6 chapters
VAE
Variational autoencoder
Filling in the Holes
Why latent gaps decode to noise, how the ELBO closes them, and the trick that makes VAEs trainable.
AI Cell · Resources
Self-contained courses that sit beside the weekly track. Open one when you want the derivation, not just the session.
VAE
Variational autoencoder
Why latent gaps decode to noise, how the ELBO closes them, and the trick that makes VAEs trainable.
GAN
Generative adversarial network
How a generator learns from a discriminator, why the loss is a game, and how mode collapse happens.