inria-00284620, version 2
Slope heuristics for variable selection and clustering via Gaussian mixtures
Cathy Maugis
1Bertrand Michel a, 1
N° RR-6550 (2008)
Résumé : Specific Gaussian mixtures are considered to solve simultaneously variable selection and clustering problems. A penalized likelihood criterion is proposed in Maugis and Michel (2008) to choose the number of mixture components and the relevant variable subset. This criterion is depending on unknown constants to be approximated in practical situations. A "slope heuristics" method is proposed and experimented to deal with this practical problem in this context. Numerical experiments on simulated datasets, a curve clustering example and a genomics application highlight the interest of the proposed heuristics.
- a – Université Paris-Sud 11
- 1 : SELECT (INRIA Saclay - Ile de France)
- INRIA – Université Paris XI - Paris Sud – CNRS : UMR
- Domaine : Mathématiques/Statistiques
Statistiques/Théorie - Mots-clés : Model-based clustering – Variable selection – Penalized likelihood criterion – Slope heuristics – Curve clustering
- Référence interne : RR-6550
- Versions disponibles : v1 (03-06-2008) v2 (04-06-2008)
- inria-00284620, version 2
- http://hal.inria.fr/inria-00284620
- oai:hal.inria.fr:inria-00284620
- Contributeur : Cathy Maugis
- Soumis le : Mercredi 4 Juin 2008, 12:59:10
- Dernière modification le : Mercredi 4 Juin 2008, 16:08:20
Documents associés
Exporter