Re: [Help-glpk] MILP solutions using different solvers

[Andrew Makhorin]

On Fri, 2013-03-01 at 09:06 -0800, Sundaravalli Narayanaswami wrote:
> Hi all,
> 
> I am working on a MILP model for a transportation problem, with 24
> data sets. Some of the data sets are solved to integer optimality and
> others return an integer undefined by GLPK solver. 
> 
> To verify and evaluate my solutions, I wanted to solve them using
> other commercial solvers. I used Gusec to translate my mod files (no
> external dat file) into lp and mps files. I ran each of my lp files on
> CPLEX and each of my mps files on Gurobi. Gurobi and CPLEX solutions
> are exactly same for each problem instance and GLPK solutions are
> significantly different (Small valued for a minimization problem)
> 
> Why are GLPK results hugely different, when all other solvers generate
> numerically the same solutions? I set similar parameters in all
> solvers. 

Do the solutions differ in the optimal objective values or only in the
values of the variables ?

> 
> TIA,
> SV
>