> I want to ask you about something i was thinking of last week: wouldn`t
> it make sense to always use much higher values with timeplace and
> sametime modules thant with the rest of them?
> I read the docs and they say you should put higher values on mandatoy
> restrictions. All my restrictions are mandatory (a lot by law) so i set
> similar weights for all modules: the result was that tablix could not
> find any solution with only two classes. After looking at the genetic
> alg and your sample data, i put much higher values in timeplace and
> sametime, and a solution was found in minutes.
> So i would like to know if this is only a coincidence (i don't think
> so), or it is what should always be done, in wich case you could add it
> to the docs.
If you had to assign higher weight values to timeplace.so and
sametime.so, that means that the timetabling constraints from these two
modules were the hardest to solve in your particular problem.
From my experience it isn't uncommon for some other module to require a
higher weight than sametime.so. For example consecutive.so and
sametimeas.so in Tablix 0.3.0.
Also it seems that assigning weight values that are too high to some
mandatory modules will somewhat decrease the quality of the final
result, if some modules were non-mandatory. As far as I know it is best
to choose weight values that are just high enough to ensure Tablix will
get a solution.
Manual used to contain a table of best weight values for some modules.
I later removed it because because the best combination of weights
depends on the specific problem and there were too many modules to test
all the combinations.
Received on Fri Nov 04 17:42:09 2005
This archive was generated by hypermail 2.1.8 : Sat Nov 05 2005 - 06:27:43 CET