[Cbc] heuristics in binary variables
Hruby Martin
hrubym at fit.vutbr.cz
Thu Apr 14 04:14:14 EDT 2011
Hello,
I am implementing a rather large optimizer working in binary variables
(columns). The rows are in form
sum variables = 1
sum variables <= 1
sum var_1 - sum var_2 >= number
so not very various and the number of rows is significantly lower then
number of columns (probably usual).
The computing time is very bad even with rather small problems (e.g. 1000
columns). There might be some help using a heuristics, but, at the moment,
I do not know how to implement it; and this is my question to You.
The modeled problem has a characteristics which can be computed before the
start of branchAndCut or even dynamically during its branching process:
setting a variable to 1 enforces setting some others to 0 (or some to 1 as
well) - see above. I assume that this can be done by modifying the lower
and upper bounds of the columns.
Will You please make some suggestions to me? I have studied the heuristics
in the Cbc library but I did not understand well the mechanism of invoking
the heuristics procedure.
thanks a lot
regards
Martin Hruby
More information about the Cbc
mailing list