Risk bounds in linear regression through PAC-Bayesian truncation

Jean-Yves Audibert 1, 2 Olivier Catoni 3, *
* Auteur correspondant
1 IMAGINE [Marne-la-Vallée]
LIGM - Laboratoire d'Informatique Gaspard-Monge, CSTB - Centre Scientifique et Technique du Bâtiment, ENPC - École des Ponts ParisTech
2 WILLOW - Models of visual object recognition and scene understanding
CNRS - Centre National de la Recherche Scientifique : UMR8548, Inria Paris-Rocquencourt, DI-ENS - Département d'informatique de l'École normale supérieure
Abstract : We consider the problem of predicting as well as the best linear combination of d given functions in least squares regression, and variants of this problem including constraints on the parameters of the linear combination. When the input distribution is known, there already exists an algorithm having an expected excess risk of order d/n, where n is the size of the training data. Without this strong assumption, standard results often contain a multiplicative log n factor, and require some additional assumptions like uniform boundedness of the d-dimensional input representation and exponential moments of the output. This work provides new risk bounds for the ridge estimator and the ordinary least squares estimator, and their variants. It also provides shrinkage procedures with convergence rate d/n (i.e., without the logarithmic factor) in expectation and in deviations, under various assumptions. The key common surprising factor of these results is the absence of exponential moment condition on the output distribution while achieving exponential deviations. All risk bounds are obtained through a PAC-Bayesian analysis on truncated differences of losses. Finally, we show that some of these results are not particular to the least squares loss, but can be generalized to similar strongly convex loss functions.
Type de document :
Pré-publication, Document de travail
78. 2010
Liste complète des métadonnées

Littérature citée [21 références]  Voir  Masquer  Télécharger

https://hal.archives-ouvertes.fr/hal-00360268
Contributeur : Jean-Yves Audibert <>
Soumis le : samedi 3 juillet 2010 - 22:12:07
Dernière modification le : lundi 28 janvier 2019 - 09:03:45
Document(s) archivé(s) le : lundi 4 octobre 2010 - 12:19:08

Fichiers

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

Identifiants

  • HAL Id : hal-00360268, version 2
  • ARXIV : 0902.1733

Citation

Jean-Yves Audibert, Olivier Catoni. Risk bounds in linear regression through PAC-Bayesian truncation. 78. 2010. 〈hal-00360268v2〉

Partager

Métriques

Consultations de la notice

1016

Téléchargements de fichiers

219