Orateur: Marco Sutti
Titre: Low-rank matrix recovery
Résumé: The problem of low-rank matrix recovery emerges naturally in several applications, such as in recommender systems, of which the Netflix problem is an instance. In this talk, I will recall the singular value decomposition and show how it relates to the low-rank approximation of a matrix. Then, we will see some simple conditions under which the problem of low-rank matrix recovery is well-posed. Finally, I will outline an algorithm that recovers the low-rank matrix and analyze its convergence behavior.

---