An Intrusion Detection Method Based on Fully Connected Recurrent Neural Network

Algorithm 2

Weight fine-tuning algorithm.

	Input: the training sample was (x1,y1) (i = 1, 2, ..., m).
	Initialization: the initialization model parameter was θ = {Whx,Whh,Wyh,bh,by}
	Output: the fine-tuned model parameter was θ = {Whx,Whh,Wyh,bh,by}
(1)	For each sample xi, input a fully connected RNN, the output of xi was calculated by Algorithm 2.1
(2)	Calculate the cross-entropy L(y:) between the output value of each sample and the label value:
(3)	For each network model parameter θi in θ, calculate the partial derivative
(4)	Make the error propagate back along the network and update each network model parameter θi in θ:
(5)	If t = k, save the model parameters and the algorithm ends
(6)	If t< k, then t = t + 1, turn to 1.