Institut für Mathematik
At the beginning of the talk I will give a short speech on what is a inverse problem and how to come up with the Iteratively Regularized Gauss Newton Method (IRGNM). This talk deals with an analysis of IRGNM in its Ivanov version in a Banach space setting; the Ivanov version is an equivalent formulation to the Tikhonov version but with a restriction on the minimization process instead of a penalty on the cost function as in Tikhonov. The choice of parameter regularization play crucial roles in all methods to solve a inverse problem; since parameters chosen in a priori fashion might depend on some knowledge of the solution, we present here a posteriori choice of parameter. I will show convergence results in its continuous and discretized version. At the end, I present a model example and show how is the IRGNM step and how to solve it.
Mario Previatti de Souza
Senka Haznadar (senka [dot] omerhodzic [at] aau [dot] at)