PAC-Bayesian Theory Meets Bayesian Inference - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2016

PAC-Bayesian Theory Meets Bayesian Inference

Résumé

We exhibit a strong link between frequentist PAC-Bayesian risk bounds and the Bayesian marginal likelihood. That is, for the negative log-likelihood loss function, we show that the minimization of PAC-Bayesian generalization risk bounds maximizes the Bayesian marginal likelihood. This provides an alternative explanation to the Bayesian Occam's razor criteria, under the assumption that the data is generated by an i.i.d distribution. Moreover, as the negative log-likelihood is an unbounded loss function, we motivate and propose a PAC-Bayesian theorem tailored for the sub-gamma loss family, and we show that our approach is sound on classical Bayesian linear regression tasks.
Fichier principal
Vignette du fichier
BPB.pdf (840.85 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-01324072 , version 1 (31-05-2016)
hal-01324072 , version 2 (01-11-2016)
hal-01324072 , version 3 (14-02-2017)

Identifiants

Citer

Pascal Germain, Francis Bach, Alexandre Lacoste, Simon Lacoste-Julien. PAC-Bayesian Theory Meets Bayesian Inference. Neural Information Processing Systems (NIPS 2016), Dec 2016, Barcelone, Spain. pp.1876-1884. ⟨hal-01324072v3⟩
601 Consultations
603 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More