[Ipopt] scaling

Stefan Vigerske stefan at math.hu-berlin.de
Mon Jun 21 13:52:34 EDT 2010


you could set the option obj_scaling_factor to -1.0 to influence this
scaling of the objective function only:

Then there should be no need to change the nlp_scaling_method.

Other way is to multiply your objective by -1.0 by hand in the
evaluation methods...


Drosos Kourounis schrieb:
> Dear Andreas,
> convergence is very sensitive to the scaling one choses. I am using user-scaling to suggest to IPOPT to optimize instead of minimize, since there is not an option whether you would like to minimize or maximize. This is the only reason I have to set user-scaling as my scaling method. But then, I am not able to use the options provided by gradient-scaling to force the norm of the objective and constraint gradients to 1. Generally I would like to be able to do something which SNOPT allows and it is very flexible. You say that you want to maximize instead of the default minimize without having to mess up with scalings. And secondly you can say to SNOPT to scale automatically anything it thinks that should be scaled. This is very convenient from side of user. Are there such options for IPOPT or anything similar?
> Best,
> Drosos.
> _______________________________________________
> Ipopt mailing list
> Ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/ipopt

Stefan Vigerske
Humboldt University Berlin, Numerical Mathematics

More information about the Ipopt mailing list