[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Help-glpk] glpk

From: Andrew Makhorin
Subject: Re: [Help-glpk] glpk
Date: Fri, 14 May 2010 20:00:34 +0400

> I send you an instance in cplex lp format as requested.
> Notice that GLPK status is OK (an optimal solution has been found), but
> the solution is actually wrong.


The file has some errors (missing variable names in some constraints).

However, I noticed that your model has tiny constraint coefficients of
order 1e-16, e.g.:

 c7: 2.22044604925031e-16 x2 + x3 - 2.22044604925031e-16 x258 - x259
     <= 13.6066766666667
 c8: - 2.22044604925031e-16 x2 - x3 - 2.22044604925031e-16 x258 - x259
     <= -9.60665666666667

Such small coefficients make your model extremely badly scaled that
may cause significant round-off errors in the solution. I would suggest
to replace such coefficients by exact zeros, i.e. eliminate them from
the model, or at least, not to use the geometric mean scaling (it is
enabled by default).

reply via email to

[Prev in Thread] Current Thread [Next in Thread]