![[2505.16952] FrontierCO: Real-World and Large-Scale Evaluation of Machine Learning Solvers for Combinatorial Optimization](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2505.16952] FrontierCO: Real-World and Large-Scale Evaluation of Machine Learning Solvers for Combinatorial Optimization
Summary
The paper presents FrontierCO, a benchmark for evaluating machine learning solvers in combinatorial optimization, emphasizing real-world applications and large-scale datasets.
Why It Matters
This research addresses the limitations of previous benchmarks that relied on synthetic data, providing a more rigorous evaluation of ML solvers in real-world scenarios. By focusing on large-scale problems, it aims to enhance the applicability of machine learning techniques in practical optimization tasks.
Key Takeaways
- FrontierCO evaluates ML solvers on real-world combinatorial optimization problems.
- The benchmark includes both easy and hard problem sets to assess solver performance.
- Results show a performance gap between ML methods and classical solvers, especially on large instances.
- ML methods can outperform classical solvers in specific cases, indicating potential for future research.
- The benchmark is publicly available, promoting further advancements in ML for combinatorial optimization.
Computer Science > Machine Learning arXiv:2505.16952 (cs) [Submitted on 22 May 2025 (v1), last revised 25 Feb 2026 (this version, v2)] Title:FrontierCO: Real-World and Large-Scale Evaluation of Machine Learning Solvers for Combinatorial Optimization Authors:Shengyu Feng, Weiwei Sun, Shanda Li, Ameet Talwalkar, Yiming Yang View a PDF of the paper titled FrontierCO: Real-World and Large-Scale Evaluation of Machine Learning Solvers for Combinatorial Optimization, by Shengyu Feng and 4 other authors View PDF HTML (experimental) Abstract:Machine learning (ML) has shown promise for tackling combinatorial optimization (CO), but much of the reported progress relies on small-scale, synthetic benchmarks that fail to capture real-world structure and scale. A core limitation is that ML methods are typically trained and evaluated on synthetic instance generators, leaving open how they perform on irregular, competition-grade, or industrial datasets. We present FrontierCO, a benchmark for evaluating ML-based CO solvers under real-world structure and extreme scale. FrontierCO spans eight CO problems, including routing, scheduling, facility location, and graph problems, with instances drawn from competitions and public repositories (e.g., DIMACS, TSPLib). Each task provides both easy sets (historically challenging but now solvable) and hard sets (open or computationally intensive), alongside standardized training/validation resources. Using FrontierCO, we evaluate 16 representative ML solv...
