Kybernetika 32 no. 2, 105-120, 1996

Efficient trust region method for nonlinear least squares

Ladislav Lukšan


The main purpose of this paper is to show that suitable transformations and decompositions lead to an efficient trust region method that uses one decomposition in each iteration only. Convergence properties of the resulting algorithm are comparable with convergence properties of the trust region method with optimal locally constrained step (OLCS) that uses more than one decomposition in each iteration and, therefore, that needs a longer time for obtaining results. This fact is demonstrated by numerical experiments.


90C20, 90C30, 65K05