Component Prediction of Antai Pills Based on One-Dimensional Convolutional Neural Network and Near-Infrared Spectroscopy

Table 1

Network structure parameters of 1DCNN.

Network layer	Antai pills
	Input	Output	Parameter
Gaussian noise	(None, 800)	(None, 800)	t = 0.05
Reshape	(None, 800)	(None, 800, 1)
Conv1	(None, 800, 1)	(None, 769, 8)	k = 8, s = 32, a = “relu”
BN1	(None, 769, 8)	(None, 769, 8)
Conv2	(None, 769, 8)	(None, 738, 16)	k = 16, s = 32, a = “relu”
BN2	(None, 738, 16)	(None, 738, 16)
Conv3	(None, 738, 16)	(None, 707, 32)	k = 32, s = 32, a = “relu”
BN3	(None, 707, 32)	(None, 707, 32)
Flatten	(None, 22624)	(None, 22624)
Dropout	(None, 34944)	(None, 34944)	r = 0.5
FC	(None, 128)	(None, 128)	d = 128, a = “relu”
Output	(None, 1)	(None, 1)	d = 1, a = “linear”