Adaptive Variance Reduction for Stochastic Optimization under Weaker Assumptions Wei Jiang 1