A new paradigm for artificial intelligence based on group equivariant non-expansive operators

The recent frantic surge of machine learning and, more broadly, artificial intelligence (AI) brings to light old and new open issues, among them, the so-called eXplainable (XAI) – AI that humans can understand as opposed black-box systems where even their designers cannot explain decisions. One major XAI questions is how design transparent incorporate prior knowledge. These topics are becoming relevant pervasive become unfathomable entangled with human factors. Recently a paradigm for has been introduced in literature, based on group equivariant non-expansive operators (GENEOs), which able inject knowledge system. Hence, use GENEOs dramatically reduces number unknown parameters be identified size related training set, providing both computational advantages an increased degree interpretability results. Here we will illustrate main characteristics encouraging results already obtained couple industrial case studies.