English

Learning-Augmented Scalable Linear Assignment Problem Optimization via Neural Dual Warm-Starts

Machine Learning 2026-05-12 v1 Computer Vision and Pattern Recognition Data Structures and Algorithms Optimization and Control

Abstract

The Linear Assignment Problem (LAP) is a fundamental combinatorial optimization task with applications ranging from computer vision to logistics. Classical exact solvers such as the Hungarian and Jonker-Volgenant (LAPJV) algorithms guarantee optimality, but their cubic time complexity O(N3)\mathcal{O}(N^{3}) becomes a bottleneck for large-scale instances. Recent learning-based approaches aim to replace these solvers with neural models, often sacrificing exactness or failing to scale due to memory constraints. We propose a learning-augmented framework that accelerates exact assignment solvers while maintaining optimality and worst-case guarantees. Our method predicts dual variables to warm-start a classical solver, with a fallback that prevents asymptotic runtime degradation when the learned advice is unreliable. We introduce RowDualNet, a lightweight row-independent architecture that avoids the O(N2)\mathcal{O}(N^{2}) memory bottleneck of graph-based models, enabling neural warm-starting at large scale (N=16,384N=16{,}384). Feasibility is ensured via a constructive mechanism based on LP duality (namely, the Min-Trick), eliminating costly iterative projection. Empirically, our approach reduces the search effort of LAPJV and achieves over 2×2{\times} speedups on challenging synthetic distributions, in addition to improving over 1.25×1.25{\times} and 1.5×1.5{\times} on real-world tracking (MOT) and transportation (LPT) datasets, respectively, while strictly maintaining full optimality, effectively yielding a robust zero-shot generalization to real-world tasks.

Keywords

Cite

@article{arxiv.2605.09382,
  title  = {Learning-Augmented Scalable Linear Assignment Problem Optimization via Neural Dual Warm-Starts},
  author = {Ilay Yavlovich and Jad Agbaria and Muhamed Mhamed and Jose Yallouz and Nir Weinberger},
  journal= {arXiv preprint arXiv:2605.09382},
  year   = {2026}
}

Comments

Accepted to ICML 2026. 20 pages, 13 figures

R2 v1 2026-07-01T13:01:22.355Z