Convergence analysis of distributed stochastic gradient descent with shuffling