: Federated Learning through Experience Replay and privacy-preserving data synthesis

Matteo Pennisi, Federica Proietto Salanitri, Giovanni Bellitto, Bruno Casella, Marco Aldinucci, Simone Palazzo, Concetto Spampinato

Abstract: Highlights•Addressing major limitations of FL: presence of a central node and model homogeneity.•Exploiting continual learning to enforce nodes’ convergence towards a shared solution.•GAN-based privacy preserving mechanism to enable synthetic data sharing between nodes.•Tested on two realistic non-i.i.d. medical settings (Tuberculosis, Melanoma).•Alternative architectures to reduce privacy concerns and/or communication costs.