Inverse problems. Basics, theory and applications in geophysics (Q332234)

This textbook deals with an introduction to the stable approximate solution of discretized versions of inverse problems, which are of interest in particular in geophysics. In my opinion there is one problem with this text: The author believes that it is possible to present the regularization theory including convergence rates for linear and nonlinear inverse problems without requiring more mathematical knowledge than is taught in undergraduate math courses for scientists and engineers. This seems to be extremely problematic, because the author has only linear and nonlinear systems of a finite number of equations in a finite number of unknowns as forward operators available. However, it is well-known that such finite-dimensional problems derived by discretization from inverse problems formulated as operator equations in infinite-dimensional Hilbert or Banach spaces are ill-conditioned, but not ill-posed. The proper phenomenon of ill-posedness does not occur there and cannot be explained by means of discretized problems. So, for example, the attempt in Chapter 4 to explain the complicated Engl/Kunisch/Neubauer theory including convergence rates for the Tikhonov regularization of nonlinear inverse problems must fail. As crucial detail in regularization theory, the character of source conditions as smoothness conditions of the solution with respect to the forward operator degenerates after discretization. To discuss this for simplicity in the case of linear problems: for source conditions in properly ill-posed problems it is typical that the solution must belong to the range of some aggregate of the forward operator, and this range is a non-closed infinite-dimensional subspace of the image space. After discretization, however, and in finite-dimensional spaces all subspaces and hence all such ranges are closed, and the basic character of source conditions disappears. One should clearly say that a reader can only get acquainted with such details of regularization theory if he has studied important facts and principles of functional analysis. Other books on discretized inverse problem like the one by \textit{P. C. Hansen} [Rank-deficient and discrete ill-posed problems. Numerical aspects of linear inversion. Philadelphia, PA: SIAM (1997; Zbl 0890.65037)] have focused only on relevant phenomena of the finite-dimensional equations and did not try to emulate functional analysis without requiring knowledge of this subject. However, the present author presents with Chapter 2 of the book a very valuable contribution to the practice of inverse problems by describing how to perform in detail the discretization of an ill-posed operator equation formulated in infinite-dimensional abstract spaces leading to ill-conditioned systems. Moreover, it is important and helpful for the reader that all the steps to find regularized solutions, selecting regularization parameters and solving associated optimization problems are outlined in the different parts of the book, sometimes not very detailed. As a specialist in this field, the author can make good assertions on geophysical models like inverse gravimetry and seismic tomography and he can present convincing examples.