Current browse context:
stat.ML
Change to browse by:
References & Citations
Statistics > Machine Learning
Title: Optimal Transport Graph Neural Networks
(Submitted on 8 Jun 2020 (v1), last revised 8 Oct 2021 (this version, v6))
Abstract: Current graph neural network (GNN) architectures naively average or sum node embeddings into an aggregated graph representation -- potentially losing structural or semantic information. We here introduce OT-GNN, a model that computes graph embeddings using parametric prototypes that highlight key facets of different graph aspects. Towards this goal, we successfully combine optimal transport (OT) with parametric graph models. Graph representations are obtained from Wasserstein distances between the set of GNN node embeddings and ``prototype'' point clouds as free parameters. We theoretically prove that, unlike traditional sum aggregation, our function class on point clouds satisfies a fundamental universal approximation theorem. Empirically, we address an inherent collapse optimization issue by proposing a noise contrastive regularizer to steer the model towards truly exploiting the OT geometry. Finally, we outperform popular methods on several molecular property prediction tasks, while exhibiting smoother graph representations.
Submission history
From: Octavian-Eugen Ganea [view email][v1] Mon, 8 Jun 2020 14:57:39 GMT (4028kb,D)
[v2] Thu, 11 Jun 2020 02:11:24 GMT (4028kb,D)
[v3] Fri, 2 Oct 2020 16:12:44 GMT (4666kb,D)
[v4] Wed, 18 Nov 2020 17:40:10 GMT (6286kb,D)
[v5] Sat, 13 Feb 2021 16:32:14 GMT (6676kb,D)
[v6] Fri, 8 Oct 2021 19:54:45 GMT (3063kb,D)
Link back to: arXiv, form interface, contact.