intrinsicDimension (version 1.2.0)

dancoDimEst: Dimension Estimation With the DANCo and MIND Methods

Description

Intrinsic dimension estimation with the DANCo (Ceruti et al. 2012), MIND_MLi and MIND_MLk (Rozza et al. 2012) methods.

Usage

dancoDimEst(data, k, D, ver = "DANCo", calibration.data = NULL)

Arguments

data

a data set for which the intrinsic dimension is estimated.

k

neighborhood parameter.

D

maximal dimension.

ver

possible values: 'DANCo', 'MIND_MLi', 'MIND_MLk'.

calibration.data

precomputed calibration data.

Value

A DimEst object with slots:

dim.est

the intrinsic dimension estimate.

kl.divergence

the KL divergence between data and reference data for the estimated dimension (if ver == 'DANCo').

calibration.data

calibration data that can be reused when applying DANCo to data sets of the same size with the same neighborhood parameter k.

%% ...

Details

If cal = NULL or the cal$maxdim < D new calibration data will be computed as needed.

References

Ceruti, C. et al. (2012) DANCo: Dimensionality from Angle and Norm Concentration. arXiv preprint 1206.3881.

Rozza, A et al. (2012) Novel high intrinsic dimensionality estimators. Machine learning 89, 37-65.

Examples

# NOT RUN {
data <- hyperBall(50, 10)
res <- dancoDimEst(data, 8, 20)
print(res)

## Reusing calibration data
data2 <- hyperBall(50, 5)
dancoDimEst(data2, 8, 20, calibration.data=res$calibration.data)
# }