1 min to read
Stochastic Blockmodels meet Graph Neural Networks
Latent Feature Modeling via Indian Buffet Process with Graph Neural Networks.
Video
Learn sparse node embeddings for graphs. These embeddings can be used to identify the community membership(s) of each node in the graph, as well as for tasks such as link prediction.
In this work, they unify these two directions by developing a sparse variational autoencoder for graphs, that retains the interpretability of SBMs, while also enjoying the excellent predictive performance of graph neural nets
Comments