arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2103.05934math.FAmath.MG
收藏

目标测度变化时最优输运映射的定量稳定性

Quantitative Stability of Optimal Transport Maps under Variations of the Target Measure

Alex Delalande, Quentin Merigot

摘要:

研究了R^d上固定概率密度$ρ$和概率测度$μ$之间的二次最优输运映射的定量稳定性,记为T $μ$ .假设源密度$ρ$在紧凸集上上下有界,我们证明了映射$μ$ $\rightarrow$ T $μ$在大的概率测度族上是bi-H{<$}lder连续的,例如阶矩p > d的概率测度集是由某个常数有界的.这些稳定性估计表明,线性化的最优输运度量W2,$ρ$($μ$,$ν$)= T $μ$ -- T $ν$ L 2($ρ$,Rd)是bi-H{ö}lder等效于此类集合上的2-Wasserstein距离,证明其在应用中的使用是合理的。

英文摘要:

This work studies the quantitative stability of the quadratic optimal transport map between a fixed probability density $ρ$ and a probability measure $μ$ on R^d , which we denote T$μ$. Assuming that the source density $ρ$ is bounded from above and below on a compact convex set, we prove that the map $μ$ $\rightarrow$ T$μ$ is bi-H{ö}lder continuous on large families of probability measures, such as the set of probability measures whose moment of order p > d is bounded by some constant. These stability estimates show that the linearized optimal transport metric W2,$ρ$($μ$, $ν$) = T$μ$ -- T$ν$ L 2 ($ρ$,R d) is bi-H{ö}lder equivalent to the 2-Wasserstein distance on such sets, justifiying its use in applications.

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。

↑