Hello Yash,

I would like to ask you a question regarding your implementation (link attached) to the article- "Variational Autoencoders with Jointly Optimized Latent Dependency Structure".
According to the code, you calculated the KL for each node using normal gaussian prior with zero mean and unit variance, while the paper calculates the KL with another learned prior, which depends on the gates, c. as you can see in equation 8:

Appreciate your help.
Thanks.