Reliable Fault Diagnosis of Rotary Machine Bearings Using a Stacked Sparse Autoencoder-Based Deep Neural Network

Table 4

Specifications for the training and design parameters of the stacked autoencoders.

Input size (sample length)	Number of layers	Number of hidden nodes	Sparsity constraint	Transfer function
1024	3	100, 50, 25	0.15, 0.15, 0.15	Logistic sigmoid