Re: [Help-glpk] Cannot get integer case to work

[William Gordon Rutherdale]

On 08/25/2009 08:41 AM, Andrew Makhorin wrote:

Glp_intopt requires an optimal solution to lp relaxation, i.e. in case
of mip you need to call glp_simplex first and then call glp_intopt.

  

That's funny, I specifically issued the statement
control_params.presolve = GLP_ON;
before calling glp_intopt().  My understanding from the documentation is that this would tell it to do the simplex pre-solve on its own.  Furthermore, the trace it showed on the screen indicated that it had in fact found the LP solution already, before tackling the integer programming problem.

Please see partial screen output below, from the original attempt.  Does the "Crashing..." part mean anything bad?

As a follow-up attempt I worked in a glp_simplex() call prior to the glp_intopt() call, just to see if that made a difference.  In that scenario it gave the LP solution as the integer solution, with fractional values for x and y.  I still haven't found a path that seems to work.

-Will

Solving problem with integer constraints.
ipp_basic_tech:  0 row(s) and 0 column(s) removed
ipp_reduce_bnds: 2 pass(es) made, 2 bound(s) reduced
ipp_basic_tech:  0 row(s) and 0 column(s) removed
ipp_reduce_coef: 1 pass(es) made, 0 coefficient(s) reduced
glp_intopt: presolved MIP has 2 rows, 2 columns, 4 non-zeros
glp_intopt: 2 integer columns, none of which are binary
Scaling...
 A: min|aij| =  2.000e+00  max|aij| =  1.000e+01  ratio =  5.000e+00
Problem data seem to be well scaled
Crashing...
Size of triangular part = 2
Solving LP relaxation...
*     0: obj =   0.000000000e+00  infeas =  0.000e+00 (0)
*     3: obj =   5.508333333e+01  infeas =  0.000e+00 (0)
OPTIMAL SOLUTION FOUND
Integer optimization begins...
+     3: mip =     not found yet <=              +inf        (1; 0)
+     4: >>>>>   5.300000000e+01 <=   5.300000000e+01   0.0% (3; 0)
+     4: mip =   5.300000000e+01 <=     tree is empty   0.0% (0; 5)
INTEGER OPTIMAL SOLUTION FOUND
result==0
z==0, x==0, y==0