arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2407.13445math.OC
收藏

约束近似最优传输映射

Constrained Approximate Optimal Transport Maps

Eloi Tanguy, Agnès Desolneux, Julie Delon

AI总结:

本文研究在函数类约束下最小化最优传输代价的映射求解问题,提出交替最小化与核方法等途径,并给出数值实验与颜色迁移应用。

摘要:

我们研究在函数类 $G$ 中寻找一个映射 $g$,使其最小化目标测度 $ν$ 与源测度 $μ$ 经 $g$ 映射后的像之间的最优传输(OT)代价。当从 $μ$ 到 $ν$ 的 OT 映射不存在或不满足 $G$ 的期望约束时,这一问题具有现实意义。我们针对 $L$-Lipschitz 函数的一般子类讨论存在性与唯一性,包括(强)凸函数的梯度以及典型的神经网络。我们探索了一种逼近传输计划的变体,并在某些情形下证明了其与映射问题的等价性。对于平方欧氏代价,我们提出在传输计划 $π$ 和映射 $g$ 上交替最小化,其中关于 $g$ 的优化是重心映射 $\overlineπ$ 在 $G$ 上的 $L^2$ 投影。在一维情形下,该全局问题等价于将 $μ$ 与 $ν$ 之间的某个 OT 计划 $π^*$ 所对应的重心映射 $\overline{π^*}$ 投影到 $G$ 上,但这一结论不能推广到更高维。我们引入一种简单的核方法,在离散情形下于再生核希尔伯特空间中寻找 $g$。我们给出了 $\ell$-强凸势函数的 $L$-Lipschitz 梯度的数值方法,并研究了神经网络随机梯度下降方法的收敛性。最后,我们以颜色迁移为例进行说明,将学习到的映射应用于新图像,并展示了其对异常值的鲁棒性。

英文摘要:

We investigate finding a map $g$ within a function class $G$ that minimises an Optimal Transport (OT) cost between a target measure $ν$ and the image by $g$ of a source measure $μ$. This is relevant when an OT map from $μ$ to $ν$ does not exist or does not satisfy the desired constraints of $G$. We address existence and uniqueness for generic subclasses of $L$-Lipschitz functions, including gradients of (strongly) convex functions and typical Neural Networks. We explore a variant that approaches a transport plan, showing equivalence to a map problem in some cases. For the squared Euclidean cost, we propose alternating minimisation over a transport plan $π$ and map $g$, with the optimisation over $g$ being the $L^2$ projection on $G$ of the barycentric mapping $\overlineπ$. In dimension one, this global problem equates the $L^2$ projection of $\overline{π^*}$ onto $G$ for an OT plan $π^*$ between $μ$ and $ν$, but this does not extend to higher dimensions. We introduce a simple kernel method to find $g$ within a Reproducing Kernel Hilbert Space in the discrete case. We present numerical methods for $L$-Lipschitz gradients of $\ell$-strongly convex potentials, and study the convergence of Stochastic Gradient Descent methods for Neural Networks. We finish with an illustration on colour transfer, applying learned maps on new images, and showcasing outlier robustness.

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。

↑