Self-Adaptive Inversion Mutation - Self-Adaptive Heuristics for Evolutionary Computation - page 82

25

SA-INV 5-1
SA-INV 10-2
SA-INV 20-5

20

15

10

5

0

0 10 20 30 40 50 60 70 80 90 100

generations

28000

SA-INV 5-1
SA-INV 10-2
INV 1
INV 10

26000

24000

22000

20000

18000

16000

14000

12000

10000

8000

0 10 20 30 40 50 60 70 80 90 100

generations

Fig. 5.3. SA-INV on problem berlin52 . Upper part: Development of SA-INV strategy
parameter k , averaged over 25 runs. Starting from different values for k , the runs show
a similar development after 20 generations. Lower part: Fitness development of SA-INV
and INV averaged over 25 runs. SA-INV 5 − 1 is the fastest variant in the first generations
until INV 1 overtakes its approximation.

mutation strength and with self-adaptation, see figure 3.5. In the first 10 gener-
ations INV 10 develops similarly to SA-INV 10 − 2 . But then SA-INV 10 − 2 reduces
k< 10. Figure 5.3, upper part, shows the development of k . The self-adaptive
processisabletocontrolthe mutation strength k independently of the initial
values ν =5 , 10 , 20: after generation 20 all developments proceed similarly to
each other. Parameter k almost constantly decreases to 1.

5.4.2 TSP Instance Bier127

Problem bier127 is an Euclidean distance problem of 127 beer gardens in Augs-
burg. Table 5.2 presents the comparison of the SA-INV and INV variants after