Table 2.

Classification results of the three datasets.

Site		Adult income dataset			Schwannoma dataset			eICU dataset
		Accuracy	AUROCa	Accuracy		AUROC	Accuracy		AUROC
Central
	Before VFLb	0.83	0.91	0.90		0.84	0.81		0.89
	After VFLc	0.82	0.90	0.82		0.84	0.80		0.88
	Differenced	–1.20	–1.10	–8.89		0	–1.23		–1.12
A
	Before VFL	0.81	0.89	0.82		0.81	0.70		0.72
	After VFL	0.77	0.83	0.78		0.86	0.70		0.72
	Difference	–4.94	–6.74	–4.88		+6.17	0		0
B
	Before VFL	0.81	0.90	0.76		0.82	0.73		0.80
	After VFL	0.77	0.83	0.78		0.83	0.72		0.79
	Difference	–4.94	–7.78	+2.63		+1.22	–1.37		–1.25
C
	Before VFL	0.67	0.73	0.48		0.60	0.55		0.57
	After VFL	0.76	0.83	0.62		0.71	0.56		0.57
	Difference	+13.43	+13.70	+29.17		+18.33	1.82		0

^aAUROC: area under the receiver operating characteristics curve.

^bVFL: vertical federated learning.

^cCorresponding to the latent representation of original data (central, A, B, or C) in the code layer.

^dThe difference is compared between AUROCs in classification tasks.