EURO 2025 Leeds
Abstract Submission

1188. Designing branching strategies with genetic programming

Invited abstract in session MC-38: Automating the Design, Generation and Control of Optimization Algorithms 2, stream Data Science meets Optimization.

Monday, 12:30-14:00
Room: Michael Sadler LG19

Authors (first author is the speaker)

1. Simon Renard
Computer sciences, Université Libre de Bruxelles
2. Bernard Fortz
HEC Liège, Management school of the University of Liège
3. Quentin Louveaux
Montefiore Institute, Université de Liège

Abstract

Branching strategies play a crucial role in the efficiency of branch-and-bound algorithms for solving combinatorial optimization problems. Traditional heuristics rely on handcrafted rules that remain fixed across different problem instances, while recent advances have used machine learning to learn problem-specific branching strategies.

In this work, we propose a novel approach to designing branching strategies using genetic programming. Our method simulates an evolutionary process in which individuals represent scoring functions that define branching decisions. This approach not only enables the discovery of new strategies without relying on existing heuristics but also provides interpretable strategies in the form of simple mathematical expressions.

Keywords

Status: accepted


Back to the list of papers