Ego-GNNs: Exploiting Ego Structures in Graph Neural Networks

• Main Authors: Sandfelder, Dylan, Vijayan, Priyesh, Hamilton, William L.
• Format: Conference Proceeding
• Language: English
• Subjects: Analytical models; Convolution; Deep learning; Graph neural networks; Image processing; Multiplexing; Performance gain
• ISSN: 2379-190X
• DOI: 10.1109/ICASSP39728.2021.9414015

Graph neural networks (GNNs) have achieved remarkable success as a framework for deep learning on graph-structured data. However, GNNs are fundamentally limited by their tree-structured inductive bias: the WL-subtree kernel formulation bounds the representational capacity of GNNs, and polynomial-time GNNs are provably incapable of recognizing triangles in a graph. In this work, we propose to augment the GNN message-passing operations with information de-fined on ego graphs (i.e., the induced subgraph surrounding each node). We term these approaches Ego-GNNs and show that Ego-GNNs are provably more powerful than standard message-passing GNNs. In particular, we show that Ego-GNNs are capable of recognizing closed triangles, which is essential given the prominence of transitivity in real-world graphs. We also motivate our approach from the perspective of graph signal processing as a form of multiplex graph convolution. Experimental results on node classification using synthetic and real data highlight the achievable performance gains using this approach.