Accelerated Primal-Dual Algorithms for Distributed Optimization over Networks

Distributed gradient methods One of the first distributed algorithms for Problem (1) was proposed in the seminal work Nedic and Ozdaglar (2009) and called Distributed Gradient Algorithm (DGD). DGD employing constant step-size can be written in compact form as: xk+1 = Wxk rf(x), (20) where W 2 WG . Defining x(tk) = xk, DGD can be rewritten in a piece-wise continuous form as x(t) = Wx(t) rf(xk), x(t) = x(t), t