Application of fast incremental LLE to bearing fault feature dimension reduction

Chengliang Li, Zhongsheng Wang, Hongkai Jiang

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

Abstract

Focus on incremental local linear embedding (LLE) operation efficiency problem, this paper proposes a fast incremental LLE algorithm. Firstly, we describe briefly the basic principle of LLE algorithm. Secondly, based on the incremental learning principle, the new samples are added, the global coordinates of affect samples are recomputed. Thirdly, the low dimensional embedding coordinates of the incremental samples are formulated by the updated global coordinate matrix and low dimensional embedding coordinates of the given samples, then, using Rayleigh-Ritz accelerated iterative algorithm calculate the global coordinate update. Experiment results show that the proposed algorithm can fastly establish the low dimensional feature for new samples.

Original languageEnglish
Title of host publicationProceedings - 2012 6th International Conference on New Trends in Information Science, Service Science and Data Mining (NISS, ICMIA and NASNIT), ISSDM 2012
Pages423-426
Number of pages4
StatePublished - 2012
Event2012 6th International Conference on New Trends in Information Science, Service Science and Data Mining (NISS, ICMIA and NASNIT), ISSDM 2012 - Taipei, Taiwan, Province of China
Duration: 23 Oct 201225 Oct 2012

Publication series

NameProceedings - 2012 6th International Conference on New Trends in Information Science, Service Science and Data Mining (NISS, ICMIA and NASNIT), ISSDM 2012

Conference

Conference2012 6th International Conference on New Trends in Information Science, Service Science and Data Mining (NISS, ICMIA and NASNIT), ISSDM 2012
Country/TerritoryTaiwan, Province of China
CityTaipei
Period23/10/1225/10/12

Keywords

  • dimension reduction
  • Incremental LLE
  • Rayleigh-Ritz

Fingerprint

Dive into the research topics of 'Application of fast incremental LLE to bearing fault feature dimension reduction'. Together they form a unique fingerprint.

Cite this