mv_splitting: Splitting of hard or soft clusters based on multi-view data

In andreamrau/maskmeans: Multi-view Aggregation and Splitting for K-Means Clustering

Description

Splitting of hard or soft clusters based on multi-view data

Usage

mv_splitting(
  X,
  mv,
  clustering_init,
  Kmax,
  gamma = 2,
  use_mv_weights = TRUE,
  delta = 2,
  perCluster_mv_weights = TRUE,
  verbose = TRUE,
  parallel = TRUE,
  BPPARAM = bpparam()
)

Arguments

X	Matrix of multi-view data, where the first view corresponds to the principal data used to obtain the partition or soft clustering in cluster_init
mv	(Optional unless X is a matrix.) If X is a matrix, vector corresponding to the size of each data view. The sum of mv should correspond to the number of columns in X.
clustering_init	Either a vector of available cluster labels (for hard clustering) or a matrix of soft classification labels (for soft clustering)
Kmax	Maximum number of clusters for splitting
gamma	Parameter that controls the distribution of view weights. Default value is 2.
use_mv_weights	If TRUE, run algorithm in weighted multi-view mode; if FALSE, the weight for each view is set to be equal. This option is only used for hard clustering.
delta	Parameter that controls the weights on the soft classifications pi(i,k)
perCluster_mv_weights	If TRUE, use cluster-specific multi-view weights. Otherwise use classic multi-view weights.
verbose	If TRUE, provide verbose output
parallel	If FALSE, no parallelization. If TRUE, parallel execution using BiocParallel (see next argument BPPARAM) for the soft splitting algorithm. A note on running in parallel using BiocParallel: it may be advantageous to remove large, unneeded objects from the current R environment before calling the function, as it is possible that R's internal garbage collection will copy these files while running on worker nodes.
BPPARAM	Optional parameter object passed internally to bplapply when parallel=TRUE. If not specified, the parameters last registered with register will be used.

Value

split_clusters	Matrix providing the history of each cluster splitting at each iteration of the algorithm
weights	Matrix of dimension v x niterations, where v is the number of views and niterations is the number of successive splits, providing the multi-view weights
criterion	Value taken on by the splitting criterion at each iteration
withnss	The within sum-of-squares for each cluster at the last iteration
ksplit	Vector identifying which cluster was split at each iteration of the algorithm
all_probapost	List of conditional probabilities for each split for the soft algorithm

andreamrau/maskmeans documentation built on Nov. 13, 2021, 7:44 a.m.