Reproduce “Personalizing Low-Rank Bayesian Neural Networks Via Federated 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
Boning Zhang, Dongzhu Liu, Osvaldo Simeone, Guanchu Wang, Dimitrios Pezaros, Guangxu Zhu
LR-BPFL combines a shared deterministic model with client-specific Bayesian low-rank corrections and adaptive rank selection. The method keeps corrections local while clients adapt their complexity. In CIFAR-10, LR-BPFL reports 83.30% accuracy and 1.74% ECE, compared with pFedBayes at 80.86% accuracy and 6.26% ECE. Its ablation improves accuracy and calibration in three tested heterogeneity conditions, without establishing that rank directly measures uncertainty.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.