V. Vinod, D. Lyu, M. Ruth, U. Kleinekathöfer, P. R. Schreiner, and P. Zaspel. Predicting molecular energies of small organic molecules with multifidelity methods. J. Comput. Chem., 46: e70056, 2025. DOI: https://doi.org/10.1002/jcc.70056; also available as chemrxiv-2024-9zz16.
Multifidelity methods for quantum chemistry (QC) is an effective machine learning (ML) tool to reduce computational costs without compromising on model accuracy. In this work, V. Vinod et al. assess the efficiency of several multifidelity methods in predicting energies of small organic molecules with coupled cluster triples (CCSD(T)) accuracy. In addition to an analysis of time-cost and model accuracy, the trained multifidelity models are used to predict CCSD(T) energies for a collection of atmospherically relevant molecules and highly conjugated molecules with high accuracy. This work is associated with the SPP 2363 on “Utilization and Development of Machine Learning for Molecular Applications – Molecular Machine Learning funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under its special priority program scheme.