Towards a General Recipe for Combinatorial Optimization with Multi-Filter GNNs

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wenkel, Frederik, Cantürk, Semih, Horoi, Stefan, Perlmutter, Michael, Wolf, Guy
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929603074850816
author Wenkel, Frederik
Cantürk, Semih
Horoi, Stefan
Perlmutter, Michael
Wolf, Guy
author_facet Wenkel, Frederik
Cantürk, Semih
Horoi, Stefan
Perlmutter, Michael
Wolf, Guy
contents Graph neural networks (GNNs) have achieved great success for a variety of tasks such as node classification, graph classification, and link prediction. However, the use of GNNs (and machine learning more generally) to solve combinatorial optimization (CO) problems is much less explored. Here, we introduce GCON, a novel GNN architecture that leverages a complex filter bank and localized attention mechanisms to solve CO problems on graphs. We show how our method differentiates itself from prior GNN-based CO solvers and how it can be effectively applied to the maximum cut, minimum dominating set, and maximum clique problems in a unsupervised learning setting. GCON is competitive across all tasks and consistently outperforms other specialized GNN-based approaches, and is on par with the powerful Gurobi solver on the max-cut problem. We provide an open-source implementation of our work at https://github.com/WenkelF/copt.
format Preprint
id arxiv_https___arxiv_org_abs_2405_20543
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Towards a General Recipe for Combinatorial Optimization with Multi-Filter GNNs
Wenkel, Frederik
Cantürk, Semih
Horoi, Stefan
Perlmutter, Michael
Wolf, Guy
Machine Learning
Artificial Intelligence
Discrete Mathematics
68T07 (Primary) 68T20, 90C35, 05C62 (Secondary)
F.2.2; I.2.6
Graph neural networks (GNNs) have achieved great success for a variety of tasks such as node classification, graph classification, and link prediction. However, the use of GNNs (and machine learning more generally) to solve combinatorial optimization (CO) problems is much less explored. Here, we introduce GCON, a novel GNN architecture that leverages a complex filter bank and localized attention mechanisms to solve CO problems on graphs. We show how our method differentiates itself from prior GNN-based CO solvers and how it can be effectively applied to the maximum cut, minimum dominating set, and maximum clique problems in a unsupervised learning setting. GCON is competitive across all tasks and consistently outperforms other specialized GNN-based approaches, and is on par with the powerful Gurobi solver on the max-cut problem. We provide an open-source implementation of our work at https://github.com/WenkelF/copt.
title Towards a General Recipe for Combinatorial Optimization with Multi-Filter GNNs
topic Machine Learning
Artificial Intelligence
Discrete Mathematics
68T07 (Primary) 68T20, 90C35, 05C62 (Secondary)
F.2.2; I.2.6
url https://arxiv.org/abs/2405.20543