[Scip] LP errors during branch and price and probing

[Timo Berthold]

Hi Frank.

Retcode -6 means SCIP_LPERROR (see type_retcode.h).
The typical case when this is returned is that the underlying LP solver was 
not able to solve the current LP, e.g., because it got stuck at a singular 
basis.

This raises three questions:
1. Which LP solver are you using? 
2. Which method in SCIP actually throws the error?
3. Have you any reason to believe that your LPs might be somehow nasty for an 
LP solver, e.g. a multi-level objective, large coefficients are highly 
degenerated vertices?

What you can try to debug is to write the problematic LP by calling 
SCIPwriteLP() when the error occurs. But probably the LP solver will solve it 
when being called from scratch, since the solution path is completely 
different.

Best, Timo

Am Thursday 02 July 2009 15:01:49 schrieb Frank Hennig:
> Hello everybody,
>
>
>
> I cannot resolve the following problem:
>
>
>
> Sometimes during solving my MIPs by branch and price I get SCIP_RETCODE
> of -6 and the whole MIP is stopped.
>
>
>
> The same can happen during probing, if I take a perfectly feasible
> fractional solution, start probing, fix one variable to 1 (which is
> always feasible) and solve the probing LP.
>
>
>
>
>
> Has anybody had the same problem before?
>
>
>
>
>
> Regards,
>
>
>
> Frank
>
>
>
> ________________________________________________________
>
>
>
> Frank Hennig  -  Naval Architect, PhD Candidate
>
>
>
> Department of Industrial Economics and Technology Management
>
> Norwegian University of Science and Technology
>
> Alfred Getz veg 3
>
> NO-7491 Trondheim, Norway
>
>
>
> Phone:   +49 30 314 23 598
>
> Email:      frank.hennig at iot.ntnu.no