[Octave-bug-tracker] [bug #31731] Test failed for sqp

[Marco Caliari]

Follow-up Comment #6, bug #31731 (project octave):

Yes, it exits because the stepsize has become too small. Nevertheless, with
my solution (tol = eps)


x =

  -1.71714349589853e+00
   1.59570960391797e+00
   1.82724589134608e+00
  -7.63643086466522e-01
  -7.63643086466523e-01


I get


phi(x) 
ans =  5.39498477702744e-02


and


g(x)
ans =

   5.32907051820075e-15
  -8.88178419700125e-16
  -1.77635683940025e-15


while with yours


phi(x_Rik)
ans =  5.39498477702752e-02


and


g(x_Rik)
ans =

   2.48689957516035e-14
   7.99360577730113e-15
  -7.99360577730113e-15


So, I'm satisfied with my solution. Still, there is a difference around
single precision (or sqrt(eps)) between my and your solution.

    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?31731>

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