1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
|
# Build a 'graph-like' object having 'nodes' nodes belonging to 'classes' classes.
# Class distribution is given by 'proba', and connectivity probabilities are given
# by 'intraproba' and 'interproba'.
generateGraph<-function(nodes,classes,proba=rep(1/classes,classes),
intraproba=0.1,crossproba=0.02)
{
mat_pi=CreateConnectivityMat(classes,intraproba,crossproba)
igraph=Fast2SimuleERMG(nodes,proba,mat_pi[1],mat_pi[2])
adjacency=get.adjacency(igraph$graph)
cmgraph=list(nodes=nodes,classes=classes,adjacency=adjacency,nodeclasses=igraph$node.classes,proba=proba,
intraproba=intraproba,crossproba=crossproba)
attr(cmgraph,'class')<-c('cmgraph')
cmgraph
}
# Return explicit member names for the different attributes of graph objects.
labels.cmgraph<-function(object,...)
{
c("Nodes","Classes","Adjacency Matrix","Node Classification","Class Probability Distribution","Intra Class Edge Probability","Cross Class Edge Probability")
}
# Override the summmary function for graph objects.
summary.cmgraph<-function(object,...)
{
cat(c("Nodes : ",object$nodes,"\n",
"Edges : ",length(which(object$adjacency!=0)),"\n",
"Classes : ",object$classes,"\n",
"Class Probability Distribution: ",object$proba,"\n"))
}
# Override the plot function for graph objects.
plot.cmgraph<-function(x,...)
{
RepresentationXGroup(x$adjacency,x$nodeclasses)
}
# Generate covariable data for the graph 'g'. Covariables are associated to vertex data, and
# their values are drawn according to 2 distributions: one for vertices joining nodes of
# the same class, and another for vertices joining nodes of different classes.
# The two distributions have different means but a single standard deviation.
generateCovariablesCondZ<-function(g,sameclustermean=0,otherclustermean=2,sigma=1)
{
mu=CreateMu(g$classes,sameclustermean,otherclustermean)
res=SimDataYcondZ(g$nodeclasses,mu,sigma)
cmcovars=list(graph=g,sameclustermean=sameclustermean,otherclustermean=otherclustermean,sigma=sigma,mu=mu,y=res)
attr(cmcovars,'class')<-c('cmcovarz','cmcovar')
cmcovars
}
# Generate covariable data for the graph 'g'. Covariables are associated to vertex data, and
# their values are drawn according to 2 distributions: one for vertices joining nodes of
# the same class, and another for vertices joining nodes of different classes.
# The two distributions have different means but a single standard deviation.
# This function generates two sets of covariables.
generateCovariablesCondXZ<-function(g,sameclustermean=c(0,3),otherclustermean=c(2,5),sigma=1)
{
mux0=CreateMu(g$classes,sameclustermean[1],otherclustermean[1])
mux1=CreateMu(g$classes,sameclustermean[2],otherclustermean[2])
res=SimDataYcondXZ(g$nodeclasses,g$adjacency,mux0,mux1,sigma)
cmcovars=list(graph=g,sameclustermean=sameclustermean,otherclustermean=otherclustermean,sigma=sigma,mu=c(mux0,mux1),y=res)
attr(cmcovars,'class')<-c('cmcovarxz','cmcovar')
cmcovars
}
# Override the print function for a cleaner covariable output.
print.cmcovar<-function(x,...)
{
cat("Classes : ",x$graph$classes,"\n",
"Intra cluster mean: ",x$sameclustermean,"\n",
"Cross cluster mean: ",x$otherclustermean,"\n",
"Variance : ",x$sigma,"\n",
"Covariables :\n",x$y,"\n")
}
# Perform parameter estimation on 'graph' given the covariables 'covars'.
estimateCondZ<-function(graph,covars,maxiterations,initialclasses,selfloops)
{
res=EMalgorithm(initialclasses,covars$y,graph$adjacency,maxiterations,FALSE,selfloops)
cmestimation=list(mean=res$MuEstimated,variance=res$VarianceEstimated,pi=res$PIEstimated,alpha=res$AlphaEstimated,tau=res$TauEstimated,jexpected=res$EJ,graph=graph)
attr(cmestimation,'class')<-c('cmestimationz')
cmestimation
}
# Private generic estimation function used to allow various call conventions for estimation functions.
privateestimate<-function(covars,graph,maxiterations,initialclasses,selfloops,...) UseMethod("privateestimate")
# Private estimation function used to allow various call conventions for estimation functions.
# Override of generic function for single covariables.
privateestimate.cmcovarz<-function(covars,graph,maxiterations,initialclasses,selfloops,...)
{
res=estimateCondZ(graph,covars,maxiterations,initialclasses,selfloops)
attr(res,'class')<-c(attr(res,'class'),'cmestimation')
res
}
# Perform parameter estimation on 'graph' given the covariables 'covars'.
estimateCondXZ<-function(graph,covars,maxiterations,initialclasses,selfloops)
{
#resSimXZ = EMalgorithmXZ(TauIni,Y2,Adjacente,30,SelfLoop=FALSE)
res=EMalgorithmXZ(initialclasses,covars$y,graph$adjacency,maxiterations,selfloops)
cmestimation=list(mean=c(res$MuEstimated1,res$MuEstimated2),variance=res$VarianceEstimated,pi=res$PIEstimated,alpha=res$AlphaEstimated,tau=res$TauEstimated,jexpected=res$EJ,graph=graph)
attr(cmestimation,'class')<-c('cmestimationxz')
cmestimation
}
# Private estimation function used to allow various call conventions for estimation functions.
# Override of generic function for multiple covariables.
privateestimate.cmcovarxz<-function(covars,graph,maxiterations,initialclasses,selfloops,...)
{
res=estimateCondXZ(graph,covars,maxiterations,initialclasses,selfloops)
attr(res,'class')<-c(attr(res,'class'),'cmestimation')
res
}
# Generic estimation function applicable to graphs with covariables.
estimate<-function(graph,covars,...) UseMethod("estimate")
# Override of the generic estimation function. Performs the actual function dispatch depending on the class of covariables.
estimate.cmgraph<-function(graph,covars,maxiterations=20,initialclasses=t(rmultinom(size=1,prob=graph$proba,n=graph$nodes)),selfloops=FALSE,method=NULL,...)
{
if (length(method) == 0) {
res=privateestimate(covars,graph,maxiterations,initialclasses,selfloops,...)
} else {
res=method(graph,covars,maxiterations,initialclasses,selfloops)
attr(res,'class')<-c(attr(res,'class'),'cmestimation')
}
res
}
# Override of the generic pliot function for estimation results.
plot.cmestimation<-function(x,...)
{
par(mfrow = c(1,2))
plot(x$jexpected)
title("Expected value of J: Convergence criterion")
map=MAP(x$tau)
gplot(x$graph$adjacency,vertex.col=map$node.classes+2)
title("Network with estimated classes")
}
# Generic private ICL computation function for graphs and covariables.
privatecomputeICL<-function(covars,graph,qmin,qmax,loops,maxiterations,selfloops) UseMethod("privatecomputeICL")
# Private ICL computation function for graphs with single covariables.
privatecomputeICL.cmcovarz<-function(covars,graph,qmin,qmax,loops,maxiterations,selfloops)
{
res=ICL(X=graph$adjacency,Y=covars$y,Qmin=qmin,Qmax=qmax,loop=loops,NbIteration=maxiterations,SelfLoop=selfloops,Plot=FALSE)
attr(res,'class')<-c('cmiclz')
res
}
# Private ICL computation function for graphs with multiple covariables.
privatecomputeICL.cmcovarxz<-function(covars,graph,qmin,qmax,loops,maxiterations,selfloops)
{
res=ICL(X=graph$adjacency,Y=covars$y,Qmin=qmin,Qmax=qmax,loop=loops,NbIteration=maxiterations,SelfLoop=selfloops,Plot=FALSE)
attr(res,'class')<-c('cmiclxz')
res
}
# Generic public ICL computation function applicable to graph objects.
computeICL<-function(graph,covars,qmin,qmax,...) UseMethod("computeICL")
# Override of ICL computation function applicable to graph objects.
# Performs the actual method dispatch to private functions depending on the type of covariables.
computeICL.cmgraph<-function(graph,covars,qmin,qmax,loops=10,maxiterations=20,selfloops=FALSE,...)
{
res=privatecomputeICL(covars,graph,qmin,qmax,loops,maxiterations,selfloops)
res$qmin=qmin
res$qmax=qmax
res$graph=graph
res$covars=covars
attr(res,'class')<-c(attr(res,'class'),'cmicl')
res
}
# Override of the plot function for results of ICL computation.
plot.cmicl<-function(x,...)
{
par(mfrow = c(1,2))
result=x$iclvalues
maxi=which(max(result)==result)
plot(seq(x$qmin,x$qmax),result,type="b",xlab="Number of classes",ylab="ICL value")
points(maxi+x$qmin-1,result[maxi],col="red")
title("ICL curve")
best=x$EMestimation[[maxi+x$qmin-1]]
tau=best$TauEstimated
map=MAP(tau)
gplot(x$graph$adjacency,vertex.col=map$node.classes+2)
title("Network with estimated classes")
}
|
