Cluster Shared Similarity From Reconstruction Weights

Usage

cluster_shared_similarity(
  recon,
  mode = c("average", "ao_reg"),
  dat_use = "ALL",
  k = NULL,
  method = c("Spectral", "PAM"),
  tune_method = "silhouette",
  gap_w = "uniform",
  similarity_type = c("second", "first"),
  gamma = 0.1,
  alpha_init = NULL,
  ao_max_iter = 20,
  ao_tol = 1e-04,
  knn_q = NULL,
  hollow = TRUE,
  ao_symm = TRUE,
  ao_verbose = FALSE,
  ...
)

Arguments

recon: Reconstruction output from get_reconstr_matrix().
mode: Shared clustering mode: "average" or "ao_reg".
dat_use: Shared weight source for "average" mode. Shared clustering now always uses W_all; this argument is kept for compatibility.
k: Optional number of clusters. Required for "ao_reg" mode.
method: Clustering backend for "average" mode.
tune_method: Tuning criterion for PAM in "average" mode when k is NULL.
gap_w: Weighting scheme for spectral eigengap in "average" mode when k is NULL.
similarity_type: Similarity structure used in "average" mode.
gamma: L2 regularization strength for "ao_reg" simplex weights.
alpha_init: Optional initial AO fusion weights.
ao_max_iter: Maximum AO iterations.
ao_tol: AO convergence tolerance.
knn_q: Optional kNN sparsification per base similarity in AO mode.
hollow: Logical; whether to zero diagonal in AO mode.
ao_symm: Logical; whether to symmetrize kNN sparsification in AO mode.
ao_verbose: Logical; whether to print AO diagnostics.
...: Additional arguments passed to clustering backends.

Value

A list with shared similarity, clustering result, and AO details (if used).