{"id":"arxiv/2608.13714","url":"https://arxiv.org/abs/2608.13714","tags":["math.NA"],"score":0,"title":"Approximating matrix functions by block Krylov methods with randomized vectors","domain":"math.NA","source":"arxiv","authors":["Josh Kane","Lucas Onisk","Lothar Reichel","Giuseppe Rodriguez"],"reviews":0,"subject":"math.NA","summary":"The need to evaluate expressions of the form $f(A)\\mathbf{b}$, where $A$ is a square matrix, $f$ is a function, and $\\mathbf{b}$ is a vector, arises in several areas of applied mathematics. When the matrix $A$ is very large, it is usually not attractive to evaluate $f(A)$. Instead, $f(A)\\mathbf{b}$ often is approximated by computing an estimate in a Krylov subspace that depends on $A$ and $\\mathbf{b}$, and only requires that $f$ be evaluated at a small matrix. This paper explores the application of several variants of randomized block Krylov methods to the approximation of $f(A)\\mathbf{b}$.…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-08-13","reactions":{}}