LinearDiscriminantClassifier {RSSL}R Documentation

Linear Discriminant Classifier

Description

Implementation of the linear discriminant classifier. Classes are modeled as Gaussians with different means but equal covariance matrices. The optimal covariance matrix and means for the classes are found using maximum likelihood, which, in this case, has a closed form solution.

Usage

LinearDiscriminantClassifier(X, y, method = "closedform", prior = NULL,
  scale = FALSE, x_center = FALSE)

Arguments

X

Design matrix, intercept term is added within the function

y

Vector or factor with class assignments

method

the method to use. Either "closedform" for the fast closed form solution or "ml" for explicit maximum likelihood maximization

prior

A matrix with class prior probabilities. If NULL, this will be estimated from the data

scale

logical; If TRUE, apply a z-transform to the design matrix X before running the regression

x_center

logical; Whether the feature vectors should be centered

Value

S4 object of class LeastSquaresClassifier with the following slots:

modelform

weight vector

prior

the prior probabilities of the classes

mean

the estimates means of the classes

sigma

The estimated covariance matrix

classnames

a vector with the classnames for each of the classes

scaling

scaling object used to transform new observations

See Also

Other RSSL classifiers: EMLeastSquaresClassifier, EMLinearDiscriminantClassifier, GRFClassifier, ICLeastSquaresClassifier, ICLinearDiscriminantClassifier, KernelLeastSquaresClassifier, LaplacianKernelLeastSquaresClassifier(), LaplacianSVM, LeastSquaresClassifier, LinearSVM, LinearTSVM(), LogisticLossClassifier, LogisticRegression, MCLinearDiscriminantClassifier, MCNearestMeanClassifier, MCPLDA, MajorityClassClassifier, NearestMeanClassifier, QuadraticDiscriminantClassifier, S4VM, SVM, SelfLearning, TSVM, USMLeastSquaresClassifier, WellSVM, svmlin()


[Package RSSL version 0.9.7 Index]