250
C. Karul . S. Soyupak
emphasize the importance of aquatic macrophytes. The final outcome of their
study was that the nitrogen was limiting nutrient for hypertrophie lakes.
Deterministic models (Jorgensen 1976; Benndorf and Recknagel 1982), time
series analysis models (Whitehead and Hornberger 1984), and fuzzy-Iogic models
(Recknagel et al. 1994) are the other major classes of models for the same
purpose.
Since the empirical models utilizes limited number of parameters to predict
chlorophyll-a concentrations, it is natural to conclude that they provide rough
estimates with low degree of precision with serious oversimplifications. The
processes that lead to eutrophication of water bodies are known to involve
extremely complex behaviors with nonlinear relations between system parameters
and the system responses that can not be readily explained with simplistic
approaches. Starting from this idea, the authors of this manuscript thought that
developing artificial neural network tools that are adequately trained with several
environmental factors could be a better approach for more precise predictions. The
characteristics of artificial network methodology allow learning complex systems
and predicting their responses with high degree of precision if adequately applied.
13.1.2
Artificial Neural Networks
The use of artificial neural networks in solving complex problems is becoming
popular in many disciplines due to their capability to 'learn' non-linear relations.
The method of artificial neural network has been inspired by biologie al nervous
system. Neural Network, in computer science, is highly interconnected network of
information-processing elements that mimics the connectivity and functioning of
the human brain. One of the most significant superiority of artificial neural
networks is their ability to learn from a limited set of examples. Artificial neural
networks are being created mimicking the structure and functioning of biological
neural networks, in an artificial way.
A properly trained and verified artificial neural network for the specific
problem of interest recognizes the data and makes predictions with desirable
accuracy. The problem of interest can be non-linear in nature and it can be at any
degree of complexity. Neural networks are composed of simple neuron-like
operating elements (neurons) and weighted connections between these elements.
The network function is determined largely by the connections between neurons.
A neural network can be trained to perform a particular job by adjusting the values
of the connections (weights) between neurons. The algorithmic approach es for
developing an artificial neural network model for a specific problem exist in
literature (Pu 1994; Mathworks 1998).
C. Karul . S. Soyupak
emphasize the importance of aquatic macrophytes. The final outcome of their
study was that the nitrogen was limiting nutrient for hypertrophie lakes.
Deterministic models (Jorgensen 1976; Benndorf and Recknagel 1982), time
series analysis models (Whitehead and Hornberger 1984), and fuzzy-Iogic models
(Recknagel et al. 1994) are the other major classes of models for the same
purpose.
Since the empirical models utilizes limited number of parameters to predict
chlorophyll-a concentrations, it is natural to conclude that they provide rough
estimates with low degree of precision with serious oversimplifications. The
processes that lead to eutrophication of water bodies are known to involve
extremely complex behaviors with nonlinear relations between system parameters
and the system responses that can not be readily explained with simplistic
approaches. Starting from this idea, the authors of this manuscript thought that
developing artificial neural network tools that are adequately trained with several
environmental factors could be a better approach for more precise predictions. The
characteristics of artificial network methodology allow learning complex systems
and predicting their responses with high degree of precision if adequately applied.
13.1.2
Artificial Neural Networks
The use of artificial neural networks in solving complex problems is becoming
popular in many disciplines due to their capability to 'learn' non-linear relations.
The method of artificial neural network has been inspired by biologie al nervous
system. Neural Network, in computer science, is highly interconnected network of
information-processing elements that mimics the connectivity and functioning of
the human brain. One of the most significant superiority of artificial neural
networks is their ability to learn from a limited set of examples. Artificial neural
networks are being created mimicking the structure and functioning of biological
neural networks, in an artificial way.
A properly trained and verified artificial neural network for the specific
problem of interest recognizes the data and makes predictions with desirable
accuracy. The problem of interest can be non-linear in nature and it can be at any
degree of complexity. Neural networks are composed of simple neuron-like
operating elements (neurons) and weighted connections between these elements.
The network function is determined largely by the connections between neurons.
A neural network can be trained to perform a particular job by adjusting the values
of the connections (weights) between neurons. The algorithmic approach es for
developing an artificial neural network model for a specific problem exist in
literature (Pu 1994; Mathworks 1998).