The development of new learning algorithms for design of the neural networks is a potential research area. The design of neural network is based on selection of the optimal parameters to achieve higher learning speed and accuracy during recall. These design parameters are weights, no. of hidden layers and no. of neurons. With this motivation, this paper presents an application of Taguchi method to meet high accuracy requirements. Three design factors are considered as control factors which are namely input representation scheme, no. of hidden layers and no. of input nodes to design the neural network. This design has been tested over three benchmark optimization functions. It is observed that the proposed network shows better accuracy.
Artificial Neural Network (ANN), Design of Experiment (DOE), Feed Forward Neural Networks (FFNN), Signal to Noise Ratios (SNRs), Taguchi Method