Limited-Memory Variable Metric Methods that use Quantities from the Preceding Iteration

Authors: Jan Vlček, Ladislav Lukšan
Publication year: 2010
Publication type: Konferenční příspěvek (mezinárodní konf.)
Periodical: Programs and Algorithms of Numerical Mathematics
Projects: Development of methods for solving large scale nonlinear programming and nonsmooth optimization problems
Bibliographic citation: VLČEK, Jan, LUKŠAN, Ladislav. Limited-Memory Variable Metric Methods that use Quantities from the Preceding Iteration. In: VEJCHODSKÝ, T., CHLEBOUN, J., PŘIKRYL, P., SEGETH, K., ŠÍSTEK, J., eds. Programs and Algorithms of Numerical Mathematics. Praha: MÚ AV ČR, v. v. i., 2010, S. 227-232. ISBN 978-80-85823-57-8. Dostupné z: http://www.math.cas.cz/~panm/Panm15/proceedings/PANM15proc.pdf
ID: 358545

Abstract

In this contribution, a new family of globally convergent limited-memory (LM) variable metric (VM) line search methods for unconstrained minimization is presented. Numerical results indicate that the new methods can save computational time substantially for certain problems in comparison with the well-known L-BFGS method.