Dresden 2020 – scientific programme

The DPG Spring Meeting in Dresden had to be cancelled! Read more ...

Parts | Days | Selection | Search | Updates | Downloads | Help

CPP: Fachverband Chemische Physik und Polymerphysik

CPP 104: Topical Session: Data Driven Materials Science - Machine Learning Applications (joint session MM/CPP)

CPP 104.5: Talk

Thursday, March 19, 2020, 18:30–18:45, BAR 205

Bayesian modeling for potential energy surface minimization — •Estefania Garijo del Rio, Sami Juhani Kaapa, and Karsten Wedel Jacobsen — CAMD, Department of Physics, Technical University of Denmark, 2800 Kgs. Lyngby, Denmark

Computational simulations using electronic structure methods of materials and molecules require the (meta-)stable structure of the system under investigation to be known. In the absence of experimental structural data, the usual procedure is to use quantum chemistry codes together with some optimization algorithm to find successive approximations of a (local) minimum of the potential energy surface under the Born-Oppenheimer approximation. In this context, methods that incorporate machine learning surrogate models that are built on the fly to reduce the number of evaluations have recently gained popularity. Here, we explore and compare how different choices for the kernels can affect the performance of the optimization when Gaussian process regression is used to fit energies and forces.