Basis selection, universal sampling discretization, and sampling recovery

DOI10.48550/ARXIV.2611.10402MaRDI QIDQ7424538arXivFDO

Authors Feng Dai, Jiashu Ran, Vladimir N. Temlyakov

Publication date 7 November 2026

Published in arXiv

Tldr We construct an orthonormal basis for any $N$-dimensional subspace that allows for efficient discrete approximation to the squared $L_2$ norm of all $v$-term combinations with at most $Cv\log(eN/v)$ samples, achieving sharp estimates up to constant factors.

Full work available at URL https://doi.org/10.48550/ARXIV.2611.10402

arXiv classification

math.FAmath.NA

Cites work

• A remark on constructive covering of a ball of finite-dimensional Banach space
• A survey of sampling discretization of integral and uniform norms
• Greedy approximation
• High-Dimensional Probability: An Introduction with Applications in Data Science
• Interlacing families. II: Mixed characteristic polynomials and the Kadison-Singer problem
• Multivariate approximation
• Optimal sampling for least-squares approximation
• Random points are good for universal discretization
• Remarks on sampling discretization of integral norms of functions
• Sampling recovery on function classes with a structural condition
• Sampling recovery on function classes with a structural condition
• Separated sets and Auerbach systems in Banach spaces
• Sparse approximation and sampling recovery on function classes with a structural condition
• Sparse recovery in bounded Riesz systems with applications to numerical methods for PDEs
• Sparse sampling recovery by greedy algorithms
• Sparse sampling recovery in integral norms on some function classes
• Towards optimal sampling for learning sparse approximation in high dimensions
• Universal sampling discretization

This page was built for publication: Basis selection, universal sampling discretization, and sampling recovery

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7424538)