Data-driven modeling
This research axis includes our preferred Machine learning applications, that have been instrumental in inspiring and validating the approaches developed in the former three directions. Applications in computational social sciences (e.g., in collaboration with Pôle Emploi and ENSAE; with INRAE) are related with fairness, causality, and most generally, dealing with highly sensitive data. Applications related to energy management (long-time application domain of choice of the team), population genetics, molecular biology, and a few others, all required some agility to map the specifics of the domain onto appropriate neural architectures, with growing emphasis on Graph Neural Networks.
These issues have been considered in particular in Guillaume Bied’s PhD (coll. INRAE and Pole Emploi); Benjamin Donnot and Balthazar Donon’s PhDs (coll. RTE; modelling and counter-factual operating of electrical networks, aka, n-1-safety); Théophile Sanchez’s PhD (coll. BioInfo-LISN; population genetics); Loris Felardos’ PhD (coll. IBBP; molecular dynamics).
… more to come
