Abstract
A modification of Marquardt's nonlinear least-squares fit strategy is proposed. The performances of the modified and original strategies have been compared for six test problems. The results shown that execution times for difficult problems are much reduced by the modification.