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

Scalable Message Passing Neural Networks: No Need for Attention in Large Graph Representation Learning

About

We propose Scalable Message Passing Neural Networks (SMPNNs) and demonstrate that, by integrating standard convolutional message passing into a Pre-Layer Normalization Transformer-style block instead of attention, we can produce high-performing deep message-passing-based Graph Neural Networks (GNNs). This modification yields results competitive with the state-of-the-art in large graph transductive learning, particularly outperforming the best Graph Transformers in the literature, without requiring the otherwise computationally and memory-expensive attention mechanism. Our architecture not only scales to large graphs but also makes it possible to construct deep message-passing networks, unlike simple GNNs, which have traditionally been constrained to shallow architectures due to oversmoothing. Moreover, we provide a new theoretical analysis of oversmoothing based on universal approximation which we use to motivate SMPNNs. We show that in the context of graph convolutions, residual connections are necessary for maintaining the universal approximation properties of downstream learners and that removing them can lead to a loss of universality.

Haitz S\'aez de Oc\'ariz Borde, Artem Lukoianov, Anastasis Kratsios, Michael Bronstein, Xiaowen Dong• 2024

Related benchmarks

TaskDatasetResultRank
Node Classificationogbn-arxiv (test)
Accuracy74.38
433
Node Classificationogbn-products (test)
Test Accuracy90.79
137
Node Classificationpokec (test)
Accuracy80.09
78
Node Classificationogbn-proteins (test)
ROC-AUC0.8365
62
Text Classification20NEWS (test)
Top-1 Accuracy65.6
32
Image ClassificationSTL (test)
Accuracy76.7
26
Image ClassificationCIFAR (test)
Accuracy78
24
Spatio-temporal dynamics predictionWikiMath (test)
MSE0.713
11
Spatio-temporal dynamics predictionCovid (test)
MSE0.756
11
Spatio-temporal dynamics predictionChickenpox (test)
MSE0.916
11
Showing 10 of 11 rows

Other info

Follow for update