arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2511.04844cs.LGmath.STstat.TH
收藏

即使对于DDPM,次线性迭代次数也可能足够

Sublinear iterations can suffice even for DDPMs

UToronto(多伦多大学);CMU(卡内基梅隆大学);Harvard(哈佛大学);Yale(耶鲁大学)

Matthew S. Zhang, Stephen Huan, Jerry Huang, Nicholas M. Boffi, Sitan Chen, Sinho Chewi

AI总结:

本文针对DDPM采样提出去噪扩散随机中点法(DDRaM),利用随机中点更好地近似SDE,首次证明纯DDPM采样只需次线性评分评估次数即可收敛,并通过实验验证了其在实际预训练图像合成模型中的有效性。

摘要:

基于SDE的方法,如去噪扩散概率模型(DDPM),在现实世界的样本生成任务中取得了显著成功。先前对DDPM的分析主要集中在指数欧拉离散化上,所给出的保证通常至少线性依赖于维度或初始Fisher信息。受对数凹采样方面工作(Shen和Lee,2019)的启发,我们分析了一种积分器——去噪扩散随机中点法(DDRaM)——它利用额外的随机中点来更好地近似SDE。利用最近发展的称为“移位复合规则”的分析框架,我们证明该算法在适当的光滑性假设下具有良好的离散化性质,只需次线性的$\widetilde{O}(\sqrt{d})$次评分函数评估即可确保收敛。这是纯DDPM采样的第一个次线性复杂度界——先前获得此类界的工作转而使用基于ODE的采样,并且不得不对采样器进行修改,这些修改偏离了实际使用方式。我们还提供了我们方法优势的实验验证,表明它在使用预训练图像合成模型时在实践中表现良好。

英文摘要:

SDE-based methods such as denoising diffusion probabilistic models (DDPMs) have shown remarkable success in real-world sample generation tasks. Prior analyses of DDPMs have been focused on the exponential Euler discretization, showing guarantees that generally depend at least linearly on the dimension or initial Fisher information. Inspired by works in log-concave sampling (Shen and Lee, 2019), we analyze an integrator -- the denoising diffusion randomized midpoint method (DDRaM) -- that leverages an additional randomized midpoint to better approximate the SDE. Using a recently-developed analytic framework called the "shifted composition rule", we show that this algorithm enjoys favorable discretization properties under appropriate smoothness assumptions, with sublinear $\widetilde{O}(\sqrt{d})$ score evaluations needed to ensure convergence. This is the first sublinear complexity bound for pure DDPM sampling -- prior works which obtained such bounds worked instead with ODE-based sampling and had to make modifications to the sampler which deviate from how they are used in practice. We also provide experimental validation of the advantages of our method, showing that it performs well in practice with pre-trained image synthesis models.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑