Diffusion Models Beat GANs on Image Synthesis - ML Coding Series - Part 2

This course explains the classifier guidance technique introduced in the Diffusion Models Beat GANs on Image Synthesis paper. It combines a paper overview with code analysis covering noise-aware classifier training, timestep conditioning, sampling, mean shifting, and a minor implementation bug.

Syllabus

Intro
Paper overview part - U-Net architecture improvements
Classifier guidance explained
Intuition behind classifier guidance
Scaling classifier guidance
Diversity vs quality tradeoff and future work
Coding part - training a noise-aware classifier
Main training loop
Visualizing timestep conditioning
Sampling using classifier guidance
Core of the sampling logic
Shifting the mean - classifier guidance
Minor bug in their code and my GitHub issue
Outro