γ-Observable neighbours for vector quantization

Michael Aupetit, Pierre Couturier, Pierre Massotte

Research output: Contribution to journalArticle

5 Citations (Scopus)

Abstract

We define the γ-observable neighbourhood and use it in soft-competitive learning for vector quantization. Considering a datum v and a set of n units wi in a Euclidean space, let vi be a point of the segment [vwi] whose position depends on γ a real number between 0 and 1, the γ-observable neighbours (γ-ON) of v are the units wi for which vi is in the Voronoï of wi, i.e. wi is the closest unit to vi. For γ=1, vi merges with wi, all the units are γ-ON of v, while for γ=0, vi merges with v, only the closest unit to v is its γ-ON. The size of the neighbourhood decreases from n to 1 while γ goes from 1 to 0. For γ lower or equal to 0.5, the γ-ON of v are also its natural neighbours, i.e. their Voronoï regions share a common boundary with that of v. We show that this neighbourhood used in Vector Quantization gives faster convergence in terms of number of epochs and similar distortion than the Neural-Gas on several benchmark databases, and we propose the fact that it does not have the dimension selection property could explain these results. We show it also presents a new self-organization property we call 'self-distribution'.

Original languageEnglish
Pages (from-to)1017-1027
Number of pages11
JournalNeural Networks
Volume15
Issue number8-9
DOIs
Publication statusPublished - Oct 2002
Externally publishedYes

    Fingerprint

Keywords

  • γ-Observable neighbours
  • Dimension selection
  • Natural neighbours
  • Neural-gas
  • Self-distribution
  • Self-organizing maps
  • Vector quantization

ASJC Scopus subject areas

  • Artificial Intelligence
  • Neuroscience(all)

Cite this