A Class Of Smoothing Method For Solving Low Rank Nearest Correlation Matrix Problem

In this paper,we consider a class of smoothing method for solving2l2lp Matrix Minimization of Nearest Correlation Matrix. First of all,we employ spectral function to approximate the original problem;then we solve the smooth parameter approximation problem instead of the original problem;Later we investigate the limiting behaviors of this method and observe the effectiveness of this method,we also give the explicit representation of the directional derivative of the smooth parameter approximation problem;then we introduce a smooth method for solving the smooth parameter ap-proximation problem,In this method we need to solve a series of smooth convex opti-mization problems instead of the smooth parameter approximation problem with the majoried penalty approach;We finally give some numerical experiments of this method.