On computation of matrix logarithm times a vector

Matrix functions have become a central topic in linear algebra, and many problems of their numerical approximation appear often in scientific computing. This thesis concerns with matrix functions times a vector with a special attention in the matrix logarithm case. In many applications the original...

ver descrição completa

Detalhes bibliográficos
Autor: Gimeno Alquézar, Joan
Formato: tesis de maestría
Fecha de publicación:2017
País:España
Recursos:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/107140
Acesso em linha:https://hdl.handle.net/2117/107140
Access Level:acceso abierto
Palavra-chave:Computer algorithms
Computer programming
Algorismes computacionals
Programació (Ordinadors)
Àrees temàtiques de la UPC::Informàtica
Descrição
Resumo:Matrix functions have become a central topic in linear algebra, and many problems of their numerical approximation appear often in scientific computing. This thesis concerns with matrix functions times a vector with a special attention in the matrix logarithm case. In many applications the original matrix may be large, sparse or structured. In this case evaluating the matrix function times a vector by first computing the full matrix function is usually unfeasible, so that it has sense to approximate the solution saving storage and computational time. Looking into the literature in numerical linear algebra, the standard approach for computing the matrix function times a vectors directly is based on a polynomial Krylov subspace approach that only requires matrix–vector products of the original matrix. This project deals with rational Krylov subspace which have been used recently in this context though it was originally presented for eigenvalue problem in the 90s.