广东工业大学学报 ›› 2015, Vol. 32 ›› Issue (3): 39-45.doi: 10.3969/j.issn.1007-7162.2015.03.008

• 综合研究 • 上一篇    下一篇

基于最小熵的流形学习排列方法

陈静1,刘洋2   

  1. 广东工业大学 1.物理与光电工程学院;2.信息工程学院,广东 广州 510006
  • 收稿日期:2014-05-13 出版日期:2015-09-22 发布日期:2015-09-22
  • 作者简介:陈静(1980-),女,副教授,博士,主要研究方向为机器学习.
  • 基金资助:

    国家自然科学基金资助项目(61305069);广东省自然科学基金资助项目(S2013040016371)

The Minimum Entropy Alignment in Manifold Learning

Chen Jing1, Liu Yang2   

  1. 1. School of Physics and Optoelectronic Engineering; 2. School of Information Engineering, Guangdong University of Technology, Guangzhou 510006, China
  • Received:2014-05-13 Online:2015-09-22 Published:2015-09-22

摘要: 提出一种渐进式的排列方法:每次只是排列当前阶段交集最大的二个分片的局部坐标.该方法具有以下特点:方法简单,避免了全局排列方法中大型稀疏矩阵的特征值问题;每次排列都保证是误差最小的排列;减轻误差的积累和传播.

关键词: 流形学习; 降维; 排列

Abstract: A new progressive alignment method is proposed, while only the local coordinates of the two patches with the largest intersection at the current stage of progressive alignment will be aligned into a larger local coordinate. The method has the following features: without needing to solve a largescale eigenvalues problem and suffering from the problem of local minima; less accumulation and propagation of alignment errors.

Key words: manifold learning; dimensionality reduction; alignment

No related articles found!
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!