Deep Learning for Bipartite Assignment Problems

Many assignment problems cannot be solved to optimality in real-time. The existing literature tends to focus on the development of handcrafted heuristics that exploit the structure of a particular assignment problem. We instead seek a general-purpose approach that can automatically learn such heuristics. In this work, we present a deep learning approach that is designed to automatically learn heuristics for a large class of assignment problems. We demonstrate the effectiveness of our approach on the weapon-target assignment (WTA) problem, which is nonlinear and NP-complete. We take non-trivial WTA problem instances and show that our approach reliably finds near-optimal solutions in milliseconds.