|
|
Multilevel methods
Vacek, Petr ; Strakoš, Zdeněk (advisor)
The analysis of the convergence behavior of the multilevel methods is in the literature typically carried out under the assumption that the problem on the coarsest level is solved exactly. The aim of this thesis is to present a description of the multilevel methods which allows inexact solve on the coarsest level and to revisit selected results presented in literature using these weaker assumptions. In particular, we focus on the derivation of the uniform bound on the rate of convergence. Moreover, we discuss the possible dependence of the convergence behavior on the mesh size of the initial triangulation. 41
|
|
| |
|
| |
|
| |
|
|
Krylov Subspace Methods - Analysis and Application
Gergelits, Tomáš ; Strakoš, Zdeněk (advisor) ; Farrell, Patrick (referee) ; Herzog, Roland (referee)
Title: Krylov Subspace Methods - Analysis and Application Author: Tomáš Gergelits Department: Department of Numerical Mathematics Supervisor: prof. Ing. Zdeněk Strakoš, DrSc., Department of Numerical Mathematics Abstract: Convergence behavior of Krylov subspace methods is often studied for linear algebraic systems with symmetric positive definite matrices in terms of the condition number of the system matrix. As recalled in the first part of this thesis, their actual convergence behavior (that can be in practice also substantially affected by rounding errors) is however determined by the whole spectrum of the system matrix, and by the projections of the initial residual to the associated invariant subspaces. The core part of this thesis investigates the spectra of infinite dimensional operators −∇ · (k(x)∇) and −∇ · (K(x)∇), where k(x) is a scalar coefficient function and K(x) is a symmetric tensor function, preconditioned by the Laplace operator. Subsequently, the focus is on the eigenvalues of the matrices that arise from the discretization using conforming finite elements. Assuming continuity of K(x), it is proved that the spectrum of the preconditi- oned infinite dimensional operator is equal to the convex hull of the ranges of the diagonal function entries of Λ(x) from the spectral decomposition K(x) =...
|
|
|
Conditions for convergence of the restarted and augmented GMRES method
Nádhera, David ; Zítko, Jan (advisor) ; Strakoš, Zdeněk (referee)
The GMRES method is one of the most useful methods for solving a system of linear algebraic equations with nonsymmetric matrix. So on, many bounds for the residual norm have been derived, that can give us information about the convergence or possible stagnation of the method. A generalization of the GMRES method is the augmented GMRES method. In this paper we will analyze the implementation of augmented GMRES method proposed by Morgan. In these consequences we will be interested in how precise harmonic Ritz vectors approximate the eigenvectors belonging to the smallest in magnitude eigenvalues. We generalize some previous results concerning the convergence of restarted GMRES method for the case of augmented GMRES method. This is the rst contribution of the work. Another main point will be numerical testing and comparing of the bounds for restarted and augmented GMRES and an attempt to state a criterion, when it is suitable to stop the improvement of augmenting vectors, i. e. apply the augmented GMRES method without additional computations.
|
|
|
Krylov Subspace Methods - Analysis and Application
Gergelits, Tomáš ; Strakoš, Zdeněk (advisor) ; Farrell, Patrick (referee) ; Herzog, Roland (referee)
Title: Krylov Subspace Methods - Analysis and Application Author: Tomáš Gergelits Department: Department of Numerical Mathematics Supervisor: prof. Ing. Zdeněk Strakoš, DrSc., Department of Numerical Mathematics Abstract: Convergence behavior of Krylov subspace methods is often studied for linear algebraic systems with symmetric positive definite matrices in terms of the condition number of the system matrix. As recalled in the first part of this thesis, their actual convergence behavior (that can be in practice also substantially affected by rounding errors) is however determined by the whole spectrum of the system matrix, and by the projections of the initial residual to the associated invariant subspaces. The core part of this thesis investigates the spectra of infinite dimensional operators −∇ · (k(x)∇) and −∇ · (K(x)∇), where k(x) is a scalar coefficient function and K(x) is a symmetric tensor function, preconditioned by the Laplace operator. Subsequently, the focus is on the eigenvalues of the matrices that arise from the discretization using conforming finite elements. Assuming continuity of K(x), it is proved that the spectrum of the preconditi- oned infinite dimensional operator is equal to the convex hull of the ranges of the diagonal function entries of Λ(x) from the spectral decomposition K(x) =...
|
|
|
Multilevel methods
Vacek, Petr ; Strakoš, Zdeněk (advisor) ; Pultarová, Ivana (referee)
The analysis of the convergence behavior of the multilevel methods is in the literature typically carried out under the assumption that the problem on the coarsest level is solved exactly. The aim of this thesis is to present a description of the multilevel methods which allows inexact solve on the coarsest level and to revisit selected results presented in literature using these weaker assumptions. In particular, we focus on the derivation of the uniform bound on the rate of convergence. Moreover, we discuss the possible dependence of the convergence behavior on the mesh size of the initial triangulation. 41
|
|
|
Multilevel incomplete factorizations
Mudroňová, Veronika ; Tůma, Miroslav (advisor) ; Strakoš, Zdeněk (referee)
In this work we present incomplete matrix decomposition algorithms including their multi-level extensions. First, we have mentioned classical standard strategies that are used for the solution of linear systems. Second, we present incomplete methods based on the hierarchical approach known as multilevel matrix decom- positions. Just this latter type of approaches is considered for numerical expe- riments. Namely, two major hierarchical algorithms, the greedy search with the strongest coupling and the greedy search with the smallest degree, are compared, using the Fortran programming language. The results presented in the Matlab figures are commented and discussed. 1
|
|
|
Approaches to analysis of Krylov subspace methods
Outrata, Michal ; Strakoš, Zdeněk (advisor) ; Knobloch, Petr (referee)
The text deals with the understanding of the convergence behaviour of the GMRES method. The first part reviews results formulated for the linear algebraic finite-dimensional problem Ax = b. The second part revisits the question on whether the algebraic GMRES behaviour can be analyzed using bounded operators on an infinite-dimensional Hilbert spaces. 1
|
guest ::
