Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Graph Navier Stokes Networks

About

Graph Neural Networks (GNNs) have emerged as a cornerstone of deep learning, with most existing methods rooted in graph signal processing and diffusion equations to model message passing. However, these approaches inherently suffer from the oversmoothing problem, where node features become indistinguishable as the network depth increases. Inspired by the Navier Stokes equations, we introduce Graph Navier Stokes Networks (GNSN), a novel architecture that transcends conventional diffusion-based message passing by incorporating convection into graph structures. GNSN defines a dynamic velocity field on the graph to govern convection, enabling more efficient and direct message propagation. By adaptively balancing convection and diffusion, GNSN is able to efficiently handle datasets with varying levels of homophily. Extensive evaluations across twelve real-world datasets demonstrate that GNSN consistently outperforms state-of-the-art baselines in classification accuracy. Moreover, experimental results further emphasize its effectiveness in alleviating the oversmoothing problem.

Zexing Zhao, Guangsi Shi, Yu Gong, Tianyu Wang, Shirui Pan, Hongye Cheng, Yuxiao Li• 2026

Related benchmarks

TaskDatasetResultRank
Node ClassificationChameleon
Accuracy71.85
867
Node ClassificationWisconsin
Accuracy90.02
864
Node ClassificationCornell
Accuracy89.19
851
Node ClassificationTexas
Accuracy0.9189
801
Node ClassificationSquirrel
Accuracy59.54
786
Node ClassificationPubmed
Accuracy90.41
627
Node ClassificationCora
Accuracy88.73
583
Node ClassificationPhoto
Mean Accuracy95.43
374
Node ClassificationCiteseer
Mean Accuracy76.87
202
Node ClassificationAmazon Computers
Accuracy88.13
167
Showing 10 of 13 rows

Other info

Follow for update