Locality preserving projections with adaptive neighborhood size

Wenjun Hu, Xinmin Cheng, Yunliang Jiang, Kup Sze Choi, Jungang Lou

Research output: Chapter in book / Conference proceedingConference article published in proceeding or bookAcademic researchpeer-review

2 Citations (Scopus)

Abstract

Feature extraction methods are widely employed to reduce dimensionality of data and enhance the discriminative information. Among the methods, manifold learning approaches have been developed to detect the underlying manifold structure of the data based on local invariants, which are usually guaranteed by an adjacent graph of the sampled data set. The performance of the manifold learning approaches is however affected by the locality of the data, i.e. what is the neighborhood size for suitably representing the locality? In this paper, we address this issue through proposing a method to adaptively select the neighborhood size. It is applied to the manifold learning approach Locality Preserving Projections (LPP) which is a popular linear reduction algorithm. The effectiveness of the adaptive neighborhood selection method is evaluated by performing classification and clustering experiments on the real-life data sets.
Original languageEnglish
Title of host publicationIntelligent Computing Theories and Application - 13th International Conference, ICIC 2017, Proceedings
PublisherSpringer Verlag
Pages223-234
Number of pages12
ISBN (Print)9783319633084
DOIs
Publication statusPublished - 1 Jan 2017
Event13th International Conference on Intelligent Computing, ICIC 2017 - Liverpool, United Kingdom
Duration: 7 Aug 201710 Aug 2017

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10361 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference13th International Conference on Intelligent Computing, ICIC 2017
CountryUnited Kingdom
CityLiverpool
Period7/08/1710/08/17

Keywords

  • Dimensionality reduction
  • Feature extraction
  • Locality preserving projections
  • Neighborhood size

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this