【摘要】 本文通过引入支撑集捕获基数及MP广义逆,提出了一种用于稀疏恢复问题的矩阵广义逆硬阈值追踪算法,并在观测误差存在的情况下给出了算法在约束等距条件(RIP)下的收敛性.数值实验表明,算法不仅极大地减少了收敛所需迭代次数,且观测误差存在的情况下稀疏恢复是强健的.更多还原

【Abstract】 In this paper, based on pseudoinverse and a newly introduced capture cardinality of support set, the new algorithm, Matrix Pseudoinverse Hard Thresholding Pursuit, is proposed to find the sparse solutions of underdetermined linear systems.The convergence is established under restricted isometry property(RIP) condition in the situation where observation error exist. Numerical experiments show that the proposed algorithm not only significantly reduce the iteration number required for convergence, but also can robustly recover the original signal when measurements contain errors.更多还原