Abstract:Consider an empirical measure $\mathbb{P}_n$ induced by $n$ iid samples from a $d$-dimensional $K$-subgaussian distribution $\mathbb{P}$ and let $\gamma = N(0,\sigma^2 I_d)$ be the isotropic Gaussian measure. We study the speed of convergence of the smoothed Wasserstein distance $W_2(\mathbb{P}_n * \gamma, \mathbb{P}*\gamma) = n^{-\alpha + o(1)}$ with $*$ being the convolution of measures. For $K<\sigma$ and in any dimension $d\ge 1$ we show that $\alpha = {1\over2}$. For $K>\sigma$ in dimension $d=1$ we show that the rate is slower and is given by $\alpha = {(\sigma^2 + K^2)^2\over 4 (\sigma^4 + K^4)} < 1/2$. This resolves several open problems in [GGNWP20], and in particular precisely identifies the amount of smoothing $\sigma$ needed to obtain a parametric rate. In addition, for any $d$-dimensional $K$-subgaussian distribution $\mathbb{P}$, we also establish that $D_{KL}(\mathbb{P}_n * \gamma \|\mathbb{P}*\gamma)$ has rate $O(1/n)$ for $K<\sigma$ but only slows down to $O({(\log n)^{d+1}\over n})$ for $K>\sigma$. The surprising difference of the behavior of $W_2^2$ and KL implies the failure of $T_{2}$-transportation inequality when $\sigma < K$. Consequently, it follows that for $K>\sigma$ the log-Sobolev inequality (LSI) for the Gaussian mixture $\mathbb{P} * N(0, \sigma^{2})$ cannot hold. This closes an open problem in [WW+16], who established the LSI under the condition $K<\sigma$ and asked if their bound can be improved.

Symmetric KL-divergence by Stein's Method

A Supplement to the Complete Convergence

Precise Asymptotics in the Law of Logarithm for U-statistics

Kernel Stein Discrepancy on Lie Groups: Theory and Applications

Central Limit Theorem on Symmetric Kullback-Leibler (KL) Divergence

Bounds for distributional approximation in the multivariate delta method by Stein's method

Three-Parameter Approximations of Sums of Locally Dependent Random Variables via Stein's Method

Stein's method and Narayana numbers

Nyström Kernel Stein Discrepancy

Precise rates in the law of the logarithm for the moment convergence in Hilbert spaces

Kac's Central Limit Theorem by Stein's Method

Convergence Rates of Spectral Distributions of Large Sample Covariance Matrices

Stein's method for comparison of univariate distributions

The Convergence of the Sums of Independent Random Variables under the Sub-Linear Expectations

Improved Finite-Particle Convergence Rates for Stein Variational Gradient Descent

Normal approximation in total variation for statistics in geometric probability

A Stein characterisation of the distribution of the product of correlated normal random variables

Stein kernels for $q$-moment measures and new bounds for the rate of convergence in the central limit theorem

Multidimensional Stein's method for Gamma approximation

Rate of convergence of the smoothed empirical Wasserstein distance

On Stein's method and mod-* convergence