Salvato in:
Dettagli Bibliografici
Autori principali: Naumov, Alexey, Rakhuba, Maxim, Ryapolov, Denis, Samsonov, Sergey
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2506.15660
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911012360290304
author Naumov, Alexey
Rakhuba, Maxim
Ryapolov, Denis
Samsonov, Sergey
author_facet Naumov, Alexey
Rakhuba, Maxim
Ryapolov, Denis
Samsonov, Sergey
contents We consider the problem of estimating the spectral norm of a matrix using only matrix-vector products. We propose a new Counterbalance estimator that provides upper bounds on the norm and derive probabilistic guarantees on its underestimation. Compared to standard approaches such as the power method, the proposed estimator produces significantly tighter upper bounds in both synthetic and real-world settings. Our method is especially effective for matrices with fast-decaying spectra, such as those arising in deep learning and inverse problems.
format Preprint
id arxiv_https___arxiv_org_abs_2506_15660
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle On the Upper Bounds for the Matrix Spectral Norm
Naumov, Alexey
Rakhuba, Maxim
Ryapolov, Denis
Samsonov, Sergey
Numerical Analysis
Machine Learning
Statistics Theory
65F35
We consider the problem of estimating the spectral norm of a matrix using only matrix-vector products. We propose a new Counterbalance estimator that provides upper bounds on the norm and derive probabilistic guarantees on its underestimation. Compared to standard approaches such as the power method, the proposed estimator produces significantly tighter upper bounds in both synthetic and real-world settings. Our method is especially effective for matrices with fast-decaying spectra, such as those arising in deep learning and inverse problems.
title On the Upper Bounds for the Matrix Spectral Norm
topic Numerical Analysis
Machine Learning
Statistics Theory
65F35
url https://arxiv.org/abs/2506.15660