Table 2 Statistics of predicted bandgaps by different DL algorithms without and with transfer learning for 6 × 6 supercell systems
From: Bandgap prediction by deep learning in configurationally hybridized graphene and boron nitride
MAE (eV) | MAEF | RMSE (eV) | RMSEF | R 2 | |
|---|---|---|---|---|---|
Without transfer learning | |||||
VCN | 0.15 | 15.7% | 0.18 | 20.6% | 0.5390 |
RCN | 0.12 | 13.6% | 0.15 | 19.4% | 0.5243 |
CCN | 0.13 | 14.5% | 0.16 | 19.8% | 0.5771 |
With transfer learning | |||||
VCN | 0.11 | 11.3% | 0.14 | 15.5% | 0.5958 |
RCN | 0.10 | 10.2% | 0.12 | 15.0% | 0.6525 |
CCN | 0.09 | 9.60% | 0.12 | 12.6% | 0.6642 |
