13 April 2022

Patrick Rubin-Delanchy, School of Mathematics, University of Bristol, UK
Annie Gray, School of Mathematics, Computational Statistics and Data Science PhD Student (Bristol)

Graph embedding is a common task in data science, in which we aim to represent the nodes of a network as a set of points in space. The point cloud can then be used for exploratory data analysis, e.g. visualisation, clustering, anomaly detection, or as feature vectors in a downstream machine-learning task. However, there remains a lot of confusion in this area. To give a few examples: should a link be interpreted as a similarity? Is graph embedding a form of dimension reduction? Are datasets linking e.g. users with items, or drugs with diseases with genes, “graphs”? And, finally, how can we incorporate time? In this talk we will describe a research programme in which we have tried to make sense of these questions, and develop a more systematic and robust approach to graph embedding. Results are illustrated with various applications, including cyber-security, anti-corruption, and some more light-hearted examples.

In cooperation with the Jean Golding Institute, University of Bristol

More information and registration on the Bristol Data Science Seminar Series