Chapter 4 . Applications of Evolutionary Computation
59
Prec:llcted ChJorophyll .. for 1886 and 1993
-
Qbserved
-
DItferenoe EquaUon
Evolved Dlfference Equatlon
100
O ~------------------------~~-------------------------1981
y ....
1"3
Figure 4.3. Evolved difference equation versus original parameter settings for
the two unseen test years: 1986 and 1993.
4.4.3
Evolving Differential Equations
A common approach in ecology is to produce a set of differential equations to
represent the relationships between variables in a system. Previous work has
shown that searching for suitable differential equations based on ecological data
are possible (Todorovski et al. 1998). In this work, Lagramge, an equation
discovery system, was used to define the space of possible model structures and to
auto mate the modelling of phytoplankton growth. Although this approach did not
use an evolutionary system to perform the search (the search was a systematic
breadth-first search of the possible equations) there is clearly an opportunity to
extend the work using evolution to examine a larger search space of possible
equations. Work that demonstrates the use of evolving differential equations is
that of Sakamoto and Iba (Sakamoto and Iba 2001). In this work, equations
representing a number of coupled differential equations were evolved using GP,
however this approach has not currently been applied to ecological data.
59
Prec:llcted ChJorophyll .. for 1886 and 1993
-
Qbserved
-
DItferenoe EquaUon
Evolved Dlfference Equatlon
100
O ~------------------------~~-------------------------1981
y ....
1"3
Figure 4.3. Evolved difference equation versus original parameter settings for
the two unseen test years: 1986 and 1993.
4.4.3
Evolving Differential Equations
A common approach in ecology is to produce a set of differential equations to
represent the relationships between variables in a system. Previous work has
shown that searching for suitable differential equations based on ecological data
are possible (Todorovski et al. 1998). In this work, Lagramge, an equation
discovery system, was used to define the space of possible model structures and to
auto mate the modelling of phytoplankton growth. Although this approach did not
use an evolutionary system to perform the search (the search was a systematic
breadth-first search of the possible equations) there is clearly an opportunity to
extend the work using evolution to examine a larger search space of possible
equations. Work that demonstrates the use of evolving differential equations is
that of Sakamoto and Iba (Sakamoto and Iba 2001). In this work, equations
representing a number of coupled differential equations were evolved using GP,
however this approach has not currently been applied to ecological data.