V okviru 335. doktorskega seminarja Grajeno okolje bo imel(-a) prof. Hermann Matthies, Carl-Friedrich-Gauss faculty, TU Braunschweig, Nemčija predavanje z naslovom:
Reduced-Order Models and Conditional Expectation
Systems may depend on uncertain parameters. This is the “Leitmotiv” for the following discussion. A reduced-order model is produced from the full-order model through projection onto a low-dimensional manifold. The reduction produces a function mapping the parameters to the manifold. One wants to examine the relation between the full and the reduced state for parameter values. In the machine learning, a function mapping the parameter set to the image space is learned from a training set. This set may be seen as a sample from some probability distribution, and thus the training is a computation of the expectation. This is also a special case of Bayesian updating. A combined view of these methods is possible.
Hermann G. Matthies is an Emeritus Professor in the Carl-Friedrich-Gauss faculty at TU Braunschweig. He has first degree from TU Berlin, where he studied mathematics, computer science, physics, and engineering, and PhD in mathematics at MIT. He worked for an extended period in industry in different places, where he was concerned with the development of algorithms for finite element analyses and plasticity, and their application in engineering and wind energy. He is the founding director of the Institute of Scientific Computing at the TU Braunschweig. He has received the IACM Fellow Award and the Gay-Lussac Humboldt Prize of the French Academy of Sciences. He is a full member of the academy “Braunschweigische Wissenschaftliche Gesellschaft”.
Predstavitev bo v petek, 3. 10. 2025, ob 12.00 v Svečani dvorani (FGG).
Vljudno vabljeni.