[SCIP] the (default?) variable pricer

[Ambros Gleixner]

Hi Myroslav,


Am 21.08.2018 um 14:44 schrieb myroslav:
> Dear SCIP community,
> 
> 
> I had a question regarding the variable pricer (see 
> http://scip.zib.de/doc/html/FAQ.php#variablepricer). If I understood it 
> correctly, it prices variables that have been added to scip but are not 
> in the current(some node in the branching tree) relaxation? What if I 
> know a smarter, problem specific pricing algorithm, how can I make the 
> variable pricer use this algorithm?

I am afraid that this is not possible, at least from all I know.


> Also I had a question regarding SCIPcreateVar() parameter initial. I 
> flagged my variables as not-initial(passed false when creating a 
> variable) and when I do my custom branching in SCIP_DECL_BRANCHEXECLP 
> the first time it executes, which is, I guess, the root node, I see that 
> all the columns obtained using SCIPgetNLPCols() are already in the LP. 
> But how could that happen if all the variables were flagged as not-initial?

Before the branching callback, SCIP solves the LP to optimality in 
price-and-cut loop.  It will keep adding non-initial variables to the LP 
until reaching an optimal LP solution, i.e., where all variables that 
are not in the LP have dual feasible reduced cost.  So it can easily 
happen the all non-initial variables are added.

Hope that helps,
Ambros




> 
> 
> Thanks,
> 
> Myroslav
> 
> _______________________________________________
> 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