HAL: hal-00687304, version 2

Detailed view

Export this paper BibTeX - EndNote - TEI - RefWorks -

Available versions:

v1 (2012-04-12)

v2 (2012-06-15)

Kernel discriminant analysis and clustering with parsimonious Gaussian process models

Charles Bouveyron 1, Mathieu Fauvel 2, Stéphane Girard 3

(2012-04-12)

This work presents a family of parsimonious Gaussian process models which allow to build, from a finite sample, a model-based classifier in an infinite dimensional space. The proposed parsimonious models are obtained by constraining the eigen-decomposition of the Gaussian processes modeling each class. This allows in particular to use non-linear mapping functions which project the observations into infinite dimensional spaces. It is also demonstrated that the building of the classifier can be directly done from the observation space through a kernel function. The proposed classification method is thus able to classify data of various types such as categorical data, functional data or networks. Furthermore, it is possible to classify mixed data by combining different kernels. The methodology is as well extended to the unsupervised classification case. Experimental results on various data sets demonstrate the effectiveness of the proposed method.

1:	Statistique, Analyse et Modélisation Multidisciplinaire (SAmos-Marin Mersenne) (SAMM)
	Université Paris I - Panthéon-Sorbonne
2:	Dynamiques Forestières dans l'Espace Rural (DYNAFOR)
	Institut national de la recherche agronomique (INRA) : UR1201 – Institut National Polytechnique de Toulouse – Ecole Nationale Supérieure Agronomique de Toulouse
3:	MISTIS (INRIA Grenoble Rhône-Alpes / LJK Laboratoire Jean Kuntzmann)
	INRIA – Laboratoire Jean Kuntzmann

Subject

:

Mathematics/Statistics Statistics/Statistics Theory

hal-00687304, version 2

http://hal.archives-ouvertes.fr/hal-00687304

oai:hal.archives-ouvertes.fr:hal-00687304

From: Charles Bouveyron <>

Submitted on: Friday, 15 June 2012 17:05:30

Updated on: Friday, 15 June 2012 21:43:26