An example of using normalised mutual information in R

normalised_mutual_information.R

#load the mutual information library
library(mpmi)

#define normalised continuous mutual information, bias corrected
ncmi<-function(cts,...){
  MIunnorm<-cmi(cts,...)
  MIbcmiself<-diag(MIunnorm$bcmi)
  MIbcminorm<-outer(MIbcmiself,MIbcmiself,FUN = "*")
  MInormed<-MIunnorm$bcmi / sqrt(MIbcminorm)
  colnames(MInormed)<-colnames(cts)
  return(MInormed)
}

#look at an iris example usage with corrplot
library(tidyverse)
library(corrplot)

iris.measurements<-iris%>%
  dplyr::select(1:4)%>%
  na.omit()

corrplot.mixed(ncmi(iris.measurements),
               cl.lim = c(0,1),
               title = "Normalised Mutual Information between Iris dimensions",
               mar=c(0,0,1,0))