Ellipsoid method for convex stochastic optimization in small dimension

Inexact Tensor Methods and Their Application to Stochastic Convex Optimization

“…Hence, under the assumption that dimension n ε −1/6 we may expect that in parallel architectures STM(3) outperform FGM + batching. An alternative approach for this case (small-dimensional or(and) high-precision) based on batchedcutting-plane methods was considered in [33].…”