Current Research Focus

This page is under construction. Our recent publications/preprints provide an idea of the current research in the group.

At a high level, the group's research interests include high-dimensional statistics and the mathematics of machine / deep learning. There is a strong focus on the following directions:

• Developing a comprehensive statistical and optimisation theory for unsupervised representation learning. You can find a broad idea in our perspective article [link]. Our interests cover statistical aspects of representation learning [link], optimisation and learning dynamics [link], and development of kernel methods with applications to interpretability [link]
• Characterisation of the statistical performance of graph neural networks (see our perspective paper [link]). Our focus is primarily on precise characterisation of generalisation error under graph models [link] and asymptotics of graph neural networks [link] [link]
• The interplay of optimisation and statistical generalisation, both in the context of high-dimensional statistics [link] and deep learning [link]