I have a setup with 1500 activities spanning over 2 weeks (odd/even scenario). If I don't use the min hours per day for all teachers constraint I get a timetable in less than 1 minute. If I use that constraints FET doesn't finish even after 30 minutes.
I would like that most teacher have a minimum hours per day constraint set, but I would like to permit a few exceptions (when needed). I could probably do by using a constraints for each teacher, but that is too complicated (and besides I don't know before hand which teacher should have it and which should not).
So, my question is: Is it possible to have 0-100 weights for the constraint "Minimum hours per day for all teachers" (instead of the mandatory weight of 100%). Would that break FET algorithm?
Thank you,
Andrei
weight with less then 100% was poor (if i remember correct.). That was why Liviu disallowed it.
if it is the last missing constraint in your dataset, then i suggested to just wait some more. maybe a few hours.
if you want to have a few exceptions only you might just add dummy activities to that teachers. So just give that teacher an activity without students.
of course this dummy activity might "increase" the max number of gaps. so you maybe should add also an exception to the max number of gaps for that teacher.
I removed the 100% restriction for min hours daily for all teachers and for min hours daily for all studentes, compiled FET 5.30.2 from sources, and tested with my timetable. I have to say that I like the resulting timetable more. I used 99% weight for 2hours/day for all students, and I got a timetable much suitable for our needs.
I don't understand why some constraints require 100% weight (except for the basic time/space ones), but maybe there could be an advance setting to allow less that 100% weights for them (like the Allow empty days) so that people can test with them.
Thank you,
Andrei
Quote from: abautu on September 05, 2016, 09:26:56 AM
I removed the 100% restriction for min hours daily for all teachers and for min hours daily for all studentes, compiled FET 5.30.2 from sources, and tested with my timetable. I have to say that I like the resulting timetable more. I used 99% weight for 2hours/day for all students, and I got a timetable much suitable for our needs.
I do not understand. Do you have developed a timetabling program of your own, with constraints allowing 99% for min hours daily for students?
Please let me know more about this. I want FET to solve timetables in a satisfying manner.
Quote
I don't understand why some constraints require 100% weight (except for the basic time/space ones), but maybe there could be an advance setting to allow less that 100% weights for them (like the Allow empty days) so that people can test with them.
It is very difficult for me to explain, if you don't know the FET algorithm. Firstly (but not a good argument) is that in other constraints I assume that some constraints have 100% weight, to make checks and generate the timetable. Secondly, and this is a good argument, is that the weight for some constraints like min hours daily and max gaps should be something like 99.9999% to make it in real something like 98%. I'll try to explain, but for max gaps per week (not min hours daily, to which they are similar): the span of hours for a students set over the week should be maximum x (x = max gaps per week + n hours per week for students set). So FET skips once the constraint and allows a larger value for x, say x+1. This can be done once and remains the same for a long time, without being checked again soon, because the span will remain the same, x+1.
As an extreme provocation, I could try to make you a custom version allowing <100% for students/teachers min hours daily. But I cannot guarantee.
PS: I saw your paper about FET :)