2026 Volume 18 Pages 121-124
We consider sparse optimization problems in which the objective is the sum of a smooth function and a possibly nonsmooth regularizer. In particular, we focus on the case where the regularizer is represented as a difference of convex (DC) function. To characterize critical points, we propose a DC-type gradient mapping, which generalizes the standard gradient mapping used in convex composite optimization. We reformulate the criticality condition as a nonlinear system and develop a semismooth Newton method for solving it. Numerical results demonstrate fast local convergence and, in particular, indicate quadratic convergence of the proposed method.