> Andrew, you suggest that  z= -max(x, y)  (negative of max) can
> be modeled  through min, not z=max(x,y), is that right?
> 

Minimizing an objective like 

   z = +max(x1,x2,...)+max(x3,x4,x5,...)+max(...)+ ...

does not need binaries, because the objective is convex. Obviously, if
the objective is the following:

   z = +max(x1,x2,...)-min(x3,x4,x5,...)+...

we can substitute -min(x3,x4,x5,...) = +max(-x3,-x4,-x5,...).
Maximization case can be derived by multiplying the latter by -1:

   maximize z = -max(x1,x2,...)+min(x3,x4,x5,...)+...

that gives exactly the same objective as in nikel's question (note that
max's have a minus sign while min's have a plus sign).


_______________________________________________
Help-glpk mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/help-glpk

Reply via email to