In this thesis, efficient methods for the solution of inverse problems, combining adaptive regularization and discretization are proposed. For the computation of a Tikhonov regularization parameter, we consider an inexact Newton method
based on Morozov's discrepancy principle. In each step, a regularized problem is solved on a different discretization level, which we control using DWR error estimators. In the second part of this thesis, we combine this method with iteratively regularized Gauss-Newton methods. For both approaches, we provide a convergence analysis as well as numerical results.
«
In this thesis, efficient methods for the solution of inverse problems, combining adaptive regularization and discretization are proposed. For the computation of a Tikhonov regularization parameter, we consider an inexact Newton method
based on Morozov's discrepancy principle. In each step, a regularized problem is solved on a different discretization level, which we control using DWR error estimators. In the second part of this thesis, we combine this method with iteratively regularized Gauss...
»