[SCIP] SCIP detected infeasibility for feasible problem
Stefan Vigerske
stefan at math.hu-berlin.de
Wed Sep 11 17:28:00 CEST 2019
Hi,
I think these messages are from the AMPL presolver, not from SCIP.
In SCIP, numbers below 1e-9 are actually considered to be 0, often.
I don't know if you can turn off AMPL presolver (put "option presolve
0;" somewhere or run AMPL with -P) via NEOS, though.
Stefan
On 9/11/19 4:26 PM, angelika.wiegele at aau.at wrote:
> Dear SCIP team,
>
> via NEOS, I tried to solve a QP (100 integer variables, 300 quadratic
> constraints) and SCIP returned: infeasible.
>
> However, the problem *is* feasible.
> I found another way to get a feasible solution. Then I simply fixed in
> the model-file the variables to the values of this solution, and SCIP
> told me:
>
> ***SCIP-AMPL***
>
> presolve, constraint Con128:
> all variables eliminated, but upper bound = -4.54747e-12 < 0
> presolve, constraint Con158:
> all variables eliminated, but lower bound = 7.27596e-12 > 0
> presolve, constraint Con34:
> all variables eliminated, but upper bound = -1.81899e-12 < 0
> presolve, constraint Con101:
> all variables eliminated, but lower bound = 2.67164e-11 > 0
> presolve, constraint Con137:
> all variables eliminated, but lower bound = 1.22782e-11 > 0
> Infeasible constraints determined by presolve.
>
>
> The numbers (in the solution) are "nice" in a sense that many of them
> are zero and the non-zeros are integers in the range -100,..,100.
>
> Question: how can I tell SCIP that 1e-11 can be considered as 0?
>
> Best regards,
> Angelika
>
>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> https://listserv.zib.de/mailman/listinfo/scip
>
More information about the Scip
mailing list