Second-order Information Promotes Mini-Batch Robustness in Variance-Reduced Gradients

Open in new window