Use este identificador para citar ou linkar para este item: http://repositorio.ufla.br/jspui/handle/1/42453
Título: Dynamic intersection of multiple implicit Dantzig–Wolfe decompositions applied to the adjacent only quadratic minimum spanning tree problem
Palavras-chave: Combinatorial optimization
Dantzig–Wolfe decomposition
Cutting planes
Lagrangian relaxation
Spanning trees
Combinatorial optimization
Otimização combinatória
Decomposição de Dantzig-Wolfe
Método de planos de corte
Relaxação Lagrangeana
Árvore de extensão
Árvore geradora mínima quadrática
Data do documento: Jul-2020
Editor: Elsevier
Citação: PEREIRA, D. L., CUNHA, A. S. da. Dynamic intersection of multiple implicit Dantzig–Wolfe decompositions applied to the adjacent only quadratic minimum spanning tree problem. European Journal of Operational Research, [S. I.], v. 284, n. 2, p. 413-426, July 2020. DOI: https://doi.org/10.1016/j.ejor.2019.12.042.
Resumo: In this paper, we introduce a dynamic Dantzig–Wolfe (DW) reformulation framework for the Adjacent Only Quadratic Minimum Spanning Tree Problem (AQMSTP). The approach is dynamic in the sense that the structures over which the DW reformulation takes place are defined on the fly and not beforehand. The idea is to dynamically convexify multiple promising regions of the domain, without explicitly formulating DW master programs over extended variable spaces and applying column generation. Instead, we use the halfspace representation of polytopes as an alternative to mathematically represent the convexified region in the original space of variables. Thus, the numerical machinery we devise for computing AQMSTP lower bounds operates in a cutting plane setting, separating projection cuts associated to the projection of the variables used in the extended DW reformulations. Our numerical experience indicates that the bounds are quite strong and the computational times are mostly spent by linear programming reoptimization and not by the separation procedures. Thus, we introduce a Lagrangian Relax-and-cut algorithm for approximating these bounds. The method is embedded in a Branch-and-Bound algorithm for the AQMSTP. Although it does not strictly dominate the previous state-of-the-art exact method, it is able to solve more instances to proven optimality and is significantly faster for the hardest AQMSTP instances in the literature.
URI: https://doi.org/10.1016/j.ejor.2019.12.042
http://repositorio.ufla.br/jspui/handle/1/42453
Aparece nas coleções:DCC - Artigos publicados em periódicos

Arquivos associados a este item:
Não existem arquivos associados a este item.


Os itens no repositório estão protegidos por copyright, com todos os direitos reservados, salvo quando é indicado o contrário.

Ferramentas do administrador