Re: [Bug-gsl] LU_inverse bug

[Brian Gough]

Georges Szegezdi writes:
 > Hi, I'm using the GSL library to do the pseudo-inverse calcul of a matrix 
 > (and 
 > then do some other stuff on it like transpose etc..).
 > I have a problem with some type of matrices, I reproduce the example above.
 > When inverting a matrix, the result sometimes comes out to NaN or Inf.

Hello,

This is the normal behavior for a singular matrix.  It comes from the
underlying blas function dtrsv which we don't have any control over,
and doesn't return an error code on division by zero.  For potentially
singular matrices, you can use the SVD to get more control over the
computation.

-- 
best regards

Brian Gough