[SCIP] MIP Pre-solver

Benjamin Müller benjamin.mueller at zib.de
Mon Jan 6 11:31:20 CET 2020


Dear Serge,

sorry for the late reply, your mail slipped through my attention.

1. Alternatively, you could extend the linear constraint handler and 
implement another presolving technique that tightens/changes some 
coefficients.

2. Do you want to link both variables via a linear equality constraint, 
e.g., x_j = a z_j + b? If that is the case, then you want to aggregate 
variables. For this you can use SCIPaggregateVars() or 
SCIPmultiaggregateVar() for multi-aggregations. If this is not what you 
want to do, then please provide more information on what you want to 
achieve. Thanks!

Best,
Benjamin

On 12/11/19 1:50 PM, Serge Kruk wrote:
> Hello all,
> 
> I am trying to write a pre-solver for MIP. I want to confirm the 
> following (or fix my misunderstanding):
> Consider a set simple linear constraints $\sum_i a_{ij} x_j \le b_i$
> 
> 1. Say that I want to change the coefficients $a_ij$ of a row (to scale, 
> for example).
> 
> I have not seen a simple way to just change the coefficient. Must I 
> delete the constraint and create a scaled constraint?
> 
> 2. Say I want to change the variable of a column.  Replacing $x_j$ by 
> $z_j$ and creating another constraint relating the two.
> 
> Again, I have not seen a simple way to do this except deleting all 
> constraints and re-creating all of them.
> 
> I trust that I am wrong and someone can enlighten me.
> 
> Cheers
> Serge
> 
> 
> 
> 
> 
>        Serge G. Kruk Ph.D.
>        Associate Professor kruk at oakland.edu <mailto:kruk at oakland.edu>
> Department of Mathematics and Statistics <http://www.oakland.edu/math>
> Oakland University <http://www.oakland.edu>,
>        Rochester, MI.  48309
> 
> _______________________________________________
> 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


More information about the Scip mailing list