A note on a globally convergent Newton method for solving monotone variational inequalities (Q1091770): Difference between revisions

It is shown that Newton's method yields a descent direction for a non- convex non-differentiable merit function, even in the absence of strong monotonicity. This result is used to modify Newton's method into a globally convergent algorithm by introducing a line-search strategy. Furthermore, under strong monotonicity the optimal face is attained after a finite number of iterations, the stepsize is eventually fixed to the value one, resulting in the usual Newton step. Computational results are presented for test problems involving 4,5,5,15 and 25 variables.