Compressive Statistical Learning with Random Feature Moments

Rémi Gribonval 1 Gilles Blanchard 2 Nicolas Keriven 1, 3 Yann Traonmilin 1
1 PANAMA - Parcimonie et Nouveaux Algorithmes pour le Signal et la Modélisation Audio
Inria Rennes – Bretagne Atlantique , IRISA_D5 - SIGNAUX ET IMAGES NUMÉRIQUES, ROBOTIQUE
2 Institut für Mathematik [Potsdam]
3 UR1 - Université de Rennes 1
Abstract : We describe a general framework –compressive statistical learning– for resource-efficient large-scale learning: the training collection is compressed in one pass into a low-dimensional sketch (a vector of random empirical generalized moments) that captures the information relevant to the considered learning task. A near-minimizer of the risk is computed from the sketch through the solution of a nonlinear least squares problem. We investigate sufficient sketch sizes to control the generalization error of this procedure. The framework is illustrated on compressive clustering, compressive Gaussian mixture Modeling with fixed known variance, and compressive PCA.
Keywords : random features excess risk control random moments sketching Kernel mean embedding statistical learning dimension reduction
Type de document :
Pré-publication, Document de travail
Main novelties compared to version 1: improved concentration bounds, improved sketch sizes for co.. 2017
Liste complète des métadonnées
https://hal.inria.fr/hal-01544609
Contributeur : Rémi Gribonval <>
Soumis le : mercredi 6 décembre 2017 - 15:26:55
Dernière modification le : jeudi 15 novembre 2018 - 11:59:00

Fichiers

Compressive_learning_preprint....
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : hal-01544609, version 2
  • ARXIV : 1706.07180

Collections

INSMI | UNIV-RENNES1 | INRIA | IRISA | INRIA2017 | UR1-UFR-ISTIC | UNIV-RENNES

Citation

Rémi Gribonval, Gilles Blanchard, Nicolas Keriven, Yann Traonmilin. Compressive Statistical Learning with Random Feature Moments. Main novelties compared to version 1: improved concentration bounds, improved sketch sizes for co.. 2017. 〈hal-01544609v2〉

Exporter

BibTeX TEI DC DCterms EndNote

Partager

Métriques

Consultations de la notice

1048

Téléchargements de fichiers

193

Contact


archive-ouverte@inria.fr