在线赌场

“七秩工大，学术领航”——70周年校庆系列学术讲座

报告题目：An Efficient ADMM-Based Algorithm for Non-convex Penalized Support Vector Regression

报  告 人：冯三营

报告时间：11月10日15:40-16:10

报告地点：莲花街校区惟德楼315会议室

报告人简介：

冯三营，统计学博士，教授，博士生导师，中国现场统计研究会理事，全国工业统计学教学研究会理事，河南省应用统计学会副秘书长。主要研究方向：非参数统计、模型和变量选择、统计机器学习、函数型数据分析等。主持完成和在研国家与省部级科研项目20余项。截至目前，在《Statistica Sinica》、《Statistics and Computing》、《Journal of Multivariate Analysis》、《Computational Statistics and Data Analysis》、《Statistics》以及《应用概率统计》等国内外学术期刊上发表论文70余篇，其中SCI 收录论文50余篇，出版学术专著《现代测量误差模型》1 部 (该书入选了科学出版社的现代数学基础丛书系列)。

报告简介：

To address the limitations of conventional Support Vector Regression (SVR) in high-dimensional settings, including high computational complexity, lack of variable selection capability, and estimation bias induced by convex regularization. This paper proposes a non-convex penalized SVR framework coupled with an efficient ADMM-based optimization algorithm. The framework incorporates the Smoothly Clipped Absolute Deviation (SCAD) and Minimax Concave Penalty (MCP) into the SVR model, leveraging the asymptotic unbiasedness of non-convex penalties and their capability for inducing sparsity. This effectively balances the bias–variance trade-off and simultaneously achieves variable selection and unbiased parameter estimation. Algorithmically, the reformulation ensures that all subproblems admit closed-form solutions, significantly enhancing computational efficiency and numerical stability. Theoretically, we establish that the proposed ADMM algorithm converges to a KKT point under mild assumptions. Extensive experiments on six synthetic datasets and real-world datasets demonstrate that, compared to the Elastic-Net SVR (En-SVR) method, the proposed non-convex penalized SVR model exhibits superior predictive accuracy, more accurate variable selection, and greater robustness to correlated features. The method is particularly suitable for high-dimensional sparse regression tasks and provides a reliable tool for high dimensional data modeling in genomics, medicine, and related fields.

在线赌场排名

2026年11月9日