Robust classification using mixtures of dependency networks

José A. Gámez, Juan L. Mateo, Thomas Dyhre Nielsen, José M. Puerta

Research output: Contribution to book/anthology/report/conference proceedingArticle in proceedingResearchpeer-review

2 Citations (Scopus)
155 Downloads (Pure)

Abstract

Dependency networks have previously been proposed as alternatives to e.g. Bayesian networks by supporting fast algorithms for automatic learning. Recently dependency networks have also been proposed as classification models, but as with e.g. general probabilistic inference, the reported speed-ups are often obtained at the expense of accuracy.
In this paper we try to address this issue through the use of mixtures of dependency networks. To reduce learning time and improve robustness when dealing with data sparse classes, we outline methods for reusing calculations across mixture components. Finally, the proposed model is empirically compared to other state-of-the-art classifiers, both in
terms of accuracy and learning time.
Original languageEnglish
Title of host publicationProceedings of the Fourth European Workshop on Probabilistic Graphical Models
Number of pages8
Publication date2008
Pages129-136
Publication statusPublished - 2008
EventThe Fourth European Workshop on Probabilistic Graphical Models - Hirtshals, Denmark
Duration: 17 Sep 200819 Sep 2008
Conference number: 4

Conference

ConferenceThe Fourth European Workshop on Probabilistic Graphical Models
Number4
CountryDenmark
CityHirtshals
Period17/09/200819/09/2008

Fingerprint

Bayesian networks

Cite this

Gámez, J. A., Mateo, J. L., Nielsen, T. D., & Puerta, J. M. (2008). Robust classification using mixtures of dependency networks. In Proceedings of the Fourth European Workshop on Probabilistic Graphical Models (pp. 129-136)
Gámez, José A. ; Mateo, Juan L. ; Nielsen, Thomas Dyhre ; Puerta, José M. / Robust classification using mixtures of dependency networks. Proceedings of the Fourth European Workshop on Probabilistic Graphical Models. 2008. pp. 129-136
@inproceedings{a04432d0893c11dd8b88000ea68e967b,
title = "Robust classification using mixtures of dependency networks",
abstract = "Dependency networks have previously been proposed as alternatives to e.g. Bayesian networks by supporting fast algorithms for automatic learning. Recently dependency networks have also been proposed as classification models, but as with e.g. general probabilistic inference, the reported speed-ups are often obtained at the expense of accuracy. In this paper we try to address this issue through the use of mixtures of dependency networks. To reduce learning time and improve robustness when dealing with data sparse classes, we outline methods for reusing calculations across mixture components. Finally, the proposed model is empirically compared to other state-of-the-art classifiers, both in terms of accuracy and learning time.",
author = "G{\'a}mez, {Jos{\'e} A.} and Mateo, {Juan L.} and Nielsen, {Thomas Dyhre} and Puerta, {Jos{\'e} M.}",
year = "2008",
language = "English",
pages = "129--136",
booktitle = "Proceedings of the Fourth European Workshop on Probabilistic Graphical Models",

}

Gámez, JA, Mateo, JL, Nielsen, TD & Puerta, JM 2008, Robust classification using mixtures of dependency networks. in Proceedings of the Fourth European Workshop on Probabilistic Graphical Models. pp. 129-136, Hirtshals, Denmark, 17/09/2008.

Robust classification using mixtures of dependency networks. / Gámez, José A.; Mateo, Juan L.; Nielsen, Thomas Dyhre; Puerta, José M.

Proceedings of the Fourth European Workshop on Probabilistic Graphical Models. 2008. p. 129-136.

Research output: Contribution to book/anthology/report/conference proceedingArticle in proceedingResearchpeer-review

TY - GEN

T1 - Robust classification using mixtures of dependency networks

AU - Gámez, José A.

AU - Mateo, Juan L.

AU - Nielsen, Thomas Dyhre

AU - Puerta, José M.

PY - 2008

Y1 - 2008

N2 - Dependency networks have previously been proposed as alternatives to e.g. Bayesian networks by supporting fast algorithms for automatic learning. Recently dependency networks have also been proposed as classification models, but as with e.g. general probabilistic inference, the reported speed-ups are often obtained at the expense of accuracy. In this paper we try to address this issue through the use of mixtures of dependency networks. To reduce learning time and improve robustness when dealing with data sparse classes, we outline methods for reusing calculations across mixture components. Finally, the proposed model is empirically compared to other state-of-the-art classifiers, both in terms of accuracy and learning time.

AB - Dependency networks have previously been proposed as alternatives to e.g. Bayesian networks by supporting fast algorithms for automatic learning. Recently dependency networks have also been proposed as classification models, but as with e.g. general probabilistic inference, the reported speed-ups are often obtained at the expense of accuracy. In this paper we try to address this issue through the use of mixtures of dependency networks. To reduce learning time and improve robustness when dealing with data sparse classes, we outline methods for reusing calculations across mixture components. Finally, the proposed model is empirically compared to other state-of-the-art classifiers, both in terms of accuracy and learning time.

M3 - Article in proceeding

SP - 129

EP - 136

BT - Proceedings of the Fourth European Workshop on Probabilistic Graphical Models

ER -

Gámez JA, Mateo JL, Nielsen TD, Puerta JM. Robust classification using mixtures of dependency networks. In Proceedings of the Fourth European Workshop on Probabilistic Graphical Models. 2008. p. 129-136