[Scip] scaling

[Tobias Achterberg]

Mahr Tamas wrote:
> Thank you for looking into this!
> 
> I wonder if the problem is really at soplex. I have tried to solve the 
> same problem with the soplex shell command, and it can solve it very 
> quickly. (I guess it solves the lp relaxation of the mip, but I am 
> guessing here. In fact I have no clue how soplex interprets a mip.)
> 
> Looking into the soplex code reveals that their cli program uses the 
> MySoPlex class that is derived from the SoPlex class. The scip interface 
> class for soplex (SPxSCIP), however, is derived from a different line, 
> namely the SPxSolver class. This class is a sibling of the SoPlex class. 
> Unfortunately, the SPxSolver class do not use the presolver or scaler 
> classes that are used in the SoPlex class. That is why the command line 
> program can solve this problem, while the scip solver cannot.
> 
> I think either the SPxSCIP class should be derived from the SoPlex class 
> and then it could make use of the scaling/presolving classes in there, 
> or the SPxSCIP class could at least mimic the SoPlex class by applying 
> the presolving/scaling classes in a similat manner. But maybe I am 
> totally wrong here, and scaling/presolving should happen already in scip 
> at the beginning of the branch-and-bound search and not in soplex every 
> time an lp relaxation is solved.

No, you are completely right. The SoPlex class is basically a wrapper around the core 
SPxSolver class to provide presolving and scaling.

However, this is not as easy as it seems. I don't know whether SoPlex already supports all 
the calls that are needed for SCIP, for example to access the simplex tableau rows (and 
automatically unscale it).


Tobias