When trained on 80% of QMOF Database v13, the testing set ML results for band gap prediction are as follows:
| Method | MAE (eV) |
|---|---|
| Mean of Dataset | 0.940 |
| Sine Matrix Eigenspectrum | 0.539 ± 0.004 |
| Stoichiometric-45 | 0.459 ± 0.005 |
| Stoichiometric-120 | 0.454 ± 0.006 |
| Orbital Field Matrix | 0.410 ± 0.004 |
| Average SOAP kernel | 0.344 ± 0.004 |
| CGCNN | 0.270 ± 0.011 |
You can download the corresponding features for QMOF Database v13 here.