[SCIP] Asking about nonlinear objective function

[Adhe Kania]

Hi,

Thanks for your reply, Ambros.
I opened an issue last week, but not yet got an answer.
I have this function as part of my constraint :
[image: image]
<https://user-images.githubusercontent.com/51048244/94549355-bf04b100-025a-11eb-8f87-9a994bff3f9a.png>

where D(j) is input data and P has a decision variable on it.

I wrote my constraint as follows :

def consFR(model, q, SS, SOT):

for t in Ts:
    Dr = quicksum(d for d in demand[t:min(t+L,T)])
    Dr = Dr + quicksum(d for d in
demand_day[min((t+L),T)*5:min((t+L)*5+SOT,T*5)])


and I got an error :

in consFR(model, q, SS, SOT)
3 for t in Ts:
4 Dr = quicksum(d for d in demand[t:min(t+L,T)])
----> 5 Dr = Dr + quicksum(d for d in demand_day[min((t+L),T)*5:min((t+L)
*5+SOT,T*5)])
6

src\pyscipopt\expr.pxi in pyscipopt.scip.Expr.richcmp()

src\pyscipopt\expr.pxi in pyscipopt.scip._expr_richcmp()

NotImplementedError:


The problem comes because I have a variable in the upper bound of quicksum
function. If I change the variable (SOT) with a number, it's run well.

Do you have any idea to solve my problem?

I can switch to SCIP if this problem is only for pyscipopt but run well in
SCIP.


Best,

Adhe



On Tue, 29 Sep 2020 at 13:18, Ambros Gleixner <gleixner at zib.de> wrote:

> Dear Adhe,
>
> For PySCIPOpt-related questions, could you please open an issue in the
> PySCIPOpt repository and describe in a minimal (not) working example,
> what fails?
>
> Thanks,
> Ambros
>
>
>
> Am 29.09.20 um 11:15 schrieb Adhe Kania:
> > Hi,
> >
> > I am a new user in SCIP.
> > I participated in the interesting online summer school, CO at Work 2020
> > last week and I am interested in using SCIP to solve my problem.
> > I use python, so I use pyscipopt.
> > My problem has a complicated nonlinear objective function with integer
> > variables.
> > I work with multiobjective optimization problem, so the objective
> > function is an achievement scalarazing function with nonlinear objective
> > functions.
> > I asked in the Q&A session that SCIP can accommodate a nonlinear
> > objective function, but when I try with the simple nonlinear objective
> > function, the error "Nonlinear objective functions are not supported"
> shown.
> > I found from the list issues of pyscipopt that I should linearize it,
> > using the so-called epigraphical representation, which set the nonlinear
> > objective as a constraint. However, it doesn't work when I use norm.cdf
> > function.
> >
> > Would you like to give me suggestion how to use my own complicated
> > function in SCIP?
> >
> > Thanks in advance,
> > Adhe
> >
> >
> > _______________________________________________
> > 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
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.zib.de/pipermail/scip/attachments/20201005/fc498f61/attachment.html>