52
P.A. Whigham . G.B. Fogel
models. Metapopulation models, based on spatial distributions, have also been
studied, where the local models are described using a differential or difference
equation that incorporate a colonization term that indicates how local populations
spread (Levins and Culver 1971; Carter and Prince 1981). These models differ
from CA in that the spatial terms are explicitly represented in the equations, rather
then an explicit model of spatial distribution being used to give spatial extension.
The difficulty with cellular models is that the production of realistic local rules
that give rise to appropriate global behaviour is a complex task with few
guidelines. Spatial structure has also been studied in relation to island
biogeography and the concept of gene flow between metapopulations (McCauley
1995), however all of these approaches are complex in nature and difficult to
formulate and test.
4.2.2
Summary
The previous discussion presented some of the basic challenges involved in
modelling ecological concepts. There are several salient points: models based on
differential and difference equations have constants that require tuning based on
measured data, and most models are simplified in order to be solved, either by
assumptions that produce linear models, or by removing the complexities of the
system. Additionally, models that incorporate spatial and temporal information,
and those based on populations of individuals, are difficult to formulate or
express. In the following sections, applications of evolutionary algorithms to
ecological modelling will be reviewed. These applications are both extensions to
traditional modelling efforts and wholly new approaches to ecological
informatics.
4.3
Evolutionary Computation
Evolutionary computation (EC) is an area of computer and information science
that uses principals from biological evolution to solve computational problems.
The concept of simulating evolution on a computer has a long history. Early
efforts focused on learning machines (Friedberg 1958; Friedberg 1959; Fogel
1962; Fogel et al. 1966), evolutionary systems dynamics (Barricelli 1954; Conrad
and Pattee 1970), engineering applications (Rechenberg 1965; Schwefel 1965),
and genetics (Fraser 1957; Bremermann 1962; Fraser 1962; Bremermann et al.
1966; Bagley 1967; Rosenberg 1967; Holland 1969; Holland 1973). This history
is reviewed in Fogel (Fogel 1998; Fogel 2000). Since natural evolution has
successfully created complex systems, and discovered novel solutions to difficult
P.A. Whigham . G.B. Fogel
models. Metapopulation models, based on spatial distributions, have also been
studied, where the local models are described using a differential or difference
equation that incorporate a colonization term that indicates how local populations
spread (Levins and Culver 1971; Carter and Prince 1981). These models differ
from CA in that the spatial terms are explicitly represented in the equations, rather
then an explicit model of spatial distribution being used to give spatial extension.
The difficulty with cellular models is that the production of realistic local rules
that give rise to appropriate global behaviour is a complex task with few
guidelines. Spatial structure has also been studied in relation to island
biogeography and the concept of gene flow between metapopulations (McCauley
1995), however all of these approaches are complex in nature and difficult to
formulate and test.
4.2.2
Summary
The previous discussion presented some of the basic challenges involved in
modelling ecological concepts. There are several salient points: models based on
differential and difference equations have constants that require tuning based on
measured data, and most models are simplified in order to be solved, either by
assumptions that produce linear models, or by removing the complexities of the
system. Additionally, models that incorporate spatial and temporal information,
and those based on populations of individuals, are difficult to formulate or
express. In the following sections, applications of evolutionary algorithms to
ecological modelling will be reviewed. These applications are both extensions to
traditional modelling efforts and wholly new approaches to ecological
informatics.
4.3
Evolutionary Computation
Evolutionary computation (EC) is an area of computer and information science
that uses principals from biological evolution to solve computational problems.
The concept of simulating evolution on a computer has a long history. Early
efforts focused on learning machines (Friedberg 1958; Friedberg 1959; Fogel
1962; Fogel et al. 1966), evolutionary systems dynamics (Barricelli 1954; Conrad
and Pattee 1970), engineering applications (Rechenberg 1965; Schwefel 1965),
and genetics (Fraser 1957; Bremermann 1962; Fraser 1962; Bremermann et al.
1966; Bagley 1967; Rosenberg 1967; Holland 1969; Holland 1973). This history
is reviewed in Fogel (Fogel 1998; Fogel 2000). Since natural evolution has
successfully created complex systems, and discovered novel solutions to difficult