# [galib] A representation question as a GAlib beginner

zhangh1@umbc.edu zhangh1 at umbc.edu
Fri Apr 15 16:21:50 EDT 2011

Hi Patrick:

Probably I should have put my problem a more straightforward way. Lets say
100 kinds of fruits, and I want to find the best combination of 10
different kinds of fruits (say to make the best jam). Now I number these
100 kinds of fruits from 1 to 100 (I called this "index"), and I want to
evolve over these index numbers. I need to make sure for each genome there
are no duplicate numbers (each fruit selected at most once). I want to use
GA1DArrayGenome by not sure how to make sure each number is identical.

Best Regards, Hailiang

> Le 15/04/2011 20:04, zhangh1 at umbc.edu a écrit :
>> Hi,
>>
>> I am new to GA, and wonder whether somebody can help me to set up the
>> representation for a common problem (should be:). Basically I have a
>> structure pool of size 10000 and try to find the best combination of 100
>> (no overlap) that can generate the ideal experimental results. I plan to
>> make the an array of size 10000 as the pool, and choose the array index
>> as
>> the parameters that I want to evolve. I am thinking about using
>> GA1DArrayGenome of integer type ranging from 1 to 10000 to do this job,
>> and have 2 questions:
>>
>> 1. How to make sure each of the 100 genes in the genome is unique (and
>> ignore the child containing 2 or more identical genes during the
>> evolution)?
> I'm not sure I really understood your goal : You plan to build an array
> of 100X100 genes (indexes). Each combination point on an experimental
> design. Your purpose is to find the best combinaison to achieve the best
> exp. design ?
>
> If yes,  you have no certitude that from generation to generation you
> could not get "2 or more identical genes". But the exploration of the
> entire space of index is then depending of the number of generations.
>> 2. Is evolving over the index a good choice at all? Or any other
>> suggestions?
> What I undestood is that the experimentals results are already fixed and
>  From my point of view, it would be preferable that they were not, but
> that the genetic algorithm should try to find (and to build and then to
> test) the best one (i.e. the exp. that optimizes the results of a
> function).
> But, may be I misundestood your problem.
>
> Best regards, Patrick.
>> Thanks a lot for any help!
>>
>> Best Regards, John
>>
>> _______________________________________________
>> galib mailing list
>> galib at mit.edu
>> http://mailman.mit.edu/mailman/listinfo/galib
>>
>
>
> --
> Le Savoir n'est pas une marchandise
> L'Université n'est pas une entreprise
> --
>
> Dr. Patrick Coquillard
> Institut Biologie et santé des Végétaux
> 400, route des Chappes
> 06 903, Sophia Antipolis
>
> Tél/fax : 04 92 38 64 30
> http://www.unice.fr/coquillard
>
>
>