Improved Hessian approximations for the limited memory BFGS method

Research output: Contribution to journalArticle

21 Citations (Scopus)

Abstract

This paper considers simple modifications of the limited memory BFGS (L-BFGS) method for large scale optimization. It describes algorithms in which alternating ways of re-using a given set of stored difference vectors are outlined. The proposed algorithms resemble the L-BFGS method, except that the initial Hessian approximation is defined implicitly like the L-BFGS Hessian in terms of some stored vectors rather than the usual choice of a multiple of the unit matrix. Numerical experiments show that the new algorithms yield desirable improvement over the L-BFGS method.

Original languageEnglish
Pages (from-to)99-112
Number of pages14
JournalNumerical Algorithms
Volume22
Issue number1
Publication statusPublished - 1999

Fingerprint

Limited Memory Method
BFGS Method
Data storage equipment
Approximation
Large-scale Optimization
Unit matrix
Numerical Experiment
Experiments

Keywords

  • BFGS updating formula
  • Large scale optimization
  • Limited memory BFGS method
  • Quasi-Newton methods

ASJC Scopus subject areas

  • Applied Mathematics

Cite this

Improved Hessian approximations for the limited memory BFGS method. / Al-Baali, Mehiddin.

In: Numerical Algorithms, Vol. 22, No. 1, 1999, p. 99-112.

Research output: Contribution to journalArticle

@article{c03ab2e34665439e818cd8b8ab3b2c09,
title = "Improved Hessian approximations for the limited memory BFGS method",
abstract = "This paper considers simple modifications of the limited memory BFGS (L-BFGS) method for large scale optimization. It describes algorithms in which alternating ways of re-using a given set of stored difference vectors are outlined. The proposed algorithms resemble the L-BFGS method, except that the initial Hessian approximation is defined implicitly like the L-BFGS Hessian in terms of some stored vectors rather than the usual choice of a multiple of the unit matrix. Numerical experiments show that the new algorithms yield desirable improvement over the L-BFGS method.",
keywords = "BFGS updating formula, Large scale optimization, Limited memory BFGS method, Quasi-Newton methods",
author = "Mehiddin Al-Baali",
year = "1999",
language = "English",
volume = "22",
pages = "99--112",
journal = "Numerical Algorithms",
issn = "1017-1398",
publisher = "Springer Netherlands",
number = "1",

}

TY - JOUR

T1 - Improved Hessian approximations for the limited memory BFGS method

AU - Al-Baali, Mehiddin

PY - 1999

Y1 - 1999

N2 - This paper considers simple modifications of the limited memory BFGS (L-BFGS) method for large scale optimization. It describes algorithms in which alternating ways of re-using a given set of stored difference vectors are outlined. The proposed algorithms resemble the L-BFGS method, except that the initial Hessian approximation is defined implicitly like the L-BFGS Hessian in terms of some stored vectors rather than the usual choice of a multiple of the unit matrix. Numerical experiments show that the new algorithms yield desirable improvement over the L-BFGS method.

AB - This paper considers simple modifications of the limited memory BFGS (L-BFGS) method for large scale optimization. It describes algorithms in which alternating ways of re-using a given set of stored difference vectors are outlined. The proposed algorithms resemble the L-BFGS method, except that the initial Hessian approximation is defined implicitly like the L-BFGS Hessian in terms of some stored vectors rather than the usual choice of a multiple of the unit matrix. Numerical experiments show that the new algorithms yield desirable improvement over the L-BFGS method.

KW - BFGS updating formula

KW - Large scale optimization

KW - Limited memory BFGS method

KW - Quasi-Newton methods

UR - http://www.scopus.com/inward/record.url?scp=0033408125&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0033408125&partnerID=8YFLogxK

M3 - Article

VL - 22

SP - 99

EP - 112

JO - Numerical Algorithms

JF - Numerical Algorithms

SN - 1017-1398

IS - 1

ER -