Local approximation using Hermite functions

TL;DR

This paper introduces a wavelet-like representation of functions in L^p(R) using Fourier-Hermite coefficients, enabling local analysis of smoothness and providing new proofs for kernel localization and classical inequalities.

Contribution

It develops a novel local approximation method based on Hermite functions and offers new proofs for kernel localization and classical inequalities.

Findings

New wavelet-like representation for L^p functions

Characterization of local smoothness via Hermite coefficients

Alternative proofs for kernel localization and classical estimates

Abstract

We develop a wavelet like representation of functions in $L^p(\mathbb{R})$ based on their Fourier--Hermite coefficients; i.e., we describe an expansion of such functions where the local behavior of the terms characterize completely the local smoothness of the target function. In the case of continuous functions, a similar expansion is given based on the values of the functions at arbitrary points on the real line. In the process, we give new proofs for the localization of certain kernels, as well as some very classical estimates such as the Markov--Bernstein inequality.