Nos partenaires

CNRS

Rechercher





Accueil du site > Français > Evénements > Agenda

Agenda

 

Agenda des manifestations organisées ou co-organisées par l’IRIT RSS

 

Séminaire

A quasi-Bayesian perspective to NMF: theory and applications

Benjamin GUEDJ - INRIA, Lille

Lundi 25 Septembre 2017, 14h00 - 15h30
INP-ENSEEIHT, salle F422 (Bâtiment F, 4e étage, accès par badge)
Version PDF :

Abstract

Quasi-Bayesian estimators are increasingly popular in statistics and machine learning, due to their generalization properties and flexibility. In a recent work (Alquier & Guedj 2017, Mathematical Methods of Statistics), we have proposed a quasi-Bayesian estimator for non-negative matrix factorization (NMF). I will present a quick overview of quasi- and PAC-Bayesian frameworks and discuss our theoretical and algorithmic contributions. A short demo of our method for digits recognition will conclude the talk : http://dx.doi.org/10.3103/S1066530717010045

Bio: Dr. Benjamin Guedj obtained his PhD in mathematics from Université Pierre et Marie Curie in 2013. Since 2014, he is a statistical machine learning researcher at Inria, assigned to the Lille - Nord Europe research center. His recent work address the design, analysis and implementation of statistical learning methods for high dimensional problems. Keywords include (but are not limited to) PAC-Bayesian theory, sparsity, optimisation theory, statistical learning theory, sequential learning, non-negative matrix factorisation, aggregation of estimators and classifiers, MCMC algorithms, concentration inequalities. More info on https://bguedj.github.io

 

Retour