-->
数学系Seminar第1733期 A revised gradient descent algorithm for linearly constrained lp minimization with p ∈ (0,1)

Created Date 12/6/2018 惠英   View Numbers  68 Return    
字号:   
 

报告主题:A revised gradient descent algorithm for linearly constrained lp minimization with p ∈ (0,1)
报告人:Shan Jiang   博士  (美国北卡州立大学)
报告时间:2018年12月19日(周三)15:00
报告地点:校本部G508
邀请人:白延琴
主办部门:理学院数学系
报告摘要:In this paper, we study the linearly constrained lp minimization problem with p ∈ (0,1). Unlike former works in the literature that propose ε-KKT points through relaxed optimality conditions, here we define a scaled KKT condition that is not relaxed. A revised gradient descent algorithm is proposed to search for points satisfying the proposed condition. The convergency proofs with complexity analysis of the proposed algorithm are provided. Computational experiments support that the proposed algorithm is capable of achieving better sparse recovery with far less computational time compared to state-of-the-art interior-point based algorithm.

 

 

欢迎教师、学生参加 !


版权所有 © 上海大学    沪ICP备09014157   地址:上海市宝山区上大路99号(周边交通)   邮编:200444   电话查询
技术支持:上海大学信息化工作办公室   联系我们