WS, Ronald Kriemann: H-Matrices: Arithmetic, Solvers and Preconditioners

Speaker
Ronald Kriemann, Max-Planck Institute for the Mathematical Sciences

Abstract
Hierarchical matrices (H-matrices) allow to represent various dense data, e.g., from integral equations, Schur complements or sparse matrix inverses, in (almost) linear storage complexity. Furthermore, full matrix arithmetic is possible with the same runtime complexity, which enables efficient handling of large scale problems.

We will introduce general properties of H-matrices, their construction and fundamental blocks of their arithmetic. Furthermore, different aspects of their application as preconditioners and direct solvers are presented.