Package no.uib.cipr.matrix

Class BandCholesky

java.lang.Object
no.uib.cipr.matrix.BandCholesky
public class BandCholesky extends Object
Banded Cholesky decomposition

  • Constructor Details

    • BandCholesky

      public BandCholesky(int n, int kd, boolean upper)
      Constructor for BandCholesky
      Parameters:
      n - Matrix size
      kd - Number of matrix bands
      upper - True for decomposing an upper symmetrical matrix, false for a lower symmetrical matrix

  • Method Details

    • factorize

      public static BandCholesky factorize(LowerSPDBandMatrix A)
      Creates a Cholesky decomposition of the given matrix
      Parameters:
      A - Matrix to decompose. Not modified
      Returns:
      A Cholesky decomposition of the matrix

    • factorize

      public static BandCholesky factorize(UpperSPDBandMatrix A)
      Creates a Cholesky decomposition of the given matrix
      Parameters:
      A - Matrix to decompose. Not modified
      Returns:
      A Cholesky decomposition of the matrix

    • factor

      public BandCholesky factor(LowerSPDBandMatrix A)
      Creates a Cholesky decomposition of the given matrix
      Parameters:
      A - Matrix to decompose. Overwritten on return
      Returns:
      The current decomposition

    • factor

      public BandCholesky factor(UpperSPDBandMatrix A)
      Creates a Cholesky decomposition of the given matrix
      Parameters:
      A - Matrix to decompose. Overwritten on return
      Returns:
      The current decomposition

    • getL

      public LowerTriangBandMatrix getL()
      Returns the decomposition matrix. Only valid for decomposition of a lower SPD matrix

    • getU

      public UpperTriangBandMatrix getU()
      Returns the decomposition matrix. Only valid for decomposition of a upper SPD matrix

    • isSPD

      public boolean isSPD()
      Returns true if the matrix decomposed is symmetrical, positive definite

    • rcond

      public double rcond(Matrix A)
      Computes the reciprocal condition number
      Parameters:
      A - The matrix this is a decomposition of
      Returns:
      The reciprocal condition number. Values close to unity indicate a well-conditioned system, while numbers close to zero do not.

    • solve

      public DenseMatrix solve(DenseMatrix B) throws MatrixNotSPDException
      Computes A\B, overwriting B
      Throws:
      MatrixNotSPDException