asPointwiseEstimator {intrinsicDimension}

Turn a local estimator into a pointwise estimator

Description

Returns a function that can be used as a pointwise estimator of intrinsic dimension that uses local data sets with a fixed number of data points.

Usage

asPointwiseEstimator(estimator, neighborhood.size, indices=NULL, eps=0.0)

Arguments

estimator

A local intrinsic dimension estimator.

neighborhood.size

The number of neighbors used for each dimension estimate.

indices

A vector with indices of the points in data (as sent to the estimator function) that should be used as center for neighborhoods.

eps

If non-zero, the relative error in distance allowed when finding nearest neighbors. See Details.

Details

The ann function of the package yaImpute is used for finding the k nearest neighbors. The eps parameter to neighborhoods is used in the ann function.

Value

A function that can be used as a pointwise dimension estimator.

Author(s)

Kerstin Johnsson, Lund University

Examples

data <- swissRoll3Sph(300, 300) 
 # the first 300 data points are on the swiss roll (ID=2) , the last 300 on the 3-sphere (ID=3)
essPointwiseDimEst <- asPointwiseEstimator(essLocalDimEst, neighborhood.size=10,
                                           indices = c(1:10, 301:310))
ess.pw.res <- essPointwiseDimEst(data)
ess.pw.res$dim.est