[Coin-symphony] help resolve an MILP problem

Ali Baharev ali.baharev at gmail.com
Fri Apr 14 17:35:55 EDT 2006


A would like to (re)solve an MILP problem, starting from a known
feasible solution. Note: I _MUST_ start from the known feasible
solution (from given values of the binary variables). It is something
like Local Branching but is a bit different.

First i call branchAndBound() with the colBounds of the binary
variables fixed as in the known feasible solution (as the given values
of the binary variables).

After branchAndBound() finishes, i set all the colBounds of the binary
variables to (0, 1), i must change the upper bound of exactly one row,
and then i try to call resolve(). I get this:

"sym_warm_solve(): SYMPHONY can not resolve for the rhs change when
cuts exist, for now!"

OK. I tried to switch off cut generation with
setSymParam("generate_cgl_cuts", 0). If i do not switch it on again
before calling resolve(), SYMPHONY gets too slow (unacceptable).

I tried the following, first i switch off cut generation, and only
switch it on with setSymParam("generate_cgl_cuts", 1) AFTER
branchAndBound finishes. I got the same error message.

Could anyone help me?

Ali




More information about the Symphony mailing list