Learning and branching
Session date: 6 May 2024
Session host: Lara Scavuzzo Montaña
Summary:
Mixed Integer Linear Programming (MILP) is a pillar of mathematical optimization that offers a powerful modeling language for a wide range of applications. During the past decades, enormous algorithmic progress has been made in solving MILPs, and many commercial and academic software packages exist. In recent years, there has been an explosive development in the use of machine learning algorithms for enhancing all main tasks involved in the branch-and-bound algorithm, such as primal heuristics, branching, cutting planes, node selection and solver configuration decisions. In this session, we will discuss the inner workings of an MILP solver, with a focus on branching and the advances in the field of learning to branch. In particular, we will talk about the tree MDP, a suitable learning formulation for the task of branching. We will also cover MILP representation and any other MILP topics that may arise during our discussion.