A hybrid method for the nonlinear least squares problem with simple bounds (Q1177199)

The authors consider the nonlinear least squares problem with simple bounds of the form: \[ f(x)=1/2\| F(x)\|^ 2_ 2 = 1/2\sum^ m_{i=1}F^ 2_ i(x) = \min_ x ! \] subject to the constraints \(1\leq x\leq u\); \(x,l,u\in \mathbb{R}^ n\) where \(F(x)=(F_ 1(x),\dots,F_ m(x))\) with \(F_ i(x)\in C^ 2\). The method proposed in the paper to solve such a problem constructs trust region constraints, which are ellipses centered at iterative elements, in such a way that they lie in the interior of the feasible region. To avoid the numerical instability implied by the position of the solutions of the problem on the boundary of the feasible region, the idea of active set strategy is incorporated into the method. The authors prove a convergence theorem for the method and show its practical efficiency by numerical experiments. The experiments support the claim that the method retains the advantage of the interior point method also in the case of boundary position of solutions and indicate the particular efficiency of the approach for large scale problems.