Parameter Optimization - Gene Expression Programming - page 315

+?/-???????340131-[16]- {1.5076, -0.039703, 1.36438, 1.80936, 1.33295}
-+++-??????302410-[17]- {1.83539, -0.208893, -0.917511, -0.502533, -1.71542}
+-/-*??????340131-[18]- {1.5076, -0.039703, 1.36438, 1.03183, -1.09851}
+-/-*??????340131-[19]- {-0.747101, -0.039703, -0.747101, 1.03183, -1.09851}
Outputs:
[ 0] = f(0.64897) = 1.64863
[ 1] = f(1.15298) = -0.147927
[ 2] = f(3.53224) = --
[ 3] = f(-0.152923) = 0.847722
[ 4] = f(-2.27484) = --
[ 5] = f(0.00946834) = 1.00278
[ 6] = f(1.60726) = 1.3634
[ 7] = f(0.282432) = 1.14809
[ 8] = f(0.156947) = 0.846776
[ 9] = f(1.06641) = 1.92784
[10] = f(-136.756) = --
[11] = f(-1.24176) = --
[12] = f(1.11759) = 0.413278
[13] = f(-0.561012) = 0.472227
[14] = f(-2.92636) = --
[15] = f(-1.62902) = --
[16] = f(2.97017) = --
[17] = f(-6.01007) = --
[18] = f(0.64897) = 1.64863
[19] = f(-1.60573) = --
The expression of the best individual of this generation (chromosome 9)
is shown in Figure 8.5. It encodes the parameter value x 0 = 1.06641, which
corresponds to the output f ( x 0 ) = 1.92784, thus a higher point in the fitness
landscape than the best of the previous generation. Note that this individual
is not related to the best of the previous generation, as they use a completely
different set of random constants and show little homology in their sequences.
Note again that almost all the constants are used to design this new param-
eter value and that constant c 1 is used twice.
In the next generation a new individual was created, chromosome 16, con-
siderably better than the best individual of the previous generation:
GENERATION N: 2
Structures:
-/?/+??????210143-[ 0]- {-0.892914, 0.250061, 0.227905, -0.816986, -0.466431}
+-/**??????340131-[ 1]- {1.5076, -1.20621, 1.36438, 1.03183, -1.09851}
/?*++??????021334-[ 2]- {-0.041564, 1.73029, -0.513245, 1.42899, -1.7045}
//+*+??????313334-[ 3]- {1.42899, 1.11624, -0.489014, -1.64124, -0.513245}
*?*????????103140-[ 4]- {-1.08029, 0.604004, -0.400848, -1.36026, -0.963868}