错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Stochastic Neighborhood Embedding

  • Wolfgang Karl Härdle,
  • Léopold Simar,
  • Matthias R. Fengler

摘要

LLE is widely appreciated as an effective dimension-reduction tool. It fares less well, however, in situations of very high-dimensional data sets. Moreover, an essential assumption of LLE is the presence of a single smooth manifold in the data. If the data contain multiple manifolds or if there are regions, where the density of observed data varies a lot, LLE has performance weaknesses.