[SCIP] What SCIP does when objective function equal to zero
Vladimir V. Voloshinov
vladimir.voloshinov at gmail.com
Fri Dec 27 22:56:30 CET 2019
Dear SCIP team,
Merry Christmas! and Happy New Year!
But still we have a question: how SCIP (and ParaSCIP) treat the problem
where objective function is absent, i.e. equal to zero?
That is the problem is actually a satisfiability problem, the search of a
feasible solution for a set of constraints given.
My colleagues ask me to run ParaSCIP with some rather hard problem with
binary variables of the above type, which took about 170 CPU*hour to be
solved on rather power cluster.
In attempt to reduce solving time I have recommended them to reformulate
the problem, i.e. to bring some constraints to the criterium via penalty
function... But as a result we got much more hard problem with dozens more
solving time (I even had to kill ParaSCIP after 1000 CPU*hours of
calculating with gap around 20%).
The same effect was for variants of the same problems of smaller dimensions
(solved by SCIP): converting some constraints to the penalty made the
problem much more hader (for SCIP).
Can you tell a few words about SCIP's approach to the problem with constant
goal function? May be there are some papers on the subject?
Sincerely yours,
Vladimir.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.zib.de/pipermail/scip/attachments/20191228/e3094a58/attachment.html>
More information about the Scip
mailing list