Re: [Help-glpk] Small MILP Problem which is Integer-Infeasible in GLPK 4.47 but Optimal in COIN-MP, LPSOLVE and XPRESS

[Kelly, Jeff (ON0F)]

Tor;

Thank you - I think this is my problem - I also had a similar situation with 
variables that are "fixed".

I believe the other solvers will internally recognize this and solve according.

All the best - Jeff

-----Original Message-----
From: Tor Myklebust [mailto:address@hidden 
Sent: Thursday, January 19, 2012 8:50 AM
To: Kelly, Jeff (ON0F)
Cc: address@hidden
Subject: Re: [Help-glpk] Small MILP Problem which is Integer-Infeasible in GLPK 
4.47 but Optimal in COIN-MP, LPSOLVE and XPRESS

On Thu, 19 Jan 2012, Kelly, Jeff (ON0F) wrote:

> I have attached 3 LP files for a small MILP problem that goes 
> integer-infeasible in GLPK 4.47 but is optimal in CBC (LP file N/A), 
> LPSOLVE and XPRESS.
>
> The optimal OBJ = 28.0 where a 3 binary should equal 1.0.
>
> Although GLPK gets the correct LP root relaxation OBJ value it fails 
> in the branch-and-bound.

If I make v1_gprofitterm(1,1) and v1_totalobjfun(5,1) free, it works for me.  
Having some bounds on the order of 1e20 and other bounds on the order of 0.0026 
is probably not what you wanted.