Non-Local Graph Neural Networks

TL;DR

This paper introduces a non-local aggregation framework for graph neural networks, significantly improving performance on disassortative graphs by addressing limitations of local aggregation methods.

Contribution

The paper proposes a novel non-local aggregation framework with attention-guided sorting, enhancing GNNs for disassortative graphs and outperforming existing methods.

Findings

Non-local GNNs outperform state-of-the-art on disassortative graph datasets.

The framework improves both model accuracy and efficiency.

Experimental analysis confirms effectiveness on seven benchmark datasets.

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.