coarsen_one_level_localvar: Single-level local variation coarsening step (Edge-based)

In bbuchsbaum/graphweights: Constructing Adjacency Matrices Based on Spatial and Feature Similarity

Single-level local variation coarsening step (Edge-based)

Description

Performs one coarsening step: - Computes edge costs - Selects edges to contract (if none meet threshold, force one contraction if force_contraction=TRUE) - Constructs P, updates L and U

Usage

coarsen_one_level_localvar(
  L,
  U,
  n,
  sigma_prime,
  force_contraction = FALSE,
  verbose = FALSE
)

Arguments

L	Current Laplacian.
U	Current subspace U (N x k).
n	Target size after coarsening (N_new <= n ideally).
sigma_prime	Variation threshold.
force_contraction	If TRUE, force at least one edge contraction if none selected normally.
verbose	If TRUE, prints messages.

Value

A list:

L_new	Reduced Laplacian.
P	Coarsening matrix.
U_new	Updated subspace.
N_new	New dimension.
sigma	Approx. variation cost (for epsilon update).

bbuchsbaum/graphweights documentation built on Dec. 14, 2024, 1:13 a.m.