Comment by p1esk

Basic autoencoders are conceptually easy, but try to understand how the VAE KL loss was derived. I remember I spent two days reading https://arxiv.org/abs/1906.02691 and I still didn't quite get it.