Sébastien Destercke (CNRS) ran a Seminar@SystemX on the following topic « Imprecise data in machine learning: challenges and opportunities», on may 11 from 11:00 a.m to 12:15 p.m. at IRT System X ( Amphitheater 33 , building 862 , CEA, Nano-INNOV) and by the webinar.
Click here to connect to the webinar
This event is organized in collaboration with scientific seminars of Confiance.ai.
Resume:
This talk concerns the general problem of modelling and handling uncertainty, imprecision and lack of information in machine learning. While such imprecision can also be integrated in the model (to provide robust inferences) or in the inferences (to provide more reliable predictions), I will be primarily interested in how to model and handle imperfect and uncertain data. In particular, after discussing the nature of data uncertainty and how it can be modelled and obtained, I will present various challenges regarding the adaptations of learning procedures to such data, and will finish by discussing in which settings using uncertain data can provide us some advantage in the learning procedure.
Biography:
Sebastien Destercke graduated in 2004 as an engineer from the Faculté Polytechnique de Mons in Belgium. In 2008, he earned a Ph.D. degree in computer science from Université Paul Sabatier, in Toulouse (France). He briefly worked in the French agricultural research centre working for international development, before becoming a CNRS researcher in the Heudiasyc Laboratory, in Compiègne. His main research interests are in the fields of decision making and uncertainty reasoning (modeling, propagating, learning) with imprecise probabilistic models. He is currently the head of the Heudiasyc AI team, and the holder of the UTC SAFE AI chair.
