Geoscience Reference
In-Depth Information
Evolve initial
population of
chromosomes
New generation
of
chromosomes
Adaptions
using genetic
operators
Execute each
program
Evaluate fitness
of individuals
Select cohort of
individuals
Termination
criteria satisfied
NO
Perfect fitness
NO
YES
YES
Key
Decision
Action
STOP
End of process
FIGURE 8.2
Flowchart of GP process.
in a GA, which are linear and binary representations. GP chromosomes are coded and processed in a
language that is specific to the software that each program is written in, for example, be that Fortran
(Diplock, 2000), Karva (Ferreira, 2001) or LISP (Koza, 1990). Structurally, chromosomes comprise
a mix of (1) independent variables; (2) mathematical; and/or (3) logic functions and/or numerical
constants. Each item is schematically represented in a tree-like format, called either a parse tree (PT)
or an expression tree (ET). These have a hierarchical structure which is ultimately processed by the
software as lines of computer code. A diagram of some different types and forms of chromosome
is provided in Figure 8.3. GP chromosomes are usually singular entities. However, in GEP, they are
more commonly evolved and expressed in groups of two or more, forming a multi-genetic solution, in
which each sub-expression tree (Sub-ET) is able to code for a particular facet of the problem and in
which each individual component is subsequently joined together by means of a simple linking func-
tion (+, −, *, \) (Ferreira, 2001), for example,
V
= ET, in which ET = Sub-ET1 + Sub-ET2 + Sub-ET3.
