A Modified Tikhonov Regularization for Linear Operator Equations

From MaRDI portal





The paper concerns the approximate solving of an ill-posed linear operator equation \(Kx=y\) under an a priori assumption on the exact solution formulated in terms of so called source condition \(x\in R[(K^*K)^\sigma]\). The cases of a noisy right hand side \(y_\delta\) as well as perturbed operators \(K_h\) are considered. A new class of regularization methods called modified Tikhonov regularization is introduced. The modification can be described as follows: It consists in replacing the ordinary Tikhonov regularized equation \((\alpha I+K_h^*K_h)z=K_h^*y_\delta\) by the following one: \((\alpha I+(K_h^*K_h)^\sigma)^{1/\sigma}z=K_h^*y_\delta\) which depends explicitly on the a priori source condition. An a priori formula for choosing the regularization parameter \(\alpha\) is derived, under which the optimal convergence order of the regularized solution is obtained. Numerical results are given for several ill-posed integral equations. They coincide with the theoretical results and show that for some cases (for Symm's integral equation) the modified method can be more accurate than the ordinary Tikhonov regularization.











This page was built for publication: A Modified Tikhonov Regularization for Linear Operator Equations

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5704526)