Breaking data silos with Federated Learning

Federated learning has been recognized as a promising technology with the potential to revolutionize the field of Artificial Intelligence (AI). By leveraging its decentralized nature, it has the potential to overcome known barriers to AI, such as data acquisition and privacy, paving the way for unprecedented advances in AI.

This dissertation argues the benefits of this technology as a catalyst for the irruption of AI both in the public and private sector. Federated learning promotes cooperation among otherwise competitive entities by enabling cooperative efforts to achieve a common goal.

In this dissertation, I investigate the goodness-of-fit of this technology in several contexts, with a focus on its application in power systems, financial institutions, and public administrations. The dissertation comprises five papers that investigate various aspects of federated learning in the aforementioned contexts. In particular, the first two papers explore promising venues in the energy sector, where federated learning offers a compelling solution to privately exploit the vast amounts of data and decentralized ownership of data by consumers. The third paper elaborates on another paradigmatic example, in which federated learning is used to foster cooperation among financial institutions to produce accurate credit risk models. The fourth paper makes a juxtaposition with the previous ones centered on the private sector. It elaborates on the use cases of federated learning for public administrations to reduce barriers to cooperation. Lastly, the fifth and last article acts as a finale of this dissertation, compiles the earlier work and elaborates on the constraints and opportunities associated with adopting this technology, as well as a framework for doing so.