[kde-edu]: Fwd: Timetable program

[Eva Brucherseifer]

On Thursday 10 April 2003 02:02, Dominique Devriese wrote:
> Birgit Schulz writes:
>
>   Birgit> Am Mittwoch, 9. April 2003 17:38 schrieb Eva Brucherseifer:
>   >> Can anyone take care of this, please?
>   >>
>   >> Thanks, eva
>
>   Birgit> I can have a look on it. But I don't have so much experiance
>   Birgit> with scheduling timetables.
>
>   Birgit> The screenshot is difficult to understand for me ... is
>   Birgit> there somebody who has more experience with scheduling.
>
> Hi,
>
> Just for the record, as I understand it, what this program does is to
> let the user enter a number of heuristics, to generate a timetable
> like
> "Teacher A teaches course B at time C in class D
> Teacher ..."
>
> I think this would be rather useful for schools etc., but imho, this
> program needs quite some UI work, a port to KDE instead of Qt, and of
> course an active maintainer in order for being of value to KDE-Edu.
>
> Lalescu, could you tell us what your intentions are with this program
> ?  Do you intend to keep it Qt-only, do you intend to continue to work
> on it ?
>
> I must say I have no idea how good the application's algorithm is, I
> once heard about how certain programs take days to calculate an
> appropriate schedule, or am i saying non-sense here ? :)

Maybe I can say something about this, since it's my field of research. The 
program uses a Genetic Algorithm, which is inspired by natural evolution. 
Solutions to a problem (which is a certain timetable here) are coded like 
genes and genetic operators like mutation and crossover are used to produce 
new solutions. The selection of those solutions who become parents depends on 
the fitness that the candidate parent solution has. 

Genetic algothims have been used before in this field. The quality of the 
output depends on 
 - the quality of the representation of solutions
 - if the fitness function really includes all aspects the user wants to see 
in the best solution
 - configuration of the algorithm (like which type of selection, rates of 
mutation and crossover, etc.)

I think choosing genetic algorithms to solve the timetable optimization 
problem is a good thing. Though I haven't seen any comparisons yet. 

Greetings,
eva

>
> cheers
> domi