org.openimaj.pgm.vb.lda.mle

Class LDAModel

java.lang.Object

org.openimaj.pgm.vb.lda.mle.LDAModel

public class LDAModel
extends Object

Holds the sufficient statistics for a maximum liklihood LDA as well as a single value for alpha (the parameter of the topic dirichlet prior)

Author:

Sina Samangooei (ss@ecs.soton.ac.uk)

Field Summary

Fields

Modifier and Type	Field and Description
double	alpha the dirichelet perameter for every dimension of the topic multinomial prior
int	ntopics number of topics
double[]	topicTotal The maximum likelihood sufficient statistics for estimation of Beta.
double[][]	topicWord The maximum likelihood sufficient statistics for estimation of Beta.

Constructor Summary

Constructors

Constructor and Description
LDAModel(int ntopics)

Method Summary

Modifier and Type	Method and Description
void	incTopicTotal(int topicIndex, double d) Increment a topic and word index by d.
void	incTopicWord(int topicIndex, int wordIndex, double d) Increment a topic and word index by d.
LDAModel	newInstance() This method also swaps the likelihoods (i.e.
void	prepare(Corpus corpus) initialises the sufficient statistic holder based on ntopics and the Corpus.vocabularySize().
void	prepare(int vocabularySize) initialises the sufficient statistic holder based on ntopics and the vocabularySize.
void	setAlpha(double initialAlpha)

Methods inherited from class java.lang.Object

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

Field Detail

alpha

public double alpha

the dirichelet perameter for every dimension of the topic multinomial prior

topicWord

public double[][] topicWord

The maximum likelihood sufficient statistics for estimation of Beta. This is the number of times a given word is in a given topic

topicTotal

public double[] topicTotal

The maximum likelihood sufficient statistics for estimation of Beta. This is the number of words total in a given topic

ntopics

public int ntopics

number of topics

Constructor Detail

LDAModel

public LDAModel(int ntopics)

Parameters:

ntopics - the number of topics in this LDA model

Method Detail

prepare

public void prepare(Corpus corpus)

initialises the sufficient statistic holder based on ntopics and the Corpus.vocabularySize(). Alpha remains at 0

Parameters:

corpus -

prepare

public void prepare(int vocabularySize)

initialises the sufficient statistic holder based on ntopics and the vocabularySize. Alpha remains at 0

Parameters:

vocabularySize -

incTopicWord

public void incTopicWord(int topicIndex,
                         int wordIndex,
                         double d)

Increment a topic and word index by d. The totals are left untouched

Parameters:

topicIndex -

wordIndex -

d -

incTopicTotal

public void incTopicTotal(int topicIndex,
                          double d)

Increment a topic and word index by d. The totals are left untouched

Parameters:

topicIndex -

d -

setAlpha

public void setAlpha(double initialAlpha)

Parameters:

initialAlpha - the alpha parameter for the topic multinomial dirichelet prior

newInstance

public LDAModel newInstance()

This method also swaps the likelihoods (i.e. oldLikelihood == likelihood, likelhood = 0)

Returns:

a blank copy with unset alpha matching the current model's configuration