Abstract
Despite matrix completion requiring the global solution of a non-convex objective, there are many computational efficient algorithms which are effective for a broad class of matrices. Based on these algorithms for matrix completion with given rank problem, we propose a class of two-stage iteration algorithms for general matrix completion in this paper. The inner iteration is the scaled alternating steepest descent algorithm for the fixed-rank matrix completion problem presented by Tanner and Wei (2016), the outer iteration is used two iteration criterions: the gradient norm and the distance between the feasible part with the corresponding part of reconstructed low-rank matrix. The feasibility of the two-stage algorithms are proved. Finally, the numerical experiments show the two-stage algorithms with shorting the distance are more effective than other algorithms.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have