Communication avoiding low rank approximation based on QR with tournament pivoting - Archive ouverte HAL Access content directly
Preprints, Working Papers, ... Year :

Communication avoiding low rank approximation based on QR with tournament pivoting

(1) , (1)
1

Abstract

We introduce a parallel algorithm for computing the low rank approximation $A_k$ of a large matrix $A$ which minimizes the number of messages exchanged between processors (modulo polylogarithmic factors) and has guarantees for the approximations of the singular values of $A$ provided by $A_k$. This operation is essential in many applications in scientific computing and data analysis when dealing with large data sets. Our algorithm is based on QR factorization that consists in selecting a subset of columns from the matrix $A$ that allow to approximate the range of $A$, and then projecting the columns of $A$ on a basis of the subspace spanned by those columns. The selection of columns is performed by using tournament pivoting, a strategy introduced previously for matrices partitioned into blocks of columns. This strategy is extended here to matrices partitioned along both dimensions that are distributed on a two-dimensional grid of processors, and also to tournaments with more general reduction trees. Performance results show that the algorithm scales well on up to $1024$ cores of $16$ nodes.
Fichier principal
Vignette du fichier
qr_with_tournament_pivoting.pdf (8.04 Mo) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-02947991 , version 1 (24-09-2020)
hal-02947991 , version 2 (18-01-2021)

Identifiers

  • HAL Id : hal-02947991 , version 2

Cite

Matthias Beaupère, Laura Grigori. Communication avoiding low rank approximation based on QR with tournament pivoting. 2021. ⟨hal-02947991v2⟩
436 View
307 Download

Share

Gmail Facebook Twitter LinkedIn More