PAC-Bayes and Domain Adaptation

Pascal Germain 1 Amaury Habrard 2 François Laviolette 3 Emilie Morvant 2
1 MODAL - MOdel for Data Analysis and Learning
LPP - Laboratoire Paul Painlevé - UMR 8524, Université de Lille, Sciences et Technologies, Inria Lille - Nord Europe, CERIM - Santé publique : épidémiologie et qualité des soins-EA 2694, Polytech Lille - École polytechnique universitaire de Lille
Abstract : 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 (recently 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.
Type de document :
Pré-publication, Document de travail
2018
Liste complète des métadonnées

https://hal.archives-ouvertes.fr/hal-01563152
Contributeur : Emilie Morvant <>
Soumis le : mardi 6 novembre 2018 - 13:34:27
Dernière modification le : mercredi 14 novembre 2018 - 14:40:11

Fichiers

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

Identifiants

  • HAL Id : hal-01563152, version 2
  • ARXIV : 1707.05712

Citation

Pascal Germain, Amaury Habrard, François Laviolette, Emilie Morvant. PAC-Bayes and Domain Adaptation. 2018. 〈hal-01563152v2〉

Partager

Métriques

Consultations de la notice

96

Téléchargements de fichiers

26