Granularity into functional networks

In the last years, many papers on Granular Neural Networks appeared, aiming at achieving a higher degree of transparency in the network architecture and a better understanding of the involved steps. This paper is a first attempt to emphasize and exploit the notion of granularity into Functional Networks. Functional Networks are a relatively recent alternative to Neural Networks and they have been proved to ensure better performances. We propose a revised scheme from a granular perspective with a different learning algorithm with respect to the original one. An application example shows the good performance of this new scheme.