---

fl::KMeans Class Reference

#include <cluster.h>

Inheritance diagram for fl::KMeans:

List of all members.

Public Methods
	KMeans (float maxSize, float minSize, int initialK, int maxK, const std::string &clusterFileName="")
	clusterFileName refers to target file for new clustering data, which is very likely to be different from input file.
	KMeans (std::istream &stream, const std::string &clusterFileName="")
	Construct from stream.
virtual void	run (const std::vector< Vector< float > > &data)
	Peform clustering on collection of points.
virtual int	classify (const Vector< float > &point)
	Determine the single best class of given point.
virtual Vector< float >	distribution (const Vector< float > &point)
	Return a probability distribution over the classes. Row number in the returned Vector corresponds to class number.
virtual int	classCount ()
	Returns the number of classes.
virtual Vector< float >	representative (int group)
	Return a representative member of group. "group" has same semantics as return value of classify (); we just can't use the word "class" because it is a keyword in C++. :).
virtual void	read (std::istream &stream)
virtual void	write (std::ostream &stream, bool withName=false)
void	initialize (const std::vector< Vector< float > > &data)
void	estimate (const std::vector< Vector< float > > &data, Matrix< float > &member, int jbegin, int jend)
float	maximize (const std::vector< Vector< float > > &data, const Matrix< float > &member, int i)
bool	convergence (const std::vector< Vector< float > > &data, const Matrix< float > &member, float largestChange)
Public Attributes
float	maxSize
	Largest length of dominant axis of covariance matrix. If any cluster exceeds this value, create a new cluster.
float	minSize
	Closest that two clusters can be before they merge.
int	initialK
	Lower bound on expected number of clusters.
int	maxK
	Largest number of clusters allowed.
std::vector< ClusterGauss >	clusters
std::vector< float >	changes
std::vector< float >	velocities
std::string	clusterFileName
time_t	clusterFileTime
	Time in seconds.
off_t	clusterFileSize

---

Constructor & Destructor Documentation

KMeans::KMeans (  float    maxSize, float    minSize, int    initialK, int    maxK, const std::string &    clusterFileName = "" )

clusterFileName refers to target file for new clustering data, which is very likely to be different from input file.

KMeans::KMeans (  std::istream &    stream, const std::string &    clusterFileName = "" )

Construct from stream.

---

Member Function Documentation

int KMeans::classCount (    )  [virtual]

Returns the number of classes. Implements fl::ClusterMethod.

int KMeans::classify (  const Vector< float > &    point )  [virtual]

Determine the single best class of given point. Implements fl::ClusterMethod.

bool KMeans::convergence (  const std::vector< Vector< float > > &    data, const Matrix< float > &    member, float    largestChange )

Vector< float > KMeans::distribution (  const Vector< float > &    point )  [virtual]

Return a probability distribution over the classes. Row number in the returned Vector corresponds to class number. Implements fl::ClusterMethod.

void KMeans::estimate (  const std::vector< Vector< float > > &    data, Matrix< float > &    member, int    jbegin, int    jend )

void KMeans::initialize (  const std::vector< Vector< float > > &    data )

float KMeans::maximize (  const std::vector< Vector< float > > &    data, const Matrix< float > &    member, int    i )

void KMeans::read (  std::istream &    stream )  [virtual]

Reimplemented from fl::ClusterMethod.

Vector< float > KMeans::representative (  int    group )  [virtual]

Return a representative member of group. "group" has same semantics as return value of classify (); we just can't use the word "class" because it is a keyword in C++. :). Implements fl::ClusterMethod.

void KMeans::run (  const std::vector< Vector< float > > &    data )  [virtual]

Peform clustering on collection of points. Implements fl::ClusterMethod. Reimplemented in fl::KMeansParallel.

void KMeans::write (  std::ostream &    stream, bool    withName = false )  [virtual]

Reimplemented from fl::ClusterMethod.

---

Member Data Documentation

std::vector<float> fl::KMeans::changes

std::string fl::KMeans::clusterFileName

off_t fl::KMeans::clusterFileSize

time_t fl::KMeans::clusterFileTime

Time in seconds.

std::vector<ClusterGauss> fl::KMeans::clusters

int fl::KMeans::initialK

Lower bound on expected number of clusters.

int fl::KMeans::maxK

Largest number of clusters allowed.

float fl::KMeans::maxSize

Largest length of dominant axis of covariance matrix. If any cluster exceeds this value, create a new cluster.

float fl::KMeans::minSize

Closest that two clusters can be before they merge.

std::vector<float> fl::KMeans::velocities

---

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

• cluster.h
• KMeans.cc

---