Accelerated Design for Distributed Asynchronous Optimization

This paper proposes an accelerated algorithm for addressing distributed optimization problems in directed networks. We develop an accelerated asynchronous distributed Nesterov gradient descent algorithm for directed networks called ASYADND. For the considered problems, unpredictable, time-varying, and bounded. Under strong convexity and smoothness conditions, we analyze the proposed algorithm by transforming the asynchronous scheme into a synchronous counterpart. We adopt an acceleration mechanism and establish its linear convergence with the aid of the small gain theorem.