[SCIP] My pricer is not being called :-(

[Diego Delle Donne]

Sorry! My bad!

I got a stupid bug in my BranchingEventHandler code which was bounding to 0
some of the variables (thus making some nodes infeasible). Now it's working
as it should.

Anyway, I don't understand why SCIP was calling my
BranchingEventHandler::scip_exec method before calling my pricing
procedure. Can this happen? I'd think that the pricing step should finish
before scip make some branching step... am I wrong on this?

Sorry for the troubles!
Diego


El dom., 7 abr. 2019 a las 16:43, Diego Delle Donne (<diegodd at gmail.com>)
escribió:

> Dear SCIP community,
>
> I'm coding a Branch & Price. As a first approach, I was starting with just
> one dummy binary variable representing a feasible solution (with a very bad
> obj function cost) and everything worked fine. SCIP starts with this
> variable as a feasible solution (primal bound), then call my pricing and at
> some point (when my pricing says so), it closes the LP relaxation and
> starts branching (if necessary).
>
> Now, I want to start with a set of binary variables (obtained by an
> initial heuristic approach) representing a better initial solution than
> this dummy variable. So, I added these variables instead of the dummy
> variable. The nice thing is that SCIP is able to get the initial feasible
> solution / primal bound (by putting all these variables in 1). The bad
> thing is that it is immediately closing the root node without calling my
> pricer (somehow it uses this solution as the dual bound, which is not
> right). It ends up saying that the problem is solved to optimality with my
> initial heuristic solution (and I know for sure that this solution is not
> optimal in my examples).
>
> Any ideas on what am I doing wrong? :-(
>
> Thanks in advance!
> Diego
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.zib.de/pipermail/scip/attachments/20190407/ef517128/attachment.html>