SC²S Colloquium - July 16, 2014

From Sccswiki
Jump to navigation Jump to search
Date: July 16, 2014
Room: 02.07.023
Time: 3 pm, s.t.

Christopher Ziegler: Approximate Inverse Preconditioner with approximated L1-Norm Minimization

In this presentation, we will discuss an algorithm, which is designed to improve a preconditioner of a general sparse linear equation system. It is based on the SPAI and inherently parallel. The sparsity pattern of the preconditioner will not be changed.It is done by iteratively minimizing the L2-norm of D(AM_{s}-I), where A is the system matrix, M_{s} is the s-th column of the preconditioner, I is the identity matrix and D is a diagonal matrix, storing the sqare root of the absolute entries of AM_{s}-I. Matrices coming from various applications are used to test the behaviour of the algorithm.