الفهرس | Only 14 pages are availabe for public view |
Abstract Artificial neural networks exhibit many useful and unique attributes. Hopfield neural networks are very good at recognizing patterns in noisy data. The variations of the techniques for training Hopfield neural networks are computationally expensive, and not always able to find a good solution. In short, Hopfield neural networks can be difficult and expensive to train. Evolutionary methods such as genetic programming (GP) are global search techniques based on the evolution of a given population. In this population every individual represents a solution for a problem that is intended to be solved. The evolution is achieved by means of selection of the best individuals although the worst ones also have a little chance of being selected. This process is developed using selection, crossover and mutation operations. After several generations, it is expected that the population might contains some good solutions for the problem. The GP encodes for the solutions is tree-shaped, so the user must specify which terminals (leaves of the tree) and functions (nodes with children) will be used by the evolutionary algorithm in order to build complex expressions. Pattern recognition is an important practical application of associative memory. The task is to produce a clear, noise-free pattern at the output when the input vectors are noisy versions of the trained patterns This thesis investigates the benefits of combining Hopfield neural networks and genetic programming into hybrid system for classification systems, especially for applications character recognition tasks. The research reported in this thesis investigates the possibility of using an evolutionary approach to improve Hopfield neural networks. The technique investigated by this thesis is the use of genetic programming to evolve the structure of the Hopfield neural network. The structure of the gene used by this technique obviates the need for real values to be encoded on the chromosome. This thesis tests the technique on real life problems. The proposed method is applied on two different data sets character recognition and spoken Arabic digits. |