[Clp] Warmstart Query

[John Forrest]

Piyush,

If you use memcpy to copy into columnSolution and then use primal(1)
then Clp will recalculate the rowActivity from the columnSolution and do
one pass through matrix using solution and moving variables into basis
or to bounds.  Then it will continue with a good basis.

It is worth trying putting some variables in initial basis but it may
not help as many of them may get thrown out.  If you want to try that
then mark all slack variables as at bound (lower) and sort column
variables depending on how far away they are from bound.  Then mark the
best numberRows variables (or all away from bound if <numberRows) as
basic.  Try the simple approach first and then see if the second one
helps at all.

Of course it is possible that dual simplex could be much better than
primal, in which case neither approach would help - it all depends on
shape and type of problem.

John Forrest

On Tue, 2010-07-13 at 19:29 -0700, Agram Piyush Shanker wrote:
> Hi,
>     I have large minimization problem (LP) with all equality constraints (Ax=b). I also happen to know a feasible solution (x_0) to the set of these equality constraints. How do I let the solver to use this feasible solution as a starting point.
> 
> I have tried memcpy to copy into rowActivity and columnSolution. How do I let the solver automatically update the other variables/ arrays based on this information?
> 
> Thanks
> Piyush
> 
> _______________________________________________
> Clp mailing list
> Clp at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/clp
>