Fig. 1From: Stochastic convex sparse principal component analysisConvergence for synthetic data. Convergence of the proposed stochastic Cvx-SPCA with (Prox-SVRG) and without variance reduction (prox-SGD). Proximal stochastic gradient with variance reduction has a faster convergence rate, since the variance caused by random sampling is bounded in Prox-SVRGBack to article page