Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd Logo
Upload

Welcome to Scribd!

  • 33 views

    Introduction To Numerical Analysis II: Solving Linear System

    Uploaded by

    Patrick Sibanda
    The document discusses methods for solving linear systems of the form Ax=b, including direct methods like Gaussian elimination and LU decomposition, and iterative methods like Jacobi, Gauss-Seidel, and successive over-relaxation (SOR). It provides the general forms of linear iterative methods, conditions for convergence, and examples of applying the methods to different types of matrices.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Introduction To Numerical Analysis II: Solving Linear System

    Uploaded by

    Patrick Sibanda
    33 views28 pages
    The document discusses methods for solving linear systems of the form Ax=b, including direct methods like Gaussian elimination and LU decomposition, and iterative methods like Jacobi, Gauss-Seidel, and successive over-relaxation (SOR). It provides the general forms of linear iterative methods, conditions for convergence, and examples of applying the methods to different types of matrices.

    Original Description:

    iterations

    Original Title

    Iter

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document discusses methods for solving linear systems of the form Ax=b, including direct methods like Gaussian elimination and LU decomposition, and iterative methods like Jacobi, Gauss-Seidel, and successive over-relaxation (SOR). It provides the general forms of linear iterative methods, conditions for convergence, and examples of applying the methods to different types of matrices.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    33 views28 pages

    Introduction To Numerical Analysis II: Solving Linear System

    Uploaded by

    Patrick Sibanda
    The document discusses methods for solving linear systems of the form Ax=b, including direct methods like Gaussian elimination and LU decomposition, and iterative methods like Jacobi, Gauss-Seidel, and successive over-relaxation (SOR). It provides the general forms of linear iterative methods, conditions for convergence, and examples of applying the methods to different types of matrices.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 28

    Introduction to Numerical Analysis II:

    Solving Linear System

    Xiaozhe Hu

    The Pennsylvania State University

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 1 / 10


    Solving Linear System

    Linear system:
    Ax = b

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 2 / 10


    Solving Linear System

    Linear system:
    Ax = b

    Direct Method
    I Gaussian Elimination
    I PA = LU
    Iterative Method
    I Linear Iterative Method
    I Krylov Subspace Method

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 2 / 10


    Basic Linear Iterative Method

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 3 / 10


    Basic Linear Iterative Method

    General Form of Linear Iterative Method


    Form Residual:
    r = b Ax k
    Solve Residual Equation Approximately:

    Ae = r e = Br , B A1

    Update:
    x k+1 = x k + e

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 3 / 10


    Basic Linear Iterative Method

    General Form of Linear Iterative Method


    Form Residual:
    r = b Ax k
    Solve Residual Equation Approximately:

    Ae = r e = Br , B A1

    Update:
    x k+1 = x k + e

    x k+1 = x k + B(b Ax k )

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 3 / 10


    Examples

    A=D +L+U

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 4 / 10


    Examples

    A=D +L+U

    Richardson: B = I
    Jacobi: B = D 1
    Damped Jacobi: B = D 1
    Forward Gauss-Seidel: B = (D + L)1
    Backward Gauss-Seidel: B = (D + U)1
    Symmetric Gauss-Seidel: B = (D + U)1 D(D + L1 )
    SOR: B = (D + L)1
    Symmetric SOR: B = w (1 w )(D + U)1 D(D + L)1

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 4 / 10


    Convergence

    e k+1 = (I BA)e k e k := x x k

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 5 / 10


    Convergence

    e k+1 = (I BA)e k e k := x x k

    Theorem (Necessary and sufficient condition)


    The linear iterative method converges if and only if

    (I BA) < 1.

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 5 / 10


    Convergence

    e k+1 = (I BA)e k e k := x x k

    Theorem (Necessary and sufficient condition)


    The linear iterative method converges if and only if

    (I BA) < 1.

    Theorem (A sufficient condition when A is SPD)


    If A is SPD, then the linear iterative method converges if

    B 1 + B T A

    is SPD

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 5 / 10


    Example

    2 2
    Richardson Method: 0 < < max (A) , and opt = min (A)+max (A) .

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 6 / 10


    Example

    2 2
    Richardson Method: 0 < < max (A) , and opt = min (A)+max (A) .

    Jacobi Method: 2D A is SPD.

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 6 / 10


    Example

    2 2
    Richardson Method: 0 < < max (A) , and opt = min (A)+max (A) .

    Jacobi Method: 2D A is SPD.

    Gauss-Seidel Method: Always converge when A is SPD.

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 6 / 10


    Example

    2 2
    Richardson Method: 0 < < max (A) , and opt = min (A)+max (A) .

    Jacobi Method: 2D A is SPD.

    Gauss-Seidel Method: Always converge when A is SPD.

    2
    SOR Method: 0 < < 2 and opt =
    1+ 1(I D 1 A)2

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 6 / 10


    Nonnegative Matrix

    Definition (Nonnegative Matrix)


    A matrix A is nonnegative, if aij 0. We use notation A 0.

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 7 / 10


    Nonnegative Matrix

    Definition (Nonnegative Matrix)


    A matrix A is nonnegative, if aij 0. We use notation A 0.

    Theorem (Perron-Frobenius Theory)


    Let A 0, then
    A has a nonnegative real eigenvalue equal to its spectral radius.
    To (A), there corresponds a nonzero eigenvector x 0.

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 7 / 10


    Convergence: General Case

    Definition (Regular Splitting of A)


    A=M N
    M is nonsingular, M 1 and N is nonnegative.

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 8 / 10


    Convergence: General Case

    Definition (Regular Splitting of A)


    A=M N
    M is nonsingular, M 1 and N is nonnegative.

    I BA = I M 1 A = M 1 N

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 8 / 10


    Convergence: General Case

    Definition (Regular Splitting of A)


    A=M N
    M is nonsingular, M 1 and N is nonnegative.

    I BA = I M 1 A = M 1 N

    Theorem (Necessary and sufficient condition)


    Let M and N be regular splitting of a matrix A, then (M 1 N) < 1 if and
    only if A is nonsingular and A1 is nonnegative.

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 8 / 10


    M-matrix

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 9 / 10


    M-matrix
    Definition (M-matrix)
    A Rnn , aij 0 for all i 6= j. A is a M-matrix if A is nonsingular and
    A1 0.

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 9 / 10


    M-matrix
    Definition (M-matrix)
    A Rnn , aij 0 for all i 6= j. A is a M-matrix if A is nonsingular and
    A1 0.

    Definition (irreducible)
    nn , if there exists a permutation matrix P, such that
    For A R 
    A11 A12
    P T AP = , then A is reducible. Otherwise, A is irreducible/
    0 A22

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 9 / 10


    M-matrix
    Definition (M-matrix)
    A Rnn , aij 0 for all i 6= j. A is a M-matrix if A is nonsingular and
    A1 0.

    Definition (irreducible)
    nn , if there exists a permutation matrix P, such that
    For A R 
    A11 A12
    P T AP = , then A is reducible. Otherwise, A is irreducible/
    0 A22

    Definition (Diagonally Dominant)


    Let A Rnn , if |aii |
    P
    i6=j |aij |, then A is diagonally dominant.

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 9 / 10


    M-matrix

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 10 / 10


    M-matrix
    Definition (Strictly Diagonally Dominant)
    X
    |aii | > |aij |, for all i
    i6=j

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 10 / 10


    M-matrix
    Definition (Strictly Diagonally Dominant)
    X
    |aii | > |aij |, for all i
    i6=j

    Definition (Irreducibly Diagonally Dominant)


    A is diagonally dominant and ireducible. And
    X
    |aii | > |aij |, for at least one i
    i6=j

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 10 / 10


    M-matrix
    Definition (Strictly Diagonally Dominant)
    X
    |aii | > |aij |, for all i
    i6=j

    Definition (Irreducibly Diagonally Dominant)


    A is diagonally dominant and ireducible. And
    X
    |aii | > |aij |, for at least one i
    i6=j

    Theorem
    If aii > 0 and aij 0, i 6= j. If A is SDD or IDD, then A is M-matrix.

    X. Hu (Penn State) MATH/CMPSC 456 Feb. 13, 2012 10 / 10

    You might also like