Modelling nonlinear relationships in ecology and biology using neural networks

Zarita Zainuddin, . and Saratha Sathasivam, . (2003) Modelling nonlinear relationships in ecology and biology using neural networks. [Proceedings Paper]

Artificial neural networks ANNs are computational systems whose internal structure and processing methods attempt to imitate some of the known features of biological nervous systems. As such they have properties and capabilities quite different from those of traditional serial algorithms processed by serial computers. Indeed the recent surge of interest in ANNs has come about because they can solve problems that are intractable with traditional serial methods. The capabilities of the multilayer perceptron MLP trained with the backpropagation algorithm stems from the nonlinearities used within the nodes. The capabilities of the MLP can be viewed from two different perspectives. Firstly is its ability to partition the pattern space for classification problems and secondly it can approximate to arbitrary accuracy almost any reasonable function Cybenko 1989 Hornik Stinchcombe and White 1989 Leshno et al. 1993. Neural network technology has been succesfully applied to a wide range of real-world applications of considerable complexity. This paper deals with neural network applications to ecological and biological problems. Two neural network applications Classification of Iris Plant and Gender Classification of Crabs exemplify different application aspects. To improve the performance of the neural networks several optimization methods are advocated.

Full text available from:

Abstract

Artificial neural networks ANNs are computational systems whose internal structure and processing methods attempt to imitate some of the known features of biological nervous systems. As such they have properties and capabilities quite different from those of traditional serial algorithms processed by serial computers. Indeed the recent surge of interest in ANNs has come about because they can solve problems that are intractable with traditional serial methods. The capabilities of the multilayer perceptron MLP trained with the backpropagation algorithm stems from the nonlinearities used within the nodes. The capabilities of the MLP can be viewed from two different perspectives. Firstly is its ability to partition the pattern space for classification problems and secondly it can approximate to arbitrary accuracy almost any reasonable function Cybenko 1989 Hornik Stinchcombe and White 1989 Leshno et al. 1993. Neural network technology has been succesfully applied to a wide range of real-world applications of considerable complexity. This paper deals with neural network applications to ecological and biological problems. Two neural network applications Classification of Iris Plant and Gender Classification of Crabs exemplify different application aspects. To improve the performance of the neural networks several optimization methods are advocated.

[error in script]

Item Type:	Proceedings Paper
AGROVOC Term:	COMPUTER APPLICATIONS
AGROVOC Term:	MODELS
AGROVOC Term:	METHODS
AGROVOC Term:	ECOLOGY
AGROVOC Term:	BIOLOGY
AGROVOC Term:	NERVOUS SYSTEM
AGROVOC Term:	DATA ANALYSIS
AGROVOC Term:	MALAYSIA
Geographical Term:	MALAYSIA
Depositing User:	Ms. Norfaezah Khomsan
Last Modified:	24 Apr 2025 05:27
URI:	http://webagris.upm.edu.my/id/eprint/16074

Actions (login required)

View Item