A Continuous Time Framework for Discrete Denoising Models
This provides a novel continuous-time approach for discrete data denoising, which is incremental but offers practical improvements and theoretical insights.
The authors tackled the problem of discrete denoising diffusion models by developing a continuous time framework using Continuous Time Markov Chains (CTMCs), resulting in high-performance samplers that outperform discrete time methods for discrete data and a theoretical error bound.
We provide the first complete continuous time framework for denoising diffusion models of discrete data. This is achieved by formulating the forward noising process and corresponding reverse time generative process as Continuous Time Markov Chains (CTMCs). The model can be efficiently trained using a continuous time version of the ELBO. We simulate the high dimensional CTMC using techniques developed in chemical physics and exploit our continuous time framework to derive high performance samplers that we show can outperform discrete time methods for discrete data. The continuous time treatment also enables us to derive a novel theoretical result bounding the error between the generated sample distribution and the true data distribution.