Role action embeddings: scalable representation of network positions
This work addresses the challenge of scalable representation learning for network positions, which is incremental as it builds on existing GNN methods with new techniques like neighbor shuffling.
The paper tackles the problem of learning role embeddings for nodes in networks by proposing RAE, an unsupervised framework that combines a within-node loss and a GNN architecture to embed nodes with similar local neighborhoods together, and it achieves competitive performance on graph and node classification tasks, sometimes matching semi-supervised methods.
We consider the question of embedding nodes with similar local neighborhoods together in embedding space, commonly referred to as "role embeddings." We propose RAE, an unsupervised framework that learns role embeddings. It combines a within-node loss function and a graph neural network (GNN) architecture to place nodes with similar local neighborhoods close in embedding space. We also propose a faster way of generating negative examples called neighbor shuffling, which quickly creates negative examples directly within batches. These techniques can be easily combined with existing GNN methods to create unsupervised role embeddings at scale. We then explore role action embeddings, which summarize the non-structural features in a node's neighborhood, leading to better performance on node classification tasks. We find that the model architecture proposed here provides strong performance on both graph and node classification tasks, in some cases competitive with semi-supervised methods.