A Survey on Generative Diffusion Models
0. Introduction This article reviews deep generative models, particularly diffusion models, and how they endow machines with human-like imagination. Diffusion models show great potential in generating realistic samples, overcoming the posterior distribution alignment obstacles in variational autoencoders and alleviating the instability of adversarial objectives in generative adversarial networks. Diffusion models consist of two interconnected processes: … Read more
