报告题目：Classical and Innovative Methods in Support Vector Machines

报告时间：2024年4月18日 下午19:00

报告地点：南湖校区教学科研楼307室

主办单位：beat365正版唯一网站app

报告人：Renato De leone

报告人简介：Renato De leone，现任意大利卡梅里诺大学教授，博士生导师，担任数学和计算机科学系主任，科学和技术学院副院长，校长助理。1981年获得意大利那不勒斯大学博士学位。1984-1993年间，曾任美国威斯康辛大学麦迪逊分校的客座教授。主要研究领域包括优化和数学规划、数学建模、机器学习和决策支持系统。

摘要：Support Vector Machines (SVMs) are a novel and very effective tool for classification and regression problems. The method requires the solution of a convex quadratic problem with linear constraints and simple bounds on the variables, for which convergence results and effective algorithms are available. In this talk, after presenting the main characteristics of classical SVM methods, some novel approaches will be presented based on a variation of the standard SVM method where two non—parallel hyperplanes are constructed. In particular, Sparse Twin Parametric Margin SVM, and Multiclass Robust Twin Parametric Margin SVM will be presented.