-------------------------------------------------------------
Klaus-Robert Müller
TU Berlin, Korea University, MPII
The talk starts by gently introducing selected machine learning (ML) concepts useful for analysing data from atomistic simulations, namely kernel methods and deep learning. Based on this and if time permits, two applications of ML usage are presented (1) ML for predicting quantum mechanical properties across chemical compound space and (2) ML for molecular dynamics. Finally, the importance of understanding of ML models obtained from training on data of atomistic simulations is stressed.