Optimization, Engineering, and Cluster Computing

Tamara G. Kolda, Sandia National Labs

IMA Workshop on Connecting Women in Mathematical Sciences to Industry
Minneapolis, MN
September 8-11, 2000

Abstract

We present the Asynchronous Parallel Pattern Search (APPS) algorithm for optimization. Its development was driven by the need to adapt the standard parallel pattern search method to a cluster computing environment, where proceesors may have different speeds and faults of individual processors are frequent. This adaptation required us to redesign the original algorithm, and resulted in an overall faster method.

Slides

postscript or pdf

Movies

APPS (kolda-apps.mpg, 10,060 kb) and APPS with Faults (kolda-apps-ft.mpg, 11,132 kb)

A note on playing the movies: The recommended player for unix based systems in mpeg_play, and the recommended player for windows based systems is vmpeg.


Maintained by: Tamara Kolda (tgkolda@sandia.gov).
Disclaimer and Acknowledgment.