Towards Weaker Variance Assumptions for Stochastic Optimization
A classical assumption for analyzing stochastic gradient algorithms where the squared norm of the stochastic subgradient is allowed to grow as fast as the squared norm of the optimization variable is revisited.