[Help-glpk] [Fwd: Help demand]

[Andrew Makhorin]

-------- Forwarded Message --------
From: Alexandre Saidi <address@hidden>
To: glpk <address@hidden>
Cc: Alexandre Saidi <address@hidden>, Andrew Makhorin
<address@hidden>
Subject: Help demand
Date: Mon, 11 Feb 2013 10:14:48 +0100

Dears,
I've been fighting some (few) quarters with the translation of the following 
constraint into GLPK.
If anybody can help with a working answer :

trying to translate efficiently the "element" CP constraint to GLPK, I've the 
following (simpler) constraints that i want to translate (but let's forget 
"element" for now) :

                (binary_y = 1) <=> (X = Z and W = K)            every thing is  
variable.

X,Z are bounded (1..60), W,K are also bounded (1..10, they are vector indices)

That's, if I ever have binary_y = 1, I want to put 2 pairs of variables equal. 
'and' is a logical connector and '<=>' stands for 'equivalent' (double 
'implies').
In fact, I'm working with a problem with 20000 variables and many many 
constraints. So the 'performance' is a critical issue.

I know that the 'element' constraint has been studied but those general 
solutions did not work (efficiently) for me.
A translation by the BigM method is welcome (apart from all cons of that 
method).

regards

Alex

        
 -------------------------------
Alexandre Saidi
Maitre de Conférences
Ecole Centrale de Lyon-Dép. MI
LIRIS-CNRS UMR 5205
Tél : 0472186530, Fax : 0472186443