Date & location: Tuesday 27th of May 2025, Lagrange Gris (L101), Inria Sophia Antipolis
Title: FedBEns: One-Shot Federated Learning based on Bayesian Ensemble
Abstract: One-Shot Federated Learning (FL) is a recent paradigm that enables multiple clients to cooperatively learn a global model in a single round of communication with a central server.In this presentation, we analyze the One-Shot FL problem through the lens of Bayesian inference and propose FedBEns, an algorithm that leverages the inherent multi-modality of local loss functions to find better global models.Our algorithm leverages a mixture of Laplace approximations for the clients’ local posteriors, which the server then aggregates to infer the global model.We conduct extensive experiments on various datasets, demonstrating that the proposed method outperforms competing baselines that typically rely on unimodal approximations of the local losses.
Paper: https://arxiv.org/abs/2503.15367
Bio: Jacopo Talpini came in NEO team for a PhD exchange under the supervision of Giovanni Neglia and defended his PhD last month at the University Milano-Bicocca. His research activity is focused on applying machine learning methods in the cyber-security field. Previously, he graduated in Astrophysics, his master thesis project was focused on applying machine learning techniques to characterize transient noise in gravitational-wave detectors.He is now a post-doctorant in Bicocca university, working with Marco Savi (his former PhD director).
