#
# Epage 126
# Epage 134
#
# Written by:                                                         
# --                                                                  
# John L. Weatherwax                2009-04-21                        
#                                                                     
# email: wax@alum.mit.edu                                             
#                                                                     
# Please send comments and especially bug reports to the              
# above email address.                                                
#                                                                     
#-----                                                                

source("load_vowel_data.R")
source("reduced_rank_LDA.R")

out    = load_vowel_data( FALSE, FALSE )
XTrain = out[[1]]
yTrain = out[[2]] 
XTest  = out[[3]]
yTest  = out[[4]] 

out = reduced_rank_LDA( XTrain, yTrain, XTest, yTest )

K = length(unique(yTrain)) # the number of classes (expect the classes to be labeled 1, 2, 3, ..., K-1, K 
p = dim( XTrain )[2] # the number of features

TrainClassification = out[[4]]
TestClassification = out[[5]]

#postscript("../../WriteUp/Graphics/Chapter4/dup_fig_4_10.eps", onefile=FALSE, horizontal=FALSE)

train_error_rate = matrix( data=0, nrow=1, ncol=p )
test_error_rate = matrix( data=0, nrow=1, ncol=p )

NTrain = dim(XTrain)[1]
NTest = dim(XTest)[1]

for( pi in 1:p ){
   train_error_rate[pi] = sum( TrainClassification[,pi] != yTrain )/NTrain
   test_error_rate[pi] = sum( TestClassification[,pi] != yTest )/NTest
}

plot( 1:p, train_error_rate, col="red", ylim=c( 0.3, 0.7 ), type="b", xlab="Dimension", ylab="Misclassification rate" ) # range( c(train_error_rate,test_error_rate) )
lines( 1:p, test_error_rate, col="blue", type="b" )

#dev.off()