Research

• Extreme Value Theory

My main research area is the study of extreme values. In my PhD I dealt with the study of the spectral measure for regularly varying random vectors. I developed an equivalent notion called sparse regular variation which relies on the Euclidean projection onto the simplex.

• High-dimensional Learning and Clustering via model selection

I developed a procedure to detect the most relevant directions of the space which gather extreme values. This approach relies on model selection, similarly to Akaike's approach.

I try to adapt some existing clustering methods to detect the extremal trends in a data set. An algorithm called MUSCLE for MUltivariate Sparse CLustering for Extremes manages to exhibits clusters of directions that are likely to be large simultaneously.

Besides, I am also working on graphical models for extremes. These models rely on hidden regular variation and provide the dependence structure of extremal clusters of directions.

• Extremal index

During my postdoc, I study the so-called extremal index which gives information on the extremal clusters of a stationnary regularly varying time series. The goal is to establish several characterizations of this index and to apply them on some standard examples.

• Subexponential distributions

My postdoc is also dedicated to develop the notion of multivariate subexponential distributions. The aim is to extend ideas from multivariate regular variation to time series models with lighter tails.

• Research with World Health Organization (WHO)

During my postdoc I have the opportunity to collaborate with researchers from WHO. The project aims at using time series analysis to model antibiotic resistance.