I have a bizarre file. FET solved it successfully in nearly 7 minutes, without any conflicts(soft conflicts=0). the random seed before generation is :
s10=2, s11=1, s12=0,
s20=4, s21=3, s22=0;
Weight of min days =99.75℅. (For activities divided into 4 subactivities)
For other activities weight=100℅.
But when I modified the weight to 100% with the same random seed cited above, FET failed to find a solution? (I launched it for 10 hours, it reached 506/507 activities in 1 hour 5 minutes and then endless!)
to tell the truth, I expected that FET would find the same solution.
file attached below ...
No, it is not exactly the same simulation. Because at a moment it might break, then comes back and not break anymore the constraint.
So a very small difference can change the behavior of FET! even if we have the same data entry, the same constraints, the "same" everywhere, except for this one!?
So I can conclude that for better FET behavior, and less generation time as possible, set the weight of min days, for some activities (or all activities), for such files to a value less than 100(maybe 99.75% is better)?
is this correct?
Quote from: Benahmed Abdelkrim on August 09, 2020, 08:30:56 AM
So a very small difference can change the behavior of FET! even if we have the same data entry, the same constraints, the "same" everywhere, except for this one!?
Yes, a very small difference counts.
Quote
So I can conclude that for better FET behavior, and less generation time as possible, set the weight of min days, for some activities (or all activities), for such files to a value less than 100(maybe 99.75% is better)?
is this correct?
No, this is not correct. I think it is just luck. Please try some more generations with 99.75% and 100%.
maybe conflicts existed during the generation which were neutralized by other permutations
Quote from: Benahmed Abdelkrim on August 09, 2020, 06:26:28 AM
I have a bizarre file. FET solved it successfully in nearly 7 minutes, without any conflicts(soft conflicts=0). the random seed before generation is :
s10=2, s11=1, s12=0,
s20=4, s21=3, s22=0;
Weight of min days =99.75℅. (For activities divided into 4 subactivities)
For other activities weight=100℅.
But when I modified the weight to 100% with the same random seed cited above, FET failed to find a solution? (I launched it for 10 hours, it reached 506/507 activities in 1 hour 5 minutes and then endless!)
to tell the truth, I expected that FET would find the same solution.
استاذ عبد الكريم كان لدي تسآءل في هذا الموضوع جاءت فرصة لطرحة : هل نفس البذور العشوائية لجدول انتجته بنجاح أستطيع استعمالها مرة أخرى مع وضع تعديلات في القيود اخرى( ليس شرط نسبة القيود في ادنى ايام بين انشطة 100- 99.75) وتعطيلي انتاج ؟؟ ... ام يرجع ذلك لطبيعة القيد
مثال : جدول انتجته بمطلق جزائري و اغير الى استثناءالاساتذة في ايام ( مغربي جزائري ) ا.
file attached below ...
Quote
No, this is not correct. I think it is just luck. Please try some more generations with 99.75% and 100%.
test: no solution found with a weight = 100%, until now. but when I try with 99.75% for all activities, FET succeeds in finding a quick solution!...
Try multiple with limit = say 15 minutes. If you have at least dual core, start two FETs, one with 99.75% and one with 100%, and let me know. Leave it over night.
Ok, I will run it now. Fortunately, I have a dual-core :-)
... just now
Quote from: Liviu Lalescu on August 09, 2020, 02:48:30 PM
Try multiple with limit = say 15 minutes. If you have at least dual-core, start two FETs, one with 99.75% and one with 100%, and let me know. Leave it overnight.
Yes, you are right, I tried multiple generations. I started with 2 FETs instances; one with weight=99.75% of min days (for all constraints of this type), and the other with weight=100% of min days. The time limit for every timetable=15 minutes. and I have a pleasure to let you know that I have got timetables with this method in record time(5~12 minutes).
Please, can we know the history of this powerful feature; MULTIPLE GENERATION?
I attach the results of this test as text files with 2 screenshots.
Quote from: Benahmed Abdelkrim on August 10, 2020, 07:16:24 AM
Please, can we know the history of this powerful feature; MULTIPLE GENERATION?
I am not sure I understand.
Quote
I am not sure I understand.
I mean, when this function was added to FET? and in which version?
In FET-5.2.0, suggested by Volker. See the ChangeLog. Probably ~2007.
Thank you, I will see it. :)