Last change
on this file since 6934 was
6152,
checked in by bfarka, 14 years ago
|
added ecj and custom statistics to communicate with the okb services #1441
|
File size:
1.3 KB
|
Line | |
---|
1 | This package implements the (mu+lambda) and (mu,lambda) evolution strategies. |
---|
2 | Use this package in conjunction with the 'simple' package (that is, |
---|
3 | generational evolution). The provided es.params file shows some more |
---|
4 | information on how to do this. |
---|
5 | |
---|
6 | These techniques are implemented with custom Breeders in combination with a |
---|
7 | special SelectionMethod which handles the dirty work of truncation selection. |
---|
8 | The SelectionMethod is: |
---|
9 | |
---|
10 | ec.es.ESSelection |
---|
11 | |
---|
12 | This method should only be used with the Breeders in this package. You can |
---|
13 | place ESSelection anywhere in your pipeline, including multiple times if you |
---|
14 | like. It will return the mu best individuals in the population, one at a time, |
---|
15 | each time it's pulsed. Individual ESSelection methods in various parts of |
---|
16 | your pipeline will return the same individual. |
---|
17 | |
---|
18 | The (mu,lambda) evolution strategy is implemented with the Breeder |
---|
19 | |
---|
20 | ec.es.MuCommaLambdaBreeder |
---|
21 | |
---|
22 | This breeder requires you to specify a 'mu' value and a 'lambda' value for |
---|
23 | each subpopulation. You need to provide a 'lambda' value instead of just |
---|
24 | the subpopulation size because the subpopulation size is used solely to set the |
---|
25 | INITIAL size of the subpopulation. |
---|
26 | |
---|
27 | The (mu+lambda) evolution strategy has exactly the same options, and is |
---|
28 | implemented in a subclass: |
---|
29 | |
---|
30 | ec.es.MuPlusLambdaBreeder |
---|
Note: See
TracBrowser
for help on using the repository browser.