Information Technology Reference
In-Depth Information
20
SA-INV 5-1
SA-INV 10-2
SA-INV 20-5
18
16
14
12
10
8
6
4
2
0
0
10
20
30
40
50
60
70
80
90
100
generations
4.4e+006
SA-INV 5-1
SA-INV 10-2
INV 1
INV 10
4.2e+006
4e+006
3.8e+006
3.6e+006
3.4e+006
3.2e+006
3e+006
2.8e+006
0
10
20
30
40
50
60
70
80
90
100
generations
Fig. 5.5. SA-INV on problem gr666 . Upper part: Average development of k over 25
runs. The mutation strength k is spread more broadly than in the previous experiments.
Lower part: Average fitness development of SA-INV and INV over 25 runs. SA-INV 5 1
and SA-INV 10 2 are the fastest algorithms, at least until t = 100. INV 10 suffers from
premature fitness stagnation.
of the instance. There are 666! possible tours and the number of inversions in
each generation until successful solutions are produced might be spread broadly
around a certain value. Here it is noticeable that the comma selection allows
deteriorations. Furthermore, we can observe an increase of k at the beginning of
the search, slower than in the previous experiments, but appearing for all three
SA-INV variants. Again, we observe the premature fitness stagnation of INV 10
and the superiority of SA-INV 5 1 and SA-INV 10 2 , at least until generation
t = 100.
 
Search WWH ::




Custom Search