Convergence and finite sample approximations of entropic regularized Wasserstein distances in Gaussian and RKHS settings

Convergence and finite sample approximations of entropic regularized Wasserstein distances in Gaussian and RKHS settings
复制标题

高斯和 RKHS 设置中熵正则化 Wasserstein 距离的收敛性和有限样本近似

DOI:
10.1142/s0219530522500142
复制
发表时间:
2023
影响因子:
2.2
通讯作者:
Ha Quang Minh
Ha Quang Minh
中科院分区:
数学3区
文献类型:
--
作者:
Yuhei Noda;Shota Saito;Shinichi Shirakawa;Ha Quang Minh

文献摘要

参考文献

被引文献

相似文献

本文研究了Hilbert空间中熵正则化Wasserstein距离的收敛性和有限样本逼近。我们的第一个主要结果是,对于无限维Hilbert空间上的Gauss测度,在2-Sinkhorn发散下的收敛严格弱于在精确2-Wasserstein距离下的收敛。具体地说,如果相应的协方差算子收敛于希尔伯特-施密特范数,则中心高斯测度序列收敛于2-Sinkhorn发散。这是在以前已知的结果,一个序列的中心高斯措施收敛于精确的2-Wasserstein距离,当且仅当协方差算子收敛于迹类规范。在再生核希尔伯特空间(RKHS)中,核高斯-辛霍恩发散是定义在RKHS上的高斯测度之间的辛霍恩发散,它定义了波兰空间上的波莱尔概率测度集上的半度量,给定该空间上的特征核。在Hilbert-Schmidt范数收敛的条件下,我们得到了核Gaussian-Sinkhorn散度的有限样本近似的维数无关收敛速度,其阶数与最大平均离散度相同.这些收敛速度特别适用于Sinkhorn发散高斯措施之间的欧几里得和无限维希尔伯特空间。欧氏空间上高斯测度之间的2-Wasserstein距离的样本复杂度,虽然依赖于维度,但比文献中的最坏情况要快。
This work studies the convergence and finite sample approximations of entropic regularized Wasserstein distances in the Hilbert space setting. Our first main result is that for Gaussian measures on an infinite-dimensional Hilbert space, convergence in the 2-Sinkhorn divergence isstrictly weakerthan convergence in the exact 2-Wasserstein distance. Specifically, a sequence of centered Gaussian measures converges in the 2-Sinkhorn divergence if the corresponding covariance operators converge in the Hilbert–Schmidt norm. This is in contrast to the previous known result that a sequence of centered Gaussian measures converges in the exact 2-Wasserstein distance if and only if the covariance operators converge in the trace class norm. In the reproducing kernel Hilbert space (RKHS) setting, thekernel Gaussian–Sinkhorn divergence, which is the Sinkhorn divergence between Gaussian measures defined on an RKHS, defines a semi-metric on the set of Borel probability measures on a Polish space, given a characteristic kernel on that space. With the Hilbert–Schmidt norm convergence, we obtaindimension-independentconvergence rates for finite sample approximations of the kernel Gaussian–Sinkhorn divergence, of the same order as the Maximum Mean Discrepancy. These convergence rates apply in particular to Sinkhorn divergence between Gaussian measures on Euclidean and infinite-dimensional Hilbert spaces. The sample complexity for the 2-Wasserstein distance between Gaussian measures on Euclidean space, whiledimension-dependent, is exponentially faster than the worst case scenario in the literature.
和空间的封闭性与广义薛定谔问题
DOI: --
发表时间: 1998
期刊:
影响因子: --
作者:
L. Rüschendorf;W. Thomsen
通讯作者: W. Thomsen
DOI: --
发表时间: 2007-12
期刊: --
影响因子: --
作者:
K. Fukumizu;A. Gretton;Xiaohai Sun;B. Scholkopf
通讯作者: K. Fukumizu;A. Gretton;Xiaohai Sun;B. Scholkopf
DOI: --
发表时间: 2007
期刊:
影响因子: --
作者:
A;Gretton;K;Fukumizu;C. -H. Teo;L;Song;B. Scholkopf;A. Smola.;Arthur Gretton
通讯作者: Arthur Gretton
权衡匹配:显示出相对于竞争特征的偏好
DOI: --
发表时间: 2009
期刊:
影响因子: --
作者:
Alfred Galichon;B. Salanié
通讯作者: B. Salanié
DOI: 10.1021/acs.jctc.9b01133
发表时间: 2019
影响因子: 5.5
作者:
Augusto Gerolin;Juri Grossi;P. Gori
通讯作者: P. Gori