Table 6.
Overview of results for individual modalities.
| Model | Labels | Difference | LoW |
|---|---|---|---|
| MLP | Individual | 0.1892 | 1.14 |
| S_MLP | Individual | 0.1642 | 0.985** |
| LIT | Individual | 0.1978 | 1.19 |
| S_LIT | Individual | 0.1681 | 1.09 |
| MLP | Group | 0.2616 | – |
| S_MLP | Group | 0.2616 | – |
| LIT | Group | 0.2484 | – |
| S_LIT | Group | 0.2304 | – |
| Modality | |||
| S_MLP | PPG | 0.1969 | 1.18 |
| GSR | 0.2160 | 1.30 | |
| NIRS | 0.2865 | 1.71 | |
| ET | 0.2159 | 1.30 | |
| S_LIT | PPG | 0.2400 | 1.44 |
| GSR | 0.1796 | 1.08* | |
| NIRS | 0.3188 | 1.91 | |
| ET | 0.2224 | 1.33 |
The first column contains the name of the model. The second column describes which labels were used. The third column reports the achieved mean difference between predicted vs. true label. The last column shows the mean difference in level of workload. An asterisk indicates the best performance in the unimodal setting; a double asterisk indicates the overall best performance.