Hi All,
I'm wondering if any cplex experts out there can tell me if it's
possible to pass a feasible LP solution (non integer solution) to a MIP
to be used as a starting point in the root node. I'm solving my
problem via ConcertTechnology/C Currently, I already have in place
code to pass an integer solution (via SetVectors) to cplex. This
integer solution, I assume will be used by cplex as the incumbent,
which is helpful for pruning purposes as the problem starts to branch.
However, I've noticed my problem does not branch very much and it's
taking a really long time to find an LP solution (for the root node).
Very often it is taking a few million iterations to find the solution
to the LP relaxation (at the root node). I've also notice that there
is a big gap between the objective value in the beginning iterations
(the first few thousand iterations) and the ending iterations. I have
a heuristic that can provide an LP solution that is sufficiently close
to the root node solution most of the time. So, I'm wondering if there
is a way to pass my MIP a starting non integer solution (basic or
non-basic) for the root node. I know that we can pass an LP a starting
basis via cplex.setBasisStatuses. Is there similar functionality for a
MIP? Thanks in advance for your help.
Regards,
Richard

rchen wrote:
Hi All,

I'm wondering if any cplex experts out there can tell me if it's
possible to pass a feasible LP solution (non integer solution) to a MIP
to be used as a starting point in the root node. I'm solving my
problem via ConcertTechnology/C Currently, I already have in place
code to pass an integer solution (via SetVectors) to cplex. This
integer solution, I assume will be used by cplex as the incumbent,
which is helpful for pruning purposes as the problem starts to branch.
However, I've noticed my problem does not branch very much and it's
taking a really long time to find an LP solution (for the root node).
Very often it is taking a few million iterations to find the solution
to the LP relaxation (at the root node). I've also notice that there
is a big gap between the objective value in the beginning iterations
(the first few thousand iterations) and the ending iterations. I have
a heuristic that can provide an LP solution that is sufficiently close
to the root node solution most of the time. So, I'm wondering if there
is a way to pass my MIP a starting non integer solution (basic or
non-basic) for the root node. I know that we can pass an LP a starting
basis via cplex.setBasisStatuses. Is there similar functionality for a
MIP? Thanks in advance for your help.

Regards,
Richard
Hi,
another suggestion: you can replace the simplex LP relaxation by an
interior point/barrier algorithm.
HM