压缩感知和稀疏优化简介.pdfVIP

2012 年 9 月 运筹学学报 第 16 卷第 3 期 Sep. , 2012 Operations Research Transactions Vo 1.16 No.3 压缩感知和稀疏优化简介 文再文 1 印卧涛 2↑ 刘敖3 张寅2 摘要介绍压缩感知和稀疏优化的基本概念、理论基础和算法概要.压缩感知利用原始信 号的稀疏性，从远少于信号元素个数的测量出发，通过求解稀疏优化问题来恢复完整的原始 稀疏信号.通过一个小例子展示这一过程，并以此说明压缩感知和稀疏优化的基本理念.接 着简要介绍用以保证已凸优化恢复稀疏信号的零空间性质和 RIP 条件.最后介绍求解稀疏 优化的几个经典算法. 关键词压缩感知，稀疏优化，零空间性质，受限正交条件，紧缩算子，线性化近似è:算 法，分裂 Bregman 方法和交替方向增广拉格朗日函数法， Bregman 方法和增广拉格朗日函 数法 中图分类号 0224， 029 数学分类号 49M29， 65K05, 90C25 , 90C06 Introduction to compressive sensing and sparse optimization 1 2 3 2 叽TEN Zaiwen YIN 矶Totao ↑ LIU Xin ZHANG Yin Abstract We briefly introduce the basic principle and theory of compressive sensing and sparse optimization. Compressive sensing is a new paradigm of signal acquisition , which senses a sparse signal by taking a set of incomplete measurements and recovers the signal by solving an optimization problem. This article first illustrates the compressive sensing paradigm through a synthetic example. Then we describe two sufficient conditions , the null space property and restricted isometry principle , for f 1 convex minimization to give the sparsest solution. Finally, we summarize a few typical algorithms for solving the optimization models arising from compressive sensing. Keywords compressive sensing, sparse optimization , null space property , RIP , shrink- age , prox-linear algorithms , split Bregma时alternati吨 direction augmented Lagragian method , Bregman/augmented Lagragian method