Date : Sept. 6, 2023, 1:30 p.m. - Room :Amphi Bruno Garcia

Nonlinear Experimental Design

Stefan Körkel - Faculté d'Informatique de l'OTH Regensburg

First I will give a short introduction to OTH Regensburg and my teaching in optimization. Main topic of this talk is an introduction to my field of research. It deals with the validation of nonlinear process models. We consider models based on natural laws, e.g. differential equation initial value problems. To estimate the values of the model parameters, nonlinear least squares problems have to be solved. Due to the randomness of the data, the estimates are afflicted with statistical uncertainty. This can be described by the variance-covariance matrix and depends on the layout and controls of the experiments. To achieve best significance, we formulate and solve nonlinear experimental design problems. I discuss our approach and software package and illustrate the methods by application to an example from chemical reaction kinetics.