Passer à la navigation principale Passer à la recherche Passer au contenu principal

Probabilistic low-rank matrix completion on finite alphabets

  • ParisTech
  • ENSAE

Résultats de recherche: Contribution à un journalArticle de conférenceRevue par des pairs

34 Citations (Scopus)

Résumé

The task of reconstructing a matrix given a sample of observed entries is known as the matrix completion problem. It arises in a wide range of problems, including recommender systems, collaborative filtering, dimensionality reduction, image processing, quantum physics or multi-class classification to name a few. Most works have focused on recovering an unknown real-valued low-rank matrix from randomly sub-sampling its entries. Here, we investigate the case where the observations take a finite number of values, corresponding for examples to ratings in recommender systems or labels in multi-class classification. We also consider a general sampling scheme (not necessarily uniform) over the matrix entries. The performance of a nuclear-norm penalized estimator is analyzed theoretically. More precisely, we derive bounds for the Kullback-Leibler divergence between the true and estimated distributions. In practice, we have also proposed an efficient algorithm based on lifted coordinate gradient descent in order to tackle potentially high dimensional settings.

langue originaleAnglais
Pages (de - à)1727-1735
Nombre de pages9
journalAdvances in Neural Information Processing Systems
Volume2
Numéro de publicationJanuary
étatPublié - 1 janv. 2014
Modification externeOui
Evénement28th Annual Conference on Neural Information Processing Systems 2014, NIPS 2014 - Montreal, Canada
Durée: 8 déc. 201413 déc. 2014

Empreinte digitale

Examiner les sujets de recherche de « Probabilistic low-rank matrix completion on finite alphabets ». Ensemble, ils forment une empreinte digitale unique.

Contient cette citation