Computer Engineering and Applications ›› 2010, Vol. 46 ›› Issue (17): 176-178.DOI: 10.3778/j.issn.1002-8331.2010.17.050

• 图形、图像、模式识别 • Previous Articles     Next Articles

Incremental locally linear embedding algorithm based on inverse iteration method

ZHU Ming-han1,LUO Da-yong2   

  1. 1.Department of Electric Engineering,Hunan University of Arts and Science,Changde,Hunan 415000,China
    2.College of Information Science and Engineering,Central South University,Changsha 410083,China
  • Received:2008-12-02 Revised:2009-08-04 Online:2010-06-11 Published:2010-06-11
  • Contact: ZHU Ming-han

基于逆迭代的增量LLE算法

朱明旱1,罗大庸2   

  1. 1.湖南文理学院 电信学院,湖南 常德 415000
    2.中南大学 信息科学与工程学院,长沙 410083
  • 通讯作者: 朱明旱

Abstract: Locally Linear Embedding(LLE) is a sort of powerful manifold learning algorithm.However,LLE operates in a batch method.If only one new sample arrives,the whole algorithm must run repeatedly and all the former computational results are discarded.In this paper,an incremental locally linear embedding learning algorithm based on inverse iteration method is proposed.The proposed method can learn manifold in an incremental way.The experimental results in Swiss roll,S-curve databases show that the difference of the coordinates between the proposed method and the LLE method is less than 0.001%,and the executing time of the proposed method is much less than the LLE method.This implies that the proposed method is a promising method.

Key words: locally linear embedding, manifold learning, inverse iteration, increment

摘要: Locally Linear Embedding(LLE)算法是一种很好的流形学习算法,但是它只能以批处理的方式进行,只要有新的样本加入,就必须重作该算法的全部内容。而原来的运算结果被全部丢弃。提出了一种基于逆迭代的增量LLE算法,实现了流形的增量学习。在Swiss roll和S-curve数据库上的实验表明,该算法与LLE算法所计算出的投影值误差小于0.001%,运行的耗时少,具有很好的应用价值。

关键词: 局部线性嵌入, 流形学习, 逆迭代, 增量

CLC Number: