Sequential predictor-corrector regularization methods and their limitations (Q4493580)

Although (discrete) sequential regularization methods for ill-posed Volterra integral equations of the first kind were introduced in the early 1960s by J. V. Beck [compare the monograph Inverse heat conduction: Ill-posed problems, Wiley-Interscience, New York (1985; Zbl 0633.73120) by \textit{J. V. Beck}, \textit{B. Backwell} and \textit{C. R. St. Clair}, for an overview], it is the recent systematic analysis by \textit{P. K. Lamm} [see, e.g., SIAM J. Appl. Math. 56, No. 2, 524-541 (1996; Zbl 0852.65127), and the survey paper in Surveys on solution methods for inverse problems (D. Colton et al., eds.), 53-82, Springer-Verlag, Vienna (2000; Zbl 0963.01010)] that has greatly increased our understanding of these methods. The present paper builds on this work: it presents stability estimates and results on the rate of convergence for equations with finitely smoothing kernels, thus extending results contained in Lamm's 1997 paper. In addition, the authors discuss the limitations of sequential regularization: if the convolution kernel is \(\nu\)-smoothing with \(\nu\geq 5\), no such methods satisfying a certain stability condition can exist. Numerical examples illustrate various aspects of the analysis.