Performance of the truncated Newton method in unconstrained nonlinear optimization

Newton's method for unconstrained nonlinear optimization can be a demanding iterative process. Combining Krylov iterative methods with different termination criteria for the inexact solving of the Newton system, a linear or a curvilinear search technique and monotone and nonmontone globalization criteria, we manage to define a set of truncated Newton algorithms. Computational experiments were carried out in order to evaluate the performance of the defined algorithms.