No label defined (Q4340908)

In the book, the author discusses the problems of recover, or producing approximations to analytic functions from measured data, for which applications are wide-ranging and include systems theory and signal processing. For example, given a subset \(k\) of the unit circle \(\mathbb{T}\) and the values of \(f\) (which is continuous on the closed unit disc \(\overline D\) and analytic in the open unit disc \(\mathbb{D}\)) on \(k\), do these values determine \(f\) uniquely? The second problem he addresses is whether one can obtain a `good' approximation to an impulse response sequence given a finite set of input values \(u(0),\dots, u(n)\), and a finite set of output values \(y(0),\dots,y(n)\). The third problem he considers is what one can do given irregularly spaced values, or noise values, for recovering a Paley-Wiener function. These all are analytic problems with applications in systems theory and signal processing. The book consists of eight chapters. In Chapter 1, the author reviews some of background material on the theory of function spaces and operators, such as the Fourier series and transforms, the Hardy spaces, and the Hankel and Toeplitz operators. In Chapter 2, the author considers classical interpolation problems for analytic functions, such as those of Nevanlinna-Pick and Carathéodory and Fejér, and their interplay with problems of extension and recovery. Chapters 3 and 4 are oriented to practical applications, namely worst-case system identification, which is seen as the recovery of analytic functions from (corrupted) measurements. Having explained why linear algorithms do not solve the worst-case identification problems, the author concentrates on two kinds of nonlinear algorithms; the first is based on the use of Fourier series and Hankel operators, and the second one on Chebyshev interpolant. Moreover, by use of the Nevanlinna-Pick theory, he presents tuned identification algorithms. Chapter 5 collects together a variety of general interpolation techniques which are best suited to Hilbert spaces, such as the Hardy space \(H_2\) and the Paley-Wiener spaces. Reproducing kernel Hilbert spaces are introduced as a tool for discussing interpolation. The very recent matching pursuit algorithm is studied; the least squares method for parameter estimation is analyzed; and finally frames and wavelets are introduced as an approximation tool. Chapter 6 is slanted towards practical applications in worst-case identification of convolutions using input/output measurements rather than frequency domain measurements. The question of input design is an important one in this problem and its solution draws upon ideas from Galois theory and classical complex analysis. The author also considers some tuned identification methods using ideas from information-based complexity theory. Sampling is the subject of Chapter 7. The Paley-Wiener spaces and the appropriate reproducing kernel Hilbert spaces are introduced for this problem. It turns out that questions of sampling are much simpler and are treated first when sampling data are obtained at regular time intervals. Then he progresses to general sampling and the ideas of Beurling and Landau on frames and balayage. In Chapter 8 he gives an identification of some of the practical applications of the analysis in the earlier chaptes. In a sentence, the book is concise and readable and is a good monograph for experts and graduated students.