报告题目：Personalized Federated Learning for High-dimensional Quantile Regression

报告人：孔令臣教授北京交通大学

报告时间：2026年08月02日上午9:00-10:00

报告地点：正新楼201

校内联系人：温金明 [email protected]

报告摘要：

  Multi-source data analysis is a hot topic in modern statistics research. Its purpose is to obtain accurate estimates by mining multiple sets of data. This existing literature in this area covers a range of approaches such as federated learning,transfer learning, and fusion learning. In practice, distributional heterogeneity across sources poses challenges to model consistency and generalization. This paper considers the high-dimensional regularized quantile regression models with distributional heterogeneity and proposes a robust personalized federated learning algorithm based on an alternating direction method of multipliers, which addresses the negative effects of distributional heterogeneity by providing client specific models. Theoretically, the statistical properties of the estimators and the basic convergence of the proposed algorithm are established. Simulations and real data analysis confirm the effectiveness of the proposed method.

报告人简介：

  孔令臣，北京交通大学教授，博士生导师，中国运筹学会数学规划分会主任委员。主要从事对称锥互补问题和最优化、高维数据分析、统计优化与学习及其应用等方面的研究。在《Mathematical Programming》《SIAM Journal on Optimization》《IEEE Transactions on Pattern Analysis and Machine Intelligence》《IEEE Transactions on Signal Processing》《Technometrics》《Statistica Sinica》《Electronic Journal of Statistics》等期刊发表论文60余篇，2012年获中国运筹学会青年奖，2022年获教育部自然科学二等奖等。