Adaptive Randomized Pivoting for Tensor Singular Value Decomposition Model

DOI10.48550/arXiv.2606.26688MaRDI QIDQ7270111arXivFDO

Authors Valentin Leplat, Ahmadsho Akdodshoev, Salman Ahmadi-Asl

Publication date 25 June 2026

Published in arXiv

Tldr We propose adaptive randomized pivoting for tensor singular value decomposition (SVD) model and derive two constructions: ARP-T-CUR and T-ARP. The first construction applies matrix ARP to frontal slices in the Fourier domain, yielding a direct expected-error bound inherited from matrix theory. The second construction selects common lateral and horizontal slices across all frequencies with explicit coupling of pivot indices, and we prove an expected-error bound under a frequency-alignment condition measuring how far the common tensor-level sampling rule is from slice-wise ARP sampling rules.

Full work available at URL https://doi.org/10.48550/arXiv.2606.26688

arXiv classification

math.NA

This page was built for publication: Adaptive Randomized Pivoting for Tensor Singular Value Decomposition Model

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7270111)