Saved in:
Bibliographic Details
Main Authors: Zeng, Shuli, Zhang, Sijia, Li, Shaoang, Wu, Feng, Li, Xiang-Yang
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2503.15847
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913747514163200
author Zeng, Shuli
Zhang, Sijia
Li, Shaoang
Wu, Feng
Li, Xiang-Yang
author_facet Zeng, Shuli
Zhang, Sijia
Li, Shaoang
Wu, Feng
Li, Xiang-Yang
contents In mixed-integer programming (MIP) solvers, cutting planes are essential for Branch-and-Cut (B&C) algorithms as they reduce the search space and accelerate the solving process. Traditional methods rely on hard-coded heuristics for cut plane selection but fail to leverage problem-specific structural features. Recent machine learning approaches use neural networks for cut selection but focus narrowly on the efficiency of single-node within the B&C algorithm, without considering the broader contextual information. To address this, we propose Global Cut Selection (GCS), which uses a bipartite graph to represent the search tree and combines graph neural networks with reinforcement learning to develop cut selection strategies. Unlike prior methods, GCS applies cutting planes across all nodes, incorporating richer contextual information. Experiments show GCS significantly improves solving efficiency for synthetic and large-scale real-world MIPs compared to traditional and learning-based methods.
format Preprint
id arxiv_https___arxiv_org_abs_2503_15847
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Beyond Local Selection: Global Cut Selection for Enhanced Mixed-Integer Programming
Zeng, Shuli
Zhang, Sijia
Li, Shaoang
Wu, Feng
Li, Xiang-Yang
Artificial Intelligence
In mixed-integer programming (MIP) solvers, cutting planes are essential for Branch-and-Cut (B&C) algorithms as they reduce the search space and accelerate the solving process. Traditional methods rely on hard-coded heuristics for cut plane selection but fail to leverage problem-specific structural features. Recent machine learning approaches use neural networks for cut selection but focus narrowly on the efficiency of single-node within the B&C algorithm, without considering the broader contextual information. To address this, we propose Global Cut Selection (GCS), which uses a bipartite graph to represent the search tree and combines graph neural networks with reinforcement learning to develop cut selection strategies. Unlike prior methods, GCS applies cutting planes across all nodes, incorporating richer contextual information. Experiments show GCS significantly improves solving efficiency for synthetic and large-scale real-world MIPs compared to traditional and learning-based methods.
title Beyond Local Selection: Global Cut Selection for Enhanced Mixed-Integer Programming
topic Artificial Intelligence
url https://arxiv.org/abs/2503.15847