Year: 2026
Type: article
Abstract: We consider matrices [Formula: see text] that depend, possibly nonlinearly, on a parameter [Formula: see text] from a compact parameter space [Formula: see text]. We present a Monte Carlo estimator for minimizing [Formula: see text] over all [Formula: see text], and determine the sampling amount so that the backward error of the estimator is bounded with high probability. We derive two types of bounds, based on epsilon nets and on generic chaining. Both types predict a small sampling amount for matrices [Formula: see text] with small offdiagonal mass, and parameter spaces [Formula: see text] of small “size.” Dependence on the matrix dimension [Formula: see text] is only weak or not explicit. The bounds based on epsilon nets are easier to evaluate and come with fully specified constants. In contrast, the bounds based on chaining depend on the Talagrand functionals, which are difficult to evaluate, except in very special cases. Comparisons between the two types of bounds are difficult, although the literature suggests that chaining bounds can be superior.
DOI: 10.1137/26m1854148
Authors: Arvind K. Saibaba, Ilse C. F. Ipsen
Institution: North Carolina State University
Language: English
Topic: Matrix Theory and Algorithms
Subfield: Computational Theory and Mathematics
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Awards: FRG: Collaborative Research: Randomization as a Resource for Rapid Prototyping, Collaborative Research: Randomized algorithms for dynamic and hierarchical Bayesian inverse problems, RTG: Randomized Numerical Analysis, A CONTEXTUALLY-AWARE SENSITIVITY ANALYSIS TO GUIDE THE DESIGN OF RANDOMIZED LEAST SQUARES SOLVERS IN APPLICATIONS, CAREER: Fast and Accurate Algorithms for Uncertainty Quantification in Large-Scale Inverse Problems, NSF-BSF: AF: Collaborative Research: Small: Randomized preconditioning of iterative processes: Theory and practice
OpenAlex ID: W4415191222
