Table 2.

Layers of the deep neural network.

MATLAB command	Description	Parameters
sequenceInputLayer	Inputs 1D data into network	250 features
lstmLayer	Long short-term memory (LSTM) layer; learns long-range dependencies between features	125 hidden units
fullyConnectedLayer	Multiplies with weight matrix and adds bias	50 output units
dropoutLayer	Sets elements to zero with given probability	p = 0.5
fullyConnectedLayer	Multiplies with weight matrix and adds bias	1 output unit
regressionLayer	Outputs scalar prediction from network	−

The network employed for DNN regression consisted of six layers which were designated for using deep learning on regression problems.