[R] A simple perceptron neural network (nnet)

[Eduardo Grajeda]

Hello,

I've started to learn about neural networks and the first examples
I've seen are the implementation of an OR logical gate, aswell as the
AND gate. I've implemented it as a perceptron with no hidden layers,
and I've done it this way because so far is the only way I've learned.
The R file with the implementation of the OR gate is:

ppton_or.R
----------------------------------------------------------------------
ppton_or <- function() {
         x = array(c(1,1,1,1,0,0,1,1,0,1,0,1), dim=c(4,3))
         y = c(0,1,1,1)
         w = c(0,0,0)
         b = 1
         n = 1
         k = 0

         while(all(as.integer((x[,] %*% w) >= 0) == y) == FALSE) {
                z = as.integer((x[n,] %*% w) >= 0)
                if(z != y[n]) {
                     w = w+b*(y[n]-z)*x[n,];
                }

                n = n%%4+1
                k = k+1
         }

         print(k)
         print(w)
}
----------------------------------------------------------------------

I've would like to know if it is possible to implement this pretty
basic neural network with the nnet package. I've tried using the
"skip=TRUE" switch with "size=0" and filling x and y with the training
data but it is not working. Neither do I know how to make it use a
heaviside function as the threshold function. If someone could give me
some hint I'll be pretty grateful.

Thanks,
Eduardo Grajeda.