Statistical Modeling and Learning

Our work is organised around the following themes:

• Statistical learning. One of our aims is to develop learning methods capable of incorporating complex biological or physical knowledge. With this in mind, we are developing Bayesian methods (extension of ABC) as well as physics-guided deep learning methods for time-series data. In response to the development of sensors in digital agriculture, we are also conducting research into change-point detection and robust time-series analysis. Finally, we are carrying out research into community detection and, more generally, clustering, with a view to identifying important patterns in large datasets.
• Statistical expertise and modelling.  We carry out research into the analysis of functional data (curve data) to identify factors influencing plant growth or fermentation kinetics in collaboration with our partners. Furthermore, we are working on coupling generative models with deterministic models (e.g. coupling climate and crop models). More generally, we develop models for analyzing heterogeneous and aggregated data. 

Our areas of application include the study of biosolutions, the analysis of winemaking processes and, more generally, agroecology, as well as digital agriculture and livestock farming.

We pay particular attention to the development of R/Python packages and tutorials for biologists.

Permanent staff:

• Meili Baragatti
• Bertrand Cloez
• Bénédicte Fontez
• David Metivier
• Isabelle Sanchez
• Nicolas Verzelen (coordinator)