Theory on Score-Mismatched Diffusion Models and Zero-Shot Conditional Samplers

TL;DR

This paper provides the first theoretical analysis and performance guarantees for score-mismatched diffusion models, especially in zero-shot conditional sampling, highlighting how distributional mismatches cause bias and guiding the design of bias-optimal samplers.

Contribution

It introduces the first explicit performance guarantees for score-mismatched diffusion models, including bias analysis and convergence bounds for zero-shot conditional sampling.

Findings

Score mismatches cause asymptotic bias proportional to distributional mismatch.

Derived convergence bounds depend explicitly on dimension and conditioning.

Numerical studies support theoretical findings.

Abstract

The denoising diffusion model has recently emerged as a powerful generative technique, capable of transforming noise into meaningful data. While theoretical convergence guarantees for diffusion models are well established when the target distribution aligns with the training distribution, practical scenarios often present mismatches. One common case is in the zero-shot conditional diffusion sampling, where the target conditional distribution is different from the (unconditional) training distribution. These score-mismatched diffusion models remain largely unexplored from a theoretical perspective. In this paper, we present the first performance guarantee with explicit dimensional dependencies for general score-mismatched diffusion samplers, focusing on target distributions with finite second moments. We show that score mismatches result in an asymptotic distributional bias between the target and sampling distributions, proportional to the accumulated mismatch between the target and training distributions. This result can be directly applied to zero-shot conditional samplers for any conditional model, irrespective of measurement noise. Interestingly, the derived convergence upper bound offers useful guidance for designing a novel bias-optimal zero-shot sampler in linear conditional models that minimizes the asymptotic bias. For such bias-optimal samplers, we further establish convergence guarantees with explicit dependencies on dimension and conditioning, applied to several interesting target distributions, including those with bounded support and Gaussian mixtures. Our findings are supported by numerical studies.