Hyperbolic Graph Embeddings: a Survey and an Evaluation on Anomaly Detection
Machine Learning
2025-12-23 v1 Artificial Intelligence
Abstract
This survey reviews hyperbolic graph embedding models, and evaluate them on anomaly detection, highlighting their advantages over Euclidean methods in capturing complex structures. Evaluating models like \textit{HGCAE}, \textit{-VAE}, and \textit{HGCN} demonstrates high performance, with \textit{-VAE} achieving an F1-score of 94\% on the \textit{Elliptic} dataset and \textit{HGCAE} scoring 80\% on \textit{Cora}. In contrast, Euclidean methods like \textit{DOMINANT} and \textit{GraphSage} struggle with complex data. The study emphasizes the potential of hyperbolic spaces for improving anomaly detection, and provides an open-source library to foster further research in this field.
Cite
@article{arxiv.2512.18826,
title = {Hyperbolic Graph Embeddings: a Survey and an Evaluation on Anomaly Detection},
author = {Souhail Abdelmouaiz Sadat and Mohamed Yacine Touahria Miliani and Khadidja Hab El Hames and Hamida Seba and Mohammed Haddad},
journal= {arXiv preprint arXiv:2512.18826},
year = {2025}
}