Time: October 15, 2009, 1125-1150 hours
Place: Erskine 415
Speaker: Brendan Bycroft

Title:
Parallelizing the Global Optimization Algorithm

Abstract:
This presentation explains the global optimisation algorithm, and looks
at parallelization as it is used improve the speed of the algorithm when
applied to computers with multiple CPU cores. The ideas and
computational methods behind parallelization are examined; in
particular, those used in the C++ library Intel Threading Building
Blocks are explored.

Both successful and unsuccessful attempts of the global optimisation
algorithm using parallelization are looked at, demonstrating the
benefits and pitfalls that the technique provides.