We are pleasure to introduce Kevin Colin, to his conference :
Title: Bayesian estimation of kinetics of mammalian cells: Gaussian process and parametric approaches.
Abstract: In this seminar, we consider the problem of kinetic identification of mammalian cells using Monod functions which are parametric rational functions depending on the concentrations of the cell metabolites. Because of the non-convex properties of these functions, local minima may be expected with non-linear least-squares algorithms and a good initialization relatively close to the global optimum is required. We propose in this presentation two approaches based on Bayesian estimation in order to compute a suitable estimate: Gaussian process regression and parametric Bayesian estimation. In the Gaussian process regression method, we develop a new kernel tailored for the estimation of Monod functions. For the parametric Bayesian estimation, we present a new Markov chain Monte Carlo sampling technique. Both methods are compared in a numerical example and we will also apply them on real-life data of Chinese hamster ovary cells. Some generalization of both techniques to other types of non-linear estimation problem will be also discussed.
Office : Salle Jacques Bordet
Date/heure
Date(s) - 18 Nov 2022
14 h 00 min - 17 h 00 min
Emplacement
LAGEPP Salle Jacques Bordet
Catégories
