Academic Report of School of Mathematical Sciences [2026] No. 074

(Series Report for High-Level University Construction No. 1333)

Title：Riemannian Optimization and a Riemannian Federated Learning Algorithm

Speaker：Wen Huang, Professor (Xiamen University )

Time：16:30-17:30, July 2, 2026

Location：Room 1420, Huiwen Building, Yuehai Campus, Shenzhen University

Abstract: Optimization on Riemannian manifolds, also called Riemannian optimization, considers finding an optimum of a real-valued function defined on a Riemannian manifold. Riemannian optimization has been a topic of much interest over the past few years due to many important applications, e.g., blind source separation, computations on symmetric positive matrices, low-rank learning, graph similarity, community detection, and elastic shape analysis. In this presentation, the framework of Riemannian optimization is introduced, and the current state of Riemannian optimization algorithms are briefly reviewed. A Riemannian federated learning algorithm (RFedAGS) is presented and analyzed. The difficulties of the generalization to the Riemannian setting are highlighted. The convergence results of the proposed RFedAGS are established. Extensive experiments conducted on synthetic and real-world data demonstrate the good performance of RFedAGS.

Speaker Profile：Professor Wen Huang received his Ph.D. in Applied and Computational Mathematics from Florida State University in 2014. From 2014 to 2016, he served as a postdoctoral researcher in the Department of Mathematical Engineering at the University of Leuven in Belgium. From 2016 to 2018, he served as the Faye Postdoctoral Lecturer in the Department of Computational and Applied Mathematics at Rice University in the United States. He joined Xiamen University in September 2018. His primary research interests lie in optimization algorithms on Riemannian manifolds and their applications, including the theory and algorithmic implementation of large-scale problems in signal processing, image processing, computer vision, network component analysis, statistics, and machine learning. His research findings have been published in leading journals such as SIOPT, SISC, MATH PROGRAM, and NUMER MATH. He developed ROPTLIB, a C++ software toolkit for solving manifold optimization problems. He has led Young Scientist and General Projects funded by the National Natural Science Foundation of China and was selected for the National High-Level Talent Program (Youth Category) in 2021.

Faculty and students are welcome to attend!

Invited by: Yuqia Wu

School of Mathematical Sciences

July 1, 2026