Reproduce “Privacy Preserving Gossip Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Erkan Bayram, Mohamed-Ali Belabbas, Tamer Başar
The method learns private examples while preserving earlier mappings in a shared model. A teacher uses a robust differentially private update, while protected agents use local projections and private gossip to constrain changes. The authors prove geometric convergence for gossip and projection; experiments find higher test accuracy than direct differential privacy on MNIST and Circle at matched per-update privacy settings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.