Escaping Saddle Points with Stochastically Controlled Stochastic Gradient Methods

Open in new window