optimal solution is found while the constraint is not satisfied

[mahraz amini]

Hi all

I am trying to solve this simple optimization question including gel functions

function gsl_cdf_ugaussian_Qinv;

var x1 >= 0 ;

minimize obj: (x1^2)-8*x1;

subject to const1: x1-4 <= gsl_cdf_ugaussian_Qinv(1- 0.05);

Then I tried to solve this in AMPL

ampl: model test.mod;

ampl: load amplgsl.dll;

ampl: option solver ipopt;

ampl: solve;

EXIT: Optimal Solution Found.

 

Ipopt 3.12.4: Optimal Solution Found

ampl: display x1;

x1 = 3.99996

subject to const1:

-gsl_cdf_ugaussian_Qinv(0.95) + x1 <= 4;

ampl: display -gsl_cdf_ugaussian_Qinv(0.95) + x1;

-gsl_cdf_ugaussian_Qinv(0.95) + x1 = 5.64481

As it is obvious, the constraint is not satisfied! I wonder why and how I can fix this.

Thanks,

Mahraz

[AMPL Google Group]

Load the extended function library before reading the model that uses the gaussian function:

ampl: load amplgsl.dll;
ampl: model test.mod;

Arguably the other order should also work, or else should cause an error message to appear; this issue has been reported.

--
Robert Fourer
am...@googlegroups.com

{#HS:559356179-4986#}

--
You received this message because you are subscribed to the Google Groups "AMPL Modeling Language" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ampl+uns...@googlegroups.com.
To post to this group, send email to am...@googlegroups.com.
Visit this group at https://groups.google.com/group/ampl.
For more options, visit https://groups.google.com/d/optout.

[AMPL Google Group]

This issue has been addressed in a new update to the AMPL program. Now the model can be read before amplgsl.dll is loaded, and the results will still be correct.

--
Robert Fourer
am...@googlegroups.com

{#HS:559356179-4986#}

On Thu, Apr 12, 2018 at 12:11 AM UTC, AMPL Google Group <am...@googlegroups.com> wrote:

Load the extended function library before reading the model that uses the gaussian function:

ampl: load amplgsl.dll;
ampl: model test.mod;

Arguably the other order should also work, or else should cause an error message to appear; this issue has been reported.

--
Robert Fourer
am...@googlegroups.com

--