Table 3 Test MAE of different models for each of the materials properties for the prediction task of “Other materials properties”.

From: Improving deep learning model performance under parametric constraints for materials informatics applications

Dataset

Property

Size

AutoML

ElemNet

IRNet

BNet

BRNet

OQMD

Band gap (eV)

345,134

0.075

0.052

0.054

0.050

0.048

Stability (eV/atom)

345,134

0.113

0.051

0.047

0.045

0.043

Volume (A\(_3\)/atom)

345,134

21.02

19.56

20.09

17.92

16.91

AFLOWLIB

Density (grams/cm\(_3\))

234,299

0.556

0.227

0.186

0.184

0.176

Volume (A\(_3\)/atom)

234,299

1.001

0.690

0.611

0.603

0.588

Band gap (eV)

14,751

0.134

0.145

0.140

0.116

0.108

MP

Band gap (eV)

89,181

0.435

0.342

0.316

0.317

0.315

Density (grams/cm\(_3\))

89,181

0.446

0.373

0.373

0.349

0.344

Volume (A\(_3\)/lattice)

89,181

205.9

248.8

238.6

233.8

227.4

JARVIS

Gap OPT (eV)

17,924

0.345

0.294

0.300

0.265

0.260

Bulk modulus (GPa)

8199

13.46

11.56

11.71

11.79

10.63

Shear modulus (GPa)

8199

10.75

10.64

10.75

11.10

9.94

Gap TBMBJ (eV)

5287

0.630

0.544

0.526

0.483

0.497

  1. The lowest MAE values in each row are highlighted in bold.