Time-Delay ART for spatio-temporal patterns

研究成果: Article査読

4 被引用数 (Scopus)

抄録

This paper proposes a new self-organizing neural network for spatio-temporal patterns called Time-Delay ART (TD-ART). The TD-ART is a nearest neighbor classifier that stores arbitrary length of spatio-temporal patterns. It has at least three layers. The Layer-1 and the Layer-2 classify input spatial patterns one by one: the algorithm is based on the Carpenter/Grossberg net algorithm. The Layer-2 and the Layer-3 classify and memorize the sequence of the spatial patterns classified in the Layer-2. The connections between the Layer-2 and the Layer-3 are adjusted both in weight and in time-delay to deal with spatio-temporal patterns. Although the construction and operation of TD-ART are very simple, it has many features such as its self-organization, classification, and memorization abilities for spatio-temporal patterns. In addition, the TD-ART can distinguish different sequences composed of same patterns such as A → C → T and C → A → T by unsupervised learning, and can deal with many sequences of different lengths.

本文言語English
ページ(範囲)513-521
ページ数9
ジャーナルNeurocomputing
6
5-6
DOI
出版ステータスPublished - 1994 10

ASJC Scopus subject areas

  • コンピュータ サイエンスの応用
  • 認知神経科学
  • 人工知能

フィンガープリント

「Time-Delay ART for spatio-temporal patterns」の研究トピックを掘り下げます。これらがまとまってユニークなフィンガープリントを構成します。

引用スタイル