Leverhulme Research Fellowship success

Natalia Bochkina's fellowship will be concerned with developing mathematical foundations of Bayesian high dimensional models, including semi-parametric models and inverse problems under realistic assumptions that include potentially misspecified models, with the aim of designing statistical procedures with optimal frequentist properties.

Miguel de Carvalho's project will aim to develop novel statistical models to tackle the rising concern of cascading extreme events - like tsunamis followed by earthquakes or heatwaves sparking wildfires, which in turn lead to further losses. While EVT (Extreme Value Theory) provides a solid mathematical framework for dealing with extremes, it overlooks triggering events and their domino effects. The project will pioneer methods for modelling cascading extremes and will develop novel neural models to learn about these from data. By integrating EVT and AI, the proposed methods will enable extrapolation beyond observed data to capture and predict complex chains of extreme events.