TY - JOUR
T1 - Capturing intensive and extensive DFT/TDDFT molecular properties with machine learning
AU - Pronobis, Wiktor
AU - Schütt, Kristof T.
AU - Tkatchenko, Alexandre
AU - Müller, Klaus Robert
N1 - Publisher Copyright:
© 2018, The Author(s).
PY - 2018/8/1
Y1 - 2018/8/1
N2 - Machine learning has been successfully applied to the prediction of chemical properties of small organic molecules such as energies or polarizabilities. Compared to these properties, the electronic excitation energies pose a much more challenging learning problem. Here, we examine the applicability of two existing machine learning methodologies to the prediction of excitation energies from time-dependent density functional theory. To this end, we systematically study the performance of various 2- and 3-body descriptors as well as the deep neural network SchNet to predict extensive as well as intensive properties such as the transition energies from the ground state to the first and second excited state. As perhaps expected current state-of-the-art machine learning techniques are more suited to predict extensive as opposed to intensive quantities. We speculate on the need to develop global descriptors that can describe both extensive and intensive properties on equal footing.
AB - Machine learning has been successfully applied to the prediction of chemical properties of small organic molecules such as energies or polarizabilities. Compared to these properties, the electronic excitation energies pose a much more challenging learning problem. Here, we examine the applicability of two existing machine learning methodologies to the prediction of excitation energies from time-dependent density functional theory. To this end, we systematically study the performance of various 2- and 3-body descriptors as well as the deep neural network SchNet to predict extensive as well as intensive properties such as the transition energies from the ground state to the first and second excited state. As perhaps expected current state-of-the-art machine learning techniques are more suited to predict extensive as opposed to intensive quantities. We speculate on the need to develop global descriptors that can describe both extensive and intensive properties on equal footing.
UR - http://www.scopus.com/inward/record.url?scp=85051125521&partnerID=8YFLogxK
U2 - 10.1140/epjb/e2018-90148-y
DO - 10.1140/epjb/e2018-90148-y
M3 - Article
AN - SCOPUS:85051125521
VL - 91
JO - European Physical Journal B
JF - European Physical Journal B
SN - 1434-6028
IS - 8
M1 - 178
ER -