Detecting Malware with an Ensemble Method Based on Deep Neural Network

Table 3

The list of LSTM network parameters.

Model parameters	LSTM
Maximum iterations	6.40 + 04
Weights initialization
Truncated BPTT length	120
Batch training samples	30
Initial learning rate	2.00 − 03
Dropout probability	0.5
Gradient regularization factor	10
Activation function	Tanh
Optimization algorithm	Adam
Propagation direction of time series	One-way
Hidden layers	2
Hidden nodes	185