Learn R Programming

TeachingDemos (version 1.2)

faces: Chernoff Faces

Description

faces represent the rows of a data matrix by faces

Usage

faces(xy, which.row, fill = FALSE, nrow, ncol, scale = TRUE, byrow = FALSE, main, labels)

Arguments

xy
xy data matrix, rows represent individuals and columns attributes
which.row
defines a permutation of the rows of the input matrix
fill
if(fill==TRUE), only the first nc attributes of the faces are transformed, nc is the number of columns of xy
nrow
number of columns of faces on graphics device
ncol
number of rows of faces
scale
if(scale==TRUE), attributes will be normalized
byrow
if(byrow==TRUE), xy will be transposed
main
title
labels
character strings to use as names for the faces

Value

  • a plot of faces is created on the graphics device, no numerical results

Details

The features paramters of this implementation are: 1-height of face, 2-width of face, 3-shape of face, 4-height of mouth, 5-width of mouth, 6-curve of smile, 7-height of eyes, 8-width of eyes, 9-height of hair, 10-width of hair, 11-styling of hair, 12-height of nose, 13-width of nose, 14-width of ears, 15-height of ears. For details look at the literate program of faces

References

Chernoff, H. (1973): The use of faces to represent statistiscal assoziation, JASA, 68, pp 361--368. The smooth curves are computed by an algorithm found in Ralston, A. and Rabinowitz, P. (1985): A first course in numerical analysis, McGraw-Hill, pp 76ff. http://www.wiwi.uni-bielefeld.de/~wolf/ : S/R - functions : faces

See Also

---

Examples

Run this code
##---- Should be DIRECTLY executable !! ----
##-- ==>  Define data, use random,
##--	or do  help(data=index)  for the standard data sets.

faces(rbind(1:3,5:3,3:5,5:7))

data(longley)
faces(longley[1:9,])

set.seed(17)
faces(matrix(sample(1:1000,128,),16,8),main="random faces")

## The function is currently defined as
function(xy=rbind(1:3,5:3,3:5,5:7),which.row,fill=FALSE,nrow,ncol,
                scale=TRUE,byrow=FALSE,main,labels){

spline<-function(a,y,m=200,plot=FALSE){
    n<-length(a)
  h<-diff(a)
  dy<-diff(y)
  sigma<-dy/h
  lambda<-h[-1]/(hh<-h[-1]+h[-length(h)])
  mu<-1-lambda
  d<-6*diff(sigma)/hh
  tri.mat<-2*diag(n-2)
  tri.mat[2+  (0:(n-4))*(n-1)] <-mu[-1]
  tri.mat[    (1:(n-3))*(n-1)] <-lambda[-(n-2)]
  M<-c(0,solve(tri.mat)%*%d,0)
  x<-seq(from=a[1],to=a[n],length=m)
  anz.kl <- hist(x,breaks=a,plot=FALSE)$counts
  adj<-function(i) i-1
  i<-rep(1:(n-1),anz.kl)+1
  S.x<-  M[i-1]*(a[i]-x          )^3 / (6*h[adj(i)])  +
         M[i]  *(x        -a[i-1])^3 / (6*h[adj(i)])  +
	 (y[i-1] - M[i-1]*h[adj(i)]^2 /6) * (a[i]-x)/ h[adj(i)] +
        (y[i]   - M[i]  *h[adj(i)]^2 /6) * (x-a[i-1]) / h[adj(i)]
  if(plot){ plot(x,S.x,type="l"); points(a,y)    }
  return(cbind(x,S.x))
  }

n.char<-15
xy<-rbind(xy)
if(byrow) xy<-t(xy)
if(!missing(which.row)&& all(  !is.na(match(which.row,1:dim(xy)[2]))  ))
       xy<-xy[,which.row,drop=FALSE]
mm<-dim(xy)[2];  n<-dim(xy)[1]
xnames<-dimnames(xy)[[1]]
if(is.null(xnames)) xnames<-as.character(1:n)
if(!missing(labels)) xnames<-labels
if(scale){
   xy<-apply(xy,2,function(x){
           x<-x-min(x); x<-if(max(x)>0) 2*x/max(x)-1 else x })
} else xy[]<-pmin(pmax(-1,xy),1)
xy<-rbind(xy);n.c<-dim(xy)[2]
xy<-xy[,(h<-rep(1:mm,ceiling(n.char/mm))),drop=FALSE]
if(fill) xy[,-(1:n.c)]<-0
  
face.orig<-list( 
      eye  =rbind(c(12,0),c(19,8),c(30,8),c(37,0),c(30,-8),c(19,-8),c(12,0))
     ,iris =rbind(c(20,0),c(24,4),c(29,0),c(24,-5),c(20,0))           
     ,lipso=rbind(c(0,-47),c( 7,-49),lipsiend=c( 16,-53),c( 7,-60),c(0,-62))
     ,lipsi=rbind(c(7,-54),c(0,-54))                  # add lipsiend
     ,nose =rbind(c(0,-6),c(3,-16),c(6,-30),c(0,-31)) 
     ,shape =rbind(c(0,44),c(29,40),c(51,22),hairend=c(54,11),earsta=c(52,-4),
                  earend=c(46,-36),c(38,-61),c(25,-83),c(0,-89))
     ,ear  =rbind(c(60,-11),c(57,-30))                # add earsta,earend
     ,hair =rbind(hair1=c(72,12),hair2=c(64,50),c(36,74),c(0,79)) # add hairend
)
lipso.refl.ind<-4:1
lipsi.refl.ind<-1
nose.refl.ind<-3:1
hair.refl.ind<-3:1
shape.refl.ind<-8:1
shape.xnotnull<-2:8
nose.xnotnull<-2:3
  
nr<-n^0.5; nc<-n^0.5
if(!missing(nrow)) nr<-nrow
if(!missing(ncol)) nc<-ncol
opar<-par(mfrow=c(ceiling(c(nr,nc))),oma=rep(6,4), mar=rep(.7,4))
on.exit(par(opar))
         
  
for(ind in 1:n){
  
factors<-xy[ind,] 
face <- face.orig
  
m<-mean(face$lipso[,2])
face$lipso[,2]<-m+(face$lipso[,2]-m)*(1+0.7*factors[4])
face$lipsi[,2]<-m+(face$lipsi[,2]-m)*(1+0.7*factors[4])
face$lipso[,1]<-face$lipso[,1]*(1+0.7*factors[5])
face$lipsi[,1]<-face$lipsi[,1]*(1+0.7*factors[5])
face$lipso["lipsiend",2]<-face$lipso["lipsiend",2]+20*factors[6]
  
m<-mean(face$eye[,2])
face$eye[,2] <-m+(face$eye[,2] -m)*(1+0.7*factors[7])
face$iris[,2]<-m+(face$iris[,2]-m)*(1+0.7*factors[7])
m<-mean(face$eye[,1])
face$eye[,1] <-m+(face$eye[,1] -m)*(1+0.7*factors[8])
face$iris[,1]<-m+(face$iris[,1]-m)*(1+0.7*factors[8])
  
m<-min(face$hair[,2])
face$hair[,2]<-m+(face$hair[,2]-m)*(1+0.2*factors[9])
m<-0
face$hair[,1]<-m+(face$hair[,1]-m)*(1+0.2*factors[10])
m<-0
face$hair[c("hair1","hair2"),2]<-face$hair[c("hair1","hair2"),2]+50*factors[11]
  
m<-mean(face$nose[,2])
face$nose[,2]<-m+(face$nose[,2]-m)*(1+0.7*factors[12])
face$nose[nose.xnotnull,1]<-face$nose[nose.xnotnull,1]*(1+factors[13])
     
  
m<-mean(face$shape[c("earsta","earend"),1])
face$ear[,1]<-m+(face$ear[,1]-m)* (1+0.7*factors[14])
m<-min(face$ear[,2])
face$ear[,2]<-m+(face$ear[,2]-m)* (1+0.7*factors[15])
  
face<-lapply(face,function(x){ x[,2]<-x[,2]*(1+0.2*factors[1]);x})
face<-lapply(face,function(x){ x[,1]<-x[,1]*(1+0.2*factors[2]);x})
face<-lapply(face,function(x){ x[,1]<-ifelse(x[,1]>0,
                                        ifelse(x[,2] > -30, x[,1], 
                  pmax(0,x[,1]+(x[,2]+50)*0.2*sin(1.5*(-factors[3])))),0);x})
  
invert<-function(x) cbind(-x[,1],x[,2])
face.obj<-list(
     eyer=face$eye
    ,eyel=invert(face$eye)
    ,irisr=face$iris
    ,irisl=invert(face$iris)
    ,lipso=rbind(face$lipso,invert(face$lipso[lipso.refl.ind,]))
    ,lipsi=rbind(face$lipso["lipsiend",],face$lipsi,
                 invert(face$lipsi[lipsi.refl.ind,,drop=FALSE]),
                 invert(face$lipso["lipsiend",,drop=FALSE]))
    ,earr=rbind(face$shape["earsta",],face$ear,face$shape["earend",])
    ,earl=invert(rbind(face$shape["earsta",],face$ear,face$shape["earend",]))
    ,nose=rbind(face$nose,invert(face$nose[nose.refl.ind,]))
    ,hair=rbind(face$shape["hairend",],face$hair,invert(face$hair[hair.refl.ind,]),
                invert(face$shape["hairend",,drop=FALSE]))
    ,shape=rbind(face$shape,invert(face$shape[shape.refl.ind,]))
)

plot(1,type="n",xlim=c(-105,105)*1.1, axes=FALSE,
     ylab="",ylim=c(-105,105)*1.3)
title(xnames[ind])
for(ind in seq(face.obj)) {
       x <-face.obj[[ind]][,1]; y<-face.obj[[ind]][,2]
       xx<-spline(1:length(x),x,40,FALSE)[,2]
       yy<-spline(1:length(y),y,40,FALSE)[,2]
       lines(xx,yy)
  }
  }

if(!missing(main)){
  par(opar);par(mfrow=c(1,1))
  mtext(main, 3, 3, TRUE, 0.5)
  title(main)
  }
  }

Run the code above in your browser using DataLab