bikm1 package

Description

This package is designed to co-cluster a contingency (resp. binary) matrix, or double binary matrices in blocks respectively under the (normalized or not) Poisson (resp binary) Latent Block Model and the Multiple Latent Block Model. It enables to automatically select the number of row and column clusters and to compare partition estimations with reference partitions.

Features

Package for the segmentation of the rows and columns inducing a co-clustering and automatically select the number of row and column clusters.

Model 1

BIKM1_LBM_Poisson . This fitting procedure produces a BIKM1_LBM_Poisson object.

Model 2

BIKM1_LBM_Binary . This fitting procedure produces a BIKM1_LBM_Binary object.

Model 3

BIKM1_MLBM_Binary . This fitting procedure produces a BIKM1_MLBM_Binary object.

Technical remarks

Display of the result with plot,BIKM1_LBM_Poisson-method and

with show,BIKM1_LBM_Poisson-method, with summary,BIKM1_LBM_Poisson-method and with print,BIKM1_LBM_Poisson-method.

Display of the result with plot,BIKM1_LBM_Binary-method and

with show,BIKM1_LBM_Binary-method, with summary,BIKM1_LBM_Binary-method and with print,BIKM1_LBM_Binary-method.

Display of the result with plot,BIKM1_MLBM_Binary-method and

with show,BIKM1_MLBM_Binary-method, with summary,BIKM1_MLBM_Binary-method and with print,BIKM1_MLBM_Binary-method.

Author(s)

Valerie Robert valerie.robert.math@gmail.com

References

Keribin, Celeux and Robert, The Latent Block Model: a useful model for high dimensional data. https://hal.inria.fr/hal-01658589/document

Govaert and Nadif. Co-clustering, Wyley (2013).

Keribin, Brault and Celeux. Estimation and Selection for the Latent Block Model on Categorical Data, Statistics and Computing (2014).

Robert. Classification croisee pour l'analyse de bases de donnees de grandes dimensions de pharmacovigilance. Thesis, Paris Saclay (2017).

Robert, Vasseur and Brault. Comparing high dimensional partitions with the Co-clustering Adjusted Rand Index, Journal of Classification, 38(1), 158-186 (2021).

ARI function for agreement between two partitions

Description

Produce a measure of agreement between two partitions. A value of 1 means a perfect match.

Usage

ARI(v,vprime)

Arguments

Value

a list including the arguments:

ari: value of the index.

nv: contingency table which the index is based on.

References

Hubert and Arabie. Comparing partitions. Journal of classification (1985).

Examples

require(bikm1)
J=200
K=120
h=3
l=2
theta=list()
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$gamma_hl=matrix(c(1, 6,4, 1, 7, 1),ncol=2)
data=PoissonBlocRnd(J,K,theta)
res=BIKM1_LBM_Poisson(data$x,4,4,4,init_choice='random')
mv=ARI(res@model_max$v, data$xrow)
mv$ari
mv$nv
mw=ARI(res@model_max$w, data$xcol)

BIKM1_LBM_Binary fitting procedure

Description

Produce a blockwise estimation of a contingency matrix of observations.

Usage

BIKM1_LBM_Binary(x,Gmax,Hmax,a=4,b=1,
Gstart=2,Hstart=2,init_choice='smallVBayes',userparam=NULL,
ntry=50,criterion_choice='ICL', mc.cores=1,verbose=TRUE)

Arguments

Value

a BIKM1_LBM_Binary object including

model_max: the selected model by the procedure with free energy W, theta, conditional probabilities (s_ig, r_jh), iter, empty_cluster, and the selected partitions z and w.

criterion_choice: the chosen criterion

init_choice: the chosen init choice

criterion tab: the matrix containing the criterion values for each selected number of row and column

W_tab: the matrix containing the free energy values for each selected number of row and column

criterion_max: the maximum of the criterion values

gopt: the selected number of rows

hopt: the selected number of columns

References

Govaert and Nadif. Co-clustering, Wyley (2013).

Keribin, Brault and Celeux. Estimation and Selection for the Latent Block Model on Categorical Data, Statistics and Computing (2014).

Robert. Classification crois\'ee pour l'analyse de bases de donn\'ees de grandes dimensions de pharmacovigilance. Paris Saclay (2017).

Examples

require(bikm1)
set.seed(42)
n=200
J=120
g=3
h=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$alpha_gh=matrix(runif(6),ncol=h)
data=BinBlocRnd_LBM(n,J,theta)
res=BIKM1_LBM_Binary(data$x,3,2,Gstart=3,Hstart=2,
init_choice='user',userparam=list(z=data$xrow,v=data$xcol))

Class "BIKM1_LBM_Binary"

Description

Class of object returned by the BIKM1_LBM_Binary function.

Slots

model_max:

The selected model by the procedure with free energy W, theta, conditional probabilities (s_ig, r_jh), iter, empty_cluster, and the selected partitions z and v.

criterion_choice:

A character string corresponding to the chosen criterion used for model selection, which can be "ICL" or "BIC".

init_choice:

A character string corresponding to the chosen initialization strategy used for the procedure, which can be "random" or "Gibbs" or "smallVBayes".

criterion_tab:

The matrix corresponding to the values of the chosen criterion for pairs of numbers of clusters visited by the BIKM1_LBM_Binary function. The matrix rows design the numbers of row clusters. If a pair is not visited, by default, the value is -Inf.

W_tab:

The matrix corresponding to the values of the free energy (minimizer of the loglikelihood in the algorithm) for pairs of numbers of clusters visited by the procedure. The matrix rows design the numbers of row clusters. If a pair is not visited, by default, the value is -Inf.

criterion_max:

Numeric indicating the maximum of the criterion values, calculated on the pairs of numbers of clusters visited by the BIKM1_LBM_Binary function.

gopt:

An integer value indicating the number of row clusters selected by the BIKM1_LBM_Binary function.

hopt:

An integer value indicating the number of column clusters selected by the BIKM1_LBM_Binary function.

Examples

require(bikm1)
n=200
J=120
g=3
h=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$alpha_gh=matrix(runif(6),ncol=h)
data=BinBlocRnd_LBM(n,J,theta)
res=BIKM1_LBM_Binary(data$x,3,3,a=4,init_choice='smallVBayes')

BIKM1_LBM_Poisson fitting procedure

Description

Produce a blockwise estimation of a contingency matrix of observations.

Usage

BIKM1_LBM_Poisson(x,Hmax,Lmax,a=4,alpha=1,beta=0.01,
Hstart=2,Lstart=2,normalization=FALSE,init_choice='smallVBayes',
userparam=NULL,ntry=50,criterion_choice='ICL', mc.cores=1,verbose=TRUE)

Arguments

Value

a BIKM1_LBM_Poisson object including

model_max: the selected model by the procedure with free energy W, theta, conditional probabilities (r_jh, t_kl), iter, empty_cluster, and the selected partitions v and w.

criterion_choice: the chosen criterion

init_choice: the chosen init choice

criterion tab: matrix containing the criterion values for each selected number of row and column

W_tab: matrix containing the free energy values for each selected number of row and column

criterion_max: maximum of the criterion values

hopt: the selected number of rows

lopt: the selected number of columns

References

Keribin, Celeux and Robert, The Latent Block Model: a useful model for high dimensional data. https://hal.inria.fr/hal-01658589/document

Govaert and Nadif. Co-clustering, Wyley (2013).

Keribin, Brault and Celeux. Estimation and Selection for the Latent Block Model on Categorical Data, Statistics and Computing (2014).

Robert. Classification crois\'ee pour l'analyse de bases de donn\'ees de grandes dimensions de pharmacovigilance. Paris Saclay (2017).

Examples

require(bikm1)
J=200
K=120
h=3
l=2
theta=list()
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$gamma_hl=matrix(c(1, 6,4, 1, 7, 1),ncol=2)
data=PoissonBlocRnd(J,K,theta)
res=BIKM1_LBM_Poisson(data$x,3,2,Hstart=3,Lstart=2,
init_choice='user',userparam=list(v=data$xrow,w=data$xcol))

Class "BIKM1_LBM_Poisson"

Description

Class of object returned by the BIKM1_LBM_Poisson function.

Slots

model_max:

The selected model by the procedure with free energy W, theta, conditional probabilities (r_jh, t_kl), iter, empty_cluster, and the selected partitions v and w.

criterion_choice:

A character string corresponding to the chosen criterion used for model selection, which can be "ICL" or "BIC".

init_choice:

A character string corresponding to the chosen initialization strategy used for the procedure, which can be "random" or "Gibbs" or "smallVBayes".

criterion_tab:

The matrix corresponding to the values of the chosen criterion for pairs of numbers of clusters visited by the BIKM1_LBM_Poisson function. The matrix rows design the numbers of row clusters. If a pair is not visited, by default, the value is -Inf.

W_tab:

The matrix corresponding to the values of the free energy (minimizer of the loglikelihood in the algorithm) for pairs of numbers of clusters visited by the procedure. The matrix rows design the numbers of row clusters. If a pair is not visited, by default, the value is -Inf.

criterion_max:

Numeric indicating the maximum of the criterion values, calculated on the pairs of numbers of clusters visited by the BIKM1_LBM_Poisson function.

lopt:

An Integer value indicating the number of row clusters selected by the BIKM1_LBM_Poisson function.

hopt:

An integer value indicating the number of column clusters selected by the BIKM1_LBM_Poisson function.

Examples

require(bikm1)
set.seed(42)
J=200
K=120
h=3
l=2
theta=list()
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$gamma_hl=matrix(floor(runif(h*l)*20+1),ncol=l)
data=PoissonBlocRnd(J,K,theta)
res=BIKM1_LBM_Poisson(data$x,3,3,4,init_choice='smallVBayes')

BIKM1_MLBM_Binary fitting procedure

Description

Produce a blockwise estimation of double matrices of observations.

Usage

BIKM1_MLBM_Binary(x,y,Gmax,Hmax,Lmax,a=4,b=1,
Gstart=2,Hstart=2,Lstart=2,init_choice='smallVBayes',userparam=NULL,
ntry=50,criterion_choice='ICL', mc.cores=1,verbose=TRUE)

Arguments

Value

a BIKM1_MLBM_Binary object including

model_max: the selected model by the procedure including free energy W, theta, conditional probabilities (s_ig, r_jh,t_kl), iter, empty_cluster, and the selected partitions z,v and w.

criterion_choice: the chosen criterion

init_choice: the chosen init_choice

criterion_tab: matrix containing the criterion values for each selected number of row and column

W_tab: matrix containing the free energy values for each selected number of row and column

criterion_max: maximum of the criterion values

gopt: the selected number of rows

hopt: the selected number of columns (1rst matrix)

lopt: the selected number of columns (2nd matrix)

References

Govaert and Nadif. Co-clustering, Wyley (2013).

Keribin, Brault and Celeux. Estimation and Selection for the Latent Block Model on Categorical Data, Statistics and Computing (2014).

Robert. Classification crois\'ee pour l'analyse de bases de donn\'ees de grandes dimensions de pharmacovigilance. Paris Saclay (2017).

Examples

require(bikm1)
set.seed(42)
n=200
J=120
K=120
g=3
h=2
l=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$alpha_gh=matrix(runif(6),ncol=h)
theta$beta_gl=matrix(runif(6),ncol=l)
data=BinBlocRnd_MLBM(n,J,K,theta)
res=BIKM1_MLBM_Binary(data$x,data$y,3,2,2,Gstart=3,Hstart=2,Lstart=2,init_choice='user',
userparam=list(z=data$xrow,v=data$xcolx,w=data$xcoly))

Class "BIKM1_MLBM_Binary"

Description

Class of object returned by the BIKM1_MLBM_Binary function.

Slots

model_max:

The selected model by the procedure with free energy W, theta, conditional probabilities (s_ig, r_jh, t_kl), iter, empty_cluster, and the selected partitions z, v and w.

criterion_choice:

A character string corresponding to the chosen criterion used for model selection, which can be "ICL" or "BIC".

init_choice:

A character string corresponding to the chosen initialization strategy used for the procedure, which can be "random" or "Gibbs" or "smallVBayes".

criterion_tab:

The matrix corresponding to the values of the chosen criterion for pairs of numbers of clusters visited by the BIKM1_MLBM_Binary function. The matrix rows design the numbers of row clusters. If a pair is not visited, by default, the value is -Inf.

W_tab:

The matrix corresponding to the values of the free energy (minimizer of the loglikelihood in the algorithm) for pairs of numbers of clusters visited by the procedure. The matrix rows design the numbers of row clusters. If a pair is not visited, by default, the value is -Inf.

criterion_max:

Numeric indicating the maximum of the criterion values, calculated on the pairs of numbers of clusters visited by the BIKM1_MLBM_Binary function.

gopt:

An integer value indicating the number of row clusters selected by the BIKM1_MLBM_Binary function.

hopt:

An integer value indicating the number of column clusters for the first matrix selected by the BIKM1_MLBM_Binary function.

lopt:

An integer value indicating the number of row clusters for the second matrix selected by the BIKM1_MLBM_Binary function.

Examples

require(bikm1)
n=200
J=120
K=120
g=3
h=2
l=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$alpha_gh=matrix(runif(6),ncol=h)
theta$beta_gl=matrix(runif(6),ncol=l)
data=BinBlocRnd_MLBM(n,J,K,theta)
res=BIKM1_MLBM_Binary(data$x,data$y,3,3,3,4,init_choice='smallVBayes')

BinBlocICL_LBM function for computation of the ICL criterion in the Binary LBM

Description

Produce a value of the ICL criterion in the Binary LBM.

Usage

BinBlocICL_LBM(a,b,x,z1,v1)

Arguments

Value

a value of the ICL criterion.

Examples

require(bikm1)
set.seed(42)
n=200
J=120
g=3
h=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$alpha_gh=matrix(runif(6),ncol=h)
data=BinBlocRnd_LBM(n,J,theta)
BinBlocICL_LBM(a=4,b=1,data$x, data$xrow,data$xcol)

BinBlocICL_MLBM function for computation of the ICL criterion in the MLBM

Description

Produce a plot object representing the resumed co-clustered data-sets.

Usage

BinBlocICL_MLBM(a,b,x,y,z1,v1,w1)

Arguments

Value

a value of the ICL criterion.

Examples

require(bikm1)
set.seed(42)
n=200
J=120
K=120
g=2
h=2
l=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$alpha_gh=matrix(runif(4),ncol=h)
theta$beta_gl=matrix(runif(4),ncol=l)
data=BinBlocRnd_MLBM(n,J,K,theta)
res=BIKM1_MLBM_Binary(data$x,data$y,2,2,2,4,init_choice='smallVBayes')
BinBlocICL_MLBM(a=4,b=1,data$x,data$y, data$xrow,data$xcolx,data$xcoly)

BinBlocRnd_LBM function for binary data matrix simulation

Description

Produce a data matrix generated under the Binary Latent Block Model.

Usage

BinBlocRnd_LBM(n,J,theta)

Arguments

Value

a list including the arguments:

x: simulated data matrix.

xrow: numeric vector specifying row partition.

xcol: numeric vector specifying column partition.

Examples

require(bikm1)
set.seed(42)
n=200
J=120
g=3
h=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$alpha_gh=matrix(runif(6),ncol=h)
data=BinBlocRnd_LBM(n,J,theta)

BinBlocRnd_MLBM function for binary double data matrix simulation

Description

Produce two simulated data matrices generated under the Binary Multiple Latent Block Model.

Usage

BinBlocRnd_MLBM(n,J,K,theta)

Arguments

Value

a list including the arguments:

x: simulated first data matrix. y: simulated second data matrix.

xrow: numeric vector specifying row partition.

xcolx: numeric vector specifying first matrix column partition.

xcoly: numeric vector specifying second matrix column partition.

Examples

require(bikm1)
set.seed(42)
n=200
J=120
K=120
g=3
h=2
l=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$alpha_gh=matrix(runif(6),ncol=h)
theta$beta_gl=matrix(runif(6),ncol=l)
data=BinBlocRnd_MLBM(n,J,K,theta)

BinBlocVisuResum_LBM function for visualization of binary matrix data-sets

Description

Produce a plot object representing the resumed co-clustered data-sets.

Usage

BinBlocVisuResum_LBM(x,z,v)

Arguments

Value

a plot object.

Examples

require(bikm1)
set.seed(42)
n=200
J=120
g=3
h=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$alpha_gh=matrix(runif(6),ncol=h)
data=BinBlocRnd_LBM(n,J,theta)
BinBlocVisuResum_LBM(data$x,data$xrow,data$xcol)

BinBlocVisuResum_MLBM function for visualization of double matrix datasets

Description

Produce a plot object representing the resumed co-clustered data-sets.

Usage

BinBlocVisuResum_MLBM(x,y,z,v,w)

Arguments

Value

a plot object.

Examples

require(bikm1)
set.seed(42)
n=200
J=120
K=120
g=3
h=2
l=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$alpha_gh=matrix(runif(6),ncol=h)
theta$beta_gl=matrix(runif(6),ncol=l)
data=BinBlocRnd_MLBM(n,J,K,theta)
BinBlocVisuResum_MLBM(data$x,data$y, data$xrow,data$xcolx,data$xcoly)

BinBlocVisu_LBM function for visualization of binary matrix datasets

Description

Produce a plot object representing the co-clustered data-sets.

Usage

BinBlocVisu_LBM(x,z,v)

Arguments

Value

a plot object

Examples

require(bikm1)
set.seed(42)
n=200
J=120
g=3
h=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$alpha_gh=matrix(runif(6),ncol=h)
data=BinBlocRnd_LBM(n,J,theta)
BinBlocVisu_LBM(data$x,data$xrow,data$xcol)

BinBlocVisu_MLBM function for visualization of double matrix datasets

Description

Produce a plot object representing the co-clustered data-sets.

Usage

BinBlocVisu_MLBM(x,y,z,v,w)

Arguments

Value

a plot object

Examples

require(bikm1)
set.seed(42)
n=200
J=120
K=120
g=3
h=2
l=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$alpha_gh=matrix(runif(6),ncol=h)
theta$beta_gl=matrix(runif(6),ncol=l)
data=BinBlocRnd_MLBM(n,J,K,theta)
BinBlocVisu_MLBM(data$x,data$y, data$xrow,data$xcolx,data$xcoly)

CARI function for agreement between co-clustering partitions

Description

Produce a measure of agreement between two pairs of partitions for co-clustering. A value of 1 means a perfect match.

Usage

CARI(v,w,vprime,wprime)

Arguments

Value

a list including the arguments:

cari: value of the index (between 0 and 1). A value of 1 corresponds to a perfect match.

nvw: contingency table which the index is based on.

References

Robert, Vasseur and Brault. Comparing high dimensional partitions with the Co-clustering Adjusted Rand Index, Journal of classification 38 (1), 158-186 (2021).

Examples

require(bikm1)
J=200
K=120
h=3
l=2
theta=list()
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$gamma_hl=matrix(c(1, 6,4, 1, 7, 1),ncol=2)
data=PoissonBlocRnd(J,K,theta)
res=BIKM1_LBM_Poisson(data$x,4,4,4,init_choice='smallVBayes')
me=CARI(res@model_max$v,res@model_max$w, data$xrow,data$xcol)
me$cari
me$nvw

CE_LBM function for agreement between co-clustering partitions

Description

Produce a measure of agreement between two pairs of partitions for co-clustering using CE_simple on columns and rows of a matrix. A value of 1 means a perfect match.

Usage

CE_LBM(v,w,vprime,wprime)

Arguments

Value

ce_vw: the value of the index (between 0 and 1). A value of 0 corresponds to a perfect match.

Examples

require(bikm1)
set.seed(42)
v=floor(runif(4)*2)
vprime=floor(runif(4)*2)
w=floor(runif(4)*3)
wprime=floor(runif(4)*3)
error=CE_LBM(v,w,vprime,wprime)

CE_MLBM function for agreement between co-clustering partitions in the MBLM

Description

Produce a measure of agreement between two triplets of partitions for co-clustering. A value of 1 means a perfect match.

Usage

CE_MLBM(z,v,w,zprime,vprime,wprime)

Arguments

Value

the value of the index (between 0 and 1). A value of 0 corresponds to a perfect match.

Examples

require(bikm1)
set.seed(42)
n=200
J=120
K=120
g=2
h=2
l=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$alpha_gh=matrix(runif(4),ncol=h)
theta$beta_gl=matrix(runif(4),ncol=l)
data=BinBlocRnd_MLBM(n,J,K,theta)
res=BIKM1_MLBM_Binary(data$x,data$y,2,2,2,4,init_choice='smallVBayes')
error=CE_MLBM(res@model_max$z,res@model_max$v,res@model_max$w,data$xrow,data$xcolx,data$xcoly)

CE_simple function for agreement between clustering partitions

Description

Produce a measure of agreement between two partitions for clustering. A value of 1 means a perfect match.

Usage

CE_simple(v,vprime)

Arguments

Value

the value of the index.

Examples

require(bikm1)
set.seed(42)
v=floor(runif(4)*3)
vprime=floor(runif(4)*3)
error=CE_simple(v,vprime)
error

CoNMI function for agreement between co-clustering partitions

Description

Produce a measure of agreement between two pairs of partitions for co-clustering. A value of 1 means a perfect match.

Usage

CoNMI(v,w,vprime,wprime)

Arguments

Value

the value of the index.

References

Robert, Vasseur and Brault. Comparing high dimensional partitions with the Co-clustering Adjusted Rand Index, Journal of Classification (2021).

Examples

require(bikm1)
J=200
K=120
h=3
l=2
theta=list()
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$gamma_hl=matrix(c(1, 6,4, 1, 7, 1),ncol=2)
data=PoissonBlocRnd(J,K,theta)
res=BIKM1_LBM_Poisson(data$x,4,4,4,init_choice='smallVBayes')
me=CoNMI(res@model_max$v,res@model_max$w, data$xrow,data$xcol)
me

ENMI function for agreement between co-clustering partitions

Description

Produce a measure of agreement between two pairs of partitions for co-clustering. A value of 1 means a perfect match.

Usage

ENMI(v,w,vprime,wprime)

Arguments

Value

the value of the index.

References

Robert, Vasseur and Brault. Comparing high dimensional partitions with the Co-clustering Adjusted Rand Index, Journal of Classification (2021).

Examples

require(bikm1)
J=200
K=120
h=3
l=2
theta=list()
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$gamma_hl=matrix(c(1, 6,4, 1, 7, 1),ncol=2)
data=PoissonBlocRnd(J,K,theta)
res=BIKM1_LBM_Poisson(data$x,4,4,4,init_choice='smallVBayes')
me=ENMI(res@model_max$v,res@model_max$w, data$xrow,data$xcol)
me

MI_simple function for agreement between two partitions

Description

Produce a measure of agreement between two partitions.(between 0 and 1). A value of 1 corresponds to a perfect match.

Usage

MI_simple(v,vprime)

Arguments

Value

the value of the index.

References

Robert, Vasseur and Brault. Comparing high-dimensional partitions with the Co-clustering Adjusted Rand Index. Journal of Classification (2021).

Examples

require(bikm1)
J=200
K=120
h=3
l=2
theta=list()
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$gamma_hl=matrix(c(1, 6,4, 1, 7, 1),ncol=2)
data=PoissonBlocRnd(J,K,theta)
res=BIKM1_LBM_Poisson(data$x,4,4,4,init_choice='random')
mi=MI_simple(res@model_max$v, data$xrow)
mi
mw=MI_simple(res@model_max$w, data$xcol)

NCE_LBM function for agreement between co-clustering partitions using NCE_simple

Description

Produce a measure of agreement between two pairs of partitions for co-clustering. A value of 1 means a perfect match.

Usage

NCE_LBM(v,w,vprime,wprime)

Arguments

Value

the value of the index.

Examples

require(bikm1)
set.seed(42)
v=floor(runif(4)*2)
vprime=floor(runif(4)*2)
w=floor(runif(4)*3)
wprime=floor(runif(4)*3)
error=NCE_LBM(v,w,vprime,wprime)

NCE_simple function for agreement between clustering partitions

Description

Produce a measure of agreement between two partitions for clustering. A value of 1 means a perfect match. It's the normalized version of CE_simple.

Usage

NCE_simple(v,vprime)

Arguments

Value

the value of the index. A value of 0 means a perfect match.

Examples

require(bikm1)
set.seed(42)
v=floor(runif(4)*3)
vprime=floor(runif(4)*3)
error=NCE_simple(v,vprime)
error

PoissonBlocBIC function for the computation of the BIC criterion in the Poisson LBM

Description

Produce a value of the BIC criterion for co-clustering partitions

Usage

PoissonBlocBIC(a,alpha,beta,v1,w1,x,res,normalization)

Arguments

Value

a value of the BIC criterion

Examples

require(bikm1)
J=200
K=120
h=3
l=2
theta=list()
theta$rho_h=1/h*matrix(1,h,1)
theta$tau_l=1/l*matrix(1,l,1)
theta$gamma_hl=matrix(c(1, 6,4, 1, 7, 1),ncol=2)
data=PoissonBlocRnd(J,K,theta)
res=BIKM1_LBM_Poisson(data$x,3,3,4,init_choice='smallVBayes')
bic=PoissonBlocBIC(v1=res@model_max$v,w1=res@model_max$w,x=data$x,res=res,normalization=TRUE)

PoissonBlocICL function for the computation of the ICL criterion in the Poisson LBM

Description

Produce a value of the ICL criterion for co-clustering partitions

Usage

PoissonBlocICL(a,alpha,beta,x,v1,w1,normalization)

Arguments

Value

a value of the ICL criterion

Examples

require(bikm1)
J=200
K=120
h=3
l=2
theta=list()
theta$rho_h=(1/h)*matrix(1,h,1)
theta$tau_l=(1/l)*matrix(1,l,1)
theta$gamma_hl=matrix(c(1, 6,4, 1, 7, 1),ncol=2)
data=PoissonBlocRnd(J,K,theta)
res=BIKM1_LBM_Poisson(data$x,4,4,4,init_choice='smallVBayes')
icl=PoissonBlocICL(4,1,0.01,data$x,res@model_max$v,res@model_max$w, normalization=FALSE)

PoissonBlocRnd function for contingency data simulation

Description

Produce a simulated data matrix generated under the Poisson Latent Block Model.

Usage

PoissonBlocRnd(J,K,theta)

Arguments

Value

a list including the arguments:

x: simulated contingency data matrix.

xrow: numeric vector specifying row partition.

xcol: numeric vector specifying column partition.

Examples

require(bikm1)
J=200
K=120
h=3
l=2
theta=list()
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$gamma_hl=matrix(c(1, 6,4, 1, 7, 1),ncol=2)
data=PoissonBlocRnd(J,K,theta)

PoissonBlocVisu function for visualization of contingency datasets

Description

Produce a plot object representing the co-clustered data-sets.

Usage

PoissonBlocVisu(x,v,w)

Arguments

Value

a plot object

Examples

require(bikm1)
J=200
K=120
h=3
l=2
theta=list()
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$gamma_hl=matrix(c(1, 6,4, 1, 7, 1),ncol=2)
data=PoissonBlocRnd(J,K,theta)
PoissonBlocVisu(data$x,data$xrow,data$xcol)

PoissonBlocVisuResum function for visualization of contingency datasets

Description

Produce a plot object representing the resumed co-clustered data-sets.

Usage

PoissonBlocVisuResum(x,v,w)

Arguments

Value

a plot object.

Examples

require(bikm1)
J=200
K=120
h=3
l=2
theta=list()
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$gamma_hl=matrix(c(1, 6,4, 1, 7, 1),ncol=2)
data=PoissonBlocRnd(J,K,theta)
PoissonBlocVisuResum(data$x,data$xrow,data$xcol)

Plot method for a BIKM1_LBM_Binary object

Description

Produce respectively one plot of two-dimensional segmentation of a BIKM1_LBM_Binary fit, a plot of evolution of the chosen criterion as a function of the number of row and column clusters, and a boxplot of conditional posteriors for each row and column cluster.

Usage

## S4 method for signature 'BIKM1_LBM_Binary'
plot(x, y, ...)

Arguments

Value

One plot (initial and estimated partitions) and three ggplot2 objects (conditional posterior in each cluster for each matrix and the graph of chosen criterion values.

Examples

require(bikm1)
g=5
h=3
theta=list()
theta$pi_g=t(1/g*rep(1,g))
theta$rho_h=t(1/h*rep(1,h))
eps=0.1
theta$alpha_gh=matrix(c(1-eps,eps,eps,eps,1-eps,eps,eps,1-eps,1-eps,
1-eps,1-eps,eps,eps,eps,eps),ncol=h,byrow=TRUE)
n=250
J=150
data=BinBlocRnd_LBM(n,J,theta)
BinBlocVisu_LBM(data$x, data$xrow,data$xcol)
res=BIKM1_LBM_Binary(data$x,8,5,4,init_choice='smallVBayes')
BinBlocVisu_LBM(data$x,res@model_max$z,res@model_max$v)
e=CARI(data$xrow,data$xcol,res@model_max$z,res@model_max$v)
plot(res,data)

Plot method for a BIKM1_LBM_Poisson object

Description

Produce respectively one plot of two-dimensional segmentation of a BIKM1_LBM_Poisson fit, an evolution of the criterion as a function of the numbers of rows and columns, and a boxplot of conditional posteriors for each row and column cluster.

Usage

## S4 method for signature 'BIKM1_LBM_Poisson'
plot(x, y, ...)

Arguments

Value

Two plots (initial matrix and block estimation) and two ggplot2 objects (conditional posterior in each cluster and the graph of chosen criterion values).

Examples

require(bikm1)
J=200
K=120
h=3
l=2
theta=list()
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$gamma_hl=matrix(c(1, 6,4, 1, 7, 1),ncol=2)
data=PoissonBlocRnd(J,K,theta)
res=BIKM1_LBM_Poisson(data$x,3,3,4,init_choice='random')
plot(res,data)

Plot method for a BIKM1_MLBM_Binary object

Description

Produce respectively a plot of two-dimensional segmentation of a BIKM1_MLBM_Binary fit, and a boxplot of conditional posteriors for each row and column cluster.

Usage

## S4 method for signature 'BIKM1_MLBM_Binary'
plot(x, y, ...)

Arguments

Value

Two plot and on ggplot2 object.

Examples

require(bikm1)
n=200
J=120
K=120
g=3
h=2
l=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$alpha_gh=matrix(runif(6),ncol=h)
theta$beta_gl=matrix(runif(6),ncol=l)
data=BinBlocRnd_MLBM(n,J,K,theta)
res=BIKM1_MLBM_Binary(data$x,data$y,3,3,3,4)
plot(res,data)

Print method for a BIKM1_LBM_Binary object

Description

Print method for a BIKM1_LBM_Binary object

Usage

## S4 method for signature 'BIKM1_LBM_Binary'
print(x, ...)

Arguments

Examples

require(bikm1)
n=200
J=120
g=3
h=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$alpha_gh=matrix(runif(6),ncol=h)
data=BinBlocRnd_LBM(n,J,theta)
res=BIKM1_LBM_Binary(data$x,3,2,4,init_choice='random')
print(res)

Print method for a BIKM1_LBM_Poisson object

Description

Print method for a BIKM1_LBM_Poisson object

Usage

## S4 method for signature 'BIKM1_LBM_Poisson'
print(x, ...)

Arguments

Examples

require(bikm1)
J=200
K=120
h=3
l=2
theta=list()
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$gamma_hl=matrix(c(1, 6,4, 1, 7, 1),ncol=2)
data=PoissonBlocRnd(J,K,theta)
res=BIKM1_LBM_Poisson(data$x,3,2,4,init_choice='random')
print(res)

Print method for a BIKM1_MLBM_Binary object

Description

Print method for a BIKM1_MLBM_Binary object

Usage

## S4 method for signature 'BIKM1_MLBM_Binary'
print(x, ...)

Arguments

Examples

require(bikm1)
n=200
J=120
K=120
g=3
h=2
l=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$alpha_gh=matrix(runif(6),ncol=h)
theta$beta_gl=matrix(runif(6),ncol=l)
data=BinBlocRnd_MLBM(n,J,K,theta)
res=BIKM1_MLBM_Binary(data$x,data$y,3,3,3,4)
print(res)

Show method for a BIKM1_LBM_Binary object

Description

show method for a BIKM1_LBM_Binary object

Usage

## S4 method for signature 'BIKM1_LBM_Binary'
show(object)

Arguments

Examples

require(bikm1)
n=200
J=120
g=3
h=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$alpha_gh=matrix(runif(6),ncol=h)
data=BinBlocRnd_LBM(n,J,theta)
res=BIKM1_LBM_Binary(data$x,4,4,4,init_choice='random')
show(res)

Show method for a BIKM1_LBM_Poisson object

Description

show method for a BIKM1_LBM_Poisson object

Usage

## S4 method for signature 'BIKM1_LBM_Poisson'
show(object)

Arguments

Examples

require(bikm1)
J=200
K=120
h=3
l=2
theta=list()
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$gamma_hl=matrix(c(1, 6,4, 1, 7, 1),ncol=2)
data=PoissonBlocRnd(J,K,theta)
res=BIKM1_LBM_Poisson(data$x,4,4,4,init_choice='random')
show(res)

Show method for a BIKM1_MLBM_Binary object

Description

show method for a BIKM1_MLBM_Binary object

Usage

## S4 method for signature 'BIKM1_MLBM_Binary'
show(object)

Arguments

Examples

require(bikm1)
n=200
J=120
K=120
g=3
h=2
l=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$alpha_gh=matrix(runif(6),ncol=h)
theta$beta_gl=matrix(runif(6),ncol=l)
data=BinBlocRnd_MLBM(n,J,K,theta)
res=BIKM1_MLBM_Binary(data$x,data$y,3,3,3,4)
show(res)

Summary method for a BIKM1_LBM_Binary object

Description

Produce a summary of informations of a BIKM1_LBM_Binary object

Usage

## S4 method for signature 'BIKM1_LBM_Binary'
summary(object, ...)

Arguments

Examples

require(bikm1)
n=200
J=120
g=3
h=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$alpha_gh=matrix(runif(6),ncol=h)
data=BinBlocRnd_LBM(n,J,theta)
res=BIKM1_LBM_Binary(data$x,3,2,4,init_choice='random')
summary(res)

Summary method for a BIKM1_LBM_Poisson object

Description

Produce a summary of informations of a BIKM1_LBM_Poisson object

Usage

## S4 method for signature 'BIKM1_LBM_Poisson'
summary(object, ...)

Arguments

Examples

require(bikm1)
J=200
K=120
h=3
l=2
theta=list()
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$gamma_hl=matrix(c(1, 6,4, 1, 7, 1),ncol=2)
data=PoissonBlocRnd(J,K,theta)
res=BIKM1_LBM_Poisson(data$x,4,4,4,init_choice='random')
summary(res)

Summary method for a BIKM1_MLBM_Binary object

Description

Produce a summary of informations of a BIKM1_MLBM_Binary object

Usage

## S4 method for signature 'BIKM1_MLBM_Binary'
summary(object, ...)

Arguments

Examples

require(bikm1)
n=200
J=120
K=120
g=3
h=2
l=2
theta=list()
theta$pi_g=1/g *matrix(1,g,1)
theta$rho_h=1/h *matrix(1,h,1)
theta$tau_l=1/l *matrix(1,l,1)
theta$alpha_gh=matrix(runif(6),ncol=h)
theta$beta_gl=matrix(runif(6),ncol=l)
data=BinBlocRnd_MLBM(n,J,K,theta)
res=BIKM1_MLBM_Binary(data$x,data$y,3,3,3,4)
summary(res)