Mime-Version: |
1.0 |
Sender: |
|
Date: |
Wed, 1 Aug 2007 09:31:46 -0400 |
Reply-To: |
|
Subject: |
|
Content-Transfer-Encoding: |
7bit |
Content-Type: |
multipart/mixed;
boundary="sZUUxzfE.Dkd5X-gR7d)XtLGT6JEkusQ-B+b5L3sHUB89I2IF2,YdARFr3v,(nBdC5'lk(" |
From: |
|
Comments: |
|
Parts/Attachments: |
|
|
Hi all,
I am experimenting at the moment on how various replacement strategies
affect gp evolution. In particular I am examining the so-popular
generational and steady state approaches which are already implemented in
ECJ. Apart from that Ive implemented two other approaches.
Briefly, in the intermediate approach, a X% of the population is produced
(0<X<=90) , then the rest of percentage is filled with members from the old
population. In the greater-than-generational approach, similarly, a X%
(200<X<600) is combined with the old population, to allow the fittest part
of them as the new population.There is no multithreading , nor elitism. Even
though the populations seem to be generated and sorted as expected, I
couldnt figure out expected success rates. Anyone that might want to have a
look, I attached the two breeders in this message. Thanks v. much.
Best Regards,
Marios.
|
|
|