K-means from a fixed initialization

Runs Lloyd's k-means from an explicitly supplied set of initial center indices, recording the full assignment trajectory. Because the initialization is fixed and no RNG is used, this is the loop-safe re-clustering routine used for per-step feasibility checks inside the MCMC samplers.

Usage

kmeans_with_fixed_init(x, init_centers_idx, max_iter = 50)

Arguments

x

A numeric data matrix with observations in rows.

init_centers_idx

Integer vector of row indices of x used as initial centers; its length determines k.

max_iter

Maximum number of Lloyd iterations.

Value

A list with elements cluster, trajectory, center_history, init_centers_idx, and n_iter.

Examples

set.seed(1)
X <- generate_exchangeable_null_data(30, 2, rho = 0.1)$dat
km <- kmeans_with_fixed_init(X, init_centers_idx = c(1, 11, 21))
km$n_iter
#> [1] 7