Simpler PAC-Bayesian Bounds for Hostile Data

Alquier, Pierre; Guedj, Benjamin

Type de document :

Article dans une revue scientifique

DOI :

10.1007/s10994-017-5690-0

URL permanente :

http://hdl.handle.net/20.500.12210/29360

Titre :

Simpler PAC-Bayesian Bounds for Hostile Data

Auteur(s) :

Alquier, Pierre [Auteur]
Guedj, Benjamin [Auteur]

Laboratoire Paul Painlevé - UMR 8524 [LPP]

Titre de la revue :

Machine Learning

Numéro :

107

Pagination :

887–902

Éditeur :

Springer Verlag

Date de publication :

2018

ISSN :

0885-6125

Mot(s)-clé(s) :

F-divergence
Oracle inequalities
Dependent and unbounded data
PAC-Bayesian theory

Discipline(s) HAL :

Statistiques [stat]/Machine Learning [stat.ML]

Résumé en anglais : [en]

PAC-Bayesian learning bounds are of the utmost interest to the learning community. Their role is to connect the generalization ability of an aggregation distribution $\\rho$ to its empirical risk and to its Kullback-Leibler ...
Lire la suite >PAC-Bayesian learning bounds are of the utmost interest to the learning community. Their role is to connect the generalization ability of an aggregation distribution $\\rho$ to its empirical risk and to its Kullback-Leibler divergence with respect to some prior distribution $\\pi$. Unfortunately, most of the available bounds typically rely on heavy assumptions such as boundedness and independence of the observations. This paper aims at relaxing these constraints and provides PAC-Bayesian learning bounds that hold for dependent, heavy-tailed observations (hereafter referred to as \\emph{hostile data}). In these bounds the Kullack-Leibler divergence is replaced with a general version of Csisz\\'ar's $f$-divergence. We prove a general PAC-Bayesian bound, and show how to use it in various hostile settings.Lire moins >

Langue :

Anglais

Audience :

Internationale

Vulgarisation :

Non

Établissement(s) :

CNRS
Université de Lille

Collections :

METRICS : Evaluation des technologies de santé et des pratiques médicales - ULR 2694

Date de dépôt :

2020-06-08T14:11:19Z
2020-06-09T09:17:15Z

Fichiers

documen
Accès libre
Accéder au document

Numéro de version	Lien	Date de modification
2	20.500.12210/29360*	2020-06-09T09:17:00Z
1	20.500.12210/29360.1	2020-06-08T14:11:19Z

Simpler PAC-Bayesian Bounds for Hostile Data

BibTeX

CSV

Excel

RIS

Fichiers

Historique des modifications

Simpler PAC-Bayesian Bounds for Hostile Data BibTeX CSV Excel RIS

Fichiers

Historique des modifications

Simpler PAC-Bayesian Bounds for Hostile Data

BibTeX

CSV

Excel

RIS