Re: [Bug-glpk] MIP false optimal

[Andrew Makhorin]

On Wed, 2018-02-07 at 23:11 -0500, address@hidden wrote:
> Also I found a problem (attached) where glpk returns: "Status:
> INTEGER OPTIMAL"  for a suboptimal solution.

> Best objective should be 45040 rather than the  46040 glpk finds. 

> -Ryan

Thank you for your bug report.

Glpsol may report a suboptimal solution as optimal because your mip is
badly scaled:

Scaling...
 A: min|aij| =  1.000e+00  max|aij| =  2.032e+09  ratio =  2.032e+09

Since glpk uses floating-point calculations, this may affect the
solution obtained.

Probably you need to reformulate your mip to avoid huge constraint
coefficients at binary variables.


Andrew Makhorin