Contractive intertwining dilations and norm approximation techniques (Q799916)

Let H and H' be Hilbert spaces, T and T' be contractions on H and H' with minimal isometric dilations U and U' on K (\(\supseteq H)\) and K' (\(\supseteq H')\), respectively. For a contraction \(A:H\to H'\) which intertwines T and T' (i.e., \(T'A=AT)\), the problem of contractive interwining dilations is to describe the set \[ CID(A)=\{B:K\to K':\| B\|\leq 1,\quad U'B=BU,\quad P^{K'}_{H'}B=AP^ K_ H\}, \] where \(P^ K_ H\) and \(P^{K'}_{H'}\) denote the orthogonal projections from K onto H and K' onto H', respectively. This note starts by briefly describing the results concerning the labelling of contractive interwining dilations which were obtained in recent years by the authors together with C. Foiaş. It presents the two ways of explicitly describing the connection between an element of CID(A) and its associated A-choice sequence: one is algorithmic and the other gives a Schur type formula. These facts are then used to describe the general solution of the classical Carathéodory-Fejér extrapolation problem. In particular, computer algorithms are presented for computing from the initial string of complex numbers \(\{a_ 1,...,a_ n\}\) to the A-choice sequence \(\{\gamma_ 1,...,\gamma_ n\}\) and conversely. These algorithms are applied to the computation of the norm of an analytic Toeplitz matrix, the detecting of reflection coefficients in seismic exploration and an improvement of the Wiener prediction.