Table 3.

Performance of deep architectural classifier on testing set with reference (performance is determined by trained personalized classifier, testing set excluding the first 300 heartbeats).

	Predicted
		N	S	V	F	Total	SE (%)
True	N	37622	68	175	119	37984	99.0
	S	448	1143	7	3	1601	71.4
	V	106	0	2644	85	2835	93.3
	F	52	0	9	292	353	82.7
	Total	38228	1211	2835	499	42773	86.6
	PPV (%)	98.4	94.4	93.3	58.5	86.1	Accuracy = 97.5%
Accuracy = (TPN + TPS + TPV + TPF)/number of testing heartbeats