[SCIP] Adding Rows using Constraint Handler in SCIP

[Ambros Gleixner]

Dear Naga,

Are the variables really continuous?  If yes, and you cannot round the 
right-hand side down to 1, then you could try decreasing the parameters

separating/minefficacy
separating/minefficacyroot
separating/minortho
separating/minorthoroot

Of course, the forcecut flag that Marc mentioned is also worth a try, 
but it has the performance disadvantage that you disable the cut selection.

Best,
Ambros


Am 03.07.20 um 17:12 schrieb Marc Pfetsch:
> 
> 
> Dear Naga,
> 
> adding rows does not mean that they are added to the LP. In between
> there is a filtering process, which might remove some rows depending on
> their violation and norm of the normal vector. Without seeing the
> details, I think that this seems to be a numerical issue: The constraint
> handler adds the row, because it thinks the row is violated, but with
> respect to the filtering, the row is not violated. Note that this
> depends on the current LP solution. You can force a row to be added to
> the LP, but I would suspect that your problem will not go away, because
> of the above difference.
> 
> Best
> 
> Marc
> 
> 
> On 03/07/2020 14:12, Naga Venkata Chaitanya Gudapati -
> nagavenkata.gudapati at studio.unibo.it wrote:
>> Hello,
>>
>> I had implemented a simple branch-and-cut solver in SCIP to solve a min-max vehicle routing problem by using the TSP example in SCIP as a starting point.
>>
>> I am able to add rows using the SCIPcreateEmptyRowConshdlr and SCIPaddRow methods as described in the example. It is adding the correct rows but from what I can observe,  the constraint handler is adding the same rows again and again at various points.
>>
>> For example I add this row: -1e+20 <= +1<t_y_1,2> +1<t_y_2,1> <= 1.999 more than 30 times. My understanding is that once I add a row, the subsequent solutions should violate this  added row. Could someone let me know if I am missing something?
>>
>>
>> Regards,
>> Naga
>> _______________________________________________
>> Scip mailing list
>> Scip at zib.de
>> https://listserv.zib.de/mailman/listinfo/scip
>>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> https://listserv.zib.de/mailman/listinfo/scip
> 

-- 
Ambros Gleixner, Research Group Mathematical Optimization Methods at 
Zuse Institute Berlin, http://www.zib.de/gleixner