Blackout Diffusion: Generative Diffusion Models in Discrete-State Spaces
This work addresses the challenge of using diffusion models in discrete-state spaces, which is common in scientific applications, by providing a generalized theoretical formulation without variational approximations.
The authors tackled the problem of applying generative diffusion models to discrete-state data by developing a theoretical framework for arbitrary discrete-state Markov processes, enabling sample generation from an empty image rather than noise. Numerical experiments on CIFAR-10, Binarized MNIST, and CelebA datasets confirmed the feasibility of this approach.
Typical generative diffusion models rely on a Gaussian diffusion process for training the backward transformations, which can then be used to generate samples from Gaussian noise. However, real world data often takes place in discrete-state spaces, including many scientific applications. Here, we develop a theoretical formulation for arbitrary discrete-state Markov processes in the forward diffusion process using exact (as opposed to variational) analysis. We relate the theory to the existing continuous-state Gaussian diffusion as well as other approaches to discrete diffusion, and identify the corresponding reverse-time stochastic process and score function in the continuous-time setting, and the reverse-time mapping in the discrete-time setting. As an example of this framework, we introduce ``Blackout Diffusion'', which learns to produce samples from an empty image instead of from noise. Numerical experiments on the CIFAR-10, Binarized MNIST, and CelebA datasets confirm the feasibility of our approach. Generalizing from specific (Gaussian) forward processes to discrete-state processes without a variational approximation sheds light on how to interpret diffusion models, which we discuss.