graphClustering {graphclust}

Hierarchical graph clustering algorithm

Description

Applies the hierarchical graph clustering algorithm to a collection of networks and fits a finite mixture model of stochastic block models to the data

Usage

graphClustering(
  allAdj,
  hyperParam = list(alpha = 0.5, eta = 0.5, zeta = 0.5, lambda = 0.5),
  returnInitial = FALSE,
  nbClust = NULL,
  nbSBMBlocks = Inf,
  initCountStat = NULL,
  initDeltaICL = NULL,
  nbCores = 1
)

Arguments

allAdj

list of adjacency matrices

hyperParam

hyperparameters of prior distributions

returnInitial

Boolean. Return SBM parameters from initialization or not. Default is FALSE.

nbClust

desired number of clusters. Default NULL, which means that the number of clusters is chosen automatically via the ICL criterion

nbSBMBlocks

upper bound for the number of blocks in the SBMs of the mixture components. Default is Inf

initCountStat

initial count statistics may be provided to the method. Default is NULL.

initDeltaICL

initial deltaICL-matrix may be provided to the method. Default is NULL.

nbCores

number of cores for parallelization

Value

list with the following fields: $graphGroups is the graph clustering, $nodeClusterings is a list with the node labels for each networks, $thetaMixSBM contains the estimated parameter of the mixture of SBMs, $ICL is the value of the ICL criterion of the final clustering, $histGraphGroups traces the history of the cluster aggregations, $histDeltaICL traces the evolution of the deltaICL value, $histFusedClusters traces the history of the aggregated cluster numbers

Examples

theta <- list(pi=c(.5,.5), gamma=matrix((1:4)/8,2,2))
obs <- rCollectSBM(rep(10,4), theta)$listGraphs
res <- graphClustering(obs, nbCores=1)