대한전기학회:학술대회논문집 (Proceedings of the KIEE Conference)

대한전기학회 (The Korean Institute of Electrical Engineers)
권호 표지

ART 모델의 multirun 횟수 감소에 관한 연구

A Study on decreasing the Number of Multirun in ART Model

  • 김미나 (이화여자대학교 전자계산학과) ;
  • 김도년 (이화여자대학교 전자계산학과) ;
  • 조동섭 (이화여자대학교 전자계산학과)
  • Kim, Mi-Na (Dept. of Computer Science, Ewha Womans University) ;
  • Kim, Do-Nyun (Dept. of Computer Science, Ewha Womans University) ;
  • Cho, Dong-Sub (Dept. of Computer Science, Ewha Womans University)
  • 발행 : 1995.07.20
PDF 다운로드
⟨ 이전 논문 다음 논문 ⟩

초록

The ART(Adaptive Resonance Theory) model is self- organized with nonstationary input patterns in real time. But there is a multirun problem caused by fault clustering, or pertubated clustering and confines the advantage of the stationary real-time processing in ART model. In this paper, we propose the incremental vigilance threshold approach to decrease the number of multiruns. The incremental vigilance threshold approach is to learn with incremental vigilance threshold and competition with clusters.

키워드