[Octave-bug-tracker] [bug #44061] U and V result matrices from svds are not othogonal

[Lachlan Andrew]

Follow-up Comment #5, bug #44061 (project octave):

Rik, yes Matlab does output orthonormal vectors:

++
U =

    0.5323   -0.4654   -0.1215    0.6966
    0.4654    0.5323   -0.6966   -0.1215
   -0.4654   -0.5323   -0.6966   -0.1215
   -0.5323    0.4654   -0.1215    0.6966

V =

   -0.9978   -0.0669    0.0000    0.0000
   -0.0669    0.9978    0.0000   -0.0000
   -0.0000    0.0000    0.0002    1.0000
   -0.0000    0.0000    1.0000   -0.0002
--

Marco, what is wrong with the value returned by eigs?  I get

++
V =

   1   0

s =  1
--

Any (non-zero) vector is an eigenvector of the identity matrix, isn't it? 
That means that the returned  V  is (the transpose of) an eigenvector
corresponding to the eigenvalue returned.

However, for  n<4,  I get a failure of

++
 [U, S, V] = svds(rand(4))
error: unrecognized value for sigma: LA
error: called from:
error:   /usr/share/octave/3.8.1/m/sparse/eigs.m at line 331, column 9
error:   /usr/share/octave/3.8.1/m/sparse/eigs.m at line 254, column 15
error:   /usr/share/octave/3.8.1/m/sparse/svds.m at line 182, column 17
--

in version 3.8.1

    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?44061>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/