Unified Presentation of the Generalized Ellipsoid Method

Publikation: Beitrag in Buch/Konferenzbericht/Sammelband/GutachtenBeitrag in Buch/Sammelband/GutachtenBeigetragenBegutachtung

Beitragende

Abstract

The paper reviews classical and some recent developments in the field of ellipsoid methods and its applications. Based on a parametrization of the generalized ellipsoid method, it is demonstrated that for certain values of the space scaling parameter λ>0, the method reduces to well-known special cases of the ellipsoid method, namely, to those of Shor, Nemirovsky and Yudin, and of Khachiyan. Moreover, convergence properties of two theoretically equivalent algorithmic versions of the parameterized generalized ellipsoid method are derived. The first version is based on updating an asymmetric matrix B, whereas the second one updates a symmetric matrix H=BB⊤. Numerical differences of these versions are highlighted. Several applications of the parametrized generalized ellipsoid method are described, e.g., convex optimization problems and finding saddle points of convex-concave functions. This also includes more detailed examples related to Lagrangian upper bounds for Boolean optimization, the minimization of ravine functions, and the numerical solution of a problem by Sylvester. Finally, a software implementation is discussed.

Details

OriginalspracheEnglisch
TitelTheory, Algorithms, and Experiments in Applied Optimization
Redakteure/-innenBoris Goldengorin
ErscheinungsortCham
Herausgeber (Verlag)Springer, Cham
Seiten345-376
Seitenumfang32
ISBN (elektronisch)978-3-031-91357-0
ISBN (Print)978-3-031-91356-3, 978-3-031-91359-4
PublikationsstatusVeröffentlicht - 2025
Peer-Review-StatusJa

Publikationsreihe

Reihe Springer Optimization and Its Applications
Band226
ISSN1931-6828

Externe IDs

Mendeley 2ab44c13-d839-3a4d-9459-cf6da6cfac7d
Scopus 105021309562

Schlagworte

Forschungsprofillinien der TU Dresden

DFG-Fachsystematik nach Fachkollegium

Fächergruppen, Lehr- und Forschungsbereiche, Fachgebiete nach Destatis

Ziele für nachhaltige Entwicklung

Schlagwörter

  • Space dilation, Lagrangian bounds, Subgradient algorithm, Nonsmooth optimization, Saddle point problem, Convex programming, Ellipsoid method