crda: Compressive Regularized Discriminant Analysis (CRDA) Method

Description Usage Arguments Details Value Author(s) See Also Examples

Description

The function crda implements Compressive Regularized Discriminant Analysis (CRDA) approach and performs simultaneous feature selection and classification of high-dimensional data. CRDA approach aims to address three facets of high-dimensional classification: namely, accuracy, computational complexity, and interpretability.

Usage

1
2
crda(X, y, q = Inf, al = NULL, K = NULL, Xt = NULL,
  prior = "uniform", centerX = FALSE)

Arguments

X

Training dataset, a pxn matrix with n-samples each having p-features.

y

Labels for training dataset, an nx1 vector of whole numbers.

q

Type of Lq,1 norm, default is Linf-norm.

al

Regularization parameter.

K

Joint-sparsity level.

Xt

Test dataset, a pxnt matrix with nt-samples each of p-features.

prior

Type of prior class probabilities, either 'uniform' (default) or 'estimated'.

centerX

Flag for grand-mean centering of test dataset using grand-mean of training dataset.

Details

crda

Value

An object obj of class crda with the following attributes:

funCall

The call to the crda function.

prior

Prior class probabilities.

varSelRate

Feature selection rate (FSR).

selVarPos

Position (i.e., index) of selected features.

coefMat

Coefficient matrix before feature selection.

shrunkenCoefMat

Shrunken (rowsparse) coefficient matrix.

const

The constant part of discriminant function for CRDA method.

predTrainLabels

Predicted labels for training dataset.

predTestLabels

Optional: Predicted labels for test dataset, if it is available.

regparam

Optional: The value of regularization parameter.

muX

Optional: Grand-mean, i.e., row (feature) wise mean of training dataset.

Author(s)

Muhammad Naveed Tabassum and Esa Ollila, 2018

See Also

crda.regparam, crda.cv

Examples

1
2
3
4
5
crda(X,y)
crda(X,y, Xt = testdata)
crda(X,y, q = 1, Xt = testdata)
crda(X,y, q = 2, Xt = Xt, centerX = TRUE)
crda(X,y, Xt = Xt, prior = 'estimated', centerX = TRUE)

mntabassm/compressiveRDA documentation built on May 31, 2019, 5:22 p.m.