Adaptation in multivariate log-concave density estimation

DOI10.1214/20-AOS1950MaRDI QIDQ2656591zbMATH OpenOpenAlexFDO

Authors Oliver Y. Feng, Adityanand Guntuboyina, Arlene K. H. Kim, Richard Samworth

Publication date 11 March 2021

Published in The Annals of Statistics (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1812.11634, https://projecteuclid.org/euclid.aos/1611889221

zbMATH Keywords

bracketing entropycontour separationlog-concavitymaximum-likelihood estimationmultivariate adaption

Mathematics Subject Classification ID

Inequalities; stochastic orderings (60E15) Density estimation (62G07) Asymptotic properties of nonparametric inference (62G20) Estimation in multivariate analysis (62H12)

Abstract: We study the adaptation properties of the multivariate log-concave maximum likelihood estimator over three subclasses of log-concave densities. The first consists of densities with polyhedral support whose logarithms are piecewise affine. The complexity of such densities~$f$ can be measured in terms of the sum $Gamma(f)$ of the numbers of facets of the subdomains in the polyhedral subdivision of the support induced by $f$. Given $n$ independent observations from a $d$-dimensional log-concave density with $din2,3$, we prove a sharp oracle inequality, which in particular implies that the Kullback--Leibler risk of the log-concave maximum likelihood estimator for such densities is bounded above by $Gamma(f)/n$, up to a polylogarithmic factor. Thus, the rate can be essentially parametric, even in this multivariate setting. For the second type of adaptation, we consider densities that are bounded away from zero on a polytopal support; we show that up to polylogarithmic factors, the log-concave maximum likelihood estimator attains the rate $n^{-4/7}$ when $d=3$, which is faster than the worst-case rate of $n^{-1/2}$. Finally, our third type of subclass consists of densities whose contours are well-separated; these new classes are constructed to be affine invariant and turn out to contain a wide variety of densities, including those that satisfy H"older regularity conditions. Here, we prove another sharp oracle inequality, which reveals in particular that the log-concave maximum likelihood estimator attains a risk bound of order $$ when $d=3$ over the class of $$-H"older log-concave densities with $$, again up to a polylogarithmic factor.

Review text

The research belongs to the domain of multivariate adaptation behavior in log-concave density estimation. Denote by \(\mathcal{F_d}\) the class of upper semicontinuous, log-concave densities on \(\mathbb{R}^d\) and assume that \(X_1,\dots,X_n\) are i.i.d random vectors with density \(\rho_0 \in \mathcal{F_d} \). The aim of the article is to investigate the potential of the log-concave maximum likelihood estimator to adopt to three different types of subclasses of \(\mathcal{F_d}\). In the second section, the structure of log-\(k\)-affine functions with polyhedral support is introduced and the results on adaptation to log-\(k\)-affine densities with polyhedral support are presented. In the third section, one considers densities from \(\mathcal{F_d}\), close to a log-concave density with polyhedral support but without being log-\(k\)-affine. A sharp oracle inequality for the log-concave maximum likelihood estimator is shown and discussed. In the fourth section, a subclass of \(\mathcal{F_d}\), the so-called densities with well-separated contours is defined. The main result in this section is then a sharp oracle inequality for the performance of the log-concave maximum likelihood estimator over this subclass. Proofs of the main results in the second section are given in the APPENDIX part. All other proofs are contained in the supplement to this paper, \url{doi:10.1214/20-A=S1950SUPP; .pdf}.

Recommendations

• Adaptation in log-concave density estimation
• Adaptive log-density estimation
• Adaptive estimation of the mode of a multivariate density
• Adaptive Estimation of a Conditional Density
• Pointwise adaptive estimation of a multivariate density under independence hypothesis
• On adaptive estimation of probability density functions
• \(L^ p\) adaptive density estimation
• Multiscale adaptive inference on conditional moment inequalities

Cites work

• A Computational Framework for Multivariate Convex Regression and Its Variants
• A Framework For Estimation Of Convex Functions
• A new perspective on least squares under convex constraint
• Adaptation in log-concave density estimation
• Adaptive estimation of convex polytopes and convex sets from noisy data
• Adaptive risk bounds in unimodal regression
• Approximation by log-concave distributions, with applications to regression
• Convergence of algorithms for reconstructing convex bodies and directional measures
• Convergence of linear functionals of the Grenander estimator under misspecification
• Convex Optimization, Shape Constraints, Compound Decisions, and Empirical Bayes Rules
• Covering Numbers for Convex Functions
• Current status linear regression
• Distribution-free properties of isotonic regression
• Estimating a density under order restrictions: Nonasymptotic minimax risk
• Faithful variable screening for high-dimensional convex regression
• Generalized additive and index models with shape constraints
• Global rates of convergence in log-concave density estimation
• Global rates of convergence of the MLEs of log-concave and \(s\)-concave densities
• scientific article; zbMATH DE number 5654889 (Why is no real title available?)
• Independent component analysis via nonparametric maximum likelihood estimation
• Inference and modeling with log-concave distributions
• Isotonic regression in general dimensions
• Limit distribution theory for maximum likelihood estimation of a log-concave density
• Living on the edge: phase transitions in convex programs with random data
• Log-concavity and strong log-concavity: a review
• Maximum likelihood estimation of a log-concave density and its distribution function: basic properties and uniform consistency
• Nonparametric estimation of multivariate convex-transformed densities
• Nonparametric estimation under shape constraints. Estimators, algorithms and asymptotics
• On adaptive minimax density estimation on R^d
• On matrix estimation under monotonicity constraints
• On risk bounds in isotonic and other shape restricted regression problems
• Optimal rates of convergence for convex set estimation from support functions
• Polytopes, Rings, and K-Theory
• Recent progress in log-concave density estimation
• Rho-estimators for shape restricted density estimation
• Risk bounds in isotonic regression
• Sharp oracle inequalities for least squares estimators in shape restricted regression
• Smoothed log-concave maximum likelihood estimation with applications
• Stochastically Transitive Models for Pairwise Comparisons: Statistical and Computational Issues

Cited in

(19)

• Geometry of log-concave density estimation
• Adaptation in log-concave density estimation
• Theoretical properties of the log-concave maximum likelihood estimator of a multidimensional density
• Local continuity of log-concave projection, with applications to estimation under model misspecification
• High-dimensional nonparametric density estimation via symmetry and shape constraints
• Confidence intervals for multiple isotonic regression and other monotone models
• A variational inference for the Lévy adaptive regression with multiple kernels
• Fast multivariate log-concave density estimation
• Adaptive estimation of the mode of a multivariate density
• Confidence Bands for a Log-Concave Density
• Density estimation for shift-invariant multidimensional distributions
• Global rates of convergence in log-concave density estimation
• Shape Constrained Kernel PDF and PMF Estimation
• Inference for Local Parameters in Convexity Constrained Models
• Nonparametric estimation of multivariate convex-transformed densities
• A new computational framework for log-concave density estimation
• Log-concave density estimation in undirected graphical models
• Robust density estimation with the \(\mathbb{L}_1\)-loss. Applications to the estimation of a density on the line satisfying a shape constraint
• A new adaptive local polynomial density estimation procedure on complicated domains

Describes a project that uses

Uses Software

• scar

This page was built for publication: Adaptation in multivariate log-concave density estimation

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2656591)