Algorithm 1: Training of the Stacked CBN-VAE model.

1:Input: A mini-batch of training data set S, the number of train iteration iter, the learning rate α

2:  While i < iter do

3:  Compute the reconstructed output through forward propagation

4:  Compute the model loss according to (3)

5:  Compute the gradient of parameters according to (4)

6:  Update parameters according to (5)

7:  i + +

8:end