An investigation into the feasibility of performing federated learning on social linked data servers

Arana, Nayil, Ragab, Mohamed and Tiropanis, Thanassis (2024) An investigation into the feasibility of performing federated learning on social linked data servers. In WWW '24: Companion Proceedings of the ACM on Web Conference 2024. Association for Computing Machinery. pp. 1712-1714 . (doi:10.1145/3589335.3651950).

Record type: Conference or Workshop Item (Paper)

Abstract

Federated Learning (FL) and the Social Linked Data (\textttSolid ~\footnotehttps://solidproject.org/ ) framework represent decentralized approaches to machine learning and web development, respectively, with a focus on preserving privacy. Federated learning enables the distributed training of machine learning models across datasets partitioned across multiple clients, whereas applications developed with the Solid approach store data inPersonal Online Data Stores (pods) under the control of individual users. This paper discusses the merits and challenges of executing Federated Learning on Solid pods and the readiness of the Solid server architecture to support this. We aim to detail these challenges, in addition to identifying avenues for further work to fully harness the benefits of Federated Learning in Solid environments, where users retain sovereignty over their data.

Text

arana-et-al-24 - Accepted Manuscript

Download (459kB)