1: \begin{abstract}

2: When a physical system is modeled by a nonlinear function, the unknown 

3: parameters can be estimated by fitting experimental observations by a

4: least-squares approach.

5: Newton's method and its variants are often used to solve problems of this type.

6: In this paper, we are concerned with the computation of the minimal-norm

7: solution of an underdetermined nonlinear least-squares problem.

8: We present a Gauss--Newton type method, which relies on

9: two relaxation parameters to ensure convergence, and which incorporates a

10: procedure to dynamically estimate the two parameters, as 

11: well as the rank of the Jacobian matrix, along the iterations.

12: Numerical results are presented.

13: \end{abstract}

14: