I used a small test today, huh, huh, it's not bad, summed up the results:
Test functions: ("Evolutionary Program - Genetic Algorithm and Data Coding" Science Press, 2000, P26)
F (x1, x2) = 21.5 x1 * sin (4 * pi * x1) x2 * sin (20 * pi * x2) maximum
Among them - 3.0 <= x1 <= 12.1 and 4.1 <= x2 <= 5.8, the graphic of the function is in [Photo Album - Resources] (waiting for the map, possession, huh, huh)
Now I have 5 selection operators, of which 1 and 5 are the same, but the implementation method is different (you can look at it, just tested it once, huh, huh)
These 5 (4) operators contain the best preservation, they are
1, the turntable selection operator (the most common, that is, the proportional operator): Selection1, Selection5
2, with a ranked turntable selection operator: Selection2
3, random consistency operator: Selection3
4, championship selection operator: Selection4
Now I have two algorithm performance assessment indicators ("Evolution Computing" Pan Zhengjun, etc., Tsinghua University Press, 1998, P66)
One is the average performance value (online performance), I don't say anything: AVE
One is the optimal performance value (offline performance), huh, huh, huh: max
Below this form is the test results, knocking numbers are really not good :)
Selection1
SELECTION2
SELECTION3
SELECTION4
SELECTION5
AVE
38.8503
38.3226
38.4503
38.1489
38.8503
33.5902
22.7836
33.6016
36.5246
34.1255
11.6255
11.6333
11.6255
11.1268
11.6256
5.725
5.6309
5.325
5.5251
5.725
Max
38.8474
36.3799
38.8502
38.8503
38.8493
38.1199
36.259
38.1921
38.5717
38.6797
11.6257
11.1469
11.6254
11.6255
11.6255
5.7256
4.6323
5.7251
5.725
5.7253
Note: Parameters used above:
POPSIZE = 80, the number of individuals in the group
Gen = 300, algebraic
MINVAR = [- 3.0 4.1], because this is a two-variable function, the minvar is two, represents the minimum of X1, X2, respectively, and this program can handle multiple variables.
Maxvar = [12.1 5.8], in the same, but only means the maximum value, so that each variable has a range
Scalevar = 0.0001, solving accuracy, in the table, the decimal point is the fourth point in the fourth place is 0
PC = 0.25, crossover probability
PM = 0.01, variation probability
Explain: From the up to the four digits represent: the top one is the optimal result found by the 300th generation, the second is the AVE (or Max), the third and the first and first The four sort-oriented values corresponding to the optimal results are obtained, and there are two.
Also, AVE's large-native expression algorithm convergence, while MAX represents an algorithm's trend (ON-Going "
Appendix: The figure obtained by each result:
AVE1:
Max1
AVE2
Max2
AVE3
MAX3
AVE4
Max4
AVE5
MAX5
Forget it, I will not map, everyone go to the photo album - the resources, 嘿嘿
