Global parameters estimation and convergence proof of isomorphic networks using historical data

Zhenyu Lu, Panfeng Huang

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In recent years, the wireless sensors networks raise a great attention in the world. In this paper we proposed a method - multi-innovation coupled stochastic gradient (MICSG) algorithm for the global parameters estimation of the distributed sensors. This algorithm utilizes the identified result of the previous adjacent node and the local historical data to modify own estimated parameters. Then we make a proof of parameters convergence of proposed algorithm. Two examples are presented in the simulation. The first example concerns the influence of different length of historical data to the convergence rate and error rate. The second one exhibits the method applying the structure healthy management. Simulation shows that increasing the length of multi-innovation vector can improve the convergence effect and accelerate the convergence rate in a certain range.

Original languageEnglish
Title of host publicationProceedings of 2014 International Conference on Multisensor Fusion and Information Integration for Intelligent Systems, MFI 2014
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781479967322
DOIs
StatePublished - 23 Dec 2014
Event2014 International Conference on Multisensor Fusion and Information Integration for Intelligent Systems, MFI 2014 - Beijing, China
Duration: 28 Sep 201430 Sep 2014

Publication series

NameProceedings of 2014 International Conference on Multisensor Fusion and Information Integration for Intelligent Systems, MFI 2014

Conference

Conference2014 International Conference on Multisensor Fusion and Information Integration for Intelligent Systems, MFI 2014
Country/TerritoryChina
CityBeijing
Period28/09/1430/09/14

Fingerprint

Dive into the research topics of 'Global parameters estimation and convergence proof of isomorphic networks using historical data'. Together they form a unique fingerprint.

Cite this