EPS_en

Director : Benoîte DE SAPORTA

Research Topics :

The overall competencies of EPS focus on highly prominent themes in modern statistics: Teaching, Selection/choice of models, Statistics in high dimensions. The research completed within the teams leads to theoretical developments and practical applications. Cooperations exist with UM laboratories (ISEM, LIRMM, etc.), other Montpellier research establishments (CHU, INRAE-SupAgro, CIRAD, etc.) and private organizations (SANOFI, SCHLUMBERGER, THALES, etc.). The team, very active in the structures created by PIA (Labex Numev, IBC), is broken into three main sections lead by Jean-Noël Bacro, Benoîte de Saporta, and Ali Gannoun, which are further detailed below :

The researchers in the MEGADONC group work on statistics and large scale modeling. They process functional data with machine learning approaches and convolution kernel methods. These approaches are used in non-smooth optimization, classification, inverse problems, graphs, estimation processes, forecasting, and more. Beyond theory, (parameter choice, speed convergence, etc) many applications, in presence or absence of censoring, are proposed in domains as varied as medicine, agronomy, finance, and genomics.

The members of the research group Computational Methods and Modelization work on stochastic processes, Bayesian statistics, and process statistics. The topics worked on are numerous and varied with applications in ecology, the environment, epidemiology, and reliability. The research covers approximation, control, and simulation of probabilistic models of population dynamics, methodological developments of Bayesian statistics for large datasets, aggregate detection and inference of spatial processes, and random fields and spatialized data.  

The research group Modest combines professors and researchers working on modeling dependency structures. The topics studied are numerous and varied with applications in domains such as chemometry, environmental risks, security, and medico-social. Complex dependancy modeling is at the heart of the Modest’s activity and the principle research work covers modelling hidden structures, regularization, data mining, extremes of finite and infinite dimensions, and more. Between three and five annual meetings are organized for presentations and/or exchanges between group members.