Improved Interpretation of Feature Relevances: Iterated Relevance Matrix Analysis (IRMA)

Sofie Lövdal; Michael Biehl

Improved Interpretation of Feature Relevances: Iterated Relevance Matrix Analysis (IRMA)

Sofie Lövdal^*, Michael Biehl

^*Corresponding author for this work

Intelligent Systems

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

23 Downloads (Pure)

Abstract

We introduce and investigate the iterated application of Gen-
eralized Matrix Relevance Learning for the analysis of feature relevances in
classification problems. The suggested Iterated Relevance Matrix Analysis
(IRMA), identifies a linear subspace representing the classification specific
information of the considered data sets in feature space using General-
ized Matrix Learning Vector Quantization. By iteratively determining a
new discriminative direction while projecting out all previously identified
ones, all features carrying relevant information about the classification can
be found, facilitating a detailed analysis of feature relevances. Moreover,
IRMA can be used to generate improved low-dimensional representations
and visualizations of labeled data sets.

Original language	English
Title of host publication	ESANN 2023 Proceedings
Subtitle of host publication	European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning
Editors	Michel Verleysen
Publisher	i6doc.com publication
Pages	59-64
Number of pages	6
ISBN (Print)	9782875870889
Publication status	Published - 1-Oct-2023

Access to Document

Improved Interpretation of Feature Relevances: Iterated Relevance Matrix Analysis (IRMA)Final publisher's version, 1.55 MBLicence: Taverne

Handle.net

Persistent link

Cite this

@inproceedings{569846204c48470aaf056c26bf1eb560,

title = "Improved Interpretation of Feature Relevances: Iterated Relevance Matrix Analysis (IRMA)",

abstract = "We introduce and investigate the iterated application of Gen-eralized Matrix Relevance Learning for the analysis of feature relevances inclassification problems. The suggested Iterated Relevance Matrix Analysis(IRMA), identifies a linear subspace representing the classification specificinformation of the considered data sets in feature space using General-ized Matrix Learning Vector Quantization. By iteratively determining anew discriminative direction while projecting out all previously identifiedones, all features carrying relevant information about the classification canbe found, facilitating a detailed analysis of feature relevances. Moreover,IRMA can be used to generate improved low-dimensional representationsand visualizations of labeled data sets.",

author = "Sofie L{\"o}vdal and Michael Biehl",

year = "2023",

month = oct,

day = "1",

language = "English",

isbn = "9782875870889",

pages = "59--64",

editor = "Michel Verleysen",

booktitle = "ESANN 2023 Proceedings",

publisher = "i6doc.com publication",

}

Improved Interpretation of Feature Relevances: Iterated Relevance Matrix Analysis (IRMA). / Lövdal, Sofie ; Biehl, Michael.
ESANN 2023 Proceedings: European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning. ed. / Michel Verleysen. i6doc.com publication, 2023. p. 59-64.

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

TY - GEN

T1 - Improved Interpretation of Feature Relevances

T2 - Iterated Relevance Matrix Analysis (IRMA)

AU - Lövdal, Sofie

AU - Biehl, Michael

PY - 2023/10/1

Y1 - 2023/10/1

N2 - We introduce and investigate the iterated application of Gen-eralized Matrix Relevance Learning for the analysis of feature relevances inclassification problems. The suggested Iterated Relevance Matrix Analysis(IRMA), identifies a linear subspace representing the classification specificinformation of the considered data sets in feature space using General-ized Matrix Learning Vector Quantization. By iteratively determining anew discriminative direction while projecting out all previously identifiedones, all features carrying relevant information about the classification canbe found, facilitating a detailed analysis of feature relevances. Moreover,IRMA can be used to generate improved low-dimensional representationsand visualizations of labeled data sets.

AB - We introduce and investigate the iterated application of Gen-eralized Matrix Relevance Learning for the analysis of feature relevances inclassification problems. The suggested Iterated Relevance Matrix Analysis(IRMA), identifies a linear subspace representing the classification specificinformation of the considered data sets in feature space using General-ized Matrix Learning Vector Quantization. By iteratively determining anew discriminative direction while projecting out all previously identifiedones, all features carrying relevant information about the classification canbe found, facilitating a detailed analysis of feature relevances. Moreover,IRMA can be used to generate improved low-dimensional representationsand visualizations of labeled data sets.

UR - https://www.esann.org/sites/default/files/proceedings/2023/ES2023-127.pdf

M3 - Conference contribution

SN - 9782875870889

SP - 59

EP - 64

BT - ESANN 2023 Proceedings

A2 - Verleysen, Michel

PB - i6doc.com publication

ER -

Improved Interpretation of Feature Relevances: Iterated Relevance Matrix Analysis (IRMA)

Abstract

Access to Document

Handle.net

Other files and links

Fingerprint

Iterated Relevance Matrix Analysis (IRMA) for the identification of class-discriminative subspaces

Cite this