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

Table 2 Cross-validation Cohen’s \(\kappa \) (average ± standard deviation) obtained using each algorithm and source dataset to re-train CNN

	CiCC	ISRUC	MASS	SHHS	WSC	MrOS
Head Re-train	0.689 ± 0.086	0.659 ± 0.093	0.652 ± 0.092	0.661 ± 0.087	0.669 ± 0.085	0.667 ± 0.115
Subspace alignment	0.332 ± 0.183	0.279 ± 0.151	0.146 ± 0.139	0.230 ± 0.140	0.183 ± 0.137	0.153 ± 0.136
Per-Class CORAL	0.690 ± 0.077	0.660 ± 0.090	0.666 ± 0.079	0.658 ± 0.083	0.663 ± 0.081	0.681 ± 0.080
CORAL	0.689 ± 0.085	0.663 ± 0.084	0.672 ± 0.081	0.655 ± 0.084	0.669 ± 0.080	0.679 ± 0.086
DDC	0.660 ± 0.109	0.686 ± 0.088	0.682 ± 0.111	0.703 ± 0.098	0.663 ± 0.118	0.704 ± 0.085

ISSN: 1475-925X