Time: 4:00 p.m. - 5:00 p.m., Friday, March 13, 2026

Host: Professor Zhongyi Zhu, Department of Statistics and Data Science, School of Management

Venue: Room 302, Starr Building

Speaker: Professor Baoxue Zhang (Capital University of Economics and Business)

Topic: A locally sequentially reweighted gradient descent estimator to enhance statistical efficiency for decentralized federated learning

Abstract: 

While many studies have considered the numerical convergence of federated learning algorithms, far less attention has been given to their statistical convergence.  In this paper,  to enhance statistical efficiency, we propose a novel Locally Sequentially Re-weighted Gradient Descent (LSRGD) estimator for decentralized federated learning. Furthermore, we prove that the LSRGD estimator is asymptotically normal and achieves optimal statistical efficiency.  Moreover, we also propose a parallel version of the LSRGD algorithm, referred to as LSRGD-P. Finally, extensive experiments demonstrate that LSRGD and LSRGD-P estimators exhibit superior statistical efficiency compared to existing competitors. This advantage is particularly pronounced in scenarios where the data across different clients are imbalanced.

Speaker Bio:

Professor Baoxue Zhang is a Professor and Doctoral Supervisor at the School of Statistics, Capital University of Economics and Business. He serves as the Vice President of the Chinese Association for Applied Statistics, a Member of the National Steering Committee for Postgraduate Education of Applied Statistics Professional Degree, and the Secretary-General of the Higher Education Branch of the Chinese Association for Statistical Education.