Evolving Graphs for ANN Development and Simplification

One of the most successful tools in the Artificial Intelligence (AI) world is Artificial Neural Networks (ANNs). This technique is a powerful tool used in many different environments, with many different purposes, like classification, clustering, signal modelization, or regression (Haykin, 1999). Although they are very easy to use, their creation is not a simple task, because the expert has to do much effort and spend much time on it. The development of ANNs can be divided into two parts: architecture development and training and validation. The architecture development determines not only the number of neurons of the ANN, but also the type of the connections among those neurons. The training determines the connection weights for such architecture. The architecture design task is usually performed by means of a manual process, meaning that the expert has to test different architectures to find the one able to achieve the best results. Each architecture trial means training and validating it, which can be a process that needs many computational resources, depending on the complexity of the problem. Therefore, the expert has much participation in the whole ANN development, although techniques for relatively automatic creation of ANNs have been recently developed.