36
D. Morrall
medical diagnosis (Pattichis and Schizas 1996) and various game playing
pro grams are examples of c1assifier and control strategies. During the 1990's a
shift towards hybrid approaches (e.g., GA-Neural Net combinations) began to
emerge (Fishman and Barr 1991). Hybrid models capitalize on the GAs ability to
evolve components of the hybrid system such as cellular automata or neural net
node weightings.
Although simulations of evolution in natural systems were initiated in the field
of biology, GAs were not widely used for ecological modelling until the 1990's.
The development of the more complex GA designs that Goldberg (1989) refers to
as genetic based machine learning paved the way for ecological modelling with
genetic algorithms. This paper will explore the state of the art of GAs in the field
of ecology and possibilities for future exploration of ecological systems using
evolutionary programming. Though the focus of this paper is on genetic
algorithms, many of the ideas presented in this paper are applicable to other
machine learning and hybrid approaches that are based on the theory of evolution.
The term evolutionary algorithm is sometimes used as a more general descriptor to
refer to the entire suite of methods that employ the process of evolution through
natural selection and will also be used in this text.
3.2
Ecology and Ecological Modelling
Eco- comes from the Greek "oikos" meaning horne. "Ecology. .. is concerned
with the most complex level of biological integration. It attempts to explain why
organisms live where they do and what physical and biological variables govern
their distribution, numbers, and interactions. Based on an understanding of the
fundamental principles that govern organism distributions, numbers, and
interactions, the future behavior and assemblages of organisms can be predicted.
Induction is defined as arriving at knowledge of the universal from examination of
particulars; to see what is common to a set of similars. Aristotle stated that "It is
by induction that we know uni versals and the primary premises on which
demonstrations are based" (Lloyd 1968). Ecology is largely a field of induction.
The complexity of ecological systems and the vast array of interdisciplinary data
that must be collected to understand an ecological system make the process of
"arriving at knowledge of the uni versals from the particulars" achallenging
endeavor.
To further complicate matters, non-linear relationships and dependencies in
data are common in ecological systems, which are by definition integrative.
Eugene Odum, the father of ecology, made the oft quoted statement that
ecosystems are more than the sum of their parts; inferring the presence of nonlinear interactions which can lead to unexpected emergent properties. Ecological
modelling has, since its inception, attempted to develop ways to mathematically
describe these complex, non-linear ecological systems. Traditional ecological
modelling can be thought of as a top-down or deductive technique that represents
broad ecological principles to produce the detailed patterns observed in nature.
D. Morrall
medical diagnosis (Pattichis and Schizas 1996) and various game playing
pro grams are examples of c1assifier and control strategies. During the 1990's a
shift towards hybrid approaches (e.g., GA-Neural Net combinations) began to
emerge (Fishman and Barr 1991). Hybrid models capitalize on the GAs ability to
evolve components of the hybrid system such as cellular automata or neural net
node weightings.
Although simulations of evolution in natural systems were initiated in the field
of biology, GAs were not widely used for ecological modelling until the 1990's.
The development of the more complex GA designs that Goldberg (1989) refers to
as genetic based machine learning paved the way for ecological modelling with
genetic algorithms. This paper will explore the state of the art of GAs in the field
of ecology and possibilities for future exploration of ecological systems using
evolutionary programming. Though the focus of this paper is on genetic
algorithms, many of the ideas presented in this paper are applicable to other
machine learning and hybrid approaches that are based on the theory of evolution.
The term evolutionary algorithm is sometimes used as a more general descriptor to
refer to the entire suite of methods that employ the process of evolution through
natural selection and will also be used in this text.
3.2
Ecology and Ecological Modelling
Eco- comes from the Greek "oikos" meaning horne. "Ecology. .. is concerned
with the most complex level of biological integration. It attempts to explain why
organisms live where they do and what physical and biological variables govern
their distribution, numbers, and interactions. Based on an understanding of the
fundamental principles that govern organism distributions, numbers, and
interactions, the future behavior and assemblages of organisms can be predicted.
Induction is defined as arriving at knowledge of the universal from examination of
particulars; to see what is common to a set of similars. Aristotle stated that "It is
by induction that we know uni versals and the primary premises on which
demonstrations are based" (Lloyd 1968). Ecology is largely a field of induction.
The complexity of ecological systems and the vast array of interdisciplinary data
that must be collected to understand an ecological system make the process of
"arriving at knowledge of the uni versals from the particulars" achallenging
endeavor.
To further complicate matters, non-linear relationships and dependencies in
data are common in ecological systems, which are by definition integrative.
Eugene Odum, the father of ecology, made the oft quoted statement that
ecosystems are more than the sum of their parts; inferring the presence of nonlinear interactions which can lead to unexpected emergent properties. Ecological
modelling has, since its inception, attempted to develop ways to mathematically
describe these complex, non-linear ecological systems. Traditional ecological
modelling can be thought of as a top-down or deductive technique that represents
broad ecological principles to produce the detailed patterns observed in nature.