Stable factorized quasi-Newton methods for nonlinear least-squares problems (Q5936067)

A new approach is considered for solving nonlinear least-squares problems. The proposed stable factorized quasi-Newton methods are based on a QR decomposition. The convergence property and superlinear convergence are proved. Some issues about the practical implementation of the proposed methods are also presented. A brief analysis of the computational cost of the proposed methods versus the original ones is given. Numerical experiments show that the new methods are as effective as the original ones; moreover, as the theoretical analysis shows, the new ones are more stable.