Hyperbolic Graph Neural Networks

Abstract

Learning from graph-structured data is an important task in machine learning and artificial intelligence, for which Graph Neural Networks (GNNs) have shown great promise. Motivated by recent advances in geometric representation learning, we propose a novel GNN architecture for learning representations on Riemannian manifolds with differentiable exponential and logarithmic maps. We develop a scalable algorithm for modeling the structural properties of graphs, comparing Euclidean and hyperbolic geometry. In our experiments, we show that hyperbolic GNNs can lead to substantial improvements on various benchmark datasets.

Published at NeurIPS 2019

Cited by in corpus (4)

• Hyperbolic Graph Convolutional Neural Networks
• From Trees to Continuous Embeddings and Back: Hyperbolic Hierarchical Clustering
• Hyperbolic Graph Attention Network
• Knowledge Association with Hyperbolic Knowledge Graph Embeddings