Abstract:
A novel method for selecting the appropriate architecture and learning rule of an artificial neural network for a given application is discussed in this paper. Evolutionary Artificial Neural Networks (EANN) use the adaptation capabilities of genetic algorithms in which the natural selection process is used to attain the optimum network structure and learning algorithm for a specific task. ANNEbot is a framework which allows the combined powers of learning and adaptation of EANNs to be applied in various machine learning tasks. The framework was tested on the Iris Classification problem and the Wisconsin Breast Cancer Diagnosis problem, both of which provided results with above 90% accuracy. ANNEbot was also successfully applied on a robotic application for obstacle avoidance.
