Sparse approximate inverses (SPAI) are suitable parallel preconditioners for the iterative solution of large-scale ill-conditioned linear systems of equations on supercomputers. Hence, in its first part, the thesis presents new variants as well as parallel implementations of (M)SPAI and FSPAI. The second part deals with the application of MSPAI as smoother for multigrid and as regularizing preconditioner for iterative regularization methods to reconstruct blurred and noisy signals such as images.
Using MSPAI's probing concept it is possible to construct preconditioners with a different behavior on different frequency subspaces which are favorable for the mentioned applications. Based on this, the third part focuses on the improved reconstruction of discrete ill-posed inverse problems. Further approaches are introduced with respect to direct and iterative regularization methods.
«
Sparse approximate inverses (SPAI) are suitable parallel preconditioners for the iterative solution of large-scale ill-conditioned linear systems of equations on supercomputers. Hence, in its first part, the thesis presents new variants as well as parallel implementations of (M)SPAI and FSPAI. The second part deals with the application of MSPAI as smoother for multigrid and as regularizing preconditioner for iterative regularization methods to reconstruct blurred and noisy signals such as image...
»