Video links

Work in progress: this part of the notes is still being revised.

Video links

The lecture recordings of the courses in the course links, arranged by chapter. The reading plan gives the matching slides, notes, and readings. MIT 6.S978 publishes slides but no recordings, so it appears only in the reading plan.

Video or playlistLinkNotes
Stanford CS236, Fall 2023PlaylistThe numbers below are lecture numbers.
MIT 6.S184, IAP 2026PlaylistLecture 3 has two parts, on score matching and on guidance. Lectures 5 and 6 of the IAP 2025 edition cover robotics and protein design.
UC Berkeley CS294-158, Spring 2024PlaylistThe numbers below are lecture numbers.
Stanford CME 296, Spring 2026PlaylistThe numbers below are lecture numbers.
UT Austin IFML, Fall 2025PlaylistMathematical lectures on the analysis of diffusion samplers (chapter 15).

By chapter

ChapterCS236 Fall 2023 lectures6.S184 lecturesOther recordings
1. Foundations of Generative Modeling1: Introduction · 2: Background · 4: Maximum likelihood learning——
2. Autoregressive Models3: Autoregressive models—CS294-158 2: Autoregressive models
3. Variational Autoencoders5: Variational autoencoders · 6: Variational autoencoders, continued—CS294-158 4: Latent variable models
4. Normalizing Flows7: Normalizing flows · 8: Normalizing flows, continued—CS294-158 3: Flow models
5. Generative Adversarial Networks9: Generative adversarial networks · 10: Generative adversarial networks, continued—CS294-158 5: GANs and implicit models
6. Energy-Based Models and Score Matching11: Energy-based models · 12: Energy-based models, continued · 13: Score-based models · 14: Energy-based and score-based models3, first part: Score functions and score matching—
7. Denoising Diffusion Models16: Score-based diffusion models—CS294-158 6: Diffusion models · CME 296 1: Diffusion
8. Diffusion SDEs and Fast Sampling—1: Flow and diffusion modelsCME 296 2: Score matching
9. Flow Matching—2: Flow matchingCME 296 3: Flow matching
10. Guidance and Conditional Generation—3, second part: Classifier-free guidanceCME 296 4: Latent space and guidance
11. Latent Diffusion and Large-Scale Generation—4: Latent spaces and neural network architecturesCME 296 5: Architectures · CME 296 6: Model training · CS294-158 9: Video generation
12. Discrete Tokens and Multimodal Generation17: Discrete latent variable models · 18: Diffusion models for discrete data (Aaron Lou)5: Discrete diffusion modelsCS294-158 12a: Multimodal models
13. Evaluating Generative Models15: Evaluation of generative models—CME 296 7: Evaluation
14. Generative Models for Science and Control (optional)—IAP 2025, 5: Generative robotics · IAP 2025, 6: Generative protein designCS294-158 13a: AI for science (John Ingraham)
15. The Theory of Diffusion Models (optional)——UT Austin IFML: Diffusion models for generative AI