Case Studies for Applications of Elman Recurrent Neural Networks

Artificial neural networks (ANNs) are computational modeling tools that have recently emerged and found extensive acceptance in many disciplines for modeling complex realworld problems. ANN-based models are empirical in nature, however they can provide practically accurate solutions for precisely or imprecisely formulated problems and for phenomena that are only understood through experimental data and field observations. ANNs produce complicated nonlinear models relating the inputs (the independent variables of a system) to the outputs (the dependent predictive variables). ANNs have been widely used for various tasks, such as pattern classification, time series prediction, nonlinear control, and function approximation. ANNs are desirable because (i) nonlinearity allows better fit to the data, (ii) noise-insensitivity provides accurate prediction in the presence of uncertain data and measurement errors, (iii) high parallelism implies fast processing and hardware failure-tolerance, (iv) learning and adaptivity allow the system to modify its internal structure in response to changing environment, and (v) generalization enables application of the model to unlearned data (Fausett, 1994; Haykin, 1994; Hassoun, 1995). The idea of using ANNs for pattern classification purposes has encountered, for a long time, the favour of many researchers (Miller et al., 1992; Wright et al., 1997; Wright & Gough, 1999; Saxena et al., 2002; Übeyli, 2007a; 2007b; 2008a; 2008b; 2008c). Feedforward neural networks are a basic type of neural networks capable of approximating generic classes of functions, including continuous and integrable ones. One of the most frequently used feedforward neural network for pattern classification is the multilayer perceptron neural network (MLPNN) which is trained to produce a spatial output pattern in response to an input spatial pattern (Fausett, 1994; Haykin, 1994; Hassoun, 1995). The mapping performed is static, therefore, the network is inherently not suitable for processing temporal patterns. Attempts have been made to use the MLPNN to classify temporal patterns by transforming the temporal domain into a spatial domain. An alternate neural network approach is to use recurrent neural networks (RNNs) which have memory to encode past history. Several forms of RNNs have been proposed and they may be classified as partially recurrent or fully recurrent networks (Saad et al., 1998; Gupta O pe n A cc es s D at ab as e w w w .ite ch on lin e. co m