Experiment with Wikipedia Data

load("Wikidata.RData")

# list of 4 objects, are already dissimilarity matrices
D<-list()
for(i in 1:4){D[[i]]<-as.matrix(Wikidata[[i]])/sqrt(sum(as.matrix(Wikidata[[i]])^2))}

# embed the dissimilarities
t<-proc.time();smds<-fJOFC(10,D,10,verbose=FALSE);runtime1<-proc.time()-t

## Loading required package: shapes

## Warning in rgl.init(initValue, onlyNULL): RGL: unable to open X11 display

## Warning in fun(libname, pkgname): Error in 'rgl_init'

## 
## Attaching package: 'shapes'

## The following object is masked from 'package:stats':
## 
##     sigma

## The following object is masked from 'package:igraph':
## 
##     V

library('ape')

# hclust the points
set.seed(12345)
X<-smds$conf
HC<-hclust(dist(X),method="ward.D")

# compute the dendrogram merge height
CD<-cophenetic(HC)
CD<-as.matrix(CD)
DMH<-rep(0,1382)
for(i in 1:1382){
  MHs<-c(CD[i,i+1382],CD[i,i+1382*2],CD[i,i+1382*3],CD[i+1382,i+1382*2],CD[i+1382,i+1382*3],CD[i+1382*2,i+1382*3])
  DMH[i]<-max(MHs)
}
# figure 5, left panel
hist(DMH,col='red',main=paste("Dendrogram Merge Height for the \n 1382 points"),labels = TRUE, ylim=c(0,1400),xlab=paste("Merge Height"))

# figure 5, right panel
DMH2<-DMH[DMH<50]
hist(DMH2,col='red',main=paste("Dendrogram Merge Height for the \n 1226 points with DMH<50"),labels = TRUE, ylim=c(0,1100),xlab=paste("Merge Height"))

# figure 6

h=(0:30)/10
ari<-rep(0,31)
for(i in 1:31){L<-cutree(HC,h=h[i]); ari[i]<-adjustedRandIndex(L,rep(1:1382,4))}
ari<-cbind(h,ari)
plot(ari,type="o",col='red',main=paste("ARI of HClust versus ground truth"),ylab=paste("ARI"),xlab=paste("Cut Height"))

# figure 7
# note that coloring the branches was manually done 

d <- cut(as.dendrogram(HC), h=25)
par(cex=0.2, mar=c(5, 8, 4, 1))
plot(d$lower[[25]],yaxes=NULL)
par(cex=1)
title(xlab="", ylab="", main="")

## run JOFC for time conparison
# make the weight matrix
# n=1382
# w=10
# W1 <- matrix(1,n,n) - diag(n)
# W2 <- diag(w,n)
# W <- W2
# for(i in 1:3){W<-cbind(W,W2)}
# WW<-W
# for(i in 1:3){W<-rbind(W,WW)}
# W<-W-diag(w,1382*4)
# W <- W+as.matrix(bdiag(rep(list(W1),4)))
# 
# # make the omnibus matrix
# d<-list()
# for(i in 1:4){
#   d[[i]]<-D[[i]]
#   for(l in 1:3){d[[i]]<-cbind(d[[i]],D[[i]])}
# }
# O<-Reduce(rbind,d)
# O<-O/2+t(O)/2
# 
# t<-proc.time();smds2 <- smacofSym(O,10,weightmat=W,verbose=FALSE);runtime2<-proc.time()-t