Mixture Odor Classification using Fuzzy Neural Network and Its Optimization through Genetic Algorithm

This report presents an optimized fuzzy neural network through the use of genetic algorithms. Fuzzy neural networks are widely used as it can adaptively deal with measurement of error directly, however, this neural model creates a dilemma from the fact that both large and small networks exhibit a number of disadvantages. If the network size is too small, the error rate tends to increase due to the network might not be able to approximate enough the functional relationship between the input and the target output. While, if the size is too large, the network would not be able to generalize the input data that never been learned before. The developed optimized fuzzy neural system is then applied as the pattern classifier in the artificial odor recognition system. The performance of its recognition ability is explored and compared with that of the conventional Back-Propagation neural system and the un-optimized fuzzy neural. It is shown clearly that the average recognition rate of the GA-optimized fuzzy neural system has higher recognition capability compared with that of the other neural system.