2019年度3044am永利集团欢迎您外请专家学术报告之三
时间：2019-01-05 18:02:03 来源：3044am永利集团欢迎您 作者： 阅读： 次
报告题目：A Unified Algorithm for Mixed l2,p-Minimizations and Its Application in Feature Selection
报告人： 王丽平 博士
报告时间：2019年1月13日（周日）16:30—18:00
报告地点：3044am永利集团欢迎您学术报告厅
报告摘要：In this talk, mixed l2,1 matrix norm is presented and used to find jointly sparse solutions. An efficient iterative algorithm has been designed to solve l2,1-norm involved minimizations and the convergence is also ensured. To our best of knowledge, computational studies have showed that lp-regularization (0< p<1 ) is sparser than l1-regularization, but the extension to matrix norm has been seldom considered. This paper presents a definition of mixed l2,p (p∈(0, 1]) matrix pseudo norm which is thought as both generalizations of lp vector norm to matrix and l2,1-norm to nonconvex cases (0< p<1 ). fortunately, an efficient unified algorithm is proposed to solve the induced l2,p -norm ( p∈(0, 1]) optimization problems. The convergence can also be uniformly demonstrated for all p∈(0, 1]. Typical different p∈(0, 1] are applied to select features in computational biology and the experimental results show that some choices of 0< p<1 are potentially competitive with p=1.
报告摘要：In this talk, mixed l2,1 matrix norm is presented and used to find jointly sparse solutions. An efficient iterative algorithm has been designed to solve l2,1-norm involved minimizations and the convergence is also ensured. To our best of knowledge, computational studies have showed that lp-regularization (0< p<1 ) is sparser than l1-regularization, but the extension to matrix norm has been seldom considered. This paper presents a definition of mixed l2,p (p∈(0, 1]) matrix pseudo norm which is thought as both generalizations of lp vector norm to matrix and l2,1-norm to nonconvex cases (0< p<1 ). fortunately, an efficient unified algorithm is proposed to solve the induced l2,p -norm ( p∈(0, 1]) optimization problems. The convergence can also be uniformly demonstrated for all p∈(0, 1]. Typical different p∈(0, 1] are applied to select features in computational biology and the experimental results show that some choices of 0< p<1 are potentially competitive with p=1.
报告人简介: 王丽平，南京航空航天大学理学院副教授，运筹学、计算数学专业硕士生导师，中国运筹学学会理事。2004年毕业于中科院数学与系统科学研究院，获计算数学博士学位，2009.2-2010.2在巴西巴拉纳联邦大学完成博士后项目，2013年11月和2017年8月两次访问俄罗斯科学院计算中心，2017.5-2017.6访问香港中文大学， 2017.10-2018.10访问美国路易斯安娜州立大学。主要研究领域为最优化理论与算法，数值优化方法在模式识别、人工智能和地球物理等领域的应用。在国际著名学术期刊Computational Optimization and Applications, Pattern Recognition, SCIENCE CHINA Mathematics, Scientific Reports等发表论文二十余篇。先后主持完成国家自然科学基金青年基金项目、国家自然科学基金面上项目、 国家基金委重点国际合作项目中俄合作研究项目和中俄合作交流项目。
欢迎广大师生参加！
3044am永利集团欢迎您
2019年1月5日
学院微信
学院微博