Table 5.

Classification results of Deep CORAL and DDC (fc5^a).

	Deep CORAL						DDC
S-Tb	S c			T d			S			T
	Tre(%)	Valf(%)	Teg(%)	Tr(%)	Val(%)	Te(%)	Tr(%)	Val(%)	Te(%)	Tr(%)	Val(%)	Te(%)
01–02	100.00	87.50	80.00	58.48	58.24	56.25	100.00	87.50	86.67	60.23	58.82	56.25
01–03	100.00	93.75	86.67	80.63	84.21	83.33	100.00	87.50	86.67	80.10	84.21	83.33
01–04	100.00	87.50	93.34	65.82	73.68	73.68	100.00	81.25	86.67	70.41	73.68	73.68
02–01	100.00	82.35	75.00	75.95	81.25	73.33	92.40	82.35	68.75	71.52	75.00	73.33
02–03	100.00	76.47	75.00	76.44	77.78	78.95	100.00	76.47	75.00	78.53	73.68	72.22
02–04	100.00	82.35	75.00	72.45	68.42	68.42	95.32	82.35	75.00	71.43	73.68	57.89
03–01	100.00	89.47	83.33	82.91	87.50	80.00	100.00	89.47	77.78	82.91	87.50	73.33
03–02	100.00	89.47	83.33	74.85	70.59	68.75	100.00	94.74	83.33	73.68	76.47	62.50
03–04	100.00	94.74	88.89	84.69	89.47	73.68	100.00	94.74	88.89	73.98	78.95	73.68
04–01	100.00	89.47	94.74	71.52	73.33	68.75	100.00	100.00	94.74	69.62	68.75	66.67
04–02	100.00	94.74	94.74	74.85	68.75	58.82	100.00	84.22	94.74	66.08	64.71	62.50
04–03	100.00	89.47	89.47	83.77	84.21	77.78	100.00	100.00	89.47	82.20	78.95	77.78

^afc5 means that the penultimate fully connected layer of the convolutional neural network (CNN) was used for calculating domain loss (CORAL loss and MMD loss, as shown in Figures 3 and 4); ^bS-T means the source domain-the target domain; ^cS means the source domain; ^dT means the target domain; ^eTr means the training set of the target domain; ^fVal means the validation set of the target domain; ^gTe means the test set of the target domain. The bold values means the accuracy rate of the test set was beyond 73%.