Seminar 'Star-Shaped Denoising Diffusion Probabilistic Models'

On March 24 a seminar of the Centre of Deep Learning and Bayesian Methods will be held at HSE University. Andrey Okhotin will speak on 'Star-Shaped Denoising Diffusion Probabilistic Models'.

Abstract:

Methods based on Denoising Diffusion Probabilistic Models (DDPM) became a ubiquitous tool in generative modelling. However, they are mostly limited to Gaussian and discrete diffusion processes. We propose Star-Shaped Denoising Diffusion Probabilistic Models (SS-DDPM), a model with a non-Markovian diffusion-like noising process. In the case of Gaussian distributions, this model is equivalent to Markovian DDPMs. However, it can be defined and applied with arbitrary noising distributions, and admits efficient training and sampling algorithms for a wide range of distributions that lie in the exponential family. We provide a simple recipe for designing diffusion-like models with distributions like Beta, von Mises--Fisher, Dirichlet, Wishart and others, which can be especially useful when data lies on a constrained manifold such as the unit sphere, the space of positive semi-definite matrices, the probabilistic simplex, etc. We evaluate the model in different settings and find it competitive even on image data, where Beta SS-DDPM achieves results comparable to a Gaussian DDPM.

The seminar will be held online via Zoom.

Start time: 18:00