Diffusion Models

Diffusion models are a class of generative models that learn to generate data by reversing a diffusion process. The key idea is to model the gradual transformation of data into noise and then learn to reverse this process to generate new samples.

2. Understanding diffusion model (take 1, ELBO Perspective)

3. Understanding diffusion model (take 2, Tweedie Formula Perspective)

4. Understanding diffusion model (take 3, SDE Perspective)