org.knowceans.dirichlet.sandbox
Class LdaGibbsSamplerDpa

java.lang.Object
  org.knowceans.dirichlet.sandbox.LdaGibbsSamplerDpa

All Implemented Interfaces:

java.io.Serializable

Deprecated. use LdaGibbsSamplerCps, this class only for backward compatibility and testing.

public class LdaGibbsSamplerDpa
extends java.lang.Object
implements java.io.Serializable

Gibbs sampler for estimating the best assignments of topics for words and documents in a corpus. The algorithm is introduced in Tom Griffiths' paper "Gibbs sampling in the generative model of Latent Dirichlet Allocation" (2002).

Author:

heinrich

See Also:

Serialized Form

Field Summary

protected double	alpha Deprecated. Dirichlet parameter (document--topic associations)
protected VariationOfInformationAnalyser	analyser Deprecated.
protected int	backupInterval Deprecated.
protected int	backupIteration Deprecated. iteration in the last backup
protected double	beta Deprecated. Dirichlet parameter (topic--term associations)
protected int	BURN_IN Deprecated. burn-in period
protected java.lang.String	corpusname Deprecated.
protected int	dispcol Deprecated.
protected int[][]	documents Deprecated. document data (term lists)
protected int[]	interSamples Deprecated. number of iteration at which intermediate (single) samples are taken.
protected boolean	interSave Deprecated.
protected boolean	interTopics Deprecated.
protected int	ITERATIONS Deprecated. max iterations
protected int	K Deprecated. number of topics
protected java.lang.String	messageheader Deprecated.
protected java.lang.String[]	messagerecipients Deprecated.
protected java.lang.String	messagetext Deprecated.
protected int[][]	nd Deprecated. nd[d][k] number of words in document d assigned to topic k.
protected int[]	ndsum Deprecated. ndsum[d] total number of words in document d.
protected int	numstats Deprecated. size of statistics
protected int[][]	nw Deprecated. cwt[k][j] number of instances of word j (term?)
protected int[]	nwsum Deprecated. nwsum[k] total number of words assigned to topic k.
protected java.lang.String	outfilename Deprecated.
protected double[][]	phisum Deprecated. cumulative statistics of phi
protected int	SAMPLE_LAG Deprecated. sample lag (if -1 only one sample taken)
private static long	serialVersionUID Deprecated.
protected long	t0 Deprecated.
protected double[][]	thetasum Deprecated. cumulative statistics of theta
protected int	THIN_INTERVAL Deprecated. sampling lag (?)
protected long	timeElapsed Deprecated.
protected int	V Deprecated. vocabulary size
protected int[][]	z Deprecated. topic assignments for each word.

Constructor Summary

LdaGibbsSamplerDpa(int[][] documents, int V)
Deprecated. Initialise the Gibbs sampler with data.

Method Summary

protected void	configureMessaging(java.lang.String header, java.lang.String text, java.lang.String[] recipients) Deprecated. configure the sampler for messaging
protected void	configureOutput(java.lang.String corpusname, java.lang.String outfilename, int backupInterval, int[] interSamples, VariationOfInformationAnalyser analyser, boolean interSave, boolean interTopics) Deprecated. configure the sampler output
void	configureSampler(int iterations, int burnIn, int thinInterval, int sampleLag, int K, double alpha, double beta) Deprecated. Configure the gibbs sampler
private VariationOfInformationAnalyser.DistMetric	distance(java.lang.String outname) Deprecated. perform a distance calculation on the estimated results
double	getAlpha() Deprecated.
double	getBeta() Deprecated.
int[][]	getDocuments() Deprecated.
int	getK() Deprecated.
double[][]	getPhi() Deprecated. Retrieve estimated topic--word associations.
double[][]	getTheta() Deprecated. Retrieve estimated document--topic associations.
protected long	getTimer() Deprecated. get the current value of the timer.
int	getV() Deprecated.
int[][]	getZ() Deprecated.
private void	gibbs() Deprecated. Main method: Select initial state ?
void	initialState() Deprecated. Initialisation: Must start with an assignment of observations to topics ?
static java.lang.Object	load(java.lang.String filename) Deprecated. read object from the stream
static void	main(java.lang.String[] args) Deprecated.
protected void	output(java.lang.String analysisfile, java.lang.String addheader, java.lang.String addmessage) Deprecated. Calculate distance (if doDist) and replace all occurrences of $@ and $# in strings by the complete distance information and the distance value only, respectively.
protected void	sampleCorpus() Deprecated. sample once through the corpus.
protected int	sampleLdaFullConditional(int m, int n) Deprecated. Sample a topic z_i from the full conditional distribution: p(z_i = j | z_-i, w) = (n_-i,j(w_i) + beta)/(n_-i,j(.) + W * beta) * (n_-i,j(d_i) + alpha)/(n_-i,.
void	save(java.lang.String filename) Deprecated. Object stream only for testing.
void	setAlpha(double alpha) Deprecated.
void	setBeta(double beta) Deprecated.
protected void	startTimer(long offset) Deprecated. start timer from with an initial offset.
protected void	updateParams() Deprecated. Add to the statistics the values of theta and phi for the current state.
protected static void	writeParameters(java.lang.String file, java.lang.String corpusname, int k, double alpha, double beta, int m, int v, int w, long duration, int iterations, int samplelag, int burnin, org.knowceans.util.Arguments a) Deprecated. write statistics of the current run to a text file for later review

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

serialVersionUID

private static final long serialVersionUID

Deprecated.

See Also:

Constant Field Values

documents

protected int[][] documents

Deprecated.

document data (term lists)

V

protected int V

Deprecated.

vocabulary size

K

protected int K

Deprecated.

number of topics

alpha

protected double alpha

Deprecated.

Dirichlet parameter (document--topic associations)

beta

protected double beta

Deprecated.

Dirichlet parameter (topic--term associations)

z

protected int[][] z

Deprecated.

topic assignments for each word.

nw

protected int[][] nw

Deprecated.

cwt[k][j] number of instances of word j (term?) assigned to topic k.

nd

protected int[][] nd

Deprecated.

nd[d][k] number of words in document d assigned to topic k.

nwsum

protected int[] nwsum

Deprecated.

nwsum[k] total number of words assigned to topic k.

ndsum

protected int[] ndsum

Deprecated.

ndsum[d] total number of words in document d.

thetasum

protected double[][] thetasum

Deprecated.

cumulative statistics of theta

phisum

protected double[][] phisum

Deprecated.

cumulative statistics of phi

numstats

protected int numstats

Deprecated.

size of statistics

interSamples

protected int[] interSamples

Deprecated.

number of iteration at which intermediate (single) samples are taken.

THIN_INTERVAL

protected int THIN_INTERVAL

Deprecated.

sampling lag (?)

BURN_IN

protected int BURN_IN

Deprecated.

burn-in period

ITERATIONS

protected int ITERATIONS

Deprecated.

max iterations

SAMPLE_LAG

protected int SAMPLE_LAG

Deprecated.

sample lag (if -1 only one sample taken)

dispcol

protected int dispcol

Deprecated.

corpusname

protected java.lang.String corpusname

Deprecated.

outfilename

protected java.lang.String outfilename

Deprecated.

messagetext

protected java.lang.String messagetext

Deprecated.

messageheader

protected java.lang.String messageheader

Deprecated.

messagerecipients

protected java.lang.String[] messagerecipients

Deprecated.

interSave

protected boolean interSave

Deprecated.

analyser

protected VariationOfInformationAnalyser analyser

Deprecated.

interTopics

protected boolean interTopics

Deprecated.

backupIteration

protected int backupIteration

Deprecated.

iteration in the last backup

timeElapsed

protected long timeElapsed

Deprecated.

backupInterval

protected int backupInterval

Deprecated.

t0

protected long t0

Deprecated.

Constructor Detail

LdaGibbsSamplerDpa

public LdaGibbsSamplerDpa(int[][] documents,
                          int V)

Deprecated.

Initialise the Gibbs sampler with data.

Parameters:

V - vocabulary size

data -

Method Detail

initialState

public void initialState()

Deprecated.

Initialisation: Must start with an assignment of observations to topics ? Many alternatives are possible, I chose to perform random assignments with equal probabilities

gibbs

private void gibbs()

Deprecated.

Main method: Select initial state ? Repeat a large number of times: 1. Select an element 2. Update conditional on other elements. If appropriate, output summary for each run.

sampleCorpus

protected void sampleCorpus()

Deprecated.

sample once through the corpus.

sampleLdaFullConditional

protected int sampleLdaFullConditional(int m,
                                       int n)

Deprecated.

Sample a topic z_i from the full conditional distribution: p(z_i = j | z_-i, w) = (n_-i,j(w_i) + beta)/(n_-i,j(.) + W * beta) * (n_-i,j(d_i) + alpha)/(n_-i,.(d_i) + K * alpha)

Parameters:

m - document

n - word

updateParams

protected void updateParams()

Deprecated.

Add to the statistics the values of theta and phi for the current state.

getTheta

public double[][] getTheta()

Deprecated.

Retrieve estimated document--topic associations. If sample lag > 0 then the mean value of all sampled statistics for theta[][] is taken.

Returns:

theta multinomial mixture of document topics (M x K)

getPhi

public double[][] getPhi()

Deprecated.

Retrieve estimated topic--word associations. If sample lag > 0 then the mean value of all sampled statistics for phi[][] is taken.

Returns:

phi multinomial mixture of topic words (K x V)

configureSampler

public void configureSampler(int iterations,
                             int burnIn,
                             int thinInterval,
                             int sampleLag,
                             int K,
                             double alpha,
                             double beta)

Deprecated.

Configure the gibbs sampler

Parameters:

iterations - number of total iterations

burnIn - number of burn-in iterations

thinInterval - update statistics interval

sampleLag - sample interval (-1 for just one sample at the end)

K - number of topics

alpha - symmetric prior parameter on document--topic associations

beta - symmetric prior parameter on topic--term associations

configureOutput

protected void configureOutput(java.lang.String corpusname,
                               java.lang.String outfilename,
                               int backupInterval,
                               int[] interSamples,
                               VariationOfInformationAnalyser analyser,
                               boolean interSave,
                               boolean interTopics)

Deprecated.

configure the sampler output

Parameters:

corpusname -

outfilename -

backupInterval -

interSamples -

analyser -

interSave3 -

interSave -

configureMessaging

protected void configureMessaging(java.lang.String header,
                                  java.lang.String text,
                                  java.lang.String[] recipients)

Deprecated.

configure the sampler for messaging

Parameters:

header -

text -

recipients -

main

public static void main(java.lang.String[] args)

Deprecated.

startTimer

protected void startTimer(long offset)

Deprecated.

start timer from with an initial offset. Further, the value of time elapsed is considered.

Parameters:

offset -

getTimer

protected long getTimer()

Deprecated.

get the current value of the timer.

Returns:

output

protected void output(java.lang.String analysisfile,
                      java.lang.String addheader,
                      java.lang.String addmessage)

Deprecated.

Calculate distance (if doDist) and replace all occurrences of $@ and $# in strings by the complete distance information and the distance value only, respectively. For console output, the addmessage text is replaced with distance information ($@, $#) and the message text printed. The method works in a separate thread. If mailing is enabled, the messagetext variable is appended with addmessage and header is appended with addheader and all $@ and $# occurrences replaced. Email is sent afterwards.

Parameters:

addmessage -

doDist -

save

public void save(java.lang.String filename)

Deprecated.

Object stream only for testing.

load

public static java.lang.Object load(java.lang.String filename)

Deprecated.

read object from the stream

Parameters:

filename -

Returns:

distance

private VariationOfInformationAnalyser.DistMetric distance(java.lang.String outname)

Deprecated.

perform a distance calculation on the estimated results

Returns:

writeParameters

protected static void writeParameters(java.lang.String file,
                                      java.lang.String corpusname,
                                      int k,
                                      double alpha,
                                      double beta,
                                      int m,
                                      int v,
                                      int w,
                                      long duration,
                                      int iterations,
                                      int samplelag,
                                      int burnin,
                                      org.knowceans.util.Arguments a)

Deprecated.

write statistics of the current run to a text file for later review

Parameters:

file -

corpusname -

k - topics

alpha - hyperparameter

beta - hyperparameter

m - doc count

v - vocabulary/term count

w - word count

duration - training duration

iterations - no. of total iterations

samplelag - sampling lag

burnin - burnin samples

a - Arguments object

getAlpha

public final double getAlpha()

Deprecated.

setAlpha

public final void setAlpha(double alpha)

Deprecated.

getBeta

public final double getBeta()

Deprecated.

setBeta

public final void setBeta(double beta)

Deprecated.

getDocuments

public final int[][] getDocuments()

Deprecated.

getK

public final int getK()

Deprecated.

getV

public final int getV()

Deprecated.

getZ

public final int[][] getZ()

Deprecated.