A Schur analysis approach to minimum distance problems (Q923339)

Minimum distance problems were firstly studied in probabilistic context by such authors as Olkin/Pukelsheim or Dowson/Landau. The paper under consideration contains a unified approach to minimum distance problems which are formulated in an abstract Hilbert space setting. This approach is based on Schur analysis methods. In particular, the conception of Schur analysis of non-negative Hermitian block matrices which was developed by the second and third authors has been used. This enables a quick solution of the matrix optimization problems which are associated with minimum distance problems. Moreover, the triangular systems of matrix balls and contractive parameters provide clear geometrical insights into the structure of the feasible region. From this it becomes clear that the optimal solution has to be located on the boundary of the corresponding matrix ball. What concerns the concrete computation of the optimal solution a slight extension of a method due to John von Neumann is used.