Desenvolvimento da rede perceptron multicamadas baseado na otimização por colônia de formigas

Abstract

This work proposes the development of an Intelligent Hybrid System to which the number of hidden layers, the number of neurons in each layer and the learning rate of a multilayer Perceptron network will be optimized through an algorithm based on ant colony optimization. The development took place in five trials. A priori databases of medical diagnoses were used "Breast Cancer", "Mammographic Mass" and "Thyroid" were obtained from the machine learning repository, UCI. In sequence the databases of electromyographic signals "Class of Movements" and "Levels of Contraction" were used. From the results obtained it was possible to analyze that the proposed system presents itself as a great tool to assist professionals in projects involving artificial neural networks, proving capable of adapting to different classes of problems involving the optimization of the topology of multilayer Perceptron networks.