This class implements mean shift clustering. More...

Public Member Functions
	MeanShift (const double radius=0, const size_t maxIterations=1000, const KernelType kernel=KernelType())
	Create a mean shift object and set the parameters which mean shift will be run with. More...

void	Cluster (const MatType &data, arma::Col< size_t > &assignments, arma::mat &centroids, bool useSeeds=true)
	Perform mean shift clustering on the data, returning a list of cluster assignments and centroids. More...

double	EstimateRadius (const MatType &data, const double ratio=0.2)
	Give an estimation of radius based on given dataset. More...

const KernelType &	Kernel () const
	Get the kernel. More...

KernelType &	Kernel ()
	Modify the kernel. More...

size_t	MaxIterations () const
	Get the maximum number of iterations. More...

size_t &	MaxIterations ()
	Set the maximum number of iterations. More...

double	Radius () const
	Get the radius. More...

void	Radius (double radius)
	Set the radius. More...

Private Member Functions
template<bool ApplyKernel = UseKernel>
std::enable_if< ApplyKernel, bool >::type	CalculateCentroid (const MatType &data, const std::vector< size_t > &neighbors, const std::vector< double > &distances, arma::colvec &centroid)
	Use kernel to calculate new centroid given dataset and valid neighbors. More...

template<bool ApplyKernel = UseKernel>
std::enable_if<!ApplyKernel, bool >::type	CalculateCentroid (const MatType &data, const std::vector< size_t > &neighbors, const std::vector< double > &, arma::colvec &centroid)
	Use mean to calculate new centroid given dataset and valid neighbors. More...

void	GenSeeds (const MatType &data, const double binSize, const int minFreq, MatType &seeds)
	To speed up, we can generate some seeds from data set and use them as initial centroids rather than all the points in the data set. More...

Private Attributes
KernelType	kernel
	Instantiated kernel. More...

size_t	maxIterations
	Maximum number of iterations before giving up. More...

double	radius
	If distance of two centroids is less than radius, one will be removed. More...

Detailed Description

template<bool UseKernel = false, typename KernelType = kernel::GaussianKernel, typename MatType = arma::mat>
class mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >

This class implements mean shift clustering.

For each point in dataset, apply mean shift algorithm until maximum iterations or convergence. Then remove duplicate centroids.

A simple example of how to run mean shift clustering is shown below.

extern arma::mat data; // Dataset we want to run mean shift on.
arma::Col<size_t> assignments; // Cluster assignments.
arma::mat centroids; // Cluster centroids.
MeanShift<> meanShift();
meanShift.Cluster(dataset, assignments, centroids);

Template Parameters

UseKernel	Use kernel or mean to calculate new centroid. If false, KernelType will be ignored.
KernelType	The kernel to use.
MatType	The type of matrix the data is stored in.

Definition at line 49 of file mean_shift.hpp.

Constructor & Destructor Documentation

template<bool UseKernel = false, typename KernelType = kernel::GaussianKernel, typename MatType = arma::mat>

mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::MeanShift	(	const double	radius = `0`,
		const size_t	maxIterations = `1000`,
		const KernelType	kernel = `KernelType()`
	)

Create a mean shift object and set the parameters which mean shift will be run with.

Parameters

radius	If distance of two centroids is less than it, one will be removed. If this value isn't positive, an estimation will be given when clustering.
maxIterations	Maximum number of iterations allowed before giving up iterations will terminate.
kernel	Optional KernelType object.

Member Function Documentation

template<bool UseKernel = false, typename KernelType = kernel::GaussianKernel, typename MatType = arma::mat>

template<bool ApplyKernel = UseKernel>

std::enable_if<ApplyKernel, bool>::type mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::CalculateCentroid	(	const MatType &	data,
		const std::vector< size_t > &	neighbors,
		const std::vector< double > &	distances,
		arma::colvec &	centroid
	)

private

Use kernel to calculate new centroid given dataset and valid neighbors.

Parameters

data	The whole dataset
neighbors	Valid neighbors
distances	Distances to neighbors centroid Store calculated centroid

Referenced by mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::Kernel().

template<bool UseKernel = false, typename KernelType = kernel::GaussianKernel, typename MatType = arma::mat>

template<bool ApplyKernel = UseKernel>

std::enable_if<!ApplyKernel, bool>::type mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::CalculateCentroid	(	const MatType &	data,
		const std::vector< size_t > &	neighbors,
		const std::vector< double > &	,
		arma::colvec &	centroid
	)

private

Use mean to calculate new centroid given dataset and valid neighbors.

Parameters

data	The whole dataset
neighbors	Valid neighbors
distances	Distances to neighbors centroid Store calculated centroid

template<bool UseKernel = false, typename KernelType = kernel::GaussianKernel, typename MatType = arma::mat>

void mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::Cluster	(	const MatType &	data,
		arma::Col< size_t > &	assignments,
		arma::mat &	centroids,
		bool	useSeeds = `true`
	)

Perform mean shift clustering on the data, returning a list of cluster assignments and centroids.

Template Parameters

MatType Type of matrix.

Parameters

data	Dataset to cluster.
assignments	Vector to store cluster assignments in.
centroids	Matrix in which centroids are stored.

template<bool UseKernel = false, typename KernelType = kernel::GaussianKernel, typename MatType = arma::mat>

double mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::EstimateRadius	(	const MatType &	data,
		const double	ratio = `0.2`
	)

Give an estimation of radius based on given dataset.

Parameters

data	Dataset for estimation.
ratio	Percentage of dataset to use for nearest neighbor search.

template<bool UseKernel = false, typename KernelType = kernel::GaussianKernel, typename MatType = arma::mat>

void mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::GenSeeds	(	const MatType &	data,
		const double	binSize,
		const int	minFreq,
		MatType &	seeds
	)

private

To speed up, we can generate some seeds from data set and use them as initial centroids rather than all the points in the data set.

The basic idea here is that we will place our points into hypercube bins of side length binSize, and any bins that contain fewer than minFreq points will be removed as possible seeds. Usually, 1 is a sufficient parameter for minFreq, and the bin size can be set equal to the estimated radius.

Parameters

data	The reference data set.
binSize	Width of hypercube bins.
minFreq	Minimum number of points in bin.
seed	Matrix to store generated seeds in.

Referenced by mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::Kernel().

template<bool UseKernel = false, typename KernelType = kernel::GaussianKernel, typename MatType = arma::mat>

const KernelType& mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::Kernel ( ) const

inline

Get the kernel.

Definition at line 100 of file mean_shift.hpp.

References mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::kernel.

template<bool UseKernel = false, typename KernelType = kernel::GaussianKernel, typename MatType = arma::mat>

KernelType& mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::Kernel ( )

inline

Modify the kernel.

Definition at line 102 of file mean_shift.hpp.

References mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::CalculateCentroid(), mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::GenSeeds(), and mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::kernel.

template<bool UseKernel = false, typename KernelType = kernel::GaussianKernel, typename MatType = arma::mat>

size_t mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::MaxIterations ( ) const

inline

Get the maximum number of iterations.

Definition at line 90 of file mean_shift.hpp.

References mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::maxIterations.

template<bool UseKernel = false, typename KernelType = kernel::GaussianKernel, typename MatType = arma::mat>

size_t& mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::MaxIterations ( )

inline

Set the maximum number of iterations.

Definition at line 92 of file mean_shift.hpp.

References mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::maxIterations.

template<bool UseKernel = false, typename KernelType = kernel::GaussianKernel, typename MatType = arma::mat>

double mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::Radius ( ) const

inline

Get the radius.

Definition at line 95 of file mean_shift.hpp.

References mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::radius.

template<bool UseKernel = false, typename KernelType = kernel::GaussianKernel, typename MatType = arma::mat>

void mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::Radius ( double radius )

Set the radius.

Member Data Documentation

template<bool UseKernel = false, typename KernelType = kernel::GaussianKernel, typename MatType = arma::mat>

KernelType mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::kernel

private

Instantiated kernel.

Definition at line 164 of file mean_shift.hpp.

Referenced by mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::Kernel().

template<bool UseKernel = false, typename KernelType = kernel::GaussianKernel, typename MatType = arma::mat>

size_t mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::maxIterations

private

Maximum number of iterations before giving up.

Definition at line 161 of file mean_shift.hpp.

Referenced by mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::MaxIterations().

template<bool UseKernel = false, typename KernelType = kernel::GaussianKernel, typename MatType = arma::mat>

double mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::radius

private

If distance of two centroids is less than radius, one will be removed.

Points with distance to current centroid less than radius will be used to calculate new centroid.

Definition at line 158 of file mean_shift.hpp.

Referenced by mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >::Radius().

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

src/mlpack/methods/mean_shift/mean_shift.hpp

Public Member Functions

Private Member Functions

Private Attributes

Detailed Description

template<bool UseKernel = false, typename KernelType = kernel::GaussianKernel, typename MatType = arma::mat> class mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >

Constructor & Destructor Documentation

Member Function Documentation

centroid Store calculated centroid

centroid Store calculated centroid

Member Data Documentation

template<bool UseKernel = false, typename KernelType = kernel::GaussianKernel, typename MatType = arma::mat>
class mlpack::meanshift::MeanShift< UseKernel, KernelType, MatType >