Improving Subgraph-GNNs via Edge-Level Ego-Network Encodings
This addresses expressivity limitations in graph learning for researchers and practitioners, though it appears incremental as an enhancement to existing MP-GNN frameworks.
The authors tackled the problem of limited expressivity in Message Passing Graph Neural Networks by introducing edge-level ego-network encodings, which theoretically surpass node-based subgraph MP-GNNs and empirically match or improve baselines on 10 datasets while reducing memory usage by 18.1x in some cases.
We present a novel edge-level ego-network encoding for learning on graphs that can boost Message Passing Graph Neural Networks (MP-GNNs) by providing additional node and edge features or extending message-passing formats. The proposed encoding is sufficient to distinguish Strongly Regular Graphs, a family of challenging 3-WL equivalent graphs. We show theoretically that such encoding is more expressive than node-based sub-graph MP-GNNs. In an empirical evaluation on four benchmarks with 10 graph datasets, our results match or improve previous baselines on expressivity, graph classification, graph regression, and proximity tasks -- while reducing memory usage by 18.1x in certain real-world settings.