报告人：邹斌教授（湖北大学）

报告题目：New Lagrangian Support Vector Machine Algorithm

报告摘要：In this paper, we focus on Lagrangian Support Vector Machines for classification (LSVMC) algorithm based on Markov chain samples. We firstly set up the generalization bounds of LSVMC algorithm with Markov chain samples. And then we verify the consistency of this algorithm and obtain the optimal learning rate of LSVMC algorithm with Markov chain samples. We also propose a new LSVMC algorithm with k-times Markov selective sampling (LSVMC-MSS) and show the experimental results of this new algorithm for UCI datasets. The experimental results show that the LSVMC-MSS has better generalization performance compared to the classical LSVMC algorithm based on randomly independent sampling and the known SVMC algorithm based on Markov sampling. If the sampling and training total time is a main concern of classification problems, then the LSVMC-MSS algorithm will be the preferred method.

报告人简介：邹斌，博士，湖北大学数学与统计学学院教授、博士生导师。2007年6月博士毕业于湖北大学基础数学专业，博士学位论文获2008年湖北省优秀博士学位论文。2008年1月至2009年12月在西安交通大学信息与系统科学研究所进行博士后研究工作，合作导师为徐宗本教授。当前主要研究兴趣为统计学习理论，机器学习，大数据分析等。主持省部级、国家级科学基金共计7项，以第一作者或通讯作者在《IEEE Transactions on Neural Networks and Learning Systems》、《IEEE Transactions on Cybernetics》、《Machine Learning》、《Neural Networks》、《中国科学》等国际国内知名期刊上发表论文40余篇。

报告时间：2018年12月7日（星期五）上午10:00-11:00

报告地点：科技楼南楼602室