R/PlotLikelihoods.R

In PDEnaiveBayes: Plausible Naive Bayes Classifier Using PDE

PlotLikelihoods=function(Likelihoods,Data,PlausibleLikelihoods=NULL,Epsilon=NULL,PlausibleCenters=NULL,PlotCutOff=4,xlim){
# PlotLikelihoods(Likelihoods,Data)
#Input
#Likelihoods            List of d numeric matrices, one per feature, each matrix with 1:k columns containing 
#                             the distribution of class 1:k per feature, i.e., the Likelihood per class
#Data                         numerical data matrix [1:n,1:k]
#OPTIONAL
#PlausibleLikelihoods     List of d numeric matrices of plausible Likelihood, one per feature, each matrix with [1:n,1:k] with k columns containing 
#                             the distribution of class 1:k per feature 
#Epsilon                      scalar defining epsiolon fo plausible likelihoods
#PlausibleCenters             [d:k] plausible centers used to compute plausible likelihoods
#PlotCutOff                   scalar defining the how many feature starting from 1 should ne plottet or numerical vector defining the index of features to be plottet
#                             in second case should not be too many otherwise plot yields an error
#OUTPUT
# native plots in grid

#author MCT 05/2025
#reset par after plotting
if(!is.matrix(Data)){
  Data=as.matrix(Data)
}
if(is.list(Likelihoods)){
  Likelihoods=listOfLikelihoods2Array(Likelihoods)
}
def.par <- par(no.readonly = TRUE)
on.exit(par(def.par))
if(is.list(Likelihoods)){
  #number of dimensions
  dinit = length(Likelihoods)
  class_l=ncol(Likelihoods[[1]])
  # Set layout: 2 rows per item in Likelihoods
}else{
  V=dim(Likelihoods)
  N=V[1]
  class_l=V[2]
  dinit=V[3]
}


if(!is.null(PlotCutOff)){
  if(is.numeric(PlotCutOff)){
    if(length(PlotCutOff)==1){
      d=min(c(dinit,PlotCutOff),na.rm = T)
      rangePlots=1:d
    }else{
      rangePlots=PlotCutOff
      rangePlots[rangePlots>dinit]=dinit
      rangePlots=rangePlots[rangePlots>0]
      d=max(rangePlots,na.rm = T)-min(rangePlots,na.rm = T)+1
    }
  }else{
    rangePlots=1:dinit
    d=dinit
  }
}else{
  rangePlots=1:dinit
  d=dinit
}

if(is.null(PlausibleLikelihoods)){
  par(mfrow = c(d, 1))  # d rows, 1 columns
}else{
  par(mfrow = c(d, 2))  # d rows, 2 columns
}

if(isTRUE(requireNamespace("DataVisualizations",quietly = T))){
  Colors=DataVisualizations::DefaultColorSequence[1:class_l]
}else{
  warning("PlotLikelihoods: Please install DataVisualizations package for color setting.")
  Colors=rep(c("black","blue","green","gold","magenta","red","grey","orange"),class_l)
  Colors=Colors[1:class_l]
}

#plot all dimensions

for(f in rangePlots){
  x=Data[,f]
  ind=order(x,decreasing = F)
  fnam=colnames(Data)[f]
  if(is.list(Likelihoods)){
    LL=Likelihoods[[f]]
  }else{
    LL=Likelihoods[,,f]
  }

  maxY1=apply(LL,2,function(x) max(x[is.finite(x)]))
  
  if(!is.null(PlausibleLikelihoods)){
    if(is.list(Likelihoods)){
      LL2=PlausibleLikelihoods[[f]]
    }else{
      LL2=PlausibleLikelihoods[,,f]
    }
    maxY2=apply(LL2,2,function(x) max(x[is.finite(x)]))
    ylim=c(0,max(c(maxY1,maxY2)))
  }else{
    ylim=c(0,c(max(maxY1)))
  }
  
  if(missing(xlim)){
    plot(x[ind],LL[ind,1],type="l",xlab=paste("Feature",fnam),col=Colors[1],main="Bayesian",
         ylim=ylim,lwd=2,ylab="Class Likelihoods")
  }else{
    plot(x[ind],LL[ind,1],type="l",xlab=paste("Feature",fnam),col=Colors[1],main="Bayesian",
         ylim=ylim,lwd=2,ylab="Class Likelihoods",xlim=xlim)
  }
  for(i in 2:class_l){
    points(x[ind],LL[ind,i],type="l",col=Colors[i],ylim=ylim,lwd=2)
  }
  if(!is.null(PlausibleLikelihoods)){
    if(missing(xlim)){
      plot(x[ind],LL2[ind,1],type="l",xlab=paste("Feature",fnam),col=Colors[1],ylab="Class Likelihoods",
           main=paste0("Plausible Bayes with approx. Limit ",round(Epsilon[f]/class_l,3)),ylim=ylim,lwd=2)
    }else{
      plot(x[ind],LL2[ind,1],type="l",xlab=paste("Feature",fnam),col=Colors[1],ylab="Class Likelihoods",
           main=paste0("Plausible Bayes with approx. Limit ",round(Epsilon[f]/class_l,3)),ylim=ylim,lwd=2,xlim=xlim)
    }

    if(!is.null(PlausibleCenters)){
      PlausibleCenters_cur=PlausibleCenters[f,]
      abline(v=PlausibleCenters_cur[1],col=Colors[1])
    }
    for(i in 2:class_l){
      points(x[ind],LL2[ind,i],type="l",col=Colors[i],ylim=ylim,lwd=2)
      if(!is.null(PlausibleCenters)){
        abline(v=PlausibleCenters_cur[i],col=Colors[i])
      }
    }
  }#end in case of plausible
}#end for rangePlots
}