hausdorffDistPlot: Create diagnostic plot of Hausdorff distances

View source: R/hausdorffDistPlot.R

hausdorffDistPlotR Documentation

Create diagnostic plot of Hausdorff distances

Description

Create diagnostic plot showing the Hausdorff distance between a sketch and the full data set, for varying sketch sizes. For reproducibility, seed the random number generator before calling this function using set.seed.

Usage

hausdorffDistPlot(
  mat,
  Nvec,
  Nrep = 5,
  q = 1e-04,
  methods = c("geosketch", "scsampler", "uniform"),
  extraArgs = list()
)

Arguments

mat

m x n matrix. Samples (the dimension along which to subsample) should be in the rows, features in the columns.

Nvec

Numeric vector of sketch sizes.

Nrep

Numeric scalar indicating the number of sketches to draw for each sketch size.

q

Numeric scalar in [0,1], indicating the fraction of largest minimum distances to discard when calculating the robust Hausdorff distance. Setting q=0 gives the classical Hausdorff distance. The default is 1e-4, as suggested by Hie et al (2019).

methods

Character vector, indicating which method(s) to include in the plot. Should be a subset of c("geosketch", "scsampler", "uniform"), where "uniform" randomly samples from input features with uniform probabilities.

extraArgs

Named list providing extra arguments to the respective methods (beyond the matrix and the sketch size). The names of the list should be the method names (currently, "geosketch" or "scsampler"), and each list element should be a named list of argument values. See the examples for an illustration of how to use this argument. Note that the seed argument, if provided to any of the methods, will be ignored (since it would imply providing the same seed for each repeated run of the sketching).

Value

A ggplot object.

Author(s)

Charlotte Soneson, Michael Stadler

References

Hie et al (2019): Geometric sketching compactly summarizes the single-cell transcriptomic landscape. Cell Systems 8, 483–493.

Song et al (2022): scSampler: fast diversity-preserving subsampling of large-scale single-cell transcriptomic data. bioRxiv doi:10.1101/2022.01.15.476407

Huttenlocher et al (1993): Comparing images using the Hausdorff distance. IEEE Transactions on Pattern Analysis and Machine Intelligence 15(9), 850-863.

Examples

## Generate example data matrix
mat <- matrix(rnorm(1000), nrow = 100)

## Generate diagnostic Hausdorff distance plot
## (including all available methods)
hausdorffDistPlot(mat, Nvec = c(10, 25, 50))

## Provide additional arguments for geosketch
hausdorffDistPlot(mat, Nvec = c(10, 25, 50), Nrep = 2,
                  extraArgs = list(geosketch = list(max_iter = 100)))


csoneson/sketchR documentation built on Dec. 9, 2024, 4:45 p.m.