Geoscience Reference
In-Depth Information
Calculate the
objective values
of the
individuals
Assess the
fitness of each
individual
Initialise
population
Creation of new
individuals via
crossover and
mutation
Selection of
the fittest
individuals for
reproduction
(a)
(b)
FIGURE 7.1
(a) Basic operation of EAs. (b) Pseudocode representing the basic process.
operands and internal nodes are branches. Programs are then bred by branches of the parse
trees being swapped (and often implemented through the Lisp programming language)
(Koza 1992). See Beriro et al. (2014) for further applications of GP in geography and GC.
Classifier systems
: This approach combines many methods of adaption into a single form that
possesses different types of learning and evolution (Urbanowicz and Moore 2009).
Gene expression programming
(
GEP
): This approach is used to learn about the relation-
ships between variables in data sets and builds models to try to explain these relationships
(Ferreira 2001; Zhou et al. 2003; Beriro et al. 2013).
Computational evolution
: Banzhaf et al. (2006) proposed a new field called computational
evolution in 2006, which could solve more complex problems including more open-ended
ones than current EAs. These approaches would draw upon more complex concepts in bio-
logical evolution across a range of scales from molecular to ecosystem incorporating ideas
such as feedback, self-regulation and tuning. Although still largely in the concept stage,
these approaches may benefit geographical problems in the future.
7.4
BUILDING BLOCKS OF EAs
7.4.1 i
nitial
P
oPulationS
The first stage in the operation of an EA is the initialisation or creation of a population. This ini-
tial population is simply a pool of potential solutions that have been generated either randomly or
