search

fcampelo / MOEADr

R package MOEADr, a modular implementation of the Multiobjective Evolutionary Algorithm with Decomposition (MOEA/D) framework
20 stars 7 forks source link

Finished Uniform Decomposition (minus comments, sorry!) #5

Closed caranha closed 7 years ago

caranha commented 7 years ago

I'll finish the comments (and remove my blog comments) in my next coding session -- stay tuned!

fcampelo commented 7 years ago

One thing I couldn't exactly get from your (very elegant, btw) implementation: in the calculation of S2 (row 81, within subroutine cd2), it looks to me like you kept the triple-sum, instead of the "double sum + prod" (which was corrected in the manuscript). Or is it just a misinterpretation on my part?

caranha commented 7 years ago

No, you are correct, I did not notice the change to double sum + prod. Will fix that.

On Wed, Jan 4, 2017 at 9:28 AM, Felipe Campelo notifications@github.com wrote:

One thing I couldn't exactly get from your (very elegant, btw) implementation: in the calculation of S2 (row 81, within subroutine cd2), it looks to me like you kept the triple-sum, instead of the "double sum + prod" (which was corrected in the manuscript). Or is it just a misinterpretation on my part?

-

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/fcampelo/MOEADr/pull/5#issuecomment-270264614, or mute the thread https://github.com/notifications/unsubscribe-auth/ADxs7Nyoa154C_3N-g8mggrTVlgO22Ugks5rOueqgaJpZM4LVKME .

-- Claus Aranha http://claus.castelodelego.org

caranha commented 7 years ago

Ok, finished adding the double sum + prod.

It takes 1 second when the number of subproblems is 1 and around 3 minutes when the number of subproblems is 200.

However, doing some calculations, I realized that the Uniform decomposition is O(N^3obj) (CD2 is (N^2obj), and the expected number of co-primes of N is N). So I'm not sure there is a good way to reduce this further. We must make sure that decomposition is not called more than once when repeating experiments with the same parameters.

Pls Pull!