arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2307.01050stat.MLcs.LG
收藏

传输遇见变分推断:受控蒙特卡洛扩散

Transport meets Variational Inference: Controlled Monte Carlo Diffusions

University of Cambridge(剑桥大学);KIT Karlsruhe(卡尔斯鲁厄理工学院);Kings College London(伦敦国王学院)

Francisco Vargas, Shreyas Padhy, Denis Blessing, Nikolas Nüsken

AI总结:

本文提出受控蒙特卡洛扩散采样器CMCD,通过同时自适应前向和后向动力学并结合路径空间散度,为贝叶斯计算提供了一种优于现有方法的基于分数的退火采样框架。

摘要:

将最优传输与变分推断联系起来,我们提出了一个以路径空间散度为核心的有原则且系统化的采样与生成建模框架。我们的工作最终发展为用于贝叶斯计算的受控蒙特卡洛扩散采样器(CMCD),这是一种基于分数的退火技术,关键在于同时自适应扩散模型中的前向和后向动力学。在此过程中,我们阐明了EM算法与薛定谔桥的迭代比例拟合(IPF)之间的关系,并推导出一个正则化目标,从而绕过了标准IPF更新中的迭代瓶颈。最后,我们证明CMCD在统计物理学中的Jarzynski恒等式和Crooks恒等式上具有坚实基础,并且在广泛的实验中令人信服地优于竞争方法。

英文摘要:

Connecting optimal transport and variational inference, we present a principled and systematic framework for sampling and generative modelling centred around divergences on path space. Our work culminates in the development of the \emph{Controlled Monte Carlo Diffusion} sampler (CMCD) for Bayesian computation, a score-based annealing technique that crucially adapts both forward and backward dynamics in a diffusion model. On the way, we clarify the relationship between the EM-algorithm and iterative proportional fitting (IPF) for Schr{ö}dinger bridges, deriving as well a regularised objective that bypasses the iterative bottleneck of standard IPF-updates. Finally, we show that CMCD has a strong foundation in the Jarzinsky and Crooks identities from statistical physics, and that it convincingly outperforms competing approaches across a wide array of experiments.

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。

↑