快速L1范数最小化算法的性能分析和比较

The Performance Analysis and Comparison for Fast L1-Minimization Algorithm

随着新兴压缩传感（Compressive Sensing，cs）理论的出现，使用L1范数最小化（L1-min）算法进行信号处理和优化成为近几年的热门课题．由于传统的求解方法对于大规模数据的处理效率很低，例如内点法，越来越多的快速L1－min算法被提出，这些算法在速度和处理效果上都各有优势．该文首先介绍了L1-min算法以及影响算法效率的主要因素，然后通过实验数据对五种快速L1－min算法在处理大规模数据时的性能进行分析和客观评价。

With the emergence of new compressive sensing theory, L1-norm minimization algorithm used in processing and optimizing signals has become a hot topic in recent years, because the conventional algorithms, for example, interior-point method, are ineflficient in solving large-scale data. More and more fast L1-rain algorithms have been proposed, all of which have their own advantages in speed and effect. This paper first introduces the L1 -min algorithms and the main factors affecting the efficiency of the algorithms, and then introduces the present five major fast L1-min algorithms. Finally, based on the experimental data, it analyzes and gives an objective evaluation to the performance of these fast algorithms when dealing with large-scale data.