[SCIP] Post, non-convex QCQP

[Benjamin Müller]

Dear Ning,

nonconvex QCQPs are in general very challenging, so even relatively 
small problems can take a long time to solve.

For global solvers like SCIP, it is crucial to have variable bounds that 
are as tight as possible. Probably, a first thing to check is whether 
the variables are (at least) bounded in your model.

I am no expert in using the OPTI toolbox, but is it possible to write 
your problem into an LP (or CIP) file and send it to me? Then I could 
have a look and we might think about suitable settings for your QCQPs.

Best,
Benjamin


On 3/15/19 6:13 PM, Ning Wang wrote:
> Hi
> 
> I am a PHD student from university of Alberta. I am writing to you for 
> some technical support.
> 
> Recently, I am solving a quadratically constrained quadratic program 
> problem based on MATLAB OPTI toolbox. The optimization problem has 
> 12 optimization variables and 30 non-convex constrains. The available 
> SCIP  MATLAB  code ("scip.mexw64, v4.0.0") cannot reach a satisfactory 
> performance, which displayed ''solving was interrupted [time limit 
> reached]'' even if I set the maximum time as up to 8 hours.
> 
> Is it because the defaulted parameters of SCIP are not improperly set? 
> Then how can I improve the settings? Or is it because SCIP has limited 
> capabilities in solving non-convex QCQP problem? Is there more efficient 
> code ( better MATLAB ) to my problem? I will be very grateful for your 
> help！
> 
> Thank you for your time!
> 
> Best regards
> Ning
> 
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> https://listserv.zib.de/mailman/listinfo/scip
> 

-- 
______________________________
Benjamin Müller
Zuse Institute Berlin
Takustr. 7, 14195 Berlin
benjamin.mueller at zib.de
+49 30 841 85-195