Table 3 Mean absolute errors (MAEs) of GNNs trained on Matbench dataset

From: Materials Graph Library (MatGL), an open-source graph deep learning library for materials science and chemistry

Model

Eform (eV atom-1)

log(KVRH) (log(GPa))

log(GVRH) (log(GPa))

EG (eV)

MEGNet

0.015/0.037/0.037

0.033/0.063/0.075

0.046/0.085/0.090

0.072/0.213/0.220

M3GNet

0.007/0.020/0.020

0.039/0.054/0.065

0.032/0.081/0.091

0.032/0.160/0.170

TensorNet

0.008/0.024/0.024

0.031/0.054/0.060

0.046/0.082/0.090

0.043/0.163/0.177

SO3Net

0.008/0.022/0.022

0.035/0.052/0.060

0.031/0.079/0.083

0.033/0.169/0.180

  1. Calculated MAEs of formation energy Eform, Voigt-Reuss-Hill bulk KVRH and shear modulus GVRH as well as bandgap EG with MEGNet, M3GNet, TensorNet and SO3Net. The numbers are reported in the order of training/validation/test MAEs. The dataset was divided into training, validation, and test sets with a split ratio of 0.9, 0.05, and 0.05, respectively.
Back to article page