On the acceleration of Kaczmarz's method for inconsistent linear systems (Q920573): Difference between revisions

From MaRDI portal
Import240304020342 (talk | contribs)
Set profile property.
ReferenceBot (talk | contribs)
Changed an Item
 
Property / cites work
 
Property / cites work: Q4061081 / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q3943082 / rank
 
Normal rank
Property / cites work
 
Property / cites work: Accelerated projection methods for computing pseudoinverse solutions of systems of linear equations / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q3948525 / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q3207128 / rank
 
Normal rank
Property / cites work
 
Property / cites work: Row-Action Methods for Huge and Sparse Systems and Their Applications / rank
 
Normal rank
Property / cites work
 
Property / cites work: Strong underrelaxation in Kaczmarz's method for inconsistent systems / rank
 
Normal rank
Property / cites work
 
Property / cites work: Iterative algorithms for large partitioned linear systems, with applications to image reconstruction / rank
 
Normal rank
Property / cites work
 
Property / cites work: On the solution of singular linear systems of algebraic equations by semiiterative methods / rank
 
Normal rank
Property / cites work
 
Property / cites work: Convergence properties of ART and SOR algorithms / rank
 
Normal rank
Property / cites work
 
Property / cites work: The angles between the null spaces of X rays / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q3203925 / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q3325518 / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q5768822 / rank
 
Normal rank
Property / cites work
 
Property / cites work: The Tchebychev iteration for nonsymmetric linear systems / rank
 
Normal rank
Property / cites work
 
Property / cites work: Adaptive procedure for estimating parameters for the nonsymmetric Tchebychev iteration / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q3334035 / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q3754464 / rank
 
Normal rank
Property / cites work
 
Property / cites work: Generalizations of the projection method with applications to SOR theory for Hermitian positive semidefinite linear systems / rank
 
Normal rank
Property / cites work
 
Property / cites work: Projection method for solving a singular system of linear equations and its applications / rank
 
Normal rank
Property / cites work
 
Property / cites work: Q5604166 / rank
 
Normal rank

Latest revision as of 11:14, 21 June 2024

scientific article
Language Label Description Also known as
English
On the acceleration of Kaczmarz's method for inconsistent linear systems
scientific article

    Statements

    On the acceleration of Kaczmarz's method for inconsistent linear systems (English)
    0 references
    0 references
    0 references
    1990
    0 references
    Es sei \(A\) eine rechteckige Matrix. Durch \(\| y\|_ w=(\sum^{m}_{i=1}w^ 2_ i| y_ i|^ 2)^{1/2}\) wird eine gewichtete Euklidische Norm definiert. Gesucht ist nun die gewichtete kleinste-Quadrate-Lösung \(x^+\) von \(Ax=b\) mit minimaler Euklidischer Norm. Dazu wird zunächst eine Verallgemeinerung des Iterationsverfahrens von Kaczmarz mit Iterationsmatrix \(T_{\omega}\) vorgeschlagen, das von einem Parameter \(\omega\) abhängt. Für alle \(\omega\) mit \(0<\omega <{\bar \omega}\) und z.B. Startvektor \(x_0=0\) konvergiert das Verfahren gegen einen Vektor \(x(\omega)=A^-b\), wobei \(A^-\) eine gewisse verallgemeinerte Inverse von \(A\) ist. Für \(\omega\to 0\) wird \(x(\omega)\to x^+\) gezeigt. Auf Grund einer theoretischen Aussage und numerischer Rechnungen kann man davon ausgehen, daß das Spektrum von \(T_{\omega}\) für kleine \(\omega >0\) in der komplexen Ebene sehr dicht am Intervall \([0,1]\) liegt. Dies legt es nahe, obiges Verfahren mit dem Tschebyscheffschen semiiterativen Verfahren (für nichtsymmetrische Iterationsmatrizen) zu beschleunigen. Eine Aussage über den Konvergenzfaktor des neuen Verfahrens sowie ein Vergleich mit einer variierten Beschleunigungsmethode wird gemacht.
    0 references
    0 references
    Kaczmarz's method
    0 references
    inconsistent linear systems
    0 references
    large sparse matrix
    0 references
    iterative method
    0 references
    relaxation parameters
    0 references
    strong underrelaxation
    0 references
    least- squares solution
    0 references
    convergence
    0 references
    Chebyshev acceleration
    0 references
    0 references