Re: [eclipse-clp-users] More than one optimal solution using eplex.

From: Paul Cannongeyser <paul_dice_2_at_yahoo.com>
Date: Wed, 6 Nov 2013 11:57:25 -0800 (PST)
So, if I know of a solution (in this case, an optimal solution), I can find it by adding additional 
constraints on some of the variables.

E.g., in this particular example, the additional constraint,

prob: (B2                $=   9),

enables me to get the same solution as shown in the tutorial.

?- main2(Cost, Vars).
Cost = 710.0
Vars = [0.0, 21.0, 0.0, 16.0, 9.0, 15.0, 34.0, 0.0, 0.0, 0.0, 0.0, 10.0]
Yes (0.00s cpu)

Thank you, Kish, for your answer.



On Thursday, October 31, 2013 1:23 AM, Kish Shen <kisshen_at_cisco.com> wrote:
 
Hi,



On 30/10/2013 21:27, Paul Cannongeyser wrote:

> "Note that, in general, an MP problem can have many optimal solutions, i.e. different solutions which give the
> optimal value for the objective function."
>
> The tutorial has one optimal solution.  My recent run got another.
>
> How do I get the same optimal solution as the tutorial?
>

Short answer: you cannot.

In general, for a constraint problem, if all your constraints are 
satisfied, then you cannot demand a specific solution unless you post 
extra constraints (to remove solutions you don't want). The exact 
solution returned by the solver is likely determined by implementational
details. So in the case of MP solvers, you should not expect different 
solvers, or even different versions of the same solver, to return the 
same solution.

In fact, you probably should not even expect the same solver to return 
the exact same solution for the same problem in all cases, because it is 
possible that the solving algorithm may use some randomisation 
component, although I suspect for debugging reasons, such behaviour is 
not likely to be the default.

Cheers,

Kish
Received on Wed Nov 06 2013 - 19:57:32 CET

This archive was generated by hypermail 2.2.0 : Fri Nov 08 2013 - 06:14:17 CET