INVESTIGADORES
PILOTTA Elvio Angel
artículos
Título:
A limited-memory multipoint symmetric secant method for bound
Autor/es:
OLEG BURDAKOV; JOSÉ MARIO MARTÍNEZ; ELVIO A. PILOTTA
Revista:
ANNALS OF OPERATIONS RESEARCH
Editorial:
Springer (Kluwer)
Referencias:
Lugar: New York; Año: 2002 vol. 117 p. 51 - 70
ISSN:
0254-5330
Resumen:
A new algorithm for box-constrained minimization of differentiable functions is introduced. The way of dealing with active constraints is similar to the one used by some of the recently introducedquadratic solvers. Memoryless multipoint symmetric secant Hessian approximations are used. Since these approximations do not force positive definiteness, it is possible to accumulate, and use efficiently, negative curvature information. Numerical experiments are given.
