[vtkusers] Jacobi eigenvectors incorrect?

[David Doria]

On Fri, Oct 30, 2009 at 9:05 AM, Bryn Lloyd <blloyd at vision.ee.ethz.ch> wrote:
> Hi David,
>
> I get the same result, using my test code.
>
> Matlab:
> a = [1,2,3; 2,5,9; 3,9,8]
> [v,w] = eig(a)
>
> v =
>    0.1372    0.9644    0.2259
>    0.7382   -0.2516    0.6259
>   -0.6605   -0.0809    0.7464
>
>
> w =
>   -2.6814         0         0
>         0    0.2266         0
>         0         0   16.4547
>
>
> C++:
> 16.4547         0.225898 -0.964436 0.137232
> 0.226646        0.625927 0.251645 0.738167
> -2.68139        0.746449 0.0808535 -0.660513
>
> Both are the same, except the sign difference for (ordered) Eigenvector 2.
> Note that Matlab does not order them.
>
> /Bryn


Bryn,

I also get the correct result with your code, not sure what I was
doing wrong, but I'll just use your way. Bill/Bryn - can this code
that Bryn has written be turned into a vtkMatrix class? I think it is
insane for users to have to write this:

template<class TReal>
TReal **create_matrix(long nrow, long ncol) {
  typedef TReal* TRealPointer;
  TReal **m = new TRealPointer[nrow];

  TReal* block = (TReal*)calloc(nrow*ncol, sizeof(TReal));
  m[0] = block;
  for ( int row = 1; row < nrow; ++row ) {
    m[ row ] = &block[ row * ncol ];
  }
  return m;
}

in order to get eigen vectors of a matrix.

Thanks,

David