[Clp] CLP gives infeasible while glpk and gurobi solve to optimality

[John Forrest]

Babak,

Probably Clp uses a tighter tolerance - the default primal tolerance is 
1.0e-7.

When I do "clp 2.lp", Clp's presolve says infeasible but solves it 
anyway.  If I look at solution I find -

** x2816     -3.7280722e-07 i.e. infeasible when unscaled

If I switch off scaling it says it is infeasible with sum of 
infeasibilities 7.8510045e-06.

With tolerance of 5.0e-7 it solves with or without scaling - but 
presolve still says it is infeasible as it uses its own tolerance - 
normally 1.0e-8.  So set -preTolerance 5.0e-7.

Check what tolerance glpk or gurobi is using.

John Forrest


On 21/02/15 03:32, Babak Moazzez wrote:
> I just found out that this happens in Clp 1.15. I tried Clp 1.06 and 
> it was able to solve it.
>
> Babak
>
> On Fri, Feb 20, 2015 at 7:18 PM, Babak Moazzez <bmoazzez at ucdavis.edu 
> <mailto:bmoazzez at ucdavis.edu>> wrote:
>
>     Hi
>
>     I have a model (attached) that clp says it's infeasible. But
>     gurobi and glpk solve it. Could someone help me figure this out?
>
>     Thanks.
>
>     Babak
>
>
>
>
> -- 
> Babak Moazzez
> Post Doctoral Scholar
> University of California Davis
>
> Email: bmoazzez at ucdavis.edu <mailto:bmoazzez at ucdavis.edu>
> Tel: (530) 752 5076
>
>
> _______________________________________________
> Clp mailing list
> Clp at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/clp

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/clp/attachments/20150221/9a1b8e1a/attachment.html>