# Neural Network Example: Forward Propagation for Simple Network
w2 = c(0.171, -0.0772)
b2 = c(0.732)
w1 = c(-0.265, -0.420)
b1 = c(-0.475, -0.140)
logsig = function(x) { return(1 / (1 + exp(-x))) }
purelin = function(x) { return(x) }
a0 = 1
a1 = logsig(w1 * a0 + b1)
a2 = purelin(t(w2) %*% a1 + b2)
a2




# Neural Network Example: Sonar Classification
set.seed(2^17 - 1)
library(mxnet)
require(mlbench)
data(Sonar, package = "mlbench")
Sonar[,61] = as.numeric(Sonar[,61]) - 1
n = nrow(Sonar)
index = sample(1:n)
m = round(0.8 * n)
train.x = data.matrix(Sonar[index[1:m], 1:60])
train.y = Sonar[index[1:m], 61]
test.x = data.matrix(Sonar[index[(m+1):n], 1:60])
test.y = Sonar[index[(m+1):n], 61]
model = mx.mlp(train.x, train.y, hidden_node = 10, out_node = 2, out_activation = "softmax", learning.rate = 0.1, num.round = 1000)
predictions = predict(model, test.x)
table(test.y, predictions[2,] > 0.5)
graph.viz(model$symbol)




# Genetic Algorithm for Feature Selection Example
set.seed(2^17 - 1)
library(mlbench)
library(caret)
n = 100
sigma = 1
set.seed(1)
sim = mlbench.friedman1(n, sd = sigma)
colnames(sim$x) <- c(paste("real", 1:5, sep = ""), paste("bogus", 1:5, sep = ""))
x = sim$x
y = sim$y
normalization = preProcess(x)
x = predict(normalization, x)
x = as.data.frame(x)
gafsControl = gafsControl(functions = rfGA, method = "cv", number = 5)
set.seed(2^17-1)
start = Sys.time()
model = gafs(x = x, y = y, iters = 5, gafsControl = gafsControl)
stop = Sys.time()
model




# Naive Bayes Example
set.seed(2^17-1)
library(e1071)
data(HouseVotes84, package = "mlbench")
n = nrow(HouseVotes84)
m = round(0.8 * nrow(HouseVotes84))
shuffle = sample(1:n)
trn = HouseVotes84[shuffle[1:m], c(1,4,5)]
tst = HouseVotes84[shuffle[(m+1):n], c(1,4,5)]
model = naiveBayes(Class ~ ., data = trn)
model
names(model)
model$apriori
model$tables
prior = table(trn$Class) / sum(table(trn$Class))
prior
table(trn$V3[trn$Class == "democrat"]) /
  sum(table(trn$V3[trn$Class == "democrat"]))
table(trn$V3[trn$Class == "republican"]) /
  sum(table(trn$V3[trn$Class == "republican"]))
table(trn$V4[trn$Class == "democrat"]) /
  sum(table(trn$V4[trn$Class == "democrat"]))
table(trn$V4[trn$Class == "republican"]) /
  sum(table(trn$V4[trn$Class == "republican"]))
predict(model, tst[c(1,2,4,15),])
predict(model, tst[c(1,2,4,15),], type = "raw")
observation = tst[1,]

(
 prior["democrat"] * 
   model$tables$V3["democrat",observation$V3] * 
   model$tables$V4["democrat",observation$V4]
) /
(
 prior["democrat"] *
   model$tables$V3["democrat",observation$V3] *
   model$tables$V4["democrat",observation$V4] +
 prior["republican"] *
   model$tables$V3["republican",observation$V3] *
   model$tables$V4["republican",observation$V4]
)

(
 prior["republican"] * 
   model$tables$V3["republican",observation$V3] * 
   model$tables$V4["republican",observation$V4]
) /
(
 prior["democrat"] *
   model$tables$V3["democrat",observation$V3] *
   model$tables$V4["democrat",observation$V4] +
 prior["republican"] *
   model$tables$V3["republican",observation$V3] *
   model$tables$V4["republican",observation$V4]
)