in

.NET Opensource

Community for opensource projects by Christoph Rüegg

Possible bug in CholeskyDecompostion solve

Last post 01-09-2008 20:20 by Christoph Rüegg. 3 replies.
Page 1 of 1 (4 items)
Sort Posts: Previous Next
  • 06-16-2007 17:11

    Possible bug in CholeskyDecompostion solve

     I think there is a bug is solve method of CholeskyDecomposition. (the GetL() method works fine, though)

    I tried:

    double[,] a = { { 25, -5, 10 }, { -5, 17, 10 }, { 10, 10, 62 } };
    Matrix m = new Matrix(a);
    CholeskyDecomposition cd = new CholeskyDecomposition(m);
    Matrix i = Matrix.Identity(3, 3);

    Matrix inv1 = cd.Solve(i);

    Matrix inv2 = m.Inverse();

     

    The inv1 matrix is not the same as inv2, even though they should be. 

    Filed under: , ,
  • 10-29-2007 1:54 In reply to

    Re: Possible bug in CholeskyDecompostion solve

    Filed under: ,
  • 01-08-2008 22:11 In reply to

    Re: Possible bug in CholeskyDecompostion solve

    FYI: I've just updated the linear algebra unit tests; they now repro the cholesky issue and fail (LinearAlgebraTests.AllTests and BugRegression.IRID90_CholeskySolve).

    Btw, the bug is tracked as IRID-90 (not 61).

    Filed under: , ,
  • 01-09-2008 20:20 In reply to

    Re: Possible bug in CholeskyDecompostion solve

    Answer

    This bug has been fixed today in revision 296.

    Filed under: , ,
Page 1 of 1 (4 items)
Powered by Community Server (Non-Commercial Edition), by Telligent Systems