Diffusion models with location-scale noise
This work addresses a foundational question in generative modeling for researchers, but it is incremental as it focuses on noise distribution comparisons without introducing a new paradigm.
The authors tackled the problem of determining which noise distribution yields better generated data in diffusion models, finding that Gaussian noise performs best compared to non-Gaussian alternatives like Laplace and Uniform.
Diffusion Models (DMs) are powerful generative models that add Gaussian noise to the data and learn to remove it. We wanted to determine which noise distribution (Gaussian or non-Gaussian) led to better generated data in DMs. Since DMs do not work by design with non-Gaussian noise, we built a framework that allows reversing a diffusion process with non-Gaussian location-scale noise. We use that framework to show that the Gaussian distribution performs the best over a wide range of other distributions (Laplace, Uniform, t, Generalized-Gaussian).