Massively parallel computing: a statistical application

Authors: BENN A.; KULPERGER R.

Source: Statistics and Computing, Volume 8, Number 4, December 1998 , pp. 309-318(10)

Publisher: Springer

Buy & download fulltext article:

Price: $47.00 plus tax (Refund Policy)

Abstract:

Massively parallel computing is a computing environment with thousands of subprocessors. It requires some special programming methods, but is well suited to certain imaging problems. One such statistical example is discussed in this paper. In addition there are other natural statistical problems for which this technology is well suited. This paper describes our experience, as statisticians, with a massively parallel computer in a problem of image correlation spectroscopy. Even with this computing environment some direct computations would still take in the order of a year to finish. It is shown that some of the algorithms of interest can be made parallel.

Keywords: Massively parallel computing; image; correlation spectroscopy; Fourier transform; periodogram; inference; parallel algorithms

Language: English

Document Type: Regular paper

Affiliations: 1: Department of Statistical and Actuarial Sciences, The University of Western Ontario, London, Ontario N6A 5B7, Canada

Publication date: 1998-12-01