[SCIP] Entering a node that was not created

Gerald Gamrath gamrath at zib.de
Thu Feb 27 14:21:35 CET 2020


Dear Lan,


it may happen that SCIP was not able to solve the LP at a node to 
optimality, e.g., because of numerical issues. In that case, branching 
is performed on pseudo candidates, i.e., the BRANCHEXECPS callback is 
called. Probably, you did not implement this, so that SCIP called the 
next-best branching rule which then created two child nodes.


Best,
Gerald



Am 27.02.20 um 11:53 schrieb Ngô Lan:
>
> Hi,
>
>
> A bit more info: I printed out the parents of that node and the node 
> type of its parent is SCIP_NODETYPE_JUNCTION (fork without LP solution) :
>
> curr_node: 5085
>
> parent1: 5082, type 6
> parent2 5065, type 6
> parent3 5061, type 8
>
> The latest node that was entered before 5085 is 5065. For node 5065, I 
> think scip already found an LP solution for it but somehow its type is 
> still 6. And the branching tree jumped over 5082 to 5085 (both was not 
> created by me). I would be grateful if someone can explain to me the 
> behavior. Thank you in advance.
>
> Best,
> Lan
>
> ------------------------------------------------------------------------
> *From:* Scip <scip-bounces at zib.de> on behalf of Ngô Lan <lan.ngo at aalto.fi>
> *Sent:* 25 February 2020 16:32:28
> *To:* Scip at zib.de
> *Subject:* [SCIP] Entering a node that was not created
>
> Hi all,
>
>
> During my branch and price implementation, I got a strange error: SCIP 
> tried to enter a node that was not created before. I checked again all 
> the time I used SCIPcreateChild and printed out the node number but 
> that node was never created. Also, my program is currently only at 
> depth 4 but that node is at depth 9. Do you have any idea what might 
> go wrong?
>
>
> Best,
> Lan
>
>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> https://listserv.zib.de/mailman/listinfo/scip
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.zib.de/pipermail/scip/attachments/20200227/9c47ca0c/attachment.html>


More information about the Scip mailing list