Course Study Note

Day 5 — Diffusion, LDM, and Applications

I connected forward noise to reverse denoising, then compared where DDPM, latent diffusion, and Stable Diffusion perform their computation.

Date
2025-07-11
Course
2025 Generative Models Intensive — From VAE to Stable Diffusion
01DDPM Forward and Reverse Processes

Study checkpoint for this day

02Latent Diffusion and Stable Diffusion

Study checkpoint for this day

03Diffusion Applications and Next Steps

Study checkpoint for this day

A day-level study path generated from the saved course topic order.

Topics from the day

  1. DDPM Forward and Reverse Processes

    I rewrote timestep-wise noise prediction and the sampling loop.

  2. Latent Diffusion and Stable Diffusion

    I compared why denoising in latent space reduces computation.

  3. Diffusion Applications and Next Steps

    I grouped conditioning

Forward noise, learned reverse denoising, and latent diffusion
I redrew the two diffusion processes and the latent-space boundary in one diagram.

Forward process

Small Gaussian perturbations move (x_0) toward a standard normal sample. A cumulative coefficient produces an arbitrary (x_t) directly during training instead of replaying every earlier step.

Reverse process

In DDPM, a U-Net receives (x_t) and (t) and predicts noise. Training samples one timestep, but generation walks from (T) back to zero and is therefore sequential.

Latent diffusion and DiT

Latent Diffusion performs denoising in an autoencoder’s compressed space and uses cross-attention for conditions such as text. DiT keeps the diffusion objective but replaces the U-Net backbone with a transformer over latent patches.

The course diffusion notebook retained code but no output cells. I will first rerun a small FashionMNIST DDPM and record tensor shapes and the sampling loop before moving to latent diffusion.

← Course home
Publication first-page preview