An exact penalty method for semidefinite-box constrained low-rank matrix optimization problems

发布时间：2017-12-22 点击次数：

报告题目：	An exact penalty method for semidefinite-box constrained low-rank matrix optimization problems
报告人：	刘田香博士（香港理工大学）
报告时间：	2017年12月26日 10:00--11:00
报告地点：	理学院东北楼四楼报告厅（404）
报告摘要：	In this talk, we consider a matrix optimization problem involving a semidefinitebox constraint and a rank constraint. We penalize the rank constraint by a non-Lipschitz function and prove that the corresponding penalty problem is exact with respect to the original problem. Next, we present an efficient NPG algorithm to solve the penalty problem and furthermore propose an adaptive penalty method (APM) for solving the original problem. Finally, the efficiency of APM is shown via numerical simulations.