Generative Models

The topic of our project is the paper that introduced Denoising Diffusion Probabilistic Models (DDPMs), or Diffusion Models for short. In this report, we introduce the background of variational generative models to let a reader get a good understanding of the paper’s position in the literature. Further, we go into details of the mathematical derivation of the algorithm, making the theoretical part of the paper as accessible as possible. Lastly, we provide an implementation of the diffusion model in JAX: source code