Skip to main content

Table 6 Cohen’s \(\kappa \) (average ± standard deviation) obtained for each algorithm and source when re-training additional layers of CNN

From: Comparison of deep transfer learning algorithms and transferability measures for wearable sleep staging

 

CiCC

ISRUC

MASS

SHHS

WSC

MrOS

Head Re-train

0.704 ± 0.099

0.692 ± 0.098

0.703 ± 0.098

0.694 ± 0.099

0.704 ± 0.098

0.702 ± 0.100

Subspace alignment

0.364 ± 0.003

0.389 ± 0.003

0.326 ± 0.003

0.404 ± 0.003

0.298 ± 0.004

0.305 ± 0.004

Per-Class CORAL

0.691 ± 0.058

0.657 ± 0.118

0.691 ± 0.142

0.684 ± 0.162

0.681 ± 0.180

0.681 ± 0.200

CORAL

0.718 ± 0.062

0.696 ± 0.222

0.704 ± 0.303

0.693 ± 0.362

0.702 ± 0.402

0.700 ± 0.438

DDC

0.522 ± 0.000

0.535 ± 0.001

0.469 ± 0.002

0.525 ± 0.002

0.516 ± 0.002

0.498 ± 0.003