A. Ch. Goeßmann (Fritz-Haber-Institut der Max-Planck-Gesellschaft)
Tuesday, November 13, 2018 - 13:30
Mohrenstr. 39, 10117 Berlin, Weierstraß-Hörsaal (Raum: 406), 4. Etage
Seminar Materialmodellierung
Accurate modeling of many-body systems like crystals requires to capture their quantum-mechanical nature at the atomic scale. The solution of the associated electronic structure problem is however illusional due to the number of variables, but we obtain certain properties by computational-demanding methods like density-functional theory. In this talk, I will discuss the potential of kernel-based machine learning to circumvent this computational bottleneck and predict crystal properties. A crucial preliminary step is the representation of crystals, which has to satisfy different conditions for the learning to perform optimally.
submitted by hohn (ina.hohn@wias-berlin.de, 030 20372591)