Title: A brief introduction to denoising diffusion probabilistic models - a class of powerful deep generative models. 

Speaker: Zhifeng Kong (UCSD)

Abstract: Recently denoising diffusion probabilistic models have achieved excellent quality in many generative tasks, such as image synthesis, audio synthesis, and point cloud synthesis. In the talk I will give a 20-minute brief introduction to these models so the audience will be able to understand what these models are and how they work. I will start from the background on variational inference, then introduce how denoising diffusion probabilistic models connect to variational inference, and finally talk about some downstream applications.