Agglomerative independent variable group analysis

Antti Honkela (Corresponding Author), Jeremias Seppä, Esa Alhoniemi

Research output: Contribution to journalArticleScientificpeer-review

6 Citations (Scopus)

Abstract

Independent variable group analysis (IVGA) is a method for grouping dependent variables together while keeping mutually independent or weakly dependent variables in separate groups. In this paper two variants of an agglomerative method for learning a hierarchy of IVGA groupings are presented. The method resembles hierarchical clustering, but the choice of clusters to merge is based on variational Bayesian model comparison. This is approximately equivalent to using a distance measure based on a model-based approximation of mutual information between groups of variables. The approach also allows determining optimal cutoff points for the hierarchy. The method is demonstrated to find sensible groupings of variables that can be used for feature selection and ease construction of a predictive model.

Original languageEnglish
Pages (from-to)1311-1320
Number of pages10
JournalNeurocomputing
Volume71
Issue number7-9
DOIs
Publication statusPublished - 1 Mar 2008
MoE publication typeA1 Journal article-refereed

Fingerprint

Feature extraction
Cluster Analysis
Learning

Keywords

  • Hierarchical clustering
  • Independent variable group analysis
  • Mutual information
  • Variable grouping
  • Variational Bayesian learning

Cite this

Honkela, Antti ; Seppä, Jeremias ; Alhoniemi, Esa. / Agglomerative independent variable group analysis. In: Neurocomputing. 2008 ; Vol. 71, No. 7-9. pp. 1311-1320.
@article{c1eb6c6dad00481ea943d44610ee25e2,
title = "Agglomerative independent variable group analysis",
abstract = "Independent variable group analysis (IVGA) is a method for grouping dependent variables together while keeping mutually independent or weakly dependent variables in separate groups. In this paper two variants of an agglomerative method for learning a hierarchy of IVGA groupings are presented. The method resembles hierarchical clustering, but the choice of clusters to merge is based on variational Bayesian model comparison. This is approximately equivalent to using a distance measure based on a model-based approximation of mutual information between groups of variables. The approach also allows determining optimal cutoff points for the hierarchy. The method is demonstrated to find sensible groupings of variables that can be used for feature selection and ease construction of a predictive model.",
keywords = "Hierarchical clustering, Independent variable group analysis, Mutual information, Variable grouping, Variational Bayesian learning",
author = "Antti Honkela and Jeremias Sepp{\"a} and Esa Alhoniemi",
year = "2008",
month = "3",
day = "1",
doi = "10.1016/j.neucom.2007.11.024",
language = "English",
volume = "71",
pages = "1311--1320",
journal = "Neurocomputing",
issn = "0925-2312",
publisher = "Elsevier",
number = "7-9",

}

Agglomerative independent variable group analysis. / Honkela, Antti (Corresponding Author); Seppä, Jeremias; Alhoniemi, Esa.

In: Neurocomputing, Vol. 71, No. 7-9, 01.03.2008, p. 1311-1320.

Research output: Contribution to journalArticleScientificpeer-review

TY - JOUR

T1 - Agglomerative independent variable group analysis

AU - Honkela, Antti

AU - Seppä, Jeremias

AU - Alhoniemi, Esa

PY - 2008/3/1

Y1 - 2008/3/1

N2 - Independent variable group analysis (IVGA) is a method for grouping dependent variables together while keeping mutually independent or weakly dependent variables in separate groups. In this paper two variants of an agglomerative method for learning a hierarchy of IVGA groupings are presented. The method resembles hierarchical clustering, but the choice of clusters to merge is based on variational Bayesian model comparison. This is approximately equivalent to using a distance measure based on a model-based approximation of mutual information between groups of variables. The approach also allows determining optimal cutoff points for the hierarchy. The method is demonstrated to find sensible groupings of variables that can be used for feature selection and ease construction of a predictive model.

AB - Independent variable group analysis (IVGA) is a method for grouping dependent variables together while keeping mutually independent or weakly dependent variables in separate groups. In this paper two variants of an agglomerative method for learning a hierarchy of IVGA groupings are presented. The method resembles hierarchical clustering, but the choice of clusters to merge is based on variational Bayesian model comparison. This is approximately equivalent to using a distance measure based on a model-based approximation of mutual information between groups of variables. The approach also allows determining optimal cutoff points for the hierarchy. The method is demonstrated to find sensible groupings of variables that can be used for feature selection and ease construction of a predictive model.

KW - Hierarchical clustering

KW - Independent variable group analysis

KW - Mutual information

KW - Variable grouping

KW - Variational Bayesian learning

UR - http://www.scopus.com/inward/record.url?scp=40649084009&partnerID=8YFLogxK

U2 - 10.1016/j.neucom.2007.11.024

DO - 10.1016/j.neucom.2007.11.024

M3 - Article

AN - SCOPUS:40649084009

VL - 71

SP - 1311

EP - 1320

JO - Neurocomputing

JF - Neurocomputing

SN - 0925-2312

IS - 7-9

ER -