Table 6.

The influence of main hyperparameter settings on model recognition accuracy.

Hyperparameter	Variation Range	Recognition Accuracy (%)
		Epoch = 2	Epoch = 5	Epoch = 10
learning rate	[5 × 10−5, 5 × 10−3]	98.93	99.23	99.86
	[1 × 10−4, 1 × 10−2]	98.56	98.99	99.23
	[5 × 10−4, 5 × 10−2]	85.5	95.13	98.87
	[1 × 10−3, 1 × 10−1]	38.37	68.22	86.57
batch size	BZ = 4	67.5	92.33	96.12
	BZ = 10	81.5	96.37	98.22
	BZ = 15	96.5	97.86	99.51
	BZ = 20	97.5	99.21	99.72
kernel number	(16, 32, 64)	100	100	100
	(32, 64, 128)	99.92	99.86	100
	(64, 128, 256)	99.88	98.67	100
	(128, 256, 512)	98.65	99.12	100