Choosing Neural Equalizers Using Genetic Algorithm

Genetic Algorithm and Artificial Neural Networks have been combined to solve several problems in last two decades. The first one have been used to help find parameters and topology decision of the second one, or to cope with learning algorithm limitations. Some problems demand the application of neural networks as alternative solution to solve them, but studies that develop a methodology to indicate the best neural architecture suitable for a specific application is rare to be found in the literature. In this work we apply genetic algorithm to search for neural weights and use this information to indicate the best structure and measure the efficiency of the learning algorithm. We used a channel equalization problem as an example to test the proposed methodology. The results obtained from this application are very promising.