Local convergence analysis of the Gauss-Newton method under a majorant condition (Q617661)

The aim of the paper is to present a new local convergence analysis for the Gauss-Newton method under a majorant condition as introduced by \textit{L. V. Kantorovič} [Dokl. Akad. Nauk SSSR, n. Ser. 76, 17--20 (1951; Zbl 0042.11901)], and successfully used by the first two authors and Svaiter [\textit{O. P. Ferreira}, IMA J. Numer. Anal. 29, No.~3, 746--759 (2009; Zbl 1175.65067); \textit{M. L. N. Gonçalves} and \textit{O. P. Ferreira}, Comput. Optim. Appl. 48, No.~1, 1--21 (2011; Zbl 1279.90195); \textit{O. P. Ferreira} and \textit{B. F. Svaiter}, Comput. Optim. Appl. 42, No.~2, 213--229 (2009; Zbl 1191.90095)] for studying the Newton method. In the authors' analysis, the classical Lipschitz condition is relaxed using a majorant function. This condition is equivalent to Wang's condition as introduced and used by \textit{J. Chen} and \textit{W. Li} [Appl. Math. Comput. 170, No.~1, 686--705 (2005; Zbl 1084.65058); J. Comput. Appl. Math. 191, No.~1, 143--164 (2006; Zbl 1092.65043)] for studying the Gauss-Newton and Newton methods. The presented convergence analysis provides a clear relationship between the majorant function, which relaxes the Lipschitz continuity of the derivative, and the function associated with the nonlinear least square problem. Besides, the results presented in the paper make the conditions and the proof of convergence simpler. They also allow obtaining the biggest range for the uniqueness of the stationary point and the optimal convergence radius for the method as regards the majorant function. In the beginning of the paper, some notations and basic results are listed. The main results are stated further, and some properties involving the majorant function are established. The relationships between the majorant function and the nonlinear function \(F\) are presented. Then, the optimal ball of convergence and the uniqueness of the stationary point are established. Finally, the main result is proved, and some applications of this result are given.