Test All Clustering Methods on Synthetic Data

Comprehensive test of all available clustering methods on synthetic ground truth data with cubic clusters.

Usage

test_all_clustering_methods(
  noise_levels = c(0, 0.1, 0.5),
  methods = c("flash3d", "corr_slic", "snic", "slic", "slice_msf", "supervoxels", "g3s",
    "rena"),
  n_clusters = 27,
  seed = 42
)

Arguments

noise_levels

Vector of noise standard deviations to test. Default c(0, 0.1, 0.5) tests perfect, low noise, and moderate noise.

methods

Vector of method names to test. Default tests all main methods.

n_clusters

Number of ground truth clusters (must be perfect cube). Default 27.

seed

Random seed for reproducibility.

Value

A data.frame with columns: method, noise_sd, ari, nmi, accuracy, n_clusters, time, perfect

Examples

if (FALSE) { # \dontrun{
# Test all methods with 27 clusters
results <- test_all_clustering_methods()
print(results)

# Test with 8 clusters (simpler)
results <- test_all_clustering_methods(n_clusters = 8)

# Test specific methods
results <- test_all_clustering_methods(
  methods = c("flash3d", "corr_slic", "snic"),
  noise_levels = c(0, 0.2)
)
} # }