Chapter 3 . Applications of Genetic Aigorithms
45
evolutionary capabilities needed to develop control strategies and complex
adaptive systems.
While each of these techniques will continue to be used very successfully,
hybrids using the best attributes of each technique will have the potential to offer
revolutionaryadvances. Work done by Patel et al. (1998) is an intriguing example
of a hybrid neural net and genetic algorithm. Their objective was to design new
molecules that would kill bacteria. They trained a neural network on a dataset of
chemicals and their ability to kill bacteria. They then used a genetic algorithm to
rearrange the amino acid sequences to create new chemicals. The neural network
then evaluated the efficacy of these new chemicals. This approach took advantage
of the strengths of each technique.
D' Angelo-Morrall et al. (in prep) created a hybrid statistical clustering and GP
model. The objective of this work was to predict the toxicity to aquatic organisms
of a wide variety of chemicals. Because the chemicals in the dataset were very
diverse, a single equation could not adequately predict the toxicity of all
chemicals. They used k-means clustering to group the chemicals into similar
classes. The GP was then used to evolve equations to predict the toxicity of each
groups of chemicals. The performance of the hybrid statisticaVGP was compared
with a GP that did both clustering and predictions and K-means analysis that did
both clustering and predictions. When the dataset was of sufficient size for
predictions (training set n=50; testing set n=10) the hybrid approach outperforms
the individual methods.
Whigham and Recknagel (2001) developed a novel hybrid of a process-model
and GA with the goal of optimizing the model structure. They started with a
standard process model and used the GA to optimize the process-based equations
and, where necessary, evolve new equations. This hybrid model performed better
than either a stand-alone GA or process model. This type of hybrid has great
potential for producing better process models and provides a means by which we
can question the many of the standard process equations that have become
paradigms.
These are just a few examples of hybrid architectures, which are becoming
more and more prevalent. It is expected that in the future, the powerful
evolutionary framework of GAs will be commonly used as an integral component
of hybrids. This transition from independent frameworks to coupled systems is
the next generation in the evolution of artificial intelligence programming
techniques and has the potential to open new frontiers in ecological modelling.
References
Bonnan FR, Likens GE (1979) Pattern and process in a forested ecosystem. SpringerVerlag, New York
Bouskila A, Robinson ME, Roitberg BD, Tenhumberg B (1998) Life-history decisions
under predation risk: importance of game perspective. Evolutionary Ecology, 12: 701715
Camahan BJ, Redfern MS, Nonnan B (2000) Designing safe job rotation schedules using
optimization and heuristic search. Ergonomics, 43: 543-560
45
evolutionary capabilities needed to develop control strategies and complex
adaptive systems.
While each of these techniques will continue to be used very successfully,
hybrids using the best attributes of each technique will have the potential to offer
revolutionaryadvances. Work done by Patel et al. (1998) is an intriguing example
of a hybrid neural net and genetic algorithm. Their objective was to design new
molecules that would kill bacteria. They trained a neural network on a dataset of
chemicals and their ability to kill bacteria. They then used a genetic algorithm to
rearrange the amino acid sequences to create new chemicals. The neural network
then evaluated the efficacy of these new chemicals. This approach took advantage
of the strengths of each technique.
D' Angelo-Morrall et al. (in prep) created a hybrid statistical clustering and GP
model. The objective of this work was to predict the toxicity to aquatic organisms
of a wide variety of chemicals. Because the chemicals in the dataset were very
diverse, a single equation could not adequately predict the toxicity of all
chemicals. They used k-means clustering to group the chemicals into similar
classes. The GP was then used to evolve equations to predict the toxicity of each
groups of chemicals. The performance of the hybrid statisticaVGP was compared
with a GP that did both clustering and predictions and K-means analysis that did
both clustering and predictions. When the dataset was of sufficient size for
predictions (training set n=50; testing set n=10) the hybrid approach outperforms
the individual methods.
Whigham and Recknagel (2001) developed a novel hybrid of a process-model
and GA with the goal of optimizing the model structure. They started with a
standard process model and used the GA to optimize the process-based equations
and, where necessary, evolve new equations. This hybrid model performed better
than either a stand-alone GA or process model. This type of hybrid has great
potential for producing better process models and provides a means by which we
can question the many of the standard process equations that have become
paradigms.
These are just a few examples of hybrid architectures, which are becoming
more and more prevalent. It is expected that in the future, the powerful
evolutionary framework of GAs will be commonly used as an integral component
of hybrids. This transition from independent frameworks to coupled systems is
the next generation in the evolution of artificial intelligence programming
techniques and has the potential to open new frontiers in ecological modelling.
References
Bonnan FR, Likens GE (1979) Pattern and process in a forested ecosystem. SpringerVerlag, New York
Bouskila A, Robinson ME, Roitberg BD, Tenhumberg B (1998) Life-history decisions
under predation risk: importance of game perspective. Evolutionary Ecology, 12: 701715
Camahan BJ, Redfern MS, Nonnan B (2000) Designing safe job rotation schedules using
optimization and heuristic search. Ergonomics, 43: 543-560