An Empirical Study on GAN-Based Traffic Congestion Attack Analysis: A Visualized Method

Algorithm 1

Iterative training of CycleGAN.

Input: original state image set , target congestion image set .

Output: trained model with optimized parameters.

Initialization. initialize network parameters , , , , learning rate , , , number of training iterations .

1: while and has not converged do

2: for to do

3:  //Forward cycle

4:  Generate fake image and recommend image .

5:  Calculate , loss.

6:  Update the gradient of ,:

7:  Discriminate fake image and real image.

8:  Calculate loss.

9:  Update the gradient of :

10:  //Backward cycle

11:  Generate fake image and recommend image .

12:  Calculate , loss.

13:  Update the gradient of ,:

14:  Discriminate fake image and real image.

15:  Calculate loss.

16:  Update the gradient of :.

17: end for

18:end while