Diffusion bias-compensated LMS estimation for multitask adaptive networks

Xiaoling Xu, Lijuan Jia, Tingting Xu, Hui Wan, Kanae Shunshoku

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

2 Citations (Scopus)

Abstract

In this paper, we study the problem of the least mean-square algorithm based on bias compensation in multitask diffusion adaptive networks. Nodes in networks are divided into different clusters and the nodes in the same cluster cooperatively estimate a common parameter. When regressors are corrupted by additive white noise, the estimate results of the traditional multitask diffusion least mean-square (Multi-LMS) algorithm are biased. In order to obtain the unbiased estimation, we propose two multitask diffusion bias-compensated least mean-square (Multi-BCLMS) algorithms by achieving the real-time estimation of the input noise variance, which can be denoted by Multi-BCLMS-A and Multi-BCLMS-B respectively. Simulation results show that the two algorithms perform better than the Multi-LMS algorithm in estimation accuracy and mean-square error. Furthermore, the second algorithm (Multi-BCLMS-B) is simpler to implement and the transient is faster than the first one (Multi-BCLMS-A).

Original languageEnglish
Title of host publication2015 IEEE Conference on Control and Applications, CCA 2015 - Proceedings
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages545-550
Number of pages6
ISBN (Electronic)9781479977871
DOIs
Publication statusPublished - 4 Nov 2015
EventIEEE Conference on Control and Applications, CCA 2015 - Sydney, Australia
Duration: 21 Sept 201523 Sept 2015

Publication series

Name2015 IEEE Conference on Control and Applications, CCA 2015 - Proceedings

Conference

ConferenceIEEE Conference on Control and Applications, CCA 2015
Country/TerritoryAustralia
CitySydney
Period21/09/1523/09/15

Fingerprint

Dive into the research topics of 'Diffusion bias-compensated LMS estimation for multitask adaptive networks'. Together they form a unique fingerprint.

Cite this