Re: [Help-glpk] HELP - INTEGER PROBLEM

[Andrew Makhorin]

> I'm sorry, but on this computer I do not have the availability of
> zipping files. I send you the mip in .pdf. I forgot a thing: the value
> SIRmin = 0.03125, so 1/SIRmin = 32. The set J has 10 elements, the set
> I has 50 elements, so the binary variables (all) of the problem are
> 500.  The constraints with 1/SIRmin are 10, the constraints (1) are
> 50. I wrote (2) e (3) like "bounds" of variables. The parameters, u
> (objective function) and g (constraints) are double numbers. The
> result is:
> 60 rows, 500 integer variables, 500 non-zeros
> 500 variables, all of which are binary
> Preprocessing
> 50 rows, 500 integer variables, 500 non-zeros500 integer variables,
> all of which are binary
> 500 variables, all of which are binary
> Scaling...
> A: min|aij| = 1.000e+000  max|aij| = 1.000e+000  ratio = 1.000e+000
> Problem data seem to be well scaled
> Constructing initial basis...
> Size of triangular part is 50
> Solving LP relaxation...
> GLPK Simplex Optimizer, v4.52
> 50 rows, 500 columns, 500 non-zeros
> *     0: obj =  5.000000000e+001  infeas = 0.000e+000 (0)
> OPTIMAL LP SOLUTION FOUND
> Integer optimization begins...
> +     0: mip =     not found yet <=              +inf        (1; 0)
> +     0: >>>>>  5.000000000e+001 <=  5.000000000e+001   0.0% (1; 0)
> +     0: mip =  5.000000000e+001 <=     tree is empty   0.0% (0; 1)
> INTEGER OPTIMAL SOLUTION FOUND
> obj = 50.0
> TP1 = 0.0
[...]
> TP500 = 1.0
> 
> I conclude, from this solution, that GLPK "delete" 10 constraints. I'm
> wrong? I tried to disable Preprocessor but does not work. Thank for
> your answer and you help.

Please check your code more carefully. Most likely the model you pass to
glpk is incorrect, so the solution is not what you expect.

I'd suggest you to prepare your model in the GNU MathProg modeling
language rather than programming it in Java, and then solve it with
glpsol, the stand-alone glpk solver. This would make things much more
easier.