Earticle

초록

영어

In terms of machine learning theory, the intrinsic geometrical structure of the original data space is usually embedded in the low-dimensional manifold. The extraction of optimized manifold features could improve the performance of clustering. This paper presents a new spectral clustering method called local topology preserving indexing (LTPI). In this algorithm, the data are projected into a low-dimensional feature space in which the distances between the data points in the same local patches are minimized and the distances from the data points outside these patches are maximized simultaneously. The proposed LTPI method can effectively discover the intrinsic local topologies embedded in original high-dimensional space. The comparison experiments for document clustering demonstrate its effectiveness.

목차

Abstract
 1. Introduction
 2. Spectral Clustering based on LTPI
  2.1 Constructing Graphs and Weights
  2.2 Graph Preserving Criterion
  2.3 Algorithm Derivation
  2.4 Clustering Algorithm with LTPI
 3. Experiments on Document Clustering
 4. Conclusion
 Acknowledgements
 References

키워드

저자

• Jieqing Xing [ Department of Information Technology, Qiongtai Teachers College ]
• Chuanyi Fu [ Department of Information Technology, Qiongtai Teachers College ]

참고문헌

발행기관

간행물

표지보기 관심저널 등록

• 간행물명

  International Journal of Hybrid Information Technology

• 간기

  격월간

• pISSN

  1738-9968

• 수록기간

  2008~2016

• 십진분류

  KDC 505 DDC 605