Entropic Regularization of Wasserstein Distance Between Infinite-Dimensional Gaussian Measures and Gaussian Processes

Entropic Regularization of Wasserstein Distance Between Infinite-Dimensional Gaussian Measures and Gaussian Processes
复制标题

无限维高斯测度与高斯过程之间 Wasserstein 距离的熵正则化

DOI:
10.1007/s10959-022-01165-1
复制
发表时间:
2022
影响因子:
0.8
通讯作者:
Ha Quang Minh
Ha Quang Minh
中科院分区:
数学4区
文献类型:
--
作者:
Ryosuke Shibukawa;Shoichi Ishida;Kazuki Yoshizoe;Kunihiro Wasa;Kiyosei Takasu;Yasushi Okuno;Kei Terayama;Koji Tsuda;Ha Quang Minh

文献摘要

参考文献

相似文献

这项工作研究无限维希尔伯特空间上 2-Wasserstein 距离的熵正则化公式,特别是高斯设置。我们首先提出最小互信息性质,即希尔伯特空间上具有最小互信息的两个高斯测度的联合测度是联合高斯测度。这是欧几里得空间上高斯密度最大熵性质的无限维推广。然后,我们给出了最优熵传输计划、熵 2-Wasserstein 距离和希尔伯特空间上两个高斯测度之间的 Sinkhorn 散度的封闭式公式,以及一组高斯测度的重心的定点方程。我们的公式充分利用了熵公式的正则化方面,并且在奇异和非奇异设置中均有效。在无限维设置中,熵 2-Wasserstein 距离和 Sinkhorn 散度都是 Fréchet 可微的,而精确的 2-Wasserstein 距离是不可微的。我们的 Sinkhorn 重心方程是新的,并且总是有唯一的解。相反,熵 2-Wasserstein 距离的有限维重心方程无法推广到希尔伯特空间设置。在再现核希尔伯特空间的设置中,我们的距离公式根据相应的核 Gram 矩阵明确给出,提供核最大平均差异和核 2-Wasserstein 距离之间的插值。
This work studies the entropic regularization formulation of the 2-Wasserstein distance on an infinite-dimensional Hilbert space, in particular for the Gaussian setting. We first present the minimum mutual information property, namely, the joint measures of two Gaussian measures on Hilbert space with the smallest mutual information are joint Gaussian measures. This is the infinite-dimensional generalization of the maximum entropy property of Gaussian densities on Euclidean space. We then give closed-form formulas for the optimal entropic transport plan, entropic 2-Wasserstein distance, and Sinkhorn divergence between two Gaussian measures on a Hilbert space, along with the fixed point equations for the barycenter of a set of Gaussian measures. Our formulations fully exploit the regularization aspect of the entropic formulation and are valid both insingularand innonsingularsettings. In the infinite-dimensional setting, both the entropic 2-Wasserstein distance and Sinkhorn divergence are Fréchet differentiable, in contrast to the exact 2-Wasserstein distance, which is not differentiable. Our Sinkhorn barycenter equation is new and always has a unique solution. In contrast, the finite-dimensional barycenter equation for the entropic 2-Wasserstein distance fails to generalize to the Hilbert space setting. In the setting of reproducing kernel Hilbert spaces, our distance formulas are given explicitly in terms of the corresponding kernel Gram matrices, providing an interpolation between the kernel maximum mean discrepancy and the kernel 2-Wasserstein distance.
$${ extsf {RCD}}^*(K,N)$$RCD∗(K,N) 空间上熵成本的 Benamou–Brenier 和对偶公式
DOI: --
发表时间: 2018
影响因子: 2
作者:
N. Gigli;Luca Tamanini
通讯作者: Luca Tamanini
非正曲率:希尔伯特-施密特算子的几何方法
DOI: 10.1016/j.difgeo.2007.06.016
发表时间: 2007
影响因子: 0.5
作者:
G. Larotonda
通讯作者: G. Larotonda
高斯随机函数与 DNA 小环几何形状的二阶比较
DOI: 10.1198/jasa.2010.tm09239
发表时间: 2010
影响因子: 3.7
作者:
["V. Panaretos
通讯作者: ["V. Panaretos
DOI: 10.1137/0119044
发表时间: 1970-01-01
影响因子: 1.9
作者:
BAKER, CR
通讯作者: BAKER, CR
DOI: --
发表时间: 1958
期刊:
影响因子: --
作者:
J. Feldman
通讯作者: J. Feldman