Thiele-type and Lagrange-type generalized inverse rational interpolation for rectangular complex matrices (Q1124936)

The generalized inverse matrix rational interpolants (GMRI) are defined as a generalization of the generalized inverse vector rational interpolants as introduced by \textit{P. Graves-Morris} [Numer. Math. 42, 331-348 (1983; Zbl 0525.41014)] and \textit{P. Graves-Morris} and \textit{C. Jenkins} [Constructive Approximation 2, 263-289 (1986; Zbl 0614.41016)]. Uniqueness of the rational function is proved, a Thiele-type matrix continued fraction expansion as well as explicit Lagrange-type determinantal expressions are given. Finally, as an application, a method for extrapolation to the limit of a matrix sequence is derived. The definition of GMRI used is the following. Given are \(\{(z_i,A_i)\), \(i=0,\ldots,n\}\), \(z_i\) are distinct complex numbers, \(A_i\) are general complex \(s\times t\) matrices. \(R(z)=P(z)/q(z)\) is a GMRI of type \([m/2k]\) if \(P\) is a \(s\times t\) matrix polynomial and \(q\) is a scalar polynomial, such that (i) \(R(z_i)=A_i\), \(i=0,\ldots,n\), (ii) the maximal degree of the entries in \(P\) is \(\leq n\), and the degree of \(q\) is \(2k\), (iii) \(q\) divides \(\text{ trace}[P(z)^*P(z)]\), and (iv) \(q(z)=q^*(z)\).