Publication

Back to overview

Machine learning the computational cost of quantum chemistry

Type of publication Peer-reviewed
Publikationsform Original article (peer-reviewed)
Author Heinen Stefan, Schwilk Max, von Rudorff Guido Falk, von Lilienfeld O Anatole,
Project Sampling chemical space with alchemical perturbation theory
Show all

Original article (peer-reviewed)

Journal Machine Learning: Science and Technology
Volume (Issue) 1(2)
Page(s) 025002 - 025002
Title of proceedings Machine Learning: Science and Technology
DOI 10.1088/2632-2153/ab6ac4

Open Access

URL http://doi.org/10.1088/2632-2153/ab6ac4
Type of Open Access Publisher (Gold Open Access)

Abstract

Computational quantum mechanics based molecular and materials design campaigns consume increasingly more high-performance computer resources, making improved job scheduling efficiency desirable in order to reduce carbon footprint or wasteful spending. We introduce quantum machine learning (QML) models of the computational cost of common quantum chemistry tasks. For 2D nonlinear toy systems, single point, geometry optimization, and transition state calculations the out of sample prediction error of QML models of wall times decays systematically with training set size. We present numerical evidence for a toy system containing two functions and three commonly used optimizer and for thousands of organic molecular systems including closed and open shell equilibrium structures, as well as transition states. Levels of electronic structure theory considered include B3LYP/def2-TZVP, MP2/6-311G(d), local CCSD(T)/VTZ-F12, CASSCF/VDZ-F12, and MRCISD+Q-F12/VDZ-F12. In comparison to conventional indiscriminate job treatment, QML based wall time predictions significantly improve job scheduling efficiency for all tasks after training on just thousands of molecules. Resulting reductions in CPU time overhead range from 10% to 90%.
-