[Scip] cutoffbound and early branching

[Eamonn Coughlan]

Hi everyone,

when working with a branch-and-price implementation, SCIP sets the 
primal bound as cutoffbound for the LP solver. Now i would like to 
integrate some kind of early branching.
Obviously when the lp solving stops because of the objective limit,
there is no primal solution, and i can't compute the bounds necessary 
to decide to branch early.
I would like to know if this is always beneficial (it could be that 
the pricing problem is far more expensive than a few more LP iterations 
perhaps?), and if not, what the recommended method is.
Also, how do i turn off the objective limit to test this? There is 
misc_exactsolve, but i assume there's a reason it's not a public 
parameter...

Regards,
Eamonn



===================================================================
EASY and FREE access to your email anywhere: http://Mailreader.com/
===================================================================