How one can parallelize stochastic gradient descent to obtain high probability deviations bounds from the convergence in average
Pavel Dvurechensky, Alexander Vladimirovich Gasnikov, Anastasia Lagunovskaya · arXiv (Cornell University) · 2017
In this short paper we discuss different ways to obtain high probability bounds deviations for stochastic gradient descent (SGD). We are interested in possibility of parallelization of SGD to obtain this bounds.