FEDERATED LEARNING (FL) is the paradigm that aims at solving the data access problem. The topic can be seen as the convergence of 3 areas: Data Science, Security and Distributed connected systems. Among the advantages of a Federated set-up is the exposure to multiple dataset that can help the training processes by improving the generalizability of the models being trained.
However, these benefits bring more responsibilities and challenges at different levels: 1) Data Science and Artificial Intelligence for studying model convergence and finding better ways for aggregating the models locally trained on each client participating to the federation; 2) Communication Efficiency, for addressing scalability and allowing an effective exchange of information among all the parties involved in a federation; 3) Security and Privacy, to contrast losses and stolen information in real life application where FL might be implied for sensitive data.
Riviera W., Boscolo Galazzo I. and Menegaz G.
"FeLebrities: a user-centric assessment of federated learning frameworks"
IEEE Access, 2023
https://ieeexplore.ieee.org/abstract/document/10242027
Foley, P., Sheller, M. J., Edwards, B., Pati, S., Riviera W., Sharma, M., ... and Bakas, S.
"OpenFL: the open federated learning library."
Physics in Medicine & Biology, 67(21), 214001, 2022
https://iopscience.iop.org/article/10.1088/1361-6560/ac97d9