% IMPORTANT: The following is UTF-8 encoded. This means that in the presence
% of non-ASCII characters, it will not work with BibTeX 0.99 or older.
% Instead, you should use an up-to-date BibTeX implementation like “bibtex8” or
% “biber”.
@PHDTHESIS{Witt:780278,
author = {Witt, Jonas Timon},
othercontributors = {Lübbecke, Marco and Koster, Arie Marinus and Desaulniers,
Guy},
title = {{P}olyhedral aspects of {D}antzig-{W}olfe reformulation},
school = {RWTH Aachen University},
type = {Dissertation},
address = {Aachen},
reportid = {RWTH-2020-00547},
pages = {1 Online-Ressource (xvii, 213 Seiten) : Illustrationen},
year = {2019},
note = {Veröffentlicht auf dem Publikationsserver der RWTH Aachen
University 2020; Dissertation, RWTH Aachen University, 2019},
abstract = {Although Dantzig-Wolfe reformulation for integer programs
has practically proven to yield strong relaxations in
various applications, there is only a poor theoretical
understanding of Dantzig-Wolfe reformulation in general.
This thesis addresses this issue and contributes to a better
understanding of Dantzig-Wolfe reformulation and the
underlying polyhedral theory. We investigate all
(exponentially many) Dantzig-Wolfe reformulations for a
given integer program both theoretically (for particular
problem classes) and computationally (for small models). For
the stable set problem we characterize Dantzig-Wolfe
reformulations yielding particularly strong and weak
relaxations. These findings also give new insights on the
stable set polytope and (partially) extend to related
problems like the matching and the set packing problem.
Besides, we computationally evaluate the dual bounds
obtained from all Dantzig-Wolfe reformulations for small
models of different problem classes. We observe that only
few distinct dual bounds occur and demonstrate that some
constraints have a specifically high influence on the dual
bound in these Dantzig-Wolfe reformulations. Furthermore, we
investigate whether Dantzig-Wolfe reformulations proposed in
the literature (for particular problem classes) can be
further strengthened with cutting planes from the initial
integer program. In a comprehensive computational study we
demonstrate that generic classes of cutting planes only
rarely strengthen these Dantzig-Wolfe reformulations.
Motivated by this, we prove that whole classes of generic
cutting planes are already obtained by some Dantzig-Wolfe
reformulations from the literature. In addition, we
introduce a novel class of cutting planes, which is strongly
related to classical Benders’ cuts. These cutting planes
strengthen some of the considered Dantzig-Wolfe
reformulations from the literature. With this work we shed
more light on the polyhedral aspects of Dantzig-Wolfe
reformulation and contribute to a better understanding of
the strength of Dantzig-Wolfe reformulation in general.},
cin = {813310 / 113320 / 110000},
ddc = {510},
cid = {$I:(DE-82)813310_20140620$ / $I:(DE-82)113320_20140620$ /
$I:(DE-82)110000_20140620$},
typ = {PUB:(DE-HGF)11},
doi = {10.18154/RWTH-2020-00547},
url = {https://publications.rwth-aachen.de/record/780278},
}