On the optimal estimation of probability measures in weak and strong topologies

DOI10.3150/15-BEJ713MaRDI QIDQ282569zbMATH OpenFDO

Authors Bharath K. Sriperumbudur

Publication date 12 May 2016

Published in Bernoulli (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1310.8240, https://projecteuclid.org/euclid.bj/1458133001

zbMATH Keywords

reproducing kernel Hilbert spacekernel density estimatortwo-sample testadaptive estimationtotal variation distancebounded Lipschitz metricexponential inequalityRademacher chaossmoothed empirical processesU-processesuniform central limit theorem

Mathematics Subject Classification ID

Density estimation (62G07) Asymptotic properties of nonparametric inference (62G20) Probability measures on topological spaces (60B05) Central limit and other weak theorems (60F05) Functional limit theorems; invariance principles (60F17)

Abstract: Given random samples drawn i.i.d. from a probability measure $mathbbP$ (defined on say, $mathbb{R}^d$), it is well-known that the empirical estimator is an optimal estimator of $mathbbP$ in weak topology but not even a consistent estimator of its density (if it exists) in the strong topology (induced by the total variation distance). On the other hand, various popular density estimators such as kernel and wavelet density estimators are optimal in the strong topology in the sense of achieving the minimax rate over all estimators for a Sobolev ball of densities. Recently, it has been shown in a series of papers by Gin'{e} and Nickl that these density estimators on $mathbbR$ that are optimal in strong topology are also optimal in $|cdot{|}_{mathcalF}$ for certain choices of $mathcalF$ such that $|cdot{|}_{mathcalF}$ metrizes the weak topology, where $|mathbbP{|}_{mathcalF}:=supintf,mathrmdmathbbP:finmathcalF$. In this paper, we investigate this problem of optimal estimation in weak and strong topologies by choosing $mathcalF$ to be a unit ball in a reproducing kernel Hilbert space (say $mathcal{F}_H$ defined over $mathbb{R}^d$), where this choice is both of theoretical and computational interest. Under some mild conditions on the reproducing kernel, we show that $|cdot{|}_{mathcal{F}_H}$ metrizes the weak topology and the kernel density estimator (with $L^1$ optimal bandwidth) estimates $mathbbP$ at dimension independent optimal rate of $n^{-1/2}$ in $|cdot{|}_{mathcal{F}_H}$ along with providing a uniform central limit theorem for the kernel density estimator.

Recommendations

• On the weak convergence of the kernel density estimator in the uniform topology
• Minimax estimation of kernel mean embeddings
• On the weak convergence of kernel density estimators in \(L^{p}\) spaces
• A Note on Weak Convergence of Density Estimators in Hilbert Spaces
• On weak convergence and optimality of kernel density estimates of the mode

Cites work

• scientific article; zbMATH DE number 3870398 (Why is no real title available?)
• scientific article; zbMATH DE number 4004880 (Why is no real title available?)
• scientific article; zbMATH DE number 96990 (Why is no real title available?)
• scientific article; zbMATH DE number 3576139 (Why is no real title available?)
• scientific article; zbMATH DE number 1254560 (Why is no real title available?)
• scientific article; zbMATH DE number 1324223 (Why is no real title available?)
• scientific article; zbMATH DE number 720761 (Why is no real title available?)
• scientific article; zbMATH DE number 1022519 (Why is no real title available?)
• scientific article; zbMATH DE number 1552481 (Why is no real title available?)
• scientific article; zbMATH DE number 5055767 (Why is no real title available?)
• A kernel two-sample test
• Adaptation on the space of finite signed measures
• Adaptive estimation of a distribution function and its density in sup-norm loss by wavelet and spline projections
• An exponential inequality for the distribution function of the kernel density estimator, with applications to adaptive estimation
• Asymptotic Statistics
• Donsker-type theorems for nonparametric maximum likelihood estimators
• Empirical processes indexed by Lipschitz functions
• Hilbert space embeddings and metrics on probability measures
• Local Rademacher complexities
• Neural Network Learning
• Nonparametric estimators which can be ``plugged-in.
• On the empirical estimation of integral probability metrics
• Optimal spatial adaptation to inhomogeneous smoothness: An approach based on kernel estimates with variable bandwidth selectors
• Rademacher Chaos Complexities for Learning the Kernel Problem
• Rademacher averages and phase transitions in Glivenko-Cantelli classes
• Real Analysis and Probability
• Relationships between Donsker classes and Sobolev spaces
• Scattered Data Approximation
• Support Vector Machines
• Theory of Reproducing Kernels
• Uniform Central Limit Theorems
• Uniform central limit theorems for kernel density estimators
• Uniform limit theorems for wavelet density estimators
• Universality, Characteristic Kernels and RKHS Embedding of Measures
• Wavelets, approximation, and statistical applications
• Weak convergence and empirical processes. With applications to statistics

Cited in

(13)

• Minimax estimation of kernel mean embeddings
• Kernel distribution embeddings: universal kernels, characteristic kernels and kernel metrics on distributions
• Ivanov-regularised least-squares estimators over large RKHSs and their interpolation spaces
• scientific article; zbMATH DE number 7758314 (Why is no real title available?)
• Directional differentiability for supremum-type functionals: statistical applications
• Spectral regularized Kernel two-sample tests
• Applications of optimal transportation. Abstracts from the workshop held February 4--9, 2024
• Variational methods for evolution. Abstracts from the workshop held December 5--8, 2023
• Estimation from pairwise comparisons: sharp minimax bounds with topology dependence
• Geometrical Insights for Implicit Generative Modeling
• A uniform kernel trick for high and infinite-dimensional two-sample problems
• A note on estimation and information topologies
• The topological foundation and some properties of the mixed estimator

This page was built for publication: On the optimal estimation of probability measures in weak and strong topologies

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q282569)