[SCIP] Price and cut loop

[Maher, Stephen]

Hi Albert,

I am not totally sure of your problem here. I will explain what should be happening, and hopefully this will help you find the answer.

When you set the result to SCIP_CONSADDED, then this should trigger the resolving of the LP, which should lead to an additional pricing round. The LP will first be solved to update the dual variables. Since you say that the gap is completely closed after the separation round, then this means that there is not need to enter pricing. The addition of new variables will not improve the dual bound of the LP any further. Nothing you do will trigger a new pricing round.

I suggest that you check the correctness of your separator. This could be a source of the error, since this is where you first recognise the issue. Otherwise, the pricer may be terminating early, which could also lead to this behaviour.

Regards,

Steve

________________________________
From: Scip <scip-bounces at zib.de> on behalf of Schrotenboer, Albert <a.h.schrotenboer at rug.nl>
Sent: 24 July 2020 19:50
To: scip at zib.de <scip at zib.de>
Subject: [SCIP] Price and cut loop

CAUTION: This email originated from outside of the organisation. Do not click links or open attachments unless you recognise the sender and know the content is safe.

Dear SCIP Community,

At the end of a pricing loop in the rood node (coding a branch-cut-and-price algorithm) i am adding some cuts via an ObjSepa, in which I call:

SCIP_CALL(SCIPaddConsLocal(scip_, cons, NULL));

To add a constraint to the model. Problem is that my pricer is not called again before branching. The cuts have effect, meaning they actually close (incorrectly) the complete root node gap.

How can I enforce to start pricing again after adding the cuts? I already set the *result = SCIP_CONSADDED in the separator. I seem to have forgotten how to do this...

Kind regards,

Albert Schrotenboer
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.zib.de/pipermail/scip/attachments/20200729/e00aa40d/attachment.html>