8:30 AM-9:30 AM
Chair: John R. Gilbert, Xerox Palo Alto Research Center
Room: Ballroom 3
As the power of our computers grows, so generally does our appetite for more accurate solutions to ever harder problems. In the arena of sparse symmetric eigenproblems, the size of problems solved routinely in commercial settings has grown over the last decade from finding perhaps ten eigenpairs of order ten thousand matrices to finding thousands of eigenpairs of generalized symmetric eigenproblems of several millionth order.
This talk will focus on three important characteristics of these problems and their potential solutions:
(1)the potential deleterious effects of size on guarantees, robustness and computational feasibility; (2)the characteristics of algorithms that have been proposed or are being used for these eigenproblems; and (3)the need to adapt our mathematics to today's and tomorrow's computer architectures and programming paradigms, without which we will not have sufficient cycles to solve these problems.
These problems are being solved to build products that you and I will buy or use. The solutions to these problems improve the products or ensure their safety; we have a stake in the solutions. Can we, in fact, get them right in a reasonable amount of time? Can we package our mathematics so that commercial vendors will use our work?
John G. Lewis
Boeing Information and Support Services
The Boeing Company
LA97 Homepage | Program Updates| Registration | Hotel Information | Transportation | Program-at-a-Glance |