Re: [eclipse-clp-users] Knappsack-Problem Modification

W dniu 2010-04-14 11:45, Philipp Marcus pisze:
> Hi,
>
> thanks for all your replies. I think Marco's suggestion to minimize 
> the total dissimilarity instead of the total average dissimilarities 
> is great. I've implemented a version with ic and ic_symbolic (though I 
> know it's not neccessary). Maybe I'll try to do it with eplex later, 
> but I am not familiar with that lib.
>
> Another problem that comes up are the similarities among the solutions 
> as it obviously does not matter in which knapsack several items are 
> packed, but that they are in the same.
> Example:
>
>     I1 = "B"        I1 = "C"
>     I2 = "C"        I2 = "B"
>     I3 = "C"        I3 = "B"
>     I4 = "B"        I4 = "C"
>     I5 = "D"        I5 = "D"
>     I6 = "D"        I6 = "D"
>
>
> Does anyone have a suggestion how to realize symmetry breaking during 
> search (e.g. by using ic_sbds) in my implementation? A call could be:
You gave a suggestion. I would try replacing particular knapsack with 
"being in the same group" relation and add a constraint for a number of 
those groups.

Best regards

-- 
[ Wit Jakuczun http://pl.linkedin.com/in/jakuczunwit ]