag, 1997. pp. 337-346 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{740f5e1f607841148f33d560241ae0a7,

title = "The analysis of artificial neural network data models",

abstract = "Artificial neural networks are good non-linear function approximators but their multi-layer, non-linear form gives little immediate indication of the features they have learnt. Several methods are put forward in this paper that reduce the complexity of the network or give simplified equations that are easier to interpret. Relative weight analysis and equation synthesis are summarised while correlated activity pruning is introduced and explained in detail. The former techniques use the weights of a trained network to assign importance to inputs or groups of inputs. The latter algorithm reduces complexity of a network by merging hidden units that have correlated activations. This procedure also allows the relationship between detected features to be evaluated. Data from pollutant impact studies are used but the techniques developed are applicable to many scientific data modelling environments.",

author = "Roadknight, {C. M.} and D. Palmer-Brown and Mills, {G. E.}",

note = "Publisher Copyright: {\textcopyright} Springer-Verlag Berlin Heidelberg 1997.; 2nd International Symposium on Intelligent Data Analysis, IDA 1997 ; Conference date: 04-08-1997 Through 06-08-1997",

year = "1997",

doi = "10.1007/bfb0052852",

language = "English",

isbn = "9783540633464",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "337--346",

editor = "Xiaohui Liu and Paul Cohen and Michael Berthold",

booktitle = "Advances in Intelligent Data Analysis",

address = "Germany",

}

Roadknight, CM, Palmer-Brown, D & Mills, GE 1997, The analysis of artificial neural network data models. in X Liu, P Cohen & M Berthold (eds), Advances in Intelligent Data Analysis: Reasoning about Data - 2nd International Symposium, IDA-1997, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 1280, Springer Verlag, pp. 337-346, 2nd International Symposium on Intelligent Data Analysis, IDA 1997, London, United Kingdom, 4/08/97. https://doi.org/10.1007/bfb0052852

The analysis of artificial neural network data models. / Roadknight, C. M.; Palmer-Brown, D.; Mills, G. E.
Advances in Intelligent Data Analysis: Reasoning about Data - 2nd International Symposium, IDA-1997, Proceedings. ed. / Xiaohui Liu; Paul Cohen; Michael Berthold. Springer Verlag, 1997. p. 337-346 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1280).

Research output: Chapter in Book/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - The analysis of artificial neural network data models

AU - Roadknight, C. M.

AU - Palmer-Brown, D.

AU - Mills, G. E.

N1 - Publisher Copyright: © Springer-Verlag Berlin Heidelberg 1997.

PY - 1997

Y1 - 1997

N2 - Artificial neural networks are good non-linear function approximators but their multi-layer, non-linear form gives little immediate indication of the features they have learnt. Several methods are put forward in this paper that reduce the complexity of the network or give simplified equations that are easier to interpret. Relative weight analysis and equation synthesis are summarised while correlated activity pruning is introduced and explained in detail. The former techniques use the weights of a trained network to assign importance to inputs or groups of inputs. The latter algorithm reduces complexity of a network by merging hidden units that have correlated activations. This procedure also allows the relationship between detected features to be evaluated. Data from pollutant impact studies are used but the techniques developed are applicable to many scientific data modelling environments.

AB - Artificial neural networks are good non-linear function approximators but their multi-layer, non-linear form gives little immediate indication of the features they have learnt. Several methods are put forward in this paper that reduce the complexity of the network or give simplified equations that are easier to interpret. Relative weight analysis and equation synthesis are summarised while correlated activity pruning is introduced and explained in detail. The former techniques use the weights of a trained network to assign importance to inputs or groups of inputs. The latter algorithm reduces complexity of a network by merging hidden units that have correlated activations. This procedure also allows the relationship between detected features to be evaluated. Data from pollutant impact studies are used but the techniques developed are applicable to many scientific data modelling environments.

UR - http://www.scopus.com/inward/record.url?scp=84949801873&partnerID=8YFLogxK

U2 - 10.1007/bfb0052852

DO - 10.1007/bfb0052852

M3 - Conference contribution

AN - SCOPUS:84949801873

SN - 9783540633464

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 337

EP - 346

BT - Advances in Intelligent Data Analysis

A2 - Liu, Xiaohui

A2 - Cohen, Paul

A2 - Berthold, Michael

PB - Springer Verlag

T2 - 2nd International Symposium on Intelligent Data Analysis, IDA 1997

Y2 - 4 August 1997 through 6 August 1997

ER -

Roadknight CM, Palmer-Brown D, Mills GE. The analysis of artificial neural network data models. In Liu X, Cohen P, Berthold M, editors, Advances in Intelligent Data Analysis: Reasoning about Data - 2nd International Symposium, IDA-1997, Proceedings. Springer Verlag. 1997. p. 337-346. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/bfb0052852

The analysis of artificial neural network data models

Abstract

Publication series

Conference

ASJC Scopus subject areas

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this