Geometric Learning & Uncertainty in Edinburgh

Institute for Machine Learning · School of Informatics · University of Edinburgh

We work on two primary research directions—separately, and where they meet:

Geometric learning: exploiting the structure of data (e.g., symmetries), learning on structured domains (e.g., graphs, manifolds) (to achieve data efficiency and respect constraints)
Uncertainty quantification: building models that quantify uncertainty associated with their own predictions (to support model-based decision-making and safety-critical applications)

Where these directions meet

Current focus: Leveraging implicit structure of data to build uncertainty-aware models for complex, high-dimensional settings.
Established: Geometric Gaussian processes—probabilistic models on non-Euclidean domains such as manifolds and graphs.

Additional research directions: Gaussian Processes AI for Science Machine Learning Theory