2. Denoising Diffusion Probabilistic Models (DDPMs)

0 Hierachical HVAE

Figure 1: Hierachical Variational AutoEncoder

1. Denoising Diffusion Probabilistic Models

_Figure 2: Denoising Diffusion Probabilistic Models: DDPMS consist of 2 component. where $q(.)$ is a fixed kernel adding gaussian noise to the data, and $p_{\theta }$ is a learned reverse process that try to generate new same from noise.