Paper Image

Machine learning to enhance branch-and-bound for mixed integer linear programs

Published on:

8 February 2024

Primary Category:

Optimization and Control

Paper Authors:

Lara Scavuzzo,

Karen Aardal,

Andrea Lodi,

Neil Yorke-Smith

Bullets

Key Details

Machine learning can optimize metrics of branch-and-bound efficiency

Key solver components like branching and primal heuristics can be enhanced

MILP data and structure should be represented for machine learning

Specialized benchmarks, generators, and libraries facilitate research

Integration of ML and math optimization seen as complementary

AI generated summary

Machine learning to enhance branch-and-bound for mixed integer linear programs

This paper surveys approaches that use machine learning to improve key components of branch-and-bound solvers for mixed integer linear programs. It focuses on data-driven methods to enhance primal heuristics, branching strategies, cutting planes, node selection, and configuration decisions. The integration of machine learning and optimization is positioned as complementary technologies that can benefit solving efficiency. Detailed attention is given to representing mixed integer programs for machine learning and to benchmarks and software used in this growing research area.

Answers from this paper

Comments

No comments yet, be the first to start the conversation...

Sign up to comment on this paper

Sign Up