LDA Class Reference

Linear Discriminant Analysis (LDA). More...

#include <LDA.h>

Inheritance diagram for LDA:

List of all members.

Public Member Functions
	LDA ()
	Constructor.
	~LDA ()
	Destructor.
void	init (Model &model)
	The initialization checks for compatibility of the model.
double	optimize (Model &model, ErrorFunction &error, const Array< double > &input, const Array< double > &target)
	The inherited optimize function inherits one of the specialized versions for LinearClassifier or AffineLinearFunction models.
double	optimize (LinearClassifier &model, const Array< double > &input, const Array< double > &target)
	Compute the LDA solution for a multi-class problem.
double	optimize (AffineLinearFunction &model, const Array< double > &input, const Array< double > &target)
	Compute the LDA solution for a binary classification problem with is stated as a zero-thresholded affine linear function.

---

Detailed Description

Linear Discriminant Analysis (LDA).

This optimizer trains a linear model using linear discriminant analysis.

Definition at line 63 of file LDA.h.

---

Constructor & Destructor Documentation

LDA::LDA

(

)

Constructor.

Definition at line 52 of file LDA.cpp.

LDA::~LDA

(

)

Destructor.

Definition at line 56 of file LDA.cpp.

---

Member Function Documentation

void LDA::init

(

Model &

model

)

[virtual]

The initialization checks for compatibility of the model.

It must be either a LinearClassifier or a AffineLinearFunction in case of binary classification.

Implements Optimizer.

Definition at line 61 of file LDA.cpp.

double LDA::optimize	(	AffineLinearFunction &	model,
		const Array< double > &	input,
		const Array< double > &	target
	)

Compute the LDA solution for a binary classification problem with is stated as a zero-thresholded affine linear function.

In this case the targets must be one-dimensional, taking the values +1 and -1.

Definition at line 154 of file LDA.cpp.

References Model::getInputDimension(), i, and Model::setParameter().

double LDA::optimize	(	LinearClassifier &	model,
		const Array< double > &	input,
		const Array< double > &	target
	)

Compute the LDA solution for a multi-class problem.

Let c be the number of classes. Then the targets are assumes to be vectors in padded with zeros and a single one in the component corresponding to the target class.

Definition at line 78 of file LDA.cpp.

References Model::getInputDimension(), LinearClassifier::getNumberOfClasses(), i, and LinearClassifier::setParameter().

double LDA::optimize	(	Model &	model,
		ErrorFunction &	error,
		const Array< double > &	input,
		const Array< double > &	target
	)			[virtual]

The inherited optimize function inherits one of the specialized versions for LinearClassifier or AffineLinearFunction models.

ATTENTION: The target data must be organized differently for these models!

Implements Optimizer.

Definition at line 68 of file LDA.cpp.

---

The documentation for this class was generated from the following files:

• LDA.h
• LDA.cpp