Combining and scaling descent and negative curvature directions

摘要: The aim of this paper is the study different approaches to combine and scale, in an efficient manner, descent information for solution unconstrained optimization problems. We consider situation which directions are available a given iteration, we wish analyze how these order provide method more robust than standard Newton approach. In particular, will focus on scaling process that should be carried out before combining directions. derive some theoretical results regarding conditions necessary ensure convergence combination procedures following schemes similar our proposals. Finally, conduct computational experiments compare proposals with modified Newton’s other literature information.