While there are optimal TSP solvers, as well as recent learning-based approaches, the generalization of the TSP to the Multiple Traveling Salesmen Problem is much less studied. Here, we design a neural network solution that treats the salesmen, cities and depot as three different sets of varying cardinalities. We apply a novel technique that combines elements from recent architectures that were developed for sets, as well as elements from graph networks. Coupled with new constraint enforcing output layers, a dedicated loss, and a search method, our solution is shown to outperform all the meta-heuristics of the leading solver in the field.
@article{arxiv.1803.09621,
title = {Learning the Multiple Traveling Salesmen Problem with Permutation Invariant Pooling Networks},
author = {Yoav Kaempfer and Lior Wolf},
journal= {arXiv preprint arXiv:1803.09621},
year = {2019}
}