The Chebyshev solution of certain matrix equations (Q920580)

The problem of approximating an \((n,m)\)-matrix by a matrix of the form \(XA+BY\) is considered. The problem is closely related to that of approximating a function by a function of the form \(xh+gy\), so that the language of function is used, rather than that of matrices. The described algorithm for constructing best approximations is a version of the alternating algorithm of John von Neumann. Certain convergence properties of the alternating algorithm hold in a very general setting.