Abstract:In this work, we construct a stable and fairly fast estimator for solving multidimensional non-parametric regression problems. The proposed estimator is based on the use of a novel and special system of multivariate Jacobi polynomials that generate a basis for a reduced size of variate finite dimensional polynomials space. An ANOVA decomposition trick has been used for building this space. Also, by using some results from the theory of positive definite random matrices, we show that the proposed estimator is stable under the condition that the i.i.d. dimensional random sampling training points follow a dimensional Beta distribution. In addition, we provide the reader with an estimate for the risk error of the estimator. This risk error depends on the error of the orthogonal projection error of the regression function over the considered polynomials space. An involved study of this orthogonal projection error is done under the condition that the regression function belongs to a given weighted Sobolev space. Thanks to this novel estimate of the orthogonal projection error, we give the optimal convergence rate of our estimator. Furthermore, we give a regularized extension version of our estimator, that is capable of handling random sampling training vectors drawn according to an unknown multivariate pdf. Moreover, we derive an upper bound for the empirical risk error of this regularized estimator. Finally, we give some numerical simulations that illustrate the various theoretical results of this work. In particular, we provide simulations on a real data that compares the performance of our estimator with some existing and popular NP regression estimators.

Consistent estimation with the use of orthogonal projections for a linear regression model with errors in the variables

Strong consistency of an estimator by the truncated singular value decomposition for an errors-in-variables regression model with collinearity

An Error-in-variable Projection to Latent Structure Framework for Monitoring Technical Systems with Orthogonal Signal Components

Faithful Variable Screening for High-Dimensional Convex Regression

Asymptotic Behavior for S-estimators in Random Design Linear Model with Long-Range-dependent Errors

Total Least L1- and L2-norm Estimations of a Coordinate Transformation Model with a Structured Parameter Matrix

Linear and Conic Programming Estimators in High-Dimensional Errors-in-variables Models

The rates of strong consistency for estimators in heteroscedastic partially linear errors-in-variables model for widely orthant dependent samples

Local Linear Multiple Regression with Variable Bandwidth in the Presence of Heteroscedasticity

Orthogonality-projection-based estimation for semi-varying coefficient models with heteroscedastic errors.

Local Polynomial Estimator of the Regression Function in Autoregressive Models with Errors in Variables

Correcting the estimator for the mean vectors in a multivariate errors-in-variables regression model

Variable selection using axis-aligned random projections for partial least-squares regression

On some stable linear functional regression estimators based on random projections

Asymptotics of Estimates in Constrained Nonlinear Regression with Long-Range Dependent Innovations

A multivariate Jacobi polynomials regression estimator associated with an ANOVA decomposition model

Linearization Estimates of the Backward Errors for Least Squares Problems.

$\ell_1$-Regularized Generalized Least Squares

Low-rank matrix estimation via nonconvex spectral regularized methods in errors-in-variables matrix regression

Comparison of Structured and Weighted Total Least-Squares Adjustment Methods for Linearly Structured Errors-In-Variables Models

The Analysis of Completely Perturbed Model Based on RIP Via Orthogonal Least Squares