On Ways To Improve Convergence Of Krylov Subspace Methods

Al-Kharusi, Salim Abdullah Surur (2008) On Ways To Improve Convergence Of Krylov Subspace Methods. Masters thesis, Universiti Sains Malaysia.

[img]
Preview
PDF
Download (638Kb) | Preview

    Abstract

    "SNAP" atau "Solution by Null-space Approximation and Projection" ialah salah satu cara untuk menyelesaikan sistem linear apabila pekali matrik adalah besar dan "sparse". Objektifuya ialah Ulltuk mengatasi masalah penumpuan yang perlahan atau genangan yang disebabkan oleh nilai eigen yang kecil. Dissertasi ini bertujuan untuk menyediakan satu sorotan kritikal tentang kaedah "SNAP" yang dicadangkan pada 2006 oleh M. Illic, W. Turner dan Y. Saad. Dissertasi ini memfokuskan pada idea utama "SNAP" : algoritmanya, pembinaan Penghampiran Ruang nol hampiran dan dua algoritma yang dihasilkan oleh Illic. W. Turner dan Y. Saad iaitu SNAP-JD(m) dan 'Restarted SNAP-JD(m,kmax,l) "SNAP" or "Solution by Null-space Approximation and Projection" is one of the methods for solving linear system when the matrix coefficient is large and sparse. Its objective is to overcome the problem of slow convergence or stagnation which is caused by small eigenvalues. This dissertation is aimed at providing a critical review of the SNAP method which was proposed in 2006 by M. Illic, W. Turner and Y. Saad. The dissertation focused on the main idea of SNAP: the algorithm, the construction of Approximate Null Space and two algorithms generated by Illic, W. Turner and Y. Saad which are SNAP-JD(m) and Restarted SNAP-JD(m,kmax,l ).

    Item Type: Thesis (Masters)
    Subjects: Q Science > QA Mathematics > QA1 Mathematics (General)
    Divisions: Pusat Pengajian Sains Matematik (School of Mathematical Sciences)
    Depositing User: HJ Hazwani Jamaluddin
    Date Deposited: 18 Nov 2016 15:58
    Last Modified: 18 Nov 2016 15:58
    URI: http://eprints.usm.my/id/eprint/31155

    Actions (login required)

    View Item
    Share