Distributed Stochastic Algorithm for Convex optimization Problem Over Time-varying Directed Graphs
Songsong Cheng, Yinghui Wang, Yiguang Hong · 2019
In this paper, we develop a distributed stochastic algorithm for convex optimization problem over time-varying directed communication graphs. Firstly, we introduce a surplus based method to reduce the computation and communication burden. Besides, we overcome the unbalance of the timevarying directed graphs and remove the stringent assumptions on the double stochastic matrix condition. Furthermore, by formulating row stochastic matrix for decision variable and column stochastic matrix for surplus variable, we establish the convergent performance and achieve the convergence rate O(lnk/k). Finally, we employ a numerical example to verify the effectiveness of the proposed method.