矩阵填充的可行方向逼近法

A Feasible Direction-approximating Method for Matrix Completion

以矩阵填充的子空间逼近法为基础,提出了一种矩阵填充的可行方向逼近法,该算法运用二次规划技术产生最接近可行的矩阵,且迭代矩阵逐步向低秩可行矩阵逼近,满足收敛条件产生低秩的最优填充矩阵.通过数值实验验证了新的算法比传统算法更有效.

This paper propose a new feasible direction-approximating method for matrix completion based on the subspace-approximating algorithm.And it is used quadratic program to get the closest feasible matrix.The iterative matrices generated by the new algorithm is gradually to approximating the feasible matrix, and satisfying the convergences to produce the best completion matrix.Then it shows this new algorithm is more effective than the augmented Lagrange multiplier algorithm for matrix completion by numerical experiments.