sci.op-research

Min wrote:
> Hi all-
> I have a question about solving MIP in CPLEX. Does adding an initial
> feasible solution in CPLEX help speed up?

Sometimes yes, sometimes no (which actually answers a wide variety of
questions about MIPs).

> I was suggested to do so by
> someone. But I don't think it helps since CPLEX can find a feasible
> solution of my problem very quickly (in seconds).
>

If you know a very good solution at the outset (objective value better
than that first feasible solution that CPLEX finds quickly), it may be
worth inserting it at the outset. The sooner CPLEX sees a good solution
(and the better that solution is), the faster it will fathom nodes --
maybe. (The other factor is how tight the bounds on those nodes are.)
If the initial solution you would be inserting is not better than what
CPLEX finds on its own, you won't save much time.

/Paul