A survey of numerical algorithms that can solve the Lasso problems

In statistics, the least absolute shrinkage and selection operator (Lasso) is a regression method that performs both variable regularization. There lot of literature available, discussing statistical properties coefficients estimated by Lasso method. However, there lacks comprehensive review algorithms to solve optimization problem in Lasso. this review, we summarize five representative optimize objective function Lasso, including iterative threshold algorithm (ISTA), fast shrinkage-thresholding (FISTA), coordinate gradient descent (CGDA), smooth L1 (SLA), path following (PFA). Additionally, also compare their convergence rate, as well potential strengths weakness. This article categorized under: Statistical Models > Linear Algorithms Computational Methods Numerical Complexity