On generalized parameterized inexact Uzawa methods for singular saddle-point problems (Q2356071)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | On generalized parameterized inexact Uzawa methods for singular saddle-point problems |
scientific article |
Statements
On generalized parameterized inexact Uzawa methods for singular saddle-point problems (English)
0 references
28 July 2015
0 references
The authors investigate a generalized version of the parameterized inexact Uzawa (PIU) method, named as generalized PIU (GPIU) method, for solving singular saddle point problems of the form \[ Ax\equiv \begin{bmatrix} W & B \\ -B^T & 0 \end{bmatrix} \begin{bmatrix} u \\ v \end{bmatrix} = \begin{bmatrix} f\\-g \end{bmatrix} = b, \] where \(W \in \mathbb{R}^{n \times n}\) is a symmetric positive definite matrix, \(B\in \mathbb{R}^{n \times m}\) is a rectangular matrix, \(b \in \mathbb{R}^{n+m}\) is a given vector in the range of the saddle-point matrix \(A \in \mathbb{R}^{(n+m)\times (n+m)}\). For \(B\) of the full column rank, the matrix \(A\) is nonsingular. The authors say that in this case theoretical analysis and numerical experiments have shown that the PIU method is feasible and effective for solving large sparse nonsingular saddle-point problems. When matrix \(B\) is rank deficient, the saddle-point matrix \(A\) is singular and a singular saddle-point problem is obtained. In this case, the authors suggest a generalized version GPIU of PIU algorithm, in which instead of inverse matrices the Moore-Penrose inverse is used. Sufficient and necessary conditions for the convergence of the suggested GPIU iteration scheme and the quasi-optimal choice of the iteration parameters \(\omega\) and \(\tau\) of the GPIU method are derived and discussed. The analysis of the semi-convergence properties of the GPIU method shows that the iterative sequence produced by the GPIU method converges to a solution of the singular saddle-point problem under suitable restrictions on the iteration parameters \(\omega\) and \(\tau\). Numerical results verify the efficiency of the GPIU method. However, the calculations of quasi-optimal values of parameters \(\omega\) and \(\tau\) are still too time-consuming. The authors verify the robustness and effectiveness of seven types of GPIU methods for different choices of preconditioning matrices and compare the numerical results with those of the quasi-minimal residual (QMR), the generalized minimum residual (GMRES) and the restarted GMRES methods.
0 references
singular saddle-point problems
0 references
generalized singular saddle-point problems
0 references
Uzawa method
0 references
semi-convergence property
0 references
Moore-Penrose inverse
0 references
quasi-optimal values of iteration parameters
0 references
sparse matrices
0 references
numerical experiment
0 references
algorithm
0 references
preconditioning
0 references
quasi-minimal residual method
0 references
generalized minimum residual method
0 references
restarted GMRES method
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references