Information Technology Reference
In-Depth Information
alternative modeling tool but also to shed light on the behavior of intricately
connected networks of neurons.
10.4 Evolutionary Dynamics of GEP-nets
The neural networks described in this chapter are perhaps the most complex
entities created with gene expression programming. Therefore, it would be
interesting to see if the evolutionary dynamics of these systems exhibit the
same kind of pattern observed in other less complex systems.
As shown in Figure 10.9, GEP-NN systems exhibit the same kind of dy-
namics found on other, less complex GEP systems. The particular dynamics
shown in Figure 10.9 was obtained for a successful run of the experiment
summarized in the second column of Table 10.3. Note the characteristic os-
cillatory pattern on average fitness and that the best fitness is considerably
above average fitness.
The ubiquity of these dynamics suggests that, most probably, all healthy
genotype/phenotype evolutionary systems are ruled by them.
64
56
Best Ind
Avg fitness
48
40
32
24
16
8
0
0
250
500
750
1000
1250
1500
1750
2000
Generations
Figure 10.9.
Evolutionary dynamics found in complex GEP systems, specifically,
on run 4 of the experiment summarized in the second column of Table 10.3.
