I. Federated Learning: Pushing Forward Machine Learning Without Scarifising Privacy

Federated learning is a new family of privacy-preserving machine learning, tackling at the same time the most demanding challenges set by data-driven systems. In this paper, we attempt to present a study on federated learning topics concerning the principles, architecture, and mechanisms for preserving privacy and the protection of information. We survey the challenges and bounds of FL, its communication efficiency, heterogeneity-driven issues, and tradeoffs between privacy and utility. We also relate real-world FL applications to healthcare, finance, and mobile computing. We continue to investigate current research and future directions of the federated learning research landscape, showing in this way that federated learning is really changing the machine learning field as it was known before, that it allows collaborative training of models without violating people's sensitive data.