Fig. 10: Example of how structure motifs can be extracted from a starting model with 4 metal atoms coordinated to oxygen and used as input to the GBDT model. | npj Computational Materials

Fig. 10: Example of how structure motifs can be extracted from a starting model with 4 metal atoms coordinated to oxygen and used as input to the GBDT model.

From: Extracting structural motifs from pair distribution function data of nanostructures using explainable machine learning

Fig. 10

a The metal atoms are permuted randomly by creating an array of zeros and ones, where 0 refers to a deleted atom and 1 refers to an atom that is kept in the structure. Oxygen atoms are removed if they do not bond to any metal atoms within a distance threshold that is set by the user. Note that the metal atoms (blue) are slightly distorted from the centre of the octahedra. b Example of how the four structures from panel a and Fig. 1 are given as input to the GBDT model which predicts the Rwp value.

Back to article page