RCIN and OZwRCIN projects

Object

Title: Lagrangian Relaxation Via Ballstep Subgradient Methods

Creator:

Kiwiel, Krzysztof ; Larsson, Torbjörn ; Lindberg, Per

Date issued/created:

2002

Resource type:

Text

Subtitle:

Raport Badawczy = Research Report ; RB/76/2002

Publisher:

Instytut Badań Systemowych. Polska Akademia Nauk ; Systems Research Institute. Polish Academy of Sciences

Place of publishing:

Warszawa

Description:

26 pages ; 21 cm ; Bibliography p. 23-25

Abstract:

The paper addresses the useful properties of ballstep subgradient methods for convex optimization that use level controls for estimating the optimal value. Augmented with simple averaging schemes, they asymptotically find objective and constraint subgradients involved in optimality conditions. When applied to Lagrangian relaxation of convex programs, theyfind both primal and dual solutions, and have practicable stopping criteria. Up till now, similar results have only been known for proximal bundle methods, and for subgradient methods with divergent series stepsizes, whose convergence can be słow. Encouraging numerical results arepresented for large-scale nonlinear multicommodity network flow problems.

Relation:

Raport Badawczy = Research Report

Detailed Resource Type:

Report

Resource Identifier:

oai:rcin.org.pl:139438

Source:

RB-2002-76

Language:

eng

Language of abstract:

eng

Rights:

Creative Commons Attribution BY 4.0 license

Terms of use:

Copyright-protected material. [CC BY 4.0] May be used within the scope specified in Creative Commons Attribution BY 4.0 license, full text available at: ; -

Digitizing institution:

Systems Research Institute of the Polish Academy of Sciences

Original in:

Library of Systems Research Institute PAS

Projects co-financed by:

Operational Program Digital Poland, 2014-2020, Measure 2.3: Digital accessibility and usefulness of public sector information; funds from the European Regional Development Fund and national co-financing from the state budget.

Access:

Open

×

Citation

Citation style:

This page uses 'cookies'. More information