The system discovered a grammar that accepts all the strings in the testbed. The variation of the best fitness value with respect to the count of generation is displayed in Figure 4. It is an interesting observation that the DFA that is used to generate the testbed, when submitted to the fitness function, evaluated to a lower fitness value (2.75) compare to the discovered one (2.82). This was due to the fact that the DFA discovered by the GA was in terms of the fitness criteria a `simpler' one (was consisting of less transitions). This `discovered' DFA is pictured in Figure 3.
Figure 4: Evolution of the Fitness of the Best
Reducing the penalty for loop construction and hence allowing GA to construct
loopes of minimal number produced an even more compact DFA,
which is shown in Figure 5 (with the fitness value 2.91)