“…The affine matrix rank minimization (AMRM) problem consisting of recovering a low-rank matrix that satisfies a given system of linear equality constraints is an important problem in recent years, and has attracted much attention in many applications such as machine learning [1], collaborative filtering in recommender systems [2,3], computer vision [4], network localization [5], system identification [6,7], control theory [8,9], and so on. A special case of AMRM is the matrix completion (MC) problem [10], it has been applied in the famous Netflix problem [11] and image inpainting problem [12]. Unfortunately, the problem AMRM is generally NP-hard [13] and all known algorithms for exactly solving it are doubly exponential in theory and in practice due to the combinational nature of the rank function.…”