On the iteratively regularized Gauss-Newton method in Banach spaces with applications to parameter identification problems (Q2391810)

The paper deals with a nonlinear inverse problem in Banach spaces which can be formulated as a nonlinear operator equation \(F(x)= y\). Let \(y^\delta\) be the available noise data approximating \(y\). The solution can be obtained by the iteratively regularized Gauss-Newton method. The approximate solutions of this sequence of problems are the unique minimizer of a certain quadratic functional. In this paper, a convergence theory on the iteratively generalized Gauss-Newton method with general convex penalty functions is developed. Some a posteriori stopping rules and the discrepancy principle are proposed to terminate the method and to give a detailed convergence analysis under reasonable nonlinearity conditions. Then, the rates of convergence are derived. The method enables to reconstruct special features of the solution. Three numerical experiments on parameter identification in partial differential equations are reported to test the new method.