Explaining predictions from a neural network ensemble one at a time

Robert Wall, Pádraig Cunningham, Paul Walsh

Research output: Chapter in Book/Report/Conference proceedingConference contribution

4 Citations (Scopus)

Abstract

This paper introduces a new method for explaining the predictions of ensembles of neural networks on a case by case basis. The approach of explaining individual examples differs from much of the current research which focuses on producing a global model of the phenomenon under investigation. Explaining individual results is accomplished by modelling each of the networks as a rule-set and computing the resulting coverage statistics for each rule given the data used to train the network. This coverage information is then used to choose the rule or rules that best describe the example under investigation. This approach is based on the premise that ensembles perform an implicit problem space decomposition with ensemble members specialising in different regions of the problem space. Thus explaining an ensemble involves explaining the ensemble members that best fit the example.

Original languageEnglish (US)
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages449-460
Number of pages12
Volume2431 LNAI
StatePublished - 2002
Externally publishedYes
Event6th European Conference on Principles and Practice of Knowledge Discovery in Databases, PKDD 2002 - Helsinki, Finland
Duration: Aug 19 2002Aug 23 2002

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume2431 LNAI
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other6th European Conference on Principles and Practice of Knowledge Discovery in Databases, PKDD 2002
CountryFinland
CityHelsinki
Period8/19/028/23/02

Fingerprint

Neural Network Ensemble
Ensemble
Statistics
Neural networks
Decomposition
Prediction
Coverage
Choose
Neural Networks
Decompose
Computing
Modeling

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Wall, R., Cunningham, P., & Walsh, P. (2002). Explaining predictions from a neural network ensemble one at a time. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 2431 LNAI, pp. 449-460). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 2431 LNAI).

Explaining predictions from a neural network ensemble one at a time. / Wall, Robert; Cunningham, Pádraig; Walsh, Paul.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 2431 LNAI 2002. p. 449-460 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 2431 LNAI).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Wall, R, Cunningham, P & Walsh, P 2002, Explaining predictions from a neural network ensemble one at a time. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 2431 LNAI, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 2431 LNAI, pp. 449-460, 6th European Conference on Principles and Practice of Knowledge Discovery in Databases, PKDD 2002, Helsinki, Finland, 8/19/02.
Wall R, Cunningham P, Walsh P. Explaining predictions from a neural network ensemble one at a time. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 2431 LNAI. 2002. p. 449-460. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
Wall, Robert ; Cunningham, Pádraig ; Walsh, Paul. / Explaining predictions from a neural network ensemble one at a time. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 2431 LNAI 2002. pp. 449-460 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{331459826edf4a64885072eed1033173,
title = "Explaining predictions from a neural network ensemble one at a time",
abstract = "This paper introduces a new method for explaining the predictions of ensembles of neural networks on a case by case basis. The approach of explaining individual examples differs from much of the current research which focuses on producing a global model of the phenomenon under investigation. Explaining individual results is accomplished by modelling each of the networks as a rule-set and computing the resulting coverage statistics for each rule given the data used to train the network. This coverage information is then used to choose the rule or rules that best describe the example under investigation. This approach is based on the premise that ensembles perform an implicit problem space decomposition with ensemble members specialising in different regions of the problem space. Thus explaining an ensemble involves explaining the ensemble members that best fit the example.",
author = "Robert Wall and P{\'a}draig Cunningham and Paul Walsh",
year = "2002",
language = "English (US)",
isbn = "3540440372",
volume = "2431 LNAI",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "449--460",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - Explaining predictions from a neural network ensemble one at a time

AU - Wall, Robert

AU - Cunningham, Pádraig

AU - Walsh, Paul

PY - 2002

Y1 - 2002

N2 - This paper introduces a new method for explaining the predictions of ensembles of neural networks on a case by case basis. The approach of explaining individual examples differs from much of the current research which focuses on producing a global model of the phenomenon under investigation. Explaining individual results is accomplished by modelling each of the networks as a rule-set and computing the resulting coverage statistics for each rule given the data used to train the network. This coverage information is then used to choose the rule or rules that best describe the example under investigation. This approach is based on the premise that ensembles perform an implicit problem space decomposition with ensemble members specialising in different regions of the problem space. Thus explaining an ensemble involves explaining the ensemble members that best fit the example.

AB - This paper introduces a new method for explaining the predictions of ensembles of neural networks on a case by case basis. The approach of explaining individual examples differs from much of the current research which focuses on producing a global model of the phenomenon under investigation. Explaining individual results is accomplished by modelling each of the networks as a rule-set and computing the resulting coverage statistics for each rule given the data used to train the network. This coverage information is then used to choose the rule or rules that best describe the example under investigation. This approach is based on the premise that ensembles perform an implicit problem space decomposition with ensemble members specialising in different regions of the problem space. Thus explaining an ensemble involves explaining the ensemble members that best fit the example.

UR - http://www.scopus.com/inward/record.url?scp=84864834782&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84864834782&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:84864834782

SN - 3540440372

SN - 9783540440376

VL - 2431 LNAI

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

SP - 449

EP - 460

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

ER -