Laplacian eigenmaps for dimensionality reduction and data representation

被引:5770
作者
Belkin, M [1 ]
Niyogi, P
机构
[1] Univ Chicago, Dept Math, Chicago, IL 60637 USA
[2] Univ Chicago, Dept Comp Sci & Stat, Chicago, IL 60637 USA
关键词
D O I
10.1162/089976603321780317
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
One of the central problems in machine learning and pattern recognition is to develop appropriate representations for complex data. We consider the problem of constructing a representation for data lying on a low-dimensional manifold embedded in a high-dimensional space. Drawing on the correspondence between the graph Laplacian, the Laplace Beltrami operator on the manifold, and the connections to the heat equation, we propose a geometrically motivated algorithm for representing the high-dimensional data. The algorithm provides a computationally efficient approach to nonlinear dimensionality reduction that has locality-preserving properties and a natural connection to clustering. Some potential applications and illustrative examples are discussed.
引用
收藏
页码:1373 / 1396
页数:24
相关论文
共 19 条
[1]  
BELKIN M, 2002, ADV NEURAL INFORMATI, V14
[2]  
BERNSTEIN M., 2000, Technical Report
[3]  
Chung F., 2000, COMMUN ANAL GEOM, V8, P969, DOI [DOI 10.4310/CAG.2000.V8.N5.A2, 10.4310/CAG.2000.v8.n5.a2]
[4]  
Chung F. R. K., 1997, SPECTRAL GRAPH THEOR
[5]   AN EFFICIENT EIGENVECTOR APPROACH FOR FINDING NETLIST PARTITIONS [J].
HADLEY, SW ;
MARK, BL ;
VANNELLI, A .
IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS, 1992, 11 (07) :885-892
[6]  
Haykin S., 1999, Neural Networks: A Comprehensive Foundation, V2nd ed
[7]  
HENDRICKSON B, 1993, PROCEEDINGS OF THE SIXTH SIAM CONFERENCE ON PARALLEL PROCESSING FOR SCIENTIFIC COMPUTING, VOLS 1 AND 2, P953
[8]  
INDYK P, 2000, 11 S DISCR ALG SAN F
[9]  
KONDOR RI, 2002, P ICML 2002
[10]   THE IMBEDDING PROBLEM FOR RIEMANNIAN MANIFOLDS [J].
NASH, J .
ANNALS OF MATHEMATICS, 1956, 63 (01) :20-63