RE: [Help-glpk] GLPK found optimal solution while XPRESS & CPLEX concluded to unbounded problem, for the same problem.

[Maurice.Djona]

Thanks Michael for your suggestion.
After increasing the value of the "Optimality tolerance" of CPLEX from
1e-6 to 1e-3, I obtain the same result as with GLPK.
(This may also suggest that GLPK is a more accurate LP optimizer!).

Maurice Djona

-----Original Message-----
From: Michael Hennebry [mailto:address@hidden 
Sent: October 4, 2004 10:50 AM
To: address@hidden
Cc: address@hidden
Subject: Re: [Help-glpk] GLPK found optimal solution while XPRESS &
CPLEX concluded to unbounded problem, for the same problem.


On Sun, 3 Oct 2004 address@hidden wrote:

> In the majority of cases the 3 software give the same results. 
> However, for the problem given below, GLPK (4.7) found an optimal 
> solution while XPRESS and CPLEX concluded to an unbounded problem. The

> 3 results are:

Have you checked reduced costs, especially on column C120?
A small computed reduced cost might changed in
sign depending on differences in round off error.

> FOR GLPK:
> lpx_read_cpxlp: 53 rows, 133 columns, 212 non-zeros
> lpx_read_cpxlp: 197 lines were read
> *     0:   objval =   0.000000000e+00   infeas =   0.000000000e+00
(23)
> *     8:   objval =  -1.000000010e-07   infeas =   0.000000000e+00
(19)
> OPTIMAL SOLUTION FOUND
>
>                                 *********
>
> FOR XPRESS:
> Problem Statistics
>           53 (      0 spare) rows
>          133 (      0 spare) structural columns
>          212 (      0 spare) non-zero elements
> Global Statistics
>            0 entities        0 sets        0 set members
> The problem is unbounded due to column C120
>
>                                 *********
> FOR CPLEX:
>
> CPLEX also concluded to unbounded problem.
>
>                                 *********

-- 
Mike   address@hidden
ValleyCon 30 is coming:   The Red River Fantasy Festival
October 22-24             www.valleycon.com