Geoscience Reference
In-Depth Information
Arithmetic program
Parse tree format
+
.
-
+
2
∏
x
+
3
y
5
1
Equivalent expression
2 ∏+((
x
+3)-(
y
/(5+1))
Logic program
Parse tree format
^
true
x
y
^
x
z
x
y
Equivalent expression
(
x
true) ((
x
y
) (
z
(
x
y
)))
Coding program
Parse tree format
;
=
while
<
=
1
i
+
i
20
i
Equivalent expression
i
1
i
=1;
while (
i
< 20)
{
FIGURE 8.3
Equivalent formats for different types of program: arithmetic, logic and coding.
There are four key decisions that need to be made by a user prior to undertaking GP modelling:
(1) selection of independent variables, (2) determination of a suitable function set, (3) selection of
software settings and (4) selection of a stopping point, also called the termination criterion. Choosing
which independent variables to model should be based on an understanding of the problem, gained
during the site selection and data preparation stage of the experimental design process, but could
equally be somewhat arbitrary. The functions used may be mathematical or trigonometric (e.g. +, −, *,
/, sqrt, sine, cosine) or Boolean (e.g. IF, NOT, AND). Software settings pertain to chromosome dimen-
sions, population size, probability of using a particular genetic operator, constant creation, measures
of fitness/error, stopping point, data pre-processing operations, etc. With the exception of the inde-
pendent and dependent variables, parameters may be left to default software settings, although it is
strongly recommended that modelling includes a simple sensitivity analysis to optimise such controls
by varying population size or increasing or decreasing the probability of a particular genetic operator
