Multistrategy learning using genetic algorithms and neural networks for pattern classification

This paper introduces a two-level learning algorithm which combines parallel genetic algorithm (PGA) and backpropagation algorithm (BP) in order to evolve optimal subsets of discriminatory features for robust pattern classification. In this approach, PGA is used to explore the space of all possible subsets of a large set of candidate discriminatory features. For a given subset, BP is invoked to be trained according to related training data. The individuals of population are evaluated by the classification performance of the trained BP according to the testing data. This process iterates until a satisfactory subset is attained. We use the classification of handwritten numeral and structure of ionosphere for experiment. The results show that this multistrategy methodology improves the classification accuracy rate and the speed of training.