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 2409.08311stat.MLcs.LGmath.PR
收藏

扩散流匹配在KL散度下的理论保证

Theoretical guarantees in KL for Diffusion Flow Matching

École Polytechnique(巴黎综合理工学院);Università degli studi di Padova(帕多瓦大学)

Marta Gentiloni Silveri, Giovanni Conforti, Alain Durmus

AI总结:

本文针对以布朗运动条件分布为桥接的扩散流匹配模型,在温和假设下推导了其生成分布与目标分布间KL散度的非渐近界,为该类生成模型提供理论保证。

摘要:

流匹配(Flow Matching,FM,也称为随机插值或整流流)是一类生成模型,其目标是利用固定耦合π和确定性或随机的桥接,在有限时间内将目标分布ν⋆与辅助分布μ连接起来。这两个要素定义了一条路径测度,随后可通过学习其马尔可夫投影的漂移来近似该路径测度。本文的主要贡献是在ν⋆、μ和π上提出相对温和的假设,为以布朗运动相关条件分布为桥接的扩散流匹配(Diffusion Flow Matching,DFM)模型提供非渐近保证。更具体地说,我们在ν⋆、μ和π的得分满足矩条件,以及标准L²漂移近似误差假设下,建立了目标分布与这类DFM模型生成的分布之间的Kullback-Leibler散度界。

英文摘要:

Flow Matching (FM) (also referred to as stochastic interpolants or rectified flows) stands out as a class of generative models that aims to bridge in finite time the target distribution $ν^\star$ with an auxiliary distribution $μ$, leveraging a fixed coupling $π$ and a bridge which can either be deterministic or stochastic. These two ingredients define a path measure which can then be approximated by learning the drift of its Markovian projection. The main contribution of this paper is to provide relatively mild assumptions on $ν^\star$, $μ$ and $π$ to obtain non-asymptotics guarantees for Diffusion Flow Matching (DFM) models using as bridge the conditional distribution associated with the Brownian motion. More precisely, we establish bounds on the Kullback-Leibler divergence between the target distribution and the one generated by such DFM models under moment conditions on the score of $ν^\star$, $μ$ and $π$, and a standard $L^2$-drift-approximation error assumption.

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。

↑