Matrix extrapolation algorithms (Q1355226)

In the literature, there exists a great variety of extrapolation algorithms for scalar sequences. While for linear extrapolation, the generalization to matrix sequences is obvious, this is not the case for nonlinear extrapolation. In the present paper, the author presents an approach to matrix extrapolation algorithms. In Section 1, certain important results on scalar extrapolation are recalled, in particular Brezinski's \(E\)-algorithm and the general recursive projection algorithm. For the generalization to the matrix case, results on Schur complement and on Sylvester's identity are needed; these are given in Section 2. In the main part of the paper, Section 3, matrix extrapolation algorithms are given, which are generalizations of the above ones and therefore denoted as matrix \(E\)-algorithm and matrix general recursive projection algorithm.