Table 2 Model and benchmark test performance on surface atoms only in terms of Pearson’s correlation coefficient

From: Towards atom-level understanding of metal oxide catalysts for the oxygen evolution reaction with machine learning

surface per-site property

per-site CGCNN

per-site PAINN

bond-valence method

element average in corresponding bulk

Metal d-band center

0.980

0.975

0.512

0.670 ± 0.025

O 2p-band center

0.955

0.924 ± 0.007

0.493

0.582

magnetic moment

0.954

0.969

--

0.894

Bader charge

0.995 ± 0.000

0.992

0.938

--

  1. Errors are calculated by running 3 experiments with random weight initializations and test/train splits. The best performing model is highlighted in bold for each property. Element average in corresponding bulk was not available for Bader charges. The bond-valence method for predicting magnetic moment failed due to lack of sufficient fitting.
Back to article page