Research Area:  Machine Learning
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.
Author(s) Name:  Qi Liu, Maximilian Nickel, Douwe Kiela
Journal name:  NeurIPS Proceedings
Publisher name:  arxiv
Volume Information:  Volume 2019
Paper Link:   https://arxiv.org/abs/1910.12892