压缩感知中基于坐标下降的稀疏信号恢复

The sparse signal recovery based on coordinate descent optimization for compressed sensing

目前利用压缩感知理论来解决稀疏信号的恢复,已经成为了研究的热点之一。为了提高一类坐标下降法在压缩感知中的运行速度和减少其迭代次数,提出了一种新的扫描模式来选择更新的坐标。该扫描模式能够一次选择多个坐标,从而提高算法的效率。试验结果给出了在几种不同的矩阵下,新方法运行时间比之前的单个选择坐标更新的方法要短,并且减少了迭代次数。

The research of using compressed sensing theory to recover sparse signals has been very popular. To improve the speed of a coordinate descent method, which has been applied to com- pressed sensing, a new sweep pattern is proposed. The new sweep pattern can choose multiple coordinates so that the performance of the algorithm is improved. The numerical experiments show that the new algorithm brings shorter runtime and reduces iterations than the prior methods selecting single coordinates under various matrices.