Coordinate Dual Averaging for Decentralized Online Optimization with Nonseparable Global Objectives
From MaRDI portal
Publication:6265010
DOI10.1109/TCNS.2016.2573639arXiv1508.07933MaRDI QIDQ6265010FDOQ6265010
Authors: Soomin Lee, Angelia Nedić, Maxim Raginsky 
Publication date: 31 August 2015
Abstract: We consider a decentralized online convex optimization problem in a network of agents, where each agent controls only a coordinate (or a part) of the global decision vector. For such a problem, we propose two decentralized variants (ODA-C and ODA-PS) of Nesterov's primal-dual algorithm with dual averaging. In ODA-C, to mitigate the disagreements on the primal-vector updates, the agents implement a generalization of the local information-exchange dynamics recently proposed by Li and Marden over a static undirected graph. In ODA-PS, the agents implement the broadcast-based push-sum dynamics over a time-varying sequence of uniformly connected digraphs. We show that the regret bounds in both cases have sublinear growth of , with the time horizon , when the stepsize is of the form and the objective functions are Lipschitz-continuous convex functions with Lipschitz gradients. We also implement the proposed algorithms on a sensor network to complement our theoretical analysis.
Convex programming (90C25) Decentralized systems (93A14) Multi-agent systems (93A16) Networked control (93B70)
This page was built for publication: Coordinate Dual Averaging for Decentralized Online Optimization with Nonseparable Global Objectives
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6265010)
