Review on non-iterative recovery frameworks in compressed sensing

Although theoretical results of compressed sensing (CS) have had amazing impacts on inference problems, there are two significant challenges for practical systems, e.g., real-time processing and sparsity. To this end, the latest studies using deep learning techniques have emerged. In this paper, we review CS frameworks based on deep learning. And the approaches for reconstruction in CS using deep learning techniques show superior performances in terms of recovery time and peak signal-to-noise ratio (PSNR).