BMC Medical Informatics and Decision Making

Table 3 Model Performance: Accuracy, Epoch Number and Training Time

From: Atrial fibrillation classification based on convolutional neural networks

Model	Alex Net 1	Alex Net 2	Alex Net 3	Alex Net 4	Alex Net 5	Alex Net 6
Accuracy	0.9965	0.9960	0.9970^a	0.9945	0.9950	0.9900
Epoch #	32	61	43	54	51	37
Time (Sec)	163	185	89	110	103	76
Model	Residual 1–1	Residual 1–2	Residual 1–3	Residual 1–4	Residual 1–5	Residual 1–6
Accuracy	0.9975	0.9970	0.9980	0.9970	0.9980	0.9970
Epoch #	62	51	109	56	64	41
Time (Sec)	673	309	440	172	212	162
Model	Residual 2–1	Residual 2–2	Residual 2–3	Residual 2–4	Residual 2–5	Residual 2–6
Accuracy	0.9975	0.9990^b	0.9975	0.9975	0.9975	0.9940
Epoch #	104	50	41	58	30	28
Time (Sec)	896	253	167	177	93	87
Model	Residual 3–1	Residual 3–2	Residual 3–3	Residual 3–4	Residual 3–5	Residual 3–6
Accuracy	0.9970	0.9980	0.9950	0.9955	0.9935	0.9900
Epoch #	40	55	44	42	39	44
Time (Sec)	322	278	178	129	119	133
Model	Residual 4–1	Residual 4–2	Residual 4–3	Residual 4–4	Residual 4–5	Residual 4–6
Accuracy	0.9925	0.9935	0.9915	0.9905	0.9870	0.9590
Epoch #	26	41	44	43	68	85
Time (Sec)	158	166	134	89	139	173

^aBest network with the highest accuracy among Alex 1–6
^bBest network with the highest accuracy among Residual 1–1, …, 4–6

Back to article page

ISSN: 1472-6947

Contact us

General enquiries: journalsubmissions@springernature.com