Approximation of matrices and a family of Gander methods for polar decomposition (Q2502315)

The authors consider the problem of approximating a rectangular complex matrix and provide two solutions: the approximation by subunitary matrices with respect to the unitary invariant norm and the minimal rank approximation with respect to the spectral norm. Iterative algorithms are proposed for both approximations. One of these algorithms is based on the family of Gander methods, which are investigated in details. Some well-built numerical experiments enlighten the effectiveness of the proposed methods, which are compared with methods based on the singular value decomposition.