On the Regularity of a Trust Region-CG Algorithm for Nonlinear Ill-posed Inverse Problems

In this paper we consider the regularity of the trust region-cg algorithm, when it is applied to nonlinear ill-posed iverse problems. The trust region algorithm can be viewed as a regularization method, but it diiers from the traditional regularization method, because no penalty term is need. Thus, the determing of the so-called regular-ization parameter in a standard regularization method is avoided. Theoretical analysis of the trust region-cg method is presented, convergency and regularity of the trust region algorithm are proved, and numerical tests are also given.