Abstract
Matrix completion has gained a significant interest over the past 10 years (starting by the Netflix challenge 2006). It aims at reconstructing an unknown matrix from a small subset of its (noisy) entries. In this talk, we review some state-of-the-art approaches for matrix completion, especially on Bayesian methods. Statistical warranty for Bayesian methods is discussed together with some toy numerical experiments. We further discuss the Bi-Linear matrix completion with some open windows.