library(MASS)
library(class)
library(nnet)
library(e1071)

N=200

a<-cbind(rnorm(N,mean=0,sd=0.5),rnorm(N,mean=1,sd=0.5))
b<-cbind(rnorm(N,mean=0,sd=0.5),rnorm(N,mean=-1,sd=0.5))
ab<-rbind(a,b)
ab<-data.frame(ab,rep(1,100))
names(ab)<-c("x1","x2","y")

c<-cbind(rnorm(N,mean=-1,sd=0.5),rnorm(N,mean=0,sd=0.5))
d<-cbind(rnorm(N,mean=1,sd=0.5),rnorm(N,mean=0,sd=0.5))
cd<-rbind(c,d)
cd<-data.frame(cd,rep(0,100))
names(cd)<-c("x1","x2","y")

dat<-rbind(ab,cd)

xp <- seq(min(dat$x1), max(dat$x1), length = 100); np <- length(xp)
yp <- seq(min(dat$x2), max(dat$x2), length = 100)
pt <- expand.grid(x1 = xp, x2 = yp)

dat.svm<-svm(as.matrix(dat)[,-3], as.factor(dat$y), kernel="polynomial", degree=2)
dat.svm.t <- class.ind(predict(dat.svm, pt )) 

zp.svm <- dat.svm.t[,1]-dat.svm.t[,2]
plot(dat$x1,dat$x2,col=dat$y+1)
title("SVM XOR")
mtext("polynomial, d=2")
contour(xp, yp, matrix(zp.svm, np), add = T, levels = 0, labex = 0)

dat.svm.cv<-svm(as.matrix(dat)[,-3], as.factor(dat$y), kernel="polynomial", degree=2, cross=10)