Greedy givens algorithms for computing the rank-k updating of the QR decomposition

Authors:
Erricos John Kontoghiorghes
Affiliations:
Institut d'informatique, Université de Neuchâtel, Émile-Argand 11, Case Postale 2, CH-2007 Neuchâtel, Switzerland
Venue:
Parallel Computing
Year:
2002

Citing 8
Cited 1

Parallel QR Decomposition of a rectangular matrix

Numerische Mathematik
Adaptive condition estimation for bank-one updates of QR factorizations

SIAM Journal on Matrix Analysis and Applications
Optimal algorithms for parallel Givens factorization on a coarse-grained PRAM

Journal of the ACM (JACM)
Matrix computations (3rd ed.)

Matrix computations (3rd ed.)
On Stable Parallel Linear System Solvers

Journal of the ACM (JACM)
Parallel Strategies for Solving SURE Models with Variance Inequalities and Positivity of Correlations Constraints

Computational Economics - Computational Studies at Stanford
Load balanced parallel QR decomposition on shared memory multiprocessors

Parallel Computing
Parallel Strategies for Rank-k Updating of the QR Decomposition

SIAM Journal on Matrix Analysis and Applications

Pipeline Givens sequences for computing the QR decomposition on a EREW PRAM

Parallel Computing

Quantified Score

Hi-index	0.00

Visualization

Abstract

A Greedy Givens algorithm for computing the rank-1 updating of the QR decomposition is proposed. An exclusive-read exclusive-write parallel random access machine computational model is assumed. The complexity of the algorithms is calculated in two different ways. In the unlimited parallelism case a single time unit is required to apply a compound disjoint Givens rotation of any size. In the limited parallelism case all the disjoint Givens rotations can be applied simultaneously, but one time unit is required to apply a rotation to a two-element vector. The proposed Greedy algorithm requires approximately 5/8 the number of steps performed by the conventional sequential Givens rank-1 algorithm under unlimited parallelism. A parallel implementation of the sequential Givens algorithm outperforms the Greedy one under limited parallelism. An adaptation of the Greedy algorithm to compute the rank-k updating of the QR decomposition has been developed. This algorithm outperforms a recently reported parallel method for small k, but its efficiency decreases as k increases.