In Albluca/rpgraph: r interface to VDAOEngine to build elastic principal graphs

This document describes how to produce a sequence of principal trees to that can be used to observe the dynamics of the algorithm and to produce animations.

Building a principal tree

The simple tree dataset included in the package describes points placed on a three dimensional tree and we can use it to test the usage of principal trees. The computeElasticPrincipalGraph include an optional argument NodeStep that ca be used to obtain the intermediate steps. For example setting NodeStep = 2 will save all the intermediate principal trees after 2 nodes have been added, i.e. will output tress with 2, 4, 6, 8, ... nodes.

library(rpgraph)
Results <- computeElasticPrincipalGraph(Data = simple_tree, NumNodes = 20, Method = 'DefaultPrincipalTreeConfiguration', NodeStep = 2)

Each tree is saved as a different element of the Results list. Results[[1]] contains the tree with 2 nodes, Results[[2]] the tree with 4 nodes and so on. Therefore Results can be used to observe different snapshot of the construction process:

par(mfcol = c(4,3))

for(i in 1:length(Results)){
  plotData2D(Data = simple_tree, PrintGraph = Results[[i]] ,
           GroupsLab = rep(1, nrow(simple_tree)),
           Main = paste("Pincipal tree", i),
           Xlab = "Dimension 1", Ylab = "Dimension 2")
}

or to produce a series of plot that can be used for an animation:

for(i in 1:length(Results)){
  plotData2D(Data = simple_tree, PrintGraph = Results[[i]] ,
           GroupsLab = rep(1, nrow(simple_tree)),
           PlotProjections = "onEdges",
           Main = paste("Pincipal tree", i),
           Xlab = "Dimension 1", Ylab = "Dimension 2")
  Sys.sleep(1)
}