[SCIP] Farkas pricing

[Ngô Lan]

Hi Gerald,


Thank you for your fast reply. In my implementation, I do not implement the Farkas pricing but a heuristic in the farkas callback to generate more columns. So I didn't call SCIPcolGetRedcost directly but it was called when the master is resolved after the new column is added. The error I got is "lp.c:3899: SCIPcolGetRedcost: Assertion `lp->validsollp == stat->lpcount' failed."

I was wondering why this error occurred since the master is infeasible so when it is solved again after farkas, it should not call SCIPcolGetRedcost. Or did I do something wrong that leads to that SCIPcolGetRedcost call?


Best,

Lan

________________________________
From: Scip <scip-bounces at zib.de> on behalf of Gerald Gamrath <gamrath at zib.de>
Sent: 06 November 2019 12:02:40
To: scip at zib.de
Subject: Re: [SCIP] Farkas pricing

Dear Lan,

sorry, the documentation is a bit misleading in this case. Similar to constraints, there is a separate method to get the Farkas coefficient for variables. Please use SCIPgetVarFarkasCoef() for this.

Best,
Gerald

On 06.11.19 02:10, Ngô Lan wrote:

Hi all,


I am implementing BPC and after some branchings, my master will become infeasible due to not having enough columns in the pool. Instead of Farkas pricing, I implement a heuristic to add more columns into the problem inside the SCIP_DECL_PRICERFARKAS callback. However I ran into the assertion error "lp->validsollp == stat->lpcount" in function SCIPcolGetRedcost. I can understand that this is due to the lastest solution being infeasible but then how Farkas works normally? It would be great if someone can give me insight into this. Thank you for your time.


Best,

Lan



_______________________________________________
Scip mailing list
Scip at zib.de<mailto: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/20191107/4433d941/attachment.html>