Unifying viewpoints on distributed asynchronous optimization for MISO interference channels

Abstract

We adopt the power gain description of wireless networks in order to provide conceptual insights on existing distributed asynchronous beamforming algorithms. From this common viewpoint, we compare the asynchronous distributed pricing (ADP) algorithm and the asynchronous scaled gradient projection (SGP) algorithm in terms of local subproblems, required information exchange and convergence behavior. We illustrate the underlying mechanisms that ensure the convergence to a stationary point of a sum utility function (i.e., complementary user actions, step size bounds), and we point out the algorithms' pros and limitations.