An Improved Isomap Algorithm Based on LP-Centers

DOI

10.2991/aiie-15.2015.73How to use a DOI?

Keywords

Isomap; linear patch (LP); k-nearest center (k-NC); LPC-Isomap

Abstract

The isometric feature mapping (Isomap) is one of the widely used low-dimensional embedding methods. However, Isomap has three bottlenecks: sensitivity to outliers, high computational complexity and topological instability caused by short-circuit edges. To address above problems, by introducing the concepts of linear patch (LP) and k-nearest center (k-NC), a LP-center based Isomap (LPC-Isomap) approach is proposed in this paper. Firstly, before constructing a neighborhood graph, outliers are eliminated by LPC-Isomap. Then LP-centers are selected reasonably as landmark points by LPC-Isomap to reduce computational complexity. We compute the intersections between k-NC of each pair of points. If an intersection is empty, the edge connecting the corresponding pair of points is considered as a short-circuit edge. Lastly, the useful behavior of LPC-Isomap is verified through numerical examples with several data sets.

Copyright

© 2015, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - W.W. Chen
AU  - W.J. Mao
PY  - 2015/07
DA  - 2015/07
TI  - An Improved Isomap Algorithm Based on LP-Centers
BT  - Proceedings of the 2015 International Conference on Artificial Intelligence and Industrial Engineering
PB  - Atlantis Press
SP  - 260
EP  - 263
SN  - 1951-6851
UR  - https://doi.org/10.2991/aiie-15.2015.73
DO  - 10.2991/aiie-15.2015.73
ID  - Chen2015/07
ER  -