Hi,
Is this equation really right?
Should’t we use the log-variance in the exponent as it states in tensorflow documentation?
Thanks!
Where is the equation used exactly so I can have a look at it?
Variational Autoencoders
Both, in videos (“Sampling Layer and Encoder”) and in the lab (“C4_W3_Lab_1_VAE_MNIST.ipynb”) can be found
I see, i think as long as you are in the latent space it doesnt matter much if you use either, in tensorflow they use logvar because its more stable as far as hindering negative value is concerned. Check