Hyperbolic Graph Neural Networks

TL;DR

This paper introduces hyperbolic graph neural networks that leverage Riemannian geometry to improve graph representation learning, demonstrating significant performance gains on benchmark datasets.

Contribution

The paper proposes a novel hyperbolic GNN architecture utilizing Riemannian geometry and scalable algorithms for enhanced graph representation learning.

Findings

Hyperbolic GNNs outperform Euclidean GNNs on benchmark datasets.

The proposed method effectively models structural properties of graphs.

Hyperbolic geometry provides better embeddings for complex graph data.

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.