Transport cluster values to a medoid via entropic Sinkhorn OT

Usage

dkge_transport_to_medoid_sinkhorn(
  v_list,
  A_list,
  centroids,
  sizes = NULL,
  medoid,
  lambda_emb = 1,
  lambda_spa = 0.5,
  sigma_mm = 15,
  epsilon = 0.05,
  max_iter = 200,
  tol = 1e-06,
  transport_cache = NULL
)

dkge_transport_to_medoid_sinkhorn_cpp(
  v_list,
  A_list,
  centroids,
  sizes = NULL,
  medoid,
  lambda_emb = 1,
  lambda_spa = 0.5,
  sigma_mm = 15,
  epsilon = 0.05,
  max_iter = 300,
  tol = 1e-06,
  return_plans = FALSE,
  transport_cache = NULL
)

Arguments

v_list: List of subject-level cluster values (length P_s each).
A_list: List of subject loadings (P_s x r).
centroids: List of subject cluster centroids (each P_s x 3 matrix).
sizes: Optional list of cluster masses (defaults to uniform weights).
medoid: Integer index of the reference subject (1-based).
lambda_emb, lambda_spa: Cost weights for embedding and spatial terms.
sigma_mm: Spatial rescaling (millimetres).
epsilon, max_iter, tol: Sinkhorn parameters.
transport_cache: Optional cache returned by [dkge_prepare_transport()]. When supplied, cached operators are reused and the mapper configuration stored in the cache takes precedence.
return_plans: Logical; if TRUE, include transport plans in the output.

Value

List containing summary statistics, transported subject maps, and the per-subject transport operators.