Table 2 Summary statistics for DAs for all investigated models.
From: Identifying domains of applicability of machine learning models for materials science
Global (test set) | DA (validation set) | DA (identification set) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
MAE | 95AE | R | cov | MAE | 95AE | R | cov | MAE | 95AE | R | |
MBTR | 14.2 | 54.1 | 0.83 | 44 (6) | 7.6 (1.5) | 18.8 (2.9) | 0.88 (0.03) | 44 (1) | 7.6 (0.3) | 20.7 (0.2) | 0.89 (0.01) |
SOAP | 14.1 | 51.0 | 0.84 | 78 (3) | 11.7 (1.8) | 36.6 (10.8) | 0.85 (0.01) | 76 (1) | 11.9 (0.4) | 37.8 (2.0) | 0.85 (0.00) |
n-gram | 14.7 | 51.1 | 0.83 | 52 (5) | 10.2 (0.9) | 32.6 (2.6) | 0.86 (0.02) | 54 (1) | 10.3 (0.2) | 35.5 (1.0) | 0.86 (0.00) |
Atomic | 65.5 | 154.5 | 0.24 | 85 (1) | 60.2 (7.8) | 141.6 (28.5) | 0.25 (0.09) | 85 (0) | 63.3 (1.5) | 153.9 (5.5) | 0.25 (0.02) |
