24.05.2017 um 17:15 Uhr in Raum 69/125
Prof. Dr. Thomas Kahle (Magdeburg)
Designing experiments for non-linear regression is hard because the optimal experiment to gain information about the unknown parameters depends on the unknown parameters. We pursue a geometric approach to this chicken and egg problem: In parameter space, we describe regions of optimality of certain experimental designs. It turns out that these regions are often semi-algebraic, that is, described by polynomial inequalties. Therefore the methods from computational logic and real algebraic geometry can be applied.