Welcome to François Portier, Associate Professor of Statistics 

François Portier recently joined ENSAI as an associate professor of Statistics and researcher at CREST. Born and educated in Brittany, his academic career took him to UCLouvain and Télécom Paris. He is already collaborating on research projects with members of the ENSAI faculty.

François Portier teaches Machine Learning, Statistics and Computational Statistics classes for third-year and Master in Statistics for Smart Data students.

Back to Rennes

I am a typical product of Université de Rennes 1, where I did all my studies, including a PhD on parsimonious predictive models under the direction of Bernard Delyon.

François Portier left Rennes to join UCLouvain as a postdoctoral researcher, within the team of Ingrid Van Keilegom and Johan Segers. He worked on specific survival analysis models. In 2016, he became an assistant professor at Télécom Paris, where he taught and pursued his research until 2021.

François Portier joined ENSAI in September. Shortly after, he defended his HDR thesis (Habilitation à Diriger les Recherches), which acknowledges scientific achievement and the capacity to supervise doctoral research.

A native of Vannes, François Portier appreciates being back in Brittany and seizes every occasion to go sailing, a life-long passion.

Developing mathematical tools to study algorithms

“My research interests lie in mathematical statistics. What interests me the most is to describe algorithms from a mathematical angle: looking for the very fundamental properties that explain the success or the failure of certain algorithms. I used to work on algorithms from semi-parametrical statistics and high-dimensional statistics, however, since my time at Télécom Paris, my research areas have become more linked to machine learning.”

François Portier’s research is largely based on empirical process theory applied to different domains such as dimension reduction or more recently nearest neighbor based algorithms.

The proximity of fellow ENSAI and CREST researchers sharing similar interests is already leading to common projects. François Portier should collaborate with Adrien Saumard and Valentin Patilea in the future, namely on machine learning algorithms and model selection.

Together with Valentin Patilea, we will work on a more traditional component of statistics: model evaluation using goodness of fit test. We will try to build a measure that will give us an indicator for the validity of parametric models.

François Portier is also working on a project he developed while at Télécom Paris, together with Rémi Leluc, a PhD student there. The project, “Musketeer”, is an algorithm dedicated to the optimization of high-dimensional functions for large-scale learning. François Portier and his colleague use a stochastic gradient descent method linked to algorithm efficiency. As calculating the derivatives with respect to each variable takes time, “Musketeer” selects carefully which coordinate derivative to evaluate.

The algorithm is called Musketeer, because of the motto “one for all and all for one.” The coordinates of the gradient communicate, which makes the algorithm efficient.  One of them will do the calculation for the others and then share the information. They will in turn update and know if the coordinates were useful or not.

Welcome to ENSAI!



2010 to 2013: PhD at Université de Rennes 1

2013 to 2016: postdoctoral fellow at UCLouvain

2016: Assistant professor of Statistics at Télécom Paris.

2021: Associate professor of Statistics at ENSAI


Find out more about François Portier and research at ENSAI and CREST laboratory