[SCIP] Defining binary variables for "top level" branching

[Vladimir V. Voloshinov]

Dear SCIP developers,

We have a set of MIP problems with all-binary variables and ZERO objective
function (so, we are looking for a feasible solution).
Typical dimensions are: about 6000 binary variables and 3000 constraints.
FiberSCIP with 20 threads and (almost) default settings solves these
problems in about 10 minutes but we have an idea to reduce solving time by
a "special ramp-uping".
The thing is that we know, in advance, about 30 variables, which are worth
(we think so) to be selected for creation of "search subtrees' root nodes"
by fixing their values to 0 or 1.

I searched for proper settings in UG-parameters (*.prm file generated by
fscip or parascip), but could not find.
So we have a couple of questions.
1. Is it worth trying "home-bred" rump-up rules, or should we trust UG
heuristics?
2. Is it possible to set the "rule for ramp-up phase" mentioned above
somehow? If yes, in what way: by some settings of UG-parameters or we need
UG API?

And almost the same question regarding SCIP: is it possible to set the rule
for "search tree 2nd level nodes creation", or we need SCIP API. e.g. from
https://www.scipopt.org/doc/html/group__BRANCHINGRULES.php?

Sincerely yours,
Vladimir V. Voloshinov,
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.zib.de/pipermail/scip/attachments/20200707/482d80d4/attachment.html>