A superfast algorithm for multi-dimensional Padé systems (Q1192651)

The definitions of the following approximants are given: matrix Padé, matrix Hermite-Padé and matrix simultaneous Padé. There exists a number of fast algorithms for computing these approximants, but in this paper a new superfast algorithm is presented. This algorithm is reliable and can also compute the matrix Padé approximants in the so called non- normal and non-perfect case with only a slight increase in complexity. The algorithm can be used for computing inverses of block Toeplitz and block Toeplitz-like matrices along with their Hankel counterparts.