Skip to main content
. 2021 Aug 16;121(16):10073–10141. doi: 10.1021/acs.chemrev.1c00022

Figure 8.

Figure 8

Hands-on example for atomistic GPR: learning the potential-energy surface of a single water molecule. (a) Structures in the dataset are defined by two coordinates: the asymmetric stretch coordinate ν′ and the bending coordinate ω; the sum of both bond lengths is fixed to 2 × 0.95 Å. (b) Target property to be represented by the model, spanning several eV because a very large range of distortions has been chosen for this toy example. (c) Error in the GPR-predicted molecular energy as a function of (ν′, ω). Stars indicate structures used for training; crosses indicate structures used as representative points of the sparse GPR model.