Information Technology Reference
In-Depth Information
double ADF4(double d[])
{
double dblTemp = 0.0;
dblTemp = (((exp((14-18))*(d[0]*(d[2]-14)))+75)-
(d[2]-14));
return dblTemp;
}
double ADF5(double d[])
{
double dblTemp = 0.0;
dblTemp = sqrt(88);
return dblTemp;
}
double apsModel(double d[])
{
double dblTemp = 0.0;
dblTemp = ((ADF4(d)+(ADF4(d)/(ADF3(d)+
((ADF1(d)*ADF2(d))/ADF3(d)))))-ADF1(d));
return dblTemp;
}
(6.7b)
Note how profusely the cellular system with the facility for handling random
numerical constants uses the RNCs at its disposal. Note also that structurally
this program is very different from the model (6.6) created by the cellular
system without random numerical constants. Indeed, as observed for the acel-
lular systems, the models built with RNCs are usually much more compact
and are also usually less varied in terms of the function nodes used in their
construction.
6.5 Diagnosis of Breast Cancer
In this section we are going to solve a complex, real-world classification
problem with the cellular system, both with and without random numerical
constants, in order to evaluate the evolvability of these complex learning
systems. Again, we are going to use settings very similar to the ones used in
section 5.6.4 in order to compare the complex cellular systems with the much
simpler acellular systems. Both the performance and parameters used per
run are shown in Table 6.7.
And as you can see, despite their higher complexity, the cellular systems
perform quite well at this difficult task and, as expected, the cellular system
