Reproduce “Deep Policy Dynamic Programming for Vehicle Routing Problems” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Wouter Kool, Herke van Hoof, Joaquim Gromicho, Max Welling
DPDP combines a graph neural network (GNN)’s edge predictions with dynamic programming to guide a beam search that builds routes and prunes inferior partial routes. On 10,000 realistic 100-node vehicle routing instances, it reports mean cost 18,168 and a 0.659% gap to HGS, compared with HGS’s 18,050. These results show competitive performance in tested settings, not superiority or optimality.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.