Table 8 Correlations of each transferability measure with CNN accuracy when re-training additional layers of CNN
Measure | Head Re-train, \(n = 144\) | CORAL, \(n = 144\) | Per-Class CORAL, \(n = 144\) | SA, \(n = 144\) | DDC, \(n = 144\) | Overall, \(n = 720\) |
|---|---|---|---|---|---|---|
LEEP | − 0.14 | − 0.19* | − 0.12 | 0.08 | 0.03 | − 0.07 |
H-score | 0.17* | 0.35*** | 0.31*** | − 0.03 | − 0.07 | 0.14*** |
Hypothesis margin | 0.05 | 0.01 | 0.10 | − 0.15 | − 0.05 | − 0.01 |
Silhouette score | − 0.05 | − 0.19* | − 0.09 | − 0.09 | 0.00 | − 0.08 |
MMD, \(\gamma = 0.1\Gamma \) | 0.14 | 0.04 | − 0.03 | − 0.19* | − 0.01 | − 0.01 |
MMD, \(\gamma = 1\Gamma \) | 0.14 | 0.12 | 0.03 | − 0.14 | − 0.02 | 0.03 |
MMD, \(\gamma = 10\Gamma \) | 0.06 | 0.08 | 0.01 | − 0.16 | 0.04 | 0.00 |
TDAS, \(\epsilon = 0.1m\) | − 0.07 | − 0.05 | − 0.02 | − 0.05 | 0.01 | − 0.03 |
TDAS, \(\epsilon = 1m\) | 0.24** | 0.42*** | 0.28*** | 0.15 | 0.06 | 0.15*** |
TDAS, \(\epsilon = 10m\) | − 0.10 | 0.24** | − 0.09 | 0.10 | 0.01 | − 0.06 |