Feasibility Pump Algorithm for Sparse Representation under Laplacian Noise

Figure 2

Difference in relative representation errors between RLAD and SFPreg for matrix conditioning of 1000. A positive value means that SFPreg is better.