1: \begin{abstract}

2:   This paper is concerned with the column $\ell_{2,0}$-regularized factorization

3:   model of low-rank matrix recovery problems and its computation.

4:   The column $\ell_{2,0}$-norm of factor matrices is introduced to promote

5:   column sparsity of factors and lower rank solutions. For this nonconvex

6:   nonsmooth and non-Lipschitz problem, we develop an alternating

7:   majorization-minimization (AMM) method with extrapolation,

8:   and a hybrid AMM in which a majorized alternating proximal method

9:   is proposed to seek an initial factor pair with less nonzero columns

10:   and the AMM with extrapolation is then employed to minimize of a smooth

11:   nonconvex loss. We provide the global convergence analysis

12:   for the proposed AMM methods and apply them to the matrix completion problem

13:   with non-uniform sampling schemes. Numerical experiments are conducted with

14:   synthetic and real data examples, and comparison results with

15:   the nuclear-norm regularized factorization model and the max-norm

16:   regularized convex model show that the column $\ell_{2,0}$-regularized

17:   factorization model has an advantage in offering solutions of lower error and

18:   rank within less time.

19:  \end{abstract}

20: