Re[2]: [Help-glpk] binary variables writing lp format; and MIP bounds

[Andrew Makhorin]

>Assume that the problem is maximization.  Then the maximum of the LP
>bounds at active nodes is an upper bound on the true optimal solution
>value.  Currently, GLPK uses the LP bound at the pseudo-root, which is a
>valid upper bound, but may be higher than the maximum of the active
>nodes.

Yes, thank you very much. (The point is that I implemented mip_driver
using IOS as a prototype. However, IOS allows generating columns, in
which case the best bound *cannot* be used as the global bound due to
some missing columns, i.e. only the local bound at the pseudo-root can
be used as the global bound. And working on other things I absolutely
forgot that in mip_driver there is no column generation that allows
using the best bound as the global bound :+).


Andrew Makhorin