ARCHAMBEAU, Cédric; CARON, Francois

Métadonnées

Afficher la notice complète

Licence d’utilisation du document

ARCHAMBEAU, Cédric
Xerox Research Centre Europe [Meylan]

CARON, Francois
Advanced Learning Evolutionary Algorithms [ALEA]
Institut de Mathématiques de Bordeaux [IMB]

Langue

Communication dans un congrès

Ce document a été publié dans

Conference on Uncertainty in Artificial Intelligence (UAI'2012), 2012-08, Catalina Island. 2012-08

Résumé en anglais

Multinomial logistic regression is one of the most popular models for modelling the effect of explanatory variables on a subject choice between a set of specified options. This model has found numerous applications in machine learning, psychology or economy. Bayesian inference in this model is non trivial and requires either to resort to a Metropolis-Hastings algorithm, or rejection sampling within a Gibbs sampler. In this paper, we propose an alternative model to multinomial logit. The model builds on the Plackett-Luce model, a popular model for multiple comparisons. We show that the introduction of a suitable set of auxiliary variables leads to an Expectation-Maximization algorithm to find Maximum A Posteriori estimates of the parameters. We further provide a full Bayesian treatment by deriving a Gibbs sampler, which only requires to sample from highly standard distributions, as well as a variational approximate inference scheme. All are very simple to implement. One property of our Plackett-Luce regression model is that it learns a sparse set of feature weights. We provide detailed comparisons of our method compared to sparse Bayesian multinomial logistic regression and show that it is competitive, especially in presence of polychotomous data.< Réduire

Origine

Importé de hal

Unités de recherche

Institut de Mathématiques de Bordeaux (IMB) - UMR 5251