Chapter 4 . Applications of Evolutionary Computation
57
general (parsimonious) solution. It is important to realize that the representations
and variations offered above are merely two of many potential representations and
the choice of the most appropriate method is left to the user to determine for each
problem. This is an important point to address for the researcher as different
representations may affect the utility of the evolutionary approach on the search
space in question.
4.3.2
Summary
Evolutionary algorithms use concepts based on biological evolution to evolve
solutions to problems for optimization, simulation and models of individual,
group and community interaction. These approaches allow useful solutions to
nonlinear problems to be generated in real-time and allow complex systems to be
developed and described that are difficult to model with standard mathematical
approaches.
4.4
Ecological Modelling and Evolutionary Aigorithms
Section 2 discussed various goals of ecological modelling and the traditional
frameworks used to understand and represent ecological problems. The following
sections will describe various applications of EAs to develop or supplement
models that have become standard approach es to understanding the patterns in
ecology. Areas of research will also be highlighted that can extend current
ecological theory based on evolutionary techniques.
4.4.1
Equation Discovery
Developing equations to describe the functional relationship between variables in
a system is a key goal to understanding ecological systems. Typically, a
differential or difference equation is created to describe the system, constants of
the equations are tuned and the equations used to model the system. Given a set of
data describing the independent and dependent variables, the goal is to produce an
equation that models this information. Ideally the form of the solution should be
constrained to allow only physically meaningful interpretations to be produced.
One such example of an evolutionary system that allows this to occur is based on
GP (Whigham 1995; Whigham and Crapper 1999), and uses a context-free
grammar to allow bias in the form of evolved solutions. This approach has been
successfully applied to freshwater system modelling, by creating equations that
57
general (parsimonious) solution. It is important to realize that the representations
and variations offered above are merely two of many potential representations and
the choice of the most appropriate method is left to the user to determine for each
problem. This is an important point to address for the researcher as different
representations may affect the utility of the evolutionary approach on the search
space in question.
4.3.2
Summary
Evolutionary algorithms use concepts based on biological evolution to evolve
solutions to problems for optimization, simulation and models of individual,
group and community interaction. These approaches allow useful solutions to
nonlinear problems to be generated in real-time and allow complex systems to be
developed and described that are difficult to model with standard mathematical
approaches.
4.4
Ecological Modelling and Evolutionary Aigorithms
Section 2 discussed various goals of ecological modelling and the traditional
frameworks used to understand and represent ecological problems. The following
sections will describe various applications of EAs to develop or supplement
models that have become standard approach es to understanding the patterns in
ecology. Areas of research will also be highlighted that can extend current
ecological theory based on evolutionary techniques.
4.4.1
Equation Discovery
Developing equations to describe the functional relationship between variables in
a system is a key goal to understanding ecological systems. Typically, a
differential or difference equation is created to describe the system, constants of
the equations are tuned and the equations used to model the system. Given a set of
data describing the independent and dependent variables, the goal is to produce an
equation that models this information. Ideally the form of the solution should be
constrained to allow only physically meaningful interpretations to be produced.
One such example of an evolutionary system that allows this to occur is based on
GP (Whigham 1995; Whigham and Crapper 1999), and uses a context-free
grammar to allow bias in the form of evolved solutions. This approach has been
successfully applied to freshwater system modelling, by creating equations that