This paper establishes a theoretical connection between diffusion models and the I-MMSE theorem in information theory. Specifically, we demonstrate that two widely used log-likelihood-based training objectives in diffusion models are closely related, with their relationship explicitly established via the I-MMSE theorem. As a direct consequence of this theoretical link, we derive a single-step denoising method that achieves optimal performance without iterative sampling.
Publisher
Allerton Conference on Communication, Control, and Computing
Series/Report Name or Number
2025 61st Allerton Conference on Communication, Control, and Computing Proceedings
Use this login method if you
don't
have an
@illinois.edu
email address.
(Oops, I do have one)
IDEALS migrated to a new platform on June 23, 2022. If you created
your account prior to this date, you will have to reset your password
using the forgot-password link below.
