A Preconditioner for Generalized Saddle Point Problems (Q4650981)
From MaRDI portal
!
This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:
scientific article; zbMATH DE number 2135306
| Language | Label | Description | Also known as |
|---|---|---|---|
| default for all languages | No label defined |
||
| English | A Preconditioner for Generalized Saddle Point Problems |
scientific article; zbMATH DE number 2135306 |
Statements
A Preconditioner for Generalized Saddle Point Problems (English)
0 references
21 February 2005
0 references
saddle point problem
0 references
matrix splittings
0 references
iterative methods
0 references
preconditioning
0 references
large scale sparse systems
0 references
Krylov subspace methods
0 references
GMRES
0 references
numerical experiments
0 references
alternating direction method
0 references
diagonal scaling
0 references
convergence
0 references
0.8859763145446777
0 references
0.8797867298126221
0 references
0.8766528964042664
0 references
0.8760521411895752
0 references
The authors consider the solution of large scale sparse systems of linear algebraic equations of a generalized saddle point type, i.e., of systems of a block \(2 \times 2\) structure \[ \left ( \begin{matrix} A & B^T \\ B & -C \end{matrix}\right) \left (\begin{matrix} u \\ p \end{matrix}\right) = \left (\begin{matrix} f \\ g \end{matrix}\right) , \] where \(A \in {\mathbb{R}}^{n \times n}\), \(B \in {\mathbb{R}}^{ m \times n}\), \(C \in {\mathbb{R}}^{m \times m}\), \(f \in {\mathbb{R}}^n\), \(g \in {\mathbb{R}}^m\), and \(m \leq n\) with the assumptions: \(A\) has a positive semidefinite part \(H =\frac{1}{2} (A+A^T)\); \(rank(B) = m\); \(\mathcal{N}(H) \cap \mathcal{N}(B) = \{0\}\); \(C\) is symmetric positive semidefinite. The existence and uniqueness of the solution is proved. The authors propose an alternating splitting iteration for the solution of such systems and show that this iteration can be efficiently used as a preconditioner for Krylov subspace methods like GMRES or its restarted version rather than as a stand-alone solver. Numerical experiments with matrices from several different applications illustrate that the method is not overly sensitive to the choice of an optimization parameter \(\alpha\) which has been introduced into the alternating splitting iteration similar to the classical alternating direction method. The numerical examples also show that diagonal scaling additionally improves the convergence of the outer iteration.
0 references