Hybrid methods for nonlinear least squares. (English) Zbl 0648.65051

The authors consider a recently proposed‘hybrid’ method for the nonlinear least squares minimization problem. (It is hybrid in the sense that it combines a Gauss-Newton strategy with a quasi-Newton update. A counterexample establishes that the method is not superlinearly convergent in general. However, modified methods are proposed which do possess a local superlinear convergence rate. Numerical results are provided.
Reviewer: T.F.Coleman


65K05 Numerical mathematical programming methods
65H10 Numerical computation of solutions to systems of equations
90C20 Quadratic programming
Full Text: DOI