PAC-Bayes and Domain Adaptation

Germain, Pascal; Habrard, Amaury; Laviolette, François; Morvant, Emilie

Type de document :

Article dans une revue scientifique: Article original

DOI :

10.1016/j.neucom.2019.10.105

Titre :

PAC-Bayes and Domain Adaptation

Auteur(s) :

Germain, Pascal [Auteur]
Department of Computer Science and Software Engineering [Québec]
MOdel for Data Analysis and Learning [MODAL]
Habrard, Amaury [Auteur]
Laboratoire Hubert Curien [LabHC]
Laviolette, François [Auteur]
Department of Computer Science and Software Engineering [Québec]
Morvant, Emilie [Auteur]
Laboratoire Hubert Curien [LabHC]

Titre de la revue :

Neurocomputing

Pagination :

379-397

Éditeur :

Elsevier

Date de publication :

2020

ISSN :

0925-2312

Mot(s)-clé(s) en anglais :

PAC-Bayesian Theory
Domain Adaptation

Discipline(s) HAL :

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

Résumé en anglais : [en]

We provide two main contributions in PAC-Bayesian theory for domain adaptation where the objective is to learn, from a source distribution, a well-performing majority vote on a different, but related, target distribution. ...
Lire la suite >We provide two main contributions in PAC-Bayesian theory for domain adaptation where the objective is to learn, from a source distribution, a well-performing majority vote on a different, but related, target distribution. Firstly, we propose an improvement of the previous approach we proposed in Germain et al. (2013), which relies on a novel distribution pseudodistance based on a disagreement averaging, allowing us to derive a new tighter domain adaptation bound for the target risk. While this bound stands in the spirit of common domain adaptation works, we derive a second bound (introduced in Germain et al., 2016) that brings a new perspective on domain adaptation by deriving an upper bound on the target risk where the distributions’ divergence—expressed as a ratio—controls the trade-off between a source error measure and the target voters’ disagreement. We discuss and compare both results, from which we obtain PAC-Bayesian generalization bounds. Furthermore, from the PAC-Bayesian specialization to linear classifiers, we infer two learning algorithms, and we evaluate them on real data.Lire moins >

Langue :

Anglais

Comité de lecture :

Oui

Audience :

Internationale

Vulgarisation :

Non

Projet ANR :

Apprendre avec des vues interactionnelles

Collections :