工程数学学报
工程數學學報
공정수학학보
CHINESE JOURNAL OF ENGINEERING MATHEMATICS
2012年
3期
451-461
,共11页
王天荆%杨震%郑宝玉
王天荊%楊震%鄭寶玉
왕천형%양진%정보옥
非凸优化%非光滑优化%同伦方法%极大熵方法
非凸優化%非光滑優化%同倫方法%極大熵方法
비철우화%비광활우화%동륜방법%겁대적방법
nonconvex optimization%nonsmooth optimization%homotopy method%maximum entropy function method
压缩感知可由少量观测重构K-稀疏信号.本文提出的极大熵方法克服了压缩感知中lp(0<p<1)最优化问题的非光滑性.极大熵方法构造一条同伦曲线以获得全局最优稀疏解.数值实验表明极大熵方法的信号重构性能优于l1最优化和AST算法.
壓縮感知可由少量觀測重構K-稀疏信號.本文提齣的極大熵方法剋服瞭壓縮感知中lp(0<p<1)最優化問題的非光滑性.極大熵方法構造一條同倫麯線以穫得全跼最優稀疏解.數值實驗錶明極大熵方法的信號重構性能優于l1最優化和AST算法.
압축감지가유소량관측중구K-희소신호.본문제출적겁대적방법극복료압축감지중lp(0<p<1)최우화문제적비광활성.겁대적방법구조일조동륜곡선이획득전국최우희소해.수치실험표명겁대적방법적신호중구성능우우l1최우화화AST산법.
Compressed Sensing (CS) can reconstruct K-sparse signal from remarkably few measurements.The paper provides a new maximum entropy function (MEF) method to overcome the nonsmooth problem of the nonconvex lp (0 < p < 1) optimization in CS.MEF constructs a homotopy path to obtain the global optimal sparse solution.The numerical results show that MEF has better performance of signal reconstruction than the l1 optimization and the affine scaling transformation algorithm.