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

TL;DR

This paper introduces a neural dual warm-start method for the Linear Assignment Problem that accelerates classical solvers, maintains optimality, and scales to large instances, with proven empirical speedups.

Contribution

It proposes a learning-augmented framework with a lightweight neural architecture that provides reliable warm-starts for exact solvers, enabling large-scale and real-world applications.

Findings

Achieves over 2x speedup on synthetic datasets.

Improves solution times by 1.25x to 1.5x on real-world datasets.

Maintains full optimality and scales to large problem sizes.

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 $\mathcal{O}(N^{3})$ 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 $\mathcal{O}(N^{2})$ memory bottleneck of graph-based models, enabling neural warm-starting at large scale ($N=16{,}384$). 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 $2{\times}$ speedups on challenging synthetic distributions, in addition to improving over $1.25{\times}$ and $1.5{\times}$ 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.