Decentralized online optimization with global objectives and local communication

Angelia Nedich, Soomin Lee, Maxim Raginsky

Research output: Chapter in Book/Report/Conference proceedingConference contribution

13 Citations (Scopus)

Abstract

We consider a decentralized online convex optimization problem in a static undirected network of agents, where each agent controls only a coordinate (or a part) of the global decision vector. For such a problem, we propose a decentralized variant of Nesterov's primal-dual algorithm with dual averaging. 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 [1]. We show that the regret has sublinear growth of O (√T) with the time horizon T when the stepsize is of the form 1/√t and the objective functions are Lipschitzcontinuous convex functions with Lipschitz gradients. We prove an analogous bound on the expected regret for the stochastic variant of the algorithm.

Original languageEnglish (US)
Title of host publicationACC 2015 - 2015 American Control Conference
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages4497-4503
Number of pages7
Volume2015-July
ISBN (Electronic)9781479986842
DOIs
StatePublished - Jul 28 2015
Externally publishedYes
Event2015 American Control Conference, ACC 2015 - Chicago, United States
Duration: Jul 1 2015Jul 3 2015

Other

Other2015 American Control Conference, ACC 2015
CountryUnited States
CityChicago
Period7/1/157/3/15

Fingerprint

Convex optimization
Communication

ASJC Scopus subject areas

  • Electrical and Electronic Engineering

Cite this

Nedich, A., Lee, S., & Raginsky, M. (2015). Decentralized online optimization with global objectives and local communication. In ACC 2015 - 2015 American Control Conference (Vol. 2015-July, pp. 4497-4503). [7172037] Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ACC.2015.7172037

Decentralized online optimization with global objectives and local communication. / Nedich, Angelia; Lee, Soomin; Raginsky, Maxim.

ACC 2015 - 2015 American Control Conference. Vol. 2015-July Institute of Electrical and Electronics Engineers Inc., 2015. p. 4497-4503 7172037.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Nedich, A, Lee, S & Raginsky, M 2015, Decentralized online optimization with global objectives and local communication. in ACC 2015 - 2015 American Control Conference. vol. 2015-July, 7172037, Institute of Electrical and Electronics Engineers Inc., pp. 4497-4503, 2015 American Control Conference, ACC 2015, Chicago, United States, 7/1/15. https://doi.org/10.1109/ACC.2015.7172037
Nedich A, Lee S, Raginsky M. Decentralized online optimization with global objectives and local communication. In ACC 2015 - 2015 American Control Conference. Vol. 2015-July. Institute of Electrical and Electronics Engineers Inc. 2015. p. 4497-4503. 7172037 https://doi.org/10.1109/ACC.2015.7172037
Nedich, Angelia ; Lee, Soomin ; Raginsky, Maxim. / Decentralized online optimization with global objectives and local communication. ACC 2015 - 2015 American Control Conference. Vol. 2015-July Institute of Electrical and Electronics Engineers Inc., 2015. pp. 4497-4503
@inproceedings{76734847033f4701996162552bcae39d,
title = "Decentralized online optimization with global objectives and local communication",
abstract = "We consider a decentralized online convex optimization problem in a static undirected network of agents, where each agent controls only a coordinate (or a part) of the global decision vector. For such a problem, we propose a decentralized variant of Nesterov's primal-dual algorithm with dual averaging. 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 [1]. We show that the regret has sublinear growth of O (√T) with the time horizon T when the stepsize is of the form 1/√t and the objective functions are Lipschitzcontinuous convex functions with Lipschitz gradients. We prove an analogous bound on the expected regret for the stochastic variant of the algorithm.",
author = "Angelia Nedich and Soomin Lee and Maxim Raginsky",
year = "2015",
month = "7",
day = "28",
doi = "10.1109/ACC.2015.7172037",
language = "English (US)",
volume = "2015-July",
pages = "4497--4503",
booktitle = "ACC 2015 - 2015 American Control Conference",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
address = "United States",

}

TY - GEN

T1 - Decentralized online optimization with global objectives and local communication

AU - Nedich, Angelia

AU - Lee, Soomin

AU - Raginsky, Maxim

PY - 2015/7/28

Y1 - 2015/7/28

N2 - We consider a decentralized online convex optimization problem in a static undirected network of agents, where each agent controls only a coordinate (or a part) of the global decision vector. For such a problem, we propose a decentralized variant of Nesterov's primal-dual algorithm with dual averaging. 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 [1]. We show that the regret has sublinear growth of O (√T) with the time horizon T when the stepsize is of the form 1/√t and the objective functions are Lipschitzcontinuous convex functions with Lipschitz gradients. We prove an analogous bound on the expected regret for the stochastic variant of the algorithm.

AB - We consider a decentralized online convex optimization problem in a static undirected network of agents, where each agent controls only a coordinate (or a part) of the global decision vector. For such a problem, we propose a decentralized variant of Nesterov's primal-dual algorithm with dual averaging. 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 [1]. We show that the regret has sublinear growth of O (√T) with the time horizon T when the stepsize is of the form 1/√t and the objective functions are Lipschitzcontinuous convex functions with Lipschitz gradients. We prove an analogous bound on the expected regret for the stochastic variant of the algorithm.

UR - http://www.scopus.com/inward/record.url?scp=84940930889&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84940930889&partnerID=8YFLogxK

U2 - 10.1109/ACC.2015.7172037

DO - 10.1109/ACC.2015.7172037

M3 - Conference contribution

AN - SCOPUS:84940930889

VL - 2015-July

SP - 4497

EP - 4503

BT - ACC 2015 - 2015 American Control Conference

PB - Institute of Electrical and Electronics Engineers Inc.

ER -