Re: [Help-glpk] Optimality!

[Michael Hennebry]

On Fri, 28 May 2004, Bernt-Ola Sandström wrote:

> Is there a way one can tell if (using glpsol) a solution is unique or if
> the optimum lies along an edge for example?

I think that glpsol will give you reduced costs if you ask it.
Possibly even if you don't.

> I had a test problem where I got a optimum solution, but when I tried
> the problem using the Interior point solver I got the same optimum value
> but different values for the variables. Trying Clp from coin-or gave a

When multiple optimal soultions are available,
interior point solvers will usually not give you a basic solution.

> third results for the variables but the same optimum. And when I ran
> this problem as an integer problem I got an answer but when I tried the
> same with Symphony and Sbb both said there is no integer solution.

It seems to me that the problem is so
ill-conditioned that round off errors matter.
There might, for example, be three inequalities
that combine to imply an equality.
If round off error tightens one of the constraints,
the result will be infeasiblilty.

You might want to test GLPK's answer for feasibility.

-- 
Mike   address@hidden
"Nothing says it like words if you know how to use them."
                    --  the Professional Organization of English Majors