[Octave-bug-tracker] [bug #56232] Octave crash when inverting an empty sparse matrix.

[Rik]

Follow-up Comment #49, bug #56232 (project octave):

I agree with Marco.  Although the inverse could, in general, be a fully dense
matrix.  It is only worth paying the computation time, and the memory cost, if
you are deriving information and getting a usable result.  If the result is
meaningless then we might as well pick a return value that we like.

I tried some sample singular matrices, including the empty matrix, in Matlab
and I don't see a pattern to the results.  The matrices themselves are fully
populated with entries consisting only of -Inf, Inf, NaN.  So no real guidance
there.

Two philosophical questions:

1) How high is the value of preserving the sparsity pattern in the output?

2) Do we want to poison downstream calculations, so that even if a user missed
the warning about a singular matrix, the end result is clearly incorrect? 
That would favor using NaN rather than Inf as a fill value.



    _______________________________________________________

Reply to this item at:

  <https://savannah.gnu.org/bugs/?56232>

_______________________________________________
  Message sent via Savannah
  https://savannah.gnu.org/