Case study : an effect of noise in character recognition system using neural network

There has been resurgence of interest in artificial neural networks over the past few years, as a researchers from diverse backgrounds have produced a firms theoretical foundation and demonstrated numerous applications of this rich field of study. Neural networks are useful tools for solving many type of problems. These problems may be characterized as mapping(including pattern association and pattern classification), clustering and constrained optimization. There has been great deal of work on enhancing neural network performance. Two important parameter are convergence and generalization. Convergence is the amount of time, measured in CPU operations or training epochs, required to find an acceptable solution for training. Generalization measures the ability to correctly classify new unseen data. This project studies the generalization ability of trained network to classify noisy data. The aim of this project is to develop a network that is able to recognize various inputs through a series of simulation using Neural Network simulator called MATLAB. The effect of the created network with noise are seen. This projects uses the most popular training method in character recognition problem, namely backpropagation algorithm. The theoretical foundation of this algorithm will be studied and summarized. Simulation experiment results on training and testing data will be recorded and discussed.

Text FSKTM_2003_8 IR.pdf Download (917kB)

Actions (login required)

View Item