General inertial proximal stochastic variance reduction gradient for nonconvex nonsmooth optimization

Abstract In this paper, motivated by the competitive performance of proximal stochastic variance reduction gradient (Prox-SVRG) method, a novel general inertial Prox-SVRG (GIProx-SVRG) algorithm is proposed for solving class nonconvex finite sum problems. More precisely, Nesterov’s momentum trick-based extrapolation accelerated step incorporated into framework method. The GIProx-SVRG possesses more expression and thus can potentially achieve convergence speed. Moreover, based on supermartingale theory error bound condition, we establish linear rate iterate sequence generated algorithm. We observe that there no in which technique whereas such paper. Experimental results demonstrate superiority our method over state-of-the-art methods.