lda_thomaz: Linear Discriminant Analysis using the Thomaz-Kitani-Gillies...

In ramhiser/sparsediscrim: Sparse and Regularized Discriminant Analysis

Description

Given a set of training data, this function builds the Linear Discriminant Analysis (LDA) classifier, where the distributions of each class are assumed to be multivariate normal and share a common covariance matrix. When the pooled sample covariance matrix is singular, the linear discriminant function is incalculable. This function replaces the pooled sample covariance matrix with a regularized estimator from Thomaz et al. (2006), where the smallest eigenvalues are replaced with the average eigenvalue. Specifically, small eigenvalues here means that the eigenvalues are less than the average eigenvalue.

Given a set of training data, this function builds the Linear Discriminant Analysis (LDA) classifier, where the distributions of each class are assumed to be multivariate normal and share a common covariance matrix. When the pooled sample covariance matrix is singular, the linear discriminant function is incalculable. This function replaces the pooled sample covariance matrix with a regularized estimator from Thomaz et al. (2006), where the smallest eigenvalues are replaced with the average eigenvalue. Specifically, small eigenvalues here means that the eigenvalues are less than the average eigenvalue.

Usage

lda_thomaz(x, ...)

## Default S3 method:
lda_thomaz(x, y, prior = NULL, ...)

## S3 method for class 'formula'
lda_thomaz(formula, data, prior = NULL, ...)

## S3 method for class 'lda_thomaz'
predict(object, newdata, ...)

Arguments

x	matrix containing the training data. The rows are the sample observations, and the columns are the features.
...	additional arguments
y	vector of class labels for each training observation
prior	vector with prior probabilities for each class. If NULL (default), then equal probabilities are used. See details.
formula	A formula of the form groups ~ x1 + x2 + ... That is, the response is the grouping factor and the right hand side specifies the (non-factor) discriminators.
data	data frame from which variables specified in formula are preferentially to be taken.
object	trained lda_thomaz object
newdata	matrix of observations to predict. Each row corresponds to a new observation.

Details

The matrix of training observations are given in x. The rows of x contain the sample observations, and the columns contain the features for each training observation.

The vector of class labels given in y are coerced to a factor. The length of y should match the number of rows in x.

An error is thrown if a given class has less than 2 observations because the variance for each feature within a class cannot be estimated with less than 2 observations.

The vector, prior, contains the a priori class membership for each class. If prior is NULL (default), the class membership probabilities are estimated as the sample proportion of observations belonging to each class. Otherwise, prior should be a vector with the same length as the number of classes in y. The prior probabilities should be nonnegative and sum to one.

Value

lda_thomaz object that contains the trained classifier

list predicted class memberships of each row in newdata

References

Thomaz, C. E., Kitani, E. C., and Gillies, D. F. (2006). "A maximum uncertainty LDA-based approach for limited sample size problems with application to face recognition," J. Braz. Comp. Soc., 12, 2, 7-18.

Thomaz, C. E., Kitani, E. C., and Gillies, D. F. (2006). "A maximum uncertainty LDA-based approach for limited sample size problems with application to face recognition," J. Braz. Comp. Soc., 12, 2, 7-18.

Examples

n <- nrow(iris)
train <- sample(seq_len(n), n / 2)
lda_thomaz_out <- lda_thomaz(Species ~ ., data = iris[train, ])
predicted <- predict(lda_thomaz_out, iris[-train, -5])$class

lda_thomaz_out2 <- lda_thomaz(x = iris[train, -5], y = iris[train, 5])
predicted2 <- predict(lda_thomaz_out2, iris[-train, -5])$class
all.equal(predicted, predicted2)

ramhiser/sparsediscrim documentation built on May 26, 2019, 10:14 p.m.