Convergence and applications of a gossip-based Gauss-Newton algorithm

Research output: Contribution to journalArticle

17 Citations (Scopus)

Abstract

The Gauss-Newton algorithm is a popular and efficient centralized method for solving non-linear least squares (NLLS) problems. In this paper, a multi-agent distributed version of this algorithm is proposed to solve general NLLS problems in a network, named Gossip-based Gauss-Newton (GGN) algorithm. Furthermore, we analyze and present sufficient conditions for its convergence and show numerically that the GGN algorithm achieves performance comparable to the centralized algorithm, with graceful degradation in case of network failures. More importantly, the GGN algorithm provides significant performance gains compared to other distributed first order methods.

Original languageEnglish (US)
Article number6574279
Pages (from-to)5231-5246
Number of pages16
JournalIEEE Transactions on Signal Processing
Volume61
Issue number21
DOIs
StatePublished - 2013
Externally publishedYes

Fingerprint

Degradation

Keywords

  • convergence
  • distributed
  • Gauss-Newton
  • gossip

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Signal Processing

Cite this

Convergence and applications of a gossip-based Gauss-Newton algorithm. / Li, Xiao; Scaglione, Anna.

In: IEEE Transactions on Signal Processing, Vol. 61, No. 21, 6574279, 2013, p. 5231-5246.

Research output: Contribution to journalArticle

@article{3bea96fdd77e4cf3b75316fc3b71fd6b,
title = "Convergence and applications of a gossip-based Gauss-Newton algorithm",
abstract = "The Gauss-Newton algorithm is a popular and efficient centralized method for solving non-linear least squares (NLLS) problems. In this paper, a multi-agent distributed version of this algorithm is proposed to solve general NLLS problems in a network, named Gossip-based Gauss-Newton (GGN) algorithm. Furthermore, we analyze and present sufficient conditions for its convergence and show numerically that the GGN algorithm achieves performance comparable to the centralized algorithm, with graceful degradation in case of network failures. More importantly, the GGN algorithm provides significant performance gains compared to other distributed first order methods.",
keywords = "convergence, distributed, Gauss-Newton, gossip",
author = "Xiao Li and Anna Scaglione",
year = "2013",
doi = "10.1109/TSP.2013.2276440",
language = "English (US)",
volume = "61",
pages = "5231--5246",
journal = "IEEE Transactions on Signal Processing",
issn = "1053-587X",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "21",

}

TY - JOUR

T1 - Convergence and applications of a gossip-based Gauss-Newton algorithm

AU - Li, Xiao

AU - Scaglione, Anna

PY - 2013

Y1 - 2013

N2 - The Gauss-Newton algorithm is a popular and efficient centralized method for solving non-linear least squares (NLLS) problems. In this paper, a multi-agent distributed version of this algorithm is proposed to solve general NLLS problems in a network, named Gossip-based Gauss-Newton (GGN) algorithm. Furthermore, we analyze and present sufficient conditions for its convergence and show numerically that the GGN algorithm achieves performance comparable to the centralized algorithm, with graceful degradation in case of network failures. More importantly, the GGN algorithm provides significant performance gains compared to other distributed first order methods.

AB - The Gauss-Newton algorithm is a popular and efficient centralized method for solving non-linear least squares (NLLS) problems. In this paper, a multi-agent distributed version of this algorithm is proposed to solve general NLLS problems in a network, named Gossip-based Gauss-Newton (GGN) algorithm. Furthermore, we analyze and present sufficient conditions for its convergence and show numerically that the GGN algorithm achieves performance comparable to the centralized algorithm, with graceful degradation in case of network failures. More importantly, the GGN algorithm provides significant performance gains compared to other distributed first order methods.

KW - convergence

KW - distributed

KW - Gauss-Newton

KW - gossip

UR - http://www.scopus.com/inward/record.url?scp=84884798449&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84884798449&partnerID=8YFLogxK

U2 - 10.1109/TSP.2013.2276440

DO - 10.1109/TSP.2013.2276440

M3 - Article

AN - SCOPUS:84884798449

VL - 61

SP - 5231

EP - 5246

JO - IEEE Transactions on Signal Processing

JF - IEEE Transactions on Signal Processing

SN - 1053-587X

IS - 21

M1 - 6574279

ER -