HEAT: Hyperbolic Embedding of Attributed Networks

Finding a low dimensional representation of hierarchical, structured data described by a network remains a challenging problem in the machine learning community. An emerging approach is embedding networks into hyperbolic space since it can naturally represent a network's hierarchical structure. However, existing hyperbolic embedding approaches cannot deal with attributed networks, where nodes are annotated with additional attributes. To overcome this, we propose HEAT (for Hyperbolic Embedding of Attributed Networks). HEAT first extracts training samples from the network that captures both topological and attribute node similarity and then learns a low-dimensional hyperboloid embedding using full Riemannian Stochastic Gradient Descent. We show that HEAT can outperform other network embedding algorithms on several common downstream tasks. As a general network embedding method, HEAT opens the door to hyperbolic manifold learning on a wide range of both attributed and unattributed networks.