CaltechAUTHORS
  A Caltech Library Service

Logarithmic Communication for Distributed Optimization in Multi-Agent Systems

London, Palma and Vardi, Shai and Wierman, Adam (2019) Logarithmic Communication for Distributed Optimization in Multi-Agent Systems. Proceedings of the ACM on Measurement and Analysis of Computing Systems, 3 (3). Art. No. 48. ISSN 2476-1249. https://resolver.caltech.edu/CaltechAUTHORS:20191218-160307829

[img] PDF - Published Version
See Usage Policy.

1198Kb

Use this Persistent URL to link to this item: https://resolver.caltech.edu/CaltechAUTHORS:20191218-160307829

Abstract

Classically, the design of multi-agent systems is approached using techniques from distributed optimization such as dual descent and consensus algorithms. Such algorithms depend on convergence to global consensus before any individual agent can determine its local action. This leads to challenges with respect to communication overhead and robustness, and improving algorithms with respect to these measures has been a focus of the community for decades. This paper presents a new approach for multi-agent system design based on ideas from the emerging field of local computation algorithms. The framework we develop, LOcal Convex Optimization (LOCO), is the first local computation algorithm for convex optimization problems and can be applied in a wide-variety of settings. We demonstrate the generality of the framework via applications to Network Utility Maximization (NUM) and the distributed training of Support Vector Machines (SVMs), providing numerical results illustrating the improvement compared to classical distributed optimization approaches in each case.


Item Type:Article
Related URLs:
URLURL TypeDescription
https://doi.org/10.1145/3366696DOIArticle
https://resolver.caltech.edu/CaltechAUTHORS:20200709-141943501Related ItemJournal Article
Additional Information:© 2019 Association for Computing Machinery. OPen Access. Received August 2019; revised September 2019; accepted October 2019. This work was funded by the National Research Foundation through the AitF-1637598, CNS-1518941, and CNS-1254169 grants, along with the Linde Foundation and an Amazon AWS Artificial Intelligence Fellowship.
Funders:
Funding AgencyGrant Number
NSFAitF-1637598
NSFCNS-1518941
NSFCNS-1254169
Linde Institute of Economic and Management ScienceUNSPECIFIED
Amazon Web ServicesUNSPECIFIED
Subject Keywords:distributed algorithms; distributed optimization; multi-agent systems
Issue or Number:3
Record Number:CaltechAUTHORS:20191218-160307829
Persistent URL:https://resolver.caltech.edu/CaltechAUTHORS:20191218-160307829
Official Citation:Palma London, Shai Vardi, and Adam Wierman. 2019. Logarithmic Communication for Distributed Optimization in Multi-Agent Systems. Proc. ACM Meas. Anal. Comput. Syst. 3, 3, Article 48 (December 2019), 29 pages. https://doi.org/10.1145/3366696
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:100368
Collection:CaltechAUTHORS
Deposited By: Tony Diaz
Deposited On:19 Dec 2019 00:19
Last Modified:09 Jul 2020 21:46

Repository Staff Only: item control page