We propose a flexible gradient tracking approach with adjustable computation
and communication steps for solving distributed stochastic optimization problem
over networks. The proposed method allows each node to perform multiple local
gradient updates and multiple inter-node communicat