Re: [Help-glpk] Modelling Advice Request - Project Tasks

[Jeffrey Kantor]

Hi Yingjie,

The 'BigM' piece was there in my first version as a way of having the solver calculate the position[a] data.  I was being lazy.  My later version avoided this by computing position in a post-processing step after the solver.  This problem doesn't actually need BigM at all.

I agree, this would be a nice problem to include in the example set.

Jeff

On Fri, Jan 22, 2010 at 3:21 AM, Yingjie Lan <address@hidden> wrote:

Hi Tawny，

My pleasure， surely enjoyed the problem as well. ;)
Probably should contribute this thing into the example set
that ships with GLPK, and maybe I will also add an exampel
in PyMathProg, once GLPK has it (which means GPL, hehe).

> > Big-M: M = 15 - 1 given 15 tasks.

> param BigM := 100;

Usually too big an M slows down the solver,
it might be good to make it as smaller as you can.

If you really have a huge model, say with
200 tasks, row generation suggested by others
might be a must, my rough guess.

Cheers,

Yingjie

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