[SCIP] Making sure that the same branching decision is not done twice

[Benjamin Müller]

Dear Myroslav,

as far as I understand your question, you don't want to branch on 
variables that are currently present in your formulation, right? 
Otherwise, a branching rule that cuts off the current LP solution in 
both branches should never generate the same branching again on a path 
to the root node.

Please have a look at the Binpacking example. There you find the 
samediff constraint handler that is used to store data in its 
constraints in order to update properly the pricing problem in a local 
node of the branch-and-bound tree. However, I am wondering about

     "I don't use the branching decisions for pricing"

because if you don't respect the branching decisions in the pricing 
problem, then I don't see how you can generate new variables during the 
tree search. Maybe you can clarify this a bit.

Best,
Benny

On 08/29/2018 03:48 PM, myroslav wrote:
> Dear SCIP community,
>
>
> I wondered how I could store branching decisions in the nodes of the 
> branching tree.  I don't  use the branching decisions for pricing, I 
> just want to make sure in the branching callback BRANCHEXECLP that I 
> don't do the same branching decision that was done earlier on the path 
> to the root, therefore, I was not sure if it's also a good idea to use 
> "marker constraints" as suggested in one of the FAQs.
>
>
> Sincerely,
>
> Myroslav
>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> https://listserv.zib.de/mailman/listinfo/scip