Table 1. NN training and prediction for desorption simulation results.
Sorption system | Model | Training MSE a (×10–4) | Test MSE a (×10–4) | Simulation variance (×10–4) |
MFI-C5-W | SorbNet | 3.8 ± 0.5 | 9.3 ± 0.9 | 2.8 |
Shallow | 4.3 ± 0.2 | 9.4 ± 0.4 | ||
MFI-C4-W | SorbNet | 2.5 ± 0.4 | 4.5 ± 0.7 | 1.9 |
Shallow | 4.1 ± 0.5 | 7.2 ± 1.1 | ||
MFI-C5-E | SorbNet | 1.9 ± 0.4 | 5.6 ± 0.9 | 0.7 |
Shallow | 2.7 ± 0.3 | 7.5 ± 0.7 | ||
LTA-C5-W | SorbNet | 2.8 ± 0.6 | 7.6 ± 1.6 | 1.4 |
Shallow | 3.8 ± 0.2 | 9.6 ± 0.5 |
aMean square errors were evaluated using the averaged simulation result as the true value. Standard deviations were measured from 8 training runs.