Multivariate prediction, de Branges spaces, and related extension and inverse problems (Q1684106)

The main topic of this book is the prediction of the future, given a segment of the past, of vector-valued stochastic processes which are either weakly stationary or have weakly stationary increments and have a locally absolutely continuous spectral function with spectral density \(\Delta(\mu)\) satisfying Szegő's condition \[ \int^\infty_{-\infty} \frac{|\ln\text{det}\{\Delta(\mu)\}|}{1+\mu^2}\,d\mu<\infty. \] The problem is studied purely functional analytically, an approach initiated for scalar processes by \textit{M. G. Kreĭn} [Dokl. Akad. Nauk SSSR, n. Ser. 94, 13--16 (1954; Zbl 0057.35002)]. It amounts to computing projections onto subspace of Hilbert spaces of \(p\times 1\) vector-valued functions with inner product \[ \langle f,g\rangle= \int^\infty_{-\infty} g(\mu)^*\Delta(\mu)f(\mu)\,d\mu. \] Based on a suitably restricted \(p\times p\) matrix-valued functions \(\Delta(\mu)\), the subspaces are identified as vector valued de Branges spaces and the projections in terms of the reproducing kernels of these spaces and/or in terms of a generalized Fourier transform obtained from the solution of an associated inverse spectral problem. The projection of the past onto the future and vice versa is expressed using a suitably defined Hankel operator and its adjoint. Explicit calculations are given for rational spectral densities. The specific mathematical background relied upon is extensively surveyed in the text and supplemented where necessary. It includes properties of the resolvent matrices of matrix versions of a number of extension problems first studied by Kreĭn, the theory of reproducing kernel Hilbert spaces of entire vector-valued functions introduced by L. de Branges, and the inverse spectral problem for canonical integral systems as well as Kreĭn-Dirac differential systems. Proofs for the auxiliary result are given, wherever the authors could not provide references or had to extend known results for their purpose. Benefiting the readability of the book, generality is kept down to what is actually needed for the main results.