Research Article
Instructor Activity Recognition through Deep Spatiotemporal Features and Feedforward Extreme Learning Machines
Table 6
Performance comparison of various variants of ELM at a different combination of the parameter on MuHAVi-Uncut dataset.
| | Dataset | Hidden nodes
L | Regularization parameters | Accuracy | | γ | Cost
C1 | Scatter Matrix C2 | ELM | RELM | MCV-ELM | MVELM |
| MuHAVi-Uncut
(LOCO validation scheme) | 4096 | 10−7 | 2−7 | 10−7 | 75.11% | 75.11% | 58.11% | NAN | | 10−6 | 2−6 | 10−6 | 82.04% | 82.04% | 59.76% | NAN | | 10−5 | 2−5 | 10−5 | 72.58% | 72.58% | 60.27% | 24.49% | | 10−4 | 2−4 | 10−4 | 72.96% | 72.96% | 63.69% | 65.79% | | 10−3 | 2−3 | 10−3 | 72.37% | 72.37% | 70.18% | 73.84% | | 10−2 | 2−2 | 10−2 | 73.34% | 73.34% | 73.26% | 73.52% | | 10−1 | 2−1 | 10−1 | 69.22% | 69.22% | 73.72% | 73.84% | | 100 | 20 | 100 | 72.55% | 72.55% | 73.61% | 73.61% | | 101 | 21 | 101 | 72.16% | 72.16% | 73.00% | 73.00% | | 102 | 22 | 102 | 71.16% | 71.16% | 73.74% | 73.74% | | 103 | 23 | 103 | 72.73% | 72.73% | 74.75% | 74.75% | | 104 | 24 | 104 | 73.28% | 73.28% | 73.39% | 73.39% | | 105 | 25 | 105 | 73.00% | 73.00% | 74.22% | 74.23% | | 106 | 26 | 106 | 71.72% | 71.72% | 73.33% | 73.34% | | 107 | 27 | 107 | 72.91% | 72.91% | 72.61% | 72.61% |
|
|
