Bayes' Theorem

$p(X|Z)=\frac{p(Z|X)p(X)}{p(Z)}$

• (137) Evidence Lower Bound (ELBO) - CLEARLY EXPLAINED! - YouTube
  • You can think of the P(Z) term as the normalizing constant
    • this is VERY hard to calculate in high dimensions since it’s an intractable integral:
    • $\int p(z|x)p(x)dx$
• scratchpad (not sure if correct):
  • If Z is the hidden/latent variable, and X is observed variable
  • e.g. Z = the img is a cat
  • X = the pixels of the img