2026-08-11T06:09:39Z https://drops.dagstuhl.de/oai

oai:drops-oai.dagstuhl.de:15324 2024-03-06T10:55:08Z ddc:004 open_access

Learning TSP Requires Rethinking Generalization Joshi, Chaitanya K. Cappart, Quentin Rousseau, Louis-Martin Laurent, Thomas Combinatorial Optimization Travelling Salesman Problem Graph Neural Networks Deep Learning End-to-end training of neural network solvers for combinatorial optimization problems such as the Travelling Salesman Problem is intractable and inefficient beyond a few hundreds of nodes. While state-of-the-art Machine Learning approaches perform closely to classical solvers when trained on trivially small sizes, they are unable to generalize the learnt policy to larger instances of practical scales. Towards leveraging transfer learning to solve large-scale TSPs, this paper identifies inductive biases, model architectures and learning algorithms that promote generalization to instances larger than those seen in training. Our controlled experiments provide the first principled investigation into such zero-shot generalization, revealing that extrapolating beyond training data requires rethinking the neural combinatorial optimization pipeline, from network layers and learning paradigms to evaluation protocols. Schloss Dagstuhl – Leibniz-Zentrum für Informatik Chaitanya K. Joshi and Quentin Cappart and Louis-Martin Rousseau and Thomas Laurent 2021 Is Part Of LIPIcs, Volume 210, 27th International Conference on Principles and Practice of Constraint Programming (CP 2021) InProceedings Text doc-type:ResearchArticle publishedVersion application/pdf doi:10.4230/LIPIcs.CP.2021.33 urn:nbn:de:0030-drops-153248 https://drops.dagstuhl.de/entities/document/10.4230/LIPIcs.CP.2021.33 eng https://creativecommons.org/licenses/by/4.0/legalcode