Comments on: On learning and branching: a survey 2
-
Quentin Louveaux [1]
-
[1] University of Liège
University of Liège
Arrondissement de Liège, Bélgica
-
- Localización: Top, ISSN-e 1863-8279, ISSN 1134-5764, Vol. 25, Nº. 2, 2017, págs. 239-241
- Idioma: inglés
- DOI: 10.1007/s11750-017-0453-4
- Enlaces
-
Resumen
Branch-and-bound lies at the core of the most successful exact techniques to solve mixed-integer linear optimization problems. Despite its big success, there remains a lot to be understood about it. In particular, the strategies behind two crucial steps of the branch-and-bound algorithm are still nowadays based on empirical research: the variable on which to branch also known as the variable selection and the node to explore next, also known as the node selection. The paper by Lodi and Zarpellon surveys several research carried out on these two questions. In particular, there has been some recent interest in using machine learning techniques to tackle these questions.
