A Data Distribution Information Combined SVM Model for Binary Classification Problem

报告摘要：By taking the data distribution information combined with the data sampling information into account, we establish an enhanced SVM model for binary classification problem in this talk. Since more deeper data structure information is employed, the proposed model exhibits more robustness as it not only minimizes the worst-case misclassification probability, but also minimizes the number of misclassified samples. With the aid of the projection penalty and the alternating minimization technique, we design a dual type algorithm for the model. The efficiency and performance of the proposed model and the algorithm are validated via theoretical analysis as well as some illustrative numerical examples.

报告人简介：王宜举，曲阜师范大学管理学院院长，教授，博士生导师。中国科学院博士,香港理工大学和南京师范大学博士后。主要从事最优化的理论与算法研究，发表有一定学术影响力的论文80多篇。主持（完成）国家自然科学基金5项、省部级科研项目7项。获教育部和山东省自然科学二、三等奖5项。2015年享受国务院特殊津贴。多次到香港城市大学、香港理工大学和澳大利亚的科廷大学进行学术访问和交流。