blockcluster (version 4.0.2)

coclusterStrategy: Strategy function

Description

This function is used to set all the parameters for Co-clustering. It returns an object of class strategy which can be given as input to coclusterBinary, coclusterCategorical, coclusterContingency, coclusterContinuous function.

This class contains all the input parameters to run coclustering.

Usage

coclusterStrategy(algo = "BEM", initmethod = character(), stopcriteria = "Parameter", nbiterationsxem = 50, nbiterationsXEM = 500, nbinititerations = 10, initepsilon = 0.01, nbiterations_int = 5, epsilon_int = 0.01, epsilonxem = 1e-04, epsilonXEM = 1e-10, nbtry = 2, nbxem = 5, hyperparam = c(1, 1))

Arguments

algo

The valid values for this parameter are "BEM" (Default), "BCEM" and "BSEM".

initmethod

Method to initialize model parameters. The valid values are "cemInitStep", "fuzzyCemInitStep" and "randomInit". For now only one kind of initialization exists for every model currently available in the package. Hence default value for initialization is set according to the model.

stopcriteria

It specifies the stopping criteria. It can be based on either relative change in parameters (preffered due to computation reasons) value or relative change in pseudo log-likelihood. Valid criterion values are "Parameter" and "Likelihood". Default criteria is "Parameter".

nbiterationsxem

Number of EM iterations used during xem step. Default value is 50.

nbiterationsXEM

Number of EM iterations used during XEM step. Default value is 500.

nbinititerations

Number of Global iterations used in initialization step. Default value is 10.

initepsilon

Tolerance value used while initialization. Default value is 1e-2.

nbiterations_int

Number of iterations for internal E step. Default value is 5.

epsilon_int

Tolerance value for relative change in Parameter/likelihood for internal E-step. Default value is 1e-2.

epsilonxem

Tolerance value used during xem step. Default value is 1e-4.

epsilonXEM

Tolerance value used during XEM step. Default value is 1e-10

nbtry

Number of tries (XEM steps). Default value is 2.

nbxem

Number of xem steps. Default value is 5.

hyperparam

Hyper-parameters ("a" and "b") in case of Bayesian settings (only valid for binary and categorical models). Default values are c(1,1) (no prior).

Value

Object of class strategy

Details

algo:

Algorithm to be use for co-clustering.

stopcriteria:

Stopping criteria used to stop the algorithm.

initmethod:

Method to initialize model parameters.

nbinititerations:

Number of global iterations while running initialization.

initepsilon:

Tolerance value used while initialization.

nbiterations_int:

Number of iterations for internal E-step.

epsilon_int:

Tolerance value for internal E-step.

nbtry:

Number of tries.

nbxem:

Number of xem iterations.

nbiterationsxem:

Number of EM iterations used during xem.

nbiterationsXEM:

Number of EM iterations used during XEM.

epsilonxem:

Tolerance value used during xem.

epsilonXEM:

Tolerance value used during XEM.

hyperparam:

Hyper-parameters ("a" and "b") in case of Bayesian settings.

Examples

#Default strategy values

strategy<-coclusterStrategy()
summary(strategy)