Graph Neural Networks with Precomputed Node Features
This addresses a limitation in GNNs for graph classification tasks, though it is incremental as it builds on existing methods with new features.
The paper tackled the problem of Graph Neural Networks (GNNs) being unable to distinguish certain graphs or node pairs, which hinders classification tasks, by introducing augmentations like positional node embeddings, canonical node IDs, and random features, with positional embeddings significantly outperforming others in synthetic subgraph detection tasks and showing better sample efficiency and performance across different graph distributions.
Most Graph Neural Networks (GNNs) cannot distinguish some graphs or indeed some pairs of nodes within a graph. This makes it impossible to solve certain classification tasks. However, adding additional node features to these models can resolve this problem. We introduce several such augmentations, including (i) positional node embeddings, (ii) canonical node IDs, and (iii) random features. These extensions are motivated by theoretical results and corroborated by extensive testing on synthetic subgraph detection tasks. We find that positional embeddings significantly outperform other extensions in these tasks. Moreover, positional embeddings have better sample efficiency, perform well on different graph distributions and even outperform learning with ground truth node positions. Finally, we show that the different augmentations perform competitively on established GNN benchmarks, and advise on when to use them.