Online Federated Learning via Staleness Awareness and Performance Prediction. Georgios Damaskinos, Rachid Guerraoui, Anne-Marie Kermarrec, et Rhicheek Patra [EPFL], Vlad Nitu [LIRIS à Lyon], et François Taïani [équipe WIDE à l'IRISA] ⟨10.1145/3423211.3425685⟩. ⟨hal-03043237⟩ ...
...has been awarded the Best Paper Award at the 2020 ACM/IFP International Conference on Middleware, the premier event for the discussion of innovations and recent scientific advances of middleware systems. This paper explores the system implications of combining an online learning approach (that helps improves the timeliness of inferred results) with federated learning (a distributed learning technique that keeps sensitive data on a user’s device) on mobile phones. It is the result of a collaboration between the DCL and SACS labs of EPFL (Lausanne, Switzerland), and the WIDE team, in the context of the EPFL-Inria International Lab.
Congratulations to all involved!
