Comments on: On learning and branching: a survey 3

• Bistra Dilkina ^[1] ; Elias B. Khalil ^[1] ; George L. Nemhauser ^[1]
  1. [1] Georgia Institute of Technology

     Georgia Institute of Technology

     Estados Unidos

     
• Localización: Top, ISSN-e 1863-8279, ISSN 1134-5764, Vol. 25, Nº. 2, 2017, págs. 242-246
• Idioma: inglés
• DOI: 10.1007/s11750-017-0454-3
• Enlaces
  • Texto completo (pdf)
• Resumen

  • Branch-and-bound algorithms provide a common framework for solving hard discrete optimization problems. In particular, branch-and-bound solvers for Mixed-Integer Linear Programming (MILP) have become very reliable and powerful optimization tools widely applied in fields that are critical to the global economy such as transportation, supply chain, energy, health, finance, and security. The diversity, difficulty, and scale of optimization problems continue to grow, and new advances in effective solution methods are critical to improving our ability to address them. We believe that such advances can be brought about by a recent successful research trend focusing on the integration of data-driven machine learning techniques in the design of branch-and-bound methods.