Abstract: The success and popularity of Compressed Sensing (CS) depends essentially on the ability of efficiently finding an approximated sparse solution of an underdetermined linear system. There are two categories of algorithms: convex relaxation algorithm and greedy algorithm. In this talk we present a primal dual active set strategy for both convex relaxation optimization problem and greedy algorithm. The locally superlinearly convergence for PDAS for convex relaxation problem is proved. Several numerical examples are given to show the efficiency of the PDAS algorithms, both for convex relaxation problem and l_0-minimization problem. This method can be extended to the inverse problems or other problems without RIP for the measurement matrix. 

 

报告人简介: 吕锡亮，武汉大学副教授。2006年毕业于新加坡国立大学数学系，获博士学位，2007年至今先后于美国马里兰大学数学系和奥地利科学院RICAM(Johann Radon Institute for Computational and Applied Mathematics)研究所从事访问研究和博士后研究，主要研究方向为偏微分方程数值解，约束优化控制，有限元分析，以及计算流体力学。

  

Date&Time: July 12, 2013 (Friday), 10:30 - 11:30 a.m. 
Location: 606 Conference Room