Intrinsic dimension estimation of data: An approach based on Grassberger-Procaccia's algorithm

被引：27

作者：

Camastra, F

Vinciarelli, A

机构：

[1] Elsag Spa, I-16154 Genoa, Italy

[2] IDIAP, CH-1920 Martigny, Switzerland

[3] Univ Genoa, Dept Comp Sci, I-16146 Genoa, Italy

来源：

NEURAL PROCESSING LETTERS | 2001年 / 14卷 / 01期

关键词：

correlation dimension; dimensionality estimation conjecture; Grassberger-Procaccia's algorithm; intrinsic dimension estimation; Multi-Layer Perceptron; Topology Representing Network;

D O I：

10.1023/A:1011326007550

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

In this paper the problem of estimating the intrinsic dimension of a data set is investigated. An approach based on the Grassberger-Procaccia's algorithm has been studied. Since this algorithm does not yield accurate measures in high-dimensional data sets, an empirical procedure has been developed. Grassberger-Procaccia's algorithm was tested on two different benchmarks and was compared to a TRN-based method.

引用

页码：27 / 34

页数：8

共 23 条

[1]

[Anonymous], 1993, NETWORKS CHAOS STAT

[2]

[Anonymous], 1982, Classification, Pattern Recognition and Reduction of Dimensionality, Handbook of Statistics

[3]

[Anonymous], 1988, GRUNDLEHREN MATH WIS

[4]

[Anonymous], 1993, CHAOS DYNAMICAL SYST