One-Shot Learning for MIPs with SOS1 Constraints

Charly Robinson La Rocca (), Jean-François Cordeau () and Emma Frejinger ()
Additional contact information
Charly Robinson La Rocca: Université de Montréal
Jean-François Cordeau: HEC Montréal
Emma Frejinger: Université de Montréal

SN Operations Research Forum, 2024, vol. 5, issue 3, 1-28

Abstract: Abstract Efficient algorithms and solvers are required to provide optimal or near-optimal solutions quickly and enable organizations to react promptly to dynamic situations such as supply chain disruptions or changing customer demands. State-of-the-art mixed-integer programming (MIP) solvers are crafted to tackle a wide variety of problems, yet many real-world situations are characterized by problem instances that originate from a narrow distribution. This has inspired the creation of tailored approaches that exploit historical data to inform heuristic design. Deep learning (DL) methods are typically used in this context to extract patterns from data, but they require large datasets and comprehensive hyperparameter tuning for strong performance. This article describes a one-shot learning heuristic that leverages solutions discovered within the branch-and-bound tree to construct a model with minimal overhead. We evaluate our method on the locomotive assignment problem (LAP) and sets of MIPLIB instances that contain constraints based on special ordered sets of type 1. Experimental results include a comparison with multiple primal heuristics and state-of-the-art MIP solvers. We show that the method is most effective with CPLEX in terms of the average primal gap.

Keywords: Machine learning; Mixed-integer programming; Learning heuristics; Fleet management problem; MIPLIB (search for similar items in EconPapers)
Date: 2024
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
http://link.springer.com/10.1007/s43069-024-00336-6 Abstract (text/html)
Access to the full text of the articles in this series is restricted.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:snopef:v:5:y:2024:i:3:d:10.1007_s43069-024-00336-6

Ordering information: This journal article can be ordered from
https://www.springer.com/journal/43069

DOI: 10.1007/s43069-024-00336-6

Access Statistics for this article

SN Operations Research Forum is currently edited by Marco Lübbecke

More articles in SN Operations Research Forum from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().