Distributed Optimization Over Dependent Random Networks

We study the averaging-based distributed optimization solvers over random networks. show a general result on convergence of such schemes using weight-matrices that are row-stochastic almost surely and column-stochastic in expectation for broad class dependent weight-matrix sequences. In addition to implying many previously known results this domain, our work shows robustness link-failure. Also, it provides new tool synthesizing algorithms. To prove main theorem, we establish rate analysis averaging dynamics (dependent) These secondary results, along with required martingale-type them, might be interest broader research endeavors computation