Stochastic trace estimation for positive trace-class operators

DOI10.48550/arXiv.2608.28135MaRDI QIDQ7285080arXivFDO

Authors Zvonimir Bujanović, Luka Grubišić, Daniel Kressner, Hrvoje Olić

Publication date 28 August 2026

Published in arXiv

Tldr We develop stochastic trace estimation algorithms for positive trace-class operators on separable Hilbert spaces based on Gaussian random elements whose covariance is determined by the target operator; we prove high-probability error bounds analogous to finite-dimensional matrix results, including O(ε^{-1}) sample complexity for idealized infHutch++.

Full work available at URL https://doi.org/10.48550/arXiv.2608.28135

Mathematics Subject Classification ID

Determinants, permanents, traces, other special matrix functions (15A15) Applications of functional analysis in numerical analysis (46N40) Randomized algorithms (68W20)

arXiv classification

math.NA

This page was built for publication: Stochastic trace estimation for positive trace-class operators

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7285080)