I am currently trying to figure out how D3VAE are working, but I can't seem to understand the network architecture given. The paper can be found here:
https://openreview.net/pdf?id=rG0jm74xtx
The things I don't understand in the picture can be found in the top right part (BVAE). As far as I understand z_n will be an output of size 2 while the residual-blocks will be of a much greater size. How can they be added, if shapes are mismatched? And they don't explain how "h" (far right in the picture) is calculated? Is it a randomly generated vector?
