Bio
Xue is a PhD candidate studying applied mathematics at the University of Edinburgh, supported by the MAC-MIGS CDT programme. Her research interests include learning on graphs/hypergraphs such as embedding, clustering and structure recovery, generative models that describe the interaction mechanism, and dynamics process on graphs/hypergraphs. She is also interested in the application of graph/hypergraph-based approaches to real-world problems such as social networks analysis, image classification, and natural language processing. Her current research involves analysing graphs/hypergraph embedding algorithms and deriving corresponding random graph models. Before her PhD, she obtained an MSc in Statistics from the National University of Singapore, and a Bachelor of Science in Physics from Nanyang Technological University in Singapore, during which she developed her interest in understanding interaction patterns in complex systems.
Research interests
Many real-world systems in science and society can be described and studied as networks (or graphs), where nodes represent agents and edges capture interactions between two agents. Examples include online social networks, transaction networks in finance, and brain networks in neuroscience. Typical tasks on graphs involve finding important nodes, identifying well-connected communities, and revealing structures of interest. Alongside these data-driven challenges, we may also wish to develop models that describe how connections are formed---these may be used to explain and predict interactions between the agents.
Some recent work on hypergraphs is motivated by the fact that the pairwise relationship is insufficient to reflect some real-world phenomena such as a meeting between a group of people. The hypergraph approach captures such interaction with a hyperedge that connects more than two nodes.
In the proposed project, time-stamped hypergraphs from the social sciences will be studied. The project will focus on developing, analysing and evaluating algorithms to characterise hypergraphs in a time-dependent setting.