Table 7 Accuracy, \(\kappa \), and % of cases where each algorithm outperformed others when re-training additional layers of CNN
Algorithm | Average ± standard deviation % accuracy | Average ± standard deviation Cohen’s \(\kappa \) | % of cases where algorithm was best |
|---|---|---|---|
Head Re-train | 78.7 ± 6.4 | 0.700 ± 0.087 | 39.6 |
Subspace alignment | 54.2 ± 17.0 | 0.348 ± 0.216 | 2.1 |
Per-Class CORAL | 77.4 ± 6.9 | 0.681 ± 0.096 | 12.5 |
CORAL | 78.9 ± 6.4 | 0.702 ± 0.086 | 31.3 |
DDC | 63.2 ± 20.5 | 0.511 ± 0.24 | 13.2 |