ML4MILP: A Benchmark Dataset for Machine Learning-based Mixed-Integer Linear Programming

Huigen Ye, Yaoyang Cheng, Hua Xu, Zhiguang Cao, Hanzhang Qin

Keywords: Mixed Integer Linear Programming, Machine Learning, Benchmark Dataset

TL;DR: We introduce ML4MILP, the first dataset explicitly tailored to evaluate ML-based algorithms for MILP problems comprehensively.

Abstract:

Machine learning (ML)-based approaches for solving mixed integer linear programming (MILP) problems have shown significant potential and are growing in sophistication. Despite this advancement, progress in this field is often hindered by the mixed and unsorted nature of current benchmark datasets, which typically lack carefully categorized collections of homogeneous instances.
To bridge this gap, we propose ML4MILP, a new open-source benchmark dataset specifically designed for evaluating ML-based optimization algorithms in the MILP domain. Based on the proposed structure and embedding similarity metrics, we used a novel classification algorithm to carefully categorize the collected and generated instances, resulting in a benchmark dataset encompassing 100,000 instances across more than 70 heterogeneous classes.
We demonstrate the utility of ML4MILP through extensive benchmarking against a comprehensive suite of algorithms in the baseline library, consisting of traditional exact solvers and heuristic algorithms, as well as ML-based approaches. Our ML4MILP is open-source and accessible at: https://anonymous.4open.science/r/ML4MILP-6BE0.

Primary Area: datasets and benchmarks

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 6997