kNN: Dimension Estimation from kNN Distances

In intrinsicDimension: Intrinsic Dimension Estimation

Description

Estimates the intrinsic dimension of a data set using weighted average kNN distances.

Usage

knnDimEst(data, k, ps, M, gamma = 2)

Arguments

data	data set with each row describing a data point.
k	number of distances to neighbors used at a time.
ps	vector with sample sizes; each sample size has to be larger than k and smaller than nrow(data).
M	number of bootstrap samples for each sample size.
gamma	weighting constant.

Details

This is a somewhat simplified version of the kNN dimension estimation method described by Carter et al. (2010), the difference being that block bootstrapping is not used.

Value

A DimEst object with slots:

dim.est	the intrinsic dimension estimate (integer).
residual	the residual, see Carter et al. (2010).

Author(s)

Kerstin Johnsson, Lund University.

References

Carter, K.M., Raich, R. and Hero, A.O. (2010) On local intrinsic dimension estimation and its applications. IEEE Trans. on Sig. Proc., 58(2), 650-663.

Examples

N <- 50
data <- hyperBall(N, 5)

k <- 2
ps <- seq(max(k + 1, round(N/2)), N - 1, by = 3)
knnDimEst(data, k, ps, M = 10, gamma = 2)

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