Kernel matrix regression

Abstract : We address the problem of filling missing entries in a kernel Gram matrix, given a related full Gram matrix. We attack this problem from the viewpoint of regression, assuming that the two kernel matrices can be considered as explanatory variables and response variables, respectively. We propose a variant of the regression model based on the underlying features in the reproducing kernel Hilbert space by modifying the idea of kernel canonical correlation analysis, and we estimate the missing entries by fitting this model to the existing samples. We obtain promising experimental results on gene network inference and protein 3D structure prediction from genomic datasets. We also discuss the relationship with the em-algorithm based on information geometry.
Type de document :
Pré-publication, Document de travail
2007
Liste complète des métadonnées


https://hal.archives-ouvertes.fr/hal-00133355
Contributeur : Jean-Philippe Vert <>
Soumis le : lundi 26 février 2007 - 03:52:49
Dernière modification le : mercredi 28 septembre 2016 - 13:55:51
Document(s) archivé(s) le : mardi 6 avril 2010 - 22:54:48

Fichiers

kmr.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

Collections

Citation

Yoshihiro Yamanishi, Jean-Philippe Vert. Kernel matrix regression. 2007. <hal-00133355>

Partager

Métriques

Consultations de
la notice

334

Téléchargements du document

118