lti::kMeansClustering Class Reference

This class implements two versions of k-means-clustering: batch and sequential. More...

#include <ltiKMeansClustering.h>

Inheritance diagram for lti::kMeansClustering:

[legend]

Collaboration diagram for lti::kMeansClustering:

[legend]

List of all members.

Classes
class	parameters
	Parameters for kMeansClustering. More...
Public Member Functions
	kMeansClustering ()
	kMeansClustering (const kMeansClustering &other)
virtual	~kMeansClustering ()
virtual const char *	getTypeName () const
kMeansClustering &	copy (const kMeansClustering &other)
virtual classifier *	clone () const
const parameters &	getParameters () const
virtual bool	train (const dmatrix &data)
virtual bool	train (const dmatrix &input, ivector &ids)
Protected Member Functions
bool	trainBatch (const dmatrix &data)
bool	trainSequential (const dmatrix &data)

---

Detailed Description

This class implements two versions of k-means-clustering: batch and sequential.

Both methods are initailized by drawing numberOfClusters training points and assigning them as centroids. The batch version continues by labeling each training point with the centroid it belongs to and then calculating new centroids as mean average of the data belonging to each cluster. This is continued until a convergence criterion is met.

The sequential algorithm performs one step of batch training. After that training points are analyzed in random order. Whenever one data point switches clusters the centroids of the old and new cluster of that data point are recalculated. Calculation is also stoped by a convergence criterion.

For both methods the convergence criterion is that for one run through the training set none of the training points switched clusters.

---

Constructor & Destructor Documentation

lti::kMeansClustering::kMeansClustering

(

)

default constructor

lti::kMeansClustering::kMeansClustering

(

const kMeansClustering &

other

)

copy constructor

Parameters:

other

the object to be copied

virtual lti::kMeansClustering::~kMeansClustering

(

)

[virtual]

destructor

---

Member Function Documentation

virtual classifier* lti::kMeansClustering::clone

(

)

const [virtual]

returns a pointer to a clone of this classifier.

Implements lti::classifier.

kMeansClustering& lti::kMeansClustering::copy

(

const kMeansClustering &

other

)

copy data of "other" functor.

Parameters:

other

the functor to be copied

Returns:

a reference to this functor object

Reimplemented from lti::centroidClustering.

const parameters& lti::kMeansClustering::getParameters

(

)

const

returns used parameters

Reimplemented from lti::centroidClustering.

virtual const char* lti::kMeansClustering::getTypeName

(

)

const [virtual]

returns the name of this type ("kMeansClustering")

Reimplemented from lti::centroidClustering.

virtual bool lti::kMeansClustering::train	(	const dmatrix &	input,
		ivector &	ids
	)			[virtual]

calls centroidClustering::train(const dmatrix&, ivector&)

Reimplemented from lti::centroidClustering.

virtual bool lti::kMeansClustering::train

(

const dmatrix &

data

)

[virtual]

Performs batch or sequential training according to the value of parameters::clusterMode.

Parameters:

data

the training points

Implements lti::centroidClustering.

bool lti::kMeansClustering::trainBatch

(

const dmatrix &

data

)

[protected]

returns current parameters.

(non const! -> protected) Performs batch k-means clustering, see class description.

bool lti::kMeansClustering::trainSequential

(

const dmatrix &

data

)

[protected]

Performs sequential k-means clustering, see class description.

---

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

• ltiKMeansClustering.h