Non-Local Graph Neural Networks

Abstract

Modern graph neural networks (GNNs) learn node embeddings through multilayer local aggregation and achieve great success in applications on assortative graphs. However, tasks on disassortative graphs usually require non-local aggregation. In addition, we find that local aggregation is even harmful for some disassortative graphs. In this work, we propose a simple yet effective non-local aggregation framework with an efficient attention-guided sorting for GNNs. Based on it, we develop various non-local GNNs. We perform thorough experiments to analyze disassortative graph datasets and evaluate our non-local GNNs. Experimental results demonstrate that our non-local GNNs significantly outperform previous state-of-the-art methods on seven benchmark datasets of disassortative graphs, in terms of both model performance and efficiency.

8 pages, 2 figures, accepted by TPAMI

References in corpus (2)

• Fast Graph Representation Learning with PyTorch Geometric
• Geom-GCN: Geometric Graph Convolutional Networks

Cited by in corpus (13)

• Is Heterophily A Real Nightmare For Graph Neural Networks To Do Node Classification?
• New Benchmarks for Learning on Non-Homophilous Graphs
• Beyond Low-Pass Filters: Adaptive Feature Propagation on Graphs
• Two Sides of the Same Coin: Heterophily and Oversmoothing in Graph Convolutional Neural Networks
• Complete the Missing Half: Augmenting Aggregation Filtering with Diversification for Graph Convolutional Networks
• Pointer Graph Networks
• Cold Brew: Distilling Graph Node Representations with Incomplete or Missing Neighborhoods
• Node2Seq: Towards Trainable Convolutions in Graph Neural Networks
• Parameter Prediction for Unseen Deep Architectures
• Stacked Graph Filter
• On Local Aggregation in Heterophilic Graphs
• BiTe-GCN: A New GCN Architecture via BidirectionalConvolution of Topology and Features on Text-Rich Networks
• FDGATII : Fast Dynamic Graph Attention with Initial Residual and Identity Mapping