• FRBS
• regression
• TSK
• XAI

This thesis has focused on the design, implementation and testing of federated algorithms for learning Explainable Artificial Intelligence (XAI) models aimed at solving regression problems. In particular, modified Takagi-Sugeno-Kang systems of the first-order have been considered as XAI models. In real applications, single nodes need to predict locally measured quantities, but possibly exploiting global models trained also on data collected in other nodes. Data cannot generally be transferred for privacy reasons. Thus, this thesis has investigated how Federated Learning (FL) paradigm can be employed to learn a global model by exchanging only aggregated values such as statistics and local models, thus preserving privacy of data. These values are used for learning a global model which then is transferred back to each node that uses and possibly optimizes it locally.