Implements kernel logistic regression. More...

#include <KernelLogisticRegressor.h>

Public Member Functions
	KernelLogisticRegressor (const KernelCollection &data, const std::vector< int > &labels)
	Imports input from `data` and `labels`, and computes the Gram matrix of the data. More...

	KernelLogisticRegressor (const KernelCollection &data, const GMatrix &gramMatrix, const std::vector< int > &labels)
	Imports input from `data` and `labels` and the Gram matrix `gramMatrix`. More...

void	setData (const KernelCollection &data, const std::vector< int > &labels)
	Imports input from `data` and `labels`, and computes the Gram matrix of the data. More...

void	train (const double delta=0.0001, const unsigned itrNewton=5)
	Computes KLR weights. More...

bool	isTrained ()
	Returns `true` if the classifier has been trained.

Vector2	test (const kernel::base &t) const
	Returns a pair of values which indicate the probability of classes `1` and `2`.

GMatrix	test (const KernelCollection &testSet) const
	Returns the probabilities of classes `1` and `2` for all data points in `testSet`. More...

const GMatrix &	vklr () const
	Returns KLR weights. More...

Detailed Description

Implements kernel logistic regression.

This class implements a two-class nonlinear classifier for \( SE(3) \) data. KLR is discussed in Kernel Logistic Regression.

This class is based on libklr, which is provided with Makoto Yamada's IWKLR.

Examples: klr_classify.cpp.

Definition at line 27 of file KernelLogisticRegressor.h.

Constructor & Destructor Documentation

◆ KernelLogisticRegressor() [1/2]

nuklei::KernelLogisticRegressor::KernelLogisticRegressor	(	const KernelCollection &	data,
		const std::vector< int > &	labels
	)

Imports input from data and labels, and computes the Gram matrix of the data.

The two structures data and labels should contain the same number of elements. The \( i^{\mathrm{th}} \) element of labels indicates the class of the \( i^{\mathrm{th}} \) element of data. The two allowed values in labels are 1 and 2.

Definition at line 16 of file KernelLogisticRegressor.cpp.

◆ KernelLogisticRegressor() [2/2]

nuklei::KernelLogisticRegressor::KernelLogisticRegressor	(	const KernelCollection &	data,
		const GMatrix &	gramMatrix,
		const std::vector< int > &	labels
	)

Imports input from data and labels and the Gram matrix gramMatrix.

The two structures data and labels should contain the same number of elements. The \( i^{\mathrm{th}} \) element of labels indicates the class of the \( i^{\mathrm{th}} \) element of data. The two allowed values in labels are 1 and 2.

The data held in gramMatrix should correspond to the Gram matrix of data.

Definition at line 27 of file KernelLogisticRegressor.cpp.

References NUKLEI_ASSERT.

Member Function Documentation

◆ setData()

void nuklei::KernelLogisticRegressor::setData	(	const KernelCollection &	data,
		const std::vector< int > &	labels
	)

Imports input from data and labels, and computes the Gram matrix of the data.

The two structures data and labels should contain the same number of elements. The \( i^{\mathrm{th}} \) element of labels indicates the class of the \( i^{\mathrm{th}} \) element of data. The two allowed values in labels are 1 and 2.

This method also resets all the other member variables of the class.

Definition at line 38 of file KernelLogisticRegressor.cpp.