Optimization of Neural Networks for Grain Classification

This paper discusses the optimization of back propagaton neural networks for the grain texture feature, extraction in grain classification. During classification, a multi layer feed forward neural network after optimization and another without optimization are used to map the non linearity between inputs and outputs of data respectively. The optimized back propagation neural network is compared with the basic form of back propagation neural network as well as their outputs. The differences between these results of grain classification from the two neural networks are discussed. Furthermore, the efficiency and the performance of the applied optimization methods are also evaluated.