[Symphony] Change epsilon?

[Ted Ralphs]

The tolerance is not exactly hard-coded---but it is inherited from the
underlying LP solver, so unfortunately that means there is no easy,
direct way at the moment to change it from within SYMPHONY. However,
it *should* work to go into the open_lp_solver() subroutine in
lp_solver.c and just add a line there that changes the feasibility
tolerance for the underlying LP solver before SYMPHONY sets its own
tolerance. Something like

lp_data->si->setDblParam(OsiPrimalTolerance, lp_data->lpetol);

should work. Make sure you are looking at the part of lp_solver.c for
the interface to Osi. If this is acceptable, we can try to add a
parameter to SYMPHONY to do this automatically.

Cheers,

Ted

On Fri, Jun 24, 2011 at 1:03 PM, Michael Hennebry
<hennebry at web.cs.ndsu.nodak.edu> wrote:
> On Fri, 24 Jun 2011, Terry wrote:
>
>> In one solution, SYMPHONY finds an integer variable is 2.00000989. How can I take the precision out a few more decimal places?
>>
>> What controls how close to an integer a variable must be? Is that epsilon? Can I change epsilon? It looks to be hard-coded. I tried changing granularity, but that's not it.
>
> You can.  I don't remember how.
> It might be a good idea sometimes.
> Not this time.
> Floating point computation is not exact.
> Demanding six decimal places is already pushing it.
> A better choice might be to use a callback to
> see whether rounding produces a feasible solution.
> If not, declare infeasiblilty and add a cut.
>
> Suppose
> x1 + x2 + x3 + x4 <= 3
> 0 <= x <= 1
> If all x's are outside [.5, .8],
> then rounding x will not violate
> x1 + x2 + x3 + x4 <= 3
>
> In the case of all binary variables, the range [1/(n+1), 1-1/(n+1)]
> can be useful.
> It guarantees that if rounding fails to produce a feasible solution,
> a cut is readily found.
>
> --
> Michael   hennebry at web.cs.ndsu.NoDak.edu
> "Pessimist: The glass is half empty.
> Optimist:   The glass is half full.
> Engineer:   The glass is twice as big as it needs to be."
>
> _______________________________________________
> Symphony mailing list
> Symphony at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/symphony
>



-- 
Dr. Ted Ralphs
Associate Professor, Lehigh University
(610) 628-1280
ted 'at' lehigh 'dot' edu
coral.ie.lehigh.edu/~ted