True! I've tested it and values that are too large spoil the picture.
But with more reasonable values it works like a charm.
Thank you for your help,
João
Em 10/11/10 16:00, Michael Hennebry escreveu:
On Wed, 10 Nov 2010, Suleyman Demirel wrote:
Usually, if you have an either/or constraint, you should define a binary
variable, say y, taking only 0-1 value. If y=0, the sum is less than
zero,
if y=1, the sum is greater than two.
Let M be a very large number (M=100000000000).Then, you should have two
constraints as follows.
Nyet.
M should be the smallest value which will work.
Merely large values give loose constraints.
Truly huge values can cause numerical difficulties.
s.t. condition1{(i,j) in E}: (sum{(i,j,i1,j1) in L} x2[i1,j1] <=M *y;
Each of the above M's should at most be an upper
bound on the corresponding sum on the left.
s.t. condition1{(i,j) in E}: (sum{(i,j,i1,j1) in L} x2[i1,j1] >=2*y - M
*(1-y)
Each of these M's should be at most an upper bound on
the negative of the corresponding sum on the left.
_______________________________________________
Help-glpk mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/help-glpk
