[Ipopt] Convergence of Ipopt to an acceptable value of the objective if the minimum appears to be in a "flat" region

[Lemoine, Roberto]

Dear members of the Ipopt community,

I am using Ipopt for a problem which seems to have its optimum in a relatively flat region (even after scaling). Im an interested in obtaining a point in which the objective function has been reduced "enough" (say, two orders of magnitude; this is acceptable and doable, as at the optimum the objective function is decreased actually by three orders of magnitude). However, particularly in the last iterations (say, 10 out of 30 iterations), the objective function is even below of this desired level and the primal infeasibility is already fine, but the dual infeasibility keeps decreasing very slowly (after wandering a bit).

At which Ipopt options should one look at to achieve the acceptable decrease of the objective function, but with a faster reduction of the infeasibility (particularly the dual infeasibility)?

Thanks a lot in advance.


Best regards,

Roberto Lemoine
Lehrstuhl für Systemdynamik und Prozessführung
Technical University Dortmund
+49 231 755 5809
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/ipopt/attachments/20150327/58896fdb/attachment.html>