Information Technology Reference
In-Depth Information
ends only after the linking of the sub-ETs by a particular linking function. And
therefore the “organism” in this case will consist of a multi-subunit expres-
sion tree composed of three smaller subunits.
For problems with multiple outputs, however, the different sub-ETs en-
coded in the genome are engaged in the identification of just one kind of
output and, therefore, they are not physically connected to one another: they
remain more or less autonomous agents working together to solve the prob-
lem at hand. For instance, in classification problems a particular sub-ET is
responsible for the identification of a particular class.
Consider, for instance, the chromosome below composed of three differ-
ent genes created to solve a classification task with three distinct classes:
012345678901201234567890120123456789012
-/dac/
dacaccd
//-aac
bbbabcd
-d/+c*
dbdbacd
(2.12)
It codes for three different sub-ETs, each one representing a rather complex
algebraic expression (Figure 2.4).
a.
012345678901201234567890120123456789012
-/dac/
dacaccd
//-aac
bbbabcd
-d/+c*
dbdbacd
b.
Sub-ET
1
Sub-ET
2
Sub-ET
3
/
/
d
d
a
c
a
a
c
c
b
d
b
d
Figure 2.4.
Translation of GEP genes as algebraic sub-ETs.
a)
A three-genic
chromosome with the tails shown in bold.
b)
The sub-ETs codified by each gene.
Note that, after translation, the sub-ETs might either form multi-subunit expression
trees composed of smaller sub-ETs or remain isolated as a single-unit expression
tree. For instance, in problems with just one output, the sub-ETs might be linked by
a particular linking function or, in problems with multiple outputs, each sub-ET is
responsible for the identification of a particular output.
