江西师范大学学报(自然科学版)
江西師範大學學報(自然科學版)
강서사범대학학보(자연과학판)
JOURNAL OF JIANGXI NORMAL UNIVERSITY(NATURAL SCIENCES EDITION)
2015年
3期
281-285
,共5页
l1 正则化%l0 正则化%压缩感知%稀疏恢复
l1 正則化%l0 正則化%壓縮感知%稀疏恢複
l1 정칙화%l0 정칙화%압축감지%희소회복
l1 regularization%l0 regularization%compressive sensing%sparse restoration
针对压缩感知模型,讨论了基于l0正则化的正交匹配追踪算法( OMP)与基于l1正则化的同伦算法( HM)和迭代加权最小二乘法( IRLS)。通过数值实验结果分析,验证了3种算法的有效性,且相对于2种基于l1正则化的算法,OMP算法的迭代次数与耗时更少,均方误差更小。
針對壓縮感知模型,討論瞭基于l0正則化的正交匹配追蹤算法( OMP)與基于l1正則化的同倫算法( HM)和迭代加權最小二乘法( IRLS)。通過數值實驗結果分析,驗證瞭3種算法的有效性,且相對于2種基于l1正則化的算法,OMP算法的迭代次數與耗時更少,均方誤差更小。
침대압축감지모형,토론료기우l0정칙화적정교필배추종산법( OMP)여기우l1정칙화적동륜산법( HM)화질대가권최소이승법( IRLS)。통과수치실험결과분석,험증료3충산법적유효성,차상대우2충기우l1정칙화적산법,OMP산법적질대차수여모시경소,균방오차경소。
For compressive sensing,orthogonal matching pursuit algorithm( OMP)based on l0 norm regularization, homotopy algorithm( HM)based on l1 norm regularization and iteratively reweighted least squares algorithm( IRLS) based on l1 norm regularization are introduced. In numerical experiment,the validity of three algorithms above through analysis of numerical result are proved. Furthermore,for lower CPU cost and smaller mean square error, OMP is more efficient than other two algorithms based on l1 norm regularization.