inst/scripts/Nonlinear/cv.r

In gbonte/gbcode: Code from the handbook "Statistical foundations of machine learning"

## "INFOF422 Statistical foundations of machine learning" course
## R package gbcode 
## Author: G. Bontempi




N<- 100
n<-3

set.seed(555)
X <- rnorm(N*n)
X<-array(X,c(N,n))

Y<-X[,1]^2+5*X[,3]+4*log(abs(X[,2]))

N.tr <- N/2

X.tr <-X[1:N.tr,]
Y.tr <- Y[1:N.tr]

K<-10
N.k<-N/K
cv<-rep(0,K)
if (T) {
  for (k in 1:K)
  {
    I.k.ts<-((k-1)*N.k+1):(N.k*k)
    I.k.tr<-setdiff(1:N.tr,I.k.ts)
    X.k.tr <-X[I.k.tr,]
    Y.k.tr <- Y[I.k.tr]
    set.seed(555)
    model.nn<- nnet (X.k.tr,Y.k.tr,size=25, maxit=10000,linout=T,trace=F)
    X.k.ts <-X[I.k.ts,]
    Y.k.ts <- Y[I.k.ts]
    Y.k.hat.ts <- predict(model.nn,X.k.ts)
    test.k.MSE <- mean((Y.k.ts-Y.k.hat.ts)^2)
    cv[k]<-test.k.MSE
  }
}
I.ts<-(N.tr+1):N
X.ts <-X[I.ts,]
Y.ts <- Y[I.ts]



set.seed(555)
model.nn<- nnet (X.tr,Y.tr,size=25, maxit=10000,linout=T,trace=T)

Y.hat.tr <- predict(model.nn,X.tr)
empirical.MSE <- mean((Y.tr-Y.hat.tr)^2)
print(paste("Empirical MSE=",round(empirical.MSE,2)))
Y.hat.ts <- predict(model.nn,X.ts)
test.MSE <- mean((Y.ts-Y.hat.ts)^2)

print(paste("Test MSE=",round(test.MSE,2)))


cv.MSE <-mean(cv)
cv.MSE

print(paste("Cross-validation MSE=",round(cv.MSE,2)))