Table 1 Performance of IRNet (AI) in predicting formation energy from material structure.
From: Moving closer to experimental level materials property prediction using AI
Model name | Model for TL | Training data | DFT test set | EXP test set | ||||
---|---|---|---|---|---|---|---|---|
Name | Training set size | Validation set size | Size | MAE (eV/atom) | Size | MAE (eV/atom) | ||
IRNet-JARVIS | – | JARVIS | 20388 | 2224 | 3063 | 0.144 | 137 | 0.147 |
IRNet-MP | – | MP | 101716 | 11224 | 12471 | 0.097 | 0.097 | |
IRNet-OQMD | – | OQMD | 352711 | 39142 | 43448 | 0.042 | 0.120 | |
IRNet-EXP | – | EXP | 522 | 28 | NA | NA | 0.327 | |
IRNet-JARVIS-EXP | IRNet-JARVIS | 0.087 | ||||||
IRNet-MP-EXP | IRNet-MP | 0.078 | ||||||
IRNet-OQMD-EXP | IRNet-OQMD | 0.064 |