Code used for post: http://btovar.com/2015/03/neural-networks-fit-eq/

nn_fit_eq.R

# **********************************
# Author: Benjamin Tovar
# Date: March 25, 2015
# Post: Using Neural Networks to fit equations in R
# Post url: http://btovar.com/2015/03/neural-networks-fit-eq/
# **********************************

# load libraries
library(RSNNS
library(ggplot2)


# set function 
f <- function(x) {
  # this is the equation we will
  # try to fit using a Neural Network
  y <- exp(-x/10)*sin(x)
  return(y)
}

# set the input
x <- seq(from=0,to=20,by=0.1)
# compute the output 
y <- f(x)
dataset <- data.frame(x=x,y=y)

# **************
# plot the function
# **************

ggplot(dataset) + 
  aes(x=x,y=y) + 
  geom_point(colour="#ff6f69") + 
  geom_line(colour="#ff6f69") + 
  ylim(c(-1,1)) + 
  xlim(c(0,20)) + 
  labs(title="Equation to fit: f(x) = exp(-x/10)*sin(x)")

# ********************************************
# train the NN model
# use the values of x to estimate values of y
# ********************************************

model <- mlp(dataset$x, dataset$y, size=15, 
             maxit=10000, linOut=TRUE, 
             learnFuncParams=c(0.01, 0),
             hiddenActFunc="Act_TanH")

# details about the model
# summary(model)
# model
# weightMatrix(model)
# extractNetInfo(model)

# *****************
# plot the training error of
# *****************

plotIterativeError(model,main="Training error",col="#ff6f69",ylim=c(0,25));grid()

# ***********************
# use the trained model
# to predict values of y
# ***********************

# predict y
y_pred <- predict(model,as.matrix(dataset$x))
# compute the length of predictions
l <- length(y_pred)

# ****************************************
# set the format of the dataset for ggplot
# *****************************************

# add column of predicted y to the dataset
dataset <- cbind(dataset,y_pred)
# option 1: constructing the data.frame by hand
  dataset_2 <- data.frame(x=c(dataset$x,dataset$x),
                         value=c(y,y_pred),
                         variable=c(rep("y",l),rep("y_pred",l)))

# option 2: using reshape library and melt function
  library(reshape)
  dataset_2 <- melt(dataset,id=c("x"))

# ****************************************
# plot the results (y and predicted y)
# *****************************************

ggplot(dataset_2) + 
  aes(x=x,y=value) + 
  geom_point(aes(colour=variable,fill=variable,shape=variable),alpha=0.6) + 
  geom_line(aes(colour=variable),alpha=0.4) + 
  ylim(c(-1,1)) + 
  xlim(c(0,20)) + 
  labs(title="Use of package RSNNS in R to fit an equation",y="y")

Is line 9 missing a ) ?