[Insight-developers] Problems with the vnl library

[Bill Hoffman]

--=====================_511189071==_.ALT
Content-Type: text/plain; charset="us-ascii"

At 02:33 PM 12/14/00 -0500, Paul Hughett wrote:


>I've encountered several problems using the vnl library, and wanted
>to make them known so other people can comment, and perhaps provide
>solutions.
>
>1. vnl_matrix_inverse does not check for singular matrices.  As far as
>I can tell, this function will not provide any error signal or exception
>when given a singular matrix.  In fact, trying the invert the singular
>matrix [1 2; 3 6] yields the "inverse" [ 2.78842e+15 -9.29473e+14 ; 
>-1.39421e+15 4.64737e+14 ] without any indication of a problem.


If you are interested in errors, you should use the vnl_svd<double> directly.   
I will pass your other questions on to the vxl user list.


--=====================_511189071==_.ALT
Content-Type: text/html; charset="us-ascii"

At 02:33 PM 12/14/00 -0500, Paul Hughett wrote:



> I've encountered several problems using the
> vnl library, and wanted
> to make them known so other people can comment, and perhaps provide
> solutions.
>
> 1. vnl_matrix_inverse does not check for singular matrices.  As far
> as
> I can tell, this function will not provide any error signal or
> exception
> when given a singular matrix.  In fact, trying the invert the
> singular
> matrix [1 2; 3 6] yields the "inverse" [ 2.78842e+15
> -9.29473e+14 ; 
> -1.39421e+15 4.64737e+14 ] without any indication of a problem.



If you are interested in errors, you should use the vnl_svd<double>
directly.   

I will pass your other questions on to the vxl user list.



--=====================_511189071==_.ALT--