Generate Multiple Clusters

This function generates a dataset with multiple clusters in high-dimensional space. Each cluster can have a different shape, scale, rotation, and centroid, allowing the construction of complex synthetic datasets.

Usage

gen_multicluster(
  n = c(200, 300, 500),
  k = 3,
  loc = matrix(c(0, 0, 0, 0, 5, 9, 0, 0, 3, 4, 10, 7), nrow = 3, byrow = TRUE),
  scale = c(3, 1, 2),
  shape = c("gaussian", "bluntedcorn", "unifcube"),
  rotation = NULL,
  is_bkg = FALSE,
  ...
)

Arguments

n

A numeric vector (default: c(200, 500, 300)) representing the sample sizes for each cluster. Must have length k.

k

A numeric value (default: 3) representing the number of clusters.

loc

A numeric matrix giving the centroids of the clusters. The number of rows must equal k; the number of columns should match the maximum dimensionality across all shapes.

scale

A numeric vector (default: c(3, 1, 2)) giving the scaling factors for each cluster. Must have length k.

shape

A character vector (default: c("gaussian", "cone", "unifcube")) specifying the generator function to use for each cluster. Must have length k.

rotation

A list of rotation matrices (one per cluster), or NULL. Each matrix must be square with dimension equal to the total number of structural dimensions in the dataset. Rotation matrices can be generated using gen_rotation for convenience.

is_bkg

Logical (default: FALSE). If TRUE, adds background noise sampled from a multivariate normal distribution centered on the dataset mean with standard deviations matching the observed spread.

...

Additional arguments passed to the cluster generator functions.

Value

A tibble containing all generated clusters, with columns x1, x2, ... for dimensions and a cluster label.

Examples

set.seed(20240412)

# Example rotation matrices for 4D space
rot1 <- gen_rotation(p = 4, planes_angles = list(list(plane = c(1, 2), angle = 60),
                                                list(plane = c(3, 4), angle = 90)))
rot2 <- gen_rotation(p = 4, planes_angles = list(list(plane = c(1, 3), angle = 30)))
rot3 <- gen_rotation(p = 4, planes_angles = list(list(plane = c(2, 4), angle = 45)))

clust_data <- gen_multicluster(
  n = c(200, 300, 500),
  k = 3,
  loc = matrix(c(
    0, 0, 0, 0,
    5, 9, 0, 0,
    3, 4, 10, 7
  ), nrow = 3, byrow = TRUE),
  scale = c(3, 1, 2),
  shape = c("gaussian", "cone", "unifcube"),
  rotation = list(rot1, rot2, rot3),
  is_bkg = FALSE
)
#> ✔ Data generation completed successfully!!!
#> ✔ Data generation completed successfully!!!
#> ✔ Data generation completed successfully!!!
#> ✔ Multiple clusters generation completed successfully!!!