Artificial Neural Networks (ANN) is information processing systems like human brain, which can learn from observations and able to generalize by abstraction. A well trained neural network is able to relate any arbitrary nonlinear input-output from the corresponding data. It has resulted in their use in areas such as pattern recognition, speech processing, control and biomedical engineering. Recently, ANNs have been applied to RF and microwave Computer-Aided Design (CAD) problems to obtain models for components, as designed by Christodoulou and Michael (2001). Electromagnetic (EM) simulation technique for high frequency structures has helped to bring models for components designing. The key contribution of EM simulation technique is described by Suntives et al. (2001) in the domain of accurate models for microwave components. Various authors have applied advanced structures and algorithm like multi-layer perceptron feed forward network, wavelet neural network, and back propagation with accurate results in their research work. A time domain-based optimized ANN and Support Vector Regression (SVR) models have been used by Liya et al. (2015) for the prediction of real-world power within the GSM 900, Very High Frequency (VHF), Ultra High Frequency (UHF), FM and TV Bands. Other than feed forward network, Recurrent Neural Network (RNN) has been applied by Christodoulou and Michael (2001) to learn distributed representation of users and product. Multilayer feed forward network has been used by Patnaik et al. (2005) along with back propagation to reach better accuracy in RF and microwave component design.
|
