CHAPTER 21 / 40

Variational Inference and Generative Models

3 lessons · Explore, experiment and understand.

DOJO PETYour little coding companion

DOJO / YOUR GUIDE

Let’s figure it out.

Ask me about this course or where to start.

Chapter lessons

3 lessons

21Variational Inference and Generative Models

0/33 lessons

Why this matters in ML

The ELBO, the reparameterization trick, and score matching are the three ideas behind variational autoencoders and the entire diffusion-model era — from Stable Diffusion to AlphaFold.

VAE
Stable Diffusion
DDPM
normalizing flows

Step 92
~10m13
The Evidence Lower Bound (ELBO)
The ELBO is the single objective that makes variational autoencoders, Bayesian deep learning, and most modern generative models trainable —…
Open lesson
Step 93
~10m13
The Reparameterization Trick
The reparameterization trick is a one-line rewrite that converts the random sampling inside a VAE into a deterministic function plus externa…
Open lesson
Step 94
~15m3D · 113
★Most cinematic
Score Matching and Diffusion Models
The math behind diffusion models (Stable Diffusion, DALL-E) is just three pieces: a fixed forward noising process, a learned score function…
Open lesson

Variational Inference and Generative Models

Let’s figure it out.

The Evidence Lower Bound (ELBO)

The Reparameterization Trick

Score Matching and Diffusion Models

Variational Inference and Generative Models

Let’s figure it out.

The Evidence Lower Bound (ELBO)

The Reparameterization Trick

Score Matching and Diffusion Models