arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2512.05092stat.MLcs.LG
收藏

扩散模型在一般状态空间中的基础:一个自包含的介绍

Foundations of Diffusion Models in General State Spaces: A Self-Contained Introduction

Technical University of Munich(慕尼黑工业大学);Helmholtz AI(亥姆霍兹AI);Munich Center for Machine Learning (MCML)(慕尼黑机器学习中心);Mila – Quebec AI Institute(Mila-魁北克人工智能研究所);AITHYRA

Vincent Pauline, Tobias Höppe, Kirill Neklyudov, Alexander Tong, Stefan Bauer, Andrea Dittadi

AI总结:

本文提供了一个自包含的扩散模型入门教程,统一了连续和离散状态空间的理论框架,阐述了反向动力学和ELBO的形成机制,适用于不同背景的读者。

摘要:

尽管扩散模型现在在生成建模中占据核心地位,但入门性介绍通常假设欧几里得数据,并很少阐明其与离散状态类比的联系。本文是关于在一般状态空间上扩散的自包含入门教程,将连续域和离散/分类结构统一在一个视角下。我们开发了离散时间观点(通过马尔可夫核的向前噪声和学习的反向动力学)及其连续时间极限——实数空间中的随机微分方程(SDEs)和有限字母表上的连续时间马尔可夫链(CTMCs)——并推导了相应的福克-普朗克方程和主方程。一种常见的变分处理导致了支撑标准训练损失的ELBO。我们明确展示了向前破坏选择——在连续空间中为高斯过程,在离散空间中为结构化分类转移核(均匀、掩码/吸收等)——如何影响反向动力学和ELBO。呈现是为三个受众分层的:寻求自包含直观介绍的新手;希望获得整体理论综合的扩散从业者;以及希望以类比方式进入离散扩散的连续扩散专家。结果是一个统一的路线图,涵盖了连续域和离散序列中的现代扩散方法,突显了一组可重用的证明、恒等式和核心理论原则。

英文摘要:

Although diffusion models now occupy a central place in generative modeling, introductory treatments commonly assume Euclidean data and seldom clarify their connection to discrete-state analogues. This article is a self-contained primer on diffusion over general state spaces, unifying continuous domains and discrete/categorical structures under one lens. We develop the discrete-time view (forward noising via Markov kernels and learned reverse dynamics) alongside its continuous-time limits -- stochastic differential equations (SDEs) in $\mathbb{R}^d$ and continuous-time Markov chains (CTMCs) on finite alphabets -- and derive the associated Fokker--Planck and master equations. A common variational treatment yields the ELBO that underpins standard training losses. We make explicit how forward corruption choices -- Gaussian processes in continuous spaces and structured categorical transition kernels (uniform, masking/absorbing and more) in discrete spaces -- shape reverse dynamics and the ELBO. The presentation is layered for three audiences: newcomers seeking a self-contained intuitive introduction; diffusion practitioners wanting a global theoretical synthesis; and continuous-diffusion experts looking for an analogy-first path into discrete diffusion. The result is a unified roadmap to modern diffusion methodology across continuous domains and discrete sequences, highlighting a compact set of reusable proofs, identities, and core theoretical principles.

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。

↑