Convergence rate analysis of an extrapolated proximal difference-of-convex algorithm

This paper considers a class of difference-of-convex (DC) optimization problems, whose objective function is the sum of a convex smooth function and a possibly nonsmooth DC function. We first propose a new extrapolated proximal difference-of-convex algorithm, which incorporates a more general setting of the extrapolation parameters (beta(k)). Then we prove the subsequential convergence of the proposed method to a stationary point of the DC problem. Based on the Kurdyka-Lojasiewicz inequality, the global convergence and convergence rate of the whole sequence generated by our method have been established. Finally, some numerical experiments on the DC regularized least squares problems have been performed to demonstrate the efficiency of our proposed method.