Seminars

Statistical Surrogates for Auto-Tuning in Scientific Computing

86
reads

Weichung Wang

2012-11-30
12:30:00 - 14:30:00

103 , Mathematics Research Center Building (ori. New Math. Bldg.)

Fast evolving computing technologies have kept leading us to the discoveries of novel and exciting scientific frontiers. However, to design efficient algorithms on these powerful computers remains a challenge. The challenge is partially due to the computer architecture complexities and the lack of equation-based performance models. Surrogate based auto-tuning is one promising way to overcome these obstacles. In this talk, we will demonstrate how data-driven statistical surrogates, software auto-tuning, and high-performance computational methods can benefit each other. Examples in space-filling designs, chaotic light sources, photonic crystal bandgap, medical imaging, and eigenvalue solver will be presented to illustrate the interdisciplinary nature of our approaches.