bs: Dimension Estimation With Optimally Topology Preserving Maps

In intrinsicDimension: Intrinsic Dimension Estimation

Description

Intrinsic dimension estimation with the method proposed in Bruske and Sommer (1998). A graph called optimally topology preserving map (OTPM) is constructed and on this local PCA is made with the Fukunaga-Olsen criterion to determine which eigenvalues that are significant.

Usage

 pcaOtpmPointwiseDimEst(data, N, alpha = .05) 

Arguments

data	a data set for which dimension should be estimated.
N	the number of the nodes in the OTPM.
alpha	the significance level for the Fukunaga-Olsen method.

Value

A DimEstPointwise object, inheriting data.frame, with two columns:

dim.est	The dimension estimate at each point.
nbr.nb	The number of neighboring nodes used for the dimension estimate at each point.

Author(s)

Kerstin Johnsson, Lund University

References

Bruske, J. and Sommer, G. (1998) Intrinsic dimensionality estimation with optimally topology preserving maps. IEEE Trans. on Pattern Anal. and Mach. Intell., 20(5), 572-575.

See Also

pcaLocalDimEst

Examples

data <- hyperBall(1000, 5)
pcaOtpmPointwiseDimEst(data, 400)

Example output

Loading required package: yaImpute
Dimension estimates at 1000 data points.
 min: 1 ; max: 8 
Additional data: nbr.neighbors 

intrinsicDimension documentation built on June 7, 2019, 5:02 p.m.