RE: [Help-glpk] Hi: Question. Please help.

[address@hidden]

Hi Prasad,

If I understand correctly....you could try:

1. Let's say you have N variables (x)
2. Create a parameter of dimension N, containing the original values of
your variables (a)
3. Introduce N binary variables (u)
4. forall (n in N): -M x u_n <= x_n - a_n <= M x u_n
5. sum u_n <= maximum number of changes

M is a "big" number. This should be chosen as small as possible (creating
the smallest problem possible plus smallest change of frustrating round off
errors). I must say that with a 1000 variables, this might slow down the
solver drastically, nevertheless you could give it a try.

I hope this helps.

Regards,

Timo


Original Message:
-----------------
From: Jahnavi Prasad address@hidden
Date: Fri, 25 Jul 2008 02:18:20 +0400
To: address@hidden
Subject: [Help-glpk] Hi: Question. Please help.


Hi,

I am trying to use GLPK in a flux-balance analysis context in biology.
Once the linear constraints are defined and maximization of the
objective function is done, I often find that the solution contains too
many changes across the free variable set, and I cannot change that many
variables (over 1000 sometimes) for my engineering problem. I can at the
most take care of say a 15-20. How can I restrict the number (not
magnitude) of changes in the LP maximization?

Can someone please answer this question?

Thanks a lot,
regards,
Prasad






_______________________________________________
Help-glpk mailing list
address@hidden
http://lists.gnu.org/mailman/listinfo/help-glpk



--------------------------------------------------------------------
myhosting.com - Premium Microsoft® Windows® and Linux web and application
hosting - http://link.myhosting.com/myhosting