FitSBM function

An R6 Class object for unilevel network

a fitted level of a unilevel network once $do_vem() is done

Public fields

• vbound: vector of variational bound for convergence monitoring

Active bindings

• adjacency: Get the adjacency matrix

• mask: Get the mask matrix for dealing with NA

• nb_nodes: Get the number of nodes of the level

• nb_clusters: Get the number of blocks

• distribution: Get the distribution used for the connections

• directed: Get if the level is directed or not

• mixture_parameter: Access the block proportions

• connectivity_parameter: Access the connectivity matrix

• membership: Access the variational parameters

• entropy: Get the entropy of the model

• bound: Get the variational bound of the model

• df_mixture: Get the degree of freedom of the block proportion

• df_connect: Get the degree of freedom of the connection parameters

• connect: Get the number of observed dyads

• ICL: Get the ICL model selection criterion

• penalty: Get the penalty used for computing the ICL

• Z: Access the vector of block membership (clustering)

• X_hat: Get the connection probability matrix

• X_likelihood: adjacency part of the log likelihood

• Z_likelihood: block part of the log likelihood

• likelihood: complete log likelihood

Methods

Public methods

• list("FitSBM$new()")

• list("FitSBM$update_alpha()")

• list("FitSBM$update_pi()")

• list("FitSBM$init_clustering()")

• list("FitSBM$m_step()")

• list("FitSBM$ve_step()")

• list("FitSBM$do_vem()")

• list("FitSBM$permute_empty_class()")

• list("FitSBM$clear()")

• list("FitSBM$clone()")

Method new()

Constructor for FitSBM R6 class

Usage

FitSBM$new(
  Q = 1,
  X = NULL,
  M = NULL,
  directed = FALSE,
  distribution = "bernoulli"
)

Arguments

• Q: Number of blocks

• X: Adjacency matrix

• M: Mask matrix

• directed: boolean

• distribution: string (only "bernoulli")

Returns

A new FitSBM object

Method update_alpha()

Update the connection parameter for the M step

Usage

FitSBM$update_alpha(safeguard = 1e-06)

Arguments

• safeguard: Parameter live in a compact [safeguard, 1-safeguard]

Method update_pi()

Update the upper level mixture parameter for the M step

Usage

FitSBM$update_pi(safeguard = 1e-06)

Arguments

• safeguard: Parameter live in a compact [safeguard, 1-safeguard]

Method init_clustering()

init_clustering Initial clustering for VEM algorithm

Usage

FitSBM$init_clustering(safeguard = 1e-06, method = "hierarchical", Z = NULL)

Arguments

• safeguard: Parameter live in a compact [safeguard, 1-safeguard]

• method: Algorithm used to initiate the clustering, either "spectral", "hierarchical" or "merge_split" (if Z is provided)

• Z: Initial clustering if provided

Method m_step()

m_step Compute the M step of the VEM algorithm

Usage

FitSBM$m_step(safeguard = 1e-06)

Arguments

• safeguard: Parameter live in a compact [safeguard, 1-safeguard]

Method ve_step()

Compute the VE step of the VEM algorithm

Usage

FitSBM$ve_step(threshold = 1e-06, fixPointIter = 100, safeguard = 1e-06)

Arguments

• threshold: The convergence threshold

• fixPointIter: The maximum number of fixed point iterations

• safeguard: Parameter live in a compact [safeguard, 1-safeguard]

Method do_vem()

Launch a Variational EM algorithm

Usage

FitSBM$do_vem(
  init = "hierarchical",
  threshold = 1e-06,
  maxIter = 1000,
  fixPointIter = 100,
  safeguard = 1e-06,
  Z = NULL
)

Arguments

• init: The method for self$init_clustering

• threshold: The convergence threshold

• maxIter: The max number of VEM iterations

• fixPointIter: The max number of fixed point iterations for VE step

• safeguard: Parameter live in a compact [safeguard, 1-safeguard]

• Z: Initial clustering if provided

Method permute_empty_class()

permute_empty_class Put empty blocks numbers at the end

Usage

FitSBM$permute_empty_class()

Method clear()

Reset all parameters

Usage

FitSBM$clear()

Method clone()

The objects of this class are cloneable with this method.

Usage

FitSBM$clone(deep = FALSE)

Arguments

• deep: Whether to make a deep clone.