Rolling Bearing Fault Diagnosis Based on Stacked Autoencoder Network with Dynamic Learning Rate

Table 3

The results of different learning rates.

Different learning rates	The number of iterations required to converge or reach the number of iterations	Time required to converge or reach the number of iterations (s)	Training error value in convergence
0.2	270	437.696	0.2868
0.1	401	667.442	0.2346
0.01	600	996.857	0.4909
AdaDec	450	705.688	0.2014
Dynamic	336	549.325	0.1504