Year: 2008
Type: preprint
Abstract: In this paper we treat the problem of updating the\nQR factorization, with applications to the least\nsquares problem. Algorithms are presented that compute\nthe factorization A1 = Q1 R1, where A1 is the matrix\nA = QR after it has had a number of rows or columns\nadded or deleted. This is achieved by updating the\nfactors Q and R, and we show this can be much faster\nthan computing the factorization of A1 from scratch.\nWe consider algorithms that exploit the Level 3 BLAS\nwhere possible and place no restriction on the dimensions\nof A or the number of rows and columns added or deleted.\nFor some of our algorithms we present Fortran 77\nLAPACK-style code and show the backward error of our\nupdated factors is comparable to the error bounds of the\nQR factorization of A1.
Authors: Sven Hammarling, Craig Lucas
Institution: University of Manchester
Language: English
Topic: Matrix Theory and Algorithms
Subfield: Computational Theory and Mathematics
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W69941433
