Nonparametric estimation of composite functions

DOI10.1214/08-AOS611MaRDI QIDQ1018644zbMATH OpenOpenAlexFDO

Authors Anatoli Juditsky, Alexandre B. Tsybakov, O. V. Lepski

Publication date 20 May 2009

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

Full work available at URL https://arxiv.org/abs/0906.0865

zbMATH Keywords

minimax estimationcomposite functionslocal modelsmultidimensional nonparametric estimationpointwise selection rule

Mathematics Subject Classification ID

Nonparametric estimation (62G05) Density estimation (62G07) Asymptotic properties of nonparametric inference (62G20) Markov processes: estimation; hidden Markov models (62M05) Estimation in multivariate analysis (62H12)

Abstract: We study the problem of nonparametric estimation of a multivariate function $g:mathbb{R}^{d}omathbbR$ that can be represented as a composition of two unknown smooth functions $f:mathbbRomathbbR$ and $G:mathbb{R}^{d}omathbbR$. We suppose that $f$ and $G$ belong to known smoothness classes of functions, with smoothness $gamma$ and $$, respectively. We obtain the full description of minimax rates of estimation of $g$ in terms of $gamma$ and $$, and propose rate-optimal estimators for the sup-norm loss. For the construction of such estimators, we first prove an approximation result for composite functions that may have an independent interest, and then a result on adaptation to the local structure. Interestingly, the construction of rate-optimal estimators for composite functions (with given, fixed smoothness) needs adaptation, but not in the traditional sense: it is now adaptation to the local structure. We prove that composition models generate only two types of local structures: the local single-index model and the local model with roughness isolated to a single dimension (i.e., a model containing elements of both additive and single-index structure). We also find the zones of ($gamma$, $$) where no local structure is generated, as well as the zones where the composition modeling leads to faster rates, as compared to the classical nonparametric rates that depend only to the overall smoothness of $g$.

Recommendations

• scientific article; zbMATH DE number 2099990
• scientific article; zbMATH DE number 549934
• scientific article; zbMATH DE number 3862231
• Nonparametric estimation of homogeneous functions
• Parameter estimation based upon nonparametric function estimators
• scientific article; zbMATH DE number 850152
• scientific article; zbMATH DE number 1215456
• Nonparametric estimation of distribution functions
• Nonparametric estimates of distribution functions

Cites work

• scientific article; zbMATH DE number 3976099 (Why is no real title available?)
• scientific article; zbMATH DE number 3733065 (Why is no real title available?)
• scientific article; zbMATH DE number 177229 (Why is no real title available?)
• Additive regression and other nonparametric models
• Asymptotic equivalence for nonparametric regression with multivariate and random design
• Asymptotic minimax estimation of regression in the additive model
• Asymptotic minimax risk for sup-norm loss: Solution via optimal recovery
• Asymptotically Minimax Adaptive Estimation. I: Upper Bounds. Optimally Adaptive Estimates
• Asymptotically exact minimax estimation in sup-norm for anisotropic Hölder classes
• Asymptotically minimax adaptive estimation. II: Schemes without optimal adaption. Adaptive estimates
• Bounds for the Risks of Non-Parametric Regression Estimates
• Direct estimation of the index coefficient in a single-index model
• Hypothesis testing under composite functions alternative
• New tight frames of curvelets and optimal representations of objects with piecewise C² singularities
• Nonlinear Estimation in Anisotropic Multi-Index Denoising. Sparse Case
• Nonlinear estimation in anisotropic multi-index denoising
• Nonparametric Estimation of a Regression Function that is Smooth in a Domain in $R^k$
• Optimal global rates of convergence for nonparametric regression
• Optimal pointwise adaptive methods in nonparametric estimation
• Optimal spatial adaptation to inhomogeneous smoothness: An approach based on kernel estimates with variable bandwidth selectors
• Rate-optimal estimation for a general class of nonparametric regression models with unknown link functions
• Structural adaptation via \(\mathbb L_p\)-norm oracle inequalities
• Universal approximation bounds for superpositions of a sigmoidal function
• Wedgelets: Nearly minimax estimation of edges

Cited in

(22)

• Structural adaptation via \(\mathbb L_p\)-norm oracle inequalities
• Nonparametric regression using deep neural networks with ReLU activation function
• Minimax and minimax adaptive estimation in multiplicative regression: locally Bayesian approach
• A new approach to estimator selection
• Rate-optimal estimation for a general class of nonparametric regression models with unknown link functions
• Estimation of the transition density of a Markov chain
• Structural adaptive deconvolution under \({\mathbb{L}_p}\)-losses
• Adaptive estimation over anisotropic functional classes via oracle approach
• Statistical inference for time-changed Lévy processes via composite characteristic function estimation
• Adaptive global thresholding on the sphere
• Adaptive nonparametric regression on spin fiber bundles
• Estimating composite functions by model selection
• A generic approach to nonparametric function estimation with mixed data
• Theory of adaptive estimation
• Estimation and Calibration of Lévy Models via Fourier Methods
• \(\mathbb{L}_{p}\) adaptive estimation of an anisotropic density under independence hypothesis
• Multivariate intensity estimation via hyperbolic wavelet selection
• Capturing ridge functions in high dimensions from point queries
• Model selection for Poisson processes with covariates
• Testing composite hypotheses, Hermite polynomials and optimal estimation of a nonsmooth functional
• Structural adaptation in the density model
• Oracle-efficient nonparametric estimation of an additive model with an unknown link function

This page was built for publication: Nonparametric estimation of composite functions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1018644)