Large Language Models (LLMs) - Stanford CME296 Diffusion & Large Vision Models | Spring 2026 | Lecture 4 - Latent Space & Guidance

Learn more details about this course: https://online.stanford.edu/courses/cme296-diffusion-and-large-vision-models To follow along with the course schedule and syllabus, visit: https://cme296.stanford.edu/syllabus/ Chapters: 00:00:00 Introduction 00:07:05 Pixel space 00:12:39 Semantic vs perceptual similarity 00:14:27 Autoencoder 00:22:56 Variational autoencoders 00:31:19 ELBO derivation 00:46:43 Blurriness issue of VAEs 00:47:43 Reconstruction loss 00:48:54 KL regularization loss 00:50:17 Perceptual loss 00:54:17 Adversarial loss 00:57:01 Latent diffusion models 01:00:31 Encoder vs decoder trade-off 01:05:55 Text representation with Transformers 01:12:20 Image representation with ViT 01:18:35 Contrastive learning with CLIP 01:27:44 Classifier-based guidance 01:34:34 Classifier-free guidance For more information about Stanford’s graduate programs, visit: https://online.stanford.edu/graduate-education Afshine Amidi is an Adjunct Lecturer at Stanford University. Shervine Amidi is an Adjunct Lecturer at Stanford University. View the course playlist: https://www.youtube.com/playlist?list=PLoROMvodv4rNdy8rt2rZ4T2xM0OjADnfu