Enhancing Industrial Wireless Communication Security Using Deep Learning Architecture-Based Channel Frequency Response

Table 5

Training information and accuracy comparison between DL models.

Models	CNN	LSTM	RBMs
Number of trainable parameters	2,138,056	8,853,576	2,096,384
Training time per epoch (s)	3	5	6
Training time (s)	300	400	625
Prediction time (µs/sample)	4	1	9
Prediction rate	0.9960	1	0.8665
Prediction loss	0.0188	0.00005	1.9056