报告题目：highly efficient second-order algorithm for solving group SLOPE-regularized least squares problems

报告人：罗自炎教授（北京交通大学）

报告时间：2025年11月21日上午11：00-12：00

报告地点：数学院425报告厅

邀请人：白敏茹

报告摘要：Least squares with group SLOPE regularization have been shown to be effective for feature selection and grouping in high-dimensional statistical regression. Existing algorithms for solving the involved non-smooth optimization models are mostly of first-order type, with relatively low computational efficiency. In this talk, a semi-smooth Newton-based augmented Lagrangian method is designed to resolve the computational challenge, by appropriately leveraging the structural sparsity in generalized Jacobian matrices from the underlying essential iteration subproblems. The theoretical convergence results, along with the numerical performances that conducted on real datasets, certify both the effectiveness and efficiency of our proposed algorithm. This is a joint work with Hang Zhao.

报告专家简介：罗自炎，北京交通大学数学与统计学院教授、博士生导师，中国运筹学会数学规划分会副秘书长，中国运筹学会女性工作委员会委员，中国运筹学会算法软件与应用分会理事，中国高等教育学会教育数学专业委员会资深理事。曾访问美国斯坦福大学、新加坡国立大学、香港理工大学、英国南安普顿大学等。主要从事张量优化、稀疏优化及统计优化的理论、算法及应用研究，在SIOPT、MP、MOR、IEEE TSP、IEEE TWC、JMLR等顶级期刊发表学术论文，合著SIAM出版社英文专著1部，曾获教育部自然科学奖二等奖、中国运筹学会青年科技奖提名奖、2023年入选国家级青年人才计划。