Scientific Computing Seminar
Title: Meeting the challenge of memory-efficient robust preconditioners
for sparse symmetric linear systems
Speaker: Dr. Jennifer Scott,
STFC Research Fellow and Numerical Analysis Group Leader
Scientific Computing Department
STFC Rutherford Appleton Laboratory, England
When: 3:10 – 4:00pm, Wednesday, October 7, 2015
Where: Room 1131, Kemper Hall
Host: Zhaojun Bai, firstname.lastname@example.org
The need to solve large sparse symmetric linear systems of equations arises in numerous practical applications. In many situations, an iterative method is the method of choice but a preconditioned is normally required for this to be effective. Over the past fifty years, many different algorithms for computing incomplete factorizations have been proposed. In this talk, we focus on the development of new robust incomplete factorization algorithms that can be used to compute high quality pre conditioners for positive definite systems and for general indefinite systems, including saddle-point problems.
A limited memory approach is used in which the user can control both the sparsity density of the incomplete factor and the amount of memory used in its computation. We discuss some of the key ideas and illustrate the effectiveness of our approach using the software we have developed for the HSL mathematical software library applied to problems from a wide range of application areas. This is a joint work with Miroslav Tuma.
1131 Kemper Hall