lemur::cluster::PLSA Class Reference

Probabilistic Latent Semantic Analysis Java Reference implementation from Andrew Schein and Alexandrin Popescul (Penn). PennAspect (GPL). More...

#include <PLSA.hpp>

List of all members.

Public Member Functions
	PLSA (const lemur::api::Index &dbIndex, int numCats, lemur::utility::HashFreqVector **train, lemur::utility::HashFreqVector **validate, int numIter, int numRestarts, double betastart, double betastop, double anneal, double betaMod)
	building with provided train/test partitions
	PLSA (const lemur::api::Index &dbIndex, int testPercentage, int numCats, int numIter, int numRestarts, double betastart, double betastop, double anneal, double betaMod)
	building without provided train/test partitions
	PLSA (const lemur::api::Index &dbIndex)
	for using prebuilt tables.
virtual	~PLSA ()
void	iterateWithRestarts ()
	Start things going.
double *	get_p_z () const
double **	get_p_w_z () const
	P(w|z) matrix.
double **	get_p_d_z () const
	P(d|z) matrix.
double	getProb (int d, int w) const
	P(d,w).
int	numWords () const
	number of terms
int	numDocs () const
	number of docs
int	numCats () const
	number of categories
bool	readArrays ()
Private Types
enum	pType { P_Z = 0, P_W_Z = 1, P_D_Z = 2 }
	read/write array options. More...
Private Member Functions
void	setPrevToCurrent ()
	Copy current iteration data to previous iteration data.
void	setCurrentToBest ()
	Copy best iteration data to current iteration data.
void	setBestToCurrent ()
	Copy current iteration data to best iteration data.
void	setBestToPrev ()
	Copy previous iteration data to best iteration data.
void	setPrevToBest ()
	Copy best iteration data to previous iteration data.
double	getAverageLikelihood ()
double	getAverageLikelihoodPrev ()
double	jointEstimate (int indexD, int indexW)
	Estimates P(d,w) using previous parameter estimates.
double	jointEstimateCurrent (int indexD, int indexW)
	Estimates P(d,w) using current parameter estimates.
double	jointEstimateBest (int indexD, int indexW)
	Estimates P(d,w) using best parameter estimates.
double	jointEstimateBeta (int indexD, int indexW)
void	iterate ()
	main routine for model training.
void	initializeParameters ()
	Initialize the prev probability arrays to random values.
double	doLogLikelihood (jointfuncType, lemur::utility::HashFreqVector **&myData)
double	logLikelihood ()
	Calculate the training data log-likelihood using prev parameters.
double	validateDataLogLikelihood ()
	Calculate the hold out data log-likelihood using prev parameters.
double	validateCurrentLogLikelihood ()
	Calculate the hold out data log-likelihood using current parameters.
double	bestDataLogLikelihood ()
	Calculate the hold out data log-likelihood using the best parameters.
double	interleavedIterationEM ()
	performs one EM iteration, returns log likelihood of training data
void	selectTestTrain (int testPercent)
	Select training/test events.
void	init ()
	Initialize attributes.
void	initR ()
	Initialize R and w->d inverted list.
void	writeArrays ()
	write out all the arrays to file.
bool	readArray (ifstream &infile, enum pType which)
	Read a probability array (matrix) from a file.
void	writeArray (ofstream &ofile, enum pType which)
	Write a probability array (matrix) to a file.
Private Attributes
const lemur::api::Index &	ind
	Index to use.
int	sizeZ
	number of categories
int	sizeD
	number of documents
int	sizeW
	number of words
lemur::utility::HashFreqVector **	data
	train d->w freq list
lemur::utility::HashFreqVector **	testData
	test (validation) d->w freq list
set< int, less< int > > *	invIndex
	w->d inverted index for M step of P(w | z)
double	startBeta
	Beta for TEM.
double	beta
	Beta for TEM.
double	betaMin
	Beta for TEM.
double	betaModifier
	eta for TEM (beta = eta * beta;)
double	annealcue
	annealcue value (delta)
int	R
	used in M step for p_z
int	numberOfIterations
	How many iterations.
int	numberOfRestarts
	How many restarts.
double	bestTestLL
	Best log likelihood on the test data so far.
double	bestA
	Best average log likelihood on the test data so far.
bool	bestOnly
	have we only loaded existing tables from files
bool	ownMem
	did we allocate the test/train vectors?
double *	p_z_current
	P(z) vector current iteration.
double **	p_w_z_current
	P(w|z) matrix current iteration.
double **	p_d_z_current
	P(d|z) matrix current iteration.
double *	p_z_prev
	P(z) vector previous iteration.
double **	p_w_z_prev
	P(w|z) matrix previous iteration.
double **	p_d_z_prev
	P(d|z) matrix previous iteration.
double *	p_z_best
	P(z) vector best iteration.
double **	p_w_z_best
	P(w|z) matrix best iteration.
double **	p_d_z_best
	P(d|z) matrix best iteration.

---

Detailed Description

Probabilistic Latent Semantic Analysis Java Reference implementation from Andrew Schein and Alexandrin Popescul (Penn). PennAspect (GPL).

---

Member Enumeration Documentation

enum lemur::cluster::PLSA::pType [private]

read/write array options. Enumeration values: P_Z  P_W_Z  P_D_Z

---

Constructor & Destructor Documentation

lemur::cluster::PLSA::PLSA (  const lemur::api::Index &  dbIndex, int  numCats, lemur::utility::HashFreqVector **  train, lemur::utility::HashFreqVector **  validate, int  numIter, int  numRestarts, double  betastart, double  betastop, double  anneal, double  betaMod )

building with provided train/test partitions

lemur::cluster::PLSA::PLSA (  const lemur::api::Index &  dbIndex, int  testPercentage, int  numCats, int  numIter, int  numRestarts, double  betastart, double  betastop, double  anneal, double  betaMod )

building without provided train/test partitions

lemur::cluster::PLSA::PLSA (  const lemur::api::Index &  dbIndex  )

for using prebuilt tables. pass in a filestem.

lemur::cluster::PLSA::~PLSA (   )  [virtual]

---

Member Function Documentation

double lemur::cluster::PLSA::bestDataLogLikelihood (   )  [private]

Calculate the hold out data log-likelihood using the best parameters.

double lemur::cluster::PLSA::doLogLikelihood (  jointfuncType  , lemur::utility::HashFreqVector **&  myData )  [private]

Calculate the log likelihood of a given data set using the supplied joint estimate method.

double** lemur::cluster::PLSA::get_p_d_z (   )  const [inline]

P(d|z) matrix.

double** lemur::cluster::PLSA::get_p_w_z (   )  const [inline]

P(w|z) matrix.

double* lemur::cluster::PLSA::get_p_z (   )  const [inline]

get the final values P(z) vector

double lemur::cluster::PLSA::getAverageLikelihood (   )  [private]

Calculate the average likelihood of an event in the test data using the current iteration estimates.

double lemur::cluster::PLSA::getAverageLikelihoodPrev (   )  [private]

Calculate the average likelihood of an event in the test data using the previous iteration estimates.

double lemur::cluster::PLSA::getProb (  int  d, int  w )  const

P(d,w).

void lemur::cluster::PLSA::init (   )  [private]

Initialize attributes.

void lemur::cluster::PLSA::initializeParameters (   )  [private]

Initialize the prev probability arrays to random values.

void lemur::cluster::PLSA::initR (   )  [private]

Initialize R and w->d inverted list.

double lemur::cluster::PLSA::interleavedIterationEM (   )  [private]

performs one EM iteration, returns log likelihood of training data

void lemur::cluster::PLSA::iterate (   )  [private]

main routine for model training.

void lemur::cluster::PLSA::iterateWithRestarts (   )

Start things going.

double lemur::cluster::PLSA::jointEstimate (  int  indexD, int  indexW )  [private]

Estimates P(d,w) using previous parameter estimates.

double lemur::cluster::PLSA::jointEstimateBest (  int  indexD, int  indexW )  [private]

Estimates P(d,w) using best parameter estimates.

double lemur::cluster::PLSA::jointEstimateBeta (  int  indexD, int  indexW )  [private]

Joint estimate using previous parameter estimates where the terms in summation are raised to power beta.

double lemur::cluster::PLSA::jointEstimateCurrent (  int  indexD, int  indexW )  [private]

Estimates P(d,w) using current parameter estimates.

double lemur::cluster::PLSA::logLikelihood (   )  [private]

Calculate the training data log-likelihood using prev parameters.

int lemur::cluster::PLSA::numCats (   )  const [inline]

number of categories

int lemur::cluster::PLSA::numDocs (   )  const [inline]

number of docs

int lemur::cluster::PLSA::numWords (   )  const [inline]

number of terms

bool lemur::cluster::PLSA::readArray (  ifstream &  infile, enum pType  which )  [private]

Read a probability array (matrix) from a file.

bool lemur::cluster::PLSA::readArrays (   )

should these be public? On ctor if not constructing.

void lemur::cluster::PLSA::selectTestTrain (  int  testPercent  )  [private]

Select training/test events.

void lemur::cluster::PLSA::setBestToCurrent (   )  [private]

Copy current iteration data to best iteration data.

void lemur::cluster::PLSA::setBestToPrev (   )  [private]

Copy previous iteration data to best iteration data.

void lemur::cluster::PLSA::setCurrentToBest (   )  [private]

Copy best iteration data to current iteration data.

void lemur::cluster::PLSA::setPrevToBest (   )  [private]

Copy best iteration data to previous iteration data.

void lemur::cluster::PLSA::setPrevToCurrent (   )  [private]

Copy current iteration data to previous iteration data.

double lemur::cluster::PLSA::validateCurrentLogLikelihood (   )  [private]

Calculate the hold out data log-likelihood using current parameters.

double lemur::cluster::PLSA::validateDataLogLikelihood (   )  [private]

Calculate the hold out data log-likelihood using prev parameters.

void lemur::cluster::PLSA::writeArray (  ofstream &  ofile, enum pType  which )  [private]

Write a probability array (matrix) to a file.

void lemur::cluster::PLSA::writeArrays (   )  [private]

write out all the arrays to file.

---

Member Data Documentation

double lemur::cluster::PLSA::annealcue [private]

annealcue value (delta)

double lemur::cluster::PLSA::bestA [private]

Best average log likelihood on the test data so far.

bool lemur::cluster::PLSA::bestOnly [private]

have we only loaded existing tables from files

double lemur::cluster::PLSA::bestTestLL [private]

Best log likelihood on the test data so far.

double lemur::cluster::PLSA::beta [private]

Beta for TEM.

double lemur::cluster::PLSA::betaMin [private]

Beta for TEM.

double lemur::cluster::PLSA::betaModifier [private]

eta for TEM (beta = eta * beta;)

lemur::utility::HashFreqVector** lemur::cluster::PLSA::data [private]

train d->w freq list

const lemur::api::Index& lemur::cluster::PLSA::ind [private]

Index to use.

set<int, less<int> >* lemur::cluster::PLSA::invIndex [private]

w->d inverted index for M step of P(w | z)

int lemur::cluster::PLSA::numberOfIterations [private]

How many iterations.

int lemur::cluster::PLSA::numberOfRestarts [private]

How many restarts.

bool lemur::cluster::PLSA::ownMem [private]

did we allocate the test/train vectors?

double** lemur::cluster::PLSA::p_d_z_best [private]

P(d|z) matrix best iteration.

double** lemur::cluster::PLSA::p_d_z_current [private]

P(d|z) matrix current iteration.

double** lemur::cluster::PLSA::p_d_z_prev [private]

P(d|z) matrix previous iteration.

double** lemur::cluster::PLSA::p_w_z_best [private]

P(w|z) matrix best iteration.

double** lemur::cluster::PLSA::p_w_z_current [private]

P(w|z) matrix current iteration.

double** lemur::cluster::PLSA::p_w_z_prev [private]

P(w|z) matrix previous iteration.

double* lemur::cluster::PLSA::p_z_best [private]

P(z) vector best iteration.

double* lemur::cluster::PLSA::p_z_current [private]

P(z) vector current iteration.

double* lemur::cluster::PLSA::p_z_prev [private]

P(z) vector previous iteration.

int lemur::cluster::PLSA::R [private]

used in M step for p_z

int lemur::cluster::PLSA::sizeD [private]

number of documents

int lemur::cluster::PLSA::sizeW [private]

number of words

int lemur::cluster::PLSA::sizeZ [private]

number of categories

double lemur::cluster::PLSA::startBeta [private]

Beta for TEM.

lemur::utility::HashFreqVector** lemur::cluster::PLSA::testData [private]

test (validation) d->w freq list

---

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

• PLSA.hpp
• cluster/src/PLSA.cpp

---