NNetIterations: Neural Network algorithm using iterations.

In SixianZhang/CS499-Coding-Project-3: Neural network for regression and binary classification

Description

This neural network algorithm has one output and one hidden layer, and stops when max.iterations is reached.

Usage

NNetIterations(X.mat, y.vec, max.iterations, step.size, n.hidden.units,
  is.train)

Arguments

X.mat	numeric feature matrix of size [n_observations x n_features].
y.vec	numeric label vector of length n_observations.
max.iterations	integer scalar greater than 1.
step.size	numeric positive scalar.
n.hidden.units	number of hidden units, greater than or equal to 1.
is.train	logical vector of length n_observations, TRUE if the observation is for training, FALSE for validation

Value

result.list with named elements: pred.mat, n_observations x max.iterations matrix of predicted values. V.mat final weight matrix (n_features+1 x n.hidden.units). w.vec final weight vector (n.hidden.units+1). predict(testX.mat) a function that takes a test features matrix and returns a vector of predictions.

Examples

data(spam, package = "ElemStatLearn")
X.mat <- data.matrix(spam[,-ncol(spam)])
y.vec <- as.vector(ifelse(spam$spam == 'spam',1,0))
max.iteration <- 50L
step.size <- 0.02
n.hidden.units = 20L
temp <- sample(rep(1:2,l=length(y.vec)))
is.train <- (temp == 1)
result.list <- NNetIterations(X.mat, y.vec, max.iteration, step.size, n.hidden.units, is.train = is.train)

SixianZhang/CS499-Coding-Project-3 documentation built on May 21, 2019, 1:42 p.m.