A playground where I was implementing the algorithm detailed in the "Centrality Based Visualization of Small World Graphs"

vanham_wattenburg.R

library(igraph)

g <- watts.strogatz.game(1, 100, 5, 0.05)
#g <- erdos.renyi.game(100, 3/100)
plot(g, vertex.size=2, vertex.label=NA)

# Calculate edge betweenness
eb <- cbind(E(g),edge_betweenness(g))
ebt <- eb[order(eb[,2], decreasing = T),]
head(ebt)
plot(ebt[,2], main="distribution of edge betweenness", ylab="Edge Between Score")
abline(h=35, col="red")
abline(v=100, col="red")

el <- get.edgelist(g)[as.vector(tail(ebt, 500-50)[,1]),] #edges as edgelist
elg <- graph_from_edgelist(el, directed=FALSE)
plot(elg, vertex.size=2, vertex.label=NA)

# Use low edge between scores to make minimum span tree
mst <- minimum.spanning.tree(elg)
layout.fruchterman.reingold(mst)
plot(mst, vertex.size=2, vertex.label=NA)

# FD layout
layout.fruchterman.reingold(mst)

# reinsert high edge between graph one at a time, starting with the highest

for (i in 1:dim(ebt)[1]) {
  new_el <- get.edgelist(g)[as.vector(ebt[i,1]),]
  el <- rbind(new_el,el) # note this will recalculate layout everytime, rather than just adding edges
  elg <- graph_from_edgelist(el, directed=FALSE)
  plot(elg, vertex.size=2, vertex.label=NA)
  readline(prompt="Press [ENTER] to continue")
}


# Everything below is unfinished an experimental
entropy.binary <- function(line) {
  p <- mean(line)
  return(-p*log2(p) - (1 - p)*log2((1 - p)))
}

binary_entropy <- function(graphObject) {
  # this is node entropy. I want edge entropy!
  # perhaps:
  # 1. Calculate all paths
  # 2. See how many edges make up each path. If an edge is part of pat, then 1, else 0
  # 3. 
  matrixObject <- as.matrix(get.adjacency(graphObject, type="both"))
  t <- cbind(dim(matrixObject)[1],apply(matrixObject, MARGIN=c(1), FUN=entropy.binary))
  t <- t[order(t[,2]),]
  return(t)
}