基于NSGPBB算法的压缩感知稀疏信号重构

Compressed sensing sparse signal reconstruction based on NSGPBB algorithm

为了更好地重构原始信号,提出一种带有交替BB步长的非单调梯度投影算法(NSGPBB)。将无约束凸优化问题转化为在闭凸集上的边界约束二次规划问题,并证明了该算法的收敛性。数值实验结果表明,该算法是有效的,且收敛速度快于梯度投影算法。

To solve a key problem of sparse signal reconstruction,a nonmonotone gradient projection algorithm with the alternation of Barzilai-Borwein rules（NSGPBB）is proposed for bound-constrainted quadratic programming（BCQP）on a convex set.Global convergence of this method is proved.The numerical results show that the method is effective and faster than other spectral gradient projection algorithms.