Provable Efficiency of Guidance in Diffusion Models for General Data Distribution
This provides a theoretical foundation for guidance techniques in diffusion models, addressing a gap in existing case studies, but it is incremental as it extends analysis to more general distributions without introducing new methods.
The paper tackles the lack of theoretical understanding of guidance in diffusion models by analyzing its effect under general data distributions, proving that guidance improves average sample quality as measured by the reciprocal of classifier probability.
Diffusion models have emerged as a powerful framework for generative modeling, with guidance techniques playing a crucial role in enhancing sample quality. Despite their empirical success, a comprehensive theoretical understanding of the guidance effect remains limited. Existing studies only focus on case studies, where the distribution conditioned on each class is either isotropic Gaussian or supported on a one-dimensional interval with some extra conditions. How to analyze the guidance effect beyond these case studies remains an open question. Towards closing this gap, we make an attempt to analyze diffusion guidance under general data distributions. Rather than demonstrating uniform sample quality improvement, which does not hold in some distributions, we prove that guidance can improve the whole sample quality, in the sense that the average reciprocal of the classifier probability decreases with the existence of guidance. This aligns with the motivation of introducing guidance.