1: \begin{abstract}

2: 

3: We consider the efficient numerical minimization of Tikhonov functionals with nonlinear operators and non-smooth and non-convex penalty terms, which appear for example in variational regularization. For this, we consider a new class of SCD semismooth$^*$ Newton methods, which are based on a novel concept of graphical derivatives, and exhibit locally superlinear convergence. We present a detailed description of these methods, and provide explicit algorithms in the case of sparsity and total-variation penalty terms. The numerical performance of these methods is then illustrated on a number of tomographic imaging problems.

4: 

5: \smallskip

6: \noindent \textbf{Keywords.} Inverse and Ill-Posed Problems, Tikhonov Regularization, Convex Constraints, Sparsity and TV Regularization, Non-Differentiable Optimization

7: 

8: % 65J22 - Numerical Analysis - Inverse Problems

9: % 65J20 - Numerical Analysis - Improperly posed problems; regularization

10: % 47A52 Operator Theory - Ill-posed problems, regularization

11: % 78A10 Optics, electromagnetic theory - Physical Optics

12: \end{abstract}

13: