Statistical regularization of inverse problems

From MaRDI portal





This paper determines a physically meaningful solution using prior information. The author presents a general framework for studying linear inverse problems based on statistical measures of performance, averaging kernels and reproducing kernel Hilbert spaces. He uses discrete Fredholm integral equations as examples of ill-posed inverse problems, looks for ways to incorporate prior information into the inverse problem, and presents methods based on roughness, singular value decomposition, and wavelet denoting. These methods are similar in that they reduce to estimating coefficients of the unknown function with respect to a basis and they depend on tuning parameters. They differ in the criteria used to select the class of functions for the singular values and a basis for this class.




Cited in
(33)








This page was built for publication: Statistical regularization of inverse problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2726317)