[SCIP] wrong bounds in vbcfile?

[Marc Pfetsch]

Hi Uwe,

in your settings the parameter "visual/objextern" is set to FALSE. This
means that the output will contain the internal objective values. The
internal objective is always minimized. Thus, if you original problem
was maximized, the values are multiplied by -1. If the parameter is
TRUE, the objective is retransformed to the original sense (e.g.
maximization).

I think that indeed in your case the upper bound is not an upper bound
anymore. I need to check the file description of VBC more carefully to
determine whether this behavior is correct or not. Our implementation
just outputs "U ..." if it found a primal solution.

Best

Marc

P.S. Please be careful with large attachments that are sent to this
list, since the number of receiving people is quite large.





On 06.03.2017 15:59, Gotzes, Uwe Dr. wrote:
> Hi there,
> 
> I have a question regarding the influence of the kind of a problem (maximization or minimization) on scip's vbcfile-output.
> 
> I have a minimization problem (air04min.lp), that I transformed into an equivalent maximization problem by multiplying the objective by -1 (air04max.lp).
> 
> Depending on the settings file that I use, the vbcfile-output looks exactly the same for the max and for the min problem or the bounds in the vbcfile have the opposite sign. In the latter case the generated vbcfile is no longer correct in my opinion. For instance the lines with a "U" in the vbcfile do no longer contain upper bounds for the maximization problem.
> 
> Unfortunately, I cannot figure out which setting yield the same (and correct) bounds for both problems in the vbcfile. Can anybody help?
> 
> With the default settings (except "set vis vbc" and "set vis dis") the signs differ in the vbcfile and with the attached settings the generated vbcfiles are identical.
> 
> Best,
> Uwe
> 
> 
> 
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> http://listserv.zib.de/mailman/listinfo/scip
>