On a modification of nonlinear overrelaxation methods


Dragoslav Herceg




A numerical composite VAOR-Newton iteration, with a modified nonlinear accelerated overrelaxation (VAOR) as the primary iteration and the Newton method as a secondary iteration, is considered. Some sufficient conditions for the local convergence of this method are given. In the linear case these conditions describe the area of convergence of the VAOR method, and also, as a subcase, the area of convergence of the AOR method.