Sitzungen | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

FM: Fall Meeting

FM 90: Special Session: Quantum Physics for AI & AI for Quantum Physics

FM 90.4: Invited Talk

Freitag, 27. September 2019, 12:30–13:00, Audi Max

Response operators in Machine Learning: Response Properties in Chemical Space — •Anders Christensen — Institute of Physical Chemistry and National Center for Computational Design and Discovery of Novel Materials (MARVEL), Department of Chemistry, University of Basel, Klingelbergstrasse 80, CH-4056 Basel, Switzerland

This talk focuses on the use of response operators in machine learning models for properties of chemical compounds. The role of response operators is well-established in quantum chemistry in which they are used to calculate properties of chemical compounds using differential operators. The same response operators commonly used in quantum chemistry are here applied to a new machine learning model in order to increase its accuracy. Prediction errors for corresponding properties reach high accuracies for small training set sizes. For example, the learning rate of dipole moments is improved by a factor 20x compared to a similar model without operators. In addition, the prediction of vibrational normal modes and infrared spectra of small molecules demonstrates the applicability of this approach for chemistry. The presented operator-based approach is general and can in principle be applied to any machine learning model.