Differentiable Cutting-plane Layers for Mixed-integer Linear Optimization

TL;DR

This paper introduces differentiable cutting-plane layers (CPLs) for solving parametric mixed-integer linear problems, enabling learned, efficient cut generation that adapts to changing data and improves solution quality.

Contribution

It proposes a novel differentiable framework for generating cutting planes in MILP, allowing offline training to optimize cut generation for parametric instances.

Findings

Algorithm generalizes to unseen instances

Captures underlying parametric structures

Produces solutions with low integrality gaps

Abstract

We consider the problem of solving a family of parametric mixed-integer linear optimization problems where some entries in the input data change. We introduce the concept of cutting-plane layer (CPL), i.e., a differentiable cutting-plane generator mapping the problem data and previous iterates to cutting planes. We propose a CPL implementation to generate split cuts, and by combining several CPLs, we devise a differentiable cutting-plane algorithm that exploits the repeated nature of parametric instances. In an offline phase, we train our algorithm by updating the internal parameters controlling the CPLs, thus altering cut generation. Once trained, our algorithm computes, with predictable execution times and a fixed number of cuts, solutions with low integrality gaps. Preliminary computational tests show that our algorithm generalizes on unseen instances and captures underlying parametric structures.