This paper introduces a model-independent framework for embedding directed networks into Euclidean and hyperbolic spaces of any dimension. The approach uses dimension reduction of proximity matrices reflecting network topology, coupled with a Euclidean-to-hyperbolic coordinate conversion. A new proximity measure based on shortest path length is proposed, along with a dimension reduction technique for direct hyperbolic space mapping. The methods demonstrate high-quality embeddings across various real networks, evaluated using common quality scores.
Publisher
Communications Physics
Published On
Feb 02, 2023
Authors
Bianka Kovács, Gergely Palla
Tags
directed networks
Euclidean space
hyperbolic space
dimension reduction
proximity matrices
shortest path length
network topology
Related Publications
Explore these studies to deepen your understanding of the subject.
